diff options
Diffstat (limited to 'sys')
278 files changed, 9926 insertions, 3203 deletions
diff --git a/sys/amd64/acpica/acpi_wakeup.c b/sys/amd64/acpica/acpi_wakeup.c index 51d6d5e36840..99565fbb69ca 100644 --- a/sys/amd64/acpica/acpi_wakeup.c +++ b/sys/amd64/acpica/acpi_wakeup.c @@ -54,10 +54,8 @@ #include <x86/apicreg.h> #include <x86/apicvar.h> -#ifdef SMP #include <machine/smp.h> #include <machine/vmparam.h> -#endif #include <contrib/dev/acpica/include/acpi.h> @@ -73,19 +71,13 @@ extern int acpi_resume_beep; extern int acpi_reset_video; extern int acpi_susp_bounce; -#ifdef SMP extern struct susppcb **susppcbs; static cpuset_t suspcpus; -#else -static struct susppcb **susppcbs; -#endif static void acpi_stop_beep(void *); -#ifdef SMP static int acpi_wakeup_ap(struct acpi_softc *, int); static void acpi_wakeup_cpus(struct acpi_softc *); -#endif #define ACPI_WAKEPT_PAGES 7 @@ -103,7 +95,6 @@ acpi_stop_beep(void *arg) timer_spkr_release(); } -#ifdef SMP static int acpi_wakeup_ap(struct acpi_softc *sc, int cpu) { @@ -177,7 +168,6 @@ acpi_wakeup_cpus(struct acpi_softc *sc) outb(CMOS_DATA, mpbiosreason); } } -#endif int acpi_sleep_machdep(struct acpi_softc *sc, int state) @@ -190,10 +180,8 @@ acpi_sleep_machdep(struct acpi_softc *sc, int state) if (sc->acpi_wakeaddr == 0ul) return (-1); /* couldn't alloc wake memory */ -#ifdef SMP suspcpus = all_cpus; CPU_CLR(PCPU_GET(cpuid), &suspcpus); -#endif if (acpi_resume_beep != 0) timer_spkr_acquire(); @@ -208,12 +196,10 @@ acpi_sleep_machdep(struct acpi_softc *sc, int state) pcb = &susppcbs[0]->sp_pcb; if (savectx(pcb)) { fpususpend(susppcbs[0]->sp_fpususpend); -#ifdef SMP if (!CPU_EMPTY(&suspcpus) && suspend_cpus(suspcpus) == 0) { device_printf(sc->acpi_dev, "Failed to suspend APs\n"); return (0); /* couldn't sleep */ } -#endif hw_ibrs_ibpb_active = 0; hw_ssb_active = 0; cpu_stdext_feature3 = 0; @@ -278,16 +264,12 @@ acpi_wakeup_machdep(struct acpi_softc *sc, int state, int sleep_result, PCPU_SET(switchtime, 0); PCPU_SET(switchticks, ticks); lapic_xapic_mode(); -#ifdef SMP if (!CPU_EMPTY(&suspcpus)) acpi_wakeup_cpus(sc); -#endif } -#ifdef SMP if (!CPU_EMPTY(&suspcpus)) resume_cpus(suspcpus); -#endif /* * Re-read cpu_stdext_feature3, which was zeroed-out diff --git a/sys/amd64/amd64/apic_vector.S b/sys/amd64/amd64/apic_vector.S index 6e51ebff298a..e98bae9eb6c5 100644 --- a/sys/amd64/amd64/apic_vector.S +++ b/sys/amd64/amd64/apic_vector.S @@ -49,12 +49,6 @@ #include <machine/specialreg.h> #include <x86/apicreg.h> -#ifdef SMP -#define LK lock ; -#else -#define LK -#endif - .text SUPERALIGN_TEXT /* End Of Interrupt to APIC */ @@ -163,7 +157,6 @@ IDTVEC(spuriousint) jmp doreti #endif -#ifdef SMP /* * Global address space TLB shootdown. */ @@ -270,5 +263,3 @@ IDTVEC(justreturn) INTR_HANDLER justreturn1 call as_lapic_eoi jmp doreti - -#endif /* SMP */ diff --git a/sys/amd64/amd64/cpu_switch.S b/sys/amd64/amd64/cpu_switch.S index a053f6c70af1..d7e954f573b0 100644 --- a/sys/amd64/amd64/cpu_switch.S +++ b/sys/amd64/amd64/cpu_switch.S @@ -136,7 +136,7 @@ ctx_switch_fpusave_done: movq %r15,TD_LOCK(%r13) /* Release the old thread */ sw1: leaq TD_MD_PCB(%r12),%r8 -#if defined(SCHED_ULE) && defined(SMP) +#if defined(SCHED_ULE) movq $blocked_lock, %rdx movq TD_LOCK(%r12),%rcx cmpq %rcx, %rdx @@ -492,7 +492,7 @@ ENTRY(resumectx) END(resumectx) /* Wait for the new thread to become unblocked */ -#if defined(SCHED_ULE) && defined(SMP) +#if defined(SCHED_ULE) sw1wait: 1: pause diff --git a/sys/amd64/amd64/exec_machdep.c b/sys/amd64/amd64/exec_machdep.c index da68289e2c83..6752b716deb5 100644 --- a/sys/amd64/amd64/exec_machdep.c +++ b/sys/amd64/amd64/exec_machdep.c @@ -59,9 +59,7 @@ #include <sys/reg.h> #include <sys/rwlock.h> #include <sys/signalvar.h> -#ifdef SMP #include <sys/smp.h> -#endif #include <sys/syscallsubr.h> #include <sys/sysctl.h> #include <sys/sysent.h> diff --git a/sys/amd64/amd64/machdep.c b/sys/amd64/amd64/machdep.c index f46462b39fa3..37c7056f649c 100644 --- a/sys/amd64/amd64/machdep.c +++ b/sys/amd64/amd64/machdep.c @@ -38,7 +38,6 @@ * SUCH DAMAGE. */ -#include <sys/cdefs.h> #include "opt_atpic.h" #include "opt_cpu.h" #include "opt_ddb.h" @@ -82,9 +81,7 @@ #include <sys/rwlock.h> #include <sys/sched.h> #include <sys/signalvar.h> -#ifdef SMP #include <sys/smp.h> -#endif #include <sys/syscallsubr.h> #include <sys/sysctl.h> #include <sys/sysent.h> @@ -132,9 +129,7 @@ #include <machine/tss.h> #include <x86/ucode.h> #include <x86/ifunc.h> -#ifdef SMP #include <machine/smp.h> -#endif #ifdef FDT #include <x86/fdt.h> #endif @@ -149,6 +144,10 @@ #include <isa/rtc.h> #include <x86/init.h> +#ifndef SMP +#error amd64 requires options SMP +#endif + /* Sanity check for __curthread() */ CTASSERT(offsetof(struct pcpu, pc_curthread) == 0); diff --git a/sys/amd64/amd64/mem.c b/sys/amd64/amd64/mem.c index 413b7c74890e..851f2df0e6e1 100644 --- a/sys/amd64/amd64/mem.c +++ b/sys/amd64/amd64/mem.c @@ -105,8 +105,8 @@ memrw(struct cdev *dev, struct uio *uio, int flags) * PAGE_SIZE, the uiomove() call does not * access past the end of the direct map. */ - if (v >= DMAP_MIN_ADDRESS && - v < DMAP_MIN_ADDRESS + dmaplimit) { + if (v >= kva_layout.dmap_low && + v < kva_layout.dmap_high) { error = uiomove((void *)v, c, uio); break; } diff --git a/sys/amd64/amd64/minidump_machdep.c b/sys/amd64/amd64/minidump_machdep.c index 6d0917e16099..43bf81a991bf 100644 --- a/sys/amd64/amd64/minidump_machdep.c +++ b/sys/amd64/amd64/minidump_machdep.c @@ -186,7 +186,7 @@ cpu_minidumpsys(struct dumperinfo *di, const struct minidumpstate *state) * tables, so care must be taken to read each entry only once. */ pmapsize = 0; - for (va = VM_MIN_KERNEL_ADDRESS; va < kva_end; ) { + for (va = kva_layout.km_low; va < kva_end; ) { /* * We always write a page, even if it is zero. Each * page written corresponds to 1GB of space @@ -279,9 +279,9 @@ cpu_minidumpsys(struct dumperinfo *di, const struct minidumpstate *state) mdhdr.msgbufsize = mbp->msg_size; mdhdr.bitmapsize = round_page(BITSET_SIZE(vm_page_dump_pages)); mdhdr.pmapsize = pmapsize; - mdhdr.kernbase = VM_MIN_KERNEL_ADDRESS; - mdhdr.dmapbase = DMAP_MIN_ADDRESS; - mdhdr.dmapend = DMAP_MAX_ADDRESS; + mdhdr.kernbase = kva_layout.km_low; + mdhdr.dmapbase = kva_layout.dmap_low; + mdhdr.dmapend = kva_layout.dmap_high; mdhdr.dumpavailsize = round_page(sizeof(dump_avail)); dump_init_header(di, &kdh, KERNELDUMPMAGIC, KERNELDUMP_AMD64_VERSION, @@ -323,7 +323,7 @@ cpu_minidumpsys(struct dumperinfo *di, const struct minidumpstate *state) /* Dump kernel page directory pages */ bzero(fakepd, sizeof(fakepd)); - for (va = VM_MIN_KERNEL_ADDRESS; va < kva_end; va += NBPDP) { + for (va = kva_layout.km_low; va < kva_end; va += NBPDP) { ii = pmap_pml4e_index(va); pml4 = (uint64_t *)PHYS_TO_DMAP(KPML4phys) + ii; pdp = (uint64_t *)PHYS_TO_DMAP(*pml4 & PG_FRAME); diff --git a/sys/amd64/amd64/pmap.c b/sys/amd64/amd64/pmap.c index 9c985df13ddf..8df082f6c5dc 100644 --- a/sys/amd64/amd64/pmap.c +++ b/sys/amd64/amd64/pmap.c @@ -162,9 +162,7 @@ #include <machine/msan.h> #include <machine/pcb.h> #include <machine/specialreg.h> -#ifdef SMP #include <machine/smp.h> -#endif #include <machine/sysarch.h> #include <machine/tss.h> @@ -415,7 +413,7 @@ SYSCTL_INT(_machdep, OID_AUTO, nkpt, CTLFLAG_RD, &nkpt, 0, static int ndmpdp; vm_paddr_t dmaplimit; -vm_offset_t kernel_vm_end = VM_MIN_KERNEL_ADDRESS; +vm_offset_t kernel_vm_end = VM_MIN_KERNEL_ADDRESS_LA48; pt_entry_t pg_nx; static SYSCTL_NODE(_vm, OID_AUTO, pmap, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, @@ -475,11 +473,56 @@ _Static_assert(DMPML4I + NDMPML4E <= KMSANSHADPML4I, "direct map overflow"); static pml4_entry_t *kernel_pml4; static u_int64_t DMPDphys; /* phys addr of direct mapped level 2 */ static u_int64_t DMPDPphys; /* phys addr of direct mapped level 3 */ +static u_int64_t DMPML4phys; /* ... level 4, for la57 */ static int ndmpdpphys; /* number of DMPDPphys pages */ vm_paddr_t kernphys; /* phys addr of start of bootstrap data */ vm_paddr_t KERNend; /* and the end */ +struct kva_layout_s kva_layout = { + .kva_min = KV4ADDR(PML4PML4I, 0, 0, 0), + .kva_max = KV4ADDR(NPML4EPG - 1, NPDPEPG - 1, + NPDEPG - 1, NPTEPG - 1), + .dmap_low = KV4ADDR(DMPML4I, 0, 0, 0), + .dmap_high = KV4ADDR(DMPML4I + NDMPML4E, 0, 0, 0), + .lm_low = KV4ADDR(LMSPML4I, 0, 0, 0), + .lm_high = KV4ADDR(LMEPML4I + 1, 0, 0, 0), + .km_low = KV4ADDR(KPML4BASE, 0, 0, 0), + .km_high = KV4ADDR(KPML4BASE + NKPML4E - 1, NPDPEPG - 1, + NPDEPG - 1, NPTEPG - 1), + .rec_pt = KV4ADDR(PML4PML4I, 0, 0, 0), + .kasan_shadow_low = KV4ADDR(KASANPML4I, 0, 0, 0), + .kasan_shadow_high = KV4ADDR(KASANPML4I + NKASANPML4E, 0, 0, 0), + .kmsan_shadow_low = KV4ADDR(KMSANSHADPML4I, 0, 0, 0), + .kmsan_shadow_high = KV4ADDR(KMSANSHADPML4I + NKMSANSHADPML4E, + 0, 0, 0), + .kmsan_origin_low = KV4ADDR(KMSANORIGPML4I, 0, 0, 0), + .kmsan_origin_high = KV4ADDR(KMSANORIGPML4I + NKMSANORIGPML4E, + 0, 0, 0), +}; + +struct kva_layout_s kva_layout_la57 = { + .kva_min = KV5ADDR(NPML5EPG / 2, 0, 0, 0, 0), /* == rec_pt */ + .kva_max = KV5ADDR(NPML5EPG - 1, NPML4EPG - 1, NPDPEPG - 1, + NPDEPG - 1, NPTEPG - 1), + .dmap_low = KV5ADDR(DMPML5I, 0, 0, 0, 0), + .dmap_high = KV5ADDR(DMPML5I + NDMPML5E, 0, 0, 0, 0), + .lm_low = KV5ADDR(LMSPML5I, 0, 0, 0, 0), + .lm_high = KV5ADDR(LMEPML5I + 1, 0, 0, 0, 0), + .km_low = KV4ADDR(KPML4BASE, 0, 0, 0), + .km_high = KV4ADDR(KPML4BASE + NKPML4E - 1, NPDPEPG - 1, + NPDEPG - 1, NPTEPG - 1), + .rec_pt = KV5ADDR(PML5PML5I, 0, 0, 0, 0), + .kasan_shadow_low = KV4ADDR(KASANPML4I, 0, 0, 0), + .kasan_shadow_high = KV4ADDR(KASANPML4I + NKASANPML4E, 0, 0, 0), + .kmsan_shadow_low = KV4ADDR(KMSANSHADPML4I, 0, 0, 0), + .kmsan_shadow_high = KV4ADDR(KMSANSHADPML4I + NKMSANSHADPML4E, + 0, 0, 0), + .kmsan_origin_low = KV4ADDR(KMSANORIGPML4I, 0, 0, 0), + .kmsan_origin_high = KV4ADDR(KMSANORIGPML4I + NKMSANORIGPML4E, + 0, 0, 0), +}; + /* * pmap_mapdev support pre initialization (i.e. console) */ @@ -549,8 +592,8 @@ static int pmap_flags = PMAP_PDE_SUPERPAGE; /* flags for x86 pmaps */ static vmem_t *large_vmem; static u_int lm_ents; -#define PMAP_ADDRESS_IN_LARGEMAP(va) ((va) >= LARGEMAP_MIN_ADDRESS && \ - (va) < LARGEMAP_MIN_ADDRESS + NBPML4 * (u_long)lm_ents) +#define PMAP_ADDRESS_IN_LARGEMAP(va) ((va) >= kva_layout.lm_low && \ + (va) < kva_layout.lm_high) int pmap_pcid_enabled = 1; SYSCTL_INT(_vm_pmap, OID_AUTO, pcid_enabled, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, @@ -1336,7 +1379,7 @@ static pdp_entry_t *pmap_pti_pdpe(vm_offset_t va); static pd_entry_t *pmap_pti_pde(vm_offset_t va); static void pmap_pti_wire_pte(void *pte); static int pmap_remove_pde(pmap_t pmap, pd_entry_t *pdq, vm_offset_t sva, - bool remove_pt, struct spglist *free, struct rwlock **lockp); + bool demote_kpde, struct spglist *free, struct rwlock **lockp); static int pmap_remove_pte(pmap_t pmap, pt_entry_t *ptq, vm_offset_t sva, pd_entry_t ptepde, struct spglist *free, struct rwlock **lockp); static vm_page_t pmap_remove_pt_page(pmap_t pmap, vm_offset_t va); @@ -1722,7 +1765,7 @@ create_pagetables(vm_paddr_t *firstaddr) { pd_entry_t *pd_p; pdp_entry_t *pdp_p; - pml4_entry_t *p4_p; + pml4_entry_t *p4_p, *p4d_p; pml5_entry_t *p5_p; uint64_t DMPDkernphys; vm_paddr_t pax; @@ -1732,7 +1775,7 @@ create_pagetables(vm_paddr_t *firstaddr) vm_offset_t kasankernbase; int kasankpdpi, kasankpdi, nkasanpte; #endif - int i, j, ndm1g, nkpdpe, nkdmpde; + int i, j, ndm1g, nkpdpe, nkdmpde, ndmpml4phys; TSENTER(); /* Allocate page table pages for the direct map */ @@ -1740,15 +1783,30 @@ create_pagetables(vm_paddr_t *firstaddr) if (ndmpdp < 4) /* Minimum 4GB of dirmap */ ndmpdp = 4; ndmpdpphys = howmany(ndmpdp, NPDPEPG); - if (ndmpdpphys > NDMPML4E) { - /* - * Each NDMPML4E allows 512 GB, so limit to that, - * and then readjust ndmpdp and ndmpdpphys. - */ - printf("NDMPML4E limits system to %d GB\n", NDMPML4E * 512); - Maxmem = atop(NDMPML4E * NBPML4); - ndmpdpphys = NDMPML4E; - ndmpdp = NDMPML4E * NPDEPG; + if (la57) { + ndmpml4phys = howmany(ndmpdpphys, NPML4EPG); + if (ndmpml4phys > NDMPML5E) { + printf("NDMPML5E limits system to %ld GB\n", + (u_long)NDMPML5E * NBPML5 / 1024 / 1024 / 1024); + Maxmem = atop(NDMPML5E * NBPML5); + ndmpml4phys = NDMPML5E; + ndmpdpphys = ndmpml4phys * NPML4EPG; + ndmpdp = ndmpdpphys * NPDEPG; + } + DMPML4phys = allocpages(firstaddr, ndmpml4phys); + } else { + if (ndmpdpphys > NDMPML4E) { + /* + * Each NDMPML4E allows 512 GB, so limit to + * that, and then readjust ndmpdp and + * ndmpdpphys. + */ + printf("NDMPML4E limits system to %d GB\n", + NDMPML4E * 512); + Maxmem = atop(NDMPML4E * NBPML4); + ndmpdpphys = NDMPML4E; + ndmpdp = NDMPML4E * NPDEPG; + } } DMPDPphys = allocpages(firstaddr, ndmpdpphys); ndm1g = 0; @@ -1773,7 +1831,13 @@ create_pagetables(vm_paddr_t *firstaddr) dmaplimit = (vm_paddr_t)ndmpdp << PDPSHIFT; /* Allocate pages. */ + if (la57) { + KPML5phys = allocpages(firstaddr, 1); + p5_p = (pml5_entry_t *)KPML5phys; + } KPML4phys = allocpages(firstaddr, 1); + p4_p = (pml4_entry_t *)KPML4phys; + KPDPphys = allocpages(firstaddr, NKPML4E); #ifdef KASAN KASANPDPphys = allocpages(firstaddr, NKASANPML4E); @@ -1893,6 +1957,16 @@ create_pagetables(vm_paddr_t *firstaddr) } /* + * Connect the Direct Map slots up to the PML4. + * pml5 entries for DMAP are handled below in global pml5 loop. + */ + p4d_p = la57 ? (pml4_entry_t *)DMPML4phys : &p4_p[DMPML4I]; + for (i = 0; i < ndmpdpphys; i++) { + p4d_p[i] = (DMPDPphys + ptoa(i)) | X86_PG_RW | X86_PG_V | + pg_nx; + } + + /* * Instead of using a 1G page for the memory containing the kernel, * use 2M pages with read-only and no-execute permissions. (If using 1G * pages, this will partially overwrite the PDPEs above.) @@ -1911,11 +1985,6 @@ create_pagetables(vm_paddr_t *firstaddr) } } - /* And recursively map PML4 to itself in order to get PTmap */ - p4_p = (pml4_entry_t *)KPML4phys; - p4_p[PML4PML4I] = KPML4phys; - p4_p[PML4PML4I] |= X86_PG_RW | X86_PG_V | pg_nx; - #ifdef KASAN /* Connect the KASAN shadow map slots up to the PML4. */ for (i = 0; i < NKASANPML4E; i++) { @@ -1938,25 +2007,15 @@ create_pagetables(vm_paddr_t *firstaddr) } #endif - /* Connect the Direct Map slots up to the PML4. */ - for (i = 0; i < ndmpdpphys; i++) { - p4_p[DMPML4I + i] = DMPDPphys + ptoa(i); - p4_p[DMPML4I + i] |= X86_PG_RW | X86_PG_V | pg_nx; - } - /* Connect the KVA slots up to the PML4 */ for (i = 0; i < NKPML4E; i++) { p4_p[KPML4BASE + i] = KPDPphys + ptoa(i); p4_p[KPML4BASE + i] |= X86_PG_RW | X86_PG_V; } - kernel_pml4 = (pml4_entry_t *)PHYS_TO_DMAP(KPML4phys); - if (la57) { /* XXXKIB bootstrap KPML5phys page is lost */ - KPML5phys = allocpages(firstaddr, 1); - for (i = 0, p5_p = (pml5_entry_t *)KPML5phys; i < NPML5EPG; - i++) { + for (i = 0; i < NPML5EPG; i++) { if (i == PML5PML5I) { /* * Recursively map PML5 to itself in @@ -1964,6 +2023,10 @@ create_pagetables(vm_paddr_t *firstaddr) */ p5_p[i] = KPML5phys | X86_PG_RW | X86_PG_A | X86_PG_M | X86_PG_V | pg_nx; + } else if (i >= DMPML5I && i < DMPML5I + ndmpml4phys) { + /* Connect DMAP pml4 pages to PML5. */ + p5_p[i] = (DMPML4phys + ptoa(i - DMPML5I)) | + X86_PG_RW | X86_PG_V | pg_nx; } else if (i == pmap_pml5e_index(UPT_MAX_ADDRESS)) { p5_p[i] = KPML4phys | X86_PG_RW | X86_PG_A | X86_PG_M | X86_PG_V; @@ -1971,6 +2034,10 @@ create_pagetables(vm_paddr_t *firstaddr) p5_p[i] = 0; } } + } else { + /* Recursively map PML4 to itself in order to get PTmap */ + p4_p[PML4PML4I] = KPML4phys; + p4_p[PML4PML4I] |= X86_PG_RW | X86_PG_V | pg_nx; } TSEXIT(); } @@ -2024,7 +2091,7 @@ pmap_bootstrap(vm_paddr_t *firstaddr) */ virtual_avail = (vm_offset_t)KERNSTART + round_2mpage(KERNend - (vm_paddr_t)kernphys); - virtual_end = VM_MAX_KERNEL_ADDRESS; + virtual_end = kva_layout.km_high; /* * Enable PG_G global pages, then switch to the kernel page @@ -2046,9 +2113,13 @@ pmap_bootstrap(vm_paddr_t *firstaddr) * Initialize the kernel pmap (which is statically allocated). * Count bootstrap data as being resident in case any of this data is * later unmapped (using pmap_remove()) and freed. + * + * DMAP_TO_PHYS()/PHYS_TO_DMAP() are functional only after + * kva_layout is fixed. */ PMAP_LOCK_INIT(kernel_pmap); if (la57) { + kva_layout = kva_layout_la57; vtoptem = ((1ul << (NPTEPGSHIFT + NPDEPGSHIFT + NPDPEPGSHIFT + NPML4EPGSHIFT + NPML5EPGSHIFT)) - 1) << 3; PTmap = (vm_offset_t)P5Tmap; @@ -2059,6 +2130,7 @@ pmap_bootstrap(vm_paddr_t *firstaddr) kernel_pmap->pm_cr3 = KPML5phys; pmap_pt_page_count_adj(kernel_pmap, 1); /* top-level page */ } else { + kernel_pml4 = (pml4_entry_t *)PHYS_TO_DMAP(KPML4phys); kernel_pmap->pm_pmltop = kernel_pml4; kernel_pmap->pm_cr3 = KPML4phys; } @@ -2420,6 +2492,8 @@ pmap_init(void) { struct pmap_preinit_mapping *ppim; vm_page_t m, mpte; + pml4_entry_t *pml4e; + unsigned long lm_max; int error, i, ret, skz63; /* L1TF, reserve page @0 unconditionally */ @@ -2545,10 +2619,15 @@ pmap_init(void) lm_ents = 8; TUNABLE_INT_FETCH("vm.pmap.large_map_pml4_entries", &lm_ents); - if (lm_ents > LMEPML4I - LMSPML4I + 1) - lm_ents = LMEPML4I - LMSPML4I + 1; + lm_max = (kva_layout.lm_high - kva_layout.lm_low) / NBPML4; + if (lm_ents > lm_max) { + printf( + "pmap: shrinking large map from requested %d slots to %ld slots\n", + lm_ents, lm_max); + lm_ents = lm_max; + } #ifdef KMSAN - if (lm_ents > KMSANORIGPML4I - LMSPML4I) { + if (!la57 && lm_ents > KMSANORIGPML4I - LMSPML4I) { printf( "pmap: shrinking large map for KMSAN (%d slots to %ld slots)\n", lm_ents, KMSANORIGPML4I - LMSPML4I); @@ -2559,18 +2638,27 @@ pmap_init(void) printf("pmap: large map %u PML4 slots (%lu GB)\n", lm_ents, (u_long)lm_ents * (NBPML4 / 1024 / 1024 / 1024)); if (lm_ents != 0) { - large_vmem = vmem_create("large", LARGEMAP_MIN_ADDRESS, + large_vmem = vmem_create("large", kva_layout.lm_low, (vmem_size_t)lm_ents * NBPML4, PAGE_SIZE, 0, M_WAITOK); if (large_vmem == NULL) { printf("pmap: cannot create large map\n"); lm_ents = 0; } + if (la57) { + for (i = 0; i < howmany((vm_offset_t)NBPML4 * + lm_ents, NBPML5); i++) { + m = pmap_large_map_getptp_unlocked(); + kernel_pmap->pm_pmltop[LMSPML5I + i] = X86_PG_V | + X86_PG_RW | X86_PG_A | X86_PG_M | + pg_nx | VM_PAGE_TO_PHYS(m); + } + } for (i = 0; i < lm_ents; i++) { m = pmap_large_map_getptp_unlocked(); - /* XXXKIB la57 */ - kernel_pml4[LMSPML4I + i] = X86_PG_V | - X86_PG_RW | X86_PG_A | X86_PG_M | pg_nx | - VM_PAGE_TO_PHYS(m); + pml4e = pmap_pml4e(kernel_pmap, kva_layout.lm_low + + (u_long)i * NBPML4); + *pml4e = X86_PG_V | X86_PG_RW | X86_PG_A | X86_PG_M | + pg_nx | VM_PAGE_TO_PHYS(m); } } } @@ -2975,7 +3063,6 @@ pmap_update_pde_invalidate(pmap_t pmap, vm_offset_t va, pd_entry_t newpde) * XXX TODO */ -#ifdef SMP /* * Interrupt the cpus that are executing in the guest context. * This will force the vcpu to exit and the cached EPT mappings @@ -3433,168 +3520,6 @@ pmap_update_pde(pmap_t pmap, vm_offset_t va, pd_entry_t *pde, pd_entry_t newpde) } sched_unpin(); } -#else /* !SMP */ -/* - * Normal, non-SMP, invalidation functions. - */ -void -pmap_invalidate_page(pmap_t pmap, vm_offset_t va) -{ - struct invpcid_descr d; - struct pmap_pcid *pcidp; - uint64_t kcr3, ucr3; - uint32_t pcid; - - if (pmap->pm_type == PT_RVI || pmap->pm_type == PT_EPT) { - pmap->pm_eptgen++; - return; - } - KASSERT(pmap->pm_type == PT_X86, - ("pmap_invalidate_range: unknown type %d", pmap->pm_type)); - - if (pmap == kernel_pmap || pmap == PCPU_GET(curpmap)) { - invlpg(va); - if (pmap == PCPU_GET(curpmap) && pmap_pcid_enabled && - pmap->pm_ucr3 != PMAP_NO_CR3) { - critical_enter(); - pcid = pmap_get_pcid(pmap); - if (invpcid_works) { - d.pcid = pcid | PMAP_PCID_USER_PT; - d.pad = 0; - d.addr = va; - invpcid(&d, INVPCID_ADDR); - } else { - kcr3 = pmap->pm_cr3 | pcid | CR3_PCID_SAVE; - ucr3 = pmap->pm_ucr3 | pcid | - PMAP_PCID_USER_PT | CR3_PCID_SAVE; - pmap_pti_pcid_invlpg(ucr3, kcr3, va); - } - critical_exit(); - } - } else if (pmap_pcid_enabled) { - pcidp = zpcpu_get(pmap->pm_pcidp); - pcidp->pm_gen = 0; - } -} - -void -pmap_invalidate_range(pmap_t pmap, vm_offset_t sva, vm_offset_t eva) -{ - struct invpcid_descr d; - struct pmap_pcid *pcidp; - vm_offset_t addr; - uint64_t kcr3, ucr3; - uint32_t pcid; - - if (pmap->pm_type == PT_RVI || pmap->pm_type == PT_EPT) { - pmap->pm_eptgen++; - return; - } - KASSERT(pmap->pm_type == PT_X86, - ("pmap_invalidate_range: unknown type %d", pmap->pm_type)); - - if (pmap == kernel_pmap || pmap == PCPU_GET(curpmap)) { - for (addr = sva; addr < eva; addr += PAGE_SIZE) - invlpg(addr); - if (pmap == PCPU_GET(curpmap) && pmap_pcid_enabled && - pmap->pm_ucr3 != PMAP_NO_CR3) { - critical_enter(); - pcid = pmap_get_pcid(pmap); - if (invpcid_works) { - d.pcid = pcid | PMAP_PCID_USER_PT; - d.pad = 0; - d.addr = sva; - for (; d.addr < eva; d.addr += PAGE_SIZE) - invpcid(&d, INVPCID_ADDR); - } else { - kcr3 = pmap->pm_cr3 | pcid | CR3_PCID_SAVE; - ucr3 = pmap->pm_ucr3 | pcid | - PMAP_PCID_USER_PT | CR3_PCID_SAVE; - pmap_pti_pcid_invlrng(ucr3, kcr3, sva, eva); - } - critical_exit(); - } - } else if (pmap_pcid_enabled) { - pcidp = zpcpu_get(pmap->pm_pcidp); - pcidp->pm_gen = 0; - } -} - -void -pmap_invalidate_all(pmap_t pmap) -{ - struct invpcid_descr d; - struct pmap_pcid *pcidp; - uint64_t kcr3, ucr3; - uint32_t pcid; - - if (pmap->pm_type == PT_RVI || pmap->pm_type == PT_EPT) { - pmap->pm_eptgen++; - return; - } - KASSERT(pmap->pm_type == PT_X86, - ("pmap_invalidate_all: unknown type %d", pmap->pm_type)); - - if (pmap == kernel_pmap) { - if (pmap_pcid_enabled && invpcid_works) { - bzero(&d, sizeof(d)); - invpcid(&d, INVPCID_CTXGLOB); - } else { - invltlb_glob(); - } - } else if (pmap == PCPU_GET(curpmap)) { - if (pmap_pcid_enabled) { - critical_enter(); - pcid = pmap_get_pcid(pmap); - if (invpcid_works) { - d.pcid = pcid; - d.pad = 0; - d.addr = 0; - invpcid(&d, INVPCID_CTX); - if (pmap->pm_ucr3 != PMAP_NO_CR3) { - d.pcid |= PMAP_PCID_USER_PT; - invpcid(&d, INVPCID_CTX); - } - } else { - kcr3 = pmap->pm_cr3 | pcid; - if (pmap->pm_ucr3 != PMAP_NO_CR3) { - ucr3 = pmap->pm_ucr3 | pcid | - PMAP_PCID_USER_PT; - pmap_pti_pcid_invalidate(ucr3, kcr3); - } else - load_cr3(kcr3); - } - critical_exit(); - } else { - invltlb(); - } - } else if (pmap_pcid_enabled) { - pcidp = zpcpu_get(pmap->pm_pcidp); - pcidp->pm_gen = 0; - } -} - -void -pmap_invalidate_cache(void) -{ - - wbinvd(); -} - -static void -pmap_update_pde(pmap_t pmap, vm_offset_t va, pd_entry_t *pde, pd_entry_t newpde) -{ - struct pmap_pcid *pcidp; - - pmap_update_pde_store(pmap, pde, newpde); - if (pmap == kernel_pmap || pmap == PCPU_GET(curpmap)) - pmap_update_pde_invalidate(pmap, va, newpde); - else { - pcidp = zpcpu_get(pmap->pm_pcidp); - pcidp->pm_gen = 0; - } -} -#endif /* !SMP */ static void pmap_invalidate_pde_page(pmap_t pmap, vm_offset_t va, pd_entry_t pde) @@ -3899,7 +3824,7 @@ pmap_kextract(vm_offset_t va) pd_entry_t pde; vm_paddr_t pa; - if (va >= DMAP_MIN_ADDRESS && va < DMAP_MAX_ADDRESS) { + if (va >= kva_layout.dmap_low && va < kva_layout.dmap_high) { pa = DMAP_TO_PHYS(va); } else if (PMAP_ADDRESS_IN_LARGEMAP(va)) { pa = pmap_large_map_kextract(va); @@ -4040,7 +3965,7 @@ pmap_qremove(vm_offset_t sva, int count) * enough to one of those pmap_enter() calls for it to * be caught up in a promotion. */ - KASSERT(va >= VM_MIN_KERNEL_ADDRESS, ("usermode va %lx", va)); + KASSERT(va >= kva_layout.km_low, ("usermode va %lx", va)); KASSERT((*vtopde(va) & X86_PG_PS) == 0, ("pmap_qremove on promoted va %#lx", va)); @@ -4328,21 +4253,13 @@ void pmap_pinit_pml5(vm_page_t pml5pg) { pml5_entry_t *pm_pml5; + int i; pm_pml5 = (pml5_entry_t *)PHYS_TO_DMAP(VM_PAGE_TO_PHYS(pml5pg)); - - /* - * Add pml5 entry at top of KVA pointing to existing pml4 table, - * entering all existing kernel mappings into level 5 table. - */ - pm_pml5[pmap_pml5e_index(UPT_MAX_ADDRESS)] = KPML4phys | X86_PG_V | - X86_PG_RW | X86_PG_A | X86_PG_M; - - /* - * Install self-referential address mapping entry. - */ - pm_pml5[PML5PML5I] = VM_PAGE_TO_PHYS(pml5pg) | - X86_PG_RW | X86_PG_V | X86_PG_M | X86_PG_A; + for (i = 0; i < NPML5EPG / 2; i++) + pm_pml5[i] = 0; + for (; i < NPML5EPG; i++) + pm_pml5[i] = kernel_pmap->pm_pmltop[i]; } static void @@ -4899,8 +4816,8 @@ pmap_release(pmap_t pmap) m = PHYS_TO_VM_PAGE(DMAP_TO_PHYS((vm_offset_t)pmap->pm_pmltop)); if (pmap_is_la57(pmap)) { - pmap->pm_pmltop[pmap_pml5e_index(UPT_MAX_ADDRESS)] = 0; - pmap->pm_pmltop[PML5PML5I] = 0; + for (i = NPML5EPG / 2; i < NPML5EPG; i++) + pmap->pm_pmltop[i] = 0; } else { for (i = 0; i < NKPML4E; i++) /* KVA */ pmap->pm_pmltop[KPML4BASE + i] = 0; @@ -4942,7 +4859,7 @@ pmap_release(pmap_t pmap) static int kvm_size(SYSCTL_HANDLER_ARGS) { - unsigned long ksize = VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS; + unsigned long ksize = kva_layout.km_high - kva_layout.km_low; return sysctl_handle_long(oidp, &ksize, 0, req); } @@ -4953,7 +4870,7 @@ SYSCTL_PROC(_vm, OID_AUTO, kvm_size, CTLTYPE_LONG | CTLFLAG_RD | CTLFLAG_MPSAFE, static int kvm_free(SYSCTL_HANDLER_ARGS) { - unsigned long kfree = VM_MAX_KERNEL_ADDRESS - kernel_vm_end; + unsigned long kfree = kva_layout.km_high - kernel_vm_end; return sysctl_handle_long(oidp, &kfree, 0, req); } @@ -5031,7 +4948,7 @@ pmap_page_array_startup(long pages) vm_page_array_size = pages; - start = VM_MIN_KERNEL_ADDRESS; + start = kva_layout.km_low; end = start + pages * sizeof(struct vm_page); for (va = start; va < end; va += NBPDR) { pfn = first_page + (va - start) / sizeof(struct vm_page); @@ -6045,17 +5962,18 @@ pmap_demote_pde_mpte(pmap_t pmap, pd_entry_t *pde, vm_offset_t va, if (mpte == NULL) { /* * Invalidate the 2MB page mapping and return "failure" if the - * mapping was never accessed. + * mapping was never accessed and not wired. */ if ((oldpde & PG_A) == 0) { - KASSERT((oldpde & PG_W) == 0, - ("pmap_demote_pde: a wired mapping is missing PG_A")); - pmap_demote_pde_abort(pmap, va, pde, oldpde, lockp); - return (false); - } - - mpte = pmap_remove_pt_page(pmap, va); - if (mpte == NULL) { + if ((oldpde & PG_W) == 0) { + pmap_demote_pde_abort(pmap, va, pde, oldpde, + lockp); + return (false); + } + mpte = pmap_remove_pt_page(pmap, va); + /* Fill the PTP with PTEs that have PG_A cleared. */ + mpte->valid = 0; + } else if ((mpte = pmap_remove_pt_page(pmap, va)) == NULL) { KASSERT((oldpde & PG_W) == 0, ("pmap_demote_pde: page table page for a wired mapping is missing")); @@ -6067,8 +5985,8 @@ pmap_demote_pde_mpte(pmap_t pmap, pd_entry_t *pde, vm_offset_t va, * so the direct map region is the only part of the * kernel address space that must be handled here. */ - KASSERT(!in_kernel || (va >= DMAP_MIN_ADDRESS && - va < DMAP_MAX_ADDRESS), + KASSERT(!in_kernel || (va >= kva_layout.dmap_low && + va < kva_layout.dmap_high), ("pmap_demote_pde: No saved mpte for va %#lx", va)); /* @@ -6107,7 +6025,7 @@ pmap_demote_pde_mpte(pmap_t pmap, pd_entry_t *pde, vm_offset_t va, /* * If the PTP is not leftover from an earlier promotion or it does not * have PG_A set in every PTE, then fill it. The new PTEs will all - * have PG_A set. + * have PG_A set, unless this is a wired mapping with PG_A clear. */ if (!vm_page_all_valid(mpte)) pmap_fill_ptp(firstpte, newpte); @@ -6165,8 +6083,7 @@ pmap_demote_pde_mpte(pmap_t pmap, pd_entry_t *pde, vm_offset_t va, * pmap_remove_kernel_pde: Remove a kernel superpage mapping. */ static void -pmap_remove_kernel_pde(pmap_t pmap, pd_entry_t *pde, vm_offset_t va, - bool remove_pt) +pmap_remove_kernel_pde(pmap_t pmap, pd_entry_t *pde, vm_offset_t va) { pd_entry_t newpde; vm_paddr_t mptepa; @@ -6174,12 +6091,8 @@ pmap_remove_kernel_pde(pmap_t pmap, pd_entry_t *pde, vm_offset_t va, KASSERT(pmap == kernel_pmap, ("pmap %p is not kernel_pmap", pmap)); PMAP_LOCK_ASSERT(pmap, MA_OWNED); - if (remove_pt) - mpte = pmap_remove_pt_page(pmap, va); - else - mpte = vm_radix_lookup(&pmap->pm_root, pmap_pde_pindex(va)); - if (mpte == NULL) - panic("pmap_remove_kernel_pde: Missing pt page."); + mpte = pmap_remove_pt_page(pmap, va); + KASSERT(mpte != NULL, ("pmap_remove_kernel_pde: missing pt page")); mptepa = VM_PAGE_TO_PHYS(mpte); newpde = mptepa | X86_PG_M | X86_PG_A | X86_PG_RW | X86_PG_V; @@ -6209,7 +6122,7 @@ pmap_remove_kernel_pde(pmap_t pmap, pd_entry_t *pde, vm_offset_t va, * pmap_remove_pde: do the things to unmap a superpage in a process */ static int -pmap_remove_pde(pmap_t pmap, pd_entry_t *pdq, vm_offset_t sva, bool remove_pt, +pmap_remove_pde(pmap_t pmap, pd_entry_t *pdq, vm_offset_t sva, bool demote_kpde, struct spglist *free, struct rwlock **lockp) { struct md_page *pvh; @@ -6249,9 +6162,7 @@ pmap_remove_pde(pmap_t pmap, pd_entry_t *pdq, vm_offset_t sva, bool remove_pt, pmap_delayed_invl_page(m); } } - if (pmap == kernel_pmap) { - pmap_remove_kernel_pde(pmap, pdq, sva, remove_pt); - } else { + if (pmap != kernel_pmap) { mpte = pmap_remove_pt_page(pmap, sva); if (mpte != NULL) { KASSERT(vm_page_any_valid(mpte), @@ -6262,6 +6173,14 @@ pmap_remove_pde(pmap_t pmap, pd_entry_t *pdq, vm_offset_t sva, bool remove_pt, mpte->ref_count = 0; pmap_add_delayed_free_list(mpte, free, false); } + } else if (demote_kpde) { + pmap_remove_kernel_pde(pmap, pdq, sva); + } else { + mpte = vm_radix_lookup(&pmap->pm_root, pmap_pde_pindex(sva)); + if (vm_page_any_valid(mpte)) { + mpte->valid = 0; + pmap_zero_page(mpte); + } } return (pmap_unuse_pt(pmap, sva, *pmap_pdpe(pmap, sva), free)); } @@ -7183,7 +7102,7 @@ pmap_enter(pmap_t pmap, vm_offset_t va, vm_page_t m, vm_prot_t prot, PG_RW = pmap_rw_bit(pmap); va = trunc_page(va); - KASSERT(va <= VM_MAX_KERNEL_ADDRESS, ("pmap_enter: toobig")); + KASSERT(va <= kva_layout.km_high, ("pmap_enter: toobig")); KASSERT(va < UPT_MIN_ADDRESS || va >= UPT_MAX_ADDRESS, ("pmap_enter: invalid to pmap_enter page table pages (va: 0x%lx)", va)); @@ -7512,6 +7431,9 @@ pmap_enter_pde(pmap_t pmap, vm_offset_t va, pd_entry_t newpde, u_int flags, PG_RW = pmap_rw_bit(pmap); KASSERT((newpde & (pmap_modified_bit(pmap) | PG_RW)) != PG_RW, ("pmap_enter_pde: newpde is missing PG_M")); + KASSERT((flags & (PMAP_ENTER_NOREPLACE | PMAP_ENTER_NORECLAIM)) != + PMAP_ENTER_NORECLAIM, + ("pmap_enter_pde: flags is missing PMAP_ENTER_NOREPLACE")); PG_V = pmap_valid_bit(pmap); PMAP_LOCK_ASSERT(pmap, MA_OWNED); @@ -7573,8 +7495,8 @@ pmap_enter_pde(pmap_t pmap, vm_offset_t va, pd_entry_t newpde, u_int flags, * the mapping is not from kernel_pmap, then * a reserved PT page could be freed. */ - (void)pmap_remove_pde(pmap, pde, va, - pmap != kernel_pmap, &free, lockp); + (void)pmap_remove_pde(pmap, pde, va, false, &free, + lockp); if ((oldpde & PG_G) == 0) pmap_invalidate_pde_page(pmap, va, oldpde); } else { @@ -7584,10 +7506,9 @@ pmap_enter_pde(pmap_t pmap, vm_offset_t va, pd_entry_t newpde, u_int flags, * before any changes to mappings are * made. Abort on failure. */ - mt = PHYS_TO_VM_PAGE(*pde & PG_FRAME); - if (pmap_insert_pt_page(pmap, mt, false, false)) { - if (pdpg != NULL) - pdpg->ref_count--; + mt = PHYS_TO_VM_PAGE(oldpde & PG_FRAME); + if (pmap_insert_pt_page(pmap, mt, false, + false)) { CTR1(KTR_PMAP, "pmap_enter_pde: cannot ins kern ptp va %#lx", va); @@ -7641,6 +7562,14 @@ pmap_enter_pde(pmap_t pmap, vm_offset_t va, pd_entry_t newpde, u_int flags, if (!pmap_pv_insert_pde(pmap, va, newpde, flags, lockp)) { if (pdpg != NULL) pmap_abort_ptp(pmap, va, pdpg); + else { + KASSERT(va >= VM_MAXUSER_ADDRESS && + (*pde & (PG_PS | PG_V)) == PG_V, + ("pmap_enter_pde: invalid kernel PDE")); + mt = pmap_remove_pt_page(pmap, va); + KASSERT(mt != NULL, + ("pmap_enter_pde: missing kernel PTP")); + } if (uwptpg != NULL) { mt = pmap_remove_pt_page(pmap, va); KASSERT(mt == uwptpg, @@ -9550,7 +9479,7 @@ pmap_unmapdev(void *p, vm_size_t size) va = (vm_offset_t)p; /* If we gave a direct map region in pmap_mapdev, do nothing */ - if (va >= DMAP_MIN_ADDRESS && va < DMAP_MAX_ADDRESS) + if (va >= kva_layout.dmap_low && va < kva_layout.dmap_high) return; offset = va & PAGE_MASK; size = round_page(offset + size); @@ -9649,6 +9578,8 @@ pmap_demote_pdpe(pmap_t pmap, pdp_entry_t *pdpe, vm_offset_t va, vm_page_t m) void pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma) { + if (m->md.pat_mode == ma) + return; m->md.pat_mode = ma; @@ -9668,6 +9599,9 @@ pmap_page_set_memattr_noflush(vm_page_t m, vm_memattr_t ma) { int error; + if (m->md.pat_mode == ma) + return; + m->md.pat_mode = ma; if ((m->flags & PG_FICTITIOUS) != 0) @@ -9724,7 +9658,7 @@ pmap_change_prot(vm_offset_t va, vm_size_t size, vm_prot_t prot) int error; /* Only supported within the kernel map. */ - if (va < VM_MIN_KERNEL_ADDRESS) + if (va < kva_layout.km_low) return (EINVAL); PMAP_LOCK(kernel_pmap); @@ -9755,7 +9689,7 @@ pmap_change_props_locked(vm_offset_t va, vm_size_t size, vm_prot_t prot, * Only supported on kernel virtual addresses, including the direct * map but excluding the recursive map. */ - if (base < DMAP_MIN_ADDRESS) + if (base < kva_layout.dmap_low) return (EINVAL); /* @@ -9778,7 +9712,7 @@ pmap_change_props_locked(vm_offset_t va, vm_size_t size, vm_prot_t prot, pte_bits |= X86_PG_RW; } if ((prot & VM_PROT_EXECUTE) == 0 || - va < VM_MIN_KERNEL_ADDRESS) { + va < kva_layout.km_low) { pde_bits |= pg_nx; pte_bits |= pg_nx; } @@ -9874,7 +9808,7 @@ pmap_change_props_locked(vm_offset_t va, vm_size_t size, vm_prot_t prot, pmap_pte_props(pdpe, pde_bits, pde_mask); changed = true; } - if (tmpva >= VM_MIN_KERNEL_ADDRESS && + if (tmpva >= kva_layout.km_low && (*pdpe & PG_PS_FRAME) < dmaplimit) { if (pa_start == pa_end) { /* Start physical address run. */ @@ -9904,7 +9838,7 @@ pmap_change_props_locked(vm_offset_t va, vm_size_t size, vm_prot_t prot, pmap_pte_props(pde, pde_bits, pde_mask); changed = true; } - if (tmpva >= VM_MIN_KERNEL_ADDRESS && + if (tmpva >= kva_layout.km_low && (*pde & PG_PS_FRAME) < dmaplimit) { if (pa_start == pa_end) { /* Start physical address run. */ @@ -9932,7 +9866,7 @@ pmap_change_props_locked(vm_offset_t va, vm_size_t size, vm_prot_t prot, pmap_pte_props(pte, pte_bits, pte_mask); changed = true; } - if (tmpva >= VM_MIN_KERNEL_ADDRESS && + if (tmpva >= kva_layout.km_low && (*pte & PG_FRAME) < dmaplimit) { if (pa_start == pa_end) { /* Start physical address run. */ @@ -10280,17 +10214,9 @@ pmap_activate_sw(struct thread *td) return; } cpuid = PCPU_GET(cpuid); -#ifdef SMP CPU_SET_ATOMIC(cpuid, &pmap->pm_active); -#else - CPU_SET(cpuid, &pmap->pm_active); -#endif pmap_activate_sw_mode(td, pmap, cpuid); -#ifdef SMP CPU_CLR_ATOMIC(cpuid, &oldpmap->pm_active); -#else - CPU_CLR(cpuid, &oldpmap->pm_active); -#endif } void @@ -10331,11 +10257,7 @@ pmap_activate_boot(pmap_t pmap) MPASS(pmap != kernel_pmap); cpuid = PCPU_GET(cpuid); -#ifdef SMP CPU_SET_ATOMIC(cpuid, &pmap->pm_active); -#else - CPU_SET(cpuid, &pmap->pm_active); -#endif PCPU_SET(curpmap, pmap); if (pti) { kcr3 = pmap->pm_cr3; @@ -10699,19 +10621,28 @@ pmap_large_map_getptp(void) static pdp_entry_t * pmap_large_map_pdpe(vm_offset_t va) { + pml4_entry_t *pml4; vm_pindex_t pml4_idx; vm_paddr_t mphys; - pml4_idx = pmap_pml4e_index(va); - KASSERT(LMSPML4I <= pml4_idx && pml4_idx < LMSPML4I + lm_ents, - ("pmap_large_map_pdpe: va %#jx out of range idx %#jx LMSPML4I " - "%#jx lm_ents %d", - (uintmax_t)va, (uintmax_t)pml4_idx, LMSPML4I, lm_ents)); - KASSERT((kernel_pml4[pml4_idx] & X86_PG_V) != 0, - ("pmap_large_map_pdpe: invalid pml4 for va %#jx idx %#jx " - "LMSPML4I %#jx lm_ents %d", - (uintmax_t)va, (uintmax_t)pml4_idx, LMSPML4I, lm_ents)); - mphys = kernel_pml4[pml4_idx] & PG_FRAME; + KASSERT(va >= kva_layout.lm_low && va < kva_layout.lm_low + + (vm_offset_t)NBPML4 * lm_ents, ("va %#lx not in large map", va)); + if (la57) { + pml4 = pmap_pml4e(kernel_pmap, va); + mphys = *pml4 & PG_FRAME; + } else { + pml4_idx = pmap_pml4e_index(va); + + KASSERT(LMSPML4I <= pml4_idx && pml4_idx < LMSPML4I + lm_ents, + ("pmap_large_map_pdpe: va %#jx out of range idx %#jx " + "LMSPML4I %#jx lm_ents %d", + (uintmax_t)va, (uintmax_t)pml4_idx, LMSPML4I, lm_ents)); + KASSERT((kernel_pml4[pml4_idx] & X86_PG_V) != 0, + ("pmap_large_map_pdpe: invalid pml4 for va %#jx idx %#jx " + "LMSPML4I %#jx lm_ents %d", + (uintmax_t)va, (uintmax_t)pml4_idx, LMSPML4I, lm_ents)); + mphys = kernel_pml4[pml4_idx] & PG_FRAME; + } return ((pdp_entry_t *)PHYS_TO_DMAP(mphys) + pmap_pdpe_index(va)); } @@ -10904,8 +10835,8 @@ pmap_large_unmap(void *svaa, vm_size_t len) struct spglist spgf; sva = (vm_offset_t)svaa; - if (len == 0 || sva + len < sva || (sva >= DMAP_MIN_ADDRESS && - sva + len <= DMAP_MIN_ADDRESS + dmaplimit)) + if (len == 0 || sva + len < sva || (sva >= kva_layout.dmap_low && + sva + len < kva_layout.dmap_high)) return; SLIST_INIT(&spgf); @@ -11151,11 +11082,10 @@ pmap_large_map_wb(void *svap, vm_size_t len) sva = (vm_offset_t)svap; eva = sva + len; pmap_large_map_wb_fence(); - if (sva >= DMAP_MIN_ADDRESS && eva <= DMAP_MIN_ADDRESS + dmaplimit) { + if (sva >= kva_layout.dmap_low && eva < kva_layout.dmap_high) { pmap_large_map_flush_range(sva, len); } else { - KASSERT(sva >= LARGEMAP_MIN_ADDRESS && - eva <= LARGEMAP_MIN_ADDRESS + lm_ents * NBPML4, + KASSERT(sva >= kva_layout.lm_low && eva < kva_layout.lm_high, ("pmap_large_map_wb: not largemap %#lx %#lx", sva, len)); pmap_large_map_wb_large(sva, eva); } @@ -11196,8 +11126,8 @@ pmap_pti_init(void) VM_OBJECT_WLOCK(pti_obj); pml4_pg = pmap_pti_alloc_page(); pti_pml4 = (pml4_entry_t *)PHYS_TO_DMAP(VM_PAGE_TO_PHYS(pml4_pg)); - for (va = VM_MIN_KERNEL_ADDRESS; va <= VM_MAX_KERNEL_ADDRESS && - va >= VM_MIN_KERNEL_ADDRESS && va > NBPML4; va += NBPML4) { + for (va = kva_layout.km_low; va <= kva_layout.km_high && + va >= kva_layout.km_low && va > NBPML4; va += NBPML4) { pdpe = pmap_pti_pdpe(va); pmap_pti_wire_pte(pdpe); } @@ -11971,9 +11901,7 @@ sysctl_kmaps_dump(struct sbuf *sb, struct pmap_kernel_map_range *range, mode, range->pdpes, range->pdes, range->ptes); /* Reset to sentinel value. */ - range->sva = la57 ? KV5ADDR(NPML5EPG - 1, NPML4EPG - 1, NPDPEPG - 1, - NPDEPG - 1, NPTEPG - 1) : KV4ADDR(NPML4EPG - 1, NPDPEPG - 1, - NPDEPG - 1, NPTEPG - 1); + range->sva = kva_layout.kva_max; } /* @@ -12014,12 +11942,18 @@ sysctl_kmaps_reinit(struct pmap_kernel_map_range *range, vm_offset_t va, */ static void sysctl_kmaps_check(struct sbuf *sb, struct pmap_kernel_map_range *range, - vm_offset_t va, pml4_entry_t pml4e, pdp_entry_t pdpe, pd_entry_t pde, - pt_entry_t pte) + vm_offset_t va, pml5_entry_t pml5e, pml4_entry_t pml4e, pdp_entry_t pdpe, + pd_entry_t pde, pt_entry_t pte) { pt_entry_t attrs; - attrs = pml4e & (X86_PG_RW | X86_PG_U | pg_nx); + if (la57) { + attrs = pml5e & (X86_PG_RW | X86_PG_U | pg_nx); + attrs |= pml4e & pg_nx; + attrs &= pg_nx | (pml4e & (X86_PG_RW | X86_PG_U)); + } else { + attrs = pml4e & (X86_PG_RW | X86_PG_U | pg_nx); + } attrs |= pdpe & pg_nx; attrs &= pg_nx | (pdpe & (X86_PG_RW | X86_PG_U)); @@ -12052,13 +11986,15 @@ sysctl_kmaps(SYSCTL_HANDLER_ARGS) { struct pmap_kernel_map_range range; struct sbuf sbuf, *sb; + pml5_entry_t pml5e; pml4_entry_t pml4e; pdp_entry_t *pdp, pdpe; pd_entry_t *pd, pde; pt_entry_t *pt, pte; vm_offset_t sva; vm_paddr_t pa; - int error, i, j, k, l; + int error, j, k, l; + bool first; error = sysctl_wire_old_buffer(req, 0); if (error != 0) @@ -12067,9 +12003,8 @@ sysctl_kmaps(SYSCTL_HANDLER_ARGS) sbuf_new_for_sysctl(sb, NULL, PAGE_SIZE, req); /* Sentinel value. */ - range.sva = la57 ? KV5ADDR(NPML5EPG - 1, NPML4EPG - 1, NPDPEPG - 1, - NPDEPG - 1, NPTEPG - 1) : KV4ADDR(NPML4EPG - 1, NPDPEPG - 1, - NPDEPG - 1, NPTEPG - 1); + range.sva = kva_layout.kva_max; + pml5e = 0; /* no UB for la48 */ /* * Iterate over the kernel page tables without holding the kernel pmap @@ -12078,41 +12013,50 @@ sysctl_kmaps(SYSCTL_HANDLER_ARGS) * Within the large map, ensure that PDP and PD page addresses are * valid before descending. */ - for (sva = 0, i = pmap_pml4e_index(sva); i < NPML4EPG; i++) { - switch (i) { - case PML4PML4I: + for (first = true, sva = 0; sva != 0 || first; first = false) { + if (sva == kva_layout.rec_pt) sbuf_printf(sb, "\nRecursive map:\n"); - break; - case DMPML4I: + else if (sva == kva_layout.dmap_low) sbuf_printf(sb, "\nDirect map:\n"); - break; #ifdef KASAN - case KASANPML4I: + else if (sva == kva_layout.kasan_shadow_low) sbuf_printf(sb, "\nKASAN shadow map:\n"); - break; #endif #ifdef KMSAN - case KMSANSHADPML4I: + else if (sva == kva_layout.kmsan_shadow_low) sbuf_printf(sb, "\nKMSAN shadow map:\n"); - break; - case KMSANORIGPML4I: + else if (sva == kva_layout.kmsan_origin_low) sbuf_printf(sb, "\nKMSAN origin map:\n"); - break; #endif - case KPML4BASE: + else if (sva == kva_layout.km_low) sbuf_printf(sb, "\nKernel map:\n"); - break; - case LMSPML4I: + else if (sva == kva_layout.lm_low) sbuf_printf(sb, "\nLarge map:\n"); - break; - } /* Convert to canonical form. */ - if (sva == 1ul << 47) - sva |= -1ul << 48; + if (la57) { + if (sva == 1ul << 56) { + sva |= -1ul << 57; + continue; + } + } else { + if (sva == 1ul << 47) { + sva |= -1ul << 48; + continue; + } + } restart: - pml4e = kernel_pml4[i]; + if (la57) { + pml5e = *pmap_pml5e(kernel_pmap, sva); + if ((pml5e & X86_PG_V) == 0) { + sva = rounddown2(sva, NBPML5); + sysctl_kmaps_dump(sb, &range, sva); + sva += NBPML5; + continue; + } + } + pml4e = *pmap_pml4e(kernel_pmap, sva); if ((pml4e & X86_PG_V) == 0) { sva = rounddown2(sva, NBPML4); sysctl_kmaps_dump(sb, &range, sva); @@ -12133,8 +12077,8 @@ restart: pa = pdpe & PG_FRAME; if ((pdpe & PG_PS) != 0) { sva = rounddown2(sva, NBPDP); - sysctl_kmaps_check(sb, &range, sva, pml4e, pdpe, - 0, 0); + sysctl_kmaps_check(sb, &range, sva, pml5e, + pml4e, pdpe, 0, 0); range.pdpes++; sva += NBPDP; continue; @@ -12146,6 +12090,7 @@ restart: * freed. Validate the next-level address * before descending. */ + sva += NBPDP; goto restart; } pd = (pd_entry_t *)PHYS_TO_DMAP(pa); @@ -12162,7 +12107,7 @@ restart: if ((pde & PG_PS) != 0) { sva = rounddown2(sva, NBPDR); sysctl_kmaps_check(sb, &range, sva, - pml4e, pdpe, pde, 0); + pml5e, pml4e, pdpe, pde, 0); range.pdes++; sva += NBPDR; continue; @@ -12174,6 +12119,7 @@ restart: * may be freed. Validate the * next-level address before descending. */ + sva += NBPDR; goto restart; } pt = (pt_entry_t *)PHYS_TO_DMAP(pa); @@ -12187,7 +12133,7 @@ restart: continue; } sysctl_kmaps_check(sb, &range, sva, - pml4e, pdpe, pde, pte); + pml5e, pml4e, pdpe, pde, pte); range.ptes++; } } diff --git a/sys/amd64/amd64/support.S b/sys/amd64/amd64/support.S index c95696bbe7ef..870cd255abb7 100644 --- a/sys/amd64/amd64/support.S +++ b/sys/amd64/amd64/support.S @@ -934,10 +934,7 @@ ENTRY(casueword32_nosmap) ja fusufault movl %esi,%eax /* old */ -#ifdef SMP - lock -#endif - cmpxchgl %ecx,(%rdi) /* new = %ecx */ + lock cmpxchgl %ecx,(%rdi) /* new = %ecx */ setne %cl /* @@ -971,10 +968,7 @@ ENTRY(casueword32_smap) movl %esi,%eax /* old */ stac -#ifdef SMP - lock -#endif - cmpxchgl %ecx,(%rdi) /* new = %ecx */ + lock cmpxchgl %ecx,(%rdi) /* new = %ecx */ clac setne %cl @@ -1014,10 +1008,7 @@ ENTRY(casueword_nosmap) ja fusufault movq %rsi,%rax /* old */ -#ifdef SMP - lock -#endif - cmpxchgq %rcx,(%rdi) /* new = %rcx */ + lock cmpxchgq %rcx,(%rdi) /* new = %rcx */ setne %cl /* @@ -1045,10 +1036,7 @@ ENTRY(casueword_smap) movq %rsi,%rax /* old */ stac -#ifdef SMP - lock -#endif - cmpxchgq %rcx,(%rdi) /* new = %rcx */ + lock cmpxchgq %rcx,(%rdi) /* new = %rcx */ clac setne %cl diff --git a/sys/amd64/amd64/trap.c b/sys/amd64/amd64/trap.c index 09ac0a67dbef..f3469ed5e2bc 100644 --- a/sys/amd64/amd64/trap.c +++ b/sys/amd64/amd64/trap.c @@ -37,7 +37,6 @@ * SUCH DAMAGE. */ -#include <sys/cdefs.h> /* * AMD64 Trap and System call handling */ @@ -87,9 +86,7 @@ PMC_SOFT_DEFINE( , , page_fault, write); #include <x86/mca.h> #include <machine/md_var.h> #include <machine/pcb.h> -#ifdef SMP #include <machine/smp.h> -#endif #include <machine/stack.h> #include <machine/trap.h> #include <machine/tss.h> @@ -769,7 +766,7 @@ trap_pfault(struct trapframe *frame, bool usermode, int *signo, int *ucode) return (-1); } } - if (eva >= VM_MIN_KERNEL_ADDRESS) { + if (eva >= kva_layout.km_low) { /* * Don't allow user-mode faults in kernel address space. */ @@ -900,11 +897,9 @@ trap_diag(struct trapframe *frame, vm_offset_t eva) printf("\n\nFatal trap %d: %s while in %s mode\n", type, type < nitems(trap_msg) ? trap_msg[type] : UNKNOWN, TRAPF_USERMODE(frame) ? "user" : "kernel"); -#ifdef SMP - /* two separate prints in case of a trap on an unmapped page */ - printf("cpuid = %d; ", PCPU_GET(cpuid)); - printf("apic id = %02x\n", PCPU_GET(apic_id)); -#endif + /* Print these separately in case pcpu accesses trap. */ + printf("cpuid = %d; apic id = %02x\n", PCPU_GET(cpuid), + PCPU_GET(apic_id)); if (type == T_PAGEFLT) { printf("fault virtual address = 0x%lx\n", eva); printf("fault code = %s %s %s%s%s, %s\n", @@ -1025,11 +1020,9 @@ dblfault_handler(struct trapframe *frame) frame->tf_cs, frame->tf_ss, frame->tf_ds, frame->tf_es, frame->tf_fs, frame->tf_gs, rdmsr(MSR_FSBASE), rdmsr(MSR_GSBASE), rdmsr(MSR_KGSBASE)); -#ifdef SMP - /* two separate prints in case of a trap on an unmapped page */ - printf("cpuid = %d; ", PCPU_GET(cpuid)); - printf("apic id = %02x\n", PCPU_GET(apic_id)); -#endif + /* Print these separately in case pcpu accesses trap. */ + printf("cpuid = %d; apic id = %02x\n", PCPU_GET(cpuid), + PCPU_GET(apic_id)); panic("double fault"); } diff --git a/sys/amd64/conf/MINIMALUP b/sys/amd64/conf/MINIMALUP deleted file mode 100644 index 0dbddbe5b341..000000000000 --- a/sys/amd64/conf/MINIMALUP +++ /dev/null @@ -1,4 +0,0 @@ -include MINIMAL -ident MINIMALUP -nooptions SMP -nooptions NUMA diff --git a/sys/amd64/include/param.h b/sys/amd64/include/param.h index 8db314fa034d..5a9c3162e14c 100644 --- a/sys/amd64/include/param.h +++ b/sys/amd64/include/param.h @@ -146,11 +146,10 @@ #define amd64_btop(x) ((unsigned long)(x) >> PAGE_SHIFT) #define amd64_ptob(x) ((unsigned long)(x) << PAGE_SHIFT) -#define INKERNEL(va) (((va) >= DMAP_MIN_ADDRESS && (va) < DMAP_MAX_ADDRESS) \ - || ((va) >= VM_MIN_KERNEL_ADDRESS && (va) < VM_MAX_KERNEL_ADDRESS)) +#define INKERNEL(va) \ + (((va) >= kva_layout.dmap_low && (va) < kva_layout.dmap_high) || \ + ((va) >= kva_layout.km_low && (va) < kva_layout.km_high)) -#ifdef SMP #define SC_TABLESIZE 1024 /* Must be power of 2. */ -#endif #endif /* !_AMD64_INCLUDE_PARAM_H_ */ diff --git a/sys/amd64/include/pmap.h b/sys/amd64/include/pmap.h index 7d3e91bcd9b9..e2f97442c10f 100644 --- a/sys/amd64/include/pmap.h +++ b/sys/amd64/include/pmap.h @@ -169,11 +169,12 @@ * the recursive page table map. */ #define NDMPML4E 8 +#define NDMPML5E 32 /* - * These values control the layout of virtual memory. The starting address - * of the direct map, which is controlled by DMPML4I, must be a multiple of - * its size. (See the PHYS_TO_DMAP() and DMAP_TO_PHYS() macros.) + * These values control the layout of virtual memory. The starting + * address of the direct map is controlled by DMPML4I on LA48 and + * DMPML5I on LA57. * * Note: KPML4I is the index of the (single) level 4 page that maps * the KVA that holds KERNBASE, while KPML4BASE is the index of the @@ -191,6 +192,7 @@ #define KPML4BASE (NPML4EPG-NKPML4E) /* KVM at highest addresses */ #define DMPML4I rounddown(KPML4BASE-NDMPML4E, NDMPML4E) /* Below KVM */ +#define DMPML5I (NPML5EPG / 2 + 1) #define KPML4I (NPML4EPG-1) #define KPDPI (NPDPEPG-2) /* kernbase at -2GB */ @@ -200,9 +202,14 @@ #define KMSANSHADPML4I (KPML4BASE - NKMSANSHADPML4E) #define KMSANORIGPML4I (DMPML4I - NKMSANORIGPML4E) -/* Large map: index of the first and max last pml4 entry */ +/* + * Large map: index of the first and max last pml4/la48 and pml5/la57 + * entry. + */ #define LMSPML4I (PML4PML4I + 1) #define LMEPML4I (KASANPML4I - 1) +#define LMSPML5I (DMPML5I + NDMPML5E) +#define LMEPML5I (LMSPML5I + 32 - 1) /* 32 slots for large map */ /* * XXX doesn't really belong here I guess... @@ -548,6 +555,25 @@ pmap_pml5e_index(vm_offset_t va) return ((va >> PML5SHIFT) & ((1ul << NPML5EPGSHIFT) - 1)); } +struct kva_layout_s { + vm_offset_t kva_min; + vm_offset_t kva_max; + vm_offset_t dmap_low; /* DMAP_MIN_ADDRESS */ + vm_offset_t dmap_high; /* DMAP_MAX_ADDRESS */ + vm_offset_t lm_low; /* LARGEMAP_MIN_ADDRESS */ + vm_offset_t lm_high; /* LARGEMAP_MAX_ADDRESS */ + vm_offset_t km_low; /* VM_MIN_KERNEL_ADDRESS */ + vm_offset_t km_high; /* VM_MAX_KERNEL_ADDRESS */ + vm_offset_t rec_pt; + vm_offset_t kasan_shadow_low; /* KASAN_MIN_ADDRESS */ + vm_offset_t kasan_shadow_high; /* KASAN_MAX_ADDRESS */ + vm_offset_t kmsan_shadow_low; /* KMSAN_SHAD_MIN_ADDRESS */ + vm_offset_t kmsan_shadow_high; /* KMSAN_SHAD_MAX_ADDRESS */ + vm_offset_t kmsan_origin_low; /* KMSAN_ORIG_MIN_ADDRESS */ + vm_offset_t kmsan_origin_high; /* KMSAN_ORIG_MAX_ADDRESS */ +}; +extern struct kva_layout_s kva_layout; + #endif /* !LOCORE */ #endif /* !_MACHINE_PMAP_H_ */ diff --git a/sys/amd64/include/smp.h b/sys/amd64/include/smp.h index 26eb227211da..bff92570ff82 100644 --- a/sys/amd64/include/smp.h +++ b/sys/amd64/include/smp.h @@ -13,8 +13,6 @@ #ifdef _KERNEL -#ifdef SMP - #ifndef LOCORE #include <x86/x86_smp.h> @@ -39,7 +37,6 @@ void invlop_handler(void); int start_all_aps(void); #endif /* !LOCORE */ -#endif /* SMP */ #endif /* _KERNEL */ #endif /* _MACHINE_SMP_H_ */ diff --git a/sys/amd64/include/vmm.h b/sys/amd64/include/vmm.h index a9c73b75213b..0b3daed4f69e 100644 --- a/sys/amd64/include/vmm.h +++ b/sys/amd64/include/vmm.h @@ -649,6 +649,8 @@ struct vm_inout_str { int addrsize; enum vm_reg_name seg_name; struct seg_desc seg_desc; + int cs_d; + uint64_t cs_base; }; enum task_switch_reason { diff --git a/sys/amd64/include/vmm_dev.h b/sys/amd64/include/vmm_dev.h index 1f86538ce5f3..441330fd57b8 100644 --- a/sys/amd64/include/vmm_dev.h +++ b/sys/amd64/include/vmm_dev.h @@ -29,6 +29,8 @@ #ifndef _VMM_DEV_H_ #define _VMM_DEV_H_ +#include <sys/domainset.h> + #include <machine/vmm.h> #include <machine/vmm_snapshot.h> @@ -52,7 +54,10 @@ struct vm_munmap { struct vm_memseg { int segid; size_t len; - char name[VM_MAX_SUFFIXLEN + 1]; + char name[VM_MAX_SUFFIXLEN + 1]; + domainset_t *ds_mask; + size_t ds_mask_size; + int ds_policy; }; struct vm_register { diff --git a/sys/amd64/include/vmm_instruction_emul.h b/sys/amd64/include/vmm_instruction_emul.h index d5f0363cfb41..1fb0f97682a7 100644 --- a/sys/amd64/include/vmm_instruction_emul.h +++ b/sys/amd64/include/vmm_instruction_emul.h @@ -31,6 +31,31 @@ #include <sys/mman.h> +/* struct vie_op.op_type */ +enum { + VIE_OP_TYPE_NONE = 0, + VIE_OP_TYPE_MOV, + VIE_OP_TYPE_MOVSX, + VIE_OP_TYPE_MOVZX, + VIE_OP_TYPE_AND, + VIE_OP_TYPE_OR, + VIE_OP_TYPE_SUB, + VIE_OP_TYPE_TWO_BYTE, + VIE_OP_TYPE_PUSH, + VIE_OP_TYPE_CMP, + VIE_OP_TYPE_POP, + VIE_OP_TYPE_MOVS, + VIE_OP_TYPE_GROUP1, + VIE_OP_TYPE_STOS, + VIE_OP_TYPE_BITTEST, + VIE_OP_TYPE_TWOB_GRP15, + VIE_OP_TYPE_ADD, + VIE_OP_TYPE_TEST, + VIE_OP_TYPE_BEXTR, + VIE_OP_TYPE_OUTS, + VIE_OP_TYPE_LAST +}; + /* * Callback functions to read and write memory regions. */ diff --git a/sys/amd64/include/vmparam.h b/sys/amd64/include/vmparam.h index 0cd9bb4fa7a4..d2ac3c6648b2 100644 --- a/sys/amd64/include/vmparam.h +++ b/sys/amd64/include/vmparam.h @@ -163,6 +163,7 @@ * Virtual addresses of things. Derived from the page directory and * page table indexes from pmap.h for precision. * + * LA48: * 0x0000000000000000 - 0x00007fffffffffff user map * 0x0000800000000000 - 0xffff7fffffffffff does not exist (hole) * 0xffff800000000000 - 0xffff804020100fff recursive page table (512GB slot) @@ -175,32 +176,38 @@ * 0xfffffc0000000000 - 0xfffffdffffffffff 2TB KMSAN shadow map, optional * 0xfffffe0000000000 - 0xffffffffffffffff 2TB kernel map * + * LA57: + * 0x0000000000000000 - 0x00ffffffffffffff user map + * 0x0100000000000000 - 0xf0ffffffffffffff does not exist (hole) + * 0xff00000000000000 - 0xff00ffffffffffff recursive page table (2048TB slot) + * 0xff01000000000000 - 0xff20ffffffffffff direct map (32 x 2048TB slots) + * 0xff21000000000000 - 0xff40ffffffffffff large map + * 0xff41000000000000 - 0xffff7fffffffffff unused + * 0xffff800000000000 - 0xfffff5ffffffffff unused (start of kernel pml4 entry) + * 0xfffff60000000000 - 0xfffff7ffffffffff 2TB KMSAN origin map, optional + * 0xfffff78000000000 - 0xfffff7bfffffffff 512GB KASAN shadow map, optional + * 0xfffff80000000000 - 0xfffffbffffffffff 4TB unused + * 0xfffffc0000000000 - 0xfffffdffffffffff 2TB KMSAN shadow map, optional + * 0xfffffe0000000000 - 0xffffffffffffffff 2TB kernel map + * * Within the kernel map: * * 0xfffffe0000000000 vm_page_array * 0xffffffff80000000 KERNBASE */ -#define VM_MIN_KERNEL_ADDRESS KV4ADDR(KPML4BASE, 0, 0, 0) -#define VM_MAX_KERNEL_ADDRESS KV4ADDR(KPML4BASE + NKPML4E - 1, \ - NPDPEPG-1, NPDEPG-1, NPTEPG-1) - -#define DMAP_MIN_ADDRESS KV4ADDR(DMPML4I, 0, 0, 0) -#define DMAP_MAX_ADDRESS KV4ADDR(DMPML4I + NDMPML4E, 0, 0, 0) - -#define KASAN_MIN_ADDRESS KV4ADDR(KASANPML4I, 0, 0, 0) -#define KASAN_MAX_ADDRESS KV4ADDR(KASANPML4I + NKASANPML4E, 0, 0, 0) +#define VM_MIN_KERNEL_ADDRESS_LA48 KV4ADDR(KPML4BASE, 0, 0, 0) +#define VM_MIN_KERNEL_ADDRESS kva_layout.km_low +#define VM_MAX_KERNEL_ADDRESS kva_layout.km_high -#define KMSAN_SHAD_MIN_ADDRESS KV4ADDR(KMSANSHADPML4I, 0, 0, 0) -#define KMSAN_SHAD_MAX_ADDRESS KV4ADDR(KMSANSHADPML4I + NKMSANSHADPML4E, \ - 0, 0, 0) +#define KASAN_MIN_ADDRESS (kva_layout.kasan_shadow_low) +#define KASAN_MAX_ADDRESS (kva_layout.kasan_shadow_high) -#define KMSAN_ORIG_MIN_ADDRESS KV4ADDR(KMSANORIGPML4I, 0, 0, 0) -#define KMSAN_ORIG_MAX_ADDRESS KV4ADDR(KMSANORIGPML4I + NKMSANORIGPML4E, \ - 0, 0, 0) +#define KMSAN_SHAD_MIN_ADDRESS (kva_layout.kmsan_shadow_low) +#define KMSAN_SHAD_MAX_ADDRESS (kva_layout.kmsan_shadow_high) -#define LARGEMAP_MIN_ADDRESS KV4ADDR(LMSPML4I, 0, 0, 0) -#define LARGEMAP_MAX_ADDRESS KV4ADDR(LMEPML4I + 1, 0, 0, 0) +#define KMSAN_ORIG_MIN_ADDRESS (kva_layout.kmsan_origin_low) +#define KMSAN_ORIG_MAX_ADDRESS (kva_layout.kmsan_origin_high) /* * Formally kernel mapping starts at KERNBASE, but kernel linker @@ -239,21 +246,21 @@ * vt fb startup needs to be reworked. */ #define PHYS_IN_DMAP(pa) (dmaplimit == 0 || (pa) < dmaplimit) -#define VIRT_IN_DMAP(va) ((va) >= DMAP_MIN_ADDRESS && \ - (va) < (DMAP_MIN_ADDRESS + dmaplimit)) +#define VIRT_IN_DMAP(va) \ + ((va) >= kva_layout.dmap_low && (va) < kva_layout.dmap_low + dmaplimit) #define PMAP_HAS_DMAP 1 -#define PHYS_TO_DMAP(x) ({ \ +#define PHYS_TO_DMAP(x) __extension__ ({ \ KASSERT(PHYS_IN_DMAP(x), \ ("physical address %#jx not covered by the DMAP", \ (uintmax_t)x)); \ - (x) | DMAP_MIN_ADDRESS; }) + (x) + kva_layout.dmap_low; }) -#define DMAP_TO_PHYS(x) ({ \ +#define DMAP_TO_PHYS(x) __extension__ ({ \ KASSERT(VIRT_IN_DMAP(x), \ ("virtual address %#jx not covered by the DMAP", \ (uintmax_t)x)); \ - (x) & ~DMAP_MIN_ADDRESS; }) + (x) - kva_layout.dmap_low; }) /* * amd64 maps the page array into KVA so that it can be more easily @@ -274,7 +281,7 @@ */ #ifndef VM_KMEM_SIZE_MAX #define VM_KMEM_SIZE_MAX ((VM_MAX_KERNEL_ADDRESS - \ - VM_MIN_KERNEL_ADDRESS + 1) * 3 / 5) + kva_layout.km_low + 1) * 3 / 5) #endif /* initial pagein size of beginning of executable file */ diff --git a/sys/amd64/pt/pt.c b/sys/amd64/pt/pt.c new file mode 100644 index 000000000000..c7b75767680a --- /dev/null +++ b/sys/amd64/pt/pt.c @@ -0,0 +1,978 @@ +/* + * Copyright (c) 2025 Bojan Novković <bnovkov@freebsd.org> + * + * SPDX-License-Identifier: BSD-2-Clause + */ + +/* + * hwt(4) Intel Processor Trace (PT) backend + * + * Driver Design Overview + * + * - Since PT is configured on a per-core basis, the driver uses + * 'smp_rendezvous' to start and disable tracing on each target core. + * - PT-specific resources are stored in a 'struct pt_ctx' context structure for + * each traced CPU core or thread. Upon initialization, a ToPA configuration + * is generated for each 'pt_ctx' structure using the HWT tracing buffers. + * The HWT tracing buffer is split into 4K ToPA entries. Currently, each + * 4K ToPA entry is configured to trigger an interrupt after it is filled. + * - The PT driver uses the XSAVE/XRSTOR PT extensions to load and save all + * relevant PT registers. Every time a traced thread is switched + * out or in, its state will be saved to or loaded from its corresponding + * 'pt_ctx' context. + * - When tracing starts, the PT hardware will start writing data into the + * tracing buffer. When a TOPA_INT entry is filled, it will trigger an + * interrupt before continuing. The interrupt handler will then fetch the + * last valid tracing buffer offset and enqueue a HWT_RECORD_BUFFER record. + * The driver is currently configured to use the NMI interrupt line. + * - The userspace PT backend waits for incoming HWT_RECORD_BUFFER records + * and uses the offsets to decode data from the tracing buffer. + * + * Future improvements and limitations + * + * - We currently configure the PT hardware to trigger an interrupt whenever + * a 4K ToPA entry is filled. While this is fine when tracing smaller + * functions or infrequent code paths, this will generate too much interrupt + * traffic when tracing hotter functions. A proper solution for this issue + * should estimate the amount of data generated by the current configuration + * and use it to determine interrupt frequency. + * + * - Support for more tracing options and PT features. + * + */ + +#include <sys/systm.h> +#include <sys/hwt.h> +#include <sys/kernel.h> +#include <sys/lock.h> +#include <sys/malloc.h> +#include <sys/module.h> +#include <sys/mutex.h> +#include <sys/sdt.h> +#include <sys/smp.h> +#include <sys/taskqueue.h> + +#include <vm/vm.h> +#include <vm/vm_page.h> + +#include <machine/atomic.h> +#include <machine/cpufunc.h> +#include <machine/fpu.h> +#include <machine/smp.h> +#include <machine/specialreg.h> + +#include <x86/apicvar.h> +#include <x86/x86_var.h> + +#include <dev/hwt/hwt_context.h> +#include <dev/hwt/hwt_vm.h> +#include <dev/hwt/hwt_backend.h> +#include <dev/hwt/hwt_config.h> +#include <dev/hwt/hwt_cpu.h> +#include <dev/hwt/hwt_record.h> +#include <dev/hwt/hwt_thread.h> + +#include <amd64/pt/pt.h> + +#ifdef PT_DEBUG +#define dprintf(fmt, ...) printf(fmt, ##__VA_ARGS__) +#else +#define dprintf(fmt, ...) +#endif +#define PT_SUPPORTED_FLAGS \ + (RTIT_CTL_MTCEN | RTIT_CTL_CR3FILTER | RTIT_CTL_DIS_TNT | \ + RTIT_CTL_USER | RTIT_CTL_OS | RTIT_CTL_BRANCHEN) +#define PT_XSAVE_MASK (XFEATURE_ENABLED_X87 | XFEATURE_ENABLED_SSE) +#define PT_XSTATE_BV (PT_XSAVE_MASK | XFEATURE_ENABLED_PT) +#define PT_MAX_IP_RANGES 2 + +#define PT_TOPA_MASK_PTRS 0x7f +#define PT_TOPA_PAGE_MASK 0xffffff80 +#define PT_TOPA_PAGE_SHIFT 7 + +#define CPUID_PT_LEAF 0x14 + +MALLOC_DEFINE(M_PT, "pt", "Intel Processor Trace"); + +SDT_PROVIDER_DEFINE(pt); +SDT_PROBE_DEFINE(pt, , , topa__intr); + +TASKQUEUE_FAST_DEFINE_THREAD(pt); + +static void pt_send_buffer_record(void *arg, int pending __unused); +static int pt_topa_intr(struct trapframe *tf); + +/* + * Intel Processor Trace XSAVE-managed state. + */ +struct pt_ext_area { + uint64_t rtit_ctl; + uint64_t rtit_output_base; + uint64_t rtit_output_mask_ptrs; + uint64_t rtit_status; + uint64_t rtit_cr3_match; + uint64_t rtit_addr0_a; + uint64_t rtit_addr0_b; + uint64_t rtit_addr1_a; + uint64_t rtit_addr1_b; +}; + +struct pt_buffer { + uint64_t *topa_hw; /* ToPA table entries. */ + size_t size; + struct mtx lock; /* Lock for fields below. */ + vm_offset_t offset; + uint64_t wrap_count; + int curpage; +}; + +struct pt_ctx { + int id; + struct pt_buffer buf; /* ToPA buffer metadata */ + struct task task; /* ToPA buffer notification task */ + struct hwt_context *hwt_ctx; + uint8_t *save_area; /* PT XSAVE area */ +}; +/* PT tracing contexts used for CPU mode. */ +static struct pt_ctx *pt_pcpu_ctx; + +enum pt_cpu_state { + PT_DISABLED = 0, + PT_STOPPED, + PT_ACTIVE +}; + +static struct pt_cpu { + struct pt_ctx *ctx; /* active PT tracing context */ + enum pt_cpu_state state; /* used as part of trace stop protocol */ +} *pt_pcpu; + +/* + * PT-related CPUID bits. + */ +static struct pt_cpu_info { + uint32_t l0_eax; + uint32_t l0_ebx; + uint32_t l0_ecx; + uint32_t l1_eax; + uint32_t l1_ebx; + size_t xsave_area_size; + size_t xstate_hdr_offset; + size_t pt_xsave_offset; +} pt_info __read_mostly; + +static bool initialized = false; +static int cpu_mode_ctr = 0; + +static __inline enum pt_cpu_state +pt_cpu_get_state(int cpu_id) +{ + return (atomic_load_int(&pt_pcpu[cpu_id].state)); +} + +static __inline void +pt_cpu_set_state(int cpu_id, enum pt_cpu_state state) +{ + atomic_store_int(&pt_pcpu[cpu_id].state, state); +} + +static __inline struct xstate_hdr * +pt_ctx_get_xstate_hdr(struct pt_ctx *ctx) +{ + return ((struct xstate_hdr *)(ctx->save_area + + pt_info.xstate_hdr_offset)); +} + + +static __inline struct pt_ext_area * +pt_ctx_get_ext_area(struct pt_ctx *ctx) +{ + return ((struct pt_ext_area *)(ctx->save_area + + pt_info.pt_xsave_offset)); +} + +/* + * Updates current trace buffer offset from the + * ToPA MSRs. Records if the trace buffer wrapped. + */ +static __inline void +pt_update_buffer(struct pt_buffer *buf) +{ + uint64_t reg; + int curpage; + + /* Update buffer offset. */ + reg = rdmsr(MSR_IA32_RTIT_OUTPUT_MASK_PTRS); + curpage = (reg & PT_TOPA_PAGE_MASK) >> PT_TOPA_PAGE_SHIFT; + mtx_lock_spin(&buf->lock); + /* Check if the output wrapped. */ + if (buf->curpage > curpage) + buf->wrap_count++; + buf->curpage = curpage; + buf->offset = reg >> 32; + mtx_unlock_spin(&buf->lock); + + dprintf("%s: wrap_cnt: %lu, curpage: %d, offset: %zu\n", __func__, + buf->wrap_count, buf->curpage, buf->offset); +} + +static __inline void +pt_fill_buffer_record(int id, struct pt_buffer *buf, + struct hwt_record_entry *rec) +{ + rec->record_type = HWT_RECORD_BUFFER; + rec->buf_id = id; + rec->curpage = buf->curpage; + rec->offset = buf->offset + (buf->wrap_count * buf->size); +} + +/* + * Enables or disables tracing on curcpu + * using the XSAVE/XRSTOR PT extensions. + */ +static void +pt_cpu_toggle_local(uint8_t *save_area, bool enable) +{ + u_long xcr0, cr0; + u_long xss; + + cr0 = rcr0(); + if (cr0 & CR0_TS) + clts(); + xcr0 = rxcr(XCR0); + if ((xcr0 & PT_XSAVE_MASK) != PT_XSAVE_MASK) + load_xcr(XCR0, xcr0 | PT_XSAVE_MASK); + xss = rdmsr(MSR_IA32_XSS); + wrmsr(MSR_IA32_XSS, xss | XFEATURE_ENABLED_PT); + + if (!enable) { + KASSERT((rdmsr(MSR_IA32_RTIT_CTL) & RTIT_CTL_TRACEEN) != 0, + ("%s: PT is disabled", __func__)); + xsaves(save_area, XFEATURE_ENABLED_PT); + } else { + KASSERT((rdmsr(MSR_IA32_RTIT_CTL) & RTIT_CTL_TRACEEN) == 0, + ("%s: PT is enabled", __func__)); + xrstors(save_area, XFEATURE_ENABLED_PT); + } + wrmsr(MSR_IA32_XSS, xss); + if ((xcr0 & PT_XSAVE_MASK) != PT_XSAVE_MASK) + load_xcr(XCR0, xcr0); + if (cr0 & CR0_TS) + load_cr0(cr0); +} + +/* + * Starts PT tracing on 'curcpu'. + */ +static void +pt_cpu_start(void *dummy) +{ + struct pt_cpu *cpu; + + cpu = &pt_pcpu[curcpu]; + MPASS(cpu->ctx != NULL); + + dprintf("%s: curcpu %d\n", __func__, curcpu); + load_cr4(rcr4() | CR4_XSAVE); + wrmsr(MSR_IA32_RTIT_STATUS, 0); + pt_cpu_set_state(curcpu, PT_ACTIVE); + pt_cpu_toggle_local(cpu->ctx->save_area, true); +} + +/* + * Stops PT tracing on 'curcpu'. + * Updates trace buffer offset to ensure + * any data generated between the last interrupt + * and the trace stop gets picked up by userspace. + */ +static void +pt_cpu_stop(void *dummy) +{ + struct pt_cpu *cpu; + struct pt_ctx *ctx; + + /* Shutdown may occur before PT gets properly configured. */ + if (pt_cpu_get_state(curcpu) == PT_DISABLED) + return; + + cpu = &pt_pcpu[curcpu]; + ctx = cpu->ctx; + MPASS(ctx != NULL); + dprintf("%s: curcpu %d\n", __func__, curcpu); + + pt_cpu_set_state(curcpu, PT_STOPPED); + pt_cpu_toggle_local(cpu->ctx->save_area, false); + pt_update_buffer(&ctx->buf); +} + +/* + * Prepares the Table of Physical Addresses (ToPA) metadata for 'pt_ctx'. + * The HWT trace buffer is split into 4K ToPA table entries and used + * as a circular buffer, meaning that the last ToPA entry points to + * the first ToPA entry. Each entry is configured to raise an + * interrupt after being filled. + */ +static int +pt_topa_prepare(struct pt_ctx *ctx, struct hwt_vm *vm) +{ + struct pt_buffer *buf; + size_t topa_size; + int i; + + topa_size = TOPA_SIZE_4K; + buf = &ctx->buf; + + KASSERT(buf->topa_hw == NULL, + ("%s: ToPA info already exists", __func__)); + buf->topa_hw = mallocarray(vm->npages + 1, sizeof(uint64_t), M_PT, + M_ZERO | M_WAITOK); + dprintf("%s: ToPA virt addr %p\n", __func__, buf->topa_hw); + buf->size = vm->npages * PAGE_SIZE; + for (i = 0; i < vm->npages; i++) { + buf->topa_hw[i] = VM_PAGE_TO_PHYS(vm->pages[i]) | topa_size; + /* + * XXX: TOPA_INT should ideally be set according to + * expected amount of incoming trace data. Too few TOPA_INT + * entries will not trigger interrupts often enough when tracing + * smaller functions. + */ + buf->topa_hw[i] |= TOPA_INT; + } + buf->topa_hw[vm->npages] = (uint64_t)vtophys(buf->topa_hw) | TOPA_END; + + return (0); +} + +/* + * Configures IP filtering for trace generation. + * A maximum of 2 ranges can be specified due to + * limitations imposed by the XSAVE/XRSTOR PT extensions. + */ +static int +pt_configure_ranges(struct pt_ctx *ctx, struct pt_cpu_config *cfg) +{ + struct pt_ext_area *pt_ext; + int nranges_supp, n, error = 0; + + pt_ext = pt_ctx_get_ext_area(ctx); + if (pt_info.l0_ebx & CPUPT_IPF) { + nranges_supp = (pt_info.l1_eax & CPUPT_NADDR_M) >> + CPUPT_NADDR_S; + + if (nranges_supp > PT_IP_FILTER_MAX_RANGES) + nranges_supp = PT_IP_FILTER_MAX_RANGES; + n = cfg->nranges; + if (n > nranges_supp) { + printf("%s: %d IP filtering ranges requested, CPU " + "supports %d, truncating\n", + __func__, n, nranges_supp); + n = nranges_supp; + } + + switch (n) { + case 2: + pt_ext->rtit_ctl |= (1UL << RTIT_CTL_ADDR_CFG_S(1)); + pt_ext->rtit_addr1_a = cfg->ip_ranges[1].start; + pt_ext->rtit_addr1_b = cfg->ip_ranges[1].end; + case 1: + pt_ext->rtit_ctl |= (1UL << RTIT_CTL_ADDR_CFG_S(0)); + pt_ext->rtit_addr0_a = cfg->ip_ranges[0].start; + pt_ext->rtit_addr0_b = cfg->ip_ranges[0].end; + break; + default: + error = (EINVAL); + break; + }; + } else + error = (ENXIO); + + return (error); +} + +static int +pt_init_ctx(struct pt_ctx *pt_ctx, struct hwt_vm *vm, int ctx_id) +{ + + dprintf("%s: ctx id %d\n", __func__, ctx_id); + + KASSERT(pt_ctx->buf.topa_hw == NULL, + ("%s: active ToPA buffer in context %p\n", __func__, pt_ctx)); + + memset(pt_ctx, 0, sizeof(struct pt_ctx)); + mtx_init(&pt_ctx->buf.lock, "pttopa", NULL, MTX_SPIN); + pt_ctx->save_area = malloc_aligned(pt_info.xsave_area_size, 64, + M_PT, M_NOWAIT | M_ZERO); + if (pt_ctx->save_area == NULL) + return (ENOMEM); + dprintf("%s: preparing ToPA buffer\n", __func__); + if (pt_topa_prepare(pt_ctx, vm) != 0) { + dprintf("%s: failed to prepare ToPA buffer\n", __func__); + free(pt_ctx->save_area, M_PT); + return (ENOMEM); + } + + pt_ctx->id = ctx_id; + TASK_INIT(&pt_ctx->task, 0, pt_send_buffer_record, pt_ctx); + + return (0); +} + +static void +pt_deinit_ctx(struct pt_ctx *pt_ctx) +{ + + if (pt_ctx->buf.topa_hw != NULL) + free(pt_ctx->buf.topa_hw, M_PT); + if (pt_ctx->save_area != NULL) + free(pt_ctx->save_area, M_PT); + memset(pt_ctx, 0, sizeof(*pt_ctx)); + pt_ctx->buf.topa_hw = NULL; +} + +/* + * HWT backend configuration method. + * + * Checks and translates the user-defined configuration to a + * set of PT tracing features. Uses the feature set to initialize + * the tracing context for the target CPU or thread. + */ +static int +pt_backend_configure(struct hwt_context *ctx, int cpu_id, int thread_id) +{ + struct hwt_cpu *hwt_cpu; + struct hwt_thread *thr; + struct pt_ctx *pt_ctx; + struct pt_cpu_config *cfg; + struct pt_ext_area *pt_ext; + struct xstate_hdr *hdr; + int error; + + dprintf("%s\n", __func__); + + cfg = (struct pt_cpu_config *)ctx->config; + pt_ctx = NULL; + + /* Clear any flags we don't support yet. */ + cfg->rtit_ctl &= PT_SUPPORTED_FLAGS; + if (cfg->rtit_ctl & RTIT_CTL_MTCEN) { + if ((pt_info.l0_ebx & CPUPT_MTC) == 0) { + printf("%s: CPU does not support generating MTC " + "packets\n", __func__); + return (ENXIO); + } + } + + if (cfg->rtit_ctl & RTIT_CTL_CR3FILTER) { + if ((pt_info.l0_ebx & CPUPT_CR3) == 0) { + printf("%s: CPU does not support CR3 filtering\n", + __func__); + return (ENXIO); + } + } + + if (cfg->rtit_ctl & RTIT_CTL_DIS_TNT) { + if ((pt_info.l0_ebx & CPUPT_DIS_TNT) == 0) { + printf("%s: CPU does not support TNT\n", __func__); + return (ENXIO); + } + } + /* TODO: support for more config bits. */ + + if (ctx->mode == HWT_MODE_CPU) { + TAILQ_FOREACH(hwt_cpu, &ctx->cpus, next) { + if (hwt_cpu->cpu_id != cpu_id) + continue; + pt_ctx = &pt_pcpu_ctx[cpu_id]; + break; + } + } else { + TAILQ_FOREACH(thr, &ctx->threads, next) { + if (thr->thread_id != thread_id) + continue; + KASSERT(thr->private != NULL, + ("%s: hwt thread private" + " not set, thr %p", + __func__, thr)); + pt_ctx = (struct pt_ctx *)thr->private; + break; + } + } + if (pt_ctx == NULL) + return (ENOENT); + + dprintf("%s: preparing MSRs\n", __func__); + pt_ext = pt_ctx_get_ext_area(pt_ctx); + hdr = pt_ctx_get_xstate_hdr(pt_ctx); + + pt_ext->rtit_ctl |= cfg->rtit_ctl; + if (cfg->nranges != 0) { + dprintf("%s: preparing IPF ranges\n", __func__); + if ((error = pt_configure_ranges(pt_ctx, cfg)) != 0) + return (error); + } + pt_ctx->hwt_ctx = ctx; + pt_ext->rtit_ctl |= RTIT_CTL_TOPA; + pt_ext->rtit_output_base = (uint64_t)vtophys(pt_ctx->buf.topa_hw); + pt_ext->rtit_output_mask_ptrs = PT_TOPA_MASK_PTRS; + hdr->xstate_bv = XFEATURE_ENABLED_PT; + hdr->xstate_xcomp_bv = XFEATURE_ENABLED_PT | + XSTATE_XCOMP_BV_COMPACT; + pt_ext->rtit_ctl |= RTIT_CTL_TRACEEN; + pt_pcpu[cpu_id].ctx = pt_ctx; + pt_cpu_set_state(cpu_id, PT_STOPPED); + + return (0); +} + +/* + * hwt backend trace start operation. CPU affine. + */ +static void +pt_backend_enable(struct hwt_context *ctx, int cpu_id) +{ + if (ctx->mode == HWT_MODE_CPU) + return; + + KASSERT(curcpu == cpu_id, + ("%s: attempting to start PT on another cpu", __func__)); + pt_cpu_start(NULL); + CPU_SET(cpu_id, &ctx->cpu_map); +} + +/* + * hwt backend trace stop operation. CPU affine. + */ +static void +pt_backend_disable(struct hwt_context *ctx, int cpu_id) +{ + struct pt_cpu *cpu; + + if (ctx->mode == HWT_MODE_CPU) + return; + + KASSERT(curcpu == cpu_id, + ("%s: attempting to disable PT on another cpu", __func__)); + pt_cpu_stop(NULL); + CPU_CLR(cpu_id, &ctx->cpu_map); + cpu = &pt_pcpu[cpu_id]; + cpu->ctx = NULL; +} + +/* + * hwt backend trace start operation for remote CPUs. + */ +static int +pt_backend_enable_smp(struct hwt_context *ctx) +{ + + dprintf("%s\n", __func__); + if (ctx->mode == HWT_MODE_CPU && + atomic_swap_32(&cpu_mode_ctr, 1) != 0) + return (-1); + + KASSERT(ctx->mode == HWT_MODE_CPU, + ("%s: should only be used for CPU mode", __func__)); + smp_rendezvous_cpus(ctx->cpu_map, NULL, pt_cpu_start, NULL, NULL); + + return (0); +} + +/* + * hwt backend trace stop operation for remote CPUs. + */ +static int +pt_backend_disable_smp(struct hwt_context *ctx) +{ + + dprintf("%s\n", __func__); + if (ctx->mode == HWT_MODE_CPU && + atomic_swap_32(&cpu_mode_ctr, 0) == 0) + return (-1); + + if (CPU_EMPTY(&ctx->cpu_map)) { + dprintf("%s: empty cpu map\n", __func__); + return (-1); + } + smp_rendezvous_cpus(ctx->cpu_map, NULL, pt_cpu_stop, NULL, NULL); + + return (0); +} + +/* + * HWT backend initialization method. + * + * Installs the ToPA interrupt handler and initializes + * the tracing contexts used for HWT_MODE_CPU. + */ +static int +pt_backend_init(struct hwt_context *ctx) +{ + struct hwt_cpu *hwt_cpu; + int error; + + dprintf("%s\n", __func__); + if (ctx->mode == HWT_MODE_CPU) { + TAILQ_FOREACH(hwt_cpu, &ctx->cpus, next) { + error = pt_init_ctx(&pt_pcpu_ctx[hwt_cpu->cpu_id], + hwt_cpu->vm, hwt_cpu->cpu_id); + if (error) + return (error); + } + } + + return (0); +} + +/* + * HWT backend teardown method. + * + * Removes the ToPA interrupt handler, stops tracing on all active CPUs, + * and releases all previously allocated ToPA metadata. + */ +static int +pt_backend_deinit(struct hwt_context *ctx) +{ + struct pt_ctx *pt_ctx; + struct hwt_thread *thr; + int cpu_id; + + dprintf("%s\n", __func__); + + pt_backend_disable_smp(ctx); + if (ctx->mode == HWT_MODE_THREAD) { + TAILQ_FOREACH(thr, &ctx->threads, next) { + KASSERT(thr->private != NULL, + ("%s: thr->private not set", __func__)); + pt_ctx = (struct pt_ctx *)thr->private; + pt_deinit_ctx(pt_ctx); + } + } else { + CPU_FOREACH(cpu_id) { + if (!CPU_ISSET(cpu_id, &ctx->cpu_map)) + continue; + if (pt_pcpu[cpu_id].ctx != NULL) { + KASSERT(pt_pcpu[cpu_id].ctx == + &pt_pcpu_ctx[cpu_id], + ("%s: CPU mode tracing with non-cpu mode PT" + "context active", + __func__)); + pt_pcpu[cpu_id].ctx = NULL; + } + pt_ctx = &pt_pcpu_ctx[cpu_id]; + pt_deinit_ctx(pt_ctx); + memset(&pt_pcpu[cpu_id], 0, sizeof(struct pt_cpu)); + } + } + + return (0); +} + +/* + * Fetches current offset into the tracing buffer. + */ +static int +pt_backend_read(struct hwt_vm *vm, int *curpage, vm_offset_t *curpage_offset, + uint64_t *data) +{ + struct pt_buffer *buf; + + if (vm->ctx->mode == HWT_MODE_THREAD) + buf = &((struct pt_ctx *)vm->thr->private)->buf; + else + buf = &pt_pcpu[vm->cpu->cpu_id].ctx->buf; + mtx_lock_spin(&buf->lock); + *curpage = buf->curpage; + *curpage_offset = buf->offset + (buf->wrap_count * vm->ctx->bufsize); + mtx_unlock_spin(&buf->lock); + + return (0); +} + +/* + * HWT thread creation hook. + * Allocates and associates a 'struct pt_ctx' for a given hwt thread. + */ +static int +pt_backend_alloc_thread(struct hwt_thread *thr) +{ + struct pt_ctx *pt_ctx; + int error; + + /* Omit M_WAITOK since this might get invoked a non-sleepable context */ + pt_ctx = malloc(sizeof(*pt_ctx), M_PT, M_NOWAIT | M_ZERO); + if (pt_ctx == NULL) + return (ENOMEM); + + error = pt_init_ctx(pt_ctx, thr->vm, thr->thread_id); + if (error) + return (error); + + thr->private = pt_ctx; + return (0); +} +/* + * HWT thread teardown hook. + */ +static void +pt_backend_free_thread(struct hwt_thread *thr) +{ + struct pt_ctx *ctx; + + ctx = (struct pt_ctx *)thr->private; + + pt_deinit_ctx(ctx); + free(ctx, M_PT); +} + +static void +pt_backend_dump(int cpu_id) +{ +} + +static struct hwt_backend_ops pt_ops = { + .hwt_backend_init = pt_backend_init, + .hwt_backend_deinit = pt_backend_deinit, + + .hwt_backend_configure = pt_backend_configure, + + .hwt_backend_enable = pt_backend_enable, + .hwt_backend_disable = pt_backend_disable, + +#ifdef SMP + .hwt_backend_enable_smp = pt_backend_enable_smp, + .hwt_backend_disable_smp = pt_backend_disable_smp, +#endif + + .hwt_backend_read = pt_backend_read, + .hwt_backend_dump = pt_backend_dump, + + .hwt_backend_thread_alloc = pt_backend_alloc_thread, + .hwt_backend_thread_free = pt_backend_free_thread, +}; + +static struct hwt_backend backend = { + .ops = &pt_ops, + .name = "pt", + .kva_req = 1, +}; + +/* + * Reads the latest valid trace buffer offset and enqueues + * a HWT_RECORD_BUFFER record. + * Used as a taskqueue routine from the ToPA interrupt handler. + */ +static void +pt_send_buffer_record(void *arg, int pending __unused) +{ + struct hwt_record_entry record; + struct pt_ctx *ctx = (struct pt_ctx *)arg; + + /* Prepare buffer record. */ + mtx_lock_spin(&ctx->buf.lock); + pt_fill_buffer_record(ctx->id, &ctx->buf, &record); + mtx_unlock_spin(&ctx->buf.lock); + hwt_record_ctx(ctx->hwt_ctx, &record, M_ZERO | M_NOWAIT); +} +static void +pt_topa_status_clear(void) +{ + uint64_t reg; + + reg = rdmsr(MSR_IA_GLOBAL_STATUS_RESET); + reg &= ~GLOBAL_STATUS_FLAG_TRACETOPAPMI; + reg |= GLOBAL_STATUS_FLAG_TRACETOPAPMI; + wrmsr(MSR_IA_GLOBAL_STATUS_RESET, reg); +} + +/* + * ToPA PMI handler. + * + * Invoked every time a ToPA entry marked with TOPA_INT is filled. + * Uses taskqueue to enqueue a buffer record for userspace. + * Re-enables the PC interrupt line as long as tracing is active. + */ +static int +pt_topa_intr(struct trapframe *tf) +{ + struct pt_buffer *buf; + struct pt_ctx *ctx; + uint64_t reg; + + SDT_PROBE0(pt, , , topa__intr); + + if (pt_cpu_get_state(curcpu) != PT_ACTIVE) { + return (0); + } + reg = rdmsr(MSR_IA_GLOBAL_STATUS); + if ((reg & GLOBAL_STATUS_FLAG_TRACETOPAPMI) == 0) { + /* ACK spurious or leftover interrupt. */ + pt_topa_status_clear(); + return (1); + } + + ctx = pt_pcpu[curcpu].ctx; + buf = &ctx->buf; + KASSERT(buf->topa_hw != NULL, + ("%s: ToPA PMI interrupt with invalid buffer", __func__)); + + pt_cpu_toggle_local(ctx->save_area, false); + pt_update_buffer(buf); + pt_topa_status_clear(); + taskqueue_enqueue_flags(taskqueue_pt, &ctx->task, + TASKQUEUE_FAIL_IF_PENDING); + + if (pt_cpu_get_state(curcpu) == PT_ACTIVE) { + pt_cpu_toggle_local(ctx->save_area, true); + lapic_reenable_pcint(); + } + return (1); +} + +/* + * Module initialization. + * + * Saves all PT-related cpuid info, registers itself as a HWT backend, + * and allocates metadata required to keep track of tracing operations + * on each CPU. + */ +static int +pt_init(void) +{ + u_int cp[4]; + int error; + + dprintf("pt: Enumerating part 1\n"); + cpuid_count(CPUID_PT_LEAF, 0, cp); + dprintf("pt: Maximum valid sub-leaf Index: %x\n", cp[0]); + dprintf("pt: ebx %x\n", cp[1]); + dprintf("pt: ecx %x\n", cp[2]); + + pt_info.l0_eax = cp[0]; + pt_info.l0_ebx = cp[1]; + pt_info.l0_ecx = cp[2]; + + dprintf("pt: Enumerating part 2\n"); + cpuid_count(CPUID_PT_LEAF, 1, cp); + dprintf("pt: eax %x\n", cp[0]); + dprintf("pt: ebx %x\n", cp[1]); + + pt_info.l1_eax = cp[0]; + pt_info.l1_ebx = cp[1]; + + error = hwt_backend_register(&backend); + if (error != 0) { + printf("pt: unable to register hwt backend, error %d\n", error); + return (error); + } + pt_pcpu = mallocarray(mp_ncpus, sizeof(struct pt_cpu), M_PT, + M_ZERO | M_WAITOK); + pt_pcpu_ctx = mallocarray(mp_ncpus, sizeof(struct pt_ctx), M_PT, + M_ZERO | M_WAITOK); + + nmi_register_handler(pt_topa_intr); + if (!lapic_enable_pcint()) { + nmi_remove_handler(pt_topa_intr); + hwt_backend_unregister(&backend); + free(pt_pcpu, M_PT); + free(pt_pcpu_ctx, M_PT); + pt_pcpu = NULL; + pt_pcpu_ctx = NULL; + printf("pt: failed to setup interrupt line\n"); + return (error); + } + initialized = true; + + return (0); +} + +/* + * Checks whether the CPU support Intel PT and + * initializes XSAVE area info. + * + * The driver relies on XSAVE/XRSTOR PT extensions, + * Table of Physical Addresses (ToPA) support, and + * support for multiple ToPA entries. + */ +static bool +pt_supported(void) +{ + u_int cp[4]; + + if ((cpu_stdext_feature & CPUID_STDEXT_PROCTRACE) == 0) { + printf("pt: CPU does not support Intel Processor Trace\n"); + return (false); + } + if ((cpu_feature2 & CPUID2_XSAVE) == 0) { + printf("pt: XSAVE is not supported\n"); + return (false); + } + if (!xsave_extfeature_supported(XFEATURE_ENABLED_PT, true)) { + printf("pt: CPU does not support managing PT state using XSAVE\n"); + return (false); + } + if (!xsave_extension_supported(CPUID_EXTSTATE_XSAVEC)) { + printf("pt: XSAVE compaction is not supported\n"); + return (false); + } + if (!xsave_extension_supported(CPUID_EXTSTATE_XSAVES)) { + printf("pt: CPU does not support XSAVES/XRSTORS\n"); + return (false); + } + + /* Require ToPA support. */ + cpuid_count(CPUID_PT_LEAF, 0, cp); + if ((cp[2] & CPUPT_TOPA) == 0) { + printf("pt: ToPA is not supported\n"); + return (false); + } + if ((cp[2] & CPUPT_TOPA_MULTI) == 0) { + printf("pt: multiple ToPA outputs are not supported\n"); + return (false); + } + + pt_info.xstate_hdr_offset = xsave_area_hdr_offset(); + pt_info.xsave_area_size = xsave_area_size(PT_XSTATE_BV, true, true); + pt_info.pt_xsave_offset = xsave_area_offset(PT_XSTATE_BV, + XFEATURE_ENABLED_PT, true, true); + + return (true); +} + +static void +pt_deinit(void) +{ + if (!initialized) + return; + nmi_remove_handler(pt_topa_intr); + lapic_disable_pcint(); + hwt_backend_unregister(&backend); + free(pt_pcpu, M_PT); + free(pt_pcpu_ctx, M_PT); + pt_pcpu = NULL; + initialized = false; +} + +static int +pt_modevent(module_t mod, int type, void *data) +{ + switch (type) { + case MOD_LOAD: + if (!pt_supported() || pt_init() != 0) { + return (ENXIO); + } + break; + case MOD_UNLOAD: + pt_deinit(); + break; + default: + break; + } + + return (0); +} + +static moduledata_t pt_mod = { "intel_pt", pt_modevent, NULL }; + +DECLARE_MODULE(intel_pt, pt_mod, SI_SUB_DRIVERS, SI_ORDER_FIRST); +MODULE_DEPEND(intel_pt, hwt, 1, 1, 1); +MODULE_VERSION(intel_pt, 1); diff --git a/sys/amd64/pt/pt.h b/sys/amd64/pt/pt.h new file mode 100644 index 000000000000..2423afdf22e9 --- /dev/null +++ b/sys/amd64/pt/pt.h @@ -0,0 +1,49 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2023 Bojan Novković <bnovkov@freebsd.org> + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#ifndef _AMD64_PT_PT_H_ +#define _AMD64_PT_PT_H_ + +#include <sys/types.h> + +#include <x86/include/specialreg.h> + +#define PT_IP_FILTER_MAX_RANGES (2) /* Intel SDM Vol. 3C, 33-29 */ + +struct pt_cpu_config { + uint64_t rtit_ctl; + register_t cr3_filter; + int nranges; + struct ipf_range { + vm_offset_t start; + vm_offset_t end; + } ip_ranges[PT_IP_FILTER_MAX_RANGES]; + uint32_t mtc_freq; + uint32_t cyc_thresh; + uint32_t psb_freq; +}; +#endif /* !_AMD64_PT_PT_H_ */ diff --git a/sys/amd64/vmm/amd/svm.c b/sys/amd64/vmm/amd/svm.c index 6c16daaa47c2..2fe6a5bc3584 100644 --- a/sys/amd64/vmm/amd/svm.c +++ b/sys/amd64/vmm/amd/svm.c @@ -317,6 +317,33 @@ svm_set_tsc_offset(struct svm_vcpu *vcpu, uint64_t offset) #define MSR_AMD7TH_START 0xC0010000UL #define MSR_AMD7TH_END 0xC0011FFFUL +static void +svm_get_cs_info(struct vmcb *vmcb, struct vm_guest_paging *paging, int *cs_d, + uint64_t *base) +{ + struct vmcb_segment seg; + int error __diagused; + + error = vmcb_seg(vmcb, VM_REG_GUEST_CS, &seg); + KASSERT(error == 0, ("%s: vmcb_seg error %d", __func__, error)); + + switch (paging->cpu_mode) { + case CPU_MODE_REAL: + *base = seg.base; + *cs_d = 0; + break; + case CPU_MODE_PROTECTED: + case CPU_MODE_COMPATIBILITY: + *cs_d = !!(seg.attrib & VMCB_CS_ATTRIB_D); + *base = seg.base; + break; + default: + *base = 0; + *cs_d = 0; + break; + } +} + /* * Get the index and bit position for a MSR in permission bitmap. * Two bits are used for each MSR: lower bit for read and higher bit for write. @@ -735,10 +762,29 @@ svm_inout_str_seginfo(struct svm_vcpu *vcpu, int64_t info1, int in, if (in) { vis->seg_name = VM_REG_GUEST_ES; - } else { - /* The segment field has standard encoding */ + } else if (decode_assist()) { + /* + * The effective segment number in EXITINFO1[12:10] is populated + * only if the processor has the DecodeAssist capability. + * + * XXX this is not specified explicitly in APMv2 but can be + * verified empirically. + */ s = (info1 >> 10) & 0x7; + + /* The segment field has standard encoding */ vis->seg_name = vm_segment_name(s); + } else { + /* + * The segment register need to be manually decoded by fetching + * the instructions near ip. However, we are unable to fetch it + * while the interrupts are disabled. Therefore, we leave the + * value unset until the generic ins/outs handler runs. + */ + vis->seg_name = VM_REG_LAST; + svm_get_cs_info(vcpu->vmcb, &vis->paging, &vis->cs_d, + &vis->cs_base); + return; } error = svm_getdesc(vcpu, vis->seg_name, &vis->seg_desc); @@ -798,16 +844,6 @@ svm_handle_io(struct svm_vcpu *vcpu, struct vm_exit *vmexit) info1 = ctrl->exitinfo1; inout_string = info1 & BIT(2) ? 1 : 0; - /* - * The effective segment number in EXITINFO1[12:10] is populated - * only if the processor has the DecodeAssist capability. - * - * XXX this is not specified explicitly in APMv2 but can be verified - * empirically. - */ - if (inout_string && !decode_assist()) - return (UNHANDLED); - vmexit->exitcode = VM_EXITCODE_INOUT; vmexit->u.inout.in = (info1 & BIT(0)) ? 1 : 0; vmexit->u.inout.string = inout_string; @@ -825,6 +861,8 @@ svm_handle_io(struct svm_vcpu *vcpu, struct vm_exit *vmexit) vis->index = svm_inout_str_index(regs, vmexit->u.inout.in); vis->count = svm_inout_str_count(regs, vmexit->u.inout.rep); vis->addrsize = svm_inout_str_addrsize(info1); + vis->cs_d = 0; + vis->cs_base = 0; svm_inout_str_seginfo(vcpu, info1, vmexit->u.inout.in, vis); } @@ -866,10 +904,9 @@ static void svm_handle_inst_emul(struct vmcb *vmcb, uint64_t gpa, struct vm_exit *vmexit) { struct vm_guest_paging *paging; - struct vmcb_segment seg; struct vmcb_ctrl *ctrl; char *inst_bytes; - int error __diagused, inst_len; + int inst_len; ctrl = &vmcb->ctrl; paging = &vmexit->u.inst_emul.paging; @@ -879,29 +916,8 @@ svm_handle_inst_emul(struct vmcb *vmcb, uint64_t gpa, struct vm_exit *vmexit) vmexit->u.inst_emul.gla = VIE_INVALID_GLA; svm_paging_info(vmcb, paging); - error = vmcb_seg(vmcb, VM_REG_GUEST_CS, &seg); - KASSERT(error == 0, ("%s: vmcb_seg(CS) error %d", __func__, error)); - - switch(paging->cpu_mode) { - case CPU_MODE_REAL: - vmexit->u.inst_emul.cs_base = seg.base; - vmexit->u.inst_emul.cs_d = 0; - break; - case CPU_MODE_PROTECTED: - case CPU_MODE_COMPATIBILITY: - vmexit->u.inst_emul.cs_base = seg.base; - - /* - * Section 4.8.1 of APM2, Default Operand Size or D bit. - */ - vmexit->u.inst_emul.cs_d = (seg.attrib & VMCB_CS_ATTRIB_D) ? - 1 : 0; - break; - default: - vmexit->u.inst_emul.cs_base = 0; - vmexit->u.inst_emul.cs_d = 0; - break; - } + svm_get_cs_info(vmcb, paging, &vmexit->u.inst_emul.cs_d, + &vmexit->u.inst_emul.cs_base); /* * Copy the instruction bytes into 'vie' if available. diff --git a/sys/amd64/vmm/intel/vmx.c b/sys/amd64/vmm/intel/vmx.c index 957217ab2258..842281ab862e 100644 --- a/sys/amd64/vmm/intel/vmx.c +++ b/sys/amd64/vmm/intel/vmx.c @@ -2659,6 +2659,8 @@ vmx_exit_process(struct vmx *vmx, struct vmx_vcpu *vcpu, struct vm_exit *vmexit) vis->index = inout_str_index(vcpu, in); vis->count = inout_str_count(vcpu, vis->inout.rep); vis->addrsize = inout_str_addrsize(inst_info); + vis->cs_d = 0; + vis->cs_base = 0; inout_str_seginfo(vcpu, inst_info, in, vis); } SDT_PROBE3(vmm, vmx, exit, inout, vmx, vcpuid, vmexit); diff --git a/sys/amd64/vmm/intel/vmx_support.S b/sys/amd64/vmm/intel/vmx_support.S index f393f160b101..130130b64541 100644 --- a/sys/amd64/vmm/intel/vmx_support.S +++ b/sys/amd64/vmm/intel/vmx_support.S @@ -32,12 +32,6 @@ #include "vmx_assym.h" -#ifdef SMP -#define LK lock ; -#else -#define LK -#endif - /* Be friendly to DTrace FBT's prologue/epilogue pattern matching */ #define VENTER push %rbp ; mov %rsp,%rbp #define VLEAVE pop %rbp diff --git a/sys/amd64/vmm/vmm_instruction_emul.c b/sys/amd64/vmm/vmm_instruction_emul.c index c53e32889000..c54b6e6d0074 100644 --- a/sys/amd64/vmm/vmm_instruction_emul.c +++ b/sys/amd64/vmm/vmm_instruction_emul.c @@ -65,30 +65,6 @@ #include <x86/psl.h> #include <x86/specialreg.h> -/* struct vie_op.op_type */ -enum { - VIE_OP_TYPE_NONE = 0, - VIE_OP_TYPE_MOV, - VIE_OP_TYPE_MOVSX, - VIE_OP_TYPE_MOVZX, - VIE_OP_TYPE_AND, - VIE_OP_TYPE_OR, - VIE_OP_TYPE_SUB, - VIE_OP_TYPE_TWO_BYTE, - VIE_OP_TYPE_PUSH, - VIE_OP_TYPE_CMP, - VIE_OP_TYPE_POP, - VIE_OP_TYPE_MOVS, - VIE_OP_TYPE_GROUP1, - VIE_OP_TYPE_STOS, - VIE_OP_TYPE_BITTEST, - VIE_OP_TYPE_TWOB_GRP15, - VIE_OP_TYPE_ADD, - VIE_OP_TYPE_TEST, - VIE_OP_TYPE_BEXTR, - VIE_OP_TYPE_LAST -}; - /* struct vie_op.op_flags */ #define VIE_OP_F_IMM (1 << 0) /* 16/32-bit immediate operand */ #define VIE_OP_F_IMM8 (1 << 1) /* 8-bit immediate operand */ @@ -152,6 +128,16 @@ static const struct vie_op one_byte_opcodes[256] = { .op_byte = 0x3B, .op_type = VIE_OP_TYPE_CMP, }, + [0x6E] = { + .op_byte = 0x6E, + .op_type = VIE_OP_TYPE_OUTS, + .op_flags = VIE_OP_F_NO_MODRM | VIE_OP_F_NO_GLA_VERIFICATION, + }, + [0x6F] = { + .op_byte = 0x6F, + .op_type = VIE_OP_TYPE_OUTS, + .op_flags = VIE_OP_F_NO_MODRM | VIE_OP_F_NO_GLA_VERIFICATION, + }, [0x88] = { .op_byte = 0x88, .op_type = VIE_OP_TYPE_MOV, diff --git a/sys/amd64/vmm/vmm_ioport.c b/sys/amd64/vmm/vmm_ioport.c index fc1ecab9f209..8aab28f5e68e 100644 --- a/sys/amd64/vmm/vmm_ioport.c +++ b/sys/amd64/vmm/vmm_ioport.c @@ -145,9 +145,49 @@ emulate_inout_port(struct vcpu *vcpu, struct vm_exit *vmexit, bool *retu) } static int +decode_segment(struct vcpu *vcpu, enum vm_reg_name *segment) +{ + struct vm_guest_paging *paging; + struct vie vie; + struct vm_exit *vme; + int err; + int fault; + + vme = vm_exitinfo(vcpu); + paging = &vme->u.inout_str.paging; + + vie_init(&vie, NULL, 0); + err = vmm_fetch_instruction(vcpu, paging, + vme->rip + vme->u.inout_str.cs_base, VIE_INST_SIZE, &vie, &fault); + if (err || fault) + return (err); + + err = vmm_decode_instruction(vcpu, VIE_INVALID_GLA, paging->cpu_mode, + vme->u.inout_str.cs_d, &vie); + + if (err || vie.op.op_type != VIE_OP_TYPE_OUTS) + return (EINVAL); + if (vie.segment_override) + *segment = vie.segment_register; + else + *segment = VM_REG_GUEST_DS; + + return (0); +} + +static int emulate_inout_str(struct vcpu *vcpu, struct vm_exit *vmexit, bool *retu) { + int err; + *retu = true; + if (vmexit->u.inout_str.seg_name == VM_REG_LAST) { + err = decode_segment(vcpu, &vmexit->u.inout_str.seg_name); + if (err) + return (err); + return (vm_get_seg_desc(vcpu, vmexit->u.inout_str.seg_name, + &vmexit->u.inout_str.seg_desc)); + } return (0); /* Return to userspace to finish emulation */ } diff --git a/sys/arm/allwinner/aw_mmc.c b/sys/arm/allwinner/aw_mmc.c index 6bebf5e5fb5e..a8add957dc74 100644 --- a/sys/arm/allwinner/aw_mmc.c +++ b/sys/arm/allwinner/aw_mmc.c @@ -84,21 +84,26 @@ struct aw_mmc_conf { uint32_t dma_xferlen; + uint32_t dma_desc_shift; bool mask_data0; bool can_calibrate; bool new_timing; + bool zero_is_skip; }; static const struct aw_mmc_conf a10_mmc_conf = { .dma_xferlen = 0x2000, + .dma_desc_shift = 0, }; static const struct aw_mmc_conf a13_mmc_conf = { .dma_xferlen = 0x10000, + .dma_desc_shift = 0, }; static const struct aw_mmc_conf a64_mmc_conf = { .dma_xferlen = 0x10000, + .dma_desc_shift = 0, .mask_data0 = true, .can_calibrate = true, .new_timing = true, @@ -106,13 +111,24 @@ static const struct aw_mmc_conf a64_mmc_conf = { static const struct aw_mmc_conf a64_emmc_conf = { .dma_xferlen = 0x2000, + .dma_desc_shift = 0, .can_calibrate = true, }; +static const struct aw_mmc_conf d1_mmc_conf = { + .dma_xferlen = 0x1000, + .dma_desc_shift = 2, + .mask_data0 = true, + .can_calibrate = true, + .new_timing = true, + .zero_is_skip = true, +}; + static struct ofw_compat_data compat_data[] = { {"allwinner,sun4i-a10-mmc", (uintptr_t)&a10_mmc_conf}, {"allwinner,sun5i-a13-mmc", (uintptr_t)&a13_mmc_conf}, {"allwinner,sun7i-a20-mmc", (uintptr_t)&a13_mmc_conf}, + {"allwinner,sun20i-d1-mmc", (uintptr_t)&d1_mmc_conf}, {"allwinner,sun50i-a64-mmc", (uintptr_t)&a64_mmc_conf}, {"allwinner,sun50i-a64-emmc", (uintptr_t)&a64_emmc_conf}, {NULL, 0} @@ -607,16 +623,18 @@ aw_dma_cb(void *arg, bus_dma_segment_t *segs, int nsegs, int err) dma_desc = sc->aw_dma_desc; for (i = 0; i < nsegs; i++) { - if (segs[i].ds_len == sc->aw_mmc_conf->dma_xferlen) + if ((segs[i].ds_len == sc->aw_mmc_conf->dma_xferlen) && + !sc->aw_mmc_conf->zero_is_skip) dma_desc[i].buf_size = 0; /* Size of 0 indicate max len */ else dma_desc[i].buf_size = segs[i].ds_len; - dma_desc[i].buf_addr = segs[i].ds_addr; + dma_desc[i].buf_addr = segs[i].ds_addr >> + sc->aw_mmc_conf->dma_desc_shift; dma_desc[i].config = AW_MMC_DMA_CONFIG_CH | - AW_MMC_DMA_CONFIG_OWN | AW_MMC_DMA_CONFIG_DIC; - - dma_desc[i].next = sc->aw_dma_desc_phys + - ((i + 1) * sizeof(struct aw_mmc_dma_desc)); + AW_MMC_DMA_CONFIG_OWN | AW_MMC_DMA_CONFIG_DIC; + dma_desc[i].next = (sc->aw_dma_desc_phys + + (i + 1) * sizeof(struct aw_mmc_dma_desc)) >> + sc->aw_mmc_conf->dma_desc_shift; } dma_desc[0].config |= AW_MMC_DMA_CONFIG_FD; @@ -678,7 +696,8 @@ aw_mmc_prepare_dma(struct aw_mmc_softc *sc) AW_MMC_WRITE_4(sc, AW_MMC_IDIE, val); /* Set DMA descritptor list address */ - AW_MMC_WRITE_4(sc, AW_MMC_DLBA, sc->aw_dma_desc_phys); + AW_MMC_WRITE_4(sc, AW_MMC_DLBA, sc->aw_dma_desc_phys >> + sc->aw_mmc_conf->dma_desc_shift); /* FIFO trigger level */ AW_MMC_WRITE_4(sc, AW_MMC_FWLR, AW_MMC_DMA_FTRGLEVEL); diff --git a/sys/arm/arm/pmap-v6.c b/sys/arm/arm/pmap-v6.c index 92eb0589f80b..78883296c5b7 100644 --- a/sys/arm/arm/pmap-v6.c +++ b/sys/arm/arm/pmap-v6.c @@ -5767,7 +5767,7 @@ pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma) CTR5(KTR_PMAP, "%s: page %p - 0x%08X oma: %d, ma: %d", __func__, m, VM_PAGE_TO_PHYS(m), oma, ma); - if ((m->flags & PG_FICTITIOUS) != 0) + if (ma == oma || (m->flags & PG_FICTITIOUS) != 0) return; #if 0 /* @@ -5784,22 +5784,20 @@ pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma) * If page is not mapped by sf buffer, map the page * transient and do invalidation. */ - if (ma != oma) { - pa = VM_PAGE_TO_PHYS(m); - sched_pin(); - pc = get_pcpu(); - cmap2_pte2p = pc->pc_cmap2_pte2p; - mtx_lock(&pc->pc_cmap_lock); - if (pte2_load(cmap2_pte2p) != 0) - panic("%s: CMAP2 busy", __func__); - pte2_store(cmap2_pte2p, PTE2_KERN_NG(pa, PTE2_AP_KRW, - vm_memattr_to_pte2(ma))); - dcache_wbinv_poc((vm_offset_t)pc->pc_cmap2_addr, pa, PAGE_SIZE); - pte2_clear(cmap2_pte2p); - tlb_flush((vm_offset_t)pc->pc_cmap2_addr); - sched_unpin(); - mtx_unlock(&pc->pc_cmap_lock); - } + pa = VM_PAGE_TO_PHYS(m); + sched_pin(); + pc = get_pcpu(); + cmap2_pte2p = pc->pc_cmap2_pte2p; + mtx_lock(&pc->pc_cmap_lock); + if (pte2_load(cmap2_pte2p) != 0) + panic("%s: CMAP2 busy", __func__); + pte2_store(cmap2_pte2p, PTE2_KERN_NG(pa, PTE2_AP_KRW, + vm_memattr_to_pte2(ma))); + dcache_wbinv_poc((vm_offset_t)pc->pc_cmap2_addr, pa, PAGE_SIZE); + pte2_clear(cmap2_pte2p); + tlb_flush((vm_offset_t)pc->pc_cmap2_addr); + sched_unpin(); + mtx_unlock(&pc->pc_cmap_lock); } /* diff --git a/sys/arm64/arm64/pmap.c b/sys/arm64/arm64/pmap.c index d2e56a270f54..2152f7fcc1c6 100644 --- a/sys/arm64/arm64/pmap.c +++ b/sys/arm64/arm64/pmap.c @@ -497,7 +497,8 @@ static bool pmap_pv_insert_l3c(pmap_t pmap, vm_offset_t va, vm_page_t m, struct rwlock **lockp); static void pmap_remove_kernel_l2(pmap_t pmap, pt_entry_t *l2, vm_offset_t va); static int pmap_remove_l2(pmap_t pmap, pt_entry_t *l2, vm_offset_t sva, - pd_entry_t l1e, struct spglist *free, struct rwlock **lockp); + pd_entry_t l1e, bool demote_kl2e, struct spglist *free, + struct rwlock **lockp); static int pmap_remove_l3(pmap_t pmap, pt_entry_t *l3, vm_offset_t sva, pd_entry_t l2e, struct spglist *free, struct rwlock **lockp); static bool pmap_remove_l3c(pmap_t pmap, pt_entry_t *l3p, vm_offset_t va, @@ -3847,8 +3848,7 @@ pmap_remove_kernel_l2(pmap_t pmap, pt_entry_t *l2, vm_offset_t va) PMAP_LOCK_ASSERT(pmap, MA_OWNED); ml3 = pmap_remove_pt_page(pmap, va); - if (ml3 == NULL) - panic("pmap_remove_kernel_l2: Missing pt page"); + KASSERT(ml3 != NULL, ("pmap_remove_kernel_l2: missing pt page")); ml3pa = VM_PAGE_TO_PHYS(ml3); newl2 = PHYS_TO_PTE(ml3pa) | L2_TABLE; @@ -3873,8 +3873,8 @@ pmap_remove_kernel_l2(pmap_t pmap, pt_entry_t *l2, vm_offset_t va) * pmap_remove_l2: Do the things to unmap a level 2 superpage. */ static int -pmap_remove_l2(pmap_t pmap, pt_entry_t *l2, vm_offset_t sva, - pd_entry_t l1e, struct spglist *free, struct rwlock **lockp) +pmap_remove_l2(pmap_t pmap, pt_entry_t *l2, vm_offset_t sva, pd_entry_t l1e, + bool demote_kl2e, struct spglist *free, struct rwlock **lockp) { struct md_page *pvh; pt_entry_t old_l2; @@ -3910,9 +3910,7 @@ pmap_remove_l2(pmap_t pmap, pt_entry_t *l2, vm_offset_t sva, vm_page_aflag_clear(mt, PGA_WRITEABLE); } } - if (pmap == kernel_pmap) { - pmap_remove_kernel_l2(pmap, l2, sva); - } else { + if (pmap != kernel_pmap) { ml3 = pmap_remove_pt_page(pmap, sva); if (ml3 != NULL) { KASSERT(vm_page_any_valid(ml3), @@ -3923,6 +3921,14 @@ pmap_remove_l2(pmap_t pmap, pt_entry_t *l2, vm_offset_t sva, ml3->ref_count = 0; pmap_add_delayed_free_list(ml3, free, false); } + } else if (demote_kl2e) { + pmap_remove_kernel_l2(pmap, l2, sva); + } else { + ml3 = vm_radix_lookup(&pmap->pm_root, pmap_l2_pindex(sva)); + if (vm_page_any_valid(ml3)) { + ml3->valid = 0; + pmap_zero_page(ml3); + } } return (pmap_unuse_pt(pmap, sva, l1e, free)); } @@ -4232,7 +4238,7 @@ pmap_remove1(pmap_t pmap, vm_offset_t sva, vm_offset_t eva, bool map_delete) if ((l3_paddr & ATTR_DESCR_MASK) == L2_BLOCK) { if (sva + L2_SIZE == va_next && eva >= va_next) { pmap_remove_l2(pmap, l2, sva, pmap_load(l1), - &free, &lock); + true, &free, &lock); continue; } else if (pmap_demote_l2_locked(pmap, l2, sva, &lock) == NULL) @@ -5703,6 +5709,9 @@ pmap_enter_l2(pmap_t pmap, vm_offset_t va, pd_entry_t new_l2, u_int flags, PMAP_LOCK_ASSERT(pmap, MA_OWNED); KASSERT(ADDR_IS_CANONICAL(va), ("%s: Address not in canonical form: %lx", __func__, va)); + KASSERT((flags & (PMAP_ENTER_NOREPLACE | PMAP_ENTER_NORECLAIM)) != + PMAP_ENTER_NORECLAIM, + ("pmap_enter_l2: flags is missing PMAP_ENTER_NOREPLACE")); if ((l2 = pmap_alloc_l2(pmap, va, &l2pg, (flags & PMAP_ENTER_NOSLEEP) != 0 ? NULL : lockp)) == NULL) { @@ -5747,33 +5756,51 @@ pmap_enter_l2(pmap_t pmap, vm_offset_t va, pd_entry_t new_l2, u_int flags, } } SLIST_INIT(&free); - if ((old_l2 & ATTR_DESCR_MASK) == L2_BLOCK) + if ((old_l2 & ATTR_DESCR_MASK) == L2_BLOCK) { (void)pmap_remove_l2(pmap, l2, va, - pmap_load(pmap_l1(pmap, va)), &free, lockp); - else + pmap_load(pmap_l1(pmap, va)), false, &free, lockp); + } else { + if (ADDR_IS_KERNEL(va)) { + /* + * Try to save the ptp in the trie + * before any changes to mappings are + * made. Abort on failure. + */ + mt = PTE_TO_VM_PAGE(old_l2); + if (pmap_insert_pt_page(pmap, mt, false, + false)) { + CTR1(KTR_PMAP, + "pmap_enter_l2: cannot ins kern ptp va %#lx", + va); + return (KERN_RESOURCE_SHORTAGE); + } + /* + * Both pmap_remove_l2() and + * pmap_remove_l3_range() will zero fill + * the L3 kernel page table page. + */ + } pmap_remove_l3_range(pmap, old_l2, va, va + L2_SIZE, &free, lockp); + if (ADDR_IS_KERNEL(va)) { + /* + * The TLB could have an intermediate + * entry for the L3 kernel page table + * page, so request an invalidation at + * all levels after clearing the + * L2_TABLE entry. + */ + pmap_clear(l2); + pmap_s1_invalidate_page(pmap, va, false); + } + } + KASSERT(pmap_load(l2) == 0, + ("pmap_enter_l2: non-zero L2 entry %p", l2)); if (!ADDR_IS_KERNEL(va)) { vm_page_free_pages_toq(&free, true); - KASSERT(pmap_load(l2) == 0, - ("pmap_enter_l2: non-zero L2 entry %p", l2)); } else { KASSERT(SLIST_EMPTY(&free), ("pmap_enter_l2: freed kernel page table page")); - - /* - * Both pmap_remove_l2() and pmap_remove_l3_range() - * will leave the kernel page table page zero filled. - * Nonetheless, the TLB could have an intermediate - * entry for the kernel page table page, so request - * an invalidation at all levels after clearing - * the L2_TABLE entry. - */ - mt = PTE_TO_VM_PAGE(pmap_load(l2)); - if (pmap_insert_pt_page(pmap, mt, false, false)) - panic("pmap_enter_l2: trie insert failed"); - pmap_clear(l2); - pmap_s1_invalidate_page(pmap, va, false); } } @@ -5804,6 +5831,15 @@ pmap_enter_l2(pmap_t pmap, vm_offset_t va, pd_entry_t new_l2, u_int flags, if (!pmap_pv_insert_l2(pmap, va, new_l2, flags, lockp)) { if (l2pg != NULL) pmap_abort_ptp(pmap, va, l2pg); + else { + KASSERT(ADDR_IS_KERNEL(va) && + (pmap_load(l2) & ATTR_DESCR_MASK) == + L2_TABLE, + ("pmap_enter_l2: invalid kernel L2E")); + mt = pmap_remove_pt_page(pmap, va); + KASSERT(mt != NULL, + ("pmap_enter_l2: missing kernel PTP")); + } if (uwptpg != NULL) { mt = pmap_remove_pt_page(pmap, va); KASSERT(mt == uwptpg, @@ -8045,6 +8081,8 @@ pmap_unmapbios(void *p, vm_size_t size) void pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma) { + if (m->md.pv_memattr == ma) + return; m->md.pv_memattr = ma; @@ -8424,8 +8462,8 @@ pmap_demote_l2_abort(pmap_t pmap, vm_offset_t va, pt_entry_t *l2, struct spglist free; SLIST_INIT(&free); - (void)pmap_remove_l2(pmap, l2, va, pmap_load(pmap_l1(pmap, va)), &free, - lockp); + (void)pmap_remove_l2(pmap, l2, va, pmap_load(pmap_l1(pmap, va)), true, + &free, lockp); vm_page_free_pages_toq(&free, true); } @@ -8463,18 +8501,20 @@ pmap_demote_l2_locked(pmap_t pmap, pt_entry_t *l2, vm_offset_t va, /* * Invalidate the 2MB page mapping and return "failure" if the - * mapping was never accessed. + * mapping was never accessed and not wired. */ if ((oldl2 & ATTR_AF) == 0) { - KASSERT((oldl2 & ATTR_SW_WIRED) == 0, - ("pmap_demote_l2: a wired mapping is missing ATTR_AF")); - pmap_demote_l2_abort(pmap, va, l2, lockp); - CTR2(KTR_PMAP, "pmap_demote_l2: failure for va %#lx in pmap %p", - va, pmap); - goto fail; - } - - if ((ml3 = pmap_remove_pt_page(pmap, va)) == NULL) { + if ((oldl2 & ATTR_SW_WIRED) == 0) { + pmap_demote_l2_abort(pmap, va, l2, lockp); + CTR2(KTR_PMAP, + "pmap_demote_l2: failure for va %#lx in pmap %p", + va, pmap); + goto fail; + } + ml3 = pmap_remove_pt_page(pmap, va); + /* Fill the PTP with L3Es that have ATTR_AF cleared. */ + ml3->valid = 0; + } else if ((ml3 = pmap_remove_pt_page(pmap, va)) == NULL) { KASSERT((oldl2 & ATTR_SW_WIRED) == 0, ("pmap_demote_l2: page table page for a wired mapping" " is missing")); @@ -8530,7 +8570,7 @@ pmap_demote_l2_locked(pmap_t pmap, pt_entry_t *l2, vm_offset_t va, /* * If the PTP is not leftover from an earlier promotion or it does not * have ATTR_AF set in every L3E, then fill it. The new L3Es will all - * have ATTR_AF set. + * have ATTR_AF set, unless this is a wired mapping with ATTR_AF clear. * * When pmap_update_entry() clears the old L2 mapping, it (indirectly) * performs a dsb(). That dsb() ensures that the stores for filling diff --git a/sys/arm64/broadcom/genet/if_genet.c b/sys/arm64/broadcom/genet/if_genet.c index 0602f076b257..182b5582fb7c 100644 --- a/sys/arm64/broadcom/genet/if_genet.c +++ b/sys/arm64/broadcom/genet/if_genet.c @@ -349,7 +349,7 @@ gen_attach(device_t dev) } /* If address was not found, create one based on the hostid and name. */ - if (eaddr_found == 0) + if (!eaddr_found) ether_gen_addr(sc->ifp, &eaddr); /* Attach ethernet interface */ ether_ifattach(sc->ifp, eaddr.octet); @@ -653,7 +653,7 @@ gen_bus_dma_teardown(struct gen_softc *sc) error); } - if (sc->tx_buf_tag != NULL) { + if (sc->rx_buf_tag != NULL) { for (i = 0; i < RX_DESC_COUNT; i++) { error = bus_dmamap_destroy(sc->rx_buf_tag, sc->rx_ring_ent[i].map); diff --git a/sys/arm64/include/vmm_dev.h b/sys/arm64/include/vmm_dev.h index 938bea47c7f8..219f1116c728 100644 --- a/sys/arm64/include/vmm_dev.h +++ b/sys/arm64/include/vmm_dev.h @@ -27,6 +27,8 @@ #ifndef _VMM_DEV_H_ #define _VMM_DEV_H_ +#include <sys/domainset.h> + #include <machine/vmm.h> struct vm_memmap { @@ -49,6 +51,9 @@ struct vm_memseg { int segid; size_t len; char name[VM_MAX_SUFFIXLEN + 1]; + domainset_t *ds_mask; + size_t ds_mask_size; + int ds_policy; }; struct vm_register { diff --git a/sys/cam/ata/ata_da.c b/sys/cam/ata/ata_da.c index ae7cf14c8f8e..1facab47473c 100644 --- a/sys/cam/ata/ata_da.c +++ b/sys/cam/ata/ata_da.c @@ -1359,10 +1359,7 @@ adaasync(void *callback_arg, uint32_t code, case AC_GETDEV_CHANGED: { softc = (struct ada_softc *)periph->softc; - memset(&cgd, 0, sizeof(cgd)); - xpt_setup_ccb(&cgd.ccb_h, periph->path, CAM_PRIORITY_NORMAL); - cgd.ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)&cgd); + xpt_gdev_type(&cgd, periph->path); /* * Update our information based on the new Identify data. diff --git a/sys/cam/cam_periph.c b/sys/cam/cam_periph.c index 833df6cfb99b..730656684e2a 100644 --- a/sys/cam/cam_periph.c +++ b/sys/cam/cam_periph.c @@ -767,27 +767,28 @@ camperiphfree(struct cam_periph *periph) CAM_DEBUG(periph->path, CAM_DEBUG_INFO, ("Periph destroyed\n")); if (periph->flags & CAM_PERIPH_NEW_DEV_FOUND) { - union ccb ccb; - void *arg; - - memset(&ccb, 0, sizeof(ccb)); switch (periph->deferred_ac) { - case AC_FOUND_DEVICE: - ccb.ccb_h.func_code = XPT_GDEV_TYPE; - xpt_setup_ccb(&ccb.ccb_h, periph->path, CAM_PRIORITY_NORMAL); - xpt_action(&ccb); - arg = &ccb; + case AC_FOUND_DEVICE: { + struct ccb_getdev cgd; + + xpt_gdev_type(&cgd, periph->path); + periph->deferred_callback(NULL, periph->deferred_ac, + periph->path, &cgd); break; - case AC_PATH_REGISTERED: - xpt_path_inq(&ccb.cpi, periph->path); - arg = &ccb; + } + case AC_PATH_REGISTERED: { + struct ccb_pathinq cpi; + + xpt_path_inq(&cpi, periph->path); + periph->deferred_callback(NULL, periph->deferred_ac, + periph->path, &cpi); break; + } default: - arg = NULL; + periph->deferred_callback(NULL, periph->deferred_ac, + periph->path, NULL); break; } - periph->deferred_callback(NULL, periph->deferred_ac, - periph->path, arg); } xpt_free_path(periph->path); free(periph, M_CAMPERIPH); @@ -1682,10 +1683,7 @@ camperiphscsisenseerror(union ccb *ccb, union ccb **orig, /* * Grab the inquiry data for this device. */ - memset(&cgd, 0, sizeof(cgd)); - xpt_setup_ccb(&cgd.ccb_h, ccb->ccb_h.path, CAM_PRIORITY_NORMAL); - cgd.ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)&cgd); + xpt_gdev_type(&cgd, ccb->ccb_h.path); err_action = scsi_error_action(&ccb->csio, &cgd.inq_data, sense_flags); @@ -2133,11 +2131,7 @@ cam_periph_devctl_notify(union ccb *ccb) sbuf_cat(&sb, "serial=\""); if ((cgd = (struct ccb_getdev *)xpt_alloc_ccb_nowait()) != NULL) { - xpt_setup_ccb(&cgd->ccb_h, ccb->ccb_h.path, - CAM_PRIORITY_NORMAL); - cgd->ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)cgd); - + xpt_gdev_type(cgd, ccb->ccb_h.path); if (cgd->ccb_h.status == CAM_REQ_CMP) sbuf_bcat(&sb, cgd->serial_num, cgd->serial_num_len); xpt_free_ccb((union ccb *)cgd); diff --git a/sys/cam/cam_xpt.c b/sys/cam/cam_xpt.c index 38bc82c69aad..cae29226d13c 100644 --- a/sys/cam/cam_xpt.c +++ b/sys/cam/cam_xpt.c @@ -2471,15 +2471,12 @@ xptsetasyncfunc(struct cam_ed *device, void *arg) if ((device->flags & CAM_DEV_UNCONFIGURED) != 0) return (1); - memset(&cgd, 0, sizeof(cgd)); xpt_compile_path(&path, NULL, device->target->bus->path_id, device->target->target_id, device->lun_id); - xpt_setup_ccb(&cgd.ccb_h, &path, CAM_PRIORITY_NORMAL); - cgd.ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)&cgd); + xpt_gdev_type(&cgd, &path); csa->callback(csa->callback_arg, AC_FOUND_DEVICE, &path, &cgd); @@ -2518,6 +2515,15 @@ xpt_action(union ccb *start_ccb) ("xpt_action: func %#x %s\n", start_ccb->ccb_h.func_code, xpt_action_name(start_ccb->ccb_h.func_code))); + /* + * Either it isn't queued, or it has a real priority. There still too + * many places that reuse CCBs with a real priority to do immediate + * queries to do the other side of this assert. + */ + KASSERT((start_ccb->ccb_h.func_code & XPT_FC_QUEUED) == 0 || + start_ccb->ccb_h.pinfo.priority != CAM_PRIORITY_NONE, + ("%s: queued ccb and CAM_PRIORITY_NONE illegal.", __func__)); + start_ccb->ccb_h.status = CAM_REQ_INPROG; (*(start_ccb->ccb_h.path->bus->xport->ops->action))(start_ccb); } diff --git a/sys/cam/cam_xpt.h b/sys/cam/cam_xpt.h index 57225a0ae92d..efa6c823245a 100644 --- a/sys/cam/cam_xpt.h +++ b/sys/cam/cam_xpt.h @@ -145,19 +145,31 @@ uint32_t xpt_poll_setup(union ccb *start_ccb); void xpt_sim_poll(struct cam_sim *sim); /* - * Perform a path inquiry at the request priority. bzero may be redundant for - * allocated CCBs, but for the on-stack CCBs it's required. + * Perform a path inquiry. bzero may be redundant for allocated CCBs, but for + * the on-stack CCBs it's required. */ static inline void xpt_path_inq(struct ccb_pathinq *cpi, struct cam_path *path) { - bzero(cpi, sizeof(*cpi)); xpt_setup_ccb(&cpi->ccb_h, path, CAM_PRIORITY_NONE); cpi->ccb_h.func_code = XPT_PATH_INQ; xpt_action((union ccb *)cpi); } +/* + * Perform get device type. bzero may be redundant for allocated CCBs, but for + * the on-stack CCBs it's required. + */ +static inline void +xpt_gdev_type(struct ccb_getdev *cgd, struct cam_path *path) +{ + bzero(cgd, sizeof(*cgd)); + xpt_setup_ccb(&cgd->ccb_h, path, CAM_PRIORITY_NONE); + cgd->ccb_h.func_code = XPT_GDEV_TYPE; + xpt_action((union ccb *)cgd); +} + #endif /* _KERNEL */ #endif /* _CAM_CAM_XPT_H */ diff --git a/sys/cam/mmc/mmc_da.c b/sys/cam/mmc/mmc_da.c index 3815561898bd..322141a72707 100644 --- a/sys/cam/mmc/mmc_da.c +++ b/sys/cam/mmc/mmc_da.c @@ -692,10 +692,7 @@ sddaasync(void *callback_arg, uint32_t code, case AC_GETDEV_CHANGED: { CAM_DEBUG(path, CAM_DEBUG_TRACE, ("=> AC_GETDEV_CHANGED\n")); - memset(&cgd, 0, sizeof(cgd)); - xpt_setup_ccb(&cgd.ccb_h, periph->path, CAM_PRIORITY_NORMAL); - cgd.ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)&cgd); + xpt_gdev_type(&cgd, periph->path); cam_periph_async(periph, code, path, arg); break; } @@ -789,7 +786,8 @@ sddaregister(struct cam_periph *periph, void *arg) static int mmc_exec_app_cmd(struct cam_periph *periph, union ccb *ccb, - struct mmc_command *cmd) { + struct mmc_command *cmd) +{ int err; /* Send APP_CMD first */ @@ -843,7 +841,8 @@ mmc_exec_app_cmd(struct cam_periph *periph, union ccb *ccb, } static int -mmc_app_get_scr(struct cam_periph *periph, union ccb *ccb, uint32_t *rawscr) { +mmc_app_get_scr(struct cam_periph *periph, union ccb *ccb, uint32_t *rawscr) +{ int err; struct mmc_command cmd; struct mmc_data d; @@ -869,7 +868,8 @@ mmc_app_get_scr(struct cam_periph *periph, union ccb *ccb, uint32_t *rawscr) { static int mmc_send_ext_csd(struct cam_periph *periph, union ccb *ccb, - uint8_t *rawextcsd, size_t buf_len) { + uint8_t *rawextcsd, size_t buf_len) +{ int err; struct mmc_data d; @@ -966,14 +966,16 @@ mmc_switch(struct cam_periph *periph, union ccb *ccb, } static uint32_t -mmc_get_spec_vers(struct cam_periph *periph) { +mmc_get_spec_vers(struct cam_periph *periph) +{ struct sdda_softc *softc = (struct sdda_softc *)periph->softc; return (softc->csd.spec_vers); } static uint64_t -mmc_get_media_size(struct cam_periph *periph) { +mmc_get_media_size(struct cam_periph *periph) +{ struct sdda_softc *softc = (struct sdda_softc *)periph->softc; return (softc->mediasize); @@ -992,7 +994,8 @@ mmc_get_cmd6_timeout(struct cam_periph *periph) static int mmc_sd_switch(struct cam_periph *periph, union ccb *ccb, uint8_t mode, uint8_t grp, uint8_t value, - uint8_t *res) { + uint8_t *res) +{ struct mmc_data mmc_d; uint32_t arg; int err; @@ -1069,7 +1072,8 @@ mmc_set_timing(struct cam_periph *periph, } static void -sdda_start_init_task(void *context, int pending) { +sdda_start_init_task(void *context, int pending) +{ union ccb *new_ccb; struct cam_periph *periph; @@ -1077,7 +1081,7 @@ sdda_start_init_task(void *context, int pending) { CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("sdda_start_init_task\n")); new_ccb = xpt_alloc_ccb(); xpt_setup_ccb(&new_ccb->ccb_h, periph->path, - CAM_PRIORITY_NONE); + CAM_PRIORITY_NORMAL); cam_periph_lock(periph); cam_periph_hold(periph, PRIBIO|PCATCH); @@ -1088,7 +1092,8 @@ sdda_start_init_task(void *context, int pending) { } static void -sdda_set_bus_width(struct cam_periph *periph, union ccb *ccb, int width) { +sdda_set_bus_width(struct cam_periph *periph, union ccb *ccb, int width) +{ struct sdda_softc *softc = (struct sdda_softc *)periph->softc; struct mmc_params *mmcp = &periph->path->device->mmc_ident_data; int err; diff --git a/sys/cam/mmc/mmc_xpt.c b/sys/cam/mmc/mmc_xpt.c index 4fce03004994..f5f66f5214a8 100644 --- a/sys/cam/mmc/mmc_xpt.c +++ b/sys/cam/mmc/mmc_xpt.c @@ -610,7 +610,6 @@ mmcprobe_start(struct cam_periph *periph, union ccb *start_ccb) CAM_DEBUG(start_ccb->ccb_h.path, CAM_DEBUG_PROBE, ("Start with PROBE_RESET\n")); /* FALLTHROUGH */ case PROBE_IDENTIFY: - xpt_path_inq(&start_ccb->cpi, periph->path); CAM_DEBUG(start_ccb->ccb_h.path, CAM_DEBUG_PROBE, ("Start with PROBE_IDENTIFY\n")); init_standard_ccb(start_ccb, XPT_MMC_GET_TRAN_SETTINGS); break; diff --git a/sys/cam/scsi/scsi_all.c b/sys/cam/scsi/scsi_all.c index 13a376ebb6e3..b518f84454ad 100644 --- a/sys/cam/scsi/scsi_all.c +++ b/sys/cam/scsi/scsi_all.c @@ -3708,11 +3708,7 @@ scsi_command_string(struct cam_device *device, struct ccb_scsiio *csio, /* * Get the device information. */ - xpt_setup_ccb(&cgd->ccb_h, - csio->ccb_h.path, - CAM_PRIORITY_NORMAL); - cgd->ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)cgd); + xpt_gdev_type(cgd, csio->ccb_h.path); /* * If the device is unconfigured, just pretend that it is a hard @@ -5144,11 +5140,7 @@ scsi_sense_sbuf(struct cam_device *device, struct ccb_scsiio *csio, /* * Get the device information. */ - xpt_setup_ccb(&cgd->ccb_h, - csio->ccb_h.path, - CAM_PRIORITY_NORMAL); - cgd->ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)cgd); + xpt_gdev_type(cgd, csio->ccb_h.path); /* * If the device is unconfigured, just pretend that it is a hard diff --git a/sys/cam/scsi/scsi_cd.c b/sys/cam/scsi/scsi_cd.c index 00a417f65052..e622a96ec77e 100644 --- a/sys/cam/scsi/scsi_cd.c +++ b/sys/cam/scsi/scsi_cd.c @@ -1240,13 +1240,7 @@ cddone(struct cam_periph *periph, union ccb *done_ccb) /*getcount_only*/0); status = done_ccb->ccb_h.status; - - bzero(&cgd, sizeof(cgd)); - xpt_setup_ccb(&cgd.ccb_h, - done_ccb->ccb_h.path, - CAM_PRIORITY_NORMAL); - cgd.ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)&cgd); + xpt_gdev_type(&cgd, done_ccb->ccb_h.path); if (scsi_extract_sense_ccb(done_ccb, &error_code, &sense_key, &asc, &ascq)) diff --git a/sys/cam/scsi/scsi_ch.c b/sys/cam/scsi/scsi_ch.c index 89a817c1b488..3da22ba61392 100644 --- a/sys/cam/scsi/scsi_ch.c +++ b/sys/cam/scsi/scsi_ch.c @@ -1705,11 +1705,7 @@ chscsiversion(struct cam_periph *periph) /* * Get the device information. */ - xpt_setup_ccb(&cgd->ccb_h, - periph->path, - CAM_PRIORITY_NORMAL); - cgd->ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)cgd); + xpt_gdev_type(cgd, periph->path); if (cgd->ccb_h.status != CAM_REQ_CMP) { xpt_free_ccb((union ccb *)cgd); diff --git a/sys/cam/scsi/scsi_da.c b/sys/cam/scsi/scsi_da.c index 0a2389cd9b5d..d02750aaacaf 100644 --- a/sys/cam/scsi/scsi_da.c +++ b/sys/cam/scsi/scsi_da.c @@ -5035,11 +5035,7 @@ dadone_proberc(struct cam_periph *periph, union ccb *done_ccb) /*timeout*/0, /*getcount_only*/0); - memset(&cgd, 0, sizeof(cgd)); - xpt_setup_ccb(&cgd.ccb_h, done_ccb->ccb_h.path, - CAM_PRIORITY_NORMAL); - cgd.ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)&cgd); + xpt_gdev_type(&cgd, done_ccb->ccb_h.path); if (scsi_extract_sense_ccb(done_ccb, &error_code, &sense_key, &asc, &ascq)) @@ -5077,6 +5073,18 @@ dadone_proberc(struct cam_periph *periph, union ccb *done_ccb) * behind a SATL translation that's fallen into a * terminally fatal state. * + * 4/2 happens on some HGST drives that are quite + * ill. We've already sent the start unit command (for + * which we ignore a 44/0 asc/ascq, which I'm hesitant + * to change since it's so basic and there's other error + * conditions to the START UNIT we should ignore). So to + * require initialization at this point when it should + * be fine implies to me, at least, that we should + * invalidate. Since we do read capacity in geom tasting + * a lot, and since this timeout is long, this leads to + * up to a 10 minute delay in booting. + * + * 4/2: LOGICAL UNIT NOT READY, INITIALIZING COMMAND REQUIRED * 25/0: LOGICAL UNIT NOT SUPPORTED * 44/0: INTERNAL TARGET FAILURE * 44/1: PERSISTENT RESERVATION INFORMATION LOST @@ -5084,6 +5092,7 @@ dadone_proberc(struct cam_periph *periph, union ccb *done_ccb) */ if ((have_sense) && (asc != 0x25) && (asc != 0x44) + && (asc != 0x04 && ascq != 0x02) && (error_code == SSD_CURRENT_ERROR || error_code == SSD_DESC_CURRENT_ERROR)) { const char *sense_key_desc; diff --git a/sys/cam/scsi/scsi_enc_ses.c b/sys/cam/scsi/scsi_enc_ses.c index c429e820a1fd..435874a9874a 100644 --- a/sys/cam/scsi/scsi_enc_ses.c +++ b/sys/cam/scsi/scsi_enc_ses.c @@ -979,10 +979,7 @@ ses_paths_iter(enc_softc_t *enc, enc_element_t *elm, != CAM_REQ_CMP) return; - memset(&cgd, 0, sizeof(cgd)); - xpt_setup_ccb(&cgd.ccb_h, path, CAM_PRIORITY_NORMAL); - cgd.ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)&cgd); + xpt_gdev_type(&cgd, path); if (cam_ccb_success((union ccb *)&cgd)) callback(enc, elm, path, callback_arg); diff --git a/sys/cam/scsi/scsi_sa.c b/sys/cam/scsi/scsi_sa.c index cfd48c98f30e..88147393192f 100644 --- a/sys/cam/scsi/scsi_sa.c +++ b/sys/cam/scsi/scsi_sa.c @@ -4731,12 +4731,7 @@ saextget(struct cdev *dev, struct cam_periph *periph, struct sbuf *sb, SASBADDVARSTR(sb, indent, periph->periph_name, %s, periph_name, strlen(periph->periph_name) + 1); SASBADDUINT(sb, indent, periph->unit_number, %u, unit_number); - memset(&cgd, 0, sizeof(cgd)); - xpt_setup_ccb(&cgd.ccb_h, - periph->path, - CAM_PRIORITY_NORMAL); - cgd.ccb_h.func_code = XPT_GDEV_TYPE; - xpt_action((union ccb *)&cgd); + xpt_gdev_type(&cgd, periph->path); if ((cgd.ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) { g->status = MT_EXT_GET_ERROR; snprintf(g->error_str, sizeof(g->error_str), diff --git a/sys/cddl/boot/zfs/zfsimpl.h b/sys/cddl/boot/zfs/zfsimpl.h index 0ce38384abbf..83d964360343 100644 --- a/sys/cddl/boot/zfs/zfsimpl.h +++ b/sys/cddl/boot/zfs/zfsimpl.h @@ -2019,6 +2019,7 @@ typedef struct vdev { vdev_list_t v_children; /* children of this vdev */ const char *v_name; /* vdev name */ uint64_t v_guid; /* vdev guid */ + uint64_t v_txg; /* most recent transaction */ uint64_t v_id; /* index in parent */ uint64_t v_psize; /* physical device capacity */ int v_ashift; /* offset to block shift */ @@ -2048,7 +2049,6 @@ typedef struct spa { STAILQ_ENTRY(spa) spa_link; /* link in global pool list */ char *spa_name; /* pool name */ uint64_t spa_guid; /* pool guid */ - uint64_t spa_txg; /* most recent transaction */ struct uberblock *spa_uberblock; /* best uberblock so far */ vdev_t *spa_root_vdev; /* toplevel vdev container */ objset_phys_t *spa_mos; /* MOS for this pool */ diff --git a/sys/cddl/dev/sdt/sdt.c b/sys/cddl/dev/sdt/sdt.c index a8da618204af..0a9059104671 100644 --- a/sys/cddl/dev/sdt/sdt.c +++ b/sys/cddl/dev/sdt/sdt.c @@ -72,6 +72,7 @@ static void sdt_load(void); static int sdt_unload(void); static void sdt_create_provider(struct sdt_provider *); static void sdt_create_probe(struct sdt_probe *); +static void sdt_init_probe(struct sdt_probe *, linker_file_t); static void sdt_kld_load(void *, struct linker_file *); static void sdt_kld_unload_try(void *, struct linker_file *, int *); @@ -204,6 +205,14 @@ sdt_create_probe(struct sdt_probe *probe) (void)dtrace_probe_create(prov->id, mod, func, name, aframes, probe); } +static void +sdt_init_probe(struct sdt_probe *probe, linker_file_t lf) +{ + probe->sdtp_lf = lf; + TAILQ_INIT(&probe->argtype_list); + STAILQ_INIT(&probe->tracepoint_list); +} + /* * Probes are created through the SDT module load/unload hook, so this function * has nothing to do. It only exists because the DTrace provider framework @@ -361,12 +370,19 @@ static void sdt_kld_load_providers(struct linker_file *lf) { struct sdt_provider **prov, **begin, **end; + struct sdt_probe **p_begin, **p_end; if (linker_file_lookup_set(lf, "sdt_providers_set", &begin, &end, NULL) == 0) { for (prov = begin; prov < end; prov++) sdt_create_provider(*prov); } + + if (linker_file_lookup_set(lf, "sdt_probes_set", &p_begin, &p_end, + NULL) == 0) { + for (struct sdt_probe **probe = p_begin; probe < p_end; probe++) + sdt_init_probe(*probe, lf); + } } static void @@ -378,13 +394,8 @@ sdt_kld_load_probes(struct linker_file *lf) if (linker_file_lookup_set(lf, "sdt_probes_set", &p_begin, &p_end, NULL) == 0) { - for (struct sdt_probe **probe = p_begin; probe < p_end; - probe++) { - (*probe)->sdtp_lf = lf; + for (struct sdt_probe **probe = p_begin; probe < p_end; probe++) sdt_create_probe(*probe); - TAILQ_INIT(&(*probe)->argtype_list); - STAILQ_INIT(&(*probe)->tracepoint_list); - } } if (linker_file_lookup_set(lf, "sdt_argtypes_set", &a_begin, &a_end, diff --git a/sys/compat/linprocfs/linprocfs.c b/sys/compat/linprocfs/linprocfs.c index cfb054235489..1c6d64d6b8bc 100644 --- a/sys/compat/linprocfs/linprocfs.c +++ b/sys/compat/linprocfs/linprocfs.c @@ -1911,7 +1911,7 @@ linprocfs_doproclimits(PFS_FILL_ARGS) "kern.sigqueue.max_pending_per_proc", &res, &size, 0, 0, 0, 0); if (error != 0) - goto out; + continue; rl.rlim_cur = res; rl.rlim_max = res; break; @@ -1919,7 +1919,7 @@ linprocfs_doproclimits(PFS_FILL_ARGS) error = kernel_sysctlbyname(td, "kern.ipc.msgmnb", &res, &size, 0, 0, 0, 0); if (error != 0) - goto out; + continue; rl.rlim_cur = res; rl.rlim_max = res; break; @@ -1941,9 +1941,9 @@ linprocfs_doproclimits(PFS_FILL_ARGS) li->desc, (unsigned long long)rl.rlim_cur, (unsigned long long)rl.rlim_max, li->unit); } -out: + lim_free(limp); - return (error); + return (0); } /* diff --git a/sys/compat/linux/linux_file.c b/sys/compat/linux/linux_file.c index 86834a7ecea8..a4be5313aa96 100644 --- a/sys/compat/linux/linux_file.c +++ b/sys/compat/linux/linux_file.c @@ -1792,7 +1792,7 @@ linux_memfd_create(struct thread *td, struct linux_memfd_create_args *args) if ((flags & MFD_ALLOW_SEALING) != 0) shmflags |= SHM_ALLOW_SEALING; return (kern_shm_open2(td, SHM_ANON, oflags, 0, shmflags, NULL, - memfd_name)); + memfd_name, NULL)); } int diff --git a/sys/compat/linuxkpi/common/include/linux/slab.h b/sys/compat/linuxkpi/common/include/linux/slab.h index f3a840d9bf4b..efa5c8cb67b3 100644 --- a/sys/compat/linuxkpi/common/include/linux/slab.h +++ b/sys/compat/linuxkpi/common/include/linux/slab.h @@ -45,7 +45,7 @@ MALLOC_DECLARE(M_KMALLOC); -#define kvzalloc(size, flags) kmalloc(size, (flags) | __GFP_ZERO) +#define kvzalloc(size, flags) kvmalloc(size, (flags) | __GFP_ZERO) #define kvcalloc(n, size, flags) kvmalloc_array(n, size, (flags) | __GFP_ZERO) #define kzalloc(size, flags) kmalloc(size, (flags) | __GFP_ZERO) #define kzalloc_node(size, flags, node) kmalloc_node(size, (flags) | __GFP_ZERO, node) diff --git a/sys/compat/linuxkpi/common/src/linux_page.c b/sys/compat/linuxkpi/common/src/linux_page.c index ebb92eacbf9a..628af17df853 100644 --- a/sys/compat/linuxkpi/common/src/linux_page.c +++ b/sys/compat/linuxkpi/common/src/linux_page.c @@ -106,6 +106,7 @@ linux_alloc_pages(gfp_t flags, unsigned int order) if ((flags & M_ZERO) != 0) req |= VM_ALLOC_ZERO; + if (order == 0 && (flags & GFP_DMA32) == 0) { page = vm_page_alloc_noobj(req); if (page == NULL) @@ -113,6 +114,10 @@ linux_alloc_pages(gfp_t flags, unsigned int order) } else { vm_paddr_t pmax = (flags & GFP_DMA32) ? BUS_SPACE_MAXADDR_32BIT : BUS_SPACE_MAXADDR; + + if ((flags & __GFP_NORETRY) != 0) + req |= VM_ALLOC_NORECLAIM; + retry: page = vm_page_alloc_noobj_contig(req, npages, 0, pmax, PAGE_SIZE, 0, VM_MEMATTR_DEFAULT); diff --git a/sys/conf/files b/sys/conf/files index 74d251c2b608..b7c19fae0b8e 100644 --- a/sys/conf/files +++ b/sys/conf/files @@ -3227,6 +3227,19 @@ dev/uart/uart_if.m optional uart dev/uart/uart_subr.c optional uart dev/uart/uart_tty.c optional uart # +# Universal Flash Storage Host Controller Interface drivers +# +dev/ufshci/ufshci.c optional ufshci +dev/ufshci/ufshci_ctrlr.c optional ufshci +dev/ufshci/ufshci_ctrlr_cmd.c optional ufshci +dev/ufshci/ufshci_dev.c optional ufshci +dev/ufshci/ufshci_pci.c optional ufshci +dev/ufshci/ufshci_req_queue.c optional ufshci +dev/ufshci/ufshci_req_sdb.c optional ufshci +dev/ufshci/ufshci_sim.c optional ufshci +dev/ufshci/ufshci_sysctl.c optional ufshci +dev/ufshci/ufshci_uic_cmd.c optional ufshci +# # USB controller drivers # dev/usb/controller/musb_otg.c optional musb @@ -3755,6 +3768,7 @@ gnu/gcov/gcov_subr.c optional gcov kern/bus_if.m standard kern/clock_if.m standard +kern/coredump_vnode.c standard kern/cpufreq_if.m standard kern/device_if.m standard kern/imgact_binmisc.c optional imgact_binmisc @@ -3843,6 +3857,7 @@ kern/kern_time.c standard kern/kern_timeout.c standard kern/kern_tslog.c optional tslog kern/kern_ubsan.c optional kubsan +kern/kern_ucoredump.c standard kern/kern_umtx.c standard kern/kern_uuid.c standard kern/kern_vnodedumper.c standard diff --git a/sys/conf/files.amd64 b/sys/conf/files.amd64 index 0584fc29d039..80548320c3fc 100644 --- a/sys/conf/files.amd64 +++ b/sys/conf/files.amd64 @@ -84,8 +84,8 @@ amd64/amd64/xen-locore.S optional xenhvm \ amd64/amd64/machdep.c standard amd64/amd64/mem.c optional mem amd64/amd64/minidump_machdep.c standard -amd64/amd64/mp_machdep.c optional smp -amd64/amd64/mpboot.S optional smp +amd64/amd64/mp_machdep.c standard +amd64/amd64/mpboot.S standard amd64/amd64/pmap.c standard amd64/amd64/ptrace_machdep.c standard amd64/amd64/support.S standard @@ -191,6 +191,10 @@ dev/ice/irdma_di_if.m optional ice pci \ compile-with "${NORMAL_M} -I$S/dev/ice" dev/ice/ice_ddp_common.c optional ice pci \ compile-with "${NORMAL_C} -I$S/dev/ice" +dev/ice/ice_iov.c optional ice pci pci_iov \ + compile-with "${NORMAL_C} -I$S/dev/ice" +dev/ice/ice_vf_mbx.c optional ice pci pci_iov \ + compile-with "${NORMAL_C} -I$S/dev/ice" ice_ddp.c optional ice_ddp \ compile-with "${AWK} -f $S/tools/fw_stub.awk ice_ddp.fw:ice_ddp:0x01032900 -mice_ddp -c${.TARGET}" \ no-ctfconvert no-implicit-rule before-depend local \ diff --git a/sys/conf/files.arm64 b/sys/conf/files.arm64 index 901da27e63f2..641001efab5e 100644 --- a/sys/conf/files.arm64 +++ b/sys/conf/files.arm64 @@ -368,6 +368,10 @@ dev/ice/irdma_di_if.m optional ice pci \ compile-with "${NORMAL_M} -I$S/dev/ice" dev/ice/ice_ddp_common.c optional ice pci \ compile-with "${NORMAL_C} -I$S/dev/ice" +dev/ice/ice_iov.c optional ice pci pci_iov \ + compile-with "${NORMAL_C} -I$S/dev/ice" +dev/ice/ice_vf_mbx.c optional ice pci pci_iov \ + compile-with "${NORMAL_C} -I$S/dev/ice" ice_ddp.c optional ice_ddp \ compile-with "${AWK} -f $S/tools/fw_stub.awk ice_ddp.fw:ice_ddp:0x01032900 -mice_ddp -c${.TARGET}" \ no-ctfconvert no-implicit-rule before-depend local \ diff --git a/sys/conf/files.x86 b/sys/conf/files.x86 index df206b314b38..9976e9cfec5d 100644 --- a/sys/conf/files.x86 +++ b/sys/conf/files.x86 @@ -62,6 +62,7 @@ dev/acpi_support/acpi_wmi_if.m standard dev/agp/agp_amd64.c optional agp dev/agp/agp_i810.c optional agp dev/agp/agp_via.c optional agp +dev/amdsmu/amdsmu.c optional amdsmu pci dev/amdsbwd/amdsbwd.c optional amdsbwd dev/amdsmn/amdsmn.c optional amdsmn | amdtemp dev/amdtemp/amdtemp.c optional amdtemp diff --git a/sys/dev/amdsmu/amdsmu.c b/sys/dev/amdsmu/amdsmu.c new file mode 100644 index 000000000000..416f875c6176 --- /dev/null +++ b/sys/dev/amdsmu/amdsmu.c @@ -0,0 +1,466 @@ +/* + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2025 The FreeBSD Foundation + * + * This software was developed by Aymeric Wibo <obiwac@freebsd.org> + * under sponsorship from the FreeBSD Foundation. + */ + +#include <sys/param.h> +#include <sys/bus.h> +#include <sys/kernel.h> +#include <sys/module.h> +#include <sys/rman.h> +#include <sys/sysctl.h> + +#include <dev/pci/pcivar.h> +#include <dev/amdsmu/amdsmu.h> + +static bool +amdsmu_match(device_t dev, const struct amdsmu_product **product_out) +{ + const uint16_t vendorid = pci_get_vendor(dev); + const uint16_t deviceid = pci_get_device(dev); + + for (size_t i = 0; i < nitems(amdsmu_products); i++) { + const struct amdsmu_product *prod = &amdsmu_products[i]; + + if (vendorid == prod->amdsmu_vendorid && + deviceid == prod->amdsmu_deviceid) { + if (product_out != NULL) + *product_out = prod; + return (true); + } + } + return (false); +} + +static void +amdsmu_identify(driver_t *driver, device_t parent) +{ + if (device_find_child(parent, "amdsmu", -1) != NULL) + return; + + if (amdsmu_match(parent, NULL)) { + if (device_add_child(parent, "amdsmu", -1) == NULL) + device_printf(parent, "add amdsmu child failed\n"); + } +} + +static int +amdsmu_probe(device_t dev) +{ + if (resource_disabled("amdsmu", 0)) + return (ENXIO); + if (!amdsmu_match(device_get_parent(dev), NULL)) + return (ENXIO); + device_set_descf(dev, "AMD System Management Unit"); + + return (BUS_PROBE_GENERIC); +} + +static enum amdsmu_res +amdsmu_wait_res(device_t dev) +{ + struct amdsmu_softc *sc = device_get_softc(dev); + enum amdsmu_res res; + + /* + * The SMU has a response ready for us when the response register is + * set. Otherwise, we must wait. + */ + for (size_t i = 0; i < SMU_RES_READ_MAX; i++) { + res = amdsmu_read4(sc, SMU_REG_RESPONSE); + if (res != SMU_RES_WAIT) + return (res); + pause_sbt("amdsmu", ustosbt(SMU_RES_READ_PERIOD_US), 0, + C_HARDCLOCK); + } + device_printf(dev, "timed out waiting for response from SMU\n"); + return (SMU_RES_WAIT); +} + +static int +amdsmu_cmd(device_t dev, enum amdsmu_msg msg, uint32_t arg, uint32_t *ret) +{ + struct amdsmu_softc *sc = device_get_softc(dev); + enum amdsmu_res res; + + /* Wait for SMU to be ready. */ + if (amdsmu_wait_res(dev) == SMU_RES_WAIT) + return (ETIMEDOUT); + + /* Clear previous response. */ + amdsmu_write4(sc, SMU_REG_RESPONSE, SMU_RES_WAIT); + + /* Write out command to registers. */ + amdsmu_write4(sc, SMU_REG_MESSAGE, msg); + amdsmu_write4(sc, SMU_REG_ARGUMENT, arg); + + /* Wait for SMU response and handle it. */ + res = amdsmu_wait_res(dev); + + switch (res) { + case SMU_RES_WAIT: + return (ETIMEDOUT); + case SMU_RES_OK: + if (ret != NULL) + *ret = amdsmu_read4(sc, SMU_REG_ARGUMENT); + return (0); + case SMU_RES_REJECT_BUSY: + device_printf(dev, "SMU is busy\n"); + return (EBUSY); + case SMU_RES_REJECT_PREREQ: + case SMU_RES_UNKNOWN: + case SMU_RES_FAILED: + device_printf(dev, "SMU error: %02x\n", res); + return (EIO); + } + + return (EINVAL); +} + +static int +amdsmu_get_vers(device_t dev) +{ + int err; + uint32_t smu_vers; + struct amdsmu_softc *sc = device_get_softc(dev); + + err = amdsmu_cmd(dev, SMU_MSG_GETSMUVERSION, 0, &smu_vers); + if (err != 0) { + device_printf(dev, "failed to get SMU version\n"); + return (err); + } + sc->smu_program = (smu_vers >> 24) & 0xFF; + sc->smu_maj = (smu_vers >> 16) & 0xFF; + sc->smu_min = (smu_vers >> 8) & 0xFF; + sc->smu_rev = smu_vers & 0xFF; + device_printf(dev, "SMU version: %d.%d.%d (program %d)\n", + sc->smu_maj, sc->smu_min, sc->smu_rev, sc->smu_program); + + return (0); +} + +static int +amdsmu_get_ip_blocks(device_t dev) +{ + struct amdsmu_softc *sc = device_get_softc(dev); + const uint16_t deviceid = pci_get_device(dev); + int err; + struct amdsmu_metrics *m = &sc->metrics; + bool active; + char sysctl_descr[32]; + + /* Get IP block count. */ + switch (deviceid) { + case PCI_DEVICEID_AMD_REMBRANDT_ROOT: + sc->ip_block_count = 12; + break; + case PCI_DEVICEID_AMD_PHOENIX_ROOT: + sc->ip_block_count = 21; + break; + /* TODO How many IP blocks does Strix Point (and the others) have? */ + case PCI_DEVICEID_AMD_STRIX_POINT_ROOT: + default: + sc->ip_block_count = nitems(amdsmu_ip_blocks_names); + } + KASSERT(sc->ip_block_count <= nitems(amdsmu_ip_blocks_names), + ("too many IP blocks for array")); + + /* Get and print out IP blocks. */ + err = amdsmu_cmd(dev, SMU_MSG_GET_SUP_CONSTRAINTS, 0, + &sc->active_ip_blocks); + if (err != 0) { + device_printf(dev, "failed to get IP blocks\n"); + return (err); + } + device_printf(dev, "Active IP blocks: "); + for (size_t i = 0; i < sc->ip_block_count; i++) { + active = (sc->active_ip_blocks & (1 << i)) != 0; + sc->ip_blocks_active[i] = active; + if (!active) + continue; + printf("%s%s", amdsmu_ip_blocks_names[i], + i + 1 < sc->ip_block_count ? " " : "\n"); + } + + /* Create a sysctl node for IP blocks. */ + sc->ip_blocks_sysctlnode = SYSCTL_ADD_NODE(sc->sysctlctx, + SYSCTL_CHILDREN(sc->sysctlnode), OID_AUTO, "ip_blocks", + CTLFLAG_RD, NULL, "SMU metrics"); + if (sc->ip_blocks_sysctlnode == NULL) { + device_printf(dev, "could not add sysctl node for IP blocks\n"); + return (ENOMEM); + } + + /* Create a sysctl node for each IP block. */ + for (size_t i = 0; i < sc->ip_block_count; i++) { + /* Create the sysctl node itself for the IP block. */ + snprintf(sysctl_descr, sizeof sysctl_descr, + "Metrics about the %s AMD IP block", + amdsmu_ip_blocks_names[i]); + sc->ip_block_sysctlnodes[i] = SYSCTL_ADD_NODE(sc->sysctlctx, + SYSCTL_CHILDREN(sc->ip_blocks_sysctlnode), OID_AUTO, + amdsmu_ip_blocks_names[i], CTLFLAG_RD, NULL, sysctl_descr); + if (sc->ip_block_sysctlnodes[i] == NULL) { + device_printf(dev, + "could not add sysctl node for \"%s\"\n", sysctl_descr); + continue; + } + /* + * Create sysctls for if the IP block is currently active, last + * active time, and total active time. + */ + SYSCTL_ADD_BOOL(sc->sysctlctx, + SYSCTL_CHILDREN(sc->ip_block_sysctlnodes[i]), OID_AUTO, + "active", CTLFLAG_RD, &sc->ip_blocks_active[i], 0, + "IP block is currently active"); + SYSCTL_ADD_U64(sc->sysctlctx, + SYSCTL_CHILDREN(sc->ip_block_sysctlnodes[i]), OID_AUTO, + "last_time", CTLFLAG_RD, &m->ip_block_last_active_time[i], + 0, "How long the IP block was active for during the last" + " sleep (us)"); +#ifdef IP_BLOCK_TOTAL_ACTIVE_TIME + SYSCTL_ADD_U64(sc->sysctlctx, + SYSCTL_CHILDREN(sc->ip_block_sysctlnodes[i]), OID_AUTO, + "total_time", CTLFLAG_RD, &m->ip_block_total_active_time[i], + 0, "How long the IP block was active for during sleep in" + " total (us)"); +#endif + } + return (0); +} + +static int +amdsmu_init_metrics(device_t dev) +{ + struct amdsmu_softc *sc = device_get_softc(dev); + int err; + uint32_t metrics_addr_lo, metrics_addr_hi; + uint64_t metrics_addr; + + /* Get physical address of logging buffer. */ + err = amdsmu_cmd(dev, SMU_MSG_LOG_GETDRAM_ADDR_LO, 0, &metrics_addr_lo); + if (err != 0) + return (err); + err = amdsmu_cmd(dev, SMU_MSG_LOG_GETDRAM_ADDR_HI, 0, &metrics_addr_hi); + if (err != 0) + return (err); + metrics_addr = ((uint64_t) metrics_addr_hi << 32) | metrics_addr_lo; + + /* Map memory of logging buffer. */ + err = bus_space_map(sc->bus_tag, metrics_addr, + sizeof(struct amdsmu_metrics), 0, &sc->metrics_space); + if (err != 0) { + device_printf(dev, "could not map bus space for SMU metrics\n"); + return (err); + } + + /* Start logging for metrics. */ + amdsmu_cmd(dev, SMU_MSG_LOG_RESET, 0, NULL); + amdsmu_cmd(dev, SMU_MSG_LOG_START, 0, NULL); + return (0); +} + +static int +amdsmu_dump_metrics(device_t dev) +{ + struct amdsmu_softc *sc = device_get_softc(dev); + int err; + + err = amdsmu_cmd(dev, SMU_MSG_LOG_DUMP_DATA, 0, NULL); + if (err != 0) { + device_printf(dev, "failed to dump metrics\n"); + return (err); + } + bus_space_read_region_4(sc->bus_tag, sc->metrics_space, 0, + (uint32_t *)&sc->metrics, sizeof(sc->metrics) / sizeof(uint32_t)); + + return (0); +} + +static void +amdsmu_fetch_idlemask(device_t dev) +{ + struct amdsmu_softc *sc = device_get_softc(dev); + + sc->idlemask = amdsmu_read4(sc, SMU_REG_IDLEMASK); +} + +static int +amdsmu_attach(device_t dev) +{ + struct amdsmu_softc *sc = device_get_softc(dev); + int err; + uint32_t physbase_addr_lo, physbase_addr_hi; + uint64_t physbase_addr; + int rid = 0; + struct sysctl_oid *node; + + /* + * Find physical base address for SMU. + * XXX I am a little confused about the masks here. I'm just copying + * what Linux does in the amd-pmc driver to get the base address. + */ + pci_write_config(dev, SMU_INDEX_ADDRESS, SMU_PHYSBASE_ADDR_LO, 4); + physbase_addr_lo = pci_read_config(dev, SMU_INDEX_DATA, 4) & 0xFFF00000; + + pci_write_config(dev, SMU_INDEX_ADDRESS, SMU_PHYSBASE_ADDR_HI, 4); + physbase_addr_hi = pci_read_config(dev, SMU_INDEX_DATA, 4) & 0x0000FFFF; + + physbase_addr = (uint64_t)physbase_addr_hi << 32 | physbase_addr_lo; + + /* Map memory for SMU and its registers. */ + sc->res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &rid, RF_ACTIVE); + if (sc->res == NULL) { + device_printf(dev, "could not allocate resource\n"); + return (ENXIO); + } + + sc->bus_tag = rman_get_bustag(sc->res); + + if (bus_space_map(sc->bus_tag, physbase_addr, + SMU_MEM_SIZE, 0, &sc->smu_space) != 0) { + device_printf(dev, "could not map bus space for SMU\n"); + err = ENXIO; + goto err_smu_space; + } + if (bus_space_map(sc->bus_tag, physbase_addr + SMU_REG_SPACE_OFF, + SMU_MEM_SIZE, 0, &sc->reg_space) != 0) { + device_printf(dev, "could not map bus space for SMU regs\n"); + err = ENXIO; + goto err_reg_space; + } + + /* sysctl stuff. */ + sc->sysctlctx = device_get_sysctl_ctx(dev); + sc->sysctlnode = device_get_sysctl_tree(dev); + + /* Get version & add sysctls. */ + if ((err = amdsmu_get_vers(dev)) != 0) + goto err_dump; + + SYSCTL_ADD_U8(sc->sysctlctx, SYSCTL_CHILDREN(sc->sysctlnode), OID_AUTO, + "program", CTLFLAG_RD, &sc->smu_program, 0, "SMU program number"); + SYSCTL_ADD_U8(sc->sysctlctx, SYSCTL_CHILDREN(sc->sysctlnode), OID_AUTO, + "version_major", CTLFLAG_RD, &sc->smu_maj, 0, + "SMU firmware major version number"); + SYSCTL_ADD_U8(sc->sysctlctx, SYSCTL_CHILDREN(sc->sysctlnode), OID_AUTO, + "version_minor", CTLFLAG_RD, &sc->smu_min, 0, + "SMU firmware minor version number"); + SYSCTL_ADD_U8(sc->sysctlctx, SYSCTL_CHILDREN(sc->sysctlnode), OID_AUTO, + "version_revision", CTLFLAG_RD, &sc->smu_rev, 0, + "SMU firmware revision number"); + + /* Set up for getting metrics & add sysctls. */ + if ((err = amdsmu_init_metrics(dev)) != 0) + goto err_dump; + if ((err = amdsmu_dump_metrics(dev)) != 0) + goto err_dump; + + node = SYSCTL_ADD_NODE(sc->sysctlctx, SYSCTL_CHILDREN(sc->sysctlnode), + OID_AUTO, "metrics", CTLFLAG_RD, NULL, "SMU metrics"); + if (node == NULL) { + device_printf(dev, "could not add sysctl node for metrics\n"); + err = ENOMEM; + goto err_dump; + } + + SYSCTL_ADD_U32(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "table_version", CTLFLAG_RD, &sc->metrics.table_version, 0, + "SMU metrics table version"); + SYSCTL_ADD_U32(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "hint_count", CTLFLAG_RD, &sc->metrics.hint_count, 0, + "How many times the sleep hint was set"); + SYSCTL_ADD_U32(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "s0i3_last_entry_status", CTLFLAG_RD, + &sc->metrics.s0i3_last_entry_status, 0, + "1 if last S0i3 entry was successful"); + SYSCTL_ADD_U32(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "time_last_in_s0i2", CTLFLAG_RD, &sc->metrics.time_last_in_s0i2, 0, + "Time spent in S0i2 during last sleep (us)"); + SYSCTL_ADD_U64(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "time_last_entering_s0i3", CTLFLAG_RD, + &sc->metrics.time_last_entering_s0i3, 0, + "Time spent entering S0i3 during last sleep (us)"); + SYSCTL_ADD_U64(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "total_time_entering_s0i3", CTLFLAG_RD, + &sc->metrics.total_time_entering_s0i3, 0, + "Total time spent entering S0i3 (us)"); + SYSCTL_ADD_U64(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "time_last_resuming", CTLFLAG_RD, &sc->metrics.time_last_resuming, + 0, "Time spent resuming from last sleep (us)"); + SYSCTL_ADD_U64(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "total_time_resuming", CTLFLAG_RD, &sc->metrics.total_time_resuming, + 0, "Total time spent resuming from sleep (us)"); + SYSCTL_ADD_U64(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "time_last_in_s0i3", CTLFLAG_RD, &sc->metrics.time_last_in_s0i3, 0, + "Time spent in S0i3 during last sleep (us)"); + SYSCTL_ADD_U64(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "total_time_in_s0i3", CTLFLAG_RD, &sc->metrics.total_time_in_s0i3, + 0, "Total time spent in S0i3 (us)"); + SYSCTL_ADD_U64(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "time_last_in_sw_drips", CTLFLAG_RD, + &sc->metrics.time_last_in_sw_drips, 0, + "Time spent in awake during last sleep (us)"); + SYSCTL_ADD_U64(sc->sysctlctx, SYSCTL_CHILDREN(node), OID_AUTO, + "total_time_in_sw_drips", CTLFLAG_RD, + &sc->metrics.total_time_in_sw_drips, 0, + "Total time spent awake (us)"); + + /* Get IP blocks & add sysctls. */ + err = amdsmu_get_ip_blocks(dev); + if (err != 0) + goto err_dump; + + /* Get idlemask & add sysctl. */ + amdsmu_fetch_idlemask(dev); + SYSCTL_ADD_U32(sc->sysctlctx, SYSCTL_CHILDREN(sc->sysctlnode), OID_AUTO, + "idlemask", CTLFLAG_RD, &sc->idlemask, 0, "SMU idlemask. This " + "value is not documented - only used to help AMD internally debug " + "issues"); + + return (0); +err_dump: + bus_space_unmap(sc->bus_tag, sc->reg_space, SMU_MEM_SIZE); +err_reg_space: + bus_space_unmap(sc->bus_tag, sc->smu_space, SMU_MEM_SIZE); +err_smu_space: + bus_release_resource(dev, SYS_RES_MEMORY, rid, sc->res); + return (err); +} + +static int +amdsmu_detach(device_t dev) +{ + struct amdsmu_softc *sc = device_get_softc(dev); + int rid = 0; + + bus_space_unmap(sc->bus_tag, sc->smu_space, SMU_MEM_SIZE); + bus_space_unmap(sc->bus_tag, sc->reg_space, SMU_MEM_SIZE); + + bus_release_resource(dev, SYS_RES_MEMORY, rid, sc->res); + return (0); +} + +static device_method_t amdsmu_methods[] = { + DEVMETHOD(device_identify, amdsmu_identify), + DEVMETHOD(device_probe, amdsmu_probe), + DEVMETHOD(device_attach, amdsmu_attach), + DEVMETHOD(device_detach, amdsmu_detach), + DEVMETHOD_END +}; + +static driver_t amdsmu_driver = { + "amdsmu", + amdsmu_methods, + sizeof(struct amdsmu_softc), +}; + +DRIVER_MODULE(amdsmu, hostb, amdsmu_driver, NULL, NULL); +MODULE_VERSION(amdsmu, 1); +MODULE_DEPEND(amdsmu, amdsmn, 1, 1, 1); +MODULE_PNP_INFO("U16:vendor;U16:device", pci, amdsmu, amdsmu_products, + nitems(amdsmu_products)); diff --git a/sys/dev/amdsmu/amdsmu.h b/sys/dev/amdsmu/amdsmu.h new file mode 100644 index 000000000000..025887f7fe5a --- /dev/null +++ b/sys/dev/amdsmu/amdsmu.h @@ -0,0 +1,95 @@ +/* + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2025 The FreeBSD Foundation + * + * This software was developed by Aymeric Wibo <obiwac@freebsd.org> + * under sponsorship from the FreeBSD Foundation. + */ +#ifndef _AMDSMU_H_ +#define _AMDSMU_H_ + +#include <sys/param.h> +#include <sys/bus.h> +#include <sys/kernel.h> +#include <machine/bus.h> +#include <x86/cputypes.h> + +#include <dev/amdsmu/amdsmu_reg.h> + +#define SMU_RES_READ_PERIOD_US 50 +#define SMU_RES_READ_MAX 20000 + +static const struct amdsmu_product { + uint16_t amdsmu_vendorid; + uint16_t amdsmu_deviceid; +} amdsmu_products[] = { + { CPU_VENDOR_AMD, PCI_DEVICEID_AMD_REMBRANDT_ROOT }, + { CPU_VENDOR_AMD, PCI_DEVICEID_AMD_PHOENIX_ROOT }, + { CPU_VENDOR_AMD, PCI_DEVICEID_AMD_STRIX_POINT_ROOT }, +}; + +static const char *const amdsmu_ip_blocks_names[] = { + "DISPLAY", + "CPU", + "GFX", + "VDD", + "ACP", + "VCN", + "ISP", + "NBIO", + "DF", + "USB3_0", + "USB3_1", + "LAPIC", + "USB3_2", + "USB3_3", + "USB3_4", + "USB4_0", + "USB4_1", + "MPM", + "JPEG", + "IPU", + "UMSCH", + "VPE", +}; + +CTASSERT(nitems(amdsmu_ip_blocks_names) <= 32); + +struct amdsmu_softc { + struct sysctl_ctx_list *sysctlctx; + struct sysctl_oid *sysctlnode; + + struct resource *res; + bus_space_tag_t bus_tag; + + bus_space_handle_t smu_space; + bus_space_handle_t reg_space; + + uint8_t smu_program; + uint8_t smu_maj, smu_min, smu_rev; + + uint32_t active_ip_blocks; + struct sysctl_oid *ip_blocks_sysctlnode; + size_t ip_block_count; + struct sysctl_oid *ip_block_sysctlnodes[nitems(amdsmu_ip_blocks_names)]; + bool ip_blocks_active[nitems(amdsmu_ip_blocks_names)]; + + bus_space_handle_t metrics_space; + struct amdsmu_metrics metrics; + uint32_t idlemask; +}; + +static inline uint32_t +amdsmu_read4(const struct amdsmu_softc *sc, bus_size_t reg) +{ + return (bus_space_read_4(sc->bus_tag, sc->reg_space, reg)); +} + +static inline void +amdsmu_write4(const struct amdsmu_softc *sc, bus_size_t reg, uint32_t val) +{ + bus_space_write_4(sc->bus_tag, sc->reg_space, reg, val); +} + +#endif /* _AMDSMU_H_ */ diff --git a/sys/dev/amdsmu/amdsmu_reg.h b/sys/dev/amdsmu/amdsmu_reg.h new file mode 100644 index 000000000000..e685b34e6883 --- /dev/null +++ b/sys/dev/amdsmu/amdsmu_reg.h @@ -0,0 +1,84 @@ +/* + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2025 The FreeBSD Foundation + * + * This software was developed by Aymeric Wibo <obiwac@freebsd.org> + * under sponsorship from the FreeBSD Foundation. + */ +#ifndef _AMDSMU_REG_H_ +#define _AMDSMU_REG_H_ + +#include <sys/types.h> + +/* + * TODO These are in common with amdtemp; should we find a way to factor these + * out? Also, there are way more of these. I couldn't find a centralized place + * which lists them though. + */ +#define PCI_DEVICEID_AMD_REMBRANDT_ROOT 0x14B5 +#define PCI_DEVICEID_AMD_PHOENIX_ROOT 0x14E8 +#define PCI_DEVICEID_AMD_STRIX_POINT_ROOT 0x14A4 + +#define SMU_INDEX_ADDRESS 0xB8 +#define SMU_INDEX_DATA 0xBC + +#define SMU_PHYSBASE_ADDR_LO 0x13B102E8 +#define SMU_PHYSBASE_ADDR_HI 0x13B102EC + +#define SMU_MEM_SIZE 0x1000 +#define SMU_REG_SPACE_OFF 0x10000 + +#define SMU_REG_MESSAGE 0x538 +#define SMU_REG_RESPONSE 0x980 +#define SMU_REG_ARGUMENT 0x9BC +#define SMU_REG_IDLEMASK 0xD14 + +enum amdsmu_res { + SMU_RES_WAIT = 0x00, + SMU_RES_OK = 0x01, + SMU_RES_REJECT_BUSY = 0xFC, + SMU_RES_REJECT_PREREQ = 0xFD, + SMU_RES_UNKNOWN = 0xFE, + SMU_RES_FAILED = 0xFF, +}; + +enum amdsmu_msg { + SMU_MSG_GETSMUVERSION = 0x02, + SMU_MSG_LOG_GETDRAM_ADDR_HI = 0x04, + SMU_MSG_LOG_GETDRAM_ADDR_LO = 0x05, + SMU_MSG_LOG_START = 0x06, + SMU_MSG_LOG_RESET = 0x07, + SMU_MSG_LOG_DUMP_DATA = 0x08, + SMU_MSG_GET_SUP_CONSTRAINTS = 0x09, +}; + +/* XXX Copied from Linux struct smu_metrics. */ +struct amdsmu_metrics { + uint32_t table_version; + uint32_t hint_count; + uint32_t s0i3_last_entry_status; + uint32_t time_last_in_s0i2; + uint64_t time_last_entering_s0i3; + uint64_t total_time_entering_s0i3; + uint64_t time_last_resuming; + uint64_t total_time_resuming; + uint64_t time_last_in_s0i3; + uint64_t total_time_in_s0i3; + uint64_t time_last_in_sw_drips; + uint64_t total_time_in_sw_drips; + /* + * This is how long each IP block was active for (us), i.e., blocking + * entry to S0i3. In Linux, these are called "timecondition_notmet_*". + * + * XXX Total active time for IP blocks seems to be buggy and reporting + * garbage (at least on Phoenix), so it's disabled for now. The last + * active time for the USB4_0 IP block also seems to be buggy. + */ + uint64_t ip_block_last_active_time[32]; +#ifdef IP_BLOCK_TOTAL_ACTIVE_TIME + uint64_t ip_block_total_active_time[32]; +#endif +} __attribute__((packed)); + +#endif /* _AMDSMU_REG_H_ */ diff --git a/sys/dev/cxgbe/tom/t4_cpl_io.c b/sys/dev/cxgbe/tom/t4_cpl_io.c index 8547f21586e1..7a6b1cbdd736 100644 --- a/sys/dev/cxgbe/tom/t4_cpl_io.c +++ b/sys/dev/cxgbe/tom/t4_cpl_io.c @@ -703,7 +703,7 @@ t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop) for (m = sndptr; m != NULL; m = m->m_next) { int n; - if ((m->m_flags & M_NOTAVAIL) != 0) + if ((m->m_flags & M_NOTREADY) != 0) break; if (m->m_flags & M_EXTPG) { #ifdef KERN_TLS @@ -787,7 +787,7 @@ t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop) /* nothing to send */ if (plen == 0) { - KASSERT(m == NULL || (m->m_flags & M_NOTAVAIL) != 0, + KASSERT(m == NULL || (m->m_flags & M_NOTREADY) != 0, ("%s: nothing to send, but m != NULL is ready", __func__)); break; @@ -880,7 +880,7 @@ t4_push_frames(struct adapter *sc, struct toepcb *toep, int drop) toep->txsd_avail--; t4_l2t_send(sc, wr, toep->l2te); - } while (m != NULL && (m->m_flags & M_NOTAVAIL) == 0); + } while (m != NULL && (m->m_flags & M_NOTREADY) == 0); /* Send a FIN if requested, but only if there's no more data to send */ if (m == NULL && toep->flags & TPF_SEND_FIN) diff --git a/sys/dev/cxgbe/tom/t4_tls.c b/sys/dev/cxgbe/tom/t4_tls.c index c6377980fca9..27c16b9988ae 100644 --- a/sys/dev/cxgbe/tom/t4_tls.c +++ b/sys/dev/cxgbe/tom/t4_tls.c @@ -563,7 +563,7 @@ t4_push_ktls(struct adapter *sc, struct toepcb *toep, int drop) * If there is no ready data to send, wait until more * data arrives. */ - if (m == NULL || (m->m_flags & M_NOTAVAIL) != 0) { + if (m == NULL || (m->m_flags & M_NOTREADY) != 0) { if (sowwakeup) sowwakeup_locked(so); else @@ -614,7 +614,7 @@ t4_push_ktls(struct adapter *sc, struct toepcb *toep, int drop) /* Shove if there is no additional data pending. */ shove = ((m->m_next == NULL || - (m->m_next->m_flags & M_NOTAVAIL) != 0)) && + (m->m_next->m_flags & M_NOTREADY) != 0)) && (tp->t_flags & TF_MORETOCOME) == 0; if (sb->sb_flags & SB_AUTOSIZE && diff --git a/sys/dev/drm2/drm_fb_helper.c b/sys/dev/drm2/drm_fb_helper.c index f67cc9f60d02..1f4abd255690 100644 --- a/sys/dev/drm2/drm_fb_helper.c +++ b/sys/dev/drm2/drm_fb_helper.c @@ -51,7 +51,7 @@ struct vt_kms_softc { struct task fb_mode_task; }; -/* Call restore out of vt(9) locks. */ +/* Call restore out of vt(4) locks. */ static void vt_restore_fbdev_mode(void *arg, int pending) { diff --git a/sys/dev/efidev/efirt.c b/sys/dev/efidev/efirt.c index b0fa33daeca7..b55c1c191077 100644 --- a/sys/dev/efidev/efirt.c +++ b/sys/dev/efidev/efirt.c @@ -107,7 +107,8 @@ static int efi_status2err[25] = { enum efi_table_type { TYPE_ESRT = 0, - TYPE_PROP + TYPE_PROP, + TYPE_MEMORY_ATTR }; static int efi_enter(void); @@ -445,6 +446,42 @@ get_table_length(enum efi_table_type type, size_t *table_len, void **taddr) free(buf, M_TEMP); return (0); } + case TYPE_MEMORY_ATTR: + { + efi_guid_t guid = EFI_MEMORY_ATTRIBUTES_TABLE; + struct efi_memory_attribute_table *tbl_addr, *mem_addr; + int error; + void *buf; + size_t len = sizeof(struct efi_memory_attribute_table); + + error = efi_get_table(&guid, (void **)&tbl_addr); + if (error) + return (error); + + buf = malloc(len, M_TEMP, M_WAITOK); + error = physcopyout((vm_paddr_t)tbl_addr, buf, len); + if (error) { + free(buf, M_TEMP); + return (error); + } + + mem_addr = (struct efi_memory_attribute_table *)buf; + if (mem_addr->version != 2) { + free(buf, M_TEMP); + return (EINVAL); + } + len += mem_addr->descriptor_size * mem_addr->num_ents; + if (len > EFI_TABLE_ALLOC_MAX) { + free(buf, M_TEMP); + return (ENOMEM); + } + + *table_len = len; + if (taddr != NULL) + *taddr = tbl_addr; + free(buf, M_TEMP); + return (0); + } } return (ENOENT); } @@ -457,7 +494,8 @@ copy_table(efi_guid_t *guid, void **buf, size_t buf_len, size_t *table_len) enum efi_table_type type; } tables[] = { { EFI_TABLE_ESRT, TYPE_ESRT }, - { EFI_PROPERTIES_TABLE, TYPE_PROP } + { EFI_PROPERTIES_TABLE, TYPE_PROP }, + { EFI_MEMORY_ATTRIBUTES_TABLE, TYPE_MEMORY_ATTR } }; size_t table_idx; void *taddr; diff --git a/sys/dev/gpio/acpi_gpiobus.c b/sys/dev/gpio/acpi_gpiobus.c index f9468e0deda0..94f4e5771266 100644 --- a/sys/dev/gpio/acpi_gpiobus.c +++ b/sys/dev/gpio/acpi_gpiobus.c @@ -357,7 +357,7 @@ acpi_gpiobus_attach(device_t dev) status = AcpiWalkResources(handle, "_AEI", acpi_gpiobus_enumerate_aei, &ctx); - if (ACPI_FAILURE(status)) + if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) device_printf(dev, "Failed to enumerate AEI resources\n"); return (0); diff --git a/sys/dev/gpio/gpiobus.c b/sys/dev/gpio/gpiobus.c index ab7f13177969..764bcb7e6ee8 100644 --- a/sys/dev/gpio/gpiobus.c +++ b/sys/dev/gpio/gpiobus.c @@ -110,10 +110,9 @@ gpio_alloc_intr_resource(device_t consumer_dev, int *rid, u_int alloc_flags, res = bus_alloc_resource(consumer_dev, SYS_RES_IRQ, rid, irq, irq, 1, alloc_flags); if (res == NULL) { - intr_free_intr_map_data((struct intr_map_data *)gpio_data); + intr_unmap_irq(irq); return (NULL); } - rman_set_virtual(res, gpio_data); return (res); } #else @@ -866,6 +865,25 @@ gpiobus_alloc_resource(device_t bus, device_t child, int type, int *rid, end, count, flags)); } +static int +gpiobus_release_resource(device_t dev, device_t child, struct resource *r) +{ + int err; +#ifdef INTRNG + u_int irq; + + irq = rman_get_start(r); + MPASS(irq == rman_get_end(r)); +#endif + err = bus_generic_rman_release_resource(dev, child, r); + if (err != 0) + return (err); +#ifdef INTRNG + intr_unmap_irq(irq); +#endif + return (0); +} + static struct resource_list * gpiobus_get_resource_list(device_t bus __unused, device_t child) { @@ -1060,7 +1078,7 @@ static device_method_t gpiobus_methods[] = { DEVMETHOD(bus_get_resource, bus_generic_rl_get_resource), DEVMETHOD(bus_set_resource, bus_generic_rl_set_resource), DEVMETHOD(bus_alloc_resource, gpiobus_alloc_resource), - DEVMETHOD(bus_release_resource, bus_generic_rman_release_resource), + DEVMETHOD(bus_release_resource, gpiobus_release_resource), DEVMETHOD(bus_activate_resource, bus_generic_rman_activate_resource), DEVMETHOD(bus_deactivate_resource, bus_generic_rman_deactivate_resource), DEVMETHOD(bus_get_resource_list, gpiobus_get_resource_list), diff --git a/sys/dev/hyperv/vmbus/vmbus_chan.c b/sys/dev/hyperv/vmbus/vmbus_chan.c index 189a3e66a039..7ea60a499c72 100644 --- a/sys/dev/hyperv/vmbus/vmbus_chan.c +++ b/sys/dev/hyperv/vmbus/vmbus_chan.c @@ -1555,7 +1555,7 @@ vmbus_event_flags_proc(struct vmbus_softc *sc, volatile u_long *event_flags, continue; flags = atomic_swap_long(&event_flags[f], 0); - chid_base = f << VMBUS_EVTFLAG_SHIFT; + chid_base = f * VMBUS_EVTFLAG_LEN; while ((chid_ofs = ffsl(flags)) != 0) { struct vmbus_channel *chan; @@ -1599,7 +1599,7 @@ vmbus_event_proc_compat(struct vmbus_softc *sc, int cpu) eventf = VMBUS_PCPU_GET(sc, event_flags, cpu) + VMBUS_SINT_MESSAGE; if (atomic_testandclear_long(&eventf->evt_flags[0], 0)) { vmbus_event_flags_proc(sc, sc->vmbus_rx_evtflags, - VMBUS_CHAN_MAX_COMPAT >> VMBUS_EVTFLAG_SHIFT); + VMBUS_CHAN_MAX_COMPAT / VMBUS_EVTFLAG_LEN); } } @@ -1903,7 +1903,7 @@ vmbus_chan_msgproc_choffer(struct vmbus_softc *sc, * Setup event flag. */ chan->ch_evtflag = - &sc->vmbus_tx_evtflags[chan->ch_id >> VMBUS_EVTFLAG_SHIFT]; + &sc->vmbus_tx_evtflags[chan->ch_id / VMBUS_EVTFLAG_LEN]; chan->ch_evtflag_mask = 1UL << (chan->ch_id & VMBUS_EVTFLAG_MASK); /* diff --git a/sys/dev/hyperv/vmbus/vmbus_reg.h b/sys/dev/hyperv/vmbus/vmbus_reg.h index 4aa729475b5d..76cdca0ebeb2 100644 --- a/sys/dev/hyperv/vmbus/vmbus_reg.h +++ b/sys/dev/hyperv/vmbus/vmbus_reg.h @@ -60,16 +60,10 @@ CTASSERT(sizeof(struct vmbus_message) == VMBUS_MSG_SIZE); * Hyper-V SynIC event flags */ -#ifdef __LP64__ -#define VMBUS_EVTFLAGS_MAX 32 -#define VMBUS_EVTFLAG_SHIFT 6 -#else -#define VMBUS_EVTFLAGS_MAX 64 -#define VMBUS_EVTFLAG_SHIFT 5 -#endif -#define VMBUS_EVTFLAG_LEN (1 << VMBUS_EVTFLAG_SHIFT) +#define VMBUS_EVTFLAG_LEN (sizeof(u_long) * 8) #define VMBUS_EVTFLAG_MASK (VMBUS_EVTFLAG_LEN - 1) #define VMBUS_EVTFLAGS_SIZE 256 +#define VMBUS_EVTFLAGS_MAX (VMBUS_EVTFLAGS_SIZE / sizeof(u_long)) struct vmbus_evtflags { u_long evt_flags[VMBUS_EVTFLAGS_MAX]; diff --git a/sys/dev/ice/ice_features.h b/sys/dev/ice/ice_features.h index 821abe4806ca..5b23757b1c98 100644 --- a/sys/dev/ice/ice_features.h +++ b/sys/dev/ice/ice_features.h @@ -91,7 +91,9 @@ enum feat_list { static inline void ice_disable_unsupported_features(ice_bitmap_t __unused *bitmap) { +#ifndef PCI_IOV ice_clear_bit(ICE_FEATURE_SRIOV, bitmap); +#endif #ifndef DEV_NETMAP ice_clear_bit(ICE_FEATURE_NETMAP, bitmap); #endif diff --git a/sys/dev/ice/ice_iflib.h b/sys/dev/ice/ice_iflib.h index 3a5dc201189a..e1d5307a9516 100644 --- a/sys/dev/ice/ice_iflib.h +++ b/sys/dev/ice/ice_iflib.h @@ -139,6 +139,9 @@ struct ice_irq_vector { * @tc: traffic class queue belongs to * @q_handle: qidx in tc; used in TXQ enable functions * + * ice_iov.c requires the following parameters (when PCI_IOV is defined): + * @itr_idx: ITR index to use for this queue + * * Other parameters may be iflib driver specific */ struct ice_tx_queue { @@ -153,6 +156,9 @@ struct ice_tx_queue { u32 me; u16 q_handle; u8 tc; +#ifdef PCI_IOV + u8 itr_idx; +#endif /* descriptor writeback status */ qidx_t *tx_rsq; @@ -175,6 +181,9 @@ struct ice_tx_queue { * @stats: queue statistics * @tc: traffic class queue belongs to * + * ice_iov.c requires the following parameters (when PCI_IOV is defined): + * @itr_idx: ITR index to use for this queue + * * Other parameters may be iflib driver specific */ struct ice_rx_queue { @@ -187,6 +196,9 @@ struct ice_rx_queue { struct ice_irq_vector *irqv; u32 me; u8 tc; +#ifdef PCI_IOV + u8 itr_idx; +#endif struct if_irq que_irq; }; @@ -332,6 +344,10 @@ struct ice_softc { ice_declare_bitmap(feat_cap, ICE_FEATURE_COUNT); ice_declare_bitmap(feat_en, ICE_FEATURE_COUNT); +#ifdef PCI_IOV + struct ice_vf *vfs; + u16 num_vfs; +#endif struct ice_resmgr os_imgr; /* For mirror interface */ struct ice_mirr_if *mirr_if; diff --git a/sys/dev/ice/ice_iov.c b/sys/dev/ice/ice_iov.c new file mode 100644 index 000000000000..c5a3e1060e44 --- /dev/null +++ b/sys/dev/ice/ice_iov.c @@ -0,0 +1,1856 @@ +/* SPDX-License-Identifier: BSD-3-Clause */ +/* Copyright (c) 2025, Intel Corporation + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * 3. Neither the name of the Intel Corporation nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +/** + * @file ice_iov.c + * @brief Virtualization support functions + * + * Contains functions for enabling and managing PCIe virtual function devices, + * including enabling new VFs, and managing VFs over the virtchnl interface. + */ + +#include "ice_iov.h" + +static struct ice_vf *ice_iov_get_vf(struct ice_softc *sc, int vf_num); +static void ice_iov_ready_vf(struct ice_softc *sc, struct ice_vf *vf); +static void ice_reset_vf(struct ice_softc *sc, struct ice_vf *vf, + bool trigger_vflr); +static void ice_iov_setup_intr_mapping(struct ice_softc *sc, struct ice_vf *vf); + +static void ice_vc_version_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_get_vf_res_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_add_eth_addr_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_del_eth_addr_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static bool ice_vc_isvalid_ring_len(u16 ring_len); +static void ice_vc_cfg_vsi_qs_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_cfg_rss_key_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_set_rss_hena_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_enable_queues_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_notify_vf_link_state(struct ice_softc *sc, struct ice_vf *vf); +static void ice_vc_disable_queues_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_cfg_irq_map_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_get_stats_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_eth_stats_to_virtchnl_eth_stats(struct ice_eth_stats *istats, + struct virtchnl_eth_stats *vstats); +static void ice_vc_cfg_rss_lut_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_cfg_promisc_mode_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_add_vlan_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static void ice_vc_del_vlan_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf); +static enum virtchnl_status_code ice_iov_err_to_virt_err(int ice_err); +static int ice_vf_validate_mac(struct ice_vf *vf, const uint8_t *addr); + +/** + * ice_iov_attach - Initialize SR-IOV PF host support + * @sc: device softc structure + * + * Initialize SR-IOV PF host support at the end of the driver attach process. + * + * @pre Must be called from sleepable context (calls malloc() w/ M_WAITOK) + * + * @returns 0 if successful, or + * - ENOMEM if there is no memory for the PF/VF schemas or iov device + * - ENXIO if the device isn't PCI-E or doesn't support the same SR-IOV + * version as the kernel + * - ENOENT if the device doesn't have the SR-IOV capability + */ +int +ice_iov_attach(struct ice_softc *sc) +{ + device_t dev = sc->dev; + nvlist_t *pf_schema, *vf_schema; + int error; + + pf_schema = pci_iov_schema_alloc_node(); + vf_schema = pci_iov_schema_alloc_node(); + + pci_iov_schema_add_unicast_mac(vf_schema, "mac-addr", 0, NULL); + pci_iov_schema_add_bool(vf_schema, "mac-anti-spoof", + IOV_SCHEMA_HASDEFAULT, TRUE); + pci_iov_schema_add_bool(vf_schema, "allow-set-mac", + IOV_SCHEMA_HASDEFAULT, FALSE); + pci_iov_schema_add_bool(vf_schema, "allow-promisc", + IOV_SCHEMA_HASDEFAULT, FALSE); + pci_iov_schema_add_uint16(vf_schema, "num-queues", + IOV_SCHEMA_HASDEFAULT, ICE_DEFAULT_VF_QUEUES); + pci_iov_schema_add_uint16(vf_schema, "mirror-src-vsi", + IOV_SCHEMA_HASDEFAULT, ICE_INVALID_MIRROR_VSI); + pci_iov_schema_add_uint16(vf_schema, "max-vlan-allowed", + IOV_SCHEMA_HASDEFAULT, ICE_DEFAULT_VF_VLAN_LIMIT); + pci_iov_schema_add_uint16(vf_schema, "max-mac-filters", + IOV_SCHEMA_HASDEFAULT, ICE_DEFAULT_VF_FILTER_LIMIT); + + error = pci_iov_attach(dev, pf_schema, vf_schema); + if (error != 0) { + device_printf(dev, + "pci_iov_attach failed (error=%s)\n", + ice_err_str(error)); + ice_clear_bit(ICE_FEATURE_SRIOV, sc->feat_en); + } else + ice_set_bit(ICE_FEATURE_SRIOV, sc->feat_en); + + return (error); +} + +/** + * ice_iov_detach - Teardown SR-IOV PF host support + * @sc: device softc structure + * + * Teardown SR-IOV PF host support at the start of the driver detach process. + * + * @returns 0 if successful or IOV support hasn't been setup, or + * - EBUSY if VFs still exist + */ +int +ice_iov_detach(struct ice_softc *sc) +{ + device_t dev = sc->dev; + int error; + + error = pci_iov_detach(dev); + if (error != 0) { + device_printf(dev, + "pci_iov_detach failed (error=%s)\n", + ice_err_str(error)); + } + + return (error); +} + +/** + * ice_iov_init - Called by the OS before the first VF is created. + * @sc: device softc structure + * @num_vfs: number of VFs to setup resources for + * @params: configuration parameters for the PF + * + * @returns 0 if successful or an error code on failure + */ +int +ice_iov_init(struct ice_softc *sc, uint16_t num_vfs, const nvlist_t *params __unused) +{ + /* Allocate array of VFs, for tracking */ + sc->vfs = (struct ice_vf *)malloc(sizeof(struct ice_vf) * num_vfs, M_ICE, M_NOWAIT | + M_ZERO); + if (sc->vfs == NULL) + return (ENOMEM); + + /* Initialize each VF with basic information */ + for (int i = 0; i < num_vfs; i++) + sc->vfs[i].vf_num = i; + + /* Save off number of configured VFs */ + sc->num_vfs = num_vfs; + + return (0); +} + +/** + * ice_iov_get_vf - Get pointer to VF at given index + * @sc: device softc structure + * @vf_num: Index of VF to retrieve + * + * @remark will throw an assertion if vf_num is not in the + * range of allocated VFs + * + * @returns a pointer to the VF structure at the given index + */ +static struct ice_vf * +ice_iov_get_vf(struct ice_softc *sc, int vf_num) +{ + MPASS(vf_num < sc->num_vfs); + + return &sc->vfs[vf_num]; +} + +/** + * ice_iov_add_vf - Called by the OS for each VF to create + * @sc: device softc structure + * @vfnum: index of VF to configure + * @params: configuration parameters for the VF + * + * @returns 0 if successful or an error code on failure + */ +int +ice_iov_add_vf(struct ice_softc *sc, uint16_t vfnum, const nvlist_t *params) +{ + struct ice_tx_queue *txq; + struct ice_rx_queue *rxq; + device_t dev = sc->dev; + struct ice_vsi *vsi; + struct ice_vf *vf; + int vf_num_queues; + const void *mac; + size_t size; + int error; + int i; + + vf = ice_iov_get_vf(sc, vfnum); + vf->vf_flags = VF_FLAG_ENABLED; + + /* This VF needs at least one VSI */ + vsi = ice_alloc_vsi(sc, ICE_VSI_VF); + if (vsi == NULL) + return (ENOMEM); + vf->vsi = vsi; + vsi->vf_num = vfnum; + + vf_num_queues = nvlist_get_number(params, "num-queues"); + /* Validate and clamp value if invalid */ + if (vf_num_queues < 1 || vf_num_queues > ICE_MAX_SCATTERED_QUEUES) + device_printf(dev, "Invalid num-queues (%d) for VF %d\n", + vf_num_queues, vf->vf_num); + if (vf_num_queues < 1) { + device_printf(dev, "Setting VF %d num-queues to 1\n", vf->vf_num); + vf_num_queues = 1; + } else if (vf_num_queues > ICE_MAX_SCATTERED_QUEUES) { + device_printf(dev, "Setting VF %d num-queues to %d\n", + vf->vf_num, ICE_MAX_SCATTERED_QUEUES); + vf_num_queues = ICE_MAX_SCATTERED_QUEUES; + } + vsi->qmap_type = ICE_RESMGR_ALLOC_SCATTERED; + + /* Reserve VF queue allocation from PF queues */ + ice_alloc_vsi_qmap(vsi, vf_num_queues, vf_num_queues); + vsi->num_tx_queues = vsi->num_rx_queues = vf_num_queues; + + /* Assign Tx queues from PF space */ + error = ice_resmgr_assign_scattered(&sc->tx_qmgr, vsi->tx_qmap, + vsi->num_tx_queues); + if (error) { + device_printf(sc->dev, "Unable to assign VF Tx queues: %s\n", + ice_err_str(error)); + goto release_vsi; + } + + /* Assign Rx queues from PF space */ + error = ice_resmgr_assign_scattered(&sc->rx_qmgr, vsi->rx_qmap, + vsi->num_rx_queues); + if (error) { + device_printf(sc->dev, "Unable to assign VF Rx queues: %s\n", + ice_err_str(error)); + goto release_vsi; + } + + vsi->max_frame_size = ICE_MAX_FRAME_SIZE; + + /* Allocate queue structure memory */ + vsi->tx_queues = (struct ice_tx_queue *) + malloc(sizeof(struct ice_tx_queue) * vsi->num_tx_queues, M_ICE, + M_NOWAIT | M_ZERO); + if (!vsi->tx_queues) { + device_printf(sc->dev, "VF-%d: Unable to allocate Tx queue memory\n", + vfnum); + error = ENOMEM; + goto release_vsi; + } + for (i = 0, txq = vsi->tx_queues; i < vsi->num_tx_queues; i++, txq++) { + txq->me = i; + txq->vsi = vsi; + } + + /* Allocate queue structure memory */ + vsi->rx_queues = (struct ice_rx_queue *) + malloc(sizeof(struct ice_rx_queue) * vsi->num_rx_queues, M_ICE, + M_NOWAIT | M_ZERO); + if (!vsi->rx_queues) { + device_printf(sc->dev, "VF-%d: Unable to allocate Rx queue memory\n", + vfnum); + error = ENOMEM; + goto free_txqs; + } + for (i = 0, rxq = vsi->rx_queues; i < vsi->num_rx_queues; i++, rxq++) { + rxq->me = i; + rxq->vsi = vsi; + } + + /* Allocate space to store the IRQ vector data */ + vf->num_irq_vectors = vf_num_queues + 1; + vf->tx_irqvs = (struct ice_irq_vector *) + malloc(sizeof(struct ice_irq_vector) * (vf->num_irq_vectors), + M_ICE, M_NOWAIT); + if (!vf->tx_irqvs) { + device_printf(sc->dev, + "Unable to allocate TX irqv memory for VF-%d's %d vectors\n", + vfnum, vf->num_irq_vectors); + error = ENOMEM; + goto free_rxqs; + } + vf->rx_irqvs = (struct ice_irq_vector *) + malloc(sizeof(struct ice_irq_vector) * (vf->num_irq_vectors), + M_ICE, M_NOWAIT); + if (!vf->rx_irqvs) { + device_printf(sc->dev, + "Unable to allocate RX irqv memory for VF-%d's %d vectors\n", + vfnum, vf->num_irq_vectors); + error = ENOMEM; + goto free_txirqvs; + } + + /* Assign VF interrupts from PF space */ + if (!(vf->vf_imap = + (u16 *)malloc(sizeof(u16) * vf->num_irq_vectors, + M_ICE, M_NOWAIT))) { + device_printf(dev, "Unable to allocate VF-%d imap memory\n", vfnum); + error = ENOMEM; + goto free_rxirqvs; + } + error = ice_resmgr_assign_contiguous(&sc->dev_imgr, vf->vf_imap, vf->num_irq_vectors); + if (error) { + device_printf(dev, "Unable to assign VF-%d interrupt mapping: %s\n", + vfnum, ice_err_str(error)); + goto free_imap; + } + + if (nvlist_exists_binary(params, "mac-addr")) { + mac = nvlist_get_binary(params, "mac-addr", &size); + memcpy(vf->mac, mac, ETHER_ADDR_LEN); + + if (nvlist_get_bool(params, "allow-set-mac")) + vf->vf_flags |= VF_FLAG_SET_MAC_CAP; + } else + /* + * If the administrator has not specified a MAC address then + * we must allow the VF to choose one. + */ + vf->vf_flags |= VF_FLAG_SET_MAC_CAP; + + if (nvlist_get_bool(params, "mac-anti-spoof")) + vf->vf_flags |= VF_FLAG_MAC_ANTI_SPOOF; + + if (nvlist_get_bool(params, "allow-promisc")) + vf->vf_flags |= VF_FLAG_PROMISC_CAP; + + vsi->mirror_src_vsi = nvlist_get_number(params, "mirror-src-vsi"); + + vf->vlan_limit = nvlist_get_number(params, "max-vlan-allowed"); + vf->mac_filter_limit = nvlist_get_number(params, "max-mac-filters"); + + vf->vf_flags |= VF_FLAG_VLAN_CAP; + + /* Create and setup VSI in HW */ + error = ice_initialize_vsi(vsi); + if (error) { + device_printf(sc->dev, "Unable to initialize VF %d VSI: %s\n", + vfnum, ice_err_str(error)); + goto release_imap; + } + + /* Add the broadcast address */ + error = ice_add_vsi_mac_filter(vsi, broadcastaddr); + if (error) { + device_printf(sc->dev, "Unable to add broadcast filter VF %d VSI: %s\n", + vfnum, ice_err_str(error)); + goto release_imap; + } + + ice_iov_ready_vf(sc, vf); + + return (0); + +release_imap: + ice_resmgr_release_map(&sc->dev_imgr, vf->vf_imap, + vf->num_irq_vectors); +free_imap: + free(vf->vf_imap, M_ICE); + vf->vf_imap = NULL; +free_rxirqvs: + free(vf->rx_irqvs, M_ICE); + vf->rx_irqvs = NULL; +free_txirqvs: + free(vf->tx_irqvs, M_ICE); + vf->tx_irqvs = NULL; +free_rxqs: + free(vsi->rx_queues, M_ICE); + vsi->rx_queues = NULL; +free_txqs: + free(vsi->tx_queues, M_ICE); + vsi->tx_queues = NULL; +release_vsi: + ice_release_vsi(vsi); + vf->vsi = NULL; + return (error); +} + +/** + * ice_iov_uninit - Called by the OS when VFs are destroyed + * @sc: device softc structure + */ +void +ice_iov_uninit(struct ice_softc *sc) +{ + struct ice_vf *vf; + struct ice_vsi *vsi; + + /* Release per-VF resources */ + for (int i = 0; i < sc->num_vfs; i++) { + vf = &sc->vfs[i]; + vsi = vf->vsi; + + /* Free VF interrupt reservation */ + if (vf->vf_imap) { + free(vf->vf_imap, M_ICE); + vf->vf_imap = NULL; + } + + /* Free queue interrupt mapping trackers */ + if (vf->tx_irqvs) { + free(vf->tx_irqvs, M_ICE); + vf->tx_irqvs = NULL; + } + if (vf->rx_irqvs) { + free(vf->rx_irqvs, M_ICE); + vf->rx_irqvs = NULL; + } + + if (!vsi) + continue; + + /* Free VSI queues */ + if (vsi->tx_queues) { + free(vsi->tx_queues, M_ICE); + vsi->tx_queues = NULL; + } + if (vsi->rx_queues) { + free(vsi->rx_queues, M_ICE); + vsi->rx_queues = NULL; + } + + ice_release_vsi(vsi); + vf->vsi = NULL; + } + + /* Release memory used for VF tracking */ + if (sc->vfs) { + free(sc->vfs, M_ICE); + sc->vfs = NULL; + } + sc->num_vfs = 0; +} + +/** + * ice_iov_handle_vflr - Process VFLR event + * @sc: device softc structure + * + * Identifys which VFs have been reset and re-configure + * them. + */ +void +ice_iov_handle_vflr(struct ice_softc *sc) +{ + struct ice_hw *hw = &sc->hw; + struct ice_vf *vf; + u32 reg, reg_idx, bit_idx; + + for (int i = 0; i < sc->num_vfs; i++) { + vf = &sc->vfs[i]; + + reg_idx = (hw->func_caps.vf_base_id + vf->vf_num) / 32; + bit_idx = (hw->func_caps.vf_base_id + vf->vf_num) % 32; + reg = rd32(hw, GLGEN_VFLRSTAT(reg_idx)); + if (reg & BIT(bit_idx)) + ice_reset_vf(sc, vf, false); + } +} + +/** + * ice_iov_ready_vf - Setup VF interrupts and mark it as ready + * @sc: device softc structure + * @vf: driver's VF structure for the VF to update + * + * Clears VF reset triggering bit, sets up the PF<->VF interrupt + * mapping and marks the VF as active in the HW so that the VF + * driver can use it. + */ +static void +ice_iov_ready_vf(struct ice_softc *sc, struct ice_vf *vf) +{ + struct ice_hw *hw = &sc->hw; + u32 reg; + + /* Clear the triggering bit */ + reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_num)); + reg &= ~VPGEN_VFRTRIG_VFSWR_M; + wr32(hw, VPGEN_VFRTRIG(vf->vf_num), reg); + + /* Setup VF interrupt allocation and mapping */ + ice_iov_setup_intr_mapping(sc, vf); + + /* Indicate to the VF that reset is done */ + wr32(hw, VFGEN_RSTAT(vf->vf_num), VIRTCHNL_VFR_VFACTIVE); + + ice_flush(hw); +} + +/** + * ice_reset_vf - Perform a hardware reset (VFR) on a VF + * @sc: device softc structure + * @vf: driver's VF structure for VF to be reset + * @trigger_vflr: trigger a reset or only handle already executed reset + * + * Performs a VFR for the given VF. This function busy waits until the + * reset completes in the HW, notifies the VF that the reset is done + * by setting a bit in a HW register, then returns. + * + * @remark This also sets up the PF<->VF interrupt mapping and allocations in + * the hardware after the hardware reset is finished, via + * ice_iov_setup_intr_mapping() + */ +static void +ice_reset_vf(struct ice_softc *sc, struct ice_vf *vf, bool trigger_vflr) +{ + u16 global_vf_num, reg_idx, bit_idx; + struct ice_hw *hw = &sc->hw; + int status; + u32 reg; + int i; + + global_vf_num = vf->vf_num + hw->func_caps.vf_base_id; + + if (trigger_vflr) { + reg = rd32(hw, VPGEN_VFRTRIG(vf->vf_num)); + reg |= VPGEN_VFRTRIG_VFSWR_M; + wr32(hw, VPGEN_VFRTRIG(vf->vf_num), reg); + } + + /* clear the VFLR bit for the VF in a GLGEN_VFLRSTAT register */ + reg_idx = (global_vf_num) / 32; + bit_idx = (global_vf_num) % 32; + wr32(hw, GLGEN_VFLRSTAT(reg_idx), BIT(bit_idx)); + ice_flush(hw); + + /* Wait until there are no pending PCI transactions */ + wr32(hw, PF_PCI_CIAA, + ICE_PCIE_DEV_STATUS | (global_vf_num << PF_PCI_CIAA_VF_NUM_S)); + + for (i = 0; i < ICE_PCI_CIAD_WAIT_COUNT; i++) { + reg = rd32(hw, PF_PCI_CIAD); + if (!(reg & PCIEM_STA_TRANSACTION_PND)) + break; + + DELAY(ICE_PCI_CIAD_WAIT_DELAY_US); + } + if (i == ICE_PCI_CIAD_WAIT_COUNT) + device_printf(sc->dev, + "VF-%d PCI transactions stuck\n", vf->vf_num); + + /* Disable TX queues, which is required during VF reset */ + status = ice_dis_vsi_txq(hw->port_info, vf->vsi->idx, 0, 0, NULL, NULL, + NULL, ICE_VF_RESET, vf->vf_num, NULL); + if (status) + device_printf(sc->dev, + "%s: Failed to disable LAN Tx queues: err %s aq_err %s\n", + __func__, ice_status_str(status), + ice_aq_str(hw->adminq.sq_last_status)); + + /* Then check for the VF reset to finish in HW */ + for (i = 0; i < ICE_VPGEN_VFRSTAT_WAIT_COUNT; i++) { + reg = rd32(hw, VPGEN_VFRSTAT(vf->vf_num)); + if ((reg & VPGEN_VFRSTAT_VFRD_M)) + break; + + DELAY(ICE_VPGEN_VFRSTAT_WAIT_DELAY_US); + } + if (i == ICE_VPGEN_VFRSTAT_WAIT_COUNT) + device_printf(sc->dev, + "VF-%d Reset is stuck\n", vf->vf_num); + + ice_iov_ready_vf(sc, vf); +} + +/** + * ice_vc_get_vf_res_msg - Handle VIRTCHNL_OP_GET_VF_RESOURCES msg from VF + * @sc: device private structure + * @vf: VF tracking structure + * @msg_buf: raw message buffer from the VF + * + * Receives a message from the VF listing its supported capabilities, and + * replies to the VF with information about what resources the PF has + * allocated for the VF. + * + * @remark This always replies to the VF with a success status; it does not + * fail. It's up to the VF driver to reject or complain about the PF's response. + */ +static void +ice_vc_get_vf_res_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct ice_hw *hw = &sc->hw; + struct virtchnl_vf_resource *vf_res; + struct virtchnl_vsi_resource *vsi_res; + u16 vf_res_len; + u32 vf_caps; + + /* XXX: Only support one VSI per VF, so this size doesn't need adjusting */ + vf_res_len = sizeof(struct virtchnl_vf_resource); + vf_res = (struct virtchnl_vf_resource *)malloc(vf_res_len, M_ICE, + M_WAITOK | M_ZERO); + + vf_res->num_vsis = 1; + vf_res->num_queue_pairs = vf->vsi->num_tx_queues; + vf_res->max_vectors = vf_res->num_queue_pairs + 1; + + vf_res->rss_key_size = ICE_GET_SET_RSS_KEY_EXTEND_KEY_SIZE; + vf_res->rss_lut_size = ICE_VSIQF_HLUT_ARRAY_SIZE; + vf_res->max_mtu = 0; + + vf_res->vf_cap_flags = VF_BASE_MODE_OFFLOADS; + if (msg_buf != NULL) { + vf_caps = *((u32 *)(msg_buf)); + + if (vf_caps & VIRTCHNL_VF_CAP_ADV_LINK_SPEED) + vf_res->vf_cap_flags |= VIRTCHNL_VF_CAP_ADV_LINK_SPEED; + + if (vf_caps & VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) + vf_res->vf_cap_flags |= VIRTCHNL_VF_OFFLOAD_WB_ON_ITR; + } + + vsi_res = &vf_res->vsi_res[0]; + vsi_res->vsi_id = vf->vsi->idx; + vsi_res->num_queue_pairs = vf->vsi->num_tx_queues; + vsi_res->vsi_type = VIRTCHNL_VSI_SRIOV; + vsi_res->qset_handle = 0; + if (!ETHER_IS_ZERO(vf->mac)) + memcpy(vsi_res->default_mac_addr, vf->mac, ETHER_ADDR_LEN); + + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_GET_VF_RESOURCES, + VIRTCHNL_STATUS_SUCCESS, (u8 *)vf_res, vf_res_len, NULL); + + free(vf_res, M_ICE); +} + +/** + * ice_vc_version_msg - Handle VIRTCHNL_OP_VERSION msg from VF + * @sc: device private structure + * @vf: VF tracking structure + * @msg_buf: raw message buffer from the VF + * + * Receives a version message from the VF, and responds to the VF with + * the version number that the PF will use. + * + * @remark This always replies to the VF with a success status; it does not + * fail. + */ +static void +ice_vc_version_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct virtchnl_version_info *recv_vf_version; + struct ice_hw *hw = &sc->hw; + device_t dev = sc->dev; + + recv_vf_version = (struct virtchnl_version_info *)msg_buf; + + /* VFs running the 1.0 API expect to get 1.0 back */ + if (VF_IS_V10(recv_vf_version)) { + vf->version.major = 1; + vf->version.minor = VIRTCHNL_VERSION_MINOR_NO_VF_CAPS; + } else { + vf->version.major = VIRTCHNL_VERSION_MAJOR; + vf->version.minor = VIRTCHNL_VERSION_MINOR; + + if ((recv_vf_version->major != VIRTCHNL_VERSION_MAJOR) || + (recv_vf_version->minor != VIRTCHNL_VERSION_MINOR)) + device_printf(dev, + "%s: VF-%d requested version (%d.%d) differs from PF version (%d.%d)\n", + __func__, vf->vf_num, + recv_vf_version->major, recv_vf_version->minor, + VIRTCHNL_VERSION_MAJOR, VIRTCHNL_VERSION_MINOR); + } + + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_VERSION, + VIRTCHNL_STATUS_SUCCESS, (u8 *)&vf->version, sizeof(vf->version), + NULL); +} + +/** + * ice_vf_validate_mac - Validate MAC address before adding it + * @vf: VF tracking structure + * @addr: MAC address to validate + * + * Validate a MAC address before adding it to a VF during the handling + * of a VIRTCHNL_OP_ADD_ETH_ADDR operation. Notably, this also checks if + * the VF is allowed to set its own arbitrary MAC addresses. + * + * Returns 0 if MAC address is valid for the given vf + */ +static int +ice_vf_validate_mac(struct ice_vf *vf, const uint8_t *addr) +{ + + if (ETHER_IS_ZERO(addr) || ETHER_IS_BROADCAST(addr)) + return (EINVAL); + + /* + * If the VF is not allowed to change its MAC address, don't let it + * set a MAC filter for an address that is not a multicast address and + * is not its assigned MAC. + */ + if (!(vf->vf_flags & VF_FLAG_SET_MAC_CAP) && + !(ETHER_IS_MULTICAST(addr) || !bcmp(addr, vf->mac, ETHER_ADDR_LEN))) + return (EPERM); + + return (0); +} + +/** + * ice_vc_add_eth_addr_msg - Handle VIRTCHNL_OP_ADD_ETH_ADDR msg from VF + * @sc: device private structure + * @vf: VF tracking structure + * @msg_buf: raw message buffer from the VF + * + * Receives a list of MAC addresses from the VF and adds those addresses + * to the VSI's filter list. + */ +static void +ice_vc_add_eth_addr_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct virtchnl_ether_addr_list *addr_list; + struct ice_hw *hw = &sc->hw; + u16 added_addr_cnt = 0; + int error = 0; + + addr_list = (struct virtchnl_ether_addr_list *)msg_buf; + + if (addr_list->num_elements > + (vf->mac_filter_limit - vf->mac_filter_cnt)) { + v_status = VIRTCHNL_STATUS_ERR_NO_MEMORY; + goto done; + } + + for (int i = 0; i < addr_list->num_elements; i++) { + u8 *addr = addr_list->list[i].addr; + + /* The type flag is currently ignored; every MAC address is + * treated as the LEGACY type + */ + + error = ice_vf_validate_mac(vf, addr); + if (error == EPERM) { + device_printf(sc->dev, + "%s: VF-%d: Not permitted to add MAC addr for VSI %d\n", + __func__, vf->vf_num, vf->vsi->idx); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + continue; + } else if (error) { + device_printf(sc->dev, + "%s: VF-%d: Did not add invalid MAC addr for VSI %d\n", + __func__, vf->vf_num, vf->vsi->idx); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + continue; + } + + error = ice_add_vsi_mac_filter(vf->vsi, addr); + if (error) { + device_printf(sc->dev, + "%s: VF-%d: Error adding MAC addr for VSI %d\n", + __func__, vf->vf_num, vf->vsi->idx); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + continue; + } + /* Don't count VF's MAC against its MAC filter limit */ + if (memcmp(addr, vf->mac, ETHER_ADDR_LEN)) + added_addr_cnt++; + } + + vf->mac_filter_cnt += added_addr_cnt; + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_ADD_ETH_ADDR, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_del_eth_addr_msg - Handle VIRTCHNL_OP_DEL_ETH_ADDR msg from VF + * @sc: device private structure + * @vf: VF tracking structure + * @msg_buf: raw message buffer from the VF + * + * Receives a list of MAC addresses from the VF and removes those addresses + * from the VSI's filter list. + */ +static void +ice_vc_del_eth_addr_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct virtchnl_ether_addr_list *addr_list; + struct ice_hw *hw = &sc->hw; + u16 deleted_addr_cnt = 0; + int error = 0; + + addr_list = (struct virtchnl_ether_addr_list *)msg_buf; + + for (int i = 0; i < addr_list->num_elements; i++) { + error = ice_remove_vsi_mac_filter(vf->vsi, addr_list->list[i].addr); + if (error) { + device_printf(sc->dev, + "%s: VF-%d: Error removing MAC addr for VSI %d\n", + __func__, vf->vf_num, vf->vsi->idx); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + continue; + } + /* Don't count VF's MAC against its MAC filter limit */ + if (memcmp(addr_list->list[i].addr, vf->mac, ETHER_ADDR_LEN)) + deleted_addr_cnt++; + } + + if (deleted_addr_cnt >= vf->mac_filter_cnt) + vf->mac_filter_cnt = 0; + else + vf->mac_filter_cnt -= deleted_addr_cnt; + + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_DEL_ETH_ADDR, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_add_vlan_msg - Handle VIRTCHNL_OP_ADD_VLAN msg from VF + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + * + * Adds the VLANs in msg_buf to the VF's VLAN filter list. + */ +static void +ice_vc_add_vlan_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct ice_hw *hw = &sc->hw; + struct virtchnl_vlan_filter_list *vlan_list; + int status = 0; + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct ice_vsi *vsi = vf->vsi; + + vlan_list = (struct virtchnl_vlan_filter_list *)msg_buf; + + if (vlan_list->vsi_id != vsi->idx) { + device_printf(sc->dev, + "VF-%d: Message has invalid VSI ID (expected %d, got %d)\n", + vf->vf_num, vsi->idx, vlan_list->vsi_id); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + if (vlan_list->num_elements > (vf->vlan_limit - vf->vlan_cnt)) { + v_status = VIRTCHNL_STATUS_ERR_NO_MEMORY; + goto done; + } + + status = ice_add_vlan_hw_filters(vsi, vlan_list->vlan_id, + vlan_list->num_elements); + if (status) { + device_printf(sc->dev, + "VF-%d: Failure adding VLANs to VSI %d, err %s aq_err %s\n", + vf->vf_num, vsi->idx, ice_status_str(status), + ice_aq_str(sc->hw.adminq.sq_last_status)); + v_status = ice_iov_err_to_virt_err(status); + goto done; + } + + vf->vlan_cnt += vlan_list->num_elements; + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_ADD_VLAN, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_del_vlan_msg - Handle VIRTCHNL_OP_DEL_VLAN msg from VF + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + * + * Removes the VLANs in msg_buf from the VF's VLAN filter list. + */ +static void +ice_vc_del_vlan_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct ice_hw *hw = &sc->hw; + struct virtchnl_vlan_filter_list *vlan_list; + int status = 0; + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct ice_vsi *vsi = vf->vsi; + + vlan_list = (struct virtchnl_vlan_filter_list *)msg_buf; + + if (vlan_list->vsi_id != vsi->idx) { + device_printf(sc->dev, + "VF-%d: Message has invalid VSI ID (expected %d, got %d)\n", + vf->vf_num, vsi->idx, vlan_list->vsi_id); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + status = ice_remove_vlan_hw_filters(vsi, vlan_list->vlan_id, + vlan_list->num_elements); + if (status) { + device_printf(sc->dev, + "VF-%d: Failure deleting VLANs from VSI %d, err %s aq_err %s\n", + vf->vf_num, vsi->idx, ice_status_str(status), + ice_aq_str(sc->hw.adminq.sq_last_status)); + v_status = ice_iov_err_to_virt_err(status); + goto done; + } + + if (vlan_list->num_elements >= vf->vlan_cnt) + vf->vlan_cnt = 0; + else + vf->vlan_cnt -= vlan_list->num_elements; + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_DEL_VLAN, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_validate_ring_len - Check to see if a descriptor ring length is valid + * @ring_len: length of ring + * + * Check whether a ring size value is valid. + * + * @returns true if given ring size is valid + */ +static bool +ice_vc_isvalid_ring_len(u16 ring_len) +{ + return (ring_len >= ICE_MIN_DESC_COUNT && + ring_len <= ICE_MAX_DESC_COUNT && + !(ring_len % ICE_DESC_COUNT_INCR)); +} + +/** + * ice_vc_cfg_vsi_qs_msg - Handle VIRTCHNL_OP_CONFIG_VSI_QUEUES msg from VF + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + */ +static void +ice_vc_cfg_vsi_qs_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + device_t dev = sc->dev; + struct ice_hw *hw = &sc->hw; + struct virtchnl_vsi_queue_config_info *vqci; + struct virtchnl_queue_pair_info *vqpi; + enum virtchnl_status_code status = VIRTCHNL_STATUS_SUCCESS; + struct ice_vsi *vsi = vf->vsi; + struct ice_tx_queue *txq; + struct ice_rx_queue *rxq; + int i, error = 0; + + vqci = (struct virtchnl_vsi_queue_config_info *)msg_buf; + + if (vqci->num_queue_pairs > vf->vsi->num_tx_queues && + vqci->num_queue_pairs > vf->vsi->num_rx_queues) { + status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + ice_vsi_disable_tx(vf->vsi); + ice_control_all_rx_queues(vf->vsi, false); + + /* + * Clear TX and RX queues config in case VF + * requests different number of queues. + */ + for (i = 0; i < vsi->num_tx_queues; i++) { + txq = &vsi->tx_queues[i]; + + txq->desc_count = 0; + txq->tx_paddr = 0; + txq->tc = 0; + } + + for (i = 0; i < vsi->num_rx_queues; i++) { + rxq = &vsi->rx_queues[i]; + + rxq->desc_count = 0; + rxq->rx_paddr = 0; + } + + vqpi = vqci->qpair; + for (i = 0; i < vqci->num_queue_pairs; i++, vqpi++) { + /* Initial parameter validation */ + if (vqpi->txq.vsi_id != vf->vsi->idx || + vqpi->rxq.vsi_id != vf->vsi->idx || + vqpi->txq.queue_id != vqpi->rxq.queue_id || + vqpi->txq.headwb_enabled || + vqpi->rxq.splithdr_enabled || + vqpi->rxq.crc_disable || + !(ice_vc_isvalid_ring_len(vqpi->txq.ring_len)) || + !(ice_vc_isvalid_ring_len(vqpi->rxq.ring_len))) { + status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + /* Copy parameters into VF's queue/VSI structs */ + txq = &vsi->tx_queues[vqpi->txq.queue_id]; + + txq->desc_count = vqpi->txq.ring_len; + txq->tx_paddr = vqpi->txq.dma_ring_addr; + txq->q_handle = vqpi->txq.queue_id; + txq->tc = 0; + + rxq = &vsi->rx_queues[vqpi->rxq.queue_id]; + + rxq->desc_count = vqpi->rxq.ring_len; + rxq->rx_paddr = vqpi->rxq.dma_ring_addr; + vsi->mbuf_sz = vqpi->rxq.databuffer_size; + } + + /* Configure TX queues in HW */ + error = ice_cfg_vsi_for_tx(vsi); + if (error) { + device_printf(dev, + "VF-%d: Unable to configure VSI for Tx: %s\n", + vf->vf_num, ice_err_str(error)); + status = VIRTCHNL_STATUS_ERR_ADMIN_QUEUE_ERROR; + goto done; + } + + /* Configure RX queues in HW */ + error = ice_cfg_vsi_for_rx(vsi); + if (error) { + device_printf(dev, + "VF-%d: Unable to configure VSI for Rx: %s\n", + vf->vf_num, ice_err_str(error)); + status = VIRTCHNL_STATUS_ERR_ADMIN_QUEUE_ERROR; + ice_vsi_disable_tx(vsi); + goto done; + } + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_CONFIG_VSI_QUEUES, + status, NULL, 0, NULL); +} + +/** + * ice_vc_cfg_rss_key_msg - Handle VIRTCHNL_OP_CONFIG_RSS_KEY msg from VF + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + * + * Sets the RSS key for the given VF, using the contents of msg_buf. + */ +static void +ice_vc_cfg_rss_key_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct ice_aqc_get_set_rss_keys keydata = + { .standard_rss_key = {0}, .extended_hash_key = {0} }; + struct ice_hw *hw = &sc->hw; + struct virtchnl_rss_key *vrk; + int status = 0; + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct ice_vsi *vsi = vf->vsi; + + vrk = (struct virtchnl_rss_key *)msg_buf; + + if (vrk->vsi_id != vsi->idx) { + device_printf(sc->dev, + "VF-%d: Message has invalid VSI ID (expected %d, got %d)\n", + vf->vf_num, vsi->idx, vrk->vsi_id); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + if ((vrk->key_len > + (ICE_AQC_GET_SET_RSS_KEY_DATA_RSS_KEY_SIZE + + ICE_AQC_GET_SET_RSS_KEY_DATA_HASH_KEY_SIZE)) || + vrk->key_len == 0) { + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + memcpy(&keydata, vrk->key, vrk->key_len); + + status = ice_aq_set_rss_key(hw, vsi->idx, &keydata); + if (status) { + device_printf(sc->dev, + "ice_aq_set_rss_key status %s, error %s\n", + ice_status_str(status), ice_aq_str(hw->adminq.sq_last_status)); + v_status = ice_iov_err_to_virt_err(status); + goto done; + } + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_CONFIG_RSS_KEY, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_cfg_rss_lut_msg - Handle VIRTCHNL_OP_CONFIG_RSS_LUT msg from VF + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + * + * Adds the LUT from the VF in msg_buf to the PF via an admin queue call. + */ +static void +ice_vc_cfg_rss_lut_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct ice_hw *hw = &sc->hw; + struct virtchnl_rss_lut *vrl; + int status = 0; + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct ice_aq_get_set_rss_lut_params lut_params = {}; + struct ice_vsi *vsi = vf->vsi; + + vrl = (struct virtchnl_rss_lut *)msg_buf; + + if (vrl->vsi_id != vsi->idx) { + device_printf(sc->dev, + "VF-%d: Message has invalid VSI ID (expected %d, got %d)\n", + vf->vf_num, vsi->idx, vrl->vsi_id); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + if (vrl->lut_entries > ICE_VSIQF_HLUT_ARRAY_SIZE) { + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + lut_params.vsi_handle = vsi->idx; + lut_params.lut_size = vsi->rss_table_size; + lut_params.lut_type = vsi->rss_lut_type; + lut_params.lut = vrl->lut; + lut_params.global_lut_id = 0; + + status = ice_aq_set_rss_lut(hw, &lut_params); + if (status) { + device_printf(sc->dev, + "VF-%d: Cannot set RSS lut, err %s aq_err %s\n", + vf->vf_num, ice_status_str(status), + ice_aq_str(hw->adminq.sq_last_status)); + v_status = ice_iov_err_to_virt_err(status); + } + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_CONFIG_RSS_LUT, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_set_rss_hena_msg - Handle VIRTCHNL_OP_SET_RSS_HENA msg from VF + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + * + * Adds the VF's hena (hash enable) bits as flow types to the PF's RSS flow + * type list. + */ +static void +ice_vc_set_rss_hena_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct ice_hw *hw = &sc->hw; + struct virtchnl_rss_hena *vrh; + int status = 0; + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct ice_vsi *vsi = vf->vsi; + + MPASS(vsi != NULL); + + vrh = (struct virtchnl_rss_hena *)msg_buf; + + /* + * Remove existing configuration to make sure only requested + * config is applied and allow VFs to disable RSS completly. + */ + status = ice_rem_vsi_rss_cfg(hw, vsi->idx); + if (vrh->hena) { + /* + * Problem with removing config is not fatal, when new one + * is requested. Warn about it but try to apply new config + * anyway. + */ + if (status) + device_printf(sc->dev, + "ice_rem_vsi_rss_cfg status %s, error %s\n", + ice_status_str(status), + ice_aq_str(hw->adminq.sq_last_status)); + status = ice_add_avf_rss_cfg(hw, vsi->idx, vrh->hena); + if (status) + device_printf(sc->dev, + "ice_add_avf_rss_cfg status %s, error %s\n", + ice_status_str(status), + ice_aq_str(hw->adminq.sq_last_status)); + } + v_status = ice_iov_err_to_virt_err(status); + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_SET_RSS_HENA, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_enable_queues_msg - Handle VIRTCHNL_OP_ENABLE_QUEUES msg from VF + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + * + * Enables VF queues selected in msg_buf for Tx/Rx traffic. + * + * @remark Only actually operates on Rx queues; Tx queues are enabled in + * CONFIG_VSI_QUEUES message handler. + */ +static void +ice_vc_enable_queues_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct ice_hw *hw = &sc->hw; + struct virtchnl_queue_select *vqs; + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct ice_vsi *vsi = vf->vsi; + int bit, error = 0; + + vqs = (struct virtchnl_queue_select *)msg_buf; + + if (vqs->vsi_id != vsi->idx) { + device_printf(sc->dev, + "%s: VF-%d: Message has invalid VSI ID (expected %d, got %d)\n", + __func__, vf->vf_num, vsi->idx, vqs->vsi_id); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + if (!vqs->rx_queues && !vqs->tx_queues) { + device_printf(sc->dev, + "%s: VF-%d: message queue masks are empty\n", + __func__, vf->vf_num); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + /* Validate rx_queue mask */ + bit = fls(vqs->rx_queues); + if (bit > vsi->num_rx_queues) { + device_printf(sc->dev, + "%s: VF-%d: message's rx_queues map (0x%08x) has invalid bit set (%d)\n", + __func__, vf->vf_num, vqs->rx_queues, bit); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + /* Tx ring enable is handled in an earlier message. */ + for_each_set_bit(bit, &vqs->rx_queues, 32) { + error = ice_control_rx_queue(vsi, bit, true); + if (error) { + device_printf(sc->dev, + "Unable to enable Rx ring %d for receive: %s\n", + bit, ice_err_str(error)); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + } + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_ENABLE_QUEUES, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_disable_queues_msg - Handle VIRTCHNL_OP_DISABLE_QUEUES msg + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + * + * Disables all VF queues for the VF's VSI. + * + * @remark Unlike the ENABLE_QUEUES handler, this operates on both + * Tx and Rx queues + */ +static void +ice_vc_disable_queues_msg(struct ice_softc *sc, struct ice_vf *vf, + u8 *msg_buf __unused) +{ + struct ice_hw *hw = &sc->hw; + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct ice_vsi *vsi = vf->vsi; + int error = 0; + + error = ice_control_all_rx_queues(vsi, false); + if (error) { + device_printf(sc->dev, + "Unable to disable Rx rings for transmit: %s\n", + ice_err_str(error)); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + error = ice_vsi_disable_tx(vsi); + if (error) { + /* Already prints an error message */ + v_status = VIRTCHNL_STATUS_ERR_PARAM; + } + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_DISABLE_QUEUES, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_cfg_irq_map_msg - Handle VIRTCHNL_OP_CFG_IRQ_MAP msg from VF + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + * + * Configures the interrupt vectors described in the message in msg_buf. The + * VF needs to send this message during init, so that queues can be allowed + * to generate interrupts. + */ +static void +ice_vc_cfg_irq_map_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ +#define ICE_VIRTCHNL_QUEUE_MAP_SIZE 16 + struct ice_hw *hw = &sc->hw; + struct virtchnl_irq_map_info *vimi; + struct virtchnl_vector_map *vvm; + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + struct ice_vsi *vsi = vf->vsi; + u16 vector; + + vimi = (struct virtchnl_irq_map_info *)msg_buf; + + if (vimi->num_vectors > vf->num_irq_vectors) { + device_printf(sc->dev, + "%s: VF-%d: message has more vectors (%d) than configured for VF (%d)\n", + __func__, vf->vf_num, vimi->num_vectors, vf->num_irq_vectors); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + vvm = vimi->vecmap; + /* Save off information from message */ + for (int i = 0; i < vimi->num_vectors; i++, vvm++) { + struct ice_tx_queue *txq; + struct ice_rx_queue *rxq; + int bit; + + if (vvm->vsi_id != vf->vsi->idx) { + device_printf(sc->dev, + "%s: VF-%d: message's VSI ID (%d) does not match VF's (%d) for vector %d\n", + __func__, vf->vf_num, vvm->vsi_id, vf->vsi->idx, i); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + /* vvm->vector_id is relative to VF space */ + vector = vvm->vector_id; + + if (vector >= vf->num_irq_vectors) { + device_printf(sc->dev, + "%s: VF-%d: message's vector ID (%d) is greater than VF's max ID (%d)\n", + __func__, vf->vf_num, vector, vf->num_irq_vectors - 1); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + /* The Misc/Admin Queue vector doesn't need mapping */ + if (vector == 0) + continue; + + /* coverity[address_of] */ + for_each_set_bit(bit, &vvm->txq_map, ICE_VIRTCHNL_QUEUE_MAP_SIZE) { + if (bit >= vsi->num_tx_queues) { + device_printf(sc->dev, + "%s: VF-%d: txq map has invalid bit set\n", + __func__, vf->vf_num); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + vf->tx_irqvs[vector].me = vector; + + txq = &vsi->tx_queues[bit]; + txq->irqv = &vf->tx_irqvs[vector]; + txq->itr_idx = vvm->txitr_idx; + } + /* coverity[address_of] */ + for_each_set_bit(bit, &vvm->rxq_map, ICE_VIRTCHNL_QUEUE_MAP_SIZE) { + if (bit >= vsi->num_rx_queues) { + device_printf(sc->dev, + "%s: VF-%d: rxq map has invalid bit set\n", + __func__, vf->vf_num); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + vf->rx_irqvs[vector].me = vector; + + rxq = &vsi->rx_queues[bit]; + rxq->irqv = &vf->rx_irqvs[vector]; + rxq->itr_idx = vvm->rxitr_idx; + } + } + + /* Write to T/RQCTL registers to actually map vectors to queues */ + for (int i = 0; i < vf->vsi->num_rx_queues; i++) + if (vsi->rx_queues[i].irqv != NULL) + ice_configure_rxq_interrupt(hw, vsi->rx_qmap[i], + vsi->rx_queues[i].irqv->me, vsi->rx_queues[i].itr_idx); + + for (int i = 0; i < vf->vsi->num_tx_queues; i++) + if (vsi->tx_queues[i].irqv != NULL) + ice_configure_txq_interrupt(hw, vsi->tx_qmap[i], + vsi->tx_queues[i].irqv->me, vsi->tx_queues[i].itr_idx); + + ice_flush(hw); + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_CONFIG_IRQ_MAP, + v_status, NULL, 0, NULL); +} + +/** + * ice_eth_stats_to_virtchnl_eth_stats - Convert stats for virtchnl + * @istats: VSI stats from HW to convert + * @vstats: stats struct to copy to + * + * This function copies all known stats in struct virtchnl_eth_stats from the + * input struct ice_eth_stats to an output struct virtchnl_eth_stats. + * + * @remark These two structure types currently have the same definition up to + * the size of struct virtchnl_eth_stats (on FreeBSD), but that could change + * in the future. + */ +static void +ice_eth_stats_to_virtchnl_eth_stats(struct ice_eth_stats *istats, + struct virtchnl_eth_stats *vstats) +{ + vstats->rx_bytes = istats->rx_bytes; + vstats->rx_unicast = istats->rx_unicast; + vstats->rx_multicast = istats->rx_multicast; + vstats->rx_broadcast = istats->rx_broadcast; + vstats->rx_discards = istats->rx_discards; + vstats->rx_unknown_protocol = istats->rx_unknown_protocol; + vstats->tx_bytes = istats->tx_bytes; + vstats->tx_unicast = istats->tx_unicast; + vstats->tx_multicast = istats->tx_multicast; + vstats->tx_broadcast = istats->tx_broadcast; + vstats->tx_discards = istats->tx_discards; + vstats->tx_errors = istats->tx_errors; +} + +/** + * ice_vc_get_stats_msg - Handle VIRTCHNL_OP_GET_STATS msg + * @sc: device private structure + * @vf: VF tracking structure + * @msg_buf: raw message buffer from the VF + * + * Updates the VF's VSI stats and sends those stats back to the VF. + */ +static void +ice_vc_get_stats_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct virtchnl_queue_select *vqs; + struct virtchnl_eth_stats stats; + struct ice_vsi *vsi = vf->vsi; + struct ice_hw *hw = &sc->hw; + + vqs = (struct virtchnl_queue_select *)msg_buf; + + if (vqs->vsi_id != vsi->idx) { + device_printf(sc->dev, + "%s: VF-%d: message has invalid VSI ID %d (VF has VSI ID %d)\n", + __func__, vf->vf_num, vqs->vsi_id, vsi->idx); + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_GET_STATS, + VIRTCHNL_STATUS_ERR_PARAM, NULL, 0, NULL); + } + + ice_update_vsi_hw_stats(vf->vsi); + ice_eth_stats_to_virtchnl_eth_stats(&vsi->hw_stats.cur, &stats); + + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_GET_STATS, + VIRTCHNL_STATUS_SUCCESS, (u8 *)&stats, + sizeof(struct virtchnl_eth_stats), NULL); +} + +/** + * ice_vc_cfg_promisc_mode_msg - Handle VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE + * @sc: PF's softc structure + * @vf: VF tracking structure + * @msg_buf: message buffer from VF + * + * Configures the promiscuous modes for the given VSI in msg_buf. + */ +static void +ice_vc_cfg_promisc_mode_msg(struct ice_softc *sc, struct ice_vf *vf, u8 *msg_buf) +{ + struct ice_hw *hw = &sc->hw; + struct virtchnl_promisc_info *vpi; + enum virtchnl_status_code v_status = VIRTCHNL_STATUS_SUCCESS; + int status = 0; + struct ice_vsi *vsi = vf->vsi; + ice_declare_bitmap(old_promisc_mask, ICE_PROMISC_MAX); + ice_declare_bitmap(req_promisc_mask, ICE_PROMISC_MAX); + ice_declare_bitmap(clear_promisc_mask, ICE_PROMISC_MAX); + ice_declare_bitmap(set_promisc_mask, ICE_PROMISC_MAX); + ice_declare_bitmap(old_req_xor_mask, ICE_PROMISC_MAX); + u16 vid; + + vpi = (struct virtchnl_promisc_info *)msg_buf; + + /* Check to see if VF has permission to configure promiscuous mode */ + if (!(vf->vf_flags & VF_FLAG_PROMISC_CAP)) { + device_printf(sc->dev, + "VF-%d: attempted to configure promiscuous mode\n", + vf->vf_num); + /* Don't reply to VF with an error */ + goto done; + } + + if (vpi->vsi_id != vsi->idx) { + device_printf(sc->dev, + "VF-%d: Message has invalid VSI ID (expected %d, got %d)\n", + vf->vf_num, vsi->idx, vpi->vsi_id); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + } + + if (vpi->flags & ~ICE_VIRTCHNL_VALID_PROMISC_FLAGS) { + device_printf(sc->dev, + "VF-%d: Message has invalid promiscuous flags set (valid 0x%02x, got 0x%02x)\n", + vf->vf_num, ICE_VIRTCHNL_VALID_PROMISC_FLAGS, + vpi->flags); + v_status = VIRTCHNL_STATUS_ERR_PARAM; + goto done; + + } + + ice_zero_bitmap(req_promisc_mask, ICE_PROMISC_MAX); + /* Convert virtchnl flags to ice AQ promiscuous mode flags */ + if (vpi->flags & FLAG_VF_UNICAST_PROMISC) { + ice_set_bit(ICE_PROMISC_UCAST_TX, req_promisc_mask); + ice_set_bit(ICE_PROMISC_UCAST_RX, req_promisc_mask); + } + if (vpi->flags & FLAG_VF_MULTICAST_PROMISC) { + ice_set_bit(ICE_PROMISC_MCAST_TX, req_promisc_mask); + ice_set_bit(ICE_PROMISC_MCAST_RX, req_promisc_mask); + } + + status = ice_get_vsi_promisc(hw, vsi->idx, old_promisc_mask, &vid); + if (status) { + device_printf(sc->dev, + "VF-%d: Failed to get promiscuous mode mask for VSI %d, err %s aq_err %s\n", + vf->vf_num, vsi->idx, + ice_status_str(status), + ice_aq_str(hw->adminq.sq_last_status)); + v_status = ice_iov_err_to_virt_err(status); + goto done; + } + + /* Figure out what got added and what got removed */ + ice_zero_bitmap(old_req_xor_mask, ICE_PROMISC_MAX); + ice_xor_bitmap(old_req_xor_mask, old_promisc_mask, req_promisc_mask, ICE_PROMISC_MAX); + ice_and_bitmap(clear_promisc_mask, old_req_xor_mask, old_promisc_mask, ICE_PROMISC_MAX); + ice_and_bitmap(set_promisc_mask, old_req_xor_mask, req_promisc_mask, ICE_PROMISC_MAX); + + if (ice_is_any_bit_set(clear_promisc_mask, ICE_PROMISC_MAX)) { + status = ice_clear_vsi_promisc(hw, vsi->idx, + clear_promisc_mask, 0); + if (status) { + device_printf(sc->dev, + "VF-%d: Failed to clear promiscuous mode for VSI %d, err %s aq_err %s\n", + vf->vf_num, vsi->idx, + ice_status_str(status), + ice_aq_str(hw->adminq.sq_last_status)); + v_status = ice_iov_err_to_virt_err(status); + goto done; + } + } + + if (ice_is_any_bit_set(set_promisc_mask, ICE_PROMISC_MAX)) { + status = ice_set_vsi_promisc(hw, vsi->idx, set_promisc_mask, 0); + if (status) { + device_printf(sc->dev, + "VF-%d: Failed to set promiscuous mode for VSI %d, err %s aq_err %s\n", + vf->vf_num, vsi->idx, + ice_status_str(status), + ice_aq_str(hw->adminq.sq_last_status)); + v_status = ice_iov_err_to_virt_err(status); + goto done; + } + } + +done: + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE, + v_status, NULL, 0, NULL); +} + +/** + * ice_vc_notify_all_vfs_link_state - Notify all VFs of PF link state + * @sc: device private structure + * + * Sends a message to all VFs about the status of the PF's link + * state. For more details, @see ice_vc_notify_vf_link_state. + */ +void +ice_vc_notify_all_vfs_link_state(struct ice_softc *sc) +{ + for (int i = 0; i < sc->num_vfs; i++) + ice_vc_notify_vf_link_state(sc, &sc->vfs[i]); +} + +/** + * ice_vc_notify_vf_link_state - Notify VF of PF link state + * @sc: device private structure + * @vf: VF tracking structure + * + * Sends an event message to the specified VF with information about + * the current link state from the PF's port. This includes whether + * link is up or down, and the link speed in 100Mbps units. + */ +static void +ice_vc_notify_vf_link_state(struct ice_softc *sc, struct ice_vf *vf) +{ + struct virtchnl_pf_event event = {}; + struct ice_hw *hw = &sc->hw; + + event.event = VIRTCHNL_EVENT_LINK_CHANGE; + event.severity = PF_EVENT_SEVERITY_INFO; + event.event_data.link_event_adv.link_status = sc->link_up; + event.event_data.link_event_adv.link_speed = + (u32)ice_conv_link_speed_to_virtchnl(true, + hw->port_info->phy.link_info.link_speed); + + ice_aq_send_msg_to_vf(hw, vf->vf_num, VIRTCHNL_OP_EVENT, + VIRTCHNL_STATUS_SUCCESS, (u8 *)&event, sizeof(event), NULL); +} + +/** + * ice_vc_handle_vf_msg - Handle a message from a VF + * @sc: device private structure + * @event: event received from the HW MBX queue + * + * Called whenever an event is received from a VF on the HW mailbox queue. + * Responsible for handling these messages as well as responding to the + * VF afterwards, depending on the received message type. + */ +void +ice_vc_handle_vf_msg(struct ice_softc *sc, struct ice_rq_event_info *event) +{ + struct ice_hw *hw = &sc->hw; + device_t dev = sc->dev; + struct ice_vf *vf; + int err = 0; + + u32 v_opcode = event->desc.cookie_high; + u16 v_id = event->desc.retval; + u8 *msg = event->msg_buf; + u16 msglen = event->msg_len; + + if (v_id >= sc->num_vfs) { + device_printf(dev, "%s: Received msg from invalid VF-%d: opcode %d, len %d\n", + __func__, v_id, v_opcode, msglen); + return; + } + + vf = &sc->vfs[v_id]; + + /* Perform basic checks on the msg */ + err = virtchnl_vc_validate_vf_msg(&vf->version, v_opcode, msg, msglen); + if (err) { + device_printf(dev, "%s: Received invalid msg from VF-%d: opcode %d, len %d, error %d\n", + __func__, vf->vf_num, v_opcode, msglen, err); + ice_aq_send_msg_to_vf(hw, v_id, v_opcode, VIRTCHNL_STATUS_ERR_PARAM, NULL, 0, NULL); + return; + } + + switch (v_opcode) { + case VIRTCHNL_OP_VERSION: + ice_vc_version_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_RESET_VF: + ice_reset_vf(sc, vf, true); + break; + case VIRTCHNL_OP_GET_VF_RESOURCES: + ice_vc_get_vf_res_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_ADD_ETH_ADDR: + ice_vc_add_eth_addr_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_DEL_ETH_ADDR: + ice_vc_del_eth_addr_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_ADD_VLAN: + ice_vc_add_vlan_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_DEL_VLAN: + ice_vc_del_vlan_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_CONFIG_VSI_QUEUES: + ice_vc_cfg_vsi_qs_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_CONFIG_RSS_KEY: + ice_vc_cfg_rss_key_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_CONFIG_RSS_LUT: + ice_vc_cfg_rss_lut_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_SET_RSS_HENA: + ice_vc_set_rss_hena_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_ENABLE_QUEUES: + ice_vc_enable_queues_msg(sc, vf, msg); + ice_vc_notify_vf_link_state(sc, vf); + break; + case VIRTCHNL_OP_DISABLE_QUEUES: + ice_vc_disable_queues_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_CONFIG_IRQ_MAP: + ice_vc_cfg_irq_map_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_GET_STATS: + ice_vc_get_stats_msg(sc, vf, msg); + break; + case VIRTCHNL_OP_CONFIG_PROMISCUOUS_MODE: + ice_vc_cfg_promisc_mode_msg(sc, vf, msg); + break; + default: + device_printf(dev, "%s: Received unknown msg from VF-%d: opcode %d, len %d\n", + __func__, vf->vf_num, v_opcode, msglen); + ice_aq_send_msg_to_vf(hw, v_id, v_opcode, + VIRTCHNL_STATUS_ERR_NOT_SUPPORTED, NULL, 0, NULL); + break; + } +} + +/** + * ice_iov_setup_intr_mapping - Setup interrupt config for a VF + * @sc: device softc structure + * @vf: driver's VF structure for VF to be configured + * + * Before a VF can be used, and after a VF reset, the PF must configure + * the VF's interrupt allocation registers. This includes allocating + * interrupts from the PF's interrupt pool to the VF using the + * VPINT_ALLOC(_PCI) registers, and setting up a mapping from PF vectors + * to VF vectors in GLINT_VECT2FUNC. + * + * As well, this sets up queue allocation registers and maps the mailbox + * interrupt for the VF. + */ +static void +ice_iov_setup_intr_mapping(struct ice_softc *sc, struct ice_vf *vf) +{ + struct ice_hw *hw = &sc->hw; + struct ice_vsi *vsi = vf->vsi; + u16 v; + + /* Calculate indices for register ops below */ + u16 vf_first_irq_idx = vf->vf_imap[0]; + u16 vf_last_irq_idx = (vf_first_irq_idx + vf->num_irq_vectors) - 1; + u16 abs_vf_first_irq_idx = hw->func_caps.common_cap.msix_vector_first_id + + vf_first_irq_idx; + u16 abs_vf_last_irq_idx = (abs_vf_first_irq_idx + vf->num_irq_vectors) - 1; + u16 abs_vf_num = vf->vf_num + hw->func_caps.vf_base_id; + + /* Map out VF interrupt allocation in global device space. Both + * VPINT_ALLOC and VPINT_ALLOC_PCI use the same values. + */ + wr32(hw, VPINT_ALLOC(vf->vf_num), + (((abs_vf_first_irq_idx << VPINT_ALLOC_FIRST_S) & VPINT_ALLOC_FIRST_M) | + ((abs_vf_last_irq_idx << VPINT_ALLOC_LAST_S) & VPINT_ALLOC_LAST_M) | + VPINT_ALLOC_VALID_M)); + wr32(hw, VPINT_ALLOC_PCI(vf->vf_num), + (((abs_vf_first_irq_idx << VPINT_ALLOC_PCI_FIRST_S) & VPINT_ALLOC_PCI_FIRST_M) | + ((abs_vf_last_irq_idx << VPINT_ALLOC_PCI_LAST_S) & VPINT_ALLOC_PCI_LAST_M) | + VPINT_ALLOC_PCI_VALID_M)); + + /* Create inverse mapping of vectors to PF/VF combinations */ + for (v = vf_first_irq_idx; v <= vf_last_irq_idx; v++) + { + wr32(hw, GLINT_VECT2FUNC(v), + (((abs_vf_num << GLINT_VECT2FUNC_VF_NUM_S) & GLINT_VECT2FUNC_VF_NUM_M) | + ((hw->pf_id << GLINT_VECT2FUNC_PF_NUM_S) & GLINT_VECT2FUNC_PF_NUM_M))); + } + + /* Map mailbox interrupt to MSI-X index 0. Disable ITR for it, too. */ + wr32(hw, VPINT_MBX_CTL(abs_vf_num), + ((0 << VPINT_MBX_CTL_MSIX_INDX_S) & VPINT_MBX_CTL_MSIX_INDX_M) | + ((0x3 << VPINT_MBX_CTL_ITR_INDX_S) & VPINT_MBX_CTL_ITR_INDX_M) | + VPINT_MBX_CTL_CAUSE_ENA_M); + + /* Mark the TX queue mapping registers as valid */ + wr32(hw, VPLAN_TXQ_MAPENA(vf->vf_num), VPLAN_TXQ_MAPENA_TX_ENA_M); + + /* Indicate to HW that VF has scattered queue allocation */ + wr32(hw, VPLAN_TX_QBASE(vf->vf_num), VPLAN_TX_QBASE_VFQTABLE_ENA_M); + for (int i = 0; i < vsi->num_tx_queues; i++) { + wr32(hw, VPLAN_TX_QTABLE(i, vf->vf_num), + (vsi->tx_qmap[i] << VPLAN_TX_QTABLE_QINDEX_S) & VPLAN_TX_QTABLE_QINDEX_M); + } + + /* Mark the RX queue mapping registers as valid */ + wr32(hw, VPLAN_RXQ_MAPENA(vf->vf_num), VPLAN_RXQ_MAPENA_RX_ENA_M); + wr32(hw, VPLAN_RX_QBASE(vf->vf_num), VPLAN_RX_QBASE_VFQTABLE_ENA_M); + for (int i = 0; i < vsi->num_rx_queues; i++) { + wr32(hw, VPLAN_RX_QTABLE(i, vf->vf_num), + (vsi->rx_qmap[i] << VPLAN_RX_QTABLE_QINDEX_S) & VPLAN_RX_QTABLE_QINDEX_M); + } +} + +/** + * ice_err_to_virt err - translate ice errors into virtchnl errors + * @ice_err: status returned from ice function + */ +static enum virtchnl_status_code +ice_iov_err_to_virt_err(int ice_err) +{ + switch (ice_err) { + case 0: + return VIRTCHNL_STATUS_SUCCESS; + case ICE_ERR_BAD_PTR: + case ICE_ERR_INVAL_SIZE: + case ICE_ERR_DEVICE_NOT_SUPPORTED: + case ICE_ERR_PARAM: + case ICE_ERR_CFG: + return VIRTCHNL_STATUS_ERR_PARAM; + case ICE_ERR_NO_MEMORY: + return VIRTCHNL_STATUS_ERR_NO_MEMORY; + case ICE_ERR_NOT_READY: + case ICE_ERR_RESET_FAILED: + case ICE_ERR_FW_API_VER: + case ICE_ERR_AQ_ERROR: + case ICE_ERR_AQ_TIMEOUT: + case ICE_ERR_AQ_FULL: + case ICE_ERR_AQ_NO_WORK: + case ICE_ERR_AQ_EMPTY: + return VIRTCHNL_STATUS_ERR_ADMIN_QUEUE_ERROR; + default: + return VIRTCHNL_STATUS_ERR_NOT_SUPPORTED; + } +} diff --git a/sys/dev/ice/ice_iov.h b/sys/dev/ice/ice_iov.h new file mode 100644 index 000000000000..c4fb3e932e3f --- /dev/null +++ b/sys/dev/ice/ice_iov.h @@ -0,0 +1,125 @@ +/* SPDX-License-Identifier: BSD-3-Clause */ +/* Copyright (c) 2025, Intel Corporation + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * 3. Neither the name of the Intel Corporation nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +/** + * @file ice_iov.h + * @brief header for IOV functionality + * + * This header includes definitions used to implement device Virtual Functions + * for the ice driver. + */ + +#ifndef _ICE_IOV_H_ +#define _ICE_IOV_H_ + +#include <sys/types.h> +#include <sys/bus.h> +#include <sys/nv.h> +#include <sys/iov_schema.h> +#include <sys/param.h> +#include <dev/pci/pcivar.h> +#include <dev/pci/pcireg.h> + +#include <dev/pci/pci_iov.h> + +#include "ice_iflib.h" +#include "ice_vf_mbx.h" + +/** + * @enum ice_vf_flags + * @brief VF state flags + * + * Used to indicate the status of a PF's VF, as well as indicating what each VF + * is capabile of. Intended to be modified only using atomic operations, so + * they can be read and modified in places that aren't locked. + * + * Used in struct ice_vf's vf_flags field. + */ +enum ice_vf_flags { + VF_FLAG_ENABLED = BIT(0), + VF_FLAG_SET_MAC_CAP = BIT(1), + VF_FLAG_VLAN_CAP = BIT(2), + VF_FLAG_PROMISC_CAP = BIT(3), + VF_FLAG_MAC_ANTI_SPOOF = BIT(4), +}; + +/** + * @struct ice_vf + * @brief PF's VF software context + * + * Represents the state and options for a VF spawned from a PF. + */ +struct ice_vf { + struct ice_vsi *vsi; + u32 vf_flags; + + u8 mac[ETHER_ADDR_LEN]; + u16 vf_num; + struct virtchnl_version_info version; + + u16 mac_filter_limit; + u16 mac_filter_cnt; + u16 vlan_limit; + u16 vlan_cnt; + + u16 num_irq_vectors; + u16 *vf_imap; + struct ice_irq_vector *tx_irqvs; + struct ice_irq_vector *rx_irqvs; +}; + +#define ICE_PCIE_DEV_STATUS 0xAA + +#define ICE_PCI_CIAD_WAIT_COUNT 100 +#define ICE_PCI_CIAD_WAIT_DELAY_US 1 +#define ICE_VPGEN_VFRSTAT_WAIT_COUNT 100 +#define ICE_VPGEN_VFRSTAT_WAIT_DELAY_US 20 + +#define ICE_VIRTCHNL_VALID_PROMISC_FLAGS (FLAG_VF_UNICAST_PROMISC | \ + FLAG_VF_MULTICAST_PROMISC) + +#define ICE_DEFAULT_VF_VLAN_LIMIT 64 +#define ICE_DEFAULT_VF_FILTER_LIMIT 16 + +int ice_iov_attach(struct ice_softc *sc); +int ice_iov_detach(struct ice_softc *sc); + +int ice_iov_init(struct ice_softc *sc, uint16_t num_vfs, const nvlist_t *params); +int ice_iov_add_vf(struct ice_softc *sc, uint16_t vfnum, const nvlist_t *params); +void ice_iov_uninit(struct ice_softc *sc); + +void ice_iov_handle_vflr(struct ice_softc *sc); + +void ice_vc_handle_vf_msg(struct ice_softc *sc, struct ice_rq_event_info *event); +void ice_vc_notify_all_vfs_link_state(struct ice_softc *sc); + +#endif /* _ICE_IOV_H_ */ + diff --git a/sys/dev/ice/ice_lib.c b/sys/dev/ice/ice_lib.c index d44ae5f37750..442111e5ffaf 100644 --- a/sys/dev/ice/ice_lib.c +++ b/sys/dev/ice/ice_lib.c @@ -42,6 +42,9 @@ #include "ice_lib.h" #include "ice_iflib.h" +#ifdef PCI_IOV +#include "ice_iov.h" +#endif #include <dev/pci/pcivar.h> #include <dev/pci/pcireg.h> #include <machine/resource.h> @@ -741,6 +744,12 @@ ice_initialize_vsi(struct ice_vsi *vsi) case ICE_VSI_VMDQ2: ctx.flags = ICE_AQ_VSI_TYPE_VMDQ2; break; +#ifdef PCI_IOV + case ICE_VSI_VF: + ctx.flags = ICE_AQ_VSI_TYPE_VF; + ctx.vf_num = vsi->vf_num; + break; +#endif default: return (ENODEV); } @@ -1607,6 +1616,12 @@ ice_setup_tx_ctx(struct ice_tx_queue *txq, struct ice_tlan_ctx *tlan_ctx, u16 pf case ICE_VSI_VMDQ2: tlan_ctx->vmvf_type = ICE_TLAN_CTX_VMVF_TYPE_VMQ; break; +#ifdef PCI_IOV + case ICE_VSI_VF: + tlan_ctx->vmvf_type = ICE_TLAN_CTX_VMVF_TYPE_VF; + tlan_ctx->vmvf_num = hw->func_caps.vf_base_id + vsi->vf_num; + break; +#endif default: return (ENODEV); } @@ -1660,6 +1675,10 @@ ice_cfg_vsi_for_tx(struct ice_vsi *vsi) struct ice_tlan_ctx tlan_ctx = { 0 }; struct ice_tx_queue *txq = &vsi->tx_queues[i]; + /* Last configured queue */ + if (txq->desc_count == 0) + break; + pf_q = vsi->tx_qmap[txq->me]; qg->txqs[0].txq_id = htole16(pf_q); @@ -1788,6 +1807,10 @@ ice_cfg_vsi_for_rx(struct ice_vsi *vsi) for (i = 0; i < vsi->num_rx_queues; i++) { MPASS(vsi->mbuf_sz > 0); + /* Last configured queue */ + if (vsi->rx_queues[i].desc_count == 0) + break; + err = ice_setup_rx_ctx(&vsi->rx_queues[i]); if (err) return err; @@ -2257,6 +2280,11 @@ ice_process_ctrlq_event(struct ice_softc *sc, const char *qname, case ice_aqc_opc_get_link_status: ice_process_link_event(sc, event); break; +#ifdef PCI_IOV + case ice_mbx_opc_send_msg_to_pf: + ice_vc_handle_vf_msg(sc, event); + break; +#endif case ice_aqc_opc_fw_logs_event: ice_handle_fw_log_event(sc, &event->desc, event->msg_buf); break; diff --git a/sys/dev/ice/ice_lib.h b/sys/dev/ice/ice_lib.h index b6b23ec82161..308b2bda2790 100644 --- a/sys/dev/ice/ice_lib.h +++ b/sys/dev/ice/ice_lib.h @@ -611,6 +611,10 @@ struct ice_vsi { u16 mirror_src_vsi; u16 rule_mir_ingress; u16 rule_mir_egress; + +#ifdef PCI_IOV + u8 vf_num; /* Index of owning VF, if applicable */ +#endif }; /** diff --git a/sys/dev/ice/ice_vf_mbx.c b/sys/dev/ice/ice_vf_mbx.c new file mode 100644 index 000000000000..387a1c6739a6 --- /dev/null +++ b/sys/dev/ice/ice_vf_mbx.c @@ -0,0 +1,471 @@ +/* SPDX-License-Identifier: BSD-3-Clause */ +/* Copyright (c) 2025, Intel Corporation + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * 3. Neither the name of the Intel Corporation nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#include "ice_common.h" +#include "ice_hw_autogen.h" +#include "ice_vf_mbx.h" + +/** + * ice_aq_send_msg_to_vf + * @hw: pointer to the hardware structure + * @vfid: VF ID to send msg + * @v_opcode: opcodes for VF-PF communication + * @v_retval: return error code + * @msg: pointer to the msg buffer + * @msglen: msg length + * @cd: pointer to command details + * + * Send message to VF driver (0x0802) using mailbox + * queue and asynchronously sending message via + * ice_sq_send_cmd() function + */ +int +ice_aq_send_msg_to_vf(struct ice_hw *hw, u16 vfid, u32 v_opcode, u32 v_retval, + u8 *msg, u16 msglen, struct ice_sq_cd *cd) +{ + struct ice_aqc_pf_vf_msg *cmd; + struct ice_aq_desc desc; + + ice_fill_dflt_direct_cmd_desc(&desc, ice_mbx_opc_send_msg_to_vf); + + cmd = &desc.params.virt; + cmd->id = CPU_TO_LE32(vfid); + + desc.cookie_high = CPU_TO_LE32(v_opcode); + desc.cookie_low = CPU_TO_LE32(v_retval); + + if (msglen) + desc.flags |= CPU_TO_LE16(ICE_AQ_FLAG_RD); + + return ice_sq_send_cmd(hw, &hw->mailboxq, &desc, msg, msglen, cd); +} + +/** + * ice_aq_send_msg_to_pf + * @hw: pointer to the hardware structure + * @v_opcode: opcodes for VF-PF communication + * @v_retval: return error code + * @msg: pointer to the msg buffer + * @msglen: msg length + * @cd: pointer to command details + * + * Send message to PF driver using mailbox queue. By default, this + * message is sent asynchronously, i.e. ice_sq_send_cmd() + * does not wait for completion before returning. + */ +int +ice_aq_send_msg_to_pf(struct ice_hw *hw, enum virtchnl_ops v_opcode, + int v_retval, u8 *msg, u16 msglen, + struct ice_sq_cd *cd) +{ + struct ice_aq_desc desc; + + ice_fill_dflt_direct_cmd_desc(&desc, ice_mbx_opc_send_msg_to_pf); + desc.cookie_high = CPU_TO_LE32(v_opcode); + desc.cookie_low = CPU_TO_LE32(v_retval); + + if (msglen) + desc.flags |= CPU_TO_LE16(ICE_AQ_FLAG_RD); + + return ice_sq_send_cmd(hw, &hw->mailboxq, &desc, msg, msglen, cd); +} + +static const u32 ice_legacy_aq_to_vc_speed[] = { + VIRTCHNL_LINK_SPEED_100MB, /* BIT(0) */ + VIRTCHNL_LINK_SPEED_100MB, + VIRTCHNL_LINK_SPEED_1GB, + VIRTCHNL_LINK_SPEED_1GB, + VIRTCHNL_LINK_SPEED_1GB, + VIRTCHNL_LINK_SPEED_10GB, + VIRTCHNL_LINK_SPEED_20GB, + VIRTCHNL_LINK_SPEED_25GB, + VIRTCHNL_LINK_SPEED_40GB, + VIRTCHNL_LINK_SPEED_40GB, + VIRTCHNL_LINK_SPEED_40GB, +}; + +/** + * ice_conv_link_speed_to_virtchnl + * @adv_link_support: determines the format of the returned link speed + * @link_speed: variable containing the link_speed to be converted + * + * Convert link speed supported by HW to link speed supported by virtchnl. + * If adv_link_support is true, then return link speed in Mbps. Else return + * link speed as a VIRTCHNL_LINK_SPEED_* casted to a u32. Note that the caller + * needs to cast back to an enum virtchnl_link_speed in the case where + * adv_link_support is false, but when adv_link_support is true the caller can + * expect the speed in Mbps. + */ +u32 ice_conv_link_speed_to_virtchnl(bool adv_link_support, u16 link_speed) +{ + /* convert a BIT() value into an array index */ + u16 index = (u16)(ice_fls(link_speed) - 1); + + if (adv_link_support) + return ice_get_link_speed(index); + else if (index < ARRAY_SIZE(ice_legacy_aq_to_vc_speed)) + /* Virtchnl speeds are not defined for every speed supported in + * the hardware. To maintain compatibility with older AVF + * drivers, while reporting the speed the new speed values are + * resolved to the closest known virtchnl speeds + */ + return ice_legacy_aq_to_vc_speed[index]; + + return VIRTCHNL_LINK_SPEED_UNKNOWN; +} + +/* The mailbox overflow detection algorithm helps to check if there + * is a possibility of a malicious VF transmitting too many MBX messages to the + * PF. + * 1. The mailbox snapshot structure, ice_mbx_snapshot, is initialized during + * driver initialization in ice_init_hw() using ice_mbx_init_snapshot(). + * The struct ice_mbx_snapshot helps to track and traverse a static window of + * messages within the mailbox queue while looking for a malicious VF. + * + * 2. When the caller starts processing its mailbox queue in response to an + * interrupt, the structure ice_mbx_snapshot is expected to be cleared before + * the algorithm can be run for the first time for that interrupt. This + * requires calling ice_mbx_reset_snapshot() as well as calling + * ice_mbx_reset_vf_info() for each VF tracking structure. + * + * 3. For every message read by the caller from the MBX Queue, the caller must + * call the detection algorithm's entry function ice_mbx_vf_state_handler(). + * Before every call to ice_mbx_vf_state_handler() the struct ice_mbx_data is + * filled as it is required to be passed to the algorithm. + * + * 4. Every time a message is read from the MBX queue, a tracking structure + * for the VF must be passed to the state handler. The boolean output + * report_malvf from ice_mbx_vf_state_handler() serves as an indicator to the + * caller whether it must report this VF as malicious or not. + * + * 5. When a VF is identified to be malicious, the caller can send a message + * to the system administrator. + * + * 6. The PF is responsible for maintaining the struct ice_mbx_vf_info + * structure for each VF. The PF should clear the VF tracking structure if the + * VF is reset. When a VF is shut down and brought back up, we will then + * assume that the new VF is not malicious and may report it again if we + * detect it again. + * + * 7. The function ice_mbx_reset_snapshot() is called to reset the information + * in ice_mbx_snapshot for every new mailbox interrupt handled. + */ +#define ICE_RQ_DATA_MASK(rq_data) ((rq_data) & PF_MBX_ARQH_ARQH_M) +/* Using the highest value for an unsigned 16-bit value 0xFFFF to indicate that + * the max messages check must be ignored in the algorithm + */ +#define ICE_IGNORE_MAX_MSG_CNT 0xFFFF + +/** + * ice_mbx_reset_snapshot - Initialize mailbox snapshot structure + * @snap: pointer to the mailbox snapshot + */ +static void ice_mbx_reset_snapshot(struct ice_mbx_snapshot *snap) +{ + struct ice_mbx_vf_info *vf_info; + + /* Clear mbx_buf in the mailbox snaphot structure and setting the + * mailbox snapshot state to a new capture. + */ + ice_memset(&snap->mbx_buf, 0, sizeof(snap->mbx_buf), ICE_NONDMA_MEM); + snap->mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT; + + /* Reset message counts for all VFs to zero */ + LIST_FOR_EACH_ENTRY(vf_info, &snap->mbx_vf, ice_mbx_vf_info, list_entry) + vf_info->msg_count = 0; +} + +/** + * ice_mbx_traverse - Pass through mailbox snapshot + * @hw: pointer to the HW struct + * @new_state: new algorithm state + * + * Traversing the mailbox static snapshot without checking + * for malicious VFs. + */ +static void +ice_mbx_traverse(struct ice_hw *hw, + enum ice_mbx_snapshot_state *new_state) +{ + struct ice_mbx_snap_buffer_data *snap_buf; + u32 num_iterations; + + snap_buf = &hw->mbx_snapshot.mbx_buf; + + /* As mailbox buffer is circular, applying a mask + * on the incremented iteration count. + */ + num_iterations = ICE_RQ_DATA_MASK(++snap_buf->num_iterations); + + /* Checking either of the below conditions to exit snapshot traversal: + * Condition-1: If the number of iterations in the mailbox is equal to + * the mailbox head which would indicate that we have reached the end + * of the static snapshot. + * Condition-2: If the maximum messages serviced in the mailbox for a + * given interrupt is the highest possible value then there is no need + * to check if the number of messages processed is equal to it. If not + * check if the number of messages processed is greater than or equal + * to the maximum number of mailbox entries serviced in current work item. + */ + if (num_iterations == snap_buf->head || + (snap_buf->max_num_msgs_mbx < ICE_IGNORE_MAX_MSG_CNT && + ++snap_buf->num_msg_proc >= snap_buf->max_num_msgs_mbx)) + *new_state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT; +} + +/** + * ice_mbx_detect_malvf - Detect malicious VF in snapshot + * @hw: pointer to the HW struct + * @vf_info: mailbox tracking structure for a VF + * @new_state: new algorithm state + * @is_malvf: boolean output to indicate if VF is malicious + * + * This function tracks the number of asynchronous messages + * sent per VF and marks the VF as malicious if it exceeds + * the permissible number of messages to send. + */ +static int +ice_mbx_detect_malvf(struct ice_hw *hw, struct ice_mbx_vf_info *vf_info, + enum ice_mbx_snapshot_state *new_state, + bool *is_malvf) +{ + /* increment the message count for this VF */ + vf_info->msg_count++; + + if (vf_info->msg_count >= ICE_ASYNC_VF_MSG_THRESHOLD) + *is_malvf = true; + + /* continue to iterate through the mailbox snapshot */ + ice_mbx_traverse(hw, new_state); + + return 0; +} + +/** + * ice_e830_mbx_vf_dec_trig - Decrements the VF mailbox queue counter + * @hw: pointer to the HW struct + * @event: pointer to the control queue receive event + * + * This function triggers to decrement the counter + * MBX_VF_IN_FLIGHT_MSGS_AT_PF_CNT when the driver replenishes + * the buffers at the PF mailbox queue. + */ +void ice_e830_mbx_vf_dec_trig(struct ice_hw *hw, + struct ice_rq_event_info *event) +{ + u16 vfid = LE16_TO_CPU(event->desc.retval); + + wr32(hw, E830_MBX_VF_DEC_TRIG(vfid), 1); +} + +/** + * ice_mbx_vf_clear_cnt_e830 - Clear the VF mailbox queue count + * @hw: pointer to the HW struct + * @vf_id: VF ID in the PF space + * + * This function clears the counter MBX_VF_IN_FLIGHT_MSGS_AT_PF_CNT, and should + * be called when a VF is created and on VF reset. + */ +void ice_mbx_vf_clear_cnt_e830(struct ice_hw *hw, u16 vf_id) +{ + u32 reg = rd32(hw, E830_MBX_VF_IN_FLIGHT_MSGS_AT_PF_CNT(vf_id)); + + wr32(hw, E830_MBX_VF_DEC_TRIG(vf_id), reg); +} + +/** + * ice_mbx_vf_state_handler - Handle states of the overflow algorithm + * @hw: pointer to the HW struct + * @mbx_data: pointer to structure containing mailbox data + * @vf_info: mailbox tracking structure for the VF in question + * @report_malvf: boolean output to indicate whether VF should be reported + * + * The function serves as an entry point for the malicious VF + * detection algorithm by handling the different states and state + * transitions of the algorithm: + * New snapshot: This state is entered when creating a new static + * snapshot. The data from any previous mailbox snapshot is + * cleared and a new capture of the mailbox head and tail is + * logged. This will be the new static snapshot to detect + * asynchronous messages sent by VFs. On capturing the snapshot + * and depending on whether the number of pending messages in that + * snapshot exceed the watermark value, the state machine enters + * traverse or detect states. + * Traverse: If pending message count is below watermark then iterate + * through the snapshot without any action on VF. + * Detect: If pending message count exceeds watermark traverse + * the static snapshot and look for a malicious VF. + */ +int +ice_mbx_vf_state_handler(struct ice_hw *hw, struct ice_mbx_data *mbx_data, + struct ice_mbx_vf_info *vf_info, bool *report_malvf) +{ + struct ice_mbx_snapshot *snap = &hw->mbx_snapshot; + struct ice_mbx_snap_buffer_data *snap_buf; + struct ice_ctl_q_info *cq = &hw->mailboxq; + enum ice_mbx_snapshot_state new_state; + int status = 0; + bool is_malvf = false; + + if (!report_malvf || !mbx_data || !vf_info) + return ICE_ERR_BAD_PTR; + + *report_malvf = false; + + /* When entering the mailbox state machine assume that the VF + * is not malicious until detected. + */ + /* Checking if max messages allowed to be processed while servicing current + * interrupt is not less than the defined AVF message threshold. + */ + if (mbx_data->max_num_msgs_mbx <= ICE_ASYNC_VF_MSG_THRESHOLD) + return ICE_ERR_INVAL_SIZE; + + /* The watermark value should not be lesser than the threshold limit + * set for the number of asynchronous messages a VF can send to mailbox + * nor should it be greater than the maximum number of messages in the + * mailbox serviced in current interrupt. + */ + if (mbx_data->async_watermark_val < ICE_ASYNC_VF_MSG_THRESHOLD || + mbx_data->async_watermark_val > mbx_data->max_num_msgs_mbx) + return ICE_ERR_PARAM; + + new_state = ICE_MAL_VF_DETECT_STATE_INVALID; + snap_buf = &snap->mbx_buf; + + switch (snap_buf->state) { + case ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT: + /* Clear any previously held data in mailbox snapshot structure. */ + ice_mbx_reset_snapshot(snap); + + /* Collect the pending ARQ count, number of messages processed and + * the maximum number of messages allowed to be processed from the + * Mailbox for current interrupt. + */ + snap_buf->num_pending_arq = mbx_data->num_pending_arq; + snap_buf->num_msg_proc = mbx_data->num_msg_proc; + snap_buf->max_num_msgs_mbx = mbx_data->max_num_msgs_mbx; + + /* Capture a new static snapshot of the mailbox by logging the + * head and tail of snapshot and set num_iterations to the tail + * value to mark the start of the iteration through the snapshot. + */ + snap_buf->head = ICE_RQ_DATA_MASK(cq->rq.next_to_clean + + mbx_data->num_pending_arq); + snap_buf->tail = ICE_RQ_DATA_MASK(cq->rq.next_to_clean - 1); + snap_buf->num_iterations = snap_buf->tail; + + /* Pending ARQ messages returned by ice_clean_rq_elem + * is the difference between the head and tail of the + * mailbox queue. Comparing this value against the watermark + * helps to check if we potentially have malicious VFs. + */ + if (snap_buf->num_pending_arq >= + mbx_data->async_watermark_val) { + new_state = ICE_MAL_VF_DETECT_STATE_DETECT; + status = ice_mbx_detect_malvf(hw, vf_info, &new_state, &is_malvf); + } else { + new_state = ICE_MAL_VF_DETECT_STATE_TRAVERSE; + ice_mbx_traverse(hw, &new_state); + } + break; + + case ICE_MAL_VF_DETECT_STATE_TRAVERSE: + new_state = ICE_MAL_VF_DETECT_STATE_TRAVERSE; + ice_mbx_traverse(hw, &new_state); + break; + + case ICE_MAL_VF_DETECT_STATE_DETECT: + new_state = ICE_MAL_VF_DETECT_STATE_DETECT; + status = ice_mbx_detect_malvf(hw, vf_info, &new_state, &is_malvf); + break; + + default: + new_state = ICE_MAL_VF_DETECT_STATE_INVALID; + status = ICE_ERR_CFG; + } + + snap_buf->state = new_state; + + /* Only report VFs as malicious the first time we detect it */ + if (is_malvf && !vf_info->malicious) { + vf_info->malicious = 1; + *report_malvf = true; + } + + return status; +} + +/** + * ice_mbx_clear_malvf - Clear VF mailbox info + * @vf_info: the mailbox tracking structure for a VF + * + * In case of a VF reset, this function shall be called to clear the VF's + * current mailbox tracking state. + */ +void ice_mbx_clear_malvf(struct ice_mbx_vf_info *vf_info) +{ + vf_info->malicious = 0; + vf_info->msg_count = 0; +} + +/** + * ice_mbx_init_vf_info - Initialize a new VF mailbox tracking info + * @hw: pointer to the hardware structure + * @vf_info: the mailbox tracking info structure for a VF + * + * Initialize a VF mailbox tracking info structure and insert it into the + * snapshot list. + * + * If you remove the VF, you must also delete the associated VF info structure + * from the linked list. + */ +void ice_mbx_init_vf_info(struct ice_hw *hw, struct ice_mbx_vf_info *vf_info) +{ + struct ice_mbx_snapshot *snap = &hw->mbx_snapshot; + + ice_mbx_clear_malvf(vf_info); + LIST_ADD(&vf_info->list_entry, &snap->mbx_vf); +} + +/** + * ice_mbx_init_snapshot - Initialize mailbox snapshot data + * @hw: pointer to the hardware structure + * + * Clear the mailbox snapshot structure and initialize the VF mailbox list. + */ +void ice_mbx_init_snapshot(struct ice_hw *hw) +{ + struct ice_mbx_snapshot *snap = &hw->mbx_snapshot; + + INIT_LIST_HEAD(&snap->mbx_vf); + ice_mbx_reset_snapshot(snap); +} diff --git a/sys/dev/ice/ice_vf_mbx.h b/sys/dev/ice/ice_vf_mbx.h new file mode 100644 index 000000000000..3b185ac89c11 --- /dev/null +++ b/sys/dev/ice/ice_vf_mbx.h @@ -0,0 +1,67 @@ +/* SPDX-License-Identifier: BSD-3-Clause */ +/* Copyright (c) 2025, Intel Corporation + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * 3. Neither the name of the Intel Corporation nor the names of its + * contributors may be used to endorse or promote products derived from + * this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE + * POSSIBILITY OF SUCH DAMAGE. + */ + +#ifndef _ICE_VF_MBX_H_ +#define _ICE_VF_MBX_H_ + +#include "ice_type.h" +#include "ice_controlq.h" + +/* Defining the mailbox message threshold as 63 asynchronous + * pending messages. Normal VF functionality does not require + * sending more than 63 asynchronous pending message. + */ + + /* Threshold value should be used to initialize + * MBX_VF_IN_FLIGHT_MSGS_AT_PF_CNT register. + */ +#define ICE_ASYNC_VF_MSG_THRESHOLD 63 + +int +ice_aq_send_msg_to_pf(struct ice_hw *hw, enum virtchnl_ops v_opcode, + int v_retval, u8 *msg, u16 msglen, + struct ice_sq_cd *cd); +int +ice_aq_send_msg_to_vf(struct ice_hw *hw, u16 vfid, u32 v_opcode, u32 v_retval, + u8 *msg, u16 msglen, struct ice_sq_cd *cd); + +u32 ice_conv_link_speed_to_virtchnl(bool adv_link_support, u16 link_speed); + +void ice_e830_mbx_vf_dec_trig(struct ice_hw *hw, + struct ice_rq_event_info *event); +void ice_mbx_vf_clear_cnt_e830(struct ice_hw *hw, u16 vf_id); +int +ice_mbx_vf_state_handler(struct ice_hw *hw, struct ice_mbx_data *mbx_data, + struct ice_mbx_vf_info *vf_info, bool *report_malvf); +void ice_mbx_clear_malvf(struct ice_mbx_vf_info *vf_info); +void ice_mbx_init_vf_info(struct ice_hw *hw, struct ice_mbx_vf_info *vf_info); +void ice_mbx_init_snapshot(struct ice_hw *hw); +#endif /* _ICE_VF_MBX_H_ */ diff --git a/sys/dev/ice/if_ice_iflib.c b/sys/dev/ice/if_ice_iflib.c index e60ee0f1c5c3..1469d2916465 100644 --- a/sys/dev/ice/if_ice_iflib.c +++ b/sys/dev/ice/if_ice_iflib.c @@ -42,6 +42,9 @@ #include "ice_drv_info.h" #include "ice_switch.h" #include "ice_sched.h" +#ifdef PCI_IOV +#include "ice_iov.h" +#endif #include <sys/module.h> #include <sys/sockio.h> @@ -85,6 +88,12 @@ static int ice_if_suspend(if_ctx_t ctx); static int ice_if_resume(if_ctx_t ctx); static bool ice_if_needs_restart(if_ctx_t ctx, enum iflib_restart_event event); static void ice_init_link(struct ice_softc *sc); +#ifdef PCI_IOV +static int ice_if_iov_init(if_ctx_t ctx, uint16_t num_vfs, const nvlist_t *params); +static void ice_if_iov_uninit(if_ctx_t ctx); +static int ice_if_iov_vf_add(if_ctx_t ctx, uint16_t vfnum, const nvlist_t *params); +static void ice_if_vflr_handle(if_ctx_t ctx); +#endif static int ice_setup_mirror_vsi(struct ice_mirr_if *mif); static int ice_wire_mirror_intrs(struct ice_mirr_if *mif); static void ice_free_irqvs_subif(struct ice_mirr_if *mif); @@ -158,6 +167,11 @@ static device_method_t ice_methods[] = { DEVMETHOD(device_shutdown, iflib_device_shutdown), DEVMETHOD(device_suspend, iflib_device_suspend), DEVMETHOD(device_resume, iflib_device_resume), +#ifdef PCI_IOV + DEVMETHOD(pci_iov_init, iflib_device_iov_init), + DEVMETHOD(pci_iov_uninit, iflib_device_iov_uninit), + DEVMETHOD(pci_iov_add_vf, iflib_device_iov_add_vf), +#endif DEVMETHOD_END }; @@ -198,6 +212,12 @@ static device_method_t ice_iflib_methods[] = { DEVMETHOD(ifdi_suspend, ice_if_suspend), DEVMETHOD(ifdi_resume, ice_if_resume), DEVMETHOD(ifdi_needs_restart, ice_if_needs_restart), +#ifdef PCI_IOV + DEVMETHOD(ifdi_iov_vf_add, ice_if_iov_vf_add), + DEVMETHOD(ifdi_iov_init, ice_if_iov_init), + DEVMETHOD(ifdi_iov_uninit, ice_if_iov_uninit), + DEVMETHOD(ifdi_vflr_handle, ice_if_vflr_handle), +#endif DEVMETHOD_END }; @@ -733,6 +753,9 @@ ice_update_link_status(struct ice_softc *sc, bool update_media) iflib_link_state_change(sc->ctx, LINK_STATE_DOWN, 0); ice_rdma_link_change(sc, LINK_STATE_DOWN, 0); } +#ifdef PCI_IOV + ice_vc_notify_all_vfs_link_state(sc); +#endif update_media = true; } @@ -831,6 +854,14 @@ ice_if_attach_post(if_ctx_t ctx) ice_add_device_sysctls(sc); +#ifdef PCI_IOV + if (ice_is_bit_set(sc->feat_cap, ICE_FEATURE_SRIOV)) { + err = ice_iov_attach(sc); + if (err == ENOMEM) + return (err); + } +#endif /* PCI_IOV */ + /* Get DCBX/LLDP state and start DCBX agent */ ice_init_dcb_setup(sc); @@ -953,6 +984,11 @@ ice_if_detach(if_ctx_t ctx) ice_destroy_mirror_interface(sc); ice_rdma_pf_detach(sc); +#ifdef PCI_IOV + if (ice_is_bit_set(sc->feat_cap, ICE_FEATURE_SRIOV)) + ice_iov_detach(sc); +#endif /* PCI_IOV */ + /* Free allocated media types */ ifmedia_removeall(sc->media); @@ -1676,6 +1712,11 @@ ice_if_msix_intr_assign(if_ctx_t ctx, int msix) /* For future interrupt assignments */ sc->last_rid = rid + sc->irdma_vectors; +#ifdef PCI_IOV + /* Create soft IRQ for handling VF resets */ + iflib_softirq_alloc_generic(ctx, NULL, IFLIB_INTR_IOV, sc, 0, "iov"); +#endif + return (0); fail: for (; i >= 0; i--, vector--) @@ -2277,7 +2318,12 @@ ice_transition_recovery_mode(struct ice_softc *sc) ice_rdma_pf_detach(sc); ice_clear_bit(ICE_FEATURE_RDMA, sc->feat_cap); +#ifdef PCI_IOV + if (ice_test_and_clear_bit(ICE_FEATURE_SRIOV, sc->feat_en)) + ice_iov_detach(sc); +#else ice_clear_bit(ICE_FEATURE_SRIOV, sc->feat_en); +#endif /* PCI_IOV */ ice_clear_bit(ICE_FEATURE_SRIOV, sc->feat_cap); ice_vsi_del_txqs_ctx(vsi); @@ -2325,7 +2371,12 @@ ice_transition_safe_mode(struct ice_softc *sc) ice_rdma_pf_detach(sc); ice_clear_bit(ICE_FEATURE_RDMA, sc->feat_cap); +#ifdef PCI_IOV + if (ice_test_and_clear_bit(ICE_FEATURE_SRIOV, sc->feat_en)) + ice_iov_detach(sc); +#else ice_clear_bit(ICE_FEATURE_SRIOV, sc->feat_en); +#endif /* PCI_IOV */ ice_clear_bit(ICE_FEATURE_SRIOV, sc->feat_cap); ice_clear_bit(ICE_FEATURE_RSS, sc->feat_cap); @@ -2410,6 +2461,15 @@ ice_if_update_admin_status(if_ctx_t ctx) /* Check and update link status */ ice_update_link_status(sc, false); +#ifdef PCI_IOV + /* + * Schedule VFs' reset handler after global resets + * and other events were processed. + */ + if (ice_testandclear_state(&sc->state, ICE_STATE_VFLR_PENDING)) + iflib_iov_intr_deferred(ctx); +#endif + /* * If there are still messages to process, we need to reschedule * ourselves. Otherwise, we can just re-enable the interrupt. We'll be @@ -3349,6 +3409,78 @@ ice_init_link(struct ice_softc *sc) } +#ifdef PCI_IOV +/** + * ice_if_iov_init - iov init handler for iflib + * @ctx: iflib context pointer + * @num_vfs: number of VFs to create + * @params: configuration parameters for the PF + * + * Configure the driver for SR-IOV mode. Used to setup things like memory + * before any VFs are created. + * + * @remark This is a wrapper for ice_iov_init + */ +static int +ice_if_iov_init(if_ctx_t ctx, uint16_t num_vfs, const nvlist_t *params) +{ + struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); + + return ice_iov_init(sc, num_vfs, params); +} + +/** + * ice_if_iov_uninit - iov uninit handler for iflib + * @ctx: iflib context pointer + * + * Destroys VFs and frees their memory and resources. + * + * @remark This is a wrapper for ice_iov_uninit + */ +static void +ice_if_iov_uninit(if_ctx_t ctx) +{ + struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); + + ice_iov_uninit(sc); +} + +/** + * ice_if_iov_vf_add - iov add vf handler for iflib + * @ctx: iflib context pointer + * @vfnum: index of VF to configure + * @params: configuration parameters for the VF + * + * Sets up the VF given by the vfnum index. This is called by the OS + * for each VF created by the PF driver after it is spawned. + * + * @remark This is a wrapper for ice_iov_vf_add + */ +static int +ice_if_iov_vf_add(if_ctx_t ctx, uint16_t vfnum, const nvlist_t *params) +{ + struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); + + return ice_iov_add_vf(sc, vfnum, params); +} + +/** + * ice_if_vflr_handle - iov VFLR handler + * @ctx: iflib context pointer + * + * Performs the necessar teardown or setup required for a VF after + * a VFLR is initiated. + * + * @remark This is a wrapper for ice_iov_handle_vflr + */ +static void +ice_if_vflr_handle(if_ctx_t ctx) +{ + struct ice_softc *sc = (struct ice_softc *)iflib_get_softc(ctx); + ice_iov_handle_vflr(sc); +} +#endif /* PCI_IOV */ + extern struct if_txrx ice_subif_txrx; /** diff --git a/sys/dev/iicbus/iichid.c b/sys/dev/iicbus/iichid.c index 9c0324a24685..3f1d7a0cefba 100644 --- a/sys/dev/iicbus/iichid.c +++ b/sys/dev/iicbus/iichid.c @@ -275,62 +275,36 @@ iichid_cmd_read(struct iichid_softc* sc, void *buf, iichid_size_t maxlen, * 6.1.3 - Retrieval of Input Reports * DEVICE returns the length (2 Bytes) and the entire Input Report. */ - uint8_t actbuf[2] = { 0, 0 }; - /* Read actual input report length. */ + + memset(buf, 0xaa, 2); // In case nothing gets read struct iic_msg msgs[] = { - { sc->addr, IIC_M_RD | IIC_M_NOSTOP, sizeof(actbuf), actbuf }, + { sc->addr, IIC_M_RD, maxlen, buf }, }; - uint16_t actlen; int error; error = iicbus_transfer(sc->dev, msgs, nitems(msgs)); if (error != 0) return (error); - actlen = actbuf[0] | actbuf[1] << 8; -#ifdef IICHID_SAMPLING - if ((actlen == 0 && sc->sampling_rate_slow < 0) || - (maxlen == 0 && sc->sampling_rate_slow >= 0)) { -#else + DPRINTFN(sc, 5, "%*D\n", msgs[0].len, msgs[0].buf, " "); + + uint16_t actlen = le16dec(buf); + if (actlen == 0) { -#endif - /* Read and discard reset command response. */ - msgs[0] = (struct iic_msg) - { sc->addr, IIC_M_RD | IIC_M_NOSTART, - le16toh(sc->desc.wMaxInputLength) - 2, sc->intr_buf }; - actlen = 0; if (!sc->reset_acked) { mtx_lock(&sc->mtx); sc->reset_acked = true; wakeup(&sc->reset_acked); mtx_unlock(&sc->mtx); } -#ifdef IICHID_SAMPLING - } else if ((actlen <= 2 || actlen == 0xFFFF) && - sc->sampling_rate_slow >= 0) { - /* Read and discard 1 byte to send I2C STOP condition. */ - msgs[0] = (struct iic_msg) - { sc->addr, IIC_M_RD | IIC_M_NOSTART, 1, actbuf }; - actlen = 0; -#endif - } else { - actlen -= 2; - if (actlen > maxlen) { - DPRINTF(sc, "input report too big. requested=%d " - "received=%d\n", maxlen, actlen); - actlen = maxlen; - } - /* Read input report itself. */ - msgs[0] = (struct iic_msg) - { sc->addr, IIC_M_RD | IIC_M_NOSTART, actlen, buf }; } - error = iicbus_transfer(sc->dev, msgs, 1); - if (error == 0 && actual_len != NULL) + if (actlen <= 2 || actlen > maxlen) { + actlen = 0; + } + if (actual_len != NULL) { *actual_len = actlen; - - DPRINTFN(sc, 5, - "%*D - %*D\n", 2, actbuf, " ", msgs[0].len, msgs[0].buf, " "); + } return (error); } @@ -566,7 +540,7 @@ iichid_sampling_task(void *context, int pending) error = iichid_cmd_read(sc, sc->intr_buf, sc->intr_bufsize, &actual); if (error == 0) { if (actual > 0) { - sc->intr_handler(sc->intr_ctx, sc->intr_buf, actual); + sc->intr_handler(sc->intr_ctx, sc->intr_buf + 2, actual); sc->missing_samples = 0; if (sc->dup_size != actual || memcmp(sc->dup_buf, sc->intr_buf, actual) != 0) { @@ -577,7 +551,7 @@ iichid_sampling_task(void *context, int pending) ++sc->dup_samples; } else { if (++sc->missing_samples == 1) - sc->intr_handler(sc->intr_ctx, sc->intr_buf, 0); + sc->intr_handler(sc->intr_ctx, sc->intr_buf + 2, 0); sc->dup_samples = 0; } } else @@ -632,7 +606,7 @@ iichid_intr(void *context) if (error == 0) { if (sc->power_on && sc->open) { if (actual != 0) - sc->intr_handler(sc->intr_ctx, sc->intr_buf, + sc->intr_handler(sc->intr_ctx, sc->intr_buf + 2, actual); else DPRINTF(sc, "no data received\n"); @@ -842,11 +816,12 @@ iichid_intr_setup(device_t dev, device_t child __unused, hid_intr_t intr, sc = device_get_softc(dev); /* - * Do not rely on wMaxInputLength, as some devices may set it to - * a wrong length. Find the longest input report in report descriptor. + * Do not rely just on wMaxInputLength, as some devices (which?) + * may set it to a wrong length. Also find the longest input report + * in report descriptor, and add two for the length field. */ - rdesc->rdsize = - MAX(rdesc->isize, le16toh(sc->desc.wMaxInputLength) - 2); + rdesc->rdsize = 2 + + MAX(rdesc->isize, le16toh(sc->desc.wMaxInputLength)); /* Write and get/set_report sizes are limited by I2C-HID protocol. */ rdesc->grsize = rdesc->srsize = IICHID_SIZE_MAX; rdesc->wrsize = IICHID_SIZE_MAX; @@ -919,7 +894,7 @@ iichid_intr_poll(device_t dev, device_t child __unused) sc = device_get_softc(dev); error = iichid_cmd_read(sc, sc->intr_buf, sc->intr_bufsize, &actual); if (error == 0 && actual != 0) - sc->intr_handler(sc->intr_ctx, sc->intr_buf, actual); + sc->intr_handler(sc->intr_ctx, sc->intr_buf + 2, actual); } /* @@ -946,6 +921,7 @@ iichid_read(device_t dev, device_t child __unused, void *buf, { struct iichid_softc *sc; device_t parent; + uint8_t *tmpbuf; int error; if (maxlen > IICHID_SIZE_MAX) @@ -954,8 +930,12 @@ iichid_read(device_t dev, device_t child __unused, void *buf, parent = device_get_parent(sc->dev); error = iicbus_request_bus(parent, sc->dev, IIC_WAIT); if (error == 0) { - error = iichid_cmd_read(sc, buf, maxlen, actlen); + tmpbuf = malloc(maxlen + 2, M_DEVBUF, M_WAITOK | M_ZERO); + error = iichid_cmd_read(sc, tmpbuf, maxlen + 2, actlen); iicbus_release_bus(parent, sc->dev); + if (*actlen > 0) + memcpy(buf, tmpbuf + 2, *actlen); + free(tmpbuf, M_DEVBUF); } return (iic2errno(error)); } diff --git a/sys/dev/md/md.c b/sys/dev/md/md.c index 741a7c013f7d..ec1664fac701 100644 --- a/sys/dev/md/md.c +++ b/sys/dev/md/md.c @@ -11,9 +11,9 @@ */ /*- - * The following functions are based on the vn(4) driver: mdstart_swap(), - * mdstart_vnode(), mdcreate_swap(), mdcreate_vnode() and mddestroy(), - * and as such under the following copyright: + * The following functions are based on the historical vn(4) driver: + * mdstart_swap(), mdstart_vnode(), mdcreate_swap(), mdcreate_vnode() + * and mddestroy(), and as such under the following copyright: * * Copyright (c) 1988 University of Utah. * Copyright (c) 1990, 1993 @@ -89,6 +89,8 @@ #include <sys/unistd.h> #include <sys/vnode.h> #include <sys/disk.h> +#include <sys/param.h> +#include <sys/bus.h> #include <geom/geom.h> #include <geom/geom_int.h> @@ -2082,8 +2084,10 @@ g_md_init(struct g_class *mp __unused) { caddr_t mod; u_char *ptr, *name, *type; + u_char scratch[40]; unsigned len; int i; + vm_offset_t paddr; /* figure out log2(NINDIR) */ for (i = NINDIR, nshift = -1; i; nshift++) @@ -2123,6 +2127,25 @@ g_md_init(struct g_class *mp __unused) sx_xunlock(&md_sx); } } + + /* + * Load up to 32 pre-loaded disks + */ + for (int i = 0; i < 32; i++) { + if (resource_long_value("md", i, "physaddr", + (long *) &paddr) != 0 || + resource_int_value("md", i, "len", &len) != 0) + break; + ptr = (char *)pmap_map(NULL, paddr, paddr + len, VM_PROT_READ); + if (ptr != NULL && len != 0) { + sprintf(scratch, "preload%d 0x%016jx", i, + (uintmax_t)paddr); + sx_xlock(&md_sx); + md_preloaded(ptr, len, scratch); + sx_xunlock(&md_sx); + } + } + status_dev = make_dev(&mdctl_cdevsw, INT_MAX, UID_ROOT, GID_WHEEL, 0600, MDCTL_NAME); g_topology_lock(); diff --git a/sys/dev/mgb/if_mgb.c b/sys/dev/mgb/if_mgb.c index 1240d0f84415..409f34167df0 100644 --- a/sys/dev/mgb/if_mgb.c +++ b/sys/dev/mgb/if_mgb.c @@ -1435,7 +1435,7 @@ mgb_hw_teardown(struct mgb_softc *sc) /* Stop MAC */ CSR_CLEAR_REG(sc, MGB_MAC_RX, MGB_MAC_ENBL); - CSR_WRITE_REG(sc, MGB_MAC_TX, MGB_MAC_ENBL); + CSR_CLEAR_REG(sc, MGB_MAC_TX, MGB_MAC_ENBL); if ((err = mgb_wait_for_bits(sc, MGB_MAC_RX, MGB_MAC_DSBL, 0))) return (err); if ((err = mgb_wait_for_bits(sc, MGB_MAC_TX, MGB_MAC_DSBL, 0))) diff --git a/sys/dev/mlx5/mlx5_accel/ipsec.h b/sys/dev/mlx5/mlx5_accel/ipsec.h index 361b9f72d873..c3f3a2372482 100644 --- a/sys/dev/mlx5/mlx5_accel/ipsec.h +++ b/sys/dev/mlx5/mlx5_accel/ipsec.h @@ -260,8 +260,8 @@ int mlx5e_accel_ipsec_fs_rx_tables_create(struct mlx5e_priv *priv); void mlx5e_accel_ipsec_fs_rx_catchall_rules_destroy(struct mlx5e_priv *priv); int mlx5e_accel_ipsec_fs_rx_catchall_rules(struct mlx5e_priv *priv); int mlx5_accel_ipsec_rx_tag_add(if_t ifp, struct mlx5e_rq_mbuf *mr); -void mlx5e_accel_ipsec_handle_rx_cqe(struct mbuf *mb, struct mlx5_cqe64 *cqe, - struct mlx5e_rq_mbuf *mr); +void mlx5e_accel_ipsec_handle_rx_cqe(if_t ifp, struct mbuf *mb, + struct mlx5_cqe64 *cqe, struct mlx5e_rq_mbuf *mr); static inline int mlx5e_accel_ipsec_flow(struct mlx5_cqe64 *cqe) { @@ -269,12 +269,12 @@ static inline int mlx5e_accel_ipsec_flow(struct mlx5_cqe64 *cqe) } static inline void -mlx5e_accel_ipsec_handle_rx(struct mbuf *mb, struct mlx5_cqe64 *cqe, +mlx5e_accel_ipsec_handle_rx(if_t ifp, struct mbuf *mb, struct mlx5_cqe64 *cqe, struct mlx5e_rq_mbuf *mr) { u32 ipsec_meta_data = be32_to_cpu(cqe->ft_metadata); if (MLX5_IPSEC_METADATA_MARKER(ipsec_meta_data)) - mlx5e_accel_ipsec_handle_rx_cqe(mb, cqe, mr); + mlx5e_accel_ipsec_handle_rx_cqe(ifp, mb, cqe, mr); } #endif /* __MLX5_ACCEL_IPSEC_H__ */ diff --git a/sys/dev/mlx5/mlx5_accel/mlx5_ipsec_rxtx.c b/sys/dev/mlx5/mlx5_accel/mlx5_ipsec_rxtx.c index 0883cfb2d510..5dccb8bc2b87 100644 --- a/sys/dev/mlx5/mlx5_accel/mlx5_ipsec_rxtx.c +++ b/sys/dev/mlx5/mlx5_accel/mlx5_ipsec_rxtx.c @@ -24,11 +24,14 @@ * */ +#include "opt_ipsec.h" + #include <sys/mbuf.h> #include <sys/socket.h> #include <netinet/in.h> #include <netipsec/keydb.h> #include <netipsec/ipsec_offload.h> +#include <netipsec/xform.h> #include <dev/mlx5/qp.h> #include <dev/mlx5/mlx5_en/en.h> #include <dev/mlx5/mlx5_accel/ipsec.h> @@ -48,7 +51,8 @@ mlx5_accel_ipsec_rx_tag_add(if_t ifp, struct mlx5e_rq_mbuf *mr) return (0); mtag = (struct ipsec_accel_in_tag *)m_tag_get( - PACKET_TAG_IPSEC_ACCEL_IN, sizeof(*mtag), M_NOWAIT); + PACKET_TAG_IPSEC_ACCEL_IN, sizeof(struct ipsec_accel_in_tag) - + __offsetof(struct ipsec_accel_in_tag, xh), M_NOWAIT); if (mtag == NULL) return (-ENOMEM); mr->ipsec_mtag = mtag; @@ -56,8 +60,8 @@ mlx5_accel_ipsec_rx_tag_add(if_t ifp, struct mlx5e_rq_mbuf *mr) } void -mlx5e_accel_ipsec_handle_rx_cqe(struct mbuf *mb, struct mlx5_cqe64 *cqe, - struct mlx5e_rq_mbuf *mr) +mlx5e_accel_ipsec_handle_rx_cqe(if_t ifp, struct mbuf *mb, + struct mlx5_cqe64 *cqe, struct mlx5e_rq_mbuf *mr) { struct ipsec_accel_in_tag *mtag; u32 drv_spi; @@ -65,10 +69,12 @@ mlx5e_accel_ipsec_handle_rx_cqe(struct mbuf *mb, struct mlx5_cqe64 *cqe, drv_spi = MLX5_IPSEC_METADATA_HANDLE(be32_to_cpu(cqe->ft_metadata)); mtag = mr->ipsec_mtag; WARN_ON(mtag == NULL); - mr->ipsec_mtag = NULL; if (mtag != NULL) { mtag->drv_spi = drv_spi; - m_tag_prepend(mb, &mtag->tag); + if (ipsec_accel_fill_xh(ifp, drv_spi, &mtag->xh)) { + m_tag_prepend(mb, &mtag->tag); + mr->ipsec_mtag = NULL; + } } } diff --git a/sys/dev/mlx5/mlx5_en/mlx5_en_hw_tls_rx.c b/sys/dev/mlx5/mlx5_en/mlx5_en_hw_tls_rx.c index 8b8f2e570245..89d2010656c5 100644 --- a/sys/dev/mlx5/mlx5_en/mlx5_en_hw_tls_rx.c +++ b/sys/dev/mlx5/mlx5_en/mlx5_en_hw_tls_rx.c @@ -42,13 +42,30 @@ static if_snd_tag_free_t mlx5e_tls_rx_snd_tag_free; static if_snd_tag_modify_t mlx5e_tls_rx_snd_tag_modify; +static if_snd_tag_status_str_t mlx5e_tls_rx_snd_tag_status_str; static const struct if_snd_tag_sw mlx5e_tls_rx_snd_tag_sw = { .snd_tag_modify = mlx5e_tls_rx_snd_tag_modify, .snd_tag_free = mlx5e_tls_rx_snd_tag_free, + .snd_tag_status_str = mlx5e_tls_rx_snd_tag_status_str, .type = IF_SND_TAG_TYPE_TLS_RX }; +static const char *mlx5e_tls_rx_progress_params_auth_state_str[] = { + [MLX5E_TLS_RX_PROGRESS_PARAMS_AUTH_STATE_NO_OFFLOAD] = "no_offload", + [MLX5E_TLS_RX_PROGRESS_PARAMS_AUTH_STATE_OFFLOAD] = "offload", + [MLX5E_TLS_RX_PROGRESS_PARAMS_AUTH_STATE_AUTHENTICATION] = + "authentication", +}; + +static const char *mlx5e_tls_rx_progress_params_record_tracker_state_str[] = { + [MLX5E_TLS_RX_PROGRESS_PARAMS_RECORD_TRACKER_STATE_START] = "start", + [MLX5E_TLS_RX_PROGRESS_PARAMS_RECORD_TRACKER_STATE_TRACKING] = + "tracking", + [MLX5E_TLS_RX_PROGRESS_PARAMS_RECORD_TRACKER_STATE_SEARCHING] = + "searching", +}; + MALLOC_DEFINE(M_MLX5E_TLS_RX, "MLX5E_TLS_RX", "MLX5 ethernet HW TLS RX"); /* software TLS RX context */ @@ -250,7 +267,8 @@ mlx5e_tls_rx_send_progress_parameters_sync(struct mlx5e_iq *iq, mtx_unlock(&iq->lock); while (1) { - if (wait_for_completion_timeout(&ptag->progress_complete, hz) != 0) + if (wait_for_completion_timeout(&ptag->progress_complete, + msecs_to_jiffies(1000)) != 0) break; priv = container_of(iq, struct mlx5e_channel, iq)->priv; if (priv->mdev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR || @@ -331,7 +349,8 @@ done: * Zero is returned upon success, else some error happened. */ static int -mlx5e_tls_rx_receive_progress_parameters(struct mlx5e_iq *iq, struct mlx5e_tls_rx_tag *ptag) +mlx5e_tls_rx_receive_progress_parameters(struct mlx5e_iq *iq, + struct mlx5e_tls_rx_tag *ptag, mlx5e_iq_callback_t *cb) { struct mlx5e_get_tls_progress_params_wqe *wqe; const u32 ds_cnt = DIV_ROUND_UP(sizeof(*wqe), MLX5_SEND_WQE_DS); @@ -367,7 +386,7 @@ mlx5e_tls_rx_receive_progress_parameters(struct mlx5e_iq *iq, struct mlx5e_tls_r memcpy(iq->doorbell.d32, &wqe->ctrl, sizeof(iq->doorbell.d32)); iq->data[pi].num_wqebbs = DIV_ROUND_UP(ds_cnt, MLX5_SEND_WQEBB_NUM_DS); - iq->data[pi].callback = &mlx5e_tls_rx_receive_progress_parameters_cb; + iq->data[pi].callback = cb; iq->data[pi].arg = ptag; m_snd_tag_ref(&ptag->tag); @@ -640,7 +659,8 @@ mlx5e_tls_rx_set_params(void *ctx, struct inpcb *inp, const struct tls_session_p return (EINVAL); MLX5_SET64(sw_tls_rx_cntx, ctx, param.initial_record_number, tls_sn_he); - MLX5_SET(sw_tls_rx_cntx, ctx, param.resync_tcp_sn, tcp_sn_he); + MLX5_SET(sw_tls_rx_cntx, ctx, param.resync_tcp_sn, 0); + MLX5_SET(sw_tls_rx_cntx, ctx, progress.next_record_tcp_sn, tcp_sn_he); return (0); } @@ -819,6 +839,7 @@ mlx5e_tls_rx_snd_tag_alloc(if_t ifp, } ptag->flow_rule = flow_rule; + init_completion(&ptag->progress_complete); return (0); @@ -968,7 +989,8 @@ mlx5e_tls_rx_snd_tag_modify(struct m_snd_tag *pmt, union if_snd_tag_modify_param params->tls_rx.tls_rec_length, params->tls_rx.tls_seq_number) && ptag->tcp_resync_pending == 0) { - err = mlx5e_tls_rx_receive_progress_parameters(iq, ptag); + err = mlx5e_tls_rx_receive_progress_parameters(iq, ptag, + &mlx5e_tls_rx_receive_progress_parameters_cb); if (err != 0) { MLX5E_TLS_RX_STAT_INC(ptag, rx_resync_err, 1); } else { @@ -1001,6 +1023,74 @@ mlx5e_tls_rx_snd_tag_free(struct m_snd_tag *pmt) queue_work(priv->tls_rx.wq, &ptag->work); } +static void +mlx5e_tls_rx_str_status_cb(void *arg) +{ + struct mlx5e_tls_rx_tag *ptag; + + ptag = (struct mlx5e_tls_rx_tag *)arg; + complete_all(&ptag->progress_complete); + m_snd_tag_rele(&ptag->tag); +} + +static int +mlx5e_tls_rx_snd_tag_status_str(struct m_snd_tag *pmt, char *buf, size_t *sz) +{ + int err, out_size; + struct mlx5e_iq *iq; + void *buffer; + uint32_t tracker_state_val; + uint32_t auth_state_val; + struct mlx5e_priv *priv; + struct mlx5e_tls_rx_tag *ptag = + container_of(pmt, struct mlx5e_tls_rx_tag, tag); + + if (buf == NULL) + return (0); + + MLX5E_TLS_RX_TAG_LOCK(ptag); + priv = container_of(ptag->tls_rx, struct mlx5e_priv, tls_rx); + iq = mlx5e_tls_rx_get_iq(priv, ptag->flowid, ptag->flowtype); + reinit_completion(&ptag->progress_complete); + err = mlx5e_tls_rx_receive_progress_parameters(iq, ptag, + &mlx5e_tls_rx_str_status_cb); + MLX5E_TLS_RX_TAG_UNLOCK(ptag); + if (err != 0) + return (err); + + for (;;) { + if (wait_for_completion_timeout(&ptag->progress_complete, + msecs_to_jiffies(1000)) != 0) + break; + if (priv->mdev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR || + pci_channel_offline(priv->mdev->pdev) != 0) + return (ENXIO); + } + buffer = mlx5e_tls_rx_get_progress_buffer(ptag); + tracker_state_val = MLX5_GET(tls_progress_params, buffer, + record_tracker_state); + auth_state_val = MLX5_GET(tls_progress_params, buffer, auth_state); + + /* Validate tracker state value is in range */ + if (tracker_state_val > + MLX5E_TLS_RX_PROGRESS_PARAMS_RECORD_TRACKER_STATE_SEARCHING) + return (EINVAL); + + /* Validate auth state value is in range */ + if (auth_state_val > + MLX5E_TLS_RX_PROGRESS_PARAMS_AUTH_STATE_AUTHENTICATION) + return (EINVAL); + + out_size = snprintf(buf, *sz, "tracker_state: %s, auth_state: %s", + mlx5e_tls_rx_progress_params_record_tracker_state_str[ + tracker_state_val], + mlx5e_tls_rx_progress_params_auth_state_str[auth_state_val]); + + if (out_size <= *sz) + *sz = out_size; + return (0); +} + #else int diff --git a/sys/dev/mlx5/mlx5_en/mlx5_en_rx.c b/sys/dev/mlx5/mlx5_en/mlx5_en_rx.c index 6b53db6fea23..eb569488631a 100644 --- a/sys/dev/mlx5/mlx5_en/mlx5_en_rx.c +++ b/sys/dev/mlx5/mlx5_en/mlx5_en_rx.c @@ -467,7 +467,7 @@ mlx5e_build_rx_mbuf(struct mlx5_cqe64 *cqe, struct mlx5e_rq *rq, break; } - mlx5e_accel_ipsec_handle_rx(mb, cqe, mr); + mlx5e_accel_ipsec_handle_rx(ifp, mb, cqe, mr); } static inline void diff --git a/sys/dev/nvme/nvme_ctrlr.c b/sys/dev/nvme/nvme_ctrlr.c index 73a7cee4aad0..fd7f00ced14b 100644 --- a/sys/dev/nvme/nvme_ctrlr.c +++ b/sys/dev/nvme/nvme_ctrlr.c @@ -48,7 +48,7 @@ #define B4_CHK_RDY_DELAY_MS 2300 /* work around controller bug */ static void nvme_ctrlr_construct_and_submit_aer(struct nvme_controller *ctrlr, - struct nvme_async_event_request *aer); + struct nvme_async_event_request *aer); static void nvme_ctrlr_barrier(struct nvme_controller *ctrlr, int flags) @@ -680,96 +680,6 @@ nvme_ctrlr_log_critical_warnings(struct nvme_controller *ctrlr, } static void -nvme_ctrlr_async_event_log_page_cb(void *arg, const struct nvme_completion *cpl) -{ - struct nvme_async_event_request *aer = arg; - struct nvme_health_information_page *health_info; - struct nvme_ns_list *nsl; - struct nvme_error_information_entry *err; - int i; - - /* - * If the log page fetch for some reason completed with an error, - * don't pass log page data to the consumers. In practice, this case - * should never happen. - */ - if (nvme_completion_is_error(cpl)) - nvme_notify_async_consumers(aer->ctrlr, &aer->cpl, - aer->log_page_id, NULL, 0); - else { - /* Convert data to host endian */ - switch (aer->log_page_id) { - case NVME_LOG_ERROR: - err = (struct nvme_error_information_entry *)aer->log_page_buffer; - for (i = 0; i < (aer->ctrlr->cdata.elpe + 1); i++) - nvme_error_information_entry_swapbytes(err++); - break; - case NVME_LOG_HEALTH_INFORMATION: - nvme_health_information_page_swapbytes( - (struct nvme_health_information_page *)aer->log_page_buffer); - break; - case NVME_LOG_CHANGED_NAMESPACE: - nvme_ns_list_swapbytes( - (struct nvme_ns_list *)aer->log_page_buffer); - break; - case NVME_LOG_COMMAND_EFFECT: - nvme_command_effects_page_swapbytes( - (struct nvme_command_effects_page *)aer->log_page_buffer); - break; - case NVME_LOG_RES_NOTIFICATION: - nvme_res_notification_page_swapbytes( - (struct nvme_res_notification_page *)aer->log_page_buffer); - break; - case NVME_LOG_SANITIZE_STATUS: - nvme_sanitize_status_page_swapbytes( - (struct nvme_sanitize_status_page *)aer->log_page_buffer); - break; - default: - break; - } - - if (aer->log_page_id == NVME_LOG_HEALTH_INFORMATION) { - health_info = (struct nvme_health_information_page *) - aer->log_page_buffer; - nvme_ctrlr_log_critical_warnings(aer->ctrlr, - health_info->critical_warning); - /* - * Critical warnings reported through the - * SMART/health log page are persistent, so - * clear the associated bits in the async event - * config so that we do not receive repeated - * notifications for the same event. - */ - aer->ctrlr->async_event_config &= - ~health_info->critical_warning; - nvme_ctrlr_cmd_set_async_event_config(aer->ctrlr, - aer->ctrlr->async_event_config, NULL, NULL); - } else if (aer->log_page_id == NVME_LOG_CHANGED_NAMESPACE && - !nvme_use_nvd) { - nsl = (struct nvme_ns_list *)aer->log_page_buffer; - for (i = 0; i < nitems(nsl->ns) && nsl->ns[i] != 0; i++) { - if (nsl->ns[i] > NVME_MAX_NAMESPACES) - break; - nvme_notify_ns(aer->ctrlr, nsl->ns[i]); - } - } - - /* - * Pass the cpl data from the original async event completion, - * not the log page fetch. - */ - nvme_notify_async_consumers(aer->ctrlr, &aer->cpl, - aer->log_page_id, aer->log_page_buffer, aer->log_page_size); - } - - /* - * Repost another asynchronous event request to replace the one - * that just completed. - */ - nvme_ctrlr_construct_and_submit_aer(aer->ctrlr, aer); -} - -static void nvme_ctrlr_async_event_cb(void *arg, const struct nvme_completion *cpl) { struct nvme_async_event_request *aer = arg; @@ -784,33 +694,18 @@ nvme_ctrlr_async_event_cb(void *arg, const struct nvme_completion *cpl) return; } - /* Associated log page is in bits 23:16 of completion entry dw0. */ + /* + * Save the completion status and associated log page is in bits 23:16 + * of completion entry dw0. Print a message and queue it for further + * processing. + */ + memcpy(&aer->cpl, cpl, sizeof(*cpl)); aer->log_page_id = NVMEV(NVME_ASYNC_EVENT_LOG_PAGE_ID, cpl->cdw0); - nvme_printf(aer->ctrlr, "async event occurred (type 0x%x, info 0x%02x," " page 0x%02x)\n", NVMEV(NVME_ASYNC_EVENT_TYPE, cpl->cdw0), NVMEV(NVME_ASYNC_EVENT_INFO, cpl->cdw0), aer->log_page_id); - - if (is_log_page_id_valid(aer->log_page_id)) { - aer->log_page_size = nvme_ctrlr_get_log_page_size(aer->ctrlr, - aer->log_page_id); - memcpy(&aer->cpl, cpl, sizeof(*cpl)); - nvme_ctrlr_cmd_get_log_page(aer->ctrlr, aer->log_page_id, - NVME_GLOBAL_NAMESPACE_TAG, aer->log_page_buffer, - aer->log_page_size, nvme_ctrlr_async_event_log_page_cb, - aer); - /* Wait to notify consumers until after log page is fetched. */ - } else { - nvme_notify_async_consumers(aer->ctrlr, cpl, aer->log_page_id, - NULL, 0); - - /* - * Repost another asynchronous event request to replace the one - * that just completed. - */ - nvme_ctrlr_construct_and_submit_aer(aer->ctrlr, aer); - } + taskqueue_enqueue(aer->ctrlr->taskqueue, &aer->task); } static void @@ -819,15 +714,21 @@ nvme_ctrlr_construct_and_submit_aer(struct nvme_controller *ctrlr, { struct nvme_request *req; - aer->ctrlr = ctrlr; /* - * XXX-MJ this should be M_WAITOK but we might be in a non-sleepable - * callback context. AER completions should be handled on a dedicated - * thread. + * We're racing the reset thread, so let that process submit this again. + * XXX does this really solve that race? And is that race even possible + * since we only reset when we've no theard from the card in a long + * time. Why would we get an AER in the middle of that just before we + * kick off the reset? */ - req = nvme_allocate_request_null(M_NOWAIT, nvme_ctrlr_async_event_cb, + if (ctrlr->is_resetting) + return; + + aer->ctrlr = ctrlr; + req = nvme_allocate_request_null(M_WAITOK, nvme_ctrlr_async_event_cb, aer); aer->req = req; + aer->log_page_id = 0; /* Not a valid page */ /* * Disable timeout here, since asynchronous event requests should by @@ -1203,6 +1104,140 @@ nvme_ctrlr_reset_task(void *arg, int pending) atomic_cmpset_32(&ctrlr->is_resetting, 1, 0); } +static void +nvme_ctrlr_aer_done(void *arg, const struct nvme_completion *cpl) +{ + struct nvme_async_event_request *aer = arg; + + mtx_lock(&aer->mtx); + if (nvme_completion_is_error(cpl)) + aer->log_page_size = (uint32_t)-1; + else + aer->log_page_size = nvme_ctrlr_get_log_page_size( + aer->ctrlr, aer->log_page_id); + wakeup(aer); + mtx_unlock(&aer->mtx); +} + +static void +nvme_ctrlr_aer_task(void *arg, int pending) +{ + struct nvme_async_event_request *aer = arg; + struct nvme_controller *ctrlr = aer->ctrlr; + uint32_t len; + + /* + * We're resetting, so just punt. + */ + if (ctrlr->is_resetting) + return; + + if (!is_log_page_id_valid(aer->log_page_id)) { + /* + * Repost another asynchronous event request to replace the one + * that just completed. + */ + nvme_notify_async_consumers(ctrlr, &aer->cpl, aer->log_page_id, + NULL, 0); + nvme_ctrlr_construct_and_submit_aer(ctrlr, aer); + goto out; + } + + aer->log_page_size = 0; + len = nvme_ctrlr_get_log_page_size(aer->ctrlr, aer->log_page_id); + nvme_ctrlr_cmd_get_log_page(aer->ctrlr, aer->log_page_id, + NVME_GLOBAL_NAMESPACE_TAG, aer->log_page_buffer, len, + nvme_ctrlr_aer_done, aer); + mtx_lock(&aer->mtx); + while (aer->log_page_size == 0) + mtx_sleep(aer, &aer->mtx, PRIBIO, "nvme_pt", 0); + mtx_unlock(&aer->mtx); + + if (aer->log_page_size != (uint32_t)-1) { + /* + * If the log page fetch for some reason completed with an + * error, don't pass log page data to the consumers. In + * practice, this case should never happen. + */ + nvme_notify_async_consumers(aer->ctrlr, &aer->cpl, + aer->log_page_id, NULL, 0); + goto out; + } + + /* Convert data to host endian */ + switch (aer->log_page_id) { + case NVME_LOG_ERROR: { + struct nvme_error_information_entry *err = + (struct nvme_error_information_entry *)aer->log_page_buffer; + for (int i = 0; i < (aer->ctrlr->cdata.elpe + 1); i++) + nvme_error_information_entry_swapbytes(err++); + break; + } + case NVME_LOG_HEALTH_INFORMATION: + nvme_health_information_page_swapbytes( + (struct nvme_health_information_page *)aer->log_page_buffer); + break; + case NVME_LOG_CHANGED_NAMESPACE: + nvme_ns_list_swapbytes( + (struct nvme_ns_list *)aer->log_page_buffer); + break; + case NVME_LOG_COMMAND_EFFECT: + nvme_command_effects_page_swapbytes( + (struct nvme_command_effects_page *)aer->log_page_buffer); + break; + case NVME_LOG_RES_NOTIFICATION: + nvme_res_notification_page_swapbytes( + (struct nvme_res_notification_page *)aer->log_page_buffer); + break; + case NVME_LOG_SANITIZE_STATUS: + nvme_sanitize_status_page_swapbytes( + (struct nvme_sanitize_status_page *)aer->log_page_buffer); + break; + default: + break; + } + + if (aer->log_page_id == NVME_LOG_HEALTH_INFORMATION) { + struct nvme_health_information_page *health_info = + (struct nvme_health_information_page *)aer->log_page_buffer; + + /* + * Critical warnings reported through the SMART/health log page + * are persistent, so clear the associated bits in the async + * event config so that we do not receive repeated notifications + * for the same event. + */ + nvme_ctrlr_log_critical_warnings(aer->ctrlr, + health_info->critical_warning); + aer->ctrlr->async_event_config &= + ~health_info->critical_warning; + nvme_ctrlr_cmd_set_async_event_config(aer->ctrlr, + aer->ctrlr->async_event_config, NULL, NULL); + } else if (aer->log_page_id == NVME_LOG_CHANGED_NAMESPACE) { + struct nvme_ns_list *nsl = + (struct nvme_ns_list *)aer->log_page_buffer; + for (int i = 0; i < nitems(nsl->ns) && nsl->ns[i] != 0; i++) { + if (nsl->ns[i] > NVME_MAX_NAMESPACES) + break; + nvme_notify_ns(aer->ctrlr, nsl->ns[i]); + } + } + + /* + * Pass the cpl data from the original async event completion, not the + * log page fetch. + */ + nvme_notify_async_consumers(aer->ctrlr, &aer->cpl, + aer->log_page_id, aer->log_page_buffer, aer->log_page_size); + + /* + * Repost another asynchronous event request to replace the one + * that just completed. + */ +out: + nvme_ctrlr_construct_and_submit_aer(ctrlr, aer); +} + /* * Poll all the queues enabled on the device for completion. */ @@ -1574,13 +1609,8 @@ nvme_ctrlr_construct(struct nvme_controller *ctrlr, device_t dev) /* * Create 2 threads for the taskqueue. The reset thread will block when * it detects that the controller has failed until all I/O has been - * failed up the stack. The fail_req task needs to be able to run in - * this case to finish the request failure for some cases. - * - * We could partially solve this race by draining the failed requeust - * queue before proceding to free the sim, though nothing would stop - * new I/O from coming in after we do that drain, but before we reach - * cam_sim_free, so this big hammer is used instead. + * failed up the stack. The second thread is used for AER events, which + * can block, but only briefly for memory and log page fetching. */ ctrlr->taskqueue = taskqueue_create("nvme_taskq", M_WAITOK, taskqueue_thread_enqueue, &ctrlr->taskqueue); @@ -1590,7 +1620,12 @@ nvme_ctrlr_construct(struct nvme_controller *ctrlr, device_t dev) ctrlr->is_initialized = false; ctrlr->notification_sent = 0; TASK_INIT(&ctrlr->reset_task, 0, nvme_ctrlr_reset_task, ctrlr); - STAILQ_INIT(&ctrlr->fail_req); + for (int i = 0; i < NVME_MAX_ASYNC_EVENTS; i++) { + struct nvme_async_event_request *aer = &ctrlr->aer[i]; + + TASK_INIT(&aer->task, 0, nvme_ctrlr_aer_task, aer); + mtx_init(&aer->mtx, "AER mutex", NULL, MTX_DEF); + } ctrlr->is_failed = false; make_dev_args_init(&md_args); @@ -1678,8 +1713,14 @@ nvme_ctrlr_destruct(struct nvme_controller *ctrlr, device_t dev) } noadminq: - if (ctrlr->taskqueue) + if (ctrlr->taskqueue) { taskqueue_free(ctrlr->taskqueue); + for (int i = 0; i < NVME_MAX_ASYNC_EVENTS; i++) { + struct nvme_async_event_request *aer = &ctrlr->aer[i]; + + mtx_destroy(&aer->mtx); + } + } if (ctrlr->tag) bus_teardown_intr(ctrlr->dev, ctrlr->res, ctrlr->tag); diff --git a/sys/dev/nvme/nvme_private.h b/sys/dev/nvme/nvme_private.h index 949e69ec9290..36f00fedc48e 100644 --- a/sys/dev/nvme/nvme_private.h +++ b/sys/dev/nvme/nvme_private.h @@ -123,6 +123,8 @@ struct nvme_request { struct nvme_async_event_request { struct nvme_controller *ctrlr; struct nvme_request *req; + struct task task; + struct mtx mtx; struct nvme_completion cpl; uint32_t log_page_id; uint32_t log_page_size; @@ -307,8 +309,6 @@ struct nvme_controller { bool isr_warned; bool is_initialized; - STAILQ_HEAD(, nvme_request) fail_req; - /* Host Memory Buffer */ int hmb_nchunks; size_t hmb_chunk; diff --git a/sys/dev/nvmf/controller/nvmft_subr.c b/sys/dev/nvmf/controller/nvmft_subr.c index bb2bc0988e81..245971813854 100644 --- a/sys/dev/nvmf/controller/nvmft_subr.c +++ b/sys/dev/nvmf/controller/nvmft_subr.c @@ -26,46 +26,6 @@ nvmf_nqn_valid(const char *nqn) len = strnlen(nqn, NVME_NQN_FIELD_SIZE); if (len == 0 || len > NVMF_NQN_MAX_LEN) return (false); - -#ifdef STRICT_CHECKS - /* - * Stricter checks from the spec. Linux does not seem to - * require these. - */ - - /* - * NVMF_NQN_MIN_LEN does not include '.', and require at least - * one character of a domain name. - */ - if (len < NVMF_NQN_MIN_LEN + 2) - return (false); - if (memcmp("nqn.", nqn, strlen("nqn.")) != 0) - return (false); - nqn += strlen("nqn."); - - /* Next 4 digits must be a year. */ - for (u_int i = 0; i < 4; i++) { - if (!isdigit(nqn[i])) - return (false); - } - nqn += 4; - - /* '-' between year and month. */ - if (nqn[0] != '-') - return (false); - nqn++; - - /* 2 digit month. */ - for (u_int i = 0; i < 2; i++) { - if (!isdigit(nqn[i])) - return (false); - } - nqn += 2; - - /* '.' between month and reverse domain name. */ - if (nqn[0] != '.') - return (false); -#endif return (true); } diff --git a/sys/dev/nvmf/host/nvmf.c b/sys/dev/nvmf/host/nvmf.c index dbdd4568bdf1..1ac0d142443b 100644 --- a/sys/dev/nvmf/host/nvmf.c +++ b/sys/dev/nvmf/host/nvmf.c @@ -27,6 +27,7 @@ #include <dev/nvmf/host/nvmf_var.h> static struct cdevsw nvmf_cdevsw; +static struct taskqueue *nvmf_tq; bool nvmf_fail_disconnect = false; SYSCTL_BOOL(_kern_nvmf, OID_AUTO, fail_on_disconnection, CTLFLAG_RWTUN, @@ -34,7 +35,10 @@ SYSCTL_BOOL(_kern_nvmf, OID_AUTO, fail_on_disconnection, CTLFLAG_RWTUN, MALLOC_DEFINE(M_NVMF, "nvmf", "NVMe over Fabrics host"); +static void nvmf_controller_loss_task(void *arg, int pending); static void nvmf_disconnect_task(void *arg, int pending); +static void nvmf_request_reconnect(struct nvmf_softc *sc); +static void nvmf_request_reconnect_task(void *arg, int pending); static void nvmf_shutdown_pre_sync(void *arg, int howto); static void nvmf_shutdown_post_sync(void *arg, int howto); @@ -294,6 +298,9 @@ nvmf_establish_connection(struct nvmf_softc *sc, nvlist_t *nvl) admin = nvlist_get_nvlist(nvl, "admin"); io = nvlist_get_nvlist_array(nvl, "io", &num_io_queues); kato = dnvlist_get_number(nvl, "kato", 0); + sc->reconnect_delay = dnvlist_get_number(nvl, "reconnect_delay", 0); + sc->controller_loss_timeout = dnvlist_get_number(nvl, + "controller_loss_timeout", 0); /* Setup the admin queue. */ sc->admin = nvmf_init_qp(sc, trtype, admin, "admin queue", 0); @@ -504,6 +511,10 @@ nvmf_attach(device_t dev) callout_init(&sc->ka_tx_timer, 1); sx_init(&sc->connection_lock, "nvmf connection"); TASK_INIT(&sc->disconnect_task, 0, nvmf_disconnect_task, sc); + TIMEOUT_TASK_INIT(nvmf_tq, &sc->controller_loss_task, 0, + nvmf_controller_loss_task, sc); + TIMEOUT_TASK_INIT(nvmf_tq, &sc->request_reconnect_task, 0, + nvmf_request_reconnect_task, sc); oid = SYSCTL_ADD_NODE(device_get_sysctl_ctx(dev), SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), OID_AUTO, "ioq", @@ -603,7 +614,9 @@ out: nvmf_destroy_aer(sc); - taskqueue_drain(taskqueue_thread, &sc->disconnect_task); + taskqueue_drain_timeout(nvmf_tq, &sc->request_reconnect_task); + taskqueue_drain_timeout(nvmf_tq, &sc->controller_loss_task); + taskqueue_drain(nvmf_tq, &sc->disconnect_task); sx_destroy(&sc->connection_lock); nvlist_destroy(sc->rparams); free(sc->cdata, M_NVMF); @@ -613,7 +626,7 @@ out: void nvmf_disconnect(struct nvmf_softc *sc) { - taskqueue_enqueue(taskqueue_thread, &sc->disconnect_task); + taskqueue_enqueue(nvmf_tq, &sc->disconnect_task); } static void @@ -676,6 +689,74 @@ nvmf_disconnect_task(void *arg, int pending __unused) nvmf_destroy_qp(sc->admin); sc->admin = NULL; + if (sc->reconnect_delay != 0) + nvmf_request_reconnect(sc); + if (sc->controller_loss_timeout != 0) + taskqueue_enqueue_timeout(nvmf_tq, + &sc->controller_loss_task, sc->controller_loss_timeout * + hz); + + sx_xunlock(&sc->connection_lock); +} + +static void +nvmf_controller_loss_task(void *arg, int pending) +{ + struct nvmf_softc *sc = arg; + device_t dev; + int error; + + bus_topo_lock(); + sx_xlock(&sc->connection_lock); + if (sc->admin != NULL || sc->detaching) { + /* Reconnected or already detaching. */ + sx_xunlock(&sc->connection_lock); + bus_topo_unlock(); + return; + } + + sc->controller_timedout = true; + sx_xunlock(&sc->connection_lock); + + /* + * XXX: Doing this from here is a bit ugly. We don't have an + * extra reference on `dev` but bus_topo_lock should block any + * concurrent device_delete_child invocations. + */ + dev = sc->dev; + error = device_delete_child(root_bus, dev); + if (error != 0) + device_printf(dev, + "failed to detach after controller loss: %d\n", error); + bus_topo_unlock(); +} + +static void +nvmf_request_reconnect(struct nvmf_softc *sc) +{ + char buf[64]; + + sx_assert(&sc->connection_lock, SX_LOCKED); + + snprintf(buf, sizeof(buf), "name=\"%s\"", device_get_nameunit(sc->dev)); + devctl_notify("nvme", "controller", "RECONNECT", buf); + taskqueue_enqueue_timeout(nvmf_tq, &sc->request_reconnect_task, + sc->reconnect_delay * hz); +} + +static void +nvmf_request_reconnect_task(void *arg, int pending) +{ + struct nvmf_softc *sc = arg; + + sx_xlock(&sc->connection_lock); + if (sc->admin != NULL || sc->detaching || sc->controller_timedout) { + /* Reconnected or already detaching. */ + sx_xunlock(&sc->connection_lock); + return; + } + + nvmf_request_reconnect(sc); sx_xunlock(&sc->connection_lock); } @@ -699,7 +780,7 @@ nvmf_reconnect_host(struct nvmf_softc *sc, struct nvmf_ioc_nv *nv) } sx_xlock(&sc->connection_lock); - if (sc->admin != NULL || sc->detaching) { + if (sc->admin != NULL || sc->detaching || sc->controller_timedout) { error = EBUSY; goto out; } @@ -745,6 +826,9 @@ nvmf_reconnect_host(struct nvmf_softc *sc, struct nvmf_ioc_nv *nv) nvmf_reconnect_sim(sc); nvmf_rescan_all_ns(sc); + + taskqueue_cancel_timeout(nvmf_tq, &sc->request_reconnect_task, NULL); + taskqueue_cancel_timeout(nvmf_tq, &sc->controller_loss_task, NULL); out: sx_xunlock(&sc->connection_lock); nvlist_destroy(nvl); @@ -852,7 +936,21 @@ nvmf_detach(device_t dev) } free(sc->io, M_NVMF); - taskqueue_drain(taskqueue_thread, &sc->disconnect_task); + taskqueue_drain(nvmf_tq, &sc->disconnect_task); + if (taskqueue_cancel_timeout(nvmf_tq, &sc->request_reconnect_task, + NULL) != 0) + taskqueue_drain_timeout(nvmf_tq, &sc->request_reconnect_task); + + /* + * Don't cancel/drain the controller loss task if that task + * has fired and is triggering the detach. + */ + if (!sc->controller_timedout) { + if (taskqueue_cancel_timeout(nvmf_tq, &sc->controller_loss_task, + NULL) != 0) + taskqueue_drain_timeout(nvmf_tq, + &sc->controller_loss_task); + } if (sc->admin != NULL) nvmf_destroy_qp(sc->admin); @@ -1154,14 +1252,25 @@ static struct cdevsw nvmf_cdevsw = { static int nvmf_modevent(module_t mod, int what, void *arg) { + int error; + switch (what) { case MOD_LOAD: - return (nvmf_ctl_load()); + error = nvmf_ctl_load(); + if (error != 0) + return (error); + + nvmf_tq = taskqueue_create("nvmf", M_WAITOK | M_ZERO, + taskqueue_thread_enqueue, &nvmf_tq); + taskqueue_start_threads(&nvmf_tq, 1, PWAIT, "nvmf taskq"); + return (0); case MOD_QUIESCE: return (0); case MOD_UNLOAD: nvmf_ctl_unload(); destroy_dev_drain(&nvmf_cdevsw); + if (nvmf_tq != NULL) + taskqueue_free(nvmf_tq); return (0); default: return (EOPNOTSUPP); diff --git a/sys/dev/nvmf/host/nvmf_var.h b/sys/dev/nvmf/host/nvmf_var.h index e45a31f413a4..606245b3969c 100644 --- a/sys/dev/nvmf/host/nvmf_var.h +++ b/sys/dev/nvmf/host/nvmf_var.h @@ -75,9 +75,15 @@ struct nvmf_softc { struct callout ka_rx_timer; sbintime_t ka_rx_sbt; + struct timeout_task request_reconnect_task; + struct timeout_task controller_loss_task; + uint32_t reconnect_delay; + uint32_t controller_loss_timeout; + struct sx connection_lock; struct task disconnect_task; bool detaching; + bool controller_timedout; u_int num_aer; struct nvmf_aer *aer; diff --git a/sys/dev/nvmf/nvmf.h b/sys/dev/nvmf/nvmf.h index d4e7b1511e9d..9b2b4c1dea40 100644 --- a/sys/dev/nvmf/nvmf.h +++ b/sys/dev/nvmf/nvmf.h @@ -27,6 +27,13 @@ #define NVMF_NN (1024) /* + * Default timeouts for Fabrics hosts. These match values used by + * Linux. + */ +#define NVMF_DEFAULT_RECONNECT_DELAY 10 +#define NVMF_DEFAULT_CONTROLLER_LOSS 600 + +/* * (data, size) is the userspace buffer for a packed nvlist. * * For requests that copyout an nvlist, len is the amount of data @@ -68,6 +75,8 @@ struct nvmf_ioc_nv { * * number trtype * number kato (optional) + * number reconnect_delay (optional) + * number controller_loss_timeout (optional) * qpair handoff nvlist admin * qpair handoff nvlist array io * binary cdata struct nvme_controller_data @@ -81,6 +90,8 @@ struct nvmf_ioc_nv { * string hostnqn * number num_io_queues * number kato (optional) + * number reconnect_delay (optional) + * number controller_loss_timeout (optional) * number io_qsize * bool sq_flow_control * diff --git a/sys/dev/ofw/ofw_bus_subr.c b/sys/dev/ofw/ofw_bus_subr.c index 4d0479dfb957..b99d784929bc 100644 --- a/sys/dev/ofw/ofw_bus_subr.c +++ b/sys/dev/ofw/ofw_bus_subr.c @@ -634,11 +634,89 @@ ofw_bus_find_iparent(phandle_t node) return (iparent); } +static phandle_t +ofw_bus_search_iparent(phandle_t node) +{ + phandle_t iparent; + + do { + if (OF_getencprop(node, "interrupt-parent", &iparent, + sizeof(iparent)) > 0) { + node = OF_node_from_xref(iparent); + } else { + node = OF_parent(node); + } + if (node == 0) + return (0); + } while (!OF_hasprop(node, "#interrupt-cells")); + + return (OF_xref_from_node(node)); +} + +static int +ofw_bus_traverse_imap(phandle_t inode, phandle_t node, uint32_t *intr, + int intrsz, pcell_t *res, int ressz, phandle_t *iparentp) +{ + struct ofw_bus_iinfo ii; + void *reg; + uint32_t *intrp; + phandle_t iparent; + int rv = 0; + + /* We already have an interrupt controller */ + if (OF_hasprop(node, "interrupt-controller")) + return (0); + + intrp = malloc(intrsz, M_OFWPROP, M_WAITOK); + memcpy(intrp, intr, intrsz); + + while (true) { + /* There is no interrupt-map to follow */ + if (!OF_hasprop(inode, "interrupt-map")) { + free(intrp, M_OFWPROP); + return (0); + } + + memset(&ii, 0, sizeof(ii)); + ofw_bus_setup_iinfo(inode, &ii, sizeof(cell_t)); + + reg = NULL; + if (ii.opi_addrc > 0) + reg = malloc(ii.opi_addrc, M_OFWPROP, M_WAITOK); + + rv = ofw_bus_lookup_imap(node, &ii, reg, ii.opi_addrc, intrp, + intrsz, res, ressz, &iparent); + + free(reg, M_OFWPROP); + free(ii.opi_imap, M_OFWPROP); + free(ii.opi_imapmsk, M_OFWPROP); + free(intrp, M_OFWPROP); + + if (rv == 0) + return (0); + + node = inode; + inode = OF_node_from_xref(iparent); + + /* Stop when we have an interrupt controller */ + if (OF_hasprop(inode, "interrupt-controller")) { + *iparentp = iparent; + return (rv); + } + + intrsz = rv * sizeof(pcell_t); + intrp = malloc(intrsz, M_OFWPROP, M_WAITOK); + memcpy(intrp, res, intrsz); + } +} + int ofw_bus_intr_to_rl(device_t dev, phandle_t node, struct resource_list *rl, int *rlen) { - phandle_t iparent; + phandle_t iparent, iparent_node; + uint32_t result[16]; + uint32_t intrpcells, *intrp; uint32_t icells, *intr; int err, i, irqnum, nintr, rid; bool extended; @@ -646,15 +724,16 @@ ofw_bus_intr_to_rl(device_t dev, phandle_t node, nintr = OF_getencprop_alloc_multi(node, "interrupts", sizeof(*intr), (void **)&intr); if (nintr > 0) { - iparent = ofw_bus_find_iparent(node); + iparent = ofw_bus_search_iparent(node); if (iparent == 0) { device_printf(dev, "No interrupt-parent found, " "assuming direct parent\n"); iparent = OF_parent(node); iparent = OF_xref_from_node(iparent); } - if (OF_searchencprop(OF_node_from_xref(iparent), - "#interrupt-cells", &icells, sizeof(icells)) == -1) { + iparent_node = OF_node_from_xref(iparent); + if (OF_searchencprop(iparent_node, "#interrupt-cells", &icells, + sizeof(icells)) == -1) { device_printf(dev, "Missing #interrupt-cells " "property, assuming <1>\n"); icells = 1; @@ -677,7 +756,8 @@ ofw_bus_intr_to_rl(device_t dev, phandle_t node, for (i = 0; i < nintr; i += icells) { if (extended) { iparent = intr[i++]; - if (OF_searchencprop(OF_node_from_xref(iparent), + iparent_node = OF_node_from_xref(iparent); + if (OF_searchencprop(iparent_node, "#interrupt-cells", &icells, sizeof(icells)) == -1) { device_printf(dev, "Missing #interrupt-cells " "property\n"); @@ -691,7 +771,16 @@ ofw_bus_intr_to_rl(device_t dev, phandle_t node, break; } } - irqnum = ofw_bus_map_intr(dev, iparent, icells, &intr[i]); + + intrp = &intr[i]; + intrpcells = ofw_bus_traverse_imap(iparent_node, node, intrp, + icells * sizeof(intr[0]), result, sizeof(result), &iparent); + if (intrpcells > 0) + intrp = result; + else + intrpcells = icells; + + irqnum = ofw_bus_map_intr(dev, iparent, intrpcells, intrp); resource_list_add(rl, SYS_RES_IRQ, rid++, irqnum, irqnum, 1); } if (rlen != NULL) diff --git a/sys/dev/pci/pci_iov.c b/sys/dev/pci/pci_iov.c index 1f72391fb6b4..0efcfeac9eff 100644 --- a/sys/dev/pci/pci_iov.c +++ b/sys/dev/pci/pci_iov.c @@ -734,11 +734,18 @@ pci_iov_config(struct cdev *cdev, struct pci_iov_arg *arg) first_rid = pci_get_rid(dev) + rid_off; last_rid = first_rid + (num_vfs - 1) * rid_stride; - /* We don't yet support allocating extra bus numbers for VFs. */ if (pci_get_bus(dev) != PCI_RID2BUS(last_rid)) { - device_printf(dev, "not enough PCIe bus numbers for VFs\n"); - error = ENOSPC; - goto out; + int rid = 0; + uint16_t last_rid_bus = PCI_RID2BUS(last_rid); + + iov->iov_bus_res = bus_alloc_resource(bus, PCI_RES_BUS, &rid, + last_rid_bus, last_rid_bus, 1, RF_ACTIVE); + if (iov->iov_bus_res == NULL) { + device_printf(dev, + "failed to allocate PCIe bus number for VFs\n"); + error = ENOSPC; + goto out; + } } if (!ari_enabled && PCI_RID2SLOT(last_rid) != 0) { @@ -786,6 +793,11 @@ out: } } + if (iov->iov_bus_res != NULL) { + bus_release_resource(bus, iov->iov_bus_res); + iov->iov_bus_res = NULL; + } + if (iov->iov_flags & IOV_RMAN_INITED) { rman_fini(&iov->rman); iov->iov_flags &= ~IOV_RMAN_INITED; @@ -896,6 +908,11 @@ pci_iov_delete_iov_children(struct pci_devinfo *dinfo) } } + if (iov->iov_bus_res != NULL) { + bus_release_resource(bus, iov->iov_bus_res); + iov->iov_bus_res = NULL; + } + if (iov->iov_flags & IOV_RMAN_INITED) { rman_fini(&iov->rman); iov->iov_flags &= ~IOV_RMAN_INITED; diff --git a/sys/dev/pci/pci_iov_private.h b/sys/dev/pci/pci_iov_private.h index 7ae2219b936d..ecf0a9b21be5 100644 --- a/sys/dev/pci/pci_iov_private.h +++ b/sys/dev/pci/pci_iov_private.h @@ -39,6 +39,8 @@ struct pcicfg_iov { struct cdev *iov_cdev; nvlist_t *iov_schema; + struct resource *iov_bus_res; + struct pci_iov_bar iov_bar[PCIR_MAX_BAR_0 + 1]; struct rman rman; char rman_name[64]; diff --git a/sys/dev/qlnx/qlnxe/qlnx_os.c b/sys/dev/qlnx/qlnxe/qlnx_os.c index 05ec69a70dfe..4ad190374f87 100644 --- a/sys/dev/qlnx/qlnxe/qlnx_os.c +++ b/sys/dev/qlnx/qlnxe/qlnx_os.c @@ -30,6 +30,8 @@ * Author : David C Somayajulu, Cavium, Inc., San Jose, CA 95131. */ +#include "opt_inet.h" + #include <sys/cdefs.h> #include "qlnx_os.h" #include "bcm_osal.h" @@ -2306,8 +2308,6 @@ qlnx_init_ifnet(device_t dev, qlnx_host_t *ha) else if (device_id == QLOGIC_PCI_DEVICE_ID_1644) if_setbaudrate(ifp, IF_Gbps(100)); - if_setcapabilities(ifp, IFCAP_LINKSTATE); - if_setinitfn(ifp, qlnx_init); if_setsoftc(ifp, ha); if_setflags(ifp, IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST); @@ -2341,7 +2341,6 @@ qlnx_init_ifnet(device_t dev, qlnx_host_t *ha) if_setcapabilities(ifp, IFCAP_HWCSUM); if_setcapabilitiesbit(ifp, IFCAP_JUMBO_MTU, 0); - if_setcapabilitiesbit(ifp, IFCAP_VLAN_MTU, 0); if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWTAGGING, 0); if_setcapabilitiesbit(ifp, IFCAP_VLAN_HWFILTER, 0); @@ -2350,6 +2349,8 @@ qlnx_init_ifnet(device_t dev, qlnx_host_t *ha) if_setcapabilitiesbit(ifp, IFCAP_TSO4, 0); if_setcapabilitiesbit(ifp, IFCAP_TSO6, 0); if_setcapabilitiesbit(ifp, IFCAP_LRO, 0); + if_setcapabilitiesbit(ifp, IFCAP_LINKSTATE, 0); + if_setcapabilitiesbit(ifp, IFCAP_HWSTATS, 0); if_sethwtsomax(ifp, QLNX_MAX_TSO_FRAME_SIZE - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN)); @@ -2778,7 +2779,7 @@ qlnx_ioctl(if_t ifp, u_long cmd, caddr_t data) if (!p_ptt) { QL_DPRINT1(ha, "ecore_ptt_acquire failed\n"); - ret = -1; + ret = ERESTART; break; } @@ -2789,7 +2790,7 @@ qlnx_ioctl(if_t ifp, u_long cmd, caddr_t data) ecore_ptt_release(p_hwfn, p_ptt); if (ret) { - ret = -1; + ret = ENODEV; break; } diff --git a/sys/dev/random/fortuna.c b/sys/dev/random/fortuna.c index c4282c723a44..8363de99a60a 100644 --- a/sys/dev/random/fortuna.c +++ b/sys/dev/random/fortuna.c @@ -341,6 +341,13 @@ random_fortuna_process_event(struct harvest_event *event) u_int pl; RANDOM_RESEED_LOCK(); + /* + * Run SP 800-90B health tests on the source if so configured. + */ + if (!random_harvest_healthtest(event)) { + RANDOM_RESEED_UNLOCK(); + return; + } /*- * FS&K - P_i = P_i|<harvested stuff> * Accumulate the event into the appropriate pool diff --git a/sys/dev/random/random_harvestq.c b/sys/dev/random/random_harvestq.c index 395310b115fb..c7762967c4fb 100644 --- a/sys/dev/random/random_harvestq.c +++ b/sys/dev/random/random_harvestq.c @@ -88,6 +88,8 @@ static void random_sources_feed(void); static __read_mostly bool epoch_inited; static __read_mostly epoch_t rs_epoch; +static const char *random_source_descr[ENTROPYSOURCE]; + /* * How many events to queue up. We create this many items in * an 'empty' queue, then transfer them to the 'harvest' queue with @@ -299,6 +301,230 @@ random_sources_feed(void) explicit_bzero(entropy, sizeof(entropy)); } +/* + * State used for conducting NIST SP 800-90B health tests on entropy sources. + */ +static struct health_test_softc { + uint32_t ht_rct_value[HARVESTSIZE + 1]; + u_int ht_rct_count; /* number of samples with the same value */ + u_int ht_rct_limit; /* constant after init */ + + uint32_t ht_apt_value[HARVESTSIZE + 1]; + u_int ht_apt_count; /* number of samples with the same value */ + u_int ht_apt_seq; /* sequence number of the last sample */ + u_int ht_apt_cutoff; /* constant after init */ + + uint64_t ht_total_samples; + bool ondemand; /* Set to true to restart the state machine */ + enum { + INIT = 0, /* initial state */ + DISABLED, /* health checking is disabled */ + STARTUP, /* doing startup tests, samples are discarded */ + STEADY, /* steady-state operation */ + FAILED, /* health check failed, discard samples */ + } ht_state; +} healthtest[ENTROPYSOURCE]; + +#define RANDOM_SELFTEST_STARTUP_SAMPLES 1024 /* 4.3, requirement 4 */ +#define RANDOM_SELFTEST_APT_WINDOW 512 /* 4.4.2 */ + +static void +copy_event(uint32_t dst[static HARVESTSIZE + 1], + const struct harvest_event *event) +{ + memset(dst, 0, sizeof(uint32_t) * (HARVESTSIZE + 1)); + memcpy(dst, event->he_entropy, event->he_size); + dst[HARVESTSIZE] = event->he_somecounter; +} + +static void +random_healthtest_rct_init(struct health_test_softc *ht, + const struct harvest_event *event) +{ + ht->ht_rct_count = 1; + copy_event(ht->ht_rct_value, event); +} + +/* + * Apply the repitition count test to a sample. + * + * Return false if the test failed, i.e., we observed >= C consecutive samples + * with the same value, and true otherwise. + */ +static bool +random_healthtest_rct_next(struct health_test_softc *ht, + const struct harvest_event *event) +{ + uint32_t val[HARVESTSIZE + 1]; + + copy_event(val, event); + if (memcmp(val, ht->ht_rct_value, sizeof(ht->ht_rct_value)) != 0) { + ht->ht_rct_count = 1; + memcpy(ht->ht_rct_value, val, sizeof(ht->ht_rct_value)); + return (true); + } else { + ht->ht_rct_count++; + return (ht->ht_rct_count < ht->ht_rct_limit); + } +} + +static void +random_healthtest_apt_init(struct health_test_softc *ht, + const struct harvest_event *event) +{ + ht->ht_apt_count = 1; + ht->ht_apt_seq = 1; + copy_event(ht->ht_apt_value, event); +} + +static bool +random_healthtest_apt_next(struct health_test_softc *ht, + const struct harvest_event *event) +{ + uint32_t val[HARVESTSIZE + 1]; + + if (ht->ht_apt_seq == 0) { + random_healthtest_apt_init(ht, event); + return (true); + } + + copy_event(val, event); + if (memcmp(val, ht->ht_apt_value, sizeof(ht->ht_apt_value)) == 0) { + ht->ht_apt_count++; + if (ht->ht_apt_count >= ht->ht_apt_cutoff) + return (false); + } + + ht->ht_apt_seq++; + if (ht->ht_apt_seq == RANDOM_SELFTEST_APT_WINDOW) + ht->ht_apt_seq = 0; + + return (true); +} + +/* + * Run the health tests for the given event. This is assumed to be called from + * a serialized context. + */ +bool +random_harvest_healthtest(const struct harvest_event *event) +{ + struct health_test_softc *ht; + + ht = &healthtest[event->he_source]; + + /* + * Was on-demand testing requested? Restart the state machine if so, + * restarting the startup tests. + */ + if (atomic_load_bool(&ht->ondemand)) { + atomic_store_bool(&ht->ondemand, false); + ht->ht_state = INIT; + } + + switch (ht->ht_state) { + case __predict_false(INIT): + /* Store the first sample and initialize test state. */ + random_healthtest_rct_init(ht, event); + random_healthtest_apt_init(ht, event); + ht->ht_total_samples = 0; + ht->ht_state = STARTUP; + return (false); + case DISABLED: + /* No health testing for this source. */ + return (true); + case STEADY: + case STARTUP: + ht->ht_total_samples++; + if (random_healthtest_rct_next(ht, event) && + random_healthtest_apt_next(ht, event)) { + if (ht->ht_state == STARTUP && + ht->ht_total_samples >= + RANDOM_SELFTEST_STARTUP_SAMPLES) { + printf( + "random: health test passed for source %s\n", + random_source_descr[event->he_source]); + ht->ht_state = STEADY; + } + return (ht->ht_state == STEADY); + } + ht->ht_state = FAILED; + printf( + "random: health test failed for source %s, discarding samples\n", + random_source_descr[event->he_source]); + /* FALLTHROUGH */ + case FAILED: + return (false); + } +} + +static bool nist_healthtest_enabled = false; +SYSCTL_BOOL(_kern_random, OID_AUTO, nist_healthtest_enabled, + CTLFLAG_RDTUN, &nist_healthtest_enabled, 0, + "Enable NIST SP 800-90B health tests for noise sources"); + +static void +random_healthtest_init(enum random_entropy_source source) +{ + struct health_test_softc *ht; + + ht = &healthtest[source]; + KASSERT(ht->ht_state == INIT, + ("%s: health test state is %d for source %d", + __func__, ht->ht_state, source)); + + /* + * If health-testing is enabled, validate all sources except CACHED and + * VMGENID: they are deterministic sources used only a small, fixed + * number of times, so statistical testing is not applicable. + */ + if (!nist_healthtest_enabled || + source == RANDOM_CACHED || source == RANDOM_PURE_VMGENID) { + ht->ht_state = DISABLED; + return; + } + + /* + * Set cutoff values for the two tests, assuming that each sample has + * min-entropy of 1 bit and allowing for an error rate of 1 in 2^{34}. + * With a sample rate of RANDOM_KTHREAD_HZ, we expect to see an false + * positive once in ~54.5 years. + * + * The RCT limit comes from the formula in section 4.4.1. + * + * The APT cutoff is calculated using the formula in section 4.4.2 + * footnote 10 with the window size changed from 512 to 511, since the + * test as written counts the number of samples equal to the first + * sample in the window, and thus tests W-1 samples. + */ + ht->ht_rct_limit = 35; + ht->ht_apt_cutoff = 330; +} + +static int +random_healthtest_ondemand(SYSCTL_HANDLER_ARGS) +{ + u_int mask, source; + int error; + + mask = 0; + error = sysctl_handle_int(oidp, &mask, 0, req); + if (error != 0 || req->newptr == NULL) + return (error); + + while (mask != 0) { + source = ffs(mask) - 1; + if (source < nitems(healthtest)) + atomic_store_bool(&healthtest[source].ondemand, true); + mask &= ~(1u << source); + } + return (0); +} +SYSCTL_PROC(_kern_random, OID_AUTO, nist_healthtest_ondemand, + CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, 0, + random_healthtest_ondemand, "I", + "Re-run NIST SP 800-90B startup health tests for a noise source"); + static int random_check_uint_harvestmask(SYSCTL_HANDLER_ARGS) { @@ -362,7 +588,8 @@ static const char *random_source_descr[ENTROPYSOURCE] = { [RANDOM_SWI] = "SWI", [RANDOM_FS_ATIME] = "FS_ATIME", [RANDOM_UMA] = "UMA", - [RANDOM_CALLOUT] = "CALLOUT", /* ENVIRONMENTAL_END */ + [RANDOM_CALLOUT] = "CALLOUT", + [RANDOM_RANDOMDEV] = "RANDOMDEV", /* ENVIRONMENTAL_END */ [RANDOM_PURE_OCTEON] = "PURE_OCTEON", /* PURE_START */ [RANDOM_PURE_SAFE] = "PURE_SAFE", [RANDOM_PURE_GLXSB] = "PURE_GLXSB", @@ -424,6 +651,9 @@ random_harvestq_init(void *unused __unused) hc_source_mask = almost_everything_mask; RANDOM_HARVEST_INIT_LOCK(); harvest_context.hc_active_buf = 0; + + for (int i = 0; i < ENTROPYSOURCE; i++) + random_healthtest_init(i); } SYSINIT(random_device_h_init, SI_SUB_RANDOM, SI_ORDER_THIRD, random_harvestq_init, NULL); diff --git a/sys/dev/random/random_harvestq.h b/sys/dev/random/random_harvestq.h index 7804bf52aa4f..1d462500df85 100644 --- a/sys/dev/random/random_harvestq.h +++ b/sys/dev/random/random_harvestq.h @@ -49,4 +49,6 @@ random_get_cyclecount(void) return ((uint32_t)get_cyclecount()); } +bool random_harvest_healthtest(const struct harvest_event *event); + #endif /* SYS_DEV_RANDOM_RANDOM_HARVESTQ_H_INCLUDED */ diff --git a/sys/dev/random/randomdev.c b/sys/dev/random/randomdev.c index 9d1c7b1167c8..ced4dd8067d9 100644 --- a/sys/dev/random/randomdev.c +++ b/sys/dev/random/randomdev.c @@ -312,7 +312,7 @@ randomdev_accumulate(uint8_t *buf, u_int count) for (i = 0; i < RANDOM_KEYSIZE_WORDS; i += sizeof(event.he_entropy)/sizeof(event.he_entropy[0])) { event.he_somecounter = random_get_cyclecount(); event.he_size = sizeof(event.he_entropy); - event.he_source = RANDOM_CACHED; + event.he_source = RANDOM_RANDOMDEV; event.he_destination = destination++; /* Harmless cheating */ memcpy(event.he_entropy, entropy_data + i, sizeof(event.he_entropy)); p_random_alg_context->ra_event_processor(&event); diff --git a/sys/dev/ufshci/ufshci_private.h b/sys/dev/ufshci/ufshci_private.h index cac743884ee6..ac58d44102a0 100644 --- a/sys/dev/ufshci/ufshci_private.h +++ b/sys/dev/ufshci/ufshci_private.h @@ -149,6 +149,8 @@ struct ufshci_hw_queue { bus_dmamap_t queuemem_map; bus_addr_t req_queue_addr; + bus_addr_t *ucd_bus_addr; + uint32_t num_entries; uint32_t num_trackers; @@ -198,8 +200,6 @@ struct ufshci_req_queue { bus_dma_tag_t dma_tag_payload; bus_dmamap_t ucdmem_map; - - bus_addr_t ucd_addr; }; struct ufshci_device { diff --git a/sys/dev/ufshci/ufshci_req_sdb.c b/sys/dev/ufshci/ufshci_req_sdb.c index 4670281d367a..b1f303afaef5 100644 --- a/sys/dev/ufshci/ufshci_req_sdb.c +++ b/sys/dev/ufshci/ufshci_req_sdb.c @@ -48,6 +48,29 @@ ufshci_req_sdb_cmd_desc_destroy(struct ufshci_req_queue *req_queue) } } +static void +ufshci_ucd_map(void *arg, bus_dma_segment_t *seg, int nseg, int error) +{ + struct ufshci_hw_queue *hwq = arg; + int i; + + if (error != 0) { + printf("ufshci: Failed to map UCD, error = %d\n", error); + return; + } + + if (hwq->num_trackers != nseg) { + printf( + "ufshci: Failed to map UCD, num_trackers = %d, nseg = %d\n", + hwq->num_trackers, nseg); + return; + } + + for (i = 0; i < nseg; i++) { + hwq->ucd_bus_addr[i] = seg[i].ds_addr; + } +} + static int ufshci_req_sdb_cmd_desc_construct(struct ufshci_req_queue *req_queue, uint32_t num_entries, struct ufshci_controller *ctrlr) @@ -55,7 +78,6 @@ ufshci_req_sdb_cmd_desc_construct(struct ufshci_req_queue *req_queue, struct ufshci_hw_queue *hwq = &req_queue->hwq[UFSHCI_SDB_Q]; struct ufshci_tracker *tr; size_t ucd_allocsz, payload_allocsz; - uint64_t ucdmem_phys; uint8_t *ucdmem; int i, error; @@ -71,10 +93,11 @@ ufshci_req_sdb_cmd_desc_construct(struct ufshci_req_queue *req_queue, * Allocate physical memory for UTP Command Descriptor (UCD) * Note: UFSHCI UCD format is restricted to 128-byte alignment. */ - error = bus_dma_tag_create(bus_get_dma_tag(ctrlr->dev), 128, - ctrlr->page_size, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, - ucd_allocsz, howmany(ucd_allocsz, ctrlr->page_size), - ctrlr->page_size, 0, NULL, NULL, &req_queue->dma_tag_ucd); + error = bus_dma_tag_create(bus_get_dma_tag(ctrlr->dev), 128, 0, + BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, ucd_allocsz, + howmany(ucd_allocsz, sizeof(struct ufshci_utp_cmd_desc)), + sizeof(struct ufshci_utp_cmd_desc), 0, NULL, NULL, + &req_queue->dma_tag_ucd); if (error != 0) { ufshci_printf(ctrlr, "request cmd desc tag create failed %d\n", error); @@ -88,7 +111,7 @@ ufshci_req_sdb_cmd_desc_construct(struct ufshci_req_queue *req_queue, } if (bus_dmamap_load(req_queue->dma_tag_ucd, req_queue->ucdmem_map, - ucdmem, ucd_allocsz, ufshci_single_map, &ucdmem_phys, 0) != 0) { + ucdmem, ucd_allocsz, ufshci_ucd_map, hwq, 0) != 0) { ufshci_printf(ctrlr, "failed to load cmd desc memory\n"); bus_dmamem_free(req_queue->dma_tag_ucd, req_queue->ucd, req_queue->ucdmem_map); @@ -96,7 +119,6 @@ ufshci_req_sdb_cmd_desc_construct(struct ufshci_req_queue *req_queue, } req_queue->ucd = (struct ufshci_utp_cmd_desc *)ucdmem; - req_queue->ucd_addr = ucdmem_phys; /* * Allocate physical memory for PRDT @@ -128,10 +150,9 @@ ufshci_req_sdb_cmd_desc_construct(struct ufshci_req_queue *req_queue, tr->slot_state = UFSHCI_SLOT_STATE_FREE; tr->ucd = (struct ufshci_utp_cmd_desc *)ucdmem; - tr->ucd_bus_addr = ucdmem_phys; + tr->ucd_bus_addr = hwq->ucd_bus_addr[i]; ucdmem += sizeof(struct ufshci_utp_cmd_desc); - ucdmem_phys += sizeof(struct ufshci_utp_cmd_desc); hwq->act_tr[i] = tr; } @@ -175,6 +196,11 @@ ufshci_req_sdb_construct(struct ufshci_controller *ctrlr, req_queue->hwq = malloc(sizeof(struct ufshci_hw_queue), M_UFSHCI, M_ZERO | M_NOWAIT); hwq = &req_queue->hwq[UFSHCI_SDB_Q]; + hwq->num_entries = req_queue->num_entries; + hwq->num_trackers = req_queue->num_trackers; + req_queue->hwq->ucd_bus_addr = malloc(sizeof(bus_addr_t) * + req_queue->num_trackers, + M_UFSHCI, M_ZERO | M_NOWAIT); mtx_init(&hwq->qlock, "ufshci req_queue lock", NULL, MTX_DEF); @@ -277,6 +303,7 @@ ufshci_req_sdb_destroy(struct ufshci_controller *ctrlr, if (mtx_initialized(&hwq->qlock)) mtx_destroy(&hwq->qlock); + free(req_queue->hwq->ucd_bus_addr, M_UFSHCI); free(req_queue->hwq, M_UFSHCI); } diff --git a/sys/dev/usb/controller/xhci_pci.c b/sys/dev/usb/controller/xhci_pci.c index b50e33ea36ce..d5cfd228a429 100644 --- a/sys/dev/usb/controller/xhci_pci.c +++ b/sys/dev/usb/controller/xhci_pci.c @@ -99,6 +99,11 @@ xhci_pci_match(device_t self) return ("AMD Starship USB 3.0 controller"); case 0x149c1022: return ("AMD Matisse USB 3.0 controller"); + case 0x15b61022: + case 0x15b71022: + return ("AMD Raphael/Granite Ridge USB 3.1 controller"); + case 0x15b81022: + return ("AMD Raphael/Granite Ridge USB 2.0 controller"); case 0x15e01022: case 0x15e11022: return ("AMD Raven USB 3.1 controller"); @@ -109,6 +114,8 @@ xhci_pci_match(device_t self) return ("AMD 300 Series USB 3.1 controller"); case 0x43d51022: return ("AMD 400 Series USB 3.1 controller"); + case 0x43f71022: + return ("AMD 600 Series USB 3.2 controller"); case 0x78121022: case 0x78141022: case 0x79141022: diff --git a/sys/dev/vmm/vmm_dev.c b/sys/dev/vmm/vmm_dev.c index 819debadd1ac..9f2b009d02ec 100644 --- a/sys/dev/vmm/vmm_dev.c +++ b/sys/dev/vmm/vmm_dev.c @@ -30,7 +30,8 @@ #include <dev/vmm/vmm_mem.h> #include <dev/vmm/vmm_stat.h> -#if defined(__amd64__) && defined(COMPAT_FREEBSD12) +#ifdef __amd64__ +#ifdef COMPAT_FREEBSD12 struct vm_memseg_12 { int segid; size_t len; @@ -42,7 +43,22 @@ _Static_assert(sizeof(struct vm_memseg_12) == 80, "COMPAT_FREEBSD12 ABI"); _IOW('v', IOCNUM_ALLOC_MEMSEG, struct vm_memseg_12) #define VM_GET_MEMSEG_12 \ _IOWR('v', IOCNUM_GET_MEMSEG, struct vm_memseg_12) -#endif +#endif /* COMPAT_FREEBSD12 */ +#ifdef COMPAT_FREEBSD14 +struct vm_memseg_14 { + int segid; + size_t len; + char name[VM_MAX_SUFFIXLEN + 1]; +}; +_Static_assert(sizeof(struct vm_memseg_14) == (VM_MAX_SUFFIXLEN + 1 + 16), + "COMPAT_FREEBSD14 ABI"); + +#define VM_ALLOC_MEMSEG_14 \ + _IOW('v', IOCNUM_ALLOC_MEMSEG, struct vm_memseg_14) +#define VM_GET_MEMSEG_14 \ + _IOWR('v', IOCNUM_GET_MEMSEG, struct vm_memseg_14) +#endif /* COMPAT_FREEBSD14 */ +#endif /* __amd64__ */ struct devmem_softc { int segid; @@ -257,7 +273,8 @@ get_memseg(struct vmmdev_softc *sc, struct vm_memseg *mseg, size_t len) } static int -alloc_memseg(struct vmmdev_softc *sc, struct vm_memseg *mseg, size_t len) +alloc_memseg(struct vmmdev_softc *sc, struct vm_memseg *mseg, size_t len, + struct domainset *domainset) { char *name; int error; @@ -278,8 +295,7 @@ alloc_memseg(struct vmmdev_softc *sc, struct vm_memseg *mseg, size_t len) if (error) goto done; } - - error = vm_alloc_memseg(sc->vm, mseg->segid, mseg->len, sysmem); + error = vm_alloc_memseg(sc->vm, mseg->segid, mseg->len, sysmem, domainset); if (error) goto done; @@ -295,6 +311,20 @@ done: return (error); } +#if defined(__amd64__) && \ + (defined(COMPAT_FREEBSD14) || defined(COMPAT_FREEBSD12)) +/* + * Translate pre-15.0 memory segment identifiers into their 15.0 counterparts. + */ +static void +adjust_segid(struct vm_memseg *mseg) +{ + if (mseg->segid != VM_SYSMEM) { + mseg->segid += (VM_BOOTROM - 1); + } +} +#endif + static int vm_get_register_set(struct vcpu *vcpu, unsigned int count, int *regnum, uint64_t *regval) @@ -353,10 +383,16 @@ static const struct vmmdev_ioctl vmmdev_ioctls[] = { VMMDEV_IOCTL(VM_STATS, VMMDEV_IOCTL_LOCK_ONE_VCPU), VMMDEV_IOCTL(VM_STAT_DESC, 0), -#if defined(__amd64__) && defined(COMPAT_FREEBSD12) +#ifdef __amd64__ +#ifdef COMPAT_FREEBSD12 VMMDEV_IOCTL(VM_ALLOC_MEMSEG_12, VMMDEV_IOCTL_XLOCK_MEMSEGS | VMMDEV_IOCTL_LOCK_ALL_VCPUS), #endif +#ifdef COMPAT_FREEBSD14 + VMMDEV_IOCTL(VM_ALLOC_MEMSEG_14, + VMMDEV_IOCTL_XLOCK_MEMSEGS | VMMDEV_IOCTL_LOCK_ALL_VCPUS), +#endif +#endif /* __amd64__ */ VMMDEV_IOCTL(VM_ALLOC_MEMSEG, VMMDEV_IOCTL_XLOCK_MEMSEGS | VMMDEV_IOCTL_LOCK_ALL_VCPUS), VMMDEV_IOCTL(VM_MMAP_MEMSEG, @@ -366,9 +402,14 @@ static const struct vmmdev_ioctl vmmdev_ioctls[] = { VMMDEV_IOCTL(VM_REINIT, VMMDEV_IOCTL_XLOCK_MEMSEGS | VMMDEV_IOCTL_LOCK_ALL_VCPUS), -#if defined(__amd64__) && defined(COMPAT_FREEBSD12) +#ifdef __amd64__ +#if defined(COMPAT_FREEBSD12) VMMDEV_IOCTL(VM_GET_MEMSEG_12, VMMDEV_IOCTL_SLOCK_MEMSEGS), #endif +#ifdef COMPAT_FREEBSD14 + VMMDEV_IOCTL(VM_GET_MEMSEG_14, VMMDEV_IOCTL_SLOCK_MEMSEGS), +#endif +#endif /* __amd64__ */ VMMDEV_IOCTL(VM_GET_MEMSEG, VMMDEV_IOCTL_SLOCK_MEMSEGS), VMMDEV_IOCTL(VM_MMAP_GETNEXT, VMMDEV_IOCTL_SLOCK_MEMSEGS), @@ -388,6 +429,7 @@ vmmdev_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag, struct vmmdev_softc *sc; struct vcpu *vcpu; const struct vmmdev_ioctl *ioctl; + struct vm_memseg *mseg; int error, vcpuid; sc = vmmdev_lookup2(cdev); @@ -499,20 +541,77 @@ vmmdev_ioctl(struct cdev *cdev, u_long cmd, caddr_t data, int fflag, error = vm_munmap_memseg(sc->vm, mu->gpa, mu->len); break; } -#if defined(__amd64__) && defined(COMPAT_FREEBSD12) +#ifdef __amd64__ +#ifdef COMPAT_FREEBSD12 case VM_ALLOC_MEMSEG_12: - error = alloc_memseg(sc, (struct vm_memseg *)data, - sizeof(((struct vm_memseg_12 *)0)->name)); + mseg = (struct vm_memseg *)data; + + adjust_segid(mseg); + error = alloc_memseg(sc, mseg, + sizeof(((struct vm_memseg_12 *)0)->name), NULL); break; case VM_GET_MEMSEG_12: - error = get_memseg(sc, (struct vm_memseg *)data, + mseg = (struct vm_memseg *)data; + + adjust_segid(mseg); + error = get_memseg(sc, mseg, sizeof(((struct vm_memseg_12 *)0)->name)); break; -#endif - case VM_ALLOC_MEMSEG: - error = alloc_memseg(sc, (struct vm_memseg *)data, - sizeof(((struct vm_memseg *)0)->name)); +#endif /* COMPAT_FREEBSD12 */ +#ifdef COMPAT_FREEBSD14 + case VM_ALLOC_MEMSEG_14: + mseg = (struct vm_memseg *)data; + + adjust_segid(mseg); + error = alloc_memseg(sc, mseg, + sizeof(((struct vm_memseg_14 *)0)->name), NULL); + break; + case VM_GET_MEMSEG_14: + mseg = (struct vm_memseg *)data; + + adjust_segid(mseg); + error = get_memseg(sc, mseg, + sizeof(((struct vm_memseg_14 *)0)->name)); + break; +#endif /* COMPAT_FREEBSD14 */ +#endif /* __amd64__ */ + case VM_ALLOC_MEMSEG: { + domainset_t *mask; + struct domainset *domainset, domain; + + domainset = NULL; + mseg = (struct vm_memseg *)data; + if (mseg->ds_policy != DOMAINSET_POLICY_INVALID && mseg->ds_mask != NULL) { + if (mseg->ds_mask_size < sizeof(domainset_t) || + mseg->ds_mask_size > DOMAINSET_MAXSIZE / NBBY) { + error = ERANGE; + break; + } + memset(&domain, 0, sizeof(domain)); + mask = malloc(mseg->ds_mask_size, M_VMMDEV, M_WAITOK); + error = copyin(mseg->ds_mask, mask, mseg->ds_mask_size); + if (error) { + free(mask, M_VMMDEV); + break; + } + error = domainset_populate(&domain, mask, mseg->ds_policy, + mseg->ds_mask_size); + if (error) { + free(mask, M_VMMDEV); + break; + } + domainset = domainset_create(&domain); + if (domainset == NULL) { + error = EINVAL; + free(mask, M_VMMDEV); + break; + } + free(mask, M_VMMDEV); + } + error = alloc_memseg(sc, mseg, sizeof(mseg->name), domainset); + break; + } case VM_GET_MEMSEG: error = get_memseg(sc, (struct vm_memseg *)data, sizeof(((struct vm_memseg *)0)->name)); @@ -820,7 +919,6 @@ sysctl_vmm_destroy(SYSCTL_HANDLER_ARGS) buflen = VM_MAX_NAMELEN + 1; buf = malloc(buflen, M_VMMDEV, M_WAITOK | M_ZERO); - strlcpy(buf, "beavis", buflen); error = sysctl_handle_string(oidp, buf, buflen, req); if (error == 0 && req->newptr != NULL) error = vmmdev_lookup_and_destroy(buf, req->td->td_ucred); @@ -830,7 +928,7 @@ sysctl_vmm_destroy(SYSCTL_HANDLER_ARGS) SYSCTL_PROC(_hw_vmm, OID_AUTO, destroy, CTLTYPE_STRING | CTLFLAG_RW | CTLFLAG_PRISON | CTLFLAG_MPSAFE, NULL, 0, sysctl_vmm_destroy, "A", - NULL); + "Destroy a vmm(4) instance (legacy interface)"); static struct cdevsw vmmdevsw = { .d_name = "vmmdev", @@ -909,7 +1007,6 @@ sysctl_vmm_create(SYSCTL_HANDLER_ARGS) buflen = VM_MAX_NAMELEN + 1; buf = malloc(buflen, M_VMMDEV, M_WAITOK | M_ZERO); - strlcpy(buf, "beavis", buflen); error = sysctl_handle_string(oidp, buf, buflen, req); if (error == 0 && req->newptr != NULL) error = vmmdev_create(buf, req->td->td_ucred); @@ -919,7 +1016,7 @@ sysctl_vmm_create(SYSCTL_HANDLER_ARGS) SYSCTL_PROC(_hw_vmm, OID_AUTO, create, CTLTYPE_STRING | CTLFLAG_RW | CTLFLAG_PRISON | CTLFLAG_MPSAFE, NULL, 0, sysctl_vmm_create, "A", - NULL); + "Create a vmm(4) instance (legacy interface)"); static int vmmctl_open(struct cdev *cdev, int flags, int fmt, struct thread *td) diff --git a/sys/dev/vmm/vmm_mem.c b/sys/dev/vmm/vmm_mem.c index c61ae2d44b96..be59e37de33d 100644 --- a/sys/dev/vmm/vmm_mem.c +++ b/sys/dev/vmm/vmm_mem.c @@ -7,6 +7,7 @@ #include <sys/types.h> #include <sys/lock.h> +#include <sys/malloc.h> #include <sys/sx.h> #include <sys/systm.h> @@ -156,10 +157,11 @@ vm_mem_allocated(struct vcpu *vcpu, vm_paddr_t gpa) } int -vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem) +vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem, + struct domainset *obj_domainset) { - struct vm_mem *mem; struct vm_mem_seg *seg; + struct vm_mem *mem; vm_object_t obj; mem = vm_mem(vm); @@ -179,13 +181,22 @@ vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem) return (EINVAL); } + /* + * When given an impossible policy, signal an + * error to the user. + */ + if (obj_domainset != NULL && domainset_empty_vm(obj_domainset)) + return (EINVAL); obj = vm_object_allocate(OBJT_SWAP, len >> PAGE_SHIFT); if (obj == NULL) return (ENOMEM); seg->len = len; seg->object = obj; + if (obj_domainset != NULL) + seg->object->domain.dr_policy = obj_domainset; seg->sysmem = sysmem; + return (0); } diff --git a/sys/dev/vmm/vmm_mem.h b/sys/dev/vmm/vmm_mem.h index a4be4c1c57aa..856470cf2590 100644 --- a/sys/dev/vmm/vmm_mem.h +++ b/sys/dev/vmm/vmm_mem.h @@ -8,6 +8,27 @@ #ifndef _DEV_VMM_MEM_H_ #define _DEV_VMM_MEM_H_ +/* Maximum number of NUMA domains in a guest. */ +#define VM_MAXMEMDOM 8 +#define VM_MAXSYSMEM VM_MAXMEMDOM + +/* + * Identifiers for memory segments. + * Each guest NUMA domain is represented by a single system + * memory segment from [VM_SYSMEM, VM_MAXSYSMEM). + * The remaining identifiers can be used to create devmem segments. + */ +enum { + VM_SYSMEM = 0, + VM_BOOTROM = VM_MAXSYSMEM, + VM_FRAMEBUFFER, + VM_PCIROM, + VM_MEMSEG_END +}; + +#define VM_MAX_MEMSEGS VM_MEMSEG_END +#define VM_MAX_MEMMAPS (VM_MAX_MEMSEGS * 2) + #ifdef _KERNEL #include <sys/types.h> @@ -31,9 +52,6 @@ struct vm_mem_map { int flags; }; -#define VM_MAX_MEMSEGS 4 -#define VM_MAX_MEMMAPS 8 - struct vm_mem { struct vm_mem_map mem_maps[VM_MAX_MEMMAPS]; struct vm_mem_seg mem_segs[VM_MAX_MEMSEGS]; @@ -55,7 +73,8 @@ void vm_assert_memseg_xlocked(struct vm *vm); int vm_mmap_memseg(struct vm *vm, vm_paddr_t gpa, int segid, vm_ooffset_t off, size_t len, int prot, int flags); int vm_munmap_memseg(struct vm *vm, vm_paddr_t gpa, size_t len); -int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem); +int vm_alloc_memseg(struct vm *vm, int ident, size_t len, bool sysmem, + struct domainset *obj_domainset); void vm_free_memseg(struct vm *vm, int ident); /* diff --git a/sys/dev/vt/hw/vga/vt_vga.c b/sys/dev/vt/hw/vga/vt_vga.c index 64039575c0ad..675c0573bd7e 100644 --- a/sys/dev/vt/hw/vga/vt_vga.c +++ b/sys/dev/vt/hw/vga/vt_vga.c @@ -1347,7 +1347,7 @@ vga_postswitch(struct vt_device *vd) /* Reinit VGA mode, to restore view after app which change mode. */ vga_initialize(vd, (vd->vd_flags & VDF_TEXTMODE)); - /* Ask vt(9) to update chars on visible area. */ + /* Ask vt(4) to update chars on visible area. */ vd->vd_flags |= VDF_INVALID; } diff --git a/sys/dev/vt/vt_core.c b/sys/dev/vt/vt_core.c index b0f58b38a6f1..b51ef6766de4 100644 --- a/sys/dev/vt/vt_core.c +++ b/sys/dev/vt/vt_core.c @@ -125,10 +125,10 @@ static const struct terminal_class vt_termclass = { (vw)->vw_number) static SYSCTL_NODE(_kern, OID_AUTO, vt, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, - "vt(9) parameters"); + "vt(4) parameters"); static VT_SYSCTL_INT(enable_altgr, 1, "Enable AltGr key (Do not assume R.Alt as Alt)"); static VT_SYSCTL_INT(enable_bell, 0, "Enable bell"); -static VT_SYSCTL_INT(debug, 0, "vt(9) debug level"); +static VT_SYSCTL_INT(debug, 0, "vt(4) debug level"); static VT_SYSCTL_INT(deadtimer, 15, "Time to wait busy process in VT_PROCESS mode"); static VT_SYSCTL_INT(suspendswitch, 1, "Switch to VT0 before suspend"); diff --git a/sys/fs/fdescfs/fdesc_vnops.c b/sys/fs/fdescfs/fdesc_vnops.c index 676ea5de12b8..58a22b8bdc50 100644 --- a/sys/fs/fdescfs/fdesc_vnops.c +++ b/sys/fs/fdescfs/fdesc_vnops.c @@ -547,6 +547,8 @@ fdesc_readdir(struct vop_readdir_args *ap) fmp = VFSTOFDESC(ap->a_vp->v_mount); if (ap->a_ncookies != NULL) *ap->a_ncookies = 0; + if (ap->a_eofflag != NULL) + *ap->a_eofflag = 0; off = (int)uio->uio_offset; if (off != uio->uio_offset || off < 0 || (u_int)off % UIO_MX != 0 || @@ -559,7 +561,12 @@ fdesc_readdir(struct vop_readdir_args *ap) fcnt = i - 2; /* The first two nodes are `.' and `..' */ FILEDESC_SLOCK(fdp); - while (i < fdp->fd_nfiles + 2 && uio->uio_resid >= UIO_MX) { + while (uio->uio_resid >= UIO_MX) { + if (i >= fdp->fd_nfiles + 2) { + if (ap->a_eofflag != NULL) + *ap->a_eofflag = 1; + break; + } bzero((caddr_t)dp, UIO_MX); switch (i) { case 0: /* `.' */ diff --git a/sys/fs/fuse/fuse_internal.h b/sys/fs/fuse/fuse_internal.h index cddf88095840..932012b5f52a 100644 --- a/sys/fs/fuse/fuse_internal.h +++ b/sys/fs/fuse/fuse_internal.h @@ -208,9 +208,9 @@ fuse_match_cred(struct ucred *basecred, struct ucred *usercred) if (basecred->cr_uid == usercred->cr_uid && basecred->cr_uid == usercred->cr_ruid && basecred->cr_uid == usercred->cr_svuid && - basecred->cr_groups[0] == usercred->cr_groups[0] && - basecred->cr_groups[0] == usercred->cr_rgid && - basecred->cr_groups[0] == usercred->cr_svgid) + basecred->cr_gid == usercred->cr_gid && + basecred->cr_gid == usercred->cr_rgid && + basecred->cr_gid == usercred->cr_svgid) return (0); return (EPERM); diff --git a/sys/fs/fuse/fuse_ipc.c b/sys/fs/fuse/fuse_ipc.c index 0b6048644d32..a751c09159ff 100644 --- a/sys/fs/fuse/fuse_ipc.c +++ b/sys/fs/fuse/fuse_ipc.c @@ -868,7 +868,7 @@ fuse_setup_ihead(struct fuse_in_header *ihead, struct fuse_ticket *ftick, ihead->pid = pid; ihead->uid = cred->cr_uid; - ihead->gid = cred->cr_groups[0]; + ihead->gid = cred->cr_gid; } /* diff --git a/sys/fs/fuse/fuse_vnops.c b/sys/fs/fuse/fuse_vnops.c index ae28617537fd..32872e8f3f3a 100644 --- a/sys/fs/fuse/fuse_vnops.c +++ b/sys/fs/fuse/fuse_vnops.c @@ -884,7 +884,7 @@ fuse_vnop_copy_file_range(struct vop_copy_file_range_args *ap) return (EXTERROR(ENOSYS, "FUSE_COPY_FILE_RANGE does not " "support different credentials for infd and outfd")); - if (incred->cr_groups[0] != outcred->cr_groups[0]) + if (incred->cr_gid != outcred->cr_gid) return (EXTERROR(ENOSYS, "FUSE_COPY_FILE_RANGE does not " "support different credentials for infd and outfd")); diff --git a/sys/fs/msdosfs/msdosfs_conv.c b/sys/fs/msdosfs/msdosfs_conv.c index da4848169173..208b64930e61 100644 --- a/sys/fs/msdosfs/msdosfs_conv.c +++ b/sys/fs/msdosfs/msdosfs_conv.c @@ -797,19 +797,24 @@ mbsadjpos(const char **instr, size_t inlen, size_t outlen, int weight, int flag, static u_char * dos2unixchr(u_char *outbuf, const u_char **instr, size_t *ilen, int lower, struct msdosfsmount *pmp) { - u_char c, *outp; - size_t len, olen; + u_char c, *outp, *outp1; + size_t i, len, olen; outp = outbuf; if (pmp->pm_flags & MSDOSFSMNT_KICONV && msdosfs_iconv) { olen = len = 4; + outp1 = outp; if (lower & (LCASE_BASE | LCASE_EXT)) msdosfs_iconv->convchr_case(pmp->pm_d2u, (const char **)instr, ilen, (char **)&outp, &olen, KICONV_LOWER); else msdosfs_iconv->convchr(pmp->pm_d2u, (const char **)instr, ilen, (char **)&outp, &olen); + for (i = 0; i < outp - outp1; i++) { + if (outp1[i] == '/') + outp1[i] = '?'; + } len -= olen; /* @@ -826,6 +831,8 @@ dos2unixchr(u_char *outbuf, const u_char **instr, size_t *ilen, int lower, struc c = dos2unix[c]; if (lower & (LCASE_BASE | LCASE_EXT)) c = u2l[c]; + if (c == '/') + c = '?'; *outp++ = c; outbuf[1] = '\0'; } diff --git a/sys/fs/msdosfs/msdosfs_vnops.c b/sys/fs/msdosfs/msdosfs_vnops.c index 5db61c8951f6..33e0d94954d7 100644 --- a/sys/fs/msdosfs/msdosfs_vnops.c +++ b/sys/fs/msdosfs/msdosfs_vnops.c @@ -1521,6 +1521,9 @@ msdosfs_readdir(struct vop_readdir_args *ap) ap->a_vp, uio, ap->a_cred, ap->a_eofflag); #endif + if (ap->a_eofflag != NULL) + *ap->a_eofflag = 0; + /* * msdosfs_readdir() won't operate properly on regular files since * it does i/o only with the filesystem vnode, and hence can @@ -1614,8 +1617,11 @@ msdosfs_readdir(struct vop_readdir_args *ap) on = (offset - bias) & pmp->pm_crbomask; n = min(pmp->pm_bpcluster - on, uio->uio_resid); diff = dep->de_FileSize - (offset - bias); - if (diff <= 0) - break; + if (diff <= 0) { + if (ap->a_eofflag != NULL) + *ap->a_eofflag = 1; + goto out; + } n = min(n, diff); error = pcbmap(dep, lbn, &bn, &cn, &blsize); if (error) @@ -1646,6 +1652,8 @@ msdosfs_readdir(struct vop_readdir_args *ap) */ if (dentp->deName[0] == SLOT_EMPTY) { brelse(bp); + if (ap->a_eofflag != NULL) + *ap->a_eofflag = 1; goto out; } /* @@ -1743,15 +1751,6 @@ out: uio->uio_offset = off; - /* - * Set the eofflag (NFS uses it) - */ - if (ap->a_eofflag) { - if (dep->de_FileSize - (offset - bias) <= 0) - *ap->a_eofflag = 1; - else - *ap->a_eofflag = 0; - } return (error); } diff --git a/sys/fs/nfs/nfs_commonport.c b/sys/fs/nfs/nfs_commonport.c index 0c94f4e7dc52..222cfc03e4b3 100644 --- a/sys/fs/nfs/nfs_commonport.c +++ b/sys/fs/nfs/nfs_commonport.c @@ -379,7 +379,8 @@ newnfs_setroot(struct ucred *cred) { cred->cr_uid = 0; - cred->cr_groups[0] = 0; + cred->cr_gid = 0; + /* XXXKE Fix this if cr_gid gets separated out. */ cred->cr_ngroups = 1; } diff --git a/sys/fs/nfs/nfs_commonsubs.c b/sys/fs/nfs/nfs_commonsubs.c index 4c498e96a3c0..a957315aaa12 100644 --- a/sys/fs/nfs/nfs_commonsubs.c +++ b/sys/fs/nfs/nfs_commonsubs.c @@ -647,7 +647,8 @@ nfscl_fillsattr(struct nfsrv_descript *nd, struct vattr *vap, NFSATTRBIT_TIMECREATE)) NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_TIMECREATE); (void) nfsv4_fillattr(nd, vp->v_mount, vp, NULL, vap, NULL, 0, - &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL); + &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL, + false, false, false); break; } } @@ -2646,7 +2647,8 @@ nfsv4_fillattr(struct nfsrv_descript *nd, struct mount *mp, vnode_t vp, NFSACL_T *saclp, struct vattr *vap, fhandle_t *fhp, int rderror, nfsattrbit_t *attrbitp, struct ucred *cred, NFSPROC_T *p, int isdgram, int reterr, int supports_nfsv4acls, int at_root, uint64_t mounted_on_fileno, - struct statfs *pnfssf) + struct statfs *pnfssf, bool xattrsupp, bool has_hiddensystem, + bool has_namedattr) { int bitpos, retnum = 0; u_int32_t *tl; @@ -2660,10 +2662,7 @@ nfsv4_fillattr(struct nfsrv_descript *nd, struct mount *mp, vnode_t vp, struct nfsfsinfo fsinf; struct timespec temptime; NFSACL_T *aclp, *naclp = NULL; - size_t atsiz; - bool xattrsupp; short irflag; - long has_pathconf; #ifdef QUOTA struct dqblk dqb; uid_t savuid; @@ -2747,18 +2746,6 @@ nfsv4_fillattr(struct nfsrv_descript *nd, struct mount *mp, vnode_t vp, } } - /* Check to see if Extended Attributes are supported. */ - xattrsupp = false; - if (NFSISSET_ATTRBIT(retbitp, NFSATTRBIT_XATTRSUPPORT)) { - if (NFSVOPLOCK(vp, LK_SHARED) == 0) { - error = VOP_GETEXTATTR(vp, EXTATTR_NAMESPACE_USER, - "xxx", NULL, &atsiz, cred, p); - NFSVOPUNLOCK(vp); - if (error != EOPNOTSUPP) - xattrsupp = true; - } - } - /* * Put out the attribute bitmap for the ones being filled in * and get the field for the number of attributes returned. @@ -2780,11 +2767,7 @@ nfsv4_fillattr(struct nfsrv_descript *nd, struct mount *mp, vnode_t vp, NFSCLRBIT_ATTRBIT(&attrbits,NFSATTRBIT_ACLSUPPORT); NFSCLRBIT_ATTRBIT(&attrbits,NFSATTRBIT_ACL); } - if (cred == NULL || p == NULL || vp == NULL || - VOP_PATHCONF(vp, _PC_HAS_HIDDENSYSTEM, - &has_pathconf) != 0) - has_pathconf = 0; - if (has_pathconf == 0) { + if (!has_hiddensystem) { NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_HIDDEN); NFSCLRBIT_ATTRBIT(&attrbits, NFSATTRBIT_SYSTEM); } @@ -2828,10 +2811,7 @@ nfsv4_fillattr(struct nfsrv_descript *nd, struct mount *mp, vnode_t vp, break; case NFSATTRBIT_NAMEDATTR: NFSM_BUILD(tl, u_int32_t *, NFSX_UNSIGNED); - if (VOP_PATHCONF(vp, _PC_HAS_NAMEDATTR, - &has_pathconf) != 0) - has_pathconf = 0; - if (has_pathconf != 0) + if (has_namedattr) *tl = newnfs_true; else *tl = newnfs_false; diff --git a/sys/fs/nfs/nfs_var.h b/sys/fs/nfs/nfs_var.h index 3b6c1ec90c06..54f60a753c50 100644 --- a/sys/fs/nfs/nfs_var.h +++ b/sys/fs/nfs/nfs_var.h @@ -395,8 +395,9 @@ int nfsrv_putopbit(struct nfsrv_descript *, nfsopbit_t *); void nfsrv_wcc(struct nfsrv_descript *, int, struct nfsvattr *, int, struct nfsvattr *); int nfsv4_fillattr(struct nfsrv_descript *, struct mount *, vnode_t, NFSACL_T *, - struct vattr *, fhandle_t *, int, nfsattrbit_t *, - struct ucred *, NFSPROC_T *, int, int, int, int, uint64_t, struct statfs *); + struct vattr *, fhandle_t *, int, nfsattrbit_t *, struct ucred *, + NFSPROC_T *, int, int, int, int, uint64_t, struct statfs *, bool, bool, + bool); void nfsrv_fillattr(struct nfsrv_descript *, struct nfsvattr *); struct mbuf *nfsrv_adj(struct mbuf *, int, int); void nfsrv_postopattr(struct nfsrv_descript *, int, struct nfsvattr *); @@ -735,7 +736,8 @@ int nfsvno_updfilerev(vnode_t, struct nfsvattr *, struct nfsrv_descript *, NFSPROC_T *); int nfsvno_fillattr(struct nfsrv_descript *, struct mount *, vnode_t, struct nfsvattr *, fhandle_t *, int, nfsattrbit_t *, - struct ucred *, NFSPROC_T *, int, int, int, int, uint64_t); + struct ucred *, NFSPROC_T *, int, int, int, int, uint64_t, bool, bool, + bool); int nfsrv_sattr(struct nfsrv_descript *, vnode_t, struct nfsvattr *, nfsattrbit_t *, NFSACL_T *, NFSPROC_T *); int nfsv4_sattr(struct nfsrv_descript *, vnode_t, struct nfsvattr *, nfsattrbit_t *, diff --git a/sys/fs/nfsclient/nfs_clrpcops.c b/sys/fs/nfsclient/nfs_clrpcops.c index e0e66baca44d..36b534be531e 100644 --- a/sys/fs/nfsclient/nfs_clrpcops.c +++ b/sys/fs/nfsclient/nfs_clrpcops.c @@ -5436,7 +5436,8 @@ nfsrpc_setaclrpc(vnode_t vp, struct ucred *cred, NFSPROC_T *p, NFSZERO_ATTRBIT(&attrbits); NFSSETBIT_ATTRBIT(&attrbits, NFSATTRBIT_ACL); (void) nfsv4_fillattr(nd, vp->v_mount, vp, aclp, NULL, NULL, 0, - &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL); + &attrbits, NULL, NULL, 0, 0, 0, 0, (uint64_t)0, NULL, false, false, + false); error = nfscl_request(nd, vp, p, cred); if (error) return (error); @@ -6932,7 +6933,8 @@ nfscl_dofflayoutio(vnode_t vp, struct uio *uiop, int *iomode, int *must_commit, if ((dp->nfsdi_flags & NFSDI_TIGHTCOUPLED) == 0) { tcred = NFSNEWCRED(cred); tcred->cr_uid = flp->nfsfl_ffm[mirror].user; - tcred->cr_groups[0] = flp->nfsfl_ffm[mirror].group; + tcred->cr_gid = flp->nfsfl_ffm[mirror].group; + /* XXXKE Fix this if cr_gid gets separated out. */ tcred->cr_ngroups = 1; } else tcred = cred; diff --git a/sys/fs/nfsclient/nfs_clstate.c b/sys/fs/nfsclient/nfs_clstate.c index 1ae5ed1a75ca..99a781640c53 100644 --- a/sys/fs/nfsclient/nfs_clstate.c +++ b/sys/fs/nfsclient/nfs_clstate.c @@ -3701,7 +3701,7 @@ nfscl_docb(struct nfsrv_descript *nd, NFSPROC_T *p) if (!error) (void) nfsv4_fillattr(nd, NULL, NULL, NULL, &va, NULL, 0, &rattrbits, NULL, p, 0, 0, 0, 0, - (uint64_t)0, NULL); + (uint64_t)0, NULL, false, false, false); break; case NFSV4OP_CBRECALL: NFSCL_DEBUG(4, "cbrecall\n"); diff --git a/sys/fs/nfsclient/nfs_clvnops.c b/sys/fs/nfsclient/nfs_clvnops.c index fbfcdafaa06b..fa451887e73e 100644 --- a/sys/fs/nfsclient/nfs_clvnops.c +++ b/sys/fs/nfsclient/nfs_clvnops.c @@ -1096,12 +1096,11 @@ nfs_setattr(struct vop_setattr_args *ap) /* * Disallow write attempts if the filesystem is mounted read-only. */ - if ((vap->va_flags != VNOVAL || vap->va_uid != (uid_t)VNOVAL || + if ((vap->va_flags != (u_long)VNOVAL || vap->va_uid != (uid_t)VNOVAL || vap->va_gid != (gid_t)VNOVAL || vap->va_atime.tv_sec != VNOVAL || vap->va_mtime.tv_sec != VNOVAL || vap->va_birthtime.tv_sec != VNOVAL || - vap->va_mode != (mode_t)VNOVAL || - vap->va_flags != (u_long)VNOVAL) && + vap->va_mode != (mode_t)VNOVAL) && (vp->v_mount->mnt_flag & MNT_RDONLY)) return (EROFS); if (vap->va_size != VNOVAL) { diff --git a/sys/fs/nfsserver/nfs_nfsdport.c b/sys/fs/nfsserver/nfs_nfsdport.c index a81f1492ef95..4f0d5946d6b9 100644 --- a/sys/fs/nfsserver/nfs_nfsdport.c +++ b/sys/fs/nfsserver/nfs_nfsdport.c @@ -1652,10 +1652,11 @@ nfsvno_rename(struct nameidata *fromndp, struct nameidata *tondp, } if (fvp == tvp) { /* - * If source and destination are the same, there is nothing to - * do. Set error to -1 to indicate this. + * If source and destination are the same, there is + * nothing to do. Set error to EJUSTRETURN to indicate + * this. */ - error = -1; + error = EJUSTRETURN; goto out; } if (nd->nd_flag & ND_NFSV4) { @@ -1697,10 +1698,26 @@ nfsvno_rename(struct nameidata *fromndp, struct nameidata *tondp, " dsdvp=%p\n", dsdvp[0]); } out: - if (!error) { + mp = NULL; + if (error == 0) { + error = VOP_GETWRITEMOUNT(tondp->ni_dvp, &mp); + if (error == 0) { + if (mp == NULL) { + error = ENOENT; + } else { + error = lockmgr(&mp->mnt_renamelock, + LK_EXCLUSIVE | LK_NOWAIT, NULL); + if (error != 0) + error = ERELOOKUP; + } + } + } + if (error == 0) { error = VOP_RENAME(fromndp->ni_dvp, fromndp->ni_vp, &fromndp->ni_cnd, tondp->ni_dvp, tondp->ni_vp, &tondp->ni_cnd); + lockmgr(&mp->mnt_renamelock, LK_RELEASE, 0); + vfs_rel(mp); } else { if (tdvp == tvp) vrele(tdvp); @@ -1710,8 +1727,13 @@ out: vput(tvp); vrele(fromndp->ni_dvp); vrele(fvp); - if (error == -1) + if (error == EJUSTRETURN) { error = 0; + } else if (error == ERELOOKUP && mp != NULL) { + lockmgr(&mp->mnt_renamelock, LK_EXCLUSIVE, 0); + lockmgr(&mp->mnt_renamelock, LK_RELEASE, 0); + vfs_rel(mp); + } } /* @@ -2090,7 +2112,8 @@ int nfsvno_fillattr(struct nfsrv_descript *nd, struct mount *mp, struct vnode *vp, struct nfsvattr *nvap, fhandle_t *fhp, int rderror, nfsattrbit_t *attrbitp, struct ucred *cred, struct thread *p, int isdgram, int reterr, - int supports_nfsv4acls, int at_root, uint64_t mounted_on_fileno) + int supports_nfsv4acls, int at_root, uint64_t mounted_on_fileno, + bool xattrsupp, bool has_hiddensystem, bool has_namedattr) { struct statfs *sf; int error; @@ -2109,7 +2132,7 @@ nfsvno_fillattr(struct nfsrv_descript *nd, struct mount *mp, struct vnode *vp, } error = nfsv4_fillattr(nd, mp, vp, NULL, &nvap->na_vattr, fhp, rderror, attrbitp, cred, p, isdgram, reterr, supports_nfsv4acls, at_root, - mounted_on_fileno, sf); + mounted_on_fileno, sf, xattrsupp, has_hiddensystem, has_namedattr); free(sf, M_TEMP); NFSEXITCODE2(0, nd); return (error); @@ -2426,7 +2449,7 @@ nfsrvd_readdirplus(struct nfsrv_descript *nd, int isdgram, struct nfsvattr nva, at, *nvap = &nva; struct mbuf *mb0, *mb1; struct nfsreferral *refp; - int nlen, r, error = 0, getret = 1, usevget = 1; + int nlen, r, error = 0, getret = 1, ret, usevget = 1; int siz, cnt, fullsiz, eofflag, ncookies, entrycnt; caddr_t bpos0, bpos1; u_int64_t off, toff, verf __unused; @@ -2440,6 +2463,9 @@ nfsrvd_readdirplus(struct nfsrv_descript *nd, int isdgram, uint64_t mounted_on_fileno; struct thread *p = curthread; int bextpg0, bextpg1, bextpgsiz0, bextpgsiz1; + size_t atsiz; + long pathval; + bool has_hiddensystem, has_namedattr, xattrsupp; if (nd->nd_repstat) { nfsrv_postopattr(nd, getret, &at); @@ -2914,9 +2940,32 @@ again: *tl++ = newnfs_true; txdr_hyper(*cookiep, tl); dirlen += nfsm_strtom(nd, dp->d_name, nlen); + xattrsupp = false; + has_hiddensystem = false; + has_namedattr = false; if (nvp != NULL) { supports_nfsv4acls = nfs_supportsnfsv4acls(nvp); + if (NFSISSET_ATTRBIT(&attrbits, + NFSATTRBIT_XATTRSUPPORT)) { + ret = VOP_GETEXTATTR(nvp, + EXTATTR_NAMESPACE_USER, + "xxx", NULL, &atsiz, + nd->nd_cred, p); + xattrsupp = ret != EOPNOTSUPP; + } + if (VOP_PATHCONF(nvp, + _PC_HAS_HIDDENSYSTEM, &pathval) != + 0) + pathval = 0; + has_hiddensystem = pathval > 0; + pathval = 0; + if (NFSISSET_ATTRBIT(&attrbits, + NFSATTRBIT_NAMEDATTR) && + VOP_PATHCONF(nvp, _PC_HAS_NAMEDATTR, + &pathval) != 0) + pathval = 0; + has_namedattr = pathval > 0; NFSVOPUNLOCK(nvp); } else supports_nfsv4acls = 0; @@ -2936,13 +2985,15 @@ again: nvp, nvap, &nfh, r, &rderrbits, nd->nd_cred, p, isdgram, 0, supports_nfsv4acls, at_root, - mounted_on_fileno); + mounted_on_fileno, xattrsupp, + has_hiddensystem, has_namedattr); } else { dirlen += nfsvno_fillattr(nd, new_mp, nvp, nvap, &nfh, r, &attrbits, nd->nd_cred, p, isdgram, 0, supports_nfsv4acls, at_root, - mounted_on_fileno); + mounted_on_fileno, xattrsupp, + has_hiddensystem, has_namedattr); } if (nvp != NULL) vrele(nvp); @@ -6334,7 +6385,7 @@ nfsrv_setacldsdorpc(fhandle_t *fhp, struct ucred *cred, NFSPROC_T *p, * the same type (VREG). */ nfsv4_fillattr(nd, NULL, vp, aclp, NULL, NULL, 0, &attrbits, NULL, - NULL, 0, 0, 0, 0, 0, NULL); + NULL, 0, 0, 0, 0, 0, NULL, false, false, false); error = newnfs_request(nd, nmp, NULL, &nmp->nm_sockreq, NULL, p, cred, NFS_PROG, NFS_VER4, NULL, 1, NULL, NULL); if (error != 0) { diff --git a/sys/fs/nfsserver/nfs_nfsdserv.c b/sys/fs/nfsserver/nfs_nfsdserv.c index f7564ade401b..9eebcda548c6 100644 --- a/sys/fs/nfsserver/nfs_nfsdserv.c +++ b/sys/fs/nfsserver/nfs_nfsdserv.c @@ -241,7 +241,7 @@ nfsrvd_getattr(struct nfsrv_descript *nd, int isdgram, { struct nfsvattr nva; fhandle_t fh; - int at_root = 0, error = 0, supports_nfsv4acls; + int at_root = 0, error = 0, ret, supports_nfsv4acls; struct nfsreferral *refp; nfsattrbit_t attrbits, tmpbits; struct mount *mp; @@ -250,6 +250,9 @@ nfsrvd_getattr(struct nfsrv_descript *nd, int isdgram, uint64_t mounted_on_fileno = 0; accmode_t accmode; struct thread *p = curthread; + size_t atsiz; + long pathval; + bool has_hiddensystem, has_namedattr, xattrsupp; if (nd->nd_repstat) goto out; @@ -307,6 +310,26 @@ nfsrvd_getattr(struct nfsrv_descript *nd, int isdgram, &nva, &attrbits, p); if (nd->nd_repstat == 0) { supports_nfsv4acls = nfs_supportsnfsv4acls(vp); + xattrsupp = false; + if (NFSISSET_ATTRBIT(&attrbits, + NFSATTRBIT_XATTRSUPPORT)) { + ret = VOP_GETEXTATTR(vp, + EXTATTR_NAMESPACE_USER, + "xxx", NULL, &atsiz, nd->nd_cred, + p); + xattrsupp = ret != EOPNOTSUPP; + } + if (VOP_PATHCONF(vp, _PC_HAS_HIDDENSYSTEM, + &pathval) != 0) + pathval = 0; + has_hiddensystem = pathval > 0; + pathval = 0; + if (NFSISSET_ATTRBIT(&attrbits, + NFSATTRBIT_NAMEDATTR) && + VOP_PATHCONF(vp, _PC_HAS_NAMEDATTR, + &pathval) != 0) + pathval = 0; + has_namedattr = pathval > 0; mp = vp->v_mount; if (nfsrv_enable_crossmntpt != 0 && vp->v_type == VDIR && @@ -340,7 +363,9 @@ nfsrvd_getattr(struct nfsrv_descript *nd, int isdgram, (void)nfsvno_fillattr(nd, mp, vp, &nva, &fh, 0, &attrbits, nd->nd_cred, p, isdgram, 1, supports_nfsv4acls, - at_root, mounted_on_fileno); + at_root, mounted_on_fileno, + xattrsupp, has_hiddensystem, + has_namedattr); vfs_unbusy(mp); } vrele(vp); @@ -4353,9 +4378,10 @@ nfsrvd_openattr(struct nfsrv_descript *nd, __unused int isdgram, int error = 0; NFSNAMEICNDSET(&cn, nd->nd_cred, LOOKUP, OPENNAMED | ISLASTCN | - NOFOLLOW); + NOFOLLOW | LOCKLEAF); cn.cn_nameptr = "."; cn.cn_namelen = 1; + cn.cn_lkflags = LK_SHARED; NFSM_DISSECT(tl, uint32_t *, NFSX_UNSIGNED); if (*tl == newnfs_true) cn.cn_flags |= CREATENAMED; @@ -4374,6 +4400,8 @@ nfsrvd_openattr(struct nfsrv_descript *nd, __unused int isdgram, if (nd->nd_repstat == ENOATTR) nd->nd_repstat = NFSERR_NOENT; } + if (nd->nd_repstat == 0) + NFSVOPUNLOCK(*vpp); vput(dp); NFSEXITCODE2(0, nd); diff --git a/sys/fs/p9fs/p9fs_vnops.c b/sys/fs/p9fs/p9fs_vnops.c index 56bf766ef801..227e2b93883e 100644 --- a/sys/fs/p9fs/p9fs_vnops.c +++ b/sys/fs/p9fs/p9fs_vnops.c @@ -1784,6 +1784,9 @@ p9fs_readdir(struct vop_readdir_args *ap) return (EBADF); } + if (ap->a_eofflag != NULL) + *ap->a_eofflag = 0; + io_buffer = uma_zalloc(p9fs_io_buffer_zone, M_WAITOK); /* We haven't reached the end yet. read more. */ @@ -1801,8 +1804,11 @@ p9fs_readdir(struct vop_readdir_args *ap) count = p9_client_readdir(vofid, (char *)io_buffer, diroffset, count); - if (count == 0) + if (count == 0) { + if (ap->a_eofflag != NULL) + *ap->a_eofflag = 1; break; + } if (count < 0) { error = EIO; diff --git a/sys/fs/pseudofs/pseudofs_vnops.c b/sys/fs/pseudofs/pseudofs_vnops.c index 0bdfedffafcb..8cd092118d0e 100644 --- a/sys/fs/pseudofs/pseudofs_vnops.c +++ b/sys/fs/pseudofs/pseudofs_vnops.c @@ -850,7 +850,7 @@ pfs_readdir(struct vop_readdir_args *va) struct uio *uio; struct pfsentry *pfsent, *pfsent2; struct pfsdirentlist lst; - off_t offset; + off_t coffset, offset; int error, i, resid; STAILQ_INIT(&lst); @@ -860,6 +860,9 @@ pfs_readdir(struct vop_readdir_args *va) PFS_TRACE(("%s pid %lu", pd->pn_name, (unsigned long)pid)); pfs_assert_not_owned(pd); + if (va->a_eofflag != NULL) + *va->a_eofflag = 0; + if (vn->v_type != VDIR) PFS_RETURN (ENOTDIR); KASSERT_PN_IS_DIR(pd); @@ -878,6 +881,10 @@ pfs_readdir(struct vop_readdir_args *va) if (pid != NO_PID && !pfs_lookup_proc(pid, &proc)) PFS_RETURN (ENOENT); + /* + * The allproc lock is required in pfs_iterate() for procdir + * directories. + */ sx_slock(&allproc_lock); pfs_lock(pd); @@ -897,23 +904,15 @@ pfs_readdir(struct vop_readdir_args *va) } } - /* skip unwanted entries */ - for (pn = NULL, p = NULL; offset > 0; offset -= PFS_DELEN) { + for (pn = NULL, p = NULL, coffset = 0; resid >= PFS_DELEN; + coffset += PFS_DELEN) { if (pfs_iterate(curthread, proc, pd, &pn, &p) == -1) { - /* nothing left... */ - if (proc != NULL) { - _PRELE(proc); - PROC_UNLOCK(proc); - } - pfs_unlock(pd); - sx_sunlock(&allproc_lock); - PFS_RETURN (0); + if (va->a_eofflag != NULL) + *va->a_eofflag = 1; + break; } - } - - /* fill in entries */ - while (pfs_iterate(curthread, proc, pd, &pn, &p) != -1 && - resid >= PFS_DELEN) { + if (coffset < offset) + continue; if ((pfsent = malloc(sizeof(struct pfsentry), M_IOV, M_NOWAIT | M_ZERO)) == NULL) { error = ENOMEM; diff --git a/sys/fs/smbfs/smbfs_io.c b/sys/fs/smbfs/smbfs_io.c index 35454998fc8e..8c484381ed59 100644 --- a/sys/fs/smbfs/smbfs_io.c +++ b/sys/fs/smbfs/smbfs_io.c @@ -71,7 +71,7 @@ SYSCTL_INT(_vfs_smbfs, OID_AUTO, fastlookup, CTLFLAG_RW, &smbfs_fastlookup, 0, " #define DE_SIZE (sizeof(struct dirent)) static int -smbfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred) +smbfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred, int *eofp) { struct dirent de; struct componentname cn; @@ -86,6 +86,8 @@ smbfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred) SMBVDEBUG("dirname='%s'\n", np->n_name); scred = smbfs_malloc_scred(); smb_makescred(scred, uio->uio_td, cred); + if (eofp != NULL) + *eofp = 0; offset = uio->uio_offset / DE_SIZE; /* offset in the directory */ limit = uio->uio_resid / DE_SIZE; if (uio->uio_resid < DE_SIZE || uio->uio_offset < 0) { @@ -138,8 +140,7 @@ smbfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred) if (error) { smbfs_findclose(np->n_dirseq, scred); np->n_dirseq = NULL; - error = ENOENT ? 0 : error; - goto out; + goto out1; } } error = 0; @@ -170,16 +171,21 @@ smbfs_readvdir(struct vnode *vp, struct uio *uio, struct ucred *cred) if (error) break; } - if (error == ENOENT) - error = 0; uio->uio_offset = offset * DE_SIZE; +out1: + if (error == ENOENT) { + if (eofp != NULL) + *eofp = 1; + error = 0; + } out: smbfs_free_scred(scred); return error; } int -smbfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred) +smbfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred, + int *eofp) { struct smbmount *smp = VFSTOSMBFS(vp->v_mount); struct smbnode *np = VTOSMB(vp); @@ -209,7 +215,7 @@ smbfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred) lks = LK_EXCLUSIVE; /* lockstatus(vp->v_vnlock); */ if (lks == LK_SHARED) vn_lock(vp, LK_UPGRADE | LK_RETRY); - error = smbfs_readvdir(vp, uiop, cred); + error = smbfs_readvdir(vp, uiop, cred, eofp); if (lks == LK_SHARED) vn_lock(vp, LK_DOWNGRADE | LK_RETRY); return error; diff --git a/sys/fs/smbfs/smbfs_node.h b/sys/fs/smbfs/smbfs_node.h index f28f0007100a..8c8ce038b913 100644 --- a/sys/fs/smbfs/smbfs_node.h +++ b/sys/fs/smbfs/smbfs_node.h @@ -93,7 +93,7 @@ u_int32_t smbfs_hash(const u_char *name, int nmlen); int smbfs_getpages(struct vop_getpages_args *); int smbfs_putpages(struct vop_putpages_args *); -int smbfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred); +int smbfs_readvnode(struct vnode *vp, struct uio *uiop, struct ucred *cred, int *eofp); int smbfs_writevnode(struct vnode *vp, struct uio *uiop, struct ucred *cred, int ioflag); void smbfs_attr_cacheenter(struct vnode *vp, struct smbfattr *fap); int smbfs_attr_cachelookup(struct vnode *vp ,struct vattr *va); diff --git a/sys/fs/smbfs/smbfs_vnops.c b/sys/fs/smbfs/smbfs_vnops.c index 5d412cabadb8..63b249c93771 100644 --- a/sys/fs/smbfs/smbfs_vnops.c +++ b/sys/fs/smbfs/smbfs_vnops.c @@ -466,7 +466,7 @@ smbfs_read(struct vop_read_args *ap) SMBVDEBUG("\n"); if (vp->v_type != VREG && vp->v_type != VDIR) return EPERM; - return smbfs_readvnode(vp, uio, ap->a_cred); + return smbfs_readvnode(vp, uio, ap->a_cred, NULL); } static int @@ -748,7 +748,6 @@ smbfs_readdir(struct vop_readdir_args *ap) { struct vnode *vp = ap->a_vp; struct uio *uio = ap->a_uio; - int error; if (vp->v_type != VDIR) return (EPERM); @@ -758,8 +757,7 @@ smbfs_readdir(struct vop_readdir_args *ap) return (EOPNOTSUPP); } #endif - error = smbfs_readvnode(vp, uio, ap->a_cred); - return error; + return (smbfs_readvnode(vp, uio, ap->a_cred, ap->a_eofflag)); } /* ARGSUSED */ diff --git a/sys/fs/udf/ecma167-udf.h b/sys/fs/udf/ecma167-udf.h index 839bbec08254..19e114763cac 100644 --- a/sys/fs/udf/ecma167-udf.h +++ b/sys/fs/udf/ecma167-udf.h @@ -243,7 +243,7 @@ struct part_map_spare { uint8_t n_st; /* Number of Sparing Tables */ uint8_t reserved1; uint32_t st_size; - uint32_t st_loc[1]; + uint32_t st_loc[]; } __packed; union udf_pmap { @@ -266,7 +266,7 @@ struct udf_sparing_table { uint16_t rt_l; /* Relocation Table len */ uint8_t reserved[2]; uint32_t seq_num; - struct spare_map_entry entries[1]; + struct spare_map_entry entries[]; } __packed; /* Partition Descriptor [3/10.5] */ diff --git a/sys/fs/udf/udf_vfsops.c b/sys/fs/udf/udf_vfsops.c index c7438147c0a0..c5ef1f686093 100644 --- a/sys/fs/udf/udf_vfsops.c +++ b/sys/fs/udf/udf_vfsops.c @@ -81,6 +81,7 @@ #include <sys/fcntl.h> #include <sys/iconv.h> #include <sys/kernel.h> +#include <sys/limits.h> #include <sys/malloc.h> #include <sys/mount.h> #include <sys/namei.h> @@ -729,7 +730,7 @@ udf_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp) struct ifid *ifhp; struct vnode *nvp; struct udf_node *np; - off_t fsize; + uint64_t fsize; int error; ifhp = (struct ifid *)fhp; @@ -741,6 +742,10 @@ udf_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp) np = VTON(nvp); fsize = le64toh(np->fentry->inf_len); + if (fsize > OFF_MAX) { + *vpp = NULLVP; + return (EIO); + } *vpp = nvp; vnode_create_vobject(*vpp, fsize, curthread); diff --git a/sys/fs/udf/udf_vnops.c b/sys/fs/udf/udf_vnops.c index 88bf4917a851..37889241e8c3 100644 --- a/sys/fs/udf/udf_vnops.c +++ b/sys/fs/udf/udf_vnops.c @@ -39,6 +39,7 @@ #include <sys/conf.h> #include <sys/buf.h> #include <sys/iconv.h> +#include <sys/limits.h> #include <sys/mount.h> #include <sys/vnode.h> #include <sys/dirent.h> @@ -182,11 +183,14 @@ udf_access(struct vop_access_args *a) } static int -udf_open(struct vop_open_args *ap) { +udf_open(struct vop_open_args *ap) +{ struct udf_node *np = VTON(ap->a_vp); - off_t fsize; + uint64_t fsize; fsize = le64toh(np->fentry->inf_len); + if (fsize > OFF_MAX) + return (EIO); vnode_create_vobject(ap->a_vp, fsize, ap->a_td); return 0; } @@ -314,12 +318,13 @@ udf_getattr(struct vop_getattr_args *a) * that directories consume at least one logical block, * make it appear so. */ - if (fentry->logblks_rec != 0) { - vap->va_size = - le64toh(fentry->logblks_rec) * node->udfmp->bsize; - } else { + vap->va_size = le64toh(fentry->logblks_rec); + if (vap->va_size == 0) vap->va_size = node->udfmp->bsize; - } + else if (vap->va_size > UINT64_MAX / node->udfmp->bsize) + vap->va_size = UINT64_MAX; + else + vap->va_size *= node->udfmp->bsize; } else { vap->va_size = le64toh(fentry->inf_len); } @@ -446,6 +451,7 @@ udf_read(struct vop_read_args *ap) struct buf *bp; uint8_t *data; daddr_t lbn, rablock; + uint64_t len; off_t diff, fsize; ssize_t n; int error = 0; @@ -471,7 +477,12 @@ udf_read(struct vop_read_args *ap) return (error); } - fsize = le64toh(node->fentry->inf_len); + len = le64toh(node->fentry->inf_len); + if (len > OFF_MAX) { + /* too big, just cap to the requested length */ + len = uio->uio_resid; + } + fsize = len; udfmp = node->udfmp; do { lbn = lblkno(udfmp, uio->uio_offset); @@ -783,6 +794,7 @@ udf_readdir(struct vop_readdir_args *a) struct udf_uiodir uiodir; struct udf_dirstream *ds; uint64_t *cookies = NULL; + uint64_t len; int ncookies; int error = 0; @@ -811,8 +823,12 @@ udf_readdir(struct vop_readdir_args *a) * Iterate through the file id descriptors. Give the parent dir * entry special attention. */ - ds = udf_opendir(node, uio->uio_offset, le64toh(node->fentry->inf_len), - node->udfmp); + len = le64toh(node->fentry->inf_len); + if (len > INT_MAX) { + /* too big, just cap to INT_MAX */ + len = INT_MAX; + } + ds = udf_opendir(node, uio->uio_offset, len, node->udfmp); while ((fid = udf_getfid(ds)) != NULL) { /* XXX Should we return an error on a bad fid? */ @@ -904,7 +920,8 @@ udf_readlink(struct vop_readlink_args *ap) struct udf_node *node; void *buf; char *cp; - int error, len, root; + uint64_t len; + int error, root; /* * A symbolic link in UDF is a list of variable-length path @@ -914,6 +931,8 @@ udf_readlink(struct vop_readlink_args *ap) vp = ap->a_vp; node = VTON(vp); len = le64toh(node->fentry->inf_len); + if (len > MAXPATHLEN) + return (EIO); buf = malloc(len, M_DEVBUF, M_WAITOK); iov[0].iov_len = len; iov[0].iov_base = buf; @@ -1116,13 +1135,14 @@ udf_lookup(struct vop_cachedlookup_args *a) struct udf_mnt *udfmp; struct fileid_desc *fid = NULL; struct udf_dirstream *ds; + uint64_t fsize; u_long nameiop; u_long flags; char *nameptr; long namelen; ino_t id = 0; int offset, error = 0; - int fsize, lkflags, ltype, numdirpasses; + int lkflags, ltype, numdirpasses; dvp = a->a_dvp; node = VTON(dvp); @@ -1133,6 +1153,10 @@ udf_lookup(struct vop_cachedlookup_args *a) nameptr = a->a_cnp->cn_nameptr; namelen = a->a_cnp->cn_namelen; fsize = le64toh(node->fentry->inf_len); + if (fsize > INT_MAX) { + /* too big, just cap to INT_MAX */ + fsize = INT_MAX; + } /* * If this is a LOOKUP and we've already partially searched through diff --git a/sys/geom/concat/g_concat.c b/sys/geom/concat/g_concat.c index 2b1cb575cac8..2173a84c7acf 100644 --- a/sys/geom/concat/g_concat.c +++ b/sys/geom/concat/g_concat.c @@ -590,6 +590,7 @@ g_concat_add_disk(struct g_concat_softc *sc, struct g_provider *pp, u_int no) strcmp(md.md_name, sc->sc_name) != 0 || md.md_id != sc->sc_id) { G_CONCAT_DEBUG(0, "Metadata on %s changed.", pp->name); + error = EINVAL; goto fail; } diff --git a/sys/geom/geom.h b/sys/geom/geom.h index dcd6f793f9f7..908ce86f03a6 100644 --- a/sys/geom/geom.h +++ b/sys/geom/geom.h @@ -282,7 +282,7 @@ void g_detach(struct g_consumer *cp); void g_error_provider(struct g_provider *pp, int error); struct g_provider *g_provider_by_name(char const *arg); int g_getattr__(const char *attr, struct g_consumer *cp, void *var, int len); -#define g_getattr(a, c, v) g_getattr__((a), (c), (v), sizeof *(v)) +#define g_getattr(a, c, v) g_getattr__((a), (c), (v), sizeof(*(v))) int g_handleattr(struct bio *bp, const char *attribute, const void *val, int len); int g_handleattr_int(struct bio *bp, const char *attribute, int val); diff --git a/sys/geom/geom_ccd.c b/sys/geom/geom_ccd.c index 5700399ee5d1..2140d005160e 100644 --- a/sys/geom/geom_ccd.c +++ b/sys/geom/geom_ccd.c @@ -730,17 +730,17 @@ g_ccd_create(struct gctl_req *req, struct g_class *mp) int i, error; g_topology_assert(); - unit = gctl_get_paraml(req, "unit", sizeof (*unit)); + unit = gctl_get_paraml(req, "unit", sizeof(*unit)); if (unit == NULL) { gctl_error(req, "unit parameter not given"); return; } - ileave = gctl_get_paraml(req, "ileave", sizeof (*ileave)); + ileave = gctl_get_paraml(req, "ileave", sizeof(*ileave)); if (ileave == NULL) { gctl_error(req, "ileave parameter not given"); return; } - nprovider = gctl_get_paraml(req, "nprovider", sizeof (*nprovider)); + nprovider = gctl_get_paraml(req, "nprovider", sizeof(*nprovider)); if (nprovider == NULL) { gctl_error(req, "nprovider parameter not given"); return; @@ -769,7 +769,7 @@ g_ccd_create(struct gctl_req *req, struct g_class *mp) } gp = g_new_geomf(mp, "ccd%d", *unit); - sc = g_malloc(sizeof *sc, M_WAITOK | M_ZERO); + sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO); gp->softc = sc; sc->sc_ndisks = *nprovider; @@ -872,7 +872,7 @@ g_ccd_list(struct gctl_req *req, struct g_class *mp) struct g_geom *gp; int i, unit, *up; - up = gctl_get_paraml(req, "unit", sizeof (*up)); + up = gctl_get_paraml(req, "unit", sizeof(*up)); if (up == NULL) { gctl_error(req, "unit parameter not given"); return; diff --git a/sys/geom/geom_event.c b/sys/geom/geom_event.c index 0a76fd6c6f57..341233a6ef47 100644 --- a/sys/geom/geom_event.c +++ b/sys/geom/geom_event.c @@ -145,7 +145,7 @@ g_attr_changed(struct g_provider *pp, const char *attr, int flag) struct g_attrchanged_args *args; int error; - args = g_malloc(sizeof *args, flag); + args = g_malloc(sizeof(*args), flag); if (args == NULL) return (ENOMEM); args->pp = pp; diff --git a/sys/geom/geom_io.c b/sys/geom/geom_io.c index 8d6b9a926e1d..247a623bf1bf 100644 --- a/sys/geom/geom_io.c +++ b/sys/geom/geom_io.c @@ -278,7 +278,7 @@ g_io_init(void) g_bioq_init(&g_bio_run_down); g_bioq_init(&g_bio_run_up); - biozone = uma_zcreate("g_bio", sizeof (struct bio), + biozone = uma_zcreate("g_bio", sizeof(struct bio), NULL, NULL, NULL, NULL, 0, 0); diff --git a/sys/geom/geom_slice.c b/sys/geom/geom_slice.c index 8cfffc478849..0491b0069be4 100644 --- a/sys/geom/geom_slice.c +++ b/sys/geom/geom_slice.c @@ -57,7 +57,7 @@ g_slice_alloc(unsigned nslice, unsigned scsize) { struct g_slicer *gsp; - gsp = g_malloc(sizeof *gsp, M_WAITOK | M_ZERO); + gsp = g_malloc(sizeof(*gsp), M_WAITOK | M_ZERO); if (scsize > 0) gsp->softc = g_malloc(scsize, M_WAITOK | M_ZERO); else @@ -463,9 +463,9 @@ g_slice_conf_hot(struct g_geom *gp, u_int idx, off_t offset, off_t length, int r } gsl = gsp->hotspot; if(idx >= gsp->nhotspot) { - gsl2 = g_malloc((idx + 1) * sizeof *gsl2, M_WAITOK | M_ZERO); + gsl2 = g_malloc((idx + 1) * sizeof(*gsl2), M_WAITOK | M_ZERO); if (gsp->hotspot != NULL) - bcopy(gsp->hotspot, gsl2, gsp->nhotspot * sizeof *gsl2); + bcopy(gsp->hotspot, gsl2, gsp->nhotspot * sizeof(*gsl2)); gsp->hotspot = gsl2; if (gsp->hotspot != NULL) g_free(gsl); diff --git a/sys/geom/geom_subr.c b/sys/geom/geom_subr.c index 41cc115225f9..1429c84942ed 100644 --- a/sys/geom/geom_subr.c +++ b/sys/geom/geom_subr.c @@ -267,7 +267,7 @@ g_modevent(module_t mod, int type, void *data) switch (type) { case MOD_LOAD: g_trace(G_T_TOPOLOGY, "g_modevent(%s, LOAD)", mp->name); - hh = g_malloc(sizeof *hh, M_WAITOK | M_ZERO); + hh = g_malloc(sizeof(*hh), M_WAITOK | M_ZERO); hh->mp = mp; /* * Once the system is not cold, MOD_LOAD calls will be @@ -351,7 +351,7 @@ g_retaste(struct g_class *mp) if (mp->taste == NULL) return (EINVAL); - hh = g_malloc(sizeof *hh, M_WAITOK | M_ZERO); + hh = g_malloc(sizeof(*hh), M_WAITOK | M_ZERO); hh->mp = mp; if (cold) { @@ -381,8 +381,8 @@ g_new_geomf(struct g_class *mp, const char *fmt, ...) sbuf_vprintf(sb, fmt, ap); va_end(ap); sbuf_finish(sb); - gp = g_malloc(sizeof *gp, M_WAITOK | M_ZERO); - gp->name = g_malloc(sbuf_len(sb) + 1, M_WAITOK | M_ZERO); + gp = g_malloc(sizeof(*gp) + sbuf_len(sb) + 1, M_WAITOK | M_ZERO); + gp->name = (char *)(gp + 1); gp->class = mp; gp->rank = 1; LIST_INIT(&gp->consumer); @@ -420,7 +420,6 @@ g_destroy_geom(struct g_geom *gp) g_cancel_event(gp); LIST_REMOVE(gp, geom); TAILQ_REMOVE(&geoms, gp, geoms); - g_free(gp->name); g_free(gp); } @@ -528,7 +527,7 @@ g_new_consumer(struct g_geom *gp) ("g_new_consumer on geom(%s) (class %s) without orphan", gp->name, gp->class->name)); - cp = g_malloc(sizeof *cp, M_WAITOK | M_ZERO); + cp = g_malloc(sizeof(*cp), M_WAITOK | M_ZERO); cp->geom = gp; cp->stat = devstat_new_entry(cp, -1, 0, DEVSTAT_ALL_SUPPORTED, DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); @@ -617,7 +616,7 @@ g_new_providerf(struct g_geom *gp, const char *fmt, ...) sbuf_vprintf(sb, fmt, ap); va_end(ap); sbuf_finish(sb); - pp = g_malloc(sizeof *pp + sbuf_len(sb) + 1, M_WAITOK | M_ZERO); + pp = g_malloc(sizeof(*pp) + sbuf_len(sb) + 1, M_WAITOK | M_ZERO); pp->name = (char *)(pp + 1); strcpy(pp->name, sbuf_data(sb)); sbuf_delete(sb); @@ -749,7 +748,7 @@ g_resize_provider(struct g_provider *pp, off_t size) if (size == pp->mediasize) return; - hh = g_malloc(sizeof *hh, M_WAITOK | M_ZERO); + hh = g_malloc(sizeof(*hh), M_WAITOK | M_ZERO); hh->pp = pp; hh->size = size; g_post_event(g_resize_provider_event, hh, M_WAITOK, NULL); @@ -1083,21 +1082,21 @@ int g_handleattr_int(struct bio *bp, const char *attribute, int val) { - return (g_handleattr(bp, attribute, &val, sizeof val)); + return (g_handleattr(bp, attribute, &val, sizeof(val))); } int g_handleattr_uint16_t(struct bio *bp, const char *attribute, uint16_t val) { - return (g_handleattr(bp, attribute, &val, sizeof val)); + return (g_handleattr(bp, attribute, &val, sizeof(val))); } int g_handleattr_off_t(struct bio *bp, const char *attribute, off_t val) { - return (g_handleattr(bp, attribute, &val, sizeof val)); + return (g_handleattr(bp, attribute, &val, sizeof(val))); } int diff --git a/sys/geom/multipath/g_multipath.c b/sys/geom/multipath/g_multipath.c index 23088c895541..a4935df7eaa1 100644 --- a/sys/geom/multipath/g_multipath.c +++ b/sys/geom/multipath/g_multipath.c @@ -321,7 +321,7 @@ g_multipath_resize(struct g_consumer *cp) if (sc->sc_uuid[0] != 0) { pp = cp->provider; strlcpy(md.md_magic, G_MULTIPATH_MAGIC, sizeof(md.md_magic)); - memcpy(md.md_uuid, sc->sc_uuid, sizeof (sc->sc_uuid)); + memcpy(md.md_uuid, sc->sc_uuid, sizeof(sc->sc_uuid)); strlcpy(md.md_name, sc->sc_name, sizeof(md.md_name)); md.md_version = G_MULTIPATH_VERSION; md.md_size = size; @@ -552,8 +552,8 @@ g_multipath_create(struct g_class *mp, struct g_multipath_metadata *md) gp = g_new_geomf(mp, "%s", md->md_name); sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO); mtx_init(&sc->sc_mtx, "multipath", NULL, MTX_DEF); - memcpy(sc->sc_uuid, md->md_uuid, sizeof (sc->sc_uuid)); - memcpy(sc->sc_name, md->md_name, sizeof (sc->sc_name)); + memcpy(sc->sc_uuid, md->md_uuid, sizeof(sc->sc_uuid)); + memcpy(sc->sc_name, md->md_name, sizeof(sc->sc_name)); sc->sc_active_active = md->md_active_active; sc->sc_size = md->md_size; gp->softc = sc; @@ -906,7 +906,7 @@ g_multipath_taste(struct g_class *mp, struct g_provider *pp, int flags __unused) char buf[16]; u_long rand = random(); - snprintf(buf, sizeof (buf), "%s-%lu", md.md_name, rand); + snprintf(buf, sizeof(buf), "%s-%lu", md.md_name, rand); printf("GEOM_MULTIPATH: geom %s/%s exists already\n", sc->sc_name, sc->sc_uuid); printf("GEOM_MULTIPATH: %s will be (temporarily) %s\n", @@ -1200,7 +1200,7 @@ g_multipath_ctl_configure(struct gctl_req *req, struct g_class *mp) cp = sc->sc_active; pp = cp->provider; strlcpy(md.md_magic, G_MULTIPATH_MAGIC, sizeof(md.md_magic)); - memcpy(md.md_uuid, sc->sc_uuid, sizeof (sc->sc_uuid)); + memcpy(md.md_uuid, sc->sc_uuid, sizeof(sc->sc_uuid)); strlcpy(md.md_name, name, sizeof(md.md_name)); md.md_version = G_MULTIPATH_VERSION; md.md_size = pp->mediasize; diff --git a/sys/geom/virstor/g_virstor.c b/sys/geom/virstor/g_virstor.c index b8cf32875660..c7d737493f11 100644 --- a/sys/geom/virstor/g_virstor.c +++ b/sys/geom/virstor/g_virstor.c @@ -202,7 +202,7 @@ virstor_ctl_stop(struct gctl_req *req, struct g_class *cp) int *force, *nargs; int i; - nargs = gctl_get_paraml(req, "nargs", sizeof *nargs); + nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs)); if (nargs == NULL) { gctl_error(req, "Error fetching argument '%s'", "nargs"); return; @@ -211,7 +211,7 @@ virstor_ctl_stop(struct gctl_req *req, struct g_class *cp) gctl_error(req, "Invalid number of arguments"); return; } - force = gctl_get_paraml(req, "force", sizeof *force); + force = gctl_get_paraml(req, "force", sizeof(*force)); if (force == NULL) { gctl_error(req, "Error fetching argument '%s'", "force"); return; @@ -315,7 +315,7 @@ virstor_ctl_add(struct gctl_req *req, struct g_class *cp) u_int nc; u_int j; - snprintf(aname, sizeof aname, "arg%d", i); + snprintf(aname, sizeof(aname), "arg%d", i); pp = gctl_get_provider(req, aname); if (pp == NULL) { /* This is the most common error so be verbose about it */ @@ -487,12 +487,12 @@ fill_metadata(struct g_virstor_softc *sc, struct g_virstor_metadata *md, { struct g_virstor_component *c; - bzero(md, sizeof *md); + bzero(md, sizeof(*md)); c = &sc->components[nc]; - strncpy(md->md_magic, G_VIRSTOR_MAGIC, sizeof md->md_magic); + strncpy(md->md_magic, G_VIRSTOR_MAGIC, sizeof(md->md_magic)); md->md_version = G_VIRSTOR_VERSION; - strncpy(md->md_name, sc->geom->name, sizeof md->md_name); + strncpy(md->md_name, sc->geom->name, sizeof(md->md_name)); md->md_id = sc->id; md->md_virsize = sc->virsize; md->md_chunk_size = sc->chunk_size; @@ -500,7 +500,7 @@ fill_metadata(struct g_virstor_softc *sc, struct g_virstor_metadata *md, if (hardcode) { strncpy(md->provider, c->gcons->provider->name, - sizeof md->provider); + sizeof(md->provider)); } md->no = nc; md->provsize = c->gcons->provider->mediasize; @@ -589,7 +589,7 @@ virstor_ctl_remove(struct gctl_req *req, struct g_class *cp) M_GVIRSTOR, M_WAITOK | M_ZERO); bcopy(sc->components, newcomp, found * sizeof(*sc->components)); bcopy(&sc->components[found + 1], newcomp + found, - found * sizeof(*sc->components)); + (sc->n_components - (found + 1)) * sizeof(*sc->components)); if ((sc->components[j].flags & VIRSTOR_PROVIDER_ALLOCATED) != 0) { LOG_MSG(LVL_ERROR, "Allocated provider %s cannot be " "removed from %s", @@ -959,7 +959,7 @@ virstor_geom_destroy(struct g_virstor_softc *sc, boolean_t force, free(sc->map, M_GVIRSTOR); free(sc->components, M_GVIRSTOR); - bzero(sc, sizeof *sc); + bzero(sc, sizeof(*sc)); free(sc, M_GVIRSTOR); pp = LIST_FIRST(&gp->provider); /* We only offer one provider */ @@ -1213,7 +1213,7 @@ virstor_check_and_run(struct g_virstor_softc *sc) sc->provider->name, sc->chunk_count * (off_t)sc->chunk_size); } - sc->map_size = sc->chunk_count * sizeof *(sc->map); + sc->map_size = sc->chunk_count * sizeof(*(sc->map)); /* The following allocation is in order of 4MB - 8MB */ sc->map = malloc(sc->map_size, M_GVIRSTOR, M_WAITOK); KASSERT(sc->map != NULL, ("%s: Memory allocation error (%zu bytes) for %s", @@ -1267,7 +1267,7 @@ virstor_check_and_run(struct g_virstor_softc *sc) bcopy(mapbuf, &sc->map[n], bs); off += bs; count += bs; - n += bs / sizeof *(sc->map); + n += bs / sizeof(*(sc->map)); g_free(mapbuf); } g_access(sc->components[0].gcons, -1, 0, 0); @@ -1306,8 +1306,8 @@ virstor_check_and_run(struct g_virstor_softc *sc) sc->components[index].chunk_next); } - sc->me_per_sector = sc->sectorsize / sizeof *(sc->map); - if (sc->sectorsize % sizeof *(sc->map) != 0) { + sc->me_per_sector = sc->sectorsize / sizeof(*(sc->map)); + if (sc->sectorsize % sizeof(*(sc->map)) != 0) { LOG_MSG(LVL_ERROR, "%s: Map entries don't fit exactly in a sector (%s)", __func__, sc->geom->name); @@ -1653,7 +1653,7 @@ g_virstor_start(struct bio *b) * XXX: this will prevent the fs from * being umounted! */ struct g_virstor_bio_q *biq; - biq = malloc(sizeof *biq, M_GVIRSTOR, + biq = malloc(sizeof(*biq), M_GVIRSTOR, M_NOWAIT); if (biq == NULL) { bioq_dismantle(&bq); @@ -1703,7 +1703,7 @@ g_virstor_start(struct bio *b) * map array. * sc_offset will end up pointing to the drive * sector. */ - s_offset = chunk_index * sizeof *me; + s_offset = chunk_index * sizeof(*me); s_offset = rounddown(s_offset, sc->sectorsize); /* data_me points to map entry sector diff --git a/sys/i386/conf/GENERIC b/sys/i386/conf/GENERIC index e7d460af21d4..f577cd07ac7c 100644 --- a/sys/i386/conf/GENERIC +++ b/sys/i386/conf/GENERIC @@ -17,6 +17,8 @@ # in NOTES. # +#NO_UNIVERSE + cpu I486_CPU cpu I586_CPU cpu I686_CPU diff --git a/sys/i386/conf/GENERIC-NODEBUG b/sys/i386/conf/GENERIC-NODEBUG index ea07613a796f..a93304481b5f 100644 --- a/sys/i386/conf/GENERIC-NODEBUG +++ b/sys/i386/conf/GENERIC-NODEBUG @@ -25,6 +25,8 @@ # in NOTES. # +#NO_UNIVERSE + include GENERIC include "std.nodebug" diff --git a/sys/i386/conf/LINT b/sys/i386/conf/LINT index 41207eb63cb9..2e947202f723 100644 --- a/sys/i386/conf/LINT +++ b/sys/i386/conf/LINT @@ -1,3 +1,4 @@ +#NO_UNIVERSE include "../../conf/NOTES" include "../../x86/conf/NOTES" diff --git a/sys/i386/conf/MINIMAL b/sys/i386/conf/MINIMAL index 2a06eb84bff8..8019617ca4d4 100644 --- a/sys/i386/conf/MINIMAL +++ b/sys/i386/conf/MINIMAL @@ -31,6 +31,8 @@ # in NOTES. # +#NO_UNIVERSE + cpu I486_CPU cpu I586_CPU cpu I686_CPU diff --git a/sys/i386/conf/PAE b/sys/i386/conf/PAE index a39d32d77106..72af9e9a9eec 100644 --- a/sys/i386/conf/PAE +++ b/sys/i386/conf/PAE @@ -2,6 +2,8 @@ # PAE -- Generic kernel configuration file for FreeBSD/i386 PAE # +#NO_UNIVERSE + include GENERIC ident PAE-GENERIC diff --git a/sys/i386/i386/pmap.c b/sys/i386/i386/pmap.c index 465b4d0f365b..b44f5e08bbcf 100644 --- a/sys/i386/i386/pmap.c +++ b/sys/i386/i386/pmap.c @@ -876,14 +876,16 @@ __CONCAT(PMTYPE, init_pat)(void) #ifdef PMAP_PAE_COMP static void * -pmap_pdpt_allocf(uma_zone_t zone, vm_size_t bytes, int domain, uint8_t *flags, - int wait) +pmap_pdpt_allocf(uma_zone_t zone, vm_size_t bytes, int domain, uint8_t *sflagsp, + int flags) { /* Inform UMA that this allocator uses kernel_map/object. */ - *flags = UMA_SLAB_KERNEL; + *sflagsp = UMA_SLAB_KERNEL; + /* contig allocations cannot be NEVERFREED */ + flags &= ~M_NEVERFREED; return ((void *)kmem_alloc_contig_domainset(DOMAINSET_FIXED(domain), - bytes, wait, 0x0ULL, 0xffffffffULL, 1, 0, VM_MEMATTR_DEFAULT)); + bytes, flags, 0x0ULL, 0xffffffffULL, 1, 0, VM_MEMATTR_DEFAULT)); } #endif @@ -5617,6 +5619,8 @@ __CONCAT(PMTYPE, unmapdev)(void *p, vm_size_t size) static void __CONCAT(PMTYPE, page_set_memattr)(vm_page_t m, vm_memattr_t ma) { + if (m->md.pat_mode == ma) + return; m->md.pat_mode = ma; if ((m->flags & PG_FICTITIOUS) != 0) diff --git a/sys/kern/coredump_vnode.c b/sys/kern/coredump_vnode.c new file mode 100644 index 000000000000..8b857e9aa4a2 --- /dev/null +++ b/sys/kern/coredump_vnode.c @@ -0,0 +1,562 @@ +/* + * SPDX-License-Identifier: BSD-3-Clause AND BSD-2-Clause + * + * Copyright (c) 1982, 1986, 1989, 1991, 1993 + * The Regents of the University of California. All rights reserved. + * (c) UNIX System Laboratories, Inc. + * All or some portions of this file are derived from material licensed + * to the University of California by American Telephone and Telegraph + * Co. or Unix System Laboratories, Inc. and are reproduced herein with + * the permission of UNIX System Laboratories, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * - kern_sig.c + */ +/* + * Copyright (c) 1993, David Greenman + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * -kern_exec.c + */ + +#include <sys/systm.h> +#include <sys/acct.h> +#include <sys/compressor.h> +#include <sys/devctl.h> +#include <sys/fcntl.h> +#include <sys/jail.h> +#include <sys/limits.h> +#include <sys/namei.h> +#include <sys/proc.h> +#include <sys/sbuf.h> +#include <sys/stat.h> +#include <sys/sysctl.h> +#include <sys/sysent.h> +#include <sys/syslog.h> +#include <sys/ucoredump.h> +#include <sys/unistd.h> +#include <sys/vnode.h> + +#include <security/audit/audit.h> + +#define GZIP_SUFFIX ".gz" +#define ZSTD_SUFFIX ".zst" + +#define MAX_NUM_CORE_FILES 100000 +#ifndef NUM_CORE_FILES +#define NUM_CORE_FILES 5 +#endif + +static coredumper_handle_fn coredump_vnode; +static struct coredumper vnode_coredumper = { + .cd_name = "vnode_coredumper", + .cd_handle = coredump_vnode, +}; + +SYSINIT(vnode_coredumper_register, SI_SUB_EXEC, SI_ORDER_ANY, + coredumper_register, &vnode_coredumper); + +_Static_assert(NUM_CORE_FILES >= 0 && NUM_CORE_FILES <= MAX_NUM_CORE_FILES, + "NUM_CORE_FILES is out of range (0 to " __STRING(MAX_NUM_CORE_FILES) ")"); +static int num_cores = NUM_CORE_FILES; + +static int capmode_coredump; +SYSCTL_INT(_kern, OID_AUTO, capmode_coredump, CTLFLAG_RWTUN, + &capmode_coredump, 0, "Allow processes in capability mode to dump core"); + +static int set_core_nodump_flag = 0; +SYSCTL_INT(_kern, OID_AUTO, nodump_coredump, CTLFLAG_RW, &set_core_nodump_flag, + 0, "Enable setting the NODUMP flag on coredump files"); + +static int coredump_devctl = 0; +SYSCTL_INT(_kern, OID_AUTO, coredump_devctl, CTLFLAG_RW, &coredump_devctl, + 0, "Generate a devctl notification when processes coredump"); + +/* + * corefilename[] is protected by the allproc_lock. + */ +static char corefilename[MAXPATHLEN] = { "%N.core" }; +TUNABLE_STR("kern.corefile", corefilename, sizeof(corefilename)); + +static int +sysctl_kern_corefile(SYSCTL_HANDLER_ARGS) +{ + int error; + + sx_xlock(&allproc_lock); + error = sysctl_handle_string(oidp, corefilename, sizeof(corefilename), + req); + sx_xunlock(&allproc_lock); + + return (error); +} +SYSCTL_PROC(_kern, OID_AUTO, corefile, CTLTYPE_STRING | CTLFLAG_RW | + CTLFLAG_MPSAFE, 0, 0, sysctl_kern_corefile, "A", + "Process corefile name format string"); + +static int +sysctl_debug_num_cores_check (SYSCTL_HANDLER_ARGS) +{ + int error; + int new_val; + + new_val = num_cores; + error = sysctl_handle_int(oidp, &new_val, 0, req); + if (error != 0 || req->newptr == NULL) + return (error); + if (new_val > MAX_NUM_CORE_FILES) + new_val = MAX_NUM_CORE_FILES; + if (new_val < 0) + new_val = 0; + num_cores = new_val; + return (0); +} +SYSCTL_PROC(_debug, OID_AUTO, ncores, + CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, sizeof(int), + sysctl_debug_num_cores_check, "I", + "Maximum number of generated process corefiles while using index format"); + +static void +vnode_close_locked(struct thread *td, struct vnode *vp) +{ + + VOP_UNLOCK(vp); + vn_close(vp, FWRITE, td->td_ucred, td); +} + +int +core_vn_write(const struct coredump_writer *cdw, const void *base, size_t len, + off_t offset, enum uio_seg seg, struct ucred *cred, size_t *resid, + struct thread *td) +{ + struct coredump_vnode_ctx *ctx = cdw->ctx; + + return (vn_rdwr_inchunks(UIO_WRITE, ctx->vp, __DECONST(void *, base), + len, offset, seg, IO_UNIT | IO_DIRECT | IO_RANGELOCKED, + cred, ctx->fcred, resid, td)); +} + +int +core_vn_extend(const struct coredump_writer *cdw, off_t newsz, + struct ucred *cred) +{ + struct coredump_vnode_ctx *ctx = cdw->ctx; + struct mount *mp; + int error; + + error = vn_start_write(ctx->vp, &mp, V_WAIT); + if (error != 0) + return (error); + vn_lock(ctx->vp, LK_EXCLUSIVE | LK_RETRY); + error = vn_truncate_locked(ctx->vp, newsz, false, cred); + VOP_UNLOCK(ctx->vp); + vn_finished_write(mp); + return (error); +} + +/* + * If the core format has a %I in it, then we need to check + * for existing corefiles before defining a name. + * To do this we iterate over 0..ncores to find a + * non-existing core file name to use. If all core files are + * already used we choose the oldest one. + */ +static int +corefile_open_last(struct thread *td, char *name, int indexpos, + int indexlen, int ncores, struct vnode **vpp) +{ + struct vnode *oldvp, *nextvp, *vp; + struct vattr vattr; + struct nameidata nd; + int error, i, flags, oflags, cmode; + char ch; + struct timespec lasttime; + + nextvp = oldvp = NULL; + cmode = S_IRUSR | S_IWUSR; + oflags = VN_OPEN_NOAUDIT | VN_OPEN_NAMECACHE | + (capmode_coredump ? VN_OPEN_NOCAPCHECK : 0); + + for (i = 0; i < ncores; i++) { + flags = O_CREAT | FWRITE | O_NOFOLLOW; + + ch = name[indexpos + indexlen]; + (void)snprintf(name + indexpos, indexlen + 1, "%.*u", indexlen, + i); + name[indexpos + indexlen] = ch; + + NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, name); + error = vn_open_cred(&nd, &flags, cmode, oflags, td->td_ucred, + NULL); + if (error != 0) + break; + + vp = nd.ni_vp; + NDFREE_PNBUF(&nd); + if ((flags & O_CREAT) == O_CREAT) { + nextvp = vp; + break; + } + + error = VOP_GETATTR(vp, &vattr, td->td_ucred); + if (error != 0) { + vnode_close_locked(td, vp); + break; + } + + if (oldvp == NULL || + lasttime.tv_sec > vattr.va_mtime.tv_sec || + (lasttime.tv_sec == vattr.va_mtime.tv_sec && + lasttime.tv_nsec >= vattr.va_mtime.tv_nsec)) { + if (oldvp != NULL) + vn_close(oldvp, FWRITE, td->td_ucred, td); + oldvp = vp; + VOP_UNLOCK(oldvp); + lasttime = vattr.va_mtime; + } else { + vnode_close_locked(td, vp); + } + } + + if (oldvp != NULL) { + if (nextvp == NULL) { + if ((td->td_proc->p_flag & P_SUGID) != 0) { + error = EFAULT; + vn_close(oldvp, FWRITE, td->td_ucred, td); + } else { + nextvp = oldvp; + error = vn_lock(nextvp, LK_EXCLUSIVE); + if (error != 0) { + vn_close(nextvp, FWRITE, td->td_ucred, + td); + nextvp = NULL; + } + } + } else { + vn_close(oldvp, FWRITE, td->td_ucred, td); + } + } + if (error != 0) { + if (nextvp != NULL) + vnode_close_locked(td, oldvp); + } else { + *vpp = nextvp; + } + + return (error); +} + +/* + * corefile_open(comm, uid, pid, td, compress, vpp, namep) + * Expand the name described in corefilename, using name, uid, and pid + * and open/create core file. + * corefilename is a printf-like string, with three format specifiers: + * %N name of process ("name") + * %P process id (pid) + * %U user id (uid) + * For example, "%N.core" is the default; they can be disabled completely + * by using "/dev/null", or all core files can be stored in "/cores/%U/%N-%P". + * This is controlled by the sysctl variable kern.corefile (see above). + */ +static int +corefile_open(const char *comm, uid_t uid, pid_t pid, struct thread *td, + int compress, int signum, struct vnode **vpp, char **namep) +{ + struct sbuf sb; + struct nameidata nd; + const char *format; + char *hostname, *name; + int cmode, error, flags, i, indexpos, indexlen, oflags, ncores; + + hostname = NULL; + format = corefilename; + name = malloc(MAXPATHLEN, M_TEMP, M_WAITOK | M_ZERO); + indexlen = 0; + indexpos = -1; + ncores = num_cores; + (void)sbuf_new(&sb, name, MAXPATHLEN, SBUF_FIXEDLEN); + sx_slock(&allproc_lock); + for (i = 0; format[i] != '\0'; i++) { + switch (format[i]) { + case '%': /* Format character */ + i++; + switch (format[i]) { + case '%': + sbuf_putc(&sb, '%'); + break; + case 'H': /* hostname */ + if (hostname == NULL) { + hostname = malloc(MAXHOSTNAMELEN, + M_TEMP, M_WAITOK); + } + getcredhostname(td->td_ucred, hostname, + MAXHOSTNAMELEN); + sbuf_cat(&sb, hostname); + break; + case 'I': /* autoincrementing index */ + if (indexpos != -1) { + sbuf_printf(&sb, "%%I"); + break; + } + + indexpos = sbuf_len(&sb); + sbuf_printf(&sb, "%u", ncores - 1); + indexlen = sbuf_len(&sb) - indexpos; + break; + case 'N': /* process name */ + sbuf_printf(&sb, "%s", comm); + break; + case 'P': /* process id */ + sbuf_printf(&sb, "%u", pid); + break; + case 'S': /* signal number */ + sbuf_printf(&sb, "%i", signum); + break; + case 'U': /* user id */ + sbuf_printf(&sb, "%u", uid); + break; + default: + log(LOG_ERR, + "Unknown format character %c in " + "corename `%s'\n", format[i], format); + break; + } + break; + default: + sbuf_putc(&sb, format[i]); + break; + } + } + sx_sunlock(&allproc_lock); + free(hostname, M_TEMP); + if (compress == COMPRESS_GZIP) + sbuf_cat(&sb, GZIP_SUFFIX); + else if (compress == COMPRESS_ZSTD) + sbuf_cat(&sb, ZSTD_SUFFIX); + if (sbuf_error(&sb) != 0) { + log(LOG_ERR, "pid %ld (%s), uid (%lu): corename is too " + "long\n", (long)pid, comm, (u_long)uid); + sbuf_delete(&sb); + free(name, M_TEMP); + return (ENOMEM); + } + sbuf_finish(&sb); + sbuf_delete(&sb); + + if (indexpos != -1) { + error = corefile_open_last(td, name, indexpos, indexlen, ncores, + vpp); + if (error != 0) { + log(LOG_ERR, + "pid %d (%s), uid (%u): Path `%s' failed " + "on initial open test, error = %d\n", + pid, comm, uid, name, error); + } + } else { + cmode = S_IRUSR | S_IWUSR; + oflags = VN_OPEN_NOAUDIT | VN_OPEN_NAMECACHE | + (capmode_coredump ? VN_OPEN_NOCAPCHECK : 0); + flags = O_CREAT | FWRITE | O_NOFOLLOW; + if ((td->td_proc->p_flag & P_SUGID) != 0) + flags |= O_EXCL; + + NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, name); + error = vn_open_cred(&nd, &flags, cmode, oflags, td->td_ucred, + NULL); + if (error == 0) { + *vpp = nd.ni_vp; + NDFREE_PNBUF(&nd); + } + } + + if (error != 0) { +#ifdef AUDIT + audit_proc_coredump(td, name, error); +#endif + free(name, M_TEMP); + return (error); + } + *namep = name; + return (0); +} + +/* + * The vnode dumper is the traditional coredump handler. Our policy and limits + * are generally checked already, so it creates the coredump name and passes on + * a vnode and a size limit to the process-specific coredump routine if there is + * one. If there _is not_ one, it returns ENOSYS; otherwise it returns the + * error from the process-specific routine. + */ +static int +coredump_vnode(struct thread *td, off_t limit) +{ + struct proc *p = td->td_proc; + struct ucred *cred = td->td_ucred; + struct vnode *vp; + struct coredump_vnode_ctx wctx; + struct coredump_writer cdw = { }; + struct flock lf; + struct vattr vattr; + size_t fullpathsize; + int error, error1, jid, locked, ppid, sig; + char *name; /* name of corefile */ + void *rl_cookie; + char *fullpath, *freepath = NULL; + struct sbuf *sb; + + PROC_LOCK_ASSERT(p, MA_OWNED); + + ppid = p->p_oppid; + sig = p->p_sig; + jid = p->p_ucred->cr_prison->pr_id; + PROC_UNLOCK(p); + + error = corefile_open(p->p_comm, cred->cr_uid, p->p_pid, td, + compress_user_cores, sig, &vp, &name); + if (error != 0) + return (error); + + /* + * Don't dump to non-regular files or files with links. + * Do not dump into system files. Effective user must own the corefile. + */ + if (vp->v_type != VREG || VOP_GETATTR(vp, &vattr, cred) != 0 || + vattr.va_nlink != 1 || (vp->v_vflag & VV_SYSTEM) != 0 || + vattr.va_uid != cred->cr_uid) { + VOP_UNLOCK(vp); + error = EFAULT; + goto out; + } + + VOP_UNLOCK(vp); + + /* Postpone other writers, including core dumps of other processes. */ + rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX); + + lf.l_whence = SEEK_SET; + lf.l_start = 0; + lf.l_len = 0; + lf.l_type = F_WRLCK; + locked = (VOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &lf, F_FLOCK) == 0); + + VATTR_NULL(&vattr); + vattr.va_size = 0; + if (set_core_nodump_flag) + vattr.va_flags = UF_NODUMP; + vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); + VOP_SETATTR(vp, &vattr, cred); + VOP_UNLOCK(vp); + PROC_LOCK(p); + p->p_acflag |= ACORE; + PROC_UNLOCK(p); + + wctx.vp = vp; + wctx.fcred = NOCRED; + + cdw.ctx = &wctx; + cdw.write_fn = core_vn_write; + cdw.extend_fn = core_vn_extend; + + if (p->p_sysent->sv_coredump != NULL) { + error = p->p_sysent->sv_coredump(td, &cdw, limit, 0); + } else { + error = ENOSYS; + } + + if (locked) { + lf.l_type = F_UNLCK; + VOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_FLOCK); + } + vn_rangelock_unlock(vp, rl_cookie); + + /* + * Notify the userland helper that a process triggered a core dump. + * This allows the helper to run an automated debugging session. + */ + if (error != 0 || coredump_devctl == 0) + goto out; + sb = sbuf_new_auto(); + if (vn_fullpath_global(p->p_textvp, &fullpath, &freepath) != 0) + goto out2; + sbuf_cat(sb, "comm=\""); + devctl_safe_quote_sb(sb, fullpath); + free(freepath, M_TEMP); + sbuf_cat(sb, "\" core=\""); + + /* + * We can't lookup core file vp directly. When we're replacing a core, and + * other random times, we flush the name cache, so it will fail. Instead, + * if the path of the core is relative, add the current dir in front if it. + */ + if (name[0] != '/') { + fullpathsize = MAXPATHLEN; + freepath = malloc(fullpathsize, M_TEMP, M_WAITOK); + if (vn_getcwd(freepath, &fullpath, &fullpathsize) != 0) { + free(freepath, M_TEMP); + goto out2; + } + devctl_safe_quote_sb(sb, fullpath); + free(freepath, M_TEMP); + sbuf_putc(sb, '/'); + } + devctl_safe_quote_sb(sb, name); + sbuf_putc(sb, '"'); + + sbuf_printf(sb, " jid=%d pid=%d ppid=%d signo=%d", + jid, p->p_pid, ppid, sig); + if (sbuf_finish(sb) == 0) + devctl_notify("kernel", "signal", "coredump", sbuf_data(sb)); +out2: + sbuf_delete(sb); +out: + error1 = vn_close(vp, FWRITE, cred, td); + if (error == 0) + error = error1; +#ifdef AUDIT + audit_proc_coredump(td, name, error); +#endif + free(name, M_TEMP); + return (error); +} diff --git a/sys/kern/imgact_elf.c b/sys/kern/imgact_elf.c index b7ffbe68b483..2690ad3b2679 100644 --- a/sys/kern/imgact_elf.c +++ b/sys/kern/imgact_elf.c @@ -64,6 +64,7 @@ #include <sys/syscall.h> #include <sys/sysctl.h> #include <sys/sysent.h> +#include <sys/ucoredump.h> #include <sys/vnode.h> #include <sys/syslog.h> #include <sys/eventhandler.h> @@ -1562,9 +1563,6 @@ struct note_info { TAILQ_HEAD(note_info_list, note_info); -extern int compress_user_cores; -extern int compress_user_cores_level; - static void cb_put_phdr(vm_map_entry_t, void *); static void cb_size_segment(vm_map_entry_t, void *); static void each_dumpable_segment(struct thread *, segment_callback, void *, @@ -1595,7 +1593,7 @@ core_compressed_write(void *base, size_t len, off_t offset, void *arg) } int -__elfN(coredump)(struct thread *td, struct vnode *vp, off_t limit, int flags) +__elfN(coredump)(struct thread *td, struct coredump_writer *cdw, off_t limit, int flags) { struct ucred *cred = td->td_ucred; int compm, error = 0; @@ -1625,9 +1623,8 @@ __elfN(coredump)(struct thread *td, struct vnode *vp, off_t limit, int flags) /* Set up core dump parameters. */ params.offset = 0; params.active_cred = cred; - params.file_cred = NOCRED; params.td = td; - params.vp = vp; + params.cdw = cdw; params.comp = NULL; #ifdef RACCT @@ -1662,6 +1659,12 @@ __elfN(coredump)(struct thread *td, struct vnode *vp, off_t limit, int flags) tmpbuf = malloc(CORE_BUF_SIZE, M_TEMP, M_WAITOK | M_ZERO); } + if (cdw->init_fn != NULL) { + error = (*cdw->init_fn)(cdw, ¶ms); + if (error != 0) + goto done; + } + /* * Allocate memory for building the header, fill it up, * and write it out following the notes. diff --git a/sys/kern/kern_cpuset.c b/sys/kern/kern_cpuset.c index 5d9e2f2f326b..d7eb82d5f259 100644 --- a/sys/kern/kern_cpuset.c +++ b/sys/kern/kern_cpuset.c @@ -530,7 +530,7 @@ _domainset_create(struct domainset *domain, struct domainlist *freelist) * remove them and update the domainset accordingly. If only empty * domains are present, we must return failure. */ -static bool +bool domainset_empty_vm(struct domainset *domain) { domainset_t empty; @@ -2409,82 +2409,92 @@ sys_cpuset_setdomain(struct thread *td, struct cpuset_setdomain_args *uap) } int -kern_cpuset_setdomain(struct thread *td, cpulevel_t level, cpuwhich_t which, - id_t id, size_t domainsetsize, const domainset_t *maskp, int policy, - const struct cpuset_copy_cb *cb) +domainset_populate(struct domainset *domain, const domainset_t *mask, int policy, + size_t mask_size) { - struct cpuset *nset; - struct cpuset *set; - struct thread *ttd; - struct proc *p; - struct domainset domain; - domainset_t *mask; - int error; - if (domainsetsize < sizeof(domainset_t) || - domainsetsize > DOMAINSET_MAXSIZE / NBBY) - return (ERANGE); if (policy <= DOMAINSET_POLICY_INVALID || - policy > DOMAINSET_POLICY_MAX) + policy > DOMAINSET_POLICY_MAX) { return (EINVAL); - error = cpuset_check_capabilities(td, level, which, id); - if (error != 0) - return (error); - memset(&domain, 0, sizeof(domain)); - mask = malloc(domainsetsize, M_TEMP, M_WAITOK | M_ZERO); - error = cb->cpuset_copyin(maskp, mask, domainsetsize); - if (error) - goto out; + } + /* * Verify that no high bits are set. */ - if (domainsetsize > sizeof(domainset_t)) { - char *end; - char *cp; + if (mask_size > sizeof(domainset_t)) { + const char *end; + const char *cp; - end = cp = (char *)&mask->__bits; - end += domainsetsize; + end = cp = (const char *)&mask->__bits; + end += mask_size; cp += sizeof(domainset_t); - while (cp != end) + while (cp != end) { if (*cp++ != 0) { - error = EINVAL; - goto out; + return (EINVAL); } + } } if (DOMAINSET_EMPTY(mask)) { - error = EDEADLK; - goto out; + return (EDEADLK); } - DOMAINSET_COPY(mask, &domain.ds_mask); - domain.ds_policy = policy; + DOMAINSET_COPY(mask, &domain->ds_mask); + domain->ds_policy = policy; /* * Sanitize the provided mask. */ - if (!DOMAINSET_SUBSET(&all_domains, &domain.ds_mask)) { - error = EINVAL; - goto out; + if (!DOMAINSET_SUBSET(&all_domains, &domain->ds_mask)) { + return (EINVAL); } /* Translate preferred policy into a mask and fallback. */ if (policy == DOMAINSET_POLICY_PREFER) { /* Only support a single preferred domain. */ - if (DOMAINSET_COUNT(&domain.ds_mask) != 1) { - error = EINVAL; - goto out; + if (DOMAINSET_COUNT(&domain->ds_mask) != 1) { + return (EINVAL); } - domain.ds_prefer = DOMAINSET_FFS(&domain.ds_mask) - 1; + domain->ds_prefer = DOMAINSET_FFS(&domain->ds_mask) - 1; /* This will be constrained by domainset_shadow(). */ - DOMAINSET_COPY(&all_domains, &domain.ds_mask); + DOMAINSET_COPY(&all_domains, &domain->ds_mask); } + return (0); +} + +int +kern_cpuset_setdomain(struct thread *td, cpulevel_t level, cpuwhich_t which, + id_t id, size_t domainsetsize, const domainset_t *maskp, int policy, + const struct cpuset_copy_cb *cb) +{ + struct cpuset *nset; + struct cpuset *set; + struct thread *ttd; + struct proc *p; + struct domainset domain; + domainset_t *mask; + int error; + + error = cpuset_check_capabilities(td, level, which, id); + if (error != 0) + return (error); + if (domainsetsize < sizeof(domainset_t) || + domainsetsize > DOMAINSET_MAXSIZE / NBBY) + return (ERANGE); + memset(&domain, 0, sizeof(domain)); + mask = malloc(domainsetsize, M_TEMP, M_WAITOK | M_ZERO); + error = cb->cpuset_copyin(maskp, mask, domainsetsize); + if (error) + goto out; + error = domainset_populate(&domain, mask, policy, domainsetsize); + if (error) + goto out; + /* * When given an impossible policy, fall back to interleaving * across all domains. */ if (domainset_empty_vm(&domain)) domainset_copy(domainset2, &domain); - switch (level) { case CPU_LEVEL_ROOT: case CPU_LEVEL_CPUSET: diff --git a/sys/kern/kern_descrip.c b/sys/kern/kern_descrip.c index 406236fc2723..a27ab33b34da 100644 --- a/sys/kern/kern_descrip.c +++ b/sys/kern/kern_descrip.c @@ -480,6 +480,92 @@ kern_fcntl_freebsd(struct thread *td, int fd, int cmd, intptr_t arg) return (error); } +struct flags_trans_elem { + u_int f; + u_int t; +}; + +static u_int +flags_trans(const struct flags_trans_elem *ftes, int nitems, u_int from_flags) +{ + u_int res; + int i; + + res = 0; + for (i = 0; i < nitems; i++) { + if ((from_flags & ftes[i].f) != 0) + res |= ftes[i].t; + } + return (res); +} + +static uint8_t +fd_to_fde_flags(int fd_flags) +{ + static const struct flags_trans_elem fd_to_fde_flags_s[] = { + { .f = FD_CLOEXEC, .t = UF_EXCLOSE }, + { .f = FD_CLOFORK, .t = UF_FOCLOSE }, + { .f = FD_RESOLVE_BENEATH, .t = UF_RESOLVE_BENEATH }, + }; + + return (flags_trans(fd_to_fde_flags_s, nitems(fd_to_fde_flags_s), + fd_flags)); +} + +static int +fde_to_fd_flags(uint8_t fde_flags) +{ + static const struct flags_trans_elem fde_to_fd_flags_s[] = { + { .f = UF_EXCLOSE, .t = FD_CLOEXEC }, + { .f = UF_FOCLOSE, .t = FD_CLOFORK }, + { .f = UF_RESOLVE_BENEATH, .t = FD_RESOLVE_BENEATH }, + }; + + return (flags_trans(fde_to_fd_flags_s, nitems(fde_to_fd_flags_s), + fde_flags)); +} + +static uint8_t +fddup_to_fde_flags(int fddup_flags) +{ + static const struct flags_trans_elem fddup_to_fde_flags_s[] = { + { .f = FDDUP_FLAG_CLOEXEC, .t = UF_EXCLOSE }, + { .f = FDDUP_FLAG_CLOFORK, .t = UF_FOCLOSE }, + }; + + return (flags_trans(fddup_to_fde_flags_s, nitems(fddup_to_fde_flags_s), + fddup_flags)); +} + +static uint8_t +close_range_to_fde_flags(int close_range_flags) +{ + static const struct flags_trans_elem close_range_to_fde_flags_s[] = { + { .f = CLOSE_RANGE_CLOEXEC, .t = UF_EXCLOSE }, + { .f = CLOSE_RANGE_CLOFORK, .t = UF_FOCLOSE }, + }; + + return (flags_trans(close_range_to_fde_flags_s, + nitems(close_range_to_fde_flags_s), close_range_flags)); +} + +static uint8_t +open_to_fde_flags(int open_flags, bool sticky_orb) +{ + static const struct flags_trans_elem open_to_fde_flags_s[] = { + { .f = O_CLOEXEC, .t = UF_EXCLOSE }, + { .f = O_CLOFORK, .t = UF_FOCLOSE }, + { .f = O_RESOLVE_BENEATH, .t = UF_RESOLVE_BENEATH }, + }; +#if defined(__clang__) && __clang_major__ >= 19 + _Static_assert(open_to_fde_flags_s[nitems(open_to_fde_flags_s) - 1].f == + O_RESOLVE_BENEATH, "O_RESOLVE_BENEATH must be last, for sticky_orb"); +#endif + + return (flags_trans(open_to_fde_flags_s, nitems(open_to_fde_flags_s) - + (sticky_orb ? 0 : 1), open_flags)); +} + int kern_fcntl(struct thread *td, int fd, int cmd, intptr_t arg) { @@ -534,11 +620,7 @@ kern_fcntl(struct thread *td, int fd, int cmd, intptr_t arg) FILEDESC_SLOCK(fdp); fde = fdeget_noref(fdp, fd); if (fde != NULL) { - td->td_retval[0] = - ((fde->fde_flags & UF_EXCLOSE) ? FD_CLOEXEC : 0) | - ((fde->fde_flags & UF_FOCLOSE) ? FD_CLOFORK : 0) | - ((fde->fde_flags & UF_RESOLVE_BENEATH) ? - FD_RESOLVE_BENEATH : 0); + td->td_retval[0] = fde_to_fd_flags(fde->fde_flags); error = 0; } FILEDESC_SUNLOCK(fdp); @@ -552,11 +634,8 @@ kern_fcntl(struct thread *td, int fd, int cmd, intptr_t arg) /* * UF_RESOLVE_BENEATH is sticky and cannot be cleared. */ - fde->fde_flags = (fde->fde_flags & ~UF_EXCLOSE) | - ((arg & FD_CLOEXEC) != 0 ? UF_EXCLOSE : 0) | - ((arg & FD_CLOFORK) != 0 ? UF_FOCLOSE : 0) | - ((arg & FD_RESOLVE_BENEATH) != 0 ? - UF_RESOLVE_BENEATH : 0); + fde->fde_flags = (fde->fde_flags & + ~(UF_EXCLOSE | UF_FOCLOSE)) | fd_to_fde_flags(arg); error = 0; } FILEDESC_XUNLOCK(fdp); @@ -991,10 +1070,7 @@ kern_dup(struct thread *td, u_int mode, int flags, int old, int new) goto unlock; if (mode == FDDUP_FIXED && old == new) { td->td_retval[0] = new; - if ((flags & FDDUP_FLAG_CLOEXEC) != 0) - fdp->fd_ofiles[new].fde_flags |= UF_EXCLOSE; - if ((flags & FDDUP_FLAG_CLOFORK) != 0) - fdp->fd_ofiles[new].fde_flags |= UF_FOCLOSE; + fdp->fd_ofiles[new].fde_flags |= fddup_to_fde_flags(flags); error = 0; goto unlock; } @@ -1070,8 +1146,7 @@ kern_dup(struct thread *td, u_int mode, int flags, int old, int new) filecaps_copy_finish(&oldfde->fde_caps, &newfde->fde_caps, nioctls); newfde->fde_flags = (oldfde->fde_flags & ~(UF_EXCLOSE | UF_FOCLOSE)) | - ((flags & FDDUP_FLAG_CLOEXEC) != 0 ? UF_EXCLOSE : 0) | - ((flags & FDDUP_FLAG_CLOFORK) != 0 ? UF_FOCLOSE : 0); + fddup_to_fde_flags(flags); #ifdef CAPABILITIES seqc_write_end(&newfde->fde_seqc); #endif @@ -1444,8 +1519,7 @@ close_range_flags(struct thread *td, u_int lowfd, u_int highfd, int flags) struct filedescent *fde; int fd, fde_flags; - fde_flags = ((flags & CLOSE_RANGE_CLOEXEC) != 0 ? UF_EXCLOSE : 0) | - ((flags & CLOSE_RANGE_CLOFORK) != 0 ? UF_FOCLOSE : 0); + fde_flags = close_range_to_fde_flags(flags); fdp = td->td_proc->p_fd; FILEDESC_XLOCK(fdp); fdt = atomic_load_ptr(&fdp->fd_files); @@ -2194,9 +2268,7 @@ _finstall(struct filedesc *fdp, struct file *fp, int fd, int flags, seqc_write_begin(&fde->fde_seqc); #endif fde->fde_file = fp; - fde->fde_flags = ((flags & O_CLOEXEC) != 0 ? UF_EXCLOSE : 0) | - ((flags & O_CLOFORK) != 0 ? UF_FOCLOSE : 0) | - ((flags & O_RESOLVE_BENEATH) != 0 ? UF_RESOLVE_BENEATH : 0); + fde->fde_flags = open_to_fde_flags(flags, true); if (fcaps != NULL) filecaps_move(fcaps, &fde->fde_caps); else diff --git a/sys/kern/kern_exec.c b/sys/kern/kern_exec.c index 03268365891e..0fc2d0e7f1bc 100644 --- a/sys/kern/kern_exec.c +++ b/sys/kern/kern_exec.c @@ -70,6 +70,7 @@ #include <sys/sysent.h> #include <sys/sysproto.h> #include <sys/timers.h> +#include <sys/ucoredump.h> #include <sys/umtxvar.h> #include <sys/vnode.h> #include <sys/wait.h> @@ -2002,10 +2003,14 @@ int core_write(struct coredump_params *cp, const void *base, size_t len, off_t offset, enum uio_seg seg, size_t *resid) { + return ((*cp->cdw->write_fn)(cp->cdw, base, len, offset, seg, + cp->active_cred, resid, cp->td)); +} - return (vn_rdwr_inchunks(UIO_WRITE, cp->vp, __DECONST(void *, base), - len, offset, seg, IO_UNIT | IO_DIRECT | IO_RANGELOCKED, - cp->active_cred, cp->file_cred, resid, cp->td)); +static int +core_extend(struct coredump_params *cp, off_t newsz) +{ + return ((*cp->cdw->extend_fn)(cp->cdw, newsz, cp->active_cred)); } int @@ -2013,7 +2018,6 @@ core_output(char *base, size_t len, off_t offset, struct coredump_params *cp, void *tmpbuf) { vm_map_t map; - struct mount *mp; size_t resid, runlen; int error; bool success; @@ -2068,14 +2072,7 @@ core_output(char *base, size_t len, off_t offset, struct coredump_params *cp, } } if (!success) { - error = vn_start_write(cp->vp, &mp, V_WAIT); - if (error != 0) - break; - vn_lock(cp->vp, LK_EXCLUSIVE | LK_RETRY); - error = vn_truncate_locked(cp->vp, offset + runlen, - false, cp->td->td_ucred); - VOP_UNLOCK(cp->vp); - vn_finished_write(mp); + error = core_extend(cp, offset + runlen); if (error != 0) break; } diff --git a/sys/kern/kern_jail.c b/sys/kern/kern_jail.c index d4529e096929..7ef1d19f0ea8 100644 --- a/sys/kern/kern_jail.c +++ b/sys/kern/kern_jail.c @@ -3466,7 +3466,7 @@ prison_check_af(struct ucred *cred, int af) pr = cred->cr_prison; #ifdef VIMAGE /* Prisons with their own network stack are not limited. */ - if (prison_owns_vnet(cred)) + if (prison_owns_vnet(pr)) return (0); #endif @@ -3531,7 +3531,7 @@ prison_if(struct ucred *cred, const struct sockaddr *sa) KASSERT(sa != NULL, ("%s: sa is NULL", __func__)); #ifdef VIMAGE - if (prison_owns_vnet(cred)) + if (prison_owns_vnet(cred->cr_prison)) return (0); #endif @@ -3648,7 +3648,7 @@ jailed_without_vnet(struct ucred *cred) if (!jailed(cred)) return (false); #ifdef VIMAGE - if (prison_owns_vnet(cred)) + if (prison_owns_vnet(cred->cr_prison)) return (false); #endif @@ -3711,20 +3711,17 @@ getjailname(struct ucred *cred, char *name, size_t len) #ifdef VIMAGE /* - * Determine whether the prison represented by cred owns - * its vnet rather than having it inherited. - * - * Returns true in case the prison owns the vnet, false otherwise. + * Determine whether the prison owns its VNET. */ bool -prison_owns_vnet(struct ucred *cred) +prison_owns_vnet(struct prison *pr) { /* * vnets cannot be added/removed after jail creation, * so no need to lock here. */ - return ((cred->cr_prison->pr_flags & PR_VNET) != 0); + return ((pr->pr_flags & PR_VNET) != 0); } #endif @@ -4425,7 +4422,7 @@ sysctl_jail_vnet(SYSCTL_HANDLER_ARGS) #ifdef VIMAGE struct ucred *cred = req->td->td_ucred; - havevnet = jailed(cred) && prison_owns_vnet(cred); + havevnet = jailed(cred) && prison_owns_vnet(cred->cr_prison); #else havevnet = 0; #endif diff --git a/sys/kern/kern_prot.c b/sys/kern/kern_prot.c index d9aeec68e620..0f0bc056cafd 100644 --- a/sys/kern/kern_prot.c +++ b/sys/kern/kern_prot.c @@ -287,7 +287,7 @@ sys_getgid(struct thread *td, struct getgid_args *uap) td->td_retval[0] = td->td_ucred->cr_rgid; #if defined(COMPAT_43) - td->td_retval[1] = td->td_ucred->cr_groups[0]; + td->td_retval[1] = td->td_ucred->cr_gid; #endif return (0); } @@ -307,7 +307,7 @@ int sys_getegid(struct thread *td, struct getegid_args *uap) { - td->td_retval[0] = td->td_ucred->cr_groups[0]; + td->td_retval[0] = td->td_ucred->cr_gid; return (0); } @@ -1080,7 +1080,7 @@ sys_setgid(struct thread *td, struct setgid_args *uap) gid != oldcred->cr_svgid && /* allow setgid(saved gid) */ #endif #ifdef POSIX_APPENDIX_B_4_2_2 /* Use BSD-compat clause from B.4.2.2 */ - gid != oldcred->cr_groups[0] && /* allow setgid(getegid()) */ + gid != oldcred->cr_gid && /* allow setgid(getegid()) */ #endif (error = priv_check_cred(oldcred, PRIV_CRED_SETGID)) != 0) goto fail; @@ -1092,7 +1092,7 @@ sys_setgid(struct thread *td, struct setgid_args *uap) */ if ( #ifdef POSIX_APPENDIX_B_4_2_2 /* use the clause from B.4.2.2 */ - gid == oldcred->cr_groups[0] || + gid == oldcred->cr_gid || #endif /* We are using privs. */ priv_check_cred(oldcred, PRIV_CRED_SETGID) == 0) @@ -1121,7 +1121,7 @@ sys_setgid(struct thread *td, struct setgid_args *uap) * In all cases permitted cases, we are changing the egid. * Copy credentials so other references do not see our changes. */ - if (oldcred->cr_groups[0] != gid) { + if (oldcred->cr_gid != gid) { change_egid(newcred, gid); setsugid(p); } @@ -1167,7 +1167,7 @@ sys_setegid(struct thread *td, struct setegid_args *uap) (error = priv_check_cred(oldcred, PRIV_CRED_SETEGID)) != 0) goto fail; - if (oldcred->cr_groups[0] != egid) { + if (oldcred->cr_gid != egid) { change_egid(newcred, egid); setsugid(p); } @@ -1393,12 +1393,12 @@ sys_setregid(struct thread *td, struct setregid_args *uap) if (((rgid != (gid_t)-1 && rgid != oldcred->cr_rgid && rgid != oldcred->cr_svgid) || - (egid != (gid_t)-1 && egid != oldcred->cr_groups[0] && + (egid != (gid_t)-1 && egid != oldcred->cr_gid && egid != oldcred->cr_rgid && egid != oldcred->cr_svgid)) && (error = priv_check_cred(oldcred, PRIV_CRED_SETREGID)) != 0) goto fail; - if (egid != (gid_t)-1 && oldcred->cr_groups[0] != egid) { + if (egid != (gid_t)-1 && oldcred->cr_gid != egid) { change_egid(newcred, egid); setsugid(p); } @@ -1406,9 +1406,9 @@ sys_setregid(struct thread *td, struct setregid_args *uap) change_rgid(newcred, rgid); setsugid(p); } - if ((rgid != (gid_t)-1 || newcred->cr_groups[0] != newcred->cr_rgid) && - newcred->cr_svgid != newcred->cr_groups[0]) { - change_svgid(newcred, newcred->cr_groups[0]); + if ((rgid != (gid_t)-1 || newcred->cr_gid != newcred->cr_rgid) && + newcred->cr_svgid != newcred->cr_gid) { + change_svgid(newcred, newcred->cr_gid); setsugid(p); } proc_set_cred(p, newcred); @@ -1547,17 +1547,17 @@ sys_setresgid(struct thread *td, struct setresgid_args *uap) if (((rgid != (gid_t)-1 && rgid != oldcred->cr_rgid && rgid != oldcred->cr_svgid && - rgid != oldcred->cr_groups[0]) || + rgid != oldcred->cr_gid) || (egid != (gid_t)-1 && egid != oldcred->cr_rgid && egid != oldcred->cr_svgid && - egid != oldcred->cr_groups[0]) || + egid != oldcred->cr_gid) || (sgid != (gid_t)-1 && sgid != oldcred->cr_rgid && sgid != oldcred->cr_svgid && - sgid != oldcred->cr_groups[0])) && + sgid != oldcred->cr_gid)) && (error = priv_check_cred(oldcred, PRIV_CRED_SETRESGID)) != 0) goto fail; - if (egid != (gid_t)-1 && oldcred->cr_groups[0] != egid) { + if (egid != (gid_t)-1 && oldcred->cr_gid != egid) { change_egid(newcred, egid); setsugid(p); } @@ -1626,8 +1626,8 @@ sys_getresgid(struct thread *td, struct getresgid_args *uap) error1 = copyout(&cred->cr_rgid, uap->rgid, sizeof(cred->cr_rgid)); if (uap->egid) - error2 = copyout(&cred->cr_groups[0], - uap->egid, sizeof(cred->cr_groups[0])); + error2 = copyout(&cred->cr_gid, + uap->egid, sizeof(cred->cr_gid)); if (uap->sgid) error3 = copyout(&cred->cr_svgid, uap->sgid, sizeof(cred->cr_svgid)); @@ -1737,7 +1737,7 @@ groupmember(gid_t gid, const struct ucred *cred) groups_check_positive_len(cred->cr_ngroups); - if (gid == cred->cr_groups[0]) + if (gid == cred->cr_gid) return (true); return (group_is_supplementary(gid, cred)); @@ -3015,7 +3015,7 @@ void change_egid(struct ucred *newcred, gid_t egid) { - newcred->cr_groups[0] = egid; + newcred->cr_gid = egid; } /*- diff --git a/sys/kern/kern_sendfile.c b/sys/kern/kern_sendfile.c index 35b258e68701..8438298afc0e 100644 --- a/sys/kern/kern_sendfile.c +++ b/sys/kern/kern_sendfile.c @@ -698,10 +698,13 @@ sendfile_wait_generic(struct socket *so, off_t need, int *space) */ error = 0; SOCK_SENDBUF_LOCK(so); - if (so->so_snd.sb_lowat < so->so_snd.sb_hiwat / 2) - so->so_snd.sb_lowat = so->so_snd.sb_hiwat / 2; - if (so->so_snd.sb_lowat < PAGE_SIZE && so->so_snd.sb_hiwat >= PAGE_SIZE) - so->so_snd.sb_lowat = PAGE_SIZE; + if (so->so_snd.sb_flags & SB_AUTOLOWAT) { + if (so->so_snd.sb_lowat < so->so_snd.sb_hiwat / 2) + so->so_snd.sb_lowat = so->so_snd.sb_hiwat / 2; + if (so->so_snd.sb_lowat < PAGE_SIZE && + so->so_snd.sb_hiwat >= PAGE_SIZE) + so->so_snd.sb_lowat = PAGE_SIZE; + } retry_space: if (so->so_snd.sb_state & SBS_CANTSENDMORE) { error = EPIPE; diff --git a/sys/kern/kern_sig.c b/sys/kern/kern_sig.c index 5d51aa675cb7..da0efac0598d 100644 --- a/sys/kern/kern_sig.c +++ b/sys/kern/kern_sig.c @@ -45,10 +45,10 @@ #include <sys/vnode.h> #include <sys/acct.h> #include <sys/capsicum.h> -#include <sys/compressor.h> #include <sys/condvar.h> #include <sys/devctl.h> #include <sys/event.h> +#include <sys/exec.h> #include <sys/fcntl.h> #include <sys/imgact.h> #include <sys/jail.h> @@ -80,6 +80,7 @@ #include <sys/syslog.h> #include <sys/sysproto.h> #include <sys/timers.h> +#include <sys/ucoredump.h> #include <sys/unistd.h> #include <sys/vmmeter.h> #include <sys/wait.h> @@ -101,7 +102,6 @@ SDT_PROBE_DEFINE2(proc, , , signal__clear, SDT_PROBE_DEFINE3(proc, , , signal__discard, "struct thread *", "struct proc *", "int"); -static int coredump(struct thread *); static int killpg1(struct thread *td, int sig, int pgid, int all, ksiginfo_t *ksi); static int issignal(struct thread *td); @@ -126,11 +126,6 @@ const struct filterops sig_filtops = { .f_event = filt_signal, }; -static int kern_logsigexit = 1; -SYSCTL_INT(_kern, KERN_LOGSIGEXIT, logsigexit, CTLFLAG_RW, - &kern_logsigexit, 0, - "Log processes quitting on abnormal signals to syslog(3)"); - static int kern_forcesigexit = 1; SYSCTL_INT(_kern, OID_AUTO, forcesigexit, CTLFLAG_RW, &kern_forcesigexit, 0, "Force trap signal to be handled"); @@ -193,26 +188,6 @@ SYSINIT(signal, SI_SUB_P1003_1B, SI_ORDER_FIRST+3, sigqueue_start, NULL); (cr1)->cr_ruid == (cr2)->cr_uid || \ (cr1)->cr_uid == (cr2)->cr_uid) -static int sugid_coredump; -SYSCTL_INT(_kern, OID_AUTO, sugid_coredump, CTLFLAG_RWTUN, - &sugid_coredump, 0, "Allow setuid and setgid processes to dump core"); - -static int capmode_coredump; -SYSCTL_INT(_kern, OID_AUTO, capmode_coredump, CTLFLAG_RWTUN, - &capmode_coredump, 0, "Allow processes in capability mode to dump core"); - -static int do_coredump = 1; -SYSCTL_INT(_kern, OID_AUTO, coredump, CTLFLAG_RW, - &do_coredump, 0, "Enable/Disable coredumps"); - -static int set_core_nodump_flag = 0; -SYSCTL_INT(_kern, OID_AUTO, nodump_coredump, CTLFLAG_RW, &set_core_nodump_flag, - 0, "Enable setting the NODUMP flag on coredump files"); - -static int coredump_devctl = 0; -SYSCTL_INT(_kern, OID_AUTO, coredump_devctl, CTLFLAG_RW, &coredump_devctl, - 0, "Generate a devctl notification when processes coredump"); - /* * Signal properties and actions. * The array below categorizes the signals and their default actions @@ -784,6 +759,13 @@ sigprop(int sig) return (0); } +bool +sig_do_core(int sig) +{ + + return ((sigprop(sig) & SIGPROP_CORE) != 0); +} + static bool sigact_flag_test(const struct sigaction *act, int flag) { @@ -2665,6 +2647,8 @@ static void ptrace_coredumpreq(struct thread *td, struct proc *p, struct thr_coredump_req *tcq) { + struct coredump_vnode_ctx wctx; + struct coredump_writer cdw; void *rl_cookie; if (p->p_sysent->sv_coredump == NULL) { @@ -2672,8 +2656,15 @@ ptrace_coredumpreq(struct thread *td, struct proc *p, return; } + wctx.vp = tcq->tc_vp; + wctx.fcred = NOCRED; + + cdw.ctx = &wctx; + cdw.write_fn = core_vn_write; + cdw.extend_fn = core_vn_extend; + rl_cookie = vn_rangelock_wlock(tcq->tc_vp, 0, OFF_MAX); - tcq->tc_error = p->p_sysent->sv_coredump(td, tcq->tc_vp, + tcq->tc_error = p->p_sysent->sv_coredump(td, &cdw, tcq->tc_limit, tcq->tc_flags); vn_rangelock_unlock(tcq->tc_vp, rl_cookie); } @@ -3635,82 +3626,6 @@ killproc(struct proc *p, const char *why) } /* - * Force the current process to exit with the specified signal, dumping core - * if appropriate. We bypass the normal tests for masked and caught signals, - * allowing unrecoverable failures to terminate the process without changing - * signal state. Mark the accounting record with the signal termination. - * If dumping core, save the signal number for the debugger. Calls exit and - * does not return. - */ -void -sigexit(struct thread *td, int sig) -{ - struct proc *p = td->td_proc; - const char *coreinfo; - int rv; - bool logexit; - - PROC_LOCK_ASSERT(p, MA_OWNED); - proc_set_p2_wexit(p); - - p->p_acflag |= AXSIG; - if ((p->p_flag2 & P2_LOGSIGEXIT_CTL) == 0) - logexit = kern_logsigexit != 0; - else - logexit = (p->p_flag2 & P2_LOGSIGEXIT_ENABLE) != 0; - - /* - * We must be single-threading to generate a core dump. This - * ensures that the registers in the core file are up-to-date. - * Also, the ELF dump handler assumes that the thread list doesn't - * change out from under it. - * - * XXX If another thread attempts to single-thread before us - * (e.g. via fork()), we won't get a dump at all. - */ - if ((sigprop(sig) & SIGPROP_CORE) && - thread_single(p, SINGLE_NO_EXIT) == 0) { - p->p_sig = sig; - /* - * Log signals which would cause core dumps - * (Log as LOG_INFO to appease those who don't want - * these messages.) - * XXX : Todo, as well as euid, write out ruid too - * Note that coredump() drops proc lock. - */ - rv = coredump(td); - switch (rv) { - case 0: - sig |= WCOREFLAG; - coreinfo = " (core dumped)"; - break; - case EFAULT: - coreinfo = " (no core dump - bad address)"; - break; - case EINVAL: - coreinfo = " (no core dump - invalid argument)"; - break; - case EFBIG: - coreinfo = " (no core dump - too large)"; - break; - default: - coreinfo = " (no core dump - other error)"; - break; - } - if (logexit) - log(LOG_INFO, - "pid %d (%s), jid %d, uid %d: exited on " - "signal %d%s\n", p->p_pid, p->p_comm, - p->p_ucred->cr_prison->pr_id, - td->td_ucred->cr_uid, - sig &~ WCOREFLAG, coreinfo); - } else - PROC_UNLOCK(p); - exit1(td, 0, sig); - /* NOTREACHED */ -} - -/* * Send queued SIGCHLD to parent when child process's state * is changed. */ @@ -3803,477 +3718,6 @@ childproc_exited(struct proc *p) sigparent(p, reason, status); } -#define MAX_NUM_CORE_FILES 100000 -#ifndef NUM_CORE_FILES -#define NUM_CORE_FILES 5 -#endif -CTASSERT(NUM_CORE_FILES >= 0 && NUM_CORE_FILES <= MAX_NUM_CORE_FILES); -static int num_cores = NUM_CORE_FILES; - -static int -sysctl_debug_num_cores_check (SYSCTL_HANDLER_ARGS) -{ - int error; - int new_val; - - new_val = num_cores; - error = sysctl_handle_int(oidp, &new_val, 0, req); - if (error != 0 || req->newptr == NULL) - return (error); - if (new_val > MAX_NUM_CORE_FILES) - new_val = MAX_NUM_CORE_FILES; - if (new_val < 0) - new_val = 0; - num_cores = new_val; - return (0); -} -SYSCTL_PROC(_debug, OID_AUTO, ncores, - CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, sizeof(int), - sysctl_debug_num_cores_check, "I", - "Maximum number of generated process corefiles while using index format"); - -#define GZIP_SUFFIX ".gz" -#define ZSTD_SUFFIX ".zst" - -int compress_user_cores = 0; - -static int -sysctl_compress_user_cores(SYSCTL_HANDLER_ARGS) -{ - int error, val; - - val = compress_user_cores; - error = sysctl_handle_int(oidp, &val, 0, req); - if (error != 0 || req->newptr == NULL) - return (error); - if (val != 0 && !compressor_avail(val)) - return (EINVAL); - compress_user_cores = val; - return (error); -} -SYSCTL_PROC(_kern, OID_AUTO, compress_user_cores, - CTLTYPE_INT | CTLFLAG_RWTUN | CTLFLAG_NEEDGIANT, 0, sizeof(int), - sysctl_compress_user_cores, "I", - "Enable compression of user corefiles (" - __XSTRING(COMPRESS_GZIP) " = gzip, " - __XSTRING(COMPRESS_ZSTD) " = zstd)"); - -int compress_user_cores_level = 6; -SYSCTL_INT(_kern, OID_AUTO, compress_user_cores_level, CTLFLAG_RWTUN, - &compress_user_cores_level, 0, - "Corefile compression level"); - -/* - * Protect the access to corefilename[] by allproc_lock. - */ -#define corefilename_lock allproc_lock - -static char corefilename[MAXPATHLEN] = {"%N.core"}; -TUNABLE_STR("kern.corefile", corefilename, sizeof(corefilename)); - -static int -sysctl_kern_corefile(SYSCTL_HANDLER_ARGS) -{ - int error; - - sx_xlock(&corefilename_lock); - error = sysctl_handle_string(oidp, corefilename, sizeof(corefilename), - req); - sx_xunlock(&corefilename_lock); - - return (error); -} -SYSCTL_PROC(_kern, OID_AUTO, corefile, CTLTYPE_STRING | CTLFLAG_RW | - CTLFLAG_MPSAFE, 0, 0, sysctl_kern_corefile, "A", - "Process corefile name format string"); - -static void -vnode_close_locked(struct thread *td, struct vnode *vp) -{ - - VOP_UNLOCK(vp); - vn_close(vp, FWRITE, td->td_ucred, td); -} - -/* - * If the core format has a %I in it, then we need to check - * for existing corefiles before defining a name. - * To do this we iterate over 0..ncores to find a - * non-existing core file name to use. If all core files are - * already used we choose the oldest one. - */ -static int -corefile_open_last(struct thread *td, char *name, int indexpos, - int indexlen, int ncores, struct vnode **vpp) -{ - struct vnode *oldvp, *nextvp, *vp; - struct vattr vattr; - struct nameidata nd; - int error, i, flags, oflags, cmode; - char ch; - struct timespec lasttime; - - nextvp = oldvp = NULL; - cmode = S_IRUSR | S_IWUSR; - oflags = VN_OPEN_NOAUDIT | VN_OPEN_NAMECACHE | - (capmode_coredump ? VN_OPEN_NOCAPCHECK : 0); - - for (i = 0; i < ncores; i++) { - flags = O_CREAT | FWRITE | O_NOFOLLOW; - - ch = name[indexpos + indexlen]; - (void)snprintf(name + indexpos, indexlen + 1, "%.*u", indexlen, - i); - name[indexpos + indexlen] = ch; - - NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, name); - error = vn_open_cred(&nd, &flags, cmode, oflags, td->td_ucred, - NULL); - if (error != 0) - break; - - vp = nd.ni_vp; - NDFREE_PNBUF(&nd); - if ((flags & O_CREAT) == O_CREAT) { - nextvp = vp; - break; - } - - error = VOP_GETATTR(vp, &vattr, td->td_ucred); - if (error != 0) { - vnode_close_locked(td, vp); - break; - } - - if (oldvp == NULL || - lasttime.tv_sec > vattr.va_mtime.tv_sec || - (lasttime.tv_sec == vattr.va_mtime.tv_sec && - lasttime.tv_nsec >= vattr.va_mtime.tv_nsec)) { - if (oldvp != NULL) - vn_close(oldvp, FWRITE, td->td_ucred, td); - oldvp = vp; - VOP_UNLOCK(oldvp); - lasttime = vattr.va_mtime; - } else { - vnode_close_locked(td, vp); - } - } - - if (oldvp != NULL) { - if (nextvp == NULL) { - if ((td->td_proc->p_flag & P_SUGID) != 0) { - error = EFAULT; - vn_close(oldvp, FWRITE, td->td_ucred, td); - } else { - nextvp = oldvp; - error = vn_lock(nextvp, LK_EXCLUSIVE); - if (error != 0) { - vn_close(nextvp, FWRITE, td->td_ucred, - td); - nextvp = NULL; - } - } - } else { - vn_close(oldvp, FWRITE, td->td_ucred, td); - } - } - if (error != 0) { - if (nextvp != NULL) - vnode_close_locked(td, oldvp); - } else { - *vpp = nextvp; - } - - return (error); -} - -/* - * corefile_open(comm, uid, pid, td, compress, vpp, namep) - * Expand the name described in corefilename, using name, uid, and pid - * and open/create core file. - * corefilename is a printf-like string, with three format specifiers: - * %N name of process ("name") - * %P process id (pid) - * %U user id (uid) - * For example, "%N.core" is the default; they can be disabled completely - * by using "/dev/null", or all core files can be stored in "/cores/%U/%N-%P". - * This is controlled by the sysctl variable kern.corefile (see above). - */ -static int -corefile_open(const char *comm, uid_t uid, pid_t pid, struct thread *td, - int compress, int signum, struct vnode **vpp, char **namep) -{ - struct sbuf sb; - struct nameidata nd; - const char *format; - char *hostname, *name; - int cmode, error, flags, i, indexpos, indexlen, oflags, ncores; - - hostname = NULL; - format = corefilename; - name = malloc(MAXPATHLEN, M_TEMP, M_WAITOK | M_ZERO); - indexlen = 0; - indexpos = -1; - ncores = num_cores; - (void)sbuf_new(&sb, name, MAXPATHLEN, SBUF_FIXEDLEN); - sx_slock(&corefilename_lock); - for (i = 0; format[i] != '\0'; i++) { - switch (format[i]) { - case '%': /* Format character */ - i++; - switch (format[i]) { - case '%': - sbuf_putc(&sb, '%'); - break; - case 'H': /* hostname */ - if (hostname == NULL) { - hostname = malloc(MAXHOSTNAMELEN, - M_TEMP, M_WAITOK); - } - getcredhostname(td->td_ucred, hostname, - MAXHOSTNAMELEN); - sbuf_cat(&sb, hostname); - break; - case 'I': /* autoincrementing index */ - if (indexpos != -1) { - sbuf_printf(&sb, "%%I"); - break; - } - - indexpos = sbuf_len(&sb); - sbuf_printf(&sb, "%u", ncores - 1); - indexlen = sbuf_len(&sb) - indexpos; - break; - case 'N': /* process name */ - sbuf_printf(&sb, "%s", comm); - break; - case 'P': /* process id */ - sbuf_printf(&sb, "%u", pid); - break; - case 'S': /* signal number */ - sbuf_printf(&sb, "%i", signum); - break; - case 'U': /* user id */ - sbuf_printf(&sb, "%u", uid); - break; - default: - log(LOG_ERR, - "Unknown format character %c in " - "corename `%s'\n", format[i], format); - break; - } - break; - default: - sbuf_putc(&sb, format[i]); - break; - } - } - sx_sunlock(&corefilename_lock); - free(hostname, M_TEMP); - if (compress == COMPRESS_GZIP) - sbuf_cat(&sb, GZIP_SUFFIX); - else if (compress == COMPRESS_ZSTD) - sbuf_cat(&sb, ZSTD_SUFFIX); - if (sbuf_error(&sb) != 0) { - log(LOG_ERR, "pid %ld (%s), uid (%lu): corename is too " - "long\n", (long)pid, comm, (u_long)uid); - sbuf_delete(&sb); - free(name, M_TEMP); - return (ENOMEM); - } - sbuf_finish(&sb); - sbuf_delete(&sb); - - if (indexpos != -1) { - error = corefile_open_last(td, name, indexpos, indexlen, ncores, - vpp); - if (error != 0) { - log(LOG_ERR, - "pid %d (%s), uid (%u): Path `%s' failed " - "on initial open test, error = %d\n", - pid, comm, uid, name, error); - } - } else { - cmode = S_IRUSR | S_IWUSR; - oflags = VN_OPEN_NOAUDIT | VN_OPEN_NAMECACHE | - (capmode_coredump ? VN_OPEN_NOCAPCHECK : 0); - flags = O_CREAT | FWRITE | O_NOFOLLOW; - if ((td->td_proc->p_flag & P_SUGID) != 0) - flags |= O_EXCL; - - NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, name); - error = vn_open_cred(&nd, &flags, cmode, oflags, td->td_ucred, - NULL); - if (error == 0) { - *vpp = nd.ni_vp; - NDFREE_PNBUF(&nd); - } - } - - if (error != 0) { -#ifdef AUDIT - audit_proc_coredump(td, name, error); -#endif - free(name, M_TEMP); - return (error); - } - *namep = name; - return (0); -} - -/* - * Dump a process' core. The main routine does some - * policy checking, and creates the name of the coredump; - * then it passes on a vnode and a size limit to the process-specific - * coredump routine if there is one; if there _is not_ one, it returns - * ENOSYS; otherwise it returns the error from the process-specific routine. - */ - -static int -coredump(struct thread *td) -{ - struct proc *p = td->td_proc; - struct ucred *cred = td->td_ucred; - struct vnode *vp; - struct flock lf; - struct vattr vattr; - size_t fullpathsize; - int error, error1, jid, locked, ppid, sig; - char *name; /* name of corefile */ - void *rl_cookie; - off_t limit; - char *fullpath, *freepath = NULL; - struct sbuf *sb; - - PROC_LOCK_ASSERT(p, MA_OWNED); - MPASS((p->p_flag & P_HADTHREADS) == 0 || p->p_singlethread == td); - - if (!do_coredump || (!sugid_coredump && (p->p_flag & P_SUGID) != 0) || - (p->p_flag2 & P2_NOTRACE) != 0) { - PROC_UNLOCK(p); - return (EFAULT); - } - - /* - * Note that the bulk of limit checking is done after - * the corefile is created. The exception is if the limit - * for corefiles is 0, in which case we don't bother - * creating the corefile at all. This layout means that - * a corefile is truncated instead of not being created, - * if it is larger than the limit. - */ - limit = (off_t)lim_cur(td, RLIMIT_CORE); - if (limit == 0 || racct_get_available(p, RACCT_CORE) == 0) { - PROC_UNLOCK(p); - return (EFBIG); - } - - ppid = p->p_oppid; - sig = p->p_sig; - jid = p->p_ucred->cr_prison->pr_id; - PROC_UNLOCK(p); - - error = corefile_open(p->p_comm, cred->cr_uid, p->p_pid, td, - compress_user_cores, p->p_sig, &vp, &name); - if (error != 0) - return (error); - - /* - * Don't dump to non-regular files or files with links. - * Do not dump into system files. Effective user must own the corefile. - */ - if (vp->v_type != VREG || VOP_GETATTR(vp, &vattr, cred) != 0 || - vattr.va_nlink != 1 || (vp->v_vflag & VV_SYSTEM) != 0 || - vattr.va_uid != cred->cr_uid) { - VOP_UNLOCK(vp); - error = EFAULT; - goto out; - } - - VOP_UNLOCK(vp); - - /* Postpone other writers, including core dumps of other processes. */ - rl_cookie = vn_rangelock_wlock(vp, 0, OFF_MAX); - - lf.l_whence = SEEK_SET; - lf.l_start = 0; - lf.l_len = 0; - lf.l_type = F_WRLCK; - locked = (VOP_ADVLOCK(vp, (caddr_t)p, F_SETLK, &lf, F_FLOCK) == 0); - - VATTR_NULL(&vattr); - vattr.va_size = 0; - if (set_core_nodump_flag) - vattr.va_flags = UF_NODUMP; - vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); - VOP_SETATTR(vp, &vattr, cred); - VOP_UNLOCK(vp); - PROC_LOCK(p); - p->p_acflag |= ACORE; - PROC_UNLOCK(p); - - if (p->p_sysent->sv_coredump != NULL) { - error = p->p_sysent->sv_coredump(td, vp, limit, 0); - } else { - error = ENOSYS; - } - - if (locked) { - lf.l_type = F_UNLCK; - VOP_ADVLOCK(vp, (caddr_t)p, F_UNLCK, &lf, F_FLOCK); - } - vn_rangelock_unlock(vp, rl_cookie); - - /* - * Notify the userland helper that a process triggered a core dump. - * This allows the helper to run an automated debugging session. - */ - if (error != 0 || coredump_devctl == 0) - goto out; - sb = sbuf_new_auto(); - if (vn_fullpath_global(p->p_textvp, &fullpath, &freepath) != 0) - goto out2; - sbuf_cat(sb, "comm=\""); - devctl_safe_quote_sb(sb, fullpath); - free(freepath, M_TEMP); - sbuf_cat(sb, "\" core=\""); - - /* - * We can't lookup core file vp directly. When we're replacing a core, and - * other random times, we flush the name cache, so it will fail. Instead, - * if the path of the core is relative, add the current dir in front if it. - */ - if (name[0] != '/') { - fullpathsize = MAXPATHLEN; - freepath = malloc(fullpathsize, M_TEMP, M_WAITOK); - if (vn_getcwd(freepath, &fullpath, &fullpathsize) != 0) { - free(freepath, M_TEMP); - goto out2; - } - devctl_safe_quote_sb(sb, fullpath); - free(freepath, M_TEMP); - sbuf_putc(sb, '/'); - } - devctl_safe_quote_sb(sb, name); - sbuf_putc(sb, '"'); - - sbuf_printf(sb, " jid=%d pid=%d ppid=%d signo=%d", - jid, p->p_pid, ppid, sig); - if (sbuf_finish(sb) == 0) - devctl_notify("kernel", "signal", "coredump", sbuf_data(sb)); -out2: - sbuf_delete(sb); -out: - error1 = vn_close(vp, FWRITE, cred, td); - if (error == 0) - error = error1; -#ifdef AUDIT - audit_proc_coredump(td, name, error); -#endif - free(name, M_TEMP); - return (error); -} - /* * Nonexistent system call-- signal process (may want to handle it). Flag * error in case process won't see signal immediately (blocked or ignored). diff --git a/sys/kern/kern_sysctl.c b/sys/kern/kern_sysctl.c index 46226cc31980..25da134661e9 100644 --- a/sys/kern/kern_sysctl.c +++ b/sys/kern/kern_sysctl.c @@ -2368,7 +2368,7 @@ sysctl_root(SYSCTL_HANDLER_ARGS) priv = PRIV_SYSCTL_WRITEJAIL; #ifdef VIMAGE else if ((oid->oid_kind & CTLFLAG_VNET) && - prison_owns_vnet(req->td->td_ucred)) + prison_owns_vnet(req->td->td_ucred->cr_prison)) priv = PRIV_SYSCTL_WRITEJAIL; #endif else diff --git a/sys/kern/kern_thread.c b/sys/kern/kern_thread.c index f853af193016..50b040132396 100644 --- a/sys/kern/kern_thread.c +++ b/sys/kern/kern_thread.c @@ -571,7 +571,7 @@ threadinit(void) /* * Thread structures are specially aligned so that (at least) the - * 5 lower bits of a pointer to 'struct thead' must be 0. These bits + * 5 lower bits of a pointer to 'struct thread' must be 0. These bits * are used by synchronization primitives to store flags in pointers to * such structures. */ diff --git a/sys/kern/kern_ucoredump.c b/sys/kern/kern_ucoredump.c new file mode 100644 index 000000000000..d425596b5f24 --- /dev/null +++ b/sys/kern/kern_ucoredump.c @@ -0,0 +1,299 @@ +/* + * SPDX-License-Identifier: BSD-3-Clause + * + * Copyright (c) 1982, 1986, 1989, 1991, 1993 + * The Regents of the University of California. All rights reserved. + * (c) UNIX System Laboratories, Inc. + * All or some portions of this file are derived from material licensed + * to the University of California by American Telephone and Telegraph + * Co. or Unix System Laboratories, Inc. and are reproduced herein with + * the permission of UNIX System Laboratories, Inc. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. Neither the name of the University nor the names of its contributors + * may be used to endorse or promote products derived from this software + * without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +#include <sys/param.h> +#include <sys/acct.h> +#include <sys/compressor.h> +#include <sys/jail.h> +#include <sys/kernel.h> +#include <sys/lock.h> +#include <sys/mutex.h> +#include <sys/proc.h> +#include <sys/signalvar.h> +#include <sys/racct.h> +#include <sys/resourcevar.h> +#include <sys/rmlock.h> +#include <sys/sysctl.h> +#include <sys/syslog.h> +#include <sys/ucoredump.h> +#include <sys/wait.h> + +static int coredump(struct thread *td, const char **); + +int compress_user_cores = 0; + +static SLIST_HEAD(, coredumper) coredumpers = + SLIST_HEAD_INITIALIZER(coredumpers); +static struct rmlock coredump_rmlock; +RM_SYSINIT(coredump_lock, &coredump_rmlock, "coredump_lock"); + +static int kern_logsigexit = 1; +SYSCTL_INT(_kern, KERN_LOGSIGEXIT, logsigexit, CTLFLAG_RW, + &kern_logsigexit, 0, + "Log processes quitting on abnormal signals to syslog(3)"); + +static int sugid_coredump; +SYSCTL_INT(_kern, OID_AUTO, sugid_coredump, CTLFLAG_RWTUN, + &sugid_coredump, 0, "Allow setuid and setgid processes to dump core"); + +static int do_coredump = 1; +SYSCTL_INT(_kern, OID_AUTO, coredump, CTLFLAG_RW, + &do_coredump, 0, "Enable/Disable coredumps"); + +static int +sysctl_compress_user_cores(SYSCTL_HANDLER_ARGS) +{ + int error, val; + + val = compress_user_cores; + error = sysctl_handle_int(oidp, &val, 0, req); + if (error != 0 || req->newptr == NULL) + return (error); + if (val != 0 && !compressor_avail(val)) + return (EINVAL); + compress_user_cores = val; + return (error); +} +SYSCTL_PROC(_kern, OID_AUTO, compress_user_cores, + CTLTYPE_INT | CTLFLAG_RWTUN | CTLFLAG_NEEDGIANT, 0, sizeof(int), + sysctl_compress_user_cores, "I", + "Enable compression of user corefiles (" + __XSTRING(COMPRESS_GZIP) " = gzip, " + __XSTRING(COMPRESS_ZSTD) " = zstd)"); + +int compress_user_cores_level = 6; +SYSCTL_INT(_kern, OID_AUTO, compress_user_cores_level, CTLFLAG_RWTUN, + &compress_user_cores_level, 0, + "Corefile compression level"); + +void +coredumper_register(struct coredumper *cd) +{ + + blockcount_init(&cd->cd_refcount); + rm_wlock(&coredump_rmlock); + SLIST_INSERT_HEAD(&coredumpers, cd, cd_entry); + rm_wunlock(&coredump_rmlock); +} + +void +coredumper_unregister(struct coredumper *cd) +{ + + rm_wlock(&coredump_rmlock); + SLIST_REMOVE(&coredumpers, cd, coredumper, cd_entry); + rm_wunlock(&coredump_rmlock); + + /* + * Wait for any in-process coredumps to finish before returning. + */ + blockcount_wait(&cd->cd_refcount, NULL, "dumpwait", 0); +} + +/* + * Force the current process to exit with the specified signal, dumping core + * if appropriate. We bypass the normal tests for masked and caught signals, + * allowing unrecoverable failures to terminate the process without changing + * signal state. Mark the accounting record with the signal termination. + * If dumping core, save the signal number for the debugger. Calls exit and + * does not return. + */ +void +sigexit(struct thread *td, int sig) +{ + struct proc *p = td->td_proc; + int rv; + bool logexit; + + PROC_LOCK_ASSERT(p, MA_OWNED); + proc_set_p2_wexit(p); + + p->p_acflag |= AXSIG; + if ((p->p_flag2 & P2_LOGSIGEXIT_CTL) == 0) + logexit = kern_logsigexit != 0; + else + logexit = (p->p_flag2 & P2_LOGSIGEXIT_ENABLE) != 0; + + /* + * We must be single-threading to generate a core dump. This + * ensures that the registers in the core file are up-to-date. + * Also, the ELF dump handler assumes that the thread list doesn't + * change out from under it. + * + * XXX If another thread attempts to single-thread before us + * (e.g. via fork()), we won't get a dump at all. + */ + if (sig_do_core(sig) && thread_single(p, SINGLE_NO_EXIT) == 0) { + const char *err = NULL; + + p->p_sig = sig; + /* + * Log signals which would cause core dumps + * (Log as LOG_INFO to appease those who don't want + * these messages.) + * XXX : Todo, as well as euid, write out ruid too + * Note that coredump() drops proc lock. + */ + rv = coredump(td, &err); + if (rv == 0) { + MPASS(err == NULL); + sig |= WCOREFLAG; + } else if (err == NULL) { + switch (rv) { + case EFAULT: + err = "bad address"; + break; + case EINVAL: + err = "invalild argument"; + break; + case EFBIG: + err = "too large"; + break; + default: + err = "other error"; + break; + } + } + if (logexit) + log(LOG_INFO, + "pid %d (%s), jid %d, uid %d: exited on " + "signal %d (%s%s)\n", p->p_pid, p->p_comm, + p->p_ucred->cr_prison->pr_id, + td->td_ucred->cr_uid, sig &~ WCOREFLAG, + err != NULL ? "no core dump - " : "core dumped", + err != NULL ? err : ""); + } else + PROC_UNLOCK(p); + exit1(td, 0, sig); + /* NOTREACHED */ +} + + +/* + * Dump a process' core. The main routine does some + * policy checking, and creates the name of the coredump; + * then it passes on a vnode and a size limit to the process-specific + * coredump routine if there is one; if there _is not_ one, it returns + * ENOSYS; otherwise it returns the error from the process-specific routine. + */ +static int +coredump(struct thread *td, const char **errmsg) +{ + struct coredumper *iter, *chosen; + struct proc *p = td->td_proc; + struct rm_priotracker tracker; + off_t limit; + int error, priority; + + PROC_LOCK_ASSERT(p, MA_OWNED); + MPASS((p->p_flag & P_HADTHREADS) == 0 || p->p_singlethread == td); + + if (!do_coredump || (!sugid_coredump && (p->p_flag & P_SUGID) != 0) || + (p->p_flag2 & P2_NOTRACE) != 0) { + PROC_UNLOCK(p); + + if (!do_coredump) + *errmsg = "denied by kern.coredump"; + else if ((p->p_flag2 & P2_NOTRACE) != 0) + *errmsg = "process has trace disabled"; + else + *errmsg = "sugid process denied by kern.sugid_coredump"; + return (EFAULT); + } + + /* + * Note that the bulk of limit checking is done after + * the corefile is created. The exception is if the limit + * for corefiles is 0, in which case we don't bother + * creating the corefile at all. This layout means that + * a corefile is truncated instead of not being created, + * if it is larger than the limit. + */ + limit = (off_t)lim_cur(td, RLIMIT_CORE); + if (limit == 0 || racct_get_available(p, RACCT_CORE) == 0) { + PROC_UNLOCK(p); + *errmsg = "coredumpsize limit is 0"; + return (EFBIG); + } + + rm_rlock(&coredump_rmlock, &tracker); + priority = -1; + chosen = NULL; + SLIST_FOREACH(iter, &coredumpers, cd_entry) { + if (iter->cd_probe == NULL) { + /* + * If we haven't found anything of a higher priority + * yet, we'll call this a GENERIC. Ideally, we want + * coredumper modules to include a probe function. + */ + if (priority < 0) { + priority = COREDUMPER_GENERIC; + chosen = iter; + } + + continue; + } + + error = (*iter->cd_probe)(td); + if (error < 0) + continue; + + /* + * Higher priority than previous options. + */ + if (error > priority) { + priority = error; + chosen = iter; + } + } + + /* + * Acquire our refcount before we drop the lock so that + * coredumper_unregister() can safely assume that the refcount will only + * go down once it's dropped the rmlock. + */ + blockcount_acquire(&chosen->cd_refcount, 1); + rm_runlock(&coredump_rmlock, &tracker); + + /* Currently, we always have the vnode dumper built in. */ + MPASS(chosen != NULL); + error = ((*chosen->cd_handle)(td, limit)); + PROC_LOCK_ASSERT(p, MA_NOTOWNED); + + blockcount_release(&chosen->cd_refcount, 1); + + return (error); +} diff --git a/sys/kern/subr_asan.c b/sys/kern/subr_asan.c index 0edb631d1475..464efda1e91a 100644 --- a/sys/kern/subr_asan.c +++ b/sys/kern/subr_asan.c @@ -263,8 +263,7 @@ kasan_mark(const void *addr, size_t size, size_t redzsize, uint8_t code) if (__predict_false(!kasan_enabled)) return; - if ((vm_offset_t)addr >= DMAP_MIN_ADDRESS && - (vm_offset_t)addr < DMAP_MAX_ADDRESS) + if (kasan_md_unsupported((vm_offset_t)addr)) return; KASSERT((vm_offset_t)addr >= VM_MIN_KERNEL_ADDRESS && diff --git a/sys/kern/subr_compressor.c b/sys/kern/subr_compressor.c index 280264881241..5d59622e0455 100644 --- a/sys/kern/subr_compressor.c +++ b/sys/kern/subr_compressor.c @@ -538,6 +538,12 @@ compressor_init(compressor_cb_t cb, int format, size_t maxiosize, int level, return (s); } +int +compressor_format(const struct compressor *stream) +{ + return (stream->methods->format); +} + void compressor_reset(struct compressor *stream) { diff --git a/sys/kern/subr_pctrie.c b/sys/kern/subr_pctrie.c index 3a3548bad52b..bb86c779b936 100644 --- a/sys/kern/subr_pctrie.c +++ b/sys/kern/subr_pctrie.c @@ -691,21 +691,23 @@ _pctrie_lookup_ge(struct pctrie *ptree, struct pctrie_node *node, */ if (node == PCTRIE_NULL || *pctrie_toval(node) < index) { /* Climb the path to find a node with a descendant > index. */ - for (node = parent; node != NULL; node = pctrie_parent(node)) { - slot = pctrie_slot(node, index) + 1; - if ((node->pn_popmap >> slot) != 0) + node = NULL; + while (parent != NULL) { + slot = pctrie_slot(parent, index) + 1; + if ((parent->pn_popmap >> slot) != 0) break; + node = parent; + parent = pctrie_parent(node); } - if (node == NULL) { + if (parent == NULL) { if (parent_out != NULL) - *parent_out = NULL; + *parent_out = node; return (NULL); } /* Step to the least child with a descendant > index. */ - slot += ffs(node->pn_popmap >> slot) - 1; - parent = node; - node = pctrie_node_load(&node->pn_child[slot], NULL, + slot += ffs(parent->pn_popmap >> slot) - 1; + node = pctrie_node_load(&parent->pn_child[slot], NULL, PCTRIE_LOCKED); } /* Descend to the least leaf of the subtrie. */ @@ -785,21 +787,23 @@ _pctrie_lookup_le(struct pctrie *ptree, struct pctrie_node *node, */ if (node == PCTRIE_NULL || *pctrie_toval(node) > index) { /* Climb the path to find a node with a descendant < index. */ - for (node = parent; node != NULL; node = pctrie_parent(node)) { - slot = pctrie_slot(node, index); - if ((node->pn_popmap & ((1 << slot) - 1)) != 0) + node = NULL; + while (parent != NULL) { + slot = pctrie_slot(parent, index); + if ((parent->pn_popmap & ((1 << slot) - 1)) != 0) break; + node = parent; + parent = pctrie_parent(node); } - if (node == NULL) { + if (parent == NULL) { if (parent_out != NULL) - *parent_out = NULL; + *parent_out = node; return (NULL); } /* Step to the greatest child with a descendant < index. */ - slot = ilog2(node->pn_popmap & ((1 << slot) - 1)); - parent = node; - node = pctrie_node_load(&node->pn_child[slot], NULL, + slot = ilog2(parent->pn_popmap & ((1 << slot) - 1)); + node = pctrie_node_load(&parent->pn_child[slot], NULL, PCTRIE_LOCKED); } /* Descend to the greatest leaf of the subtrie. */ diff --git a/sys/kern/subr_trap.c b/sys/kern/subr_trap.c index 18388ae5f232..bac7d0080c71 100644 --- a/sys/kern/subr_trap.c +++ b/sys/kern/subr_trap.c @@ -338,8 +338,9 @@ ast_handler(struct thread *td, struct trapframe *framep, bool dtor) td->td_ast = 0; } - CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, td->td_proc->p_pid, - td->td_proc->p_comm); + CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, + td->td_proc == NULL ? -1 : td->td_proc->p_pid, + td->td_proc == NULL ? "" : td->td_proc->p_comm); KASSERT(framep == NULL || TRAPF_USERMODE(framep), ("ast in kernel mode")); diff --git a/sys/kern/sys_generic.c b/sys/kern/sys_generic.c index 94e44d888181..5606b36f772f 100644 --- a/sys/kern/sys_generic.c +++ b/sys/kern/sys_generic.c @@ -2269,6 +2269,7 @@ exterr_copyout(struct thread *td) ue.error = 0; sz = sizeof(ue.error); } else { + ktrexterr(td); sz = sizeof(ue) - __offsetof(struct uexterror, error); } error = copyout(&ue.error, uloc, sz); @@ -2309,6 +2310,12 @@ sys_exterrctl(struct thread *td, struct exterrctl_args *uap) return (EINVAL); td->td_pflags2 &= ~TDP2_UEXTERR; return (0); + case EXTERRCTL_UD: + /* + * Important: this code must always return EINVAL and never any + * extended error, for testing purposes. + */ + /* FALLTHROUGH */ default: return (EINVAL); } @@ -2329,7 +2336,6 @@ exterr_set(int eerror, int category, const char *mmsg, uintptr_t pp1, td->td_kexterr.p1 = pp1; td->td_kexterr.p2 = pp2; td->td_kexterr.src_line = line; - ktrexterr(td); } return (eerror); } diff --git a/sys/kern/uipc_ktls.c b/sys/kern/uipc_ktls.c index ce09042abdac..66ce1b5a081d 100644 --- a/sys/kern/uipc_ktls.c +++ b/sys/kern/uipc_ktls.c @@ -1207,7 +1207,7 @@ sb_mark_notready(struct sockbuf *sb) for (; m != NULL; m = m->m_next) { KASSERT(m->m_nextpkt == NULL, ("%s: m_nextpkt != NULL", __func__)); - KASSERT((m->m_flags & M_NOTAVAIL) == 0, ("%s: mbuf not avail", + KASSERT((m->m_flags & M_NOTREADY) == 0, ("%s: mbuf not ready", __func__)); KASSERT(sb->sb_acc >= m->m_len, ("%s: sb_acc < m->m_len", __func__)); diff --git a/sys/kern/uipc_shm.c b/sys/kern/uipc_shm.c index 6f83b875a6b6..85fe48ddd466 100644 --- a/sys/kern/uipc_shm.c +++ b/sys/kern/uipc_shm.c @@ -1134,10 +1134,10 @@ shm_doremove(struct shm_mapping *map) int kern_shm_open2(struct thread *td, const char *userpath, int flags, mode_t mode, - int shmflags, struct filecaps *fcaps, const char *name __unused) + int shmflags, struct filecaps *fcaps, const char *name __unused, + struct shmfd *shmfd) { struct pwddesc *pdp; - struct shmfd *shmfd; struct file *fp; char *path; void *rl_cookie; @@ -1214,23 +1214,41 @@ kern_shm_open2(struct thread *td, const char *userpath, int flags, mode_t mode, if (error != 0) goto outnofp; - /* A SHM_ANON path pointer creates an anonymous object. */ + /* + * A SHM_ANON path pointer creates an anonymous object. We allow other + * parts of the kernel to pre-populate a shmfd and then materialize an + * fd for it here as a means to pass data back up to userland. This + * doesn't really make sense for named shm objects, but it makes plenty + * of sense for anonymous objects. + */ if (userpath == SHM_ANON) { - /* A read-only anonymous object is pointless. */ - if ((flags & O_ACCMODE) == O_RDONLY) { - error = EINVAL; - goto out; - } - shmfd = shm_alloc(td->td_ucred, cmode, largepage); - if (shmfd == NULL) { - error = ENOMEM; - goto out; + if (shmfd != NULL) { + shm_hold(shmfd); + } else { + /* + * A read-only anonymous object is pointless, unless it + * was pre-populated by the kernel with the expectation + * that a shmfd would later be created for userland to + * access it through. + */ + if ((flags & O_ACCMODE) == O_RDONLY) { + error = EINVAL; + goto out; + } + shmfd = shm_alloc(td->td_ucred, cmode, largepage); + if (shmfd == NULL) { + error = ENOMEM; + goto out; + } + + shmfd->shm_seals = initial_seals; + shmfd->shm_flags = shmflags; } - shmfd->shm_seals = initial_seals; - shmfd->shm_flags = shmflags; } else { fnv = fnv_32_str(path, FNV1_32_INIT); sx_xlock(&shm_dict_lock); + + MPASS(shmfd == NULL); shmfd = shm_lookup(path, fnv); if (shmfd == NULL) { /* Object does not yet exist, create it if requested. */ @@ -2173,7 +2191,7 @@ kern_shm_open(struct thread *td, const char *path, int flags, mode_t mode, struct filecaps *caps) { - return (kern_shm_open2(td, path, flags, mode, 0, caps, NULL)); + return (kern_shm_open2(td, path, flags, mode, 0, caps, NULL, NULL)); } /* @@ -2191,7 +2209,7 @@ sys_shm_open2(struct thread *td, struct shm_open2_args *uap) { return (kern_shm_open2(td, uap->path, uap->flags, uap->mode, - uap->shmflags, NULL, uap->name)); + uap->shmflags, NULL, uap->name, NULL)); } int diff --git a/sys/kern/uipc_sockbuf.c b/sys/kern/uipc_sockbuf.c index ec00878cd9a5..745702bd4a4f 100644 --- a/sys/kern/uipc_sockbuf.c +++ b/sys/kern/uipc_sockbuf.c @@ -195,14 +195,14 @@ int sbready(struct sockbuf *sb, struct mbuf *m0, int count) { struct mbuf *m; - u_int blocker; + bool blocker; SOCKBUF_LOCK_ASSERT(sb); KASSERT(sb->sb_fnrdy != NULL, ("%s: sb %p NULL fnrdy", __func__, sb)); KASSERT(count > 0, ("%s: invalid count %d", __func__, count)); m = m0; - blocker = (sb->sb_fnrdy == m) ? M_BLOCKED : 0; + blocker = (sb->sb_fnrdy == m); while (count > 0) { KASSERT(m->m_flags & M_NOTREADY, @@ -217,8 +217,7 @@ sbready(struct sockbuf *sb, struct mbuf *m0, int count) m->m_epg_nrdy = 0; } else count--; - - m->m_flags &= ~(M_NOTREADY | blocker); + m->m_flags &= ~M_NOTREADY; if (blocker) sb->sb_acc += m->m_len; m = m->m_next; @@ -240,12 +239,8 @@ sbready(struct sockbuf *sb, struct mbuf *m0, int count) } /* This one was blocking all the queue. */ - for (; m && (m->m_flags & M_NOTREADY) == 0; m = m->m_next) { - KASSERT(m->m_flags & M_BLOCKED, - ("%s: m %p !M_BLOCKED", __func__, m)); - m->m_flags &= ~M_BLOCKED; + for (; m && (m->m_flags & M_NOTREADY) == 0; m = m->m_next) sb->sb_acc += m->m_len; - } sb->sb_fnrdy = m; sbready_compress(sb, m0, m); @@ -269,8 +264,7 @@ sballoc(struct sockbuf *sb, struct mbuf *m) sb->sb_fnrdy = m; else sb->sb_acc += m->m_len; - } else - m->m_flags |= M_BLOCKED; + } if (m->m_type != MT_DATA && m->m_type != MT_OOBDATA) sb->sb_ctl += m->m_len; @@ -287,29 +281,29 @@ sballoc(struct sockbuf *sb, struct mbuf *m) void sbfree(struct sockbuf *sb, struct mbuf *m) { + struct mbuf *n; #if 0 /* XXX: not yet: soclose() call path comes here w/o lock. */ SOCKBUF_LOCK_ASSERT(sb); #endif - sb->sb_ccc -= m->m_len; - if (!(m->m_flags & M_NOTAVAIL)) - sb->sb_acc -= m->m_len; - if (m == sb->sb_fnrdy) { - struct mbuf *n; - KASSERT(m->m_flags & M_NOTREADY, ("%s: m %p !M_NOTREADY", __func__, m)); n = m->m_next; while (n != NULL && !(n->m_flags & M_NOTREADY)) { - n->m_flags &= ~M_BLOCKED; sb->sb_acc += n->m_len; n = n->m_next; } sb->sb_fnrdy = n; + } else { + /* Assert that mbuf is not behind sb_fnrdy. */ + for (n = sb->sb_fnrdy; n != NULL; n = n->m_next) + KASSERT(n != m, ("%s: sb %p freeing %p behind sb_fnrdy", + __func__, sb, m)); + sb->sb_acc -= m->m_len; } if (m->m_type != MT_DATA && m->m_type != MT_OOBDATA) @@ -779,6 +773,7 @@ sbsetopt(struct socket *so, struct sockopt *sopt) * high-water. */ *lowat = (cc > *hiwat) ? *hiwat : cc; + *flags &= ~SB_AUTOLOWAT; break; } @@ -1128,13 +1123,7 @@ sbcheck(struct sockbuf *sb, const char *file, int line) } fnrdy = m; } - if (fnrdy) { - if (!(m->m_flags & M_NOTAVAIL)) { - printf("sb %p: fnrdy %p, m %p is avail\n", - sb, sb->sb_fnrdy, m); - goto fail; - } - } else + if (fnrdy == NULL) acc += m->m_len; ccc += m->m_len; mbcnt += MSIZE; @@ -1601,8 +1590,8 @@ sbcut_internal(struct sockbuf *sb, int len) next = m->m_nextpkt; } if (m->m_len > len) { - KASSERT(!(m->m_flags & M_NOTAVAIL), - ("%s: m %p M_NOTAVAIL", __func__, m)); + KASSERT(!(m->m_flags & M_NOTREADY), + ("%s: m %p M_NOTREADY", __func__, m)); m->m_len -= len; m->m_data += len; sb->sb_ccc -= len; diff --git a/sys/kern/uipc_socket.c b/sys/kern/uipc_socket.c index 6c9eb7139cd1..fe2d8d056062 100644 --- a/sys/kern/uipc_socket.c +++ b/sys/kern/uipc_socket.c @@ -1211,7 +1211,8 @@ solisten_clone(struct socket *head) so->so_rcv.sb_timeo = head->sol_sbrcv_timeo; so->so_snd.sb_timeo = head->sol_sbsnd_timeo; so->so_rcv.sb_flags = head->sol_sbrcv_flags & SB_AUTOSIZE; - so->so_snd.sb_flags = head->sol_sbsnd_flags & SB_AUTOSIZE; + so->so_snd.sb_flags = head->sol_sbsnd_flags & + (SB_AUTOSIZE | SB_AUTOLOWAT); if ((so->so_proto->pr_flags & PR_SOCKBUF) == 0) { so->so_snd.sb_mtx = &so->so_snd_mtx; so->so_rcv.sb_mtx = &so->so_rcv_mtx; @@ -2988,8 +2989,8 @@ dontblock: */ moff = 0; offset = 0; - while (m != NULL && !(m->m_flags & M_NOTAVAIL) && uio->uio_resid > 0 - && error == 0) { + while (m != NULL && !(m->m_flags & M_NOTREADY) && uio->uio_resid > 0 && + error == 0) { /* * If the type of mbuf has changed since the last mbuf * examined ('type'), end the receive operation. @@ -3341,7 +3342,7 @@ deliver: for (m = sb->sb_mb; m != NULL && m->m_len <= len; m = m->m_next) { - KASSERT(!(m->m_flags & M_NOTAVAIL), + KASSERT(!(m->m_flags & M_NOTREADY), ("%s: m %p not available", __func__, m)); len -= m->m_len; uio->uio_resid -= m->m_len; @@ -4514,6 +4515,9 @@ sokqfilter_generic(struct socket *so, struct knote *kn) SOCK_BUF_LOCK(so, which); knlist_add(knl, kn, 1); sb->sb_flags |= SB_KNOTE; + if ((kn->kn_sfflags & NOTE_LOWAT) && + (sb->sb_flags & SB_AUTOLOWAT)) + sb->sb_flags &= ~SB_AUTOLOWAT; SOCK_BUF_UNLOCK(so, which); } SOCK_UNLOCK(so); diff --git a/sys/kern/vfs_cache.c b/sys/kern/vfs_cache.c index 3d455b3874cc..89c1d779f04c 100644 --- a/sys/kern/vfs_cache.c +++ b/sys/kern/vfs_cache.c @@ -332,7 +332,8 @@ SDT_PROBE_DEFINE2(vfs, namecache, evict_negative, done, "struct vnode *", "char *"); SDT_PROBE_DEFINE1(vfs, namecache, symlink, alloc__fail, "size_t"); -SDT_PROBE_DEFINE3(vfs, fplookup, lookup, done, "struct nameidata", "int", "bool"); +SDT_PROBE_DEFINE3(vfs, fplookup, lookup, done, "struct nameidata *", "int", + "enum cache_fpl_status"); SDT_PROBE_DECLARE(vfs, namei, lookup, entry); SDT_PROBE_DECLARE(vfs, namei, lookup, return); @@ -6420,15 +6421,11 @@ out: cache_fpl_smr_assert_not_entered(&fpl); cache_fpl_assert_status(&fpl); *status = fpl.status; - if (SDT_PROBES_ENABLED()) { - SDT_PROBE3(vfs, fplookup, lookup, done, ndp, fpl.line, fpl.status); - if (fpl.status == CACHE_FPL_STATUS_HANDLED) - SDT_PROBE4(vfs, namei, lookup, return, error, ndp->ni_vp, true, - ndp); - } - + SDT_PROBE3(vfs, fplookup, lookup, done, ndp, fpl.line, fpl.status); if (__predict_true(fpl.status == CACHE_FPL_STATUS_HANDLED)) { MPASS(error != CACHE_FPL_FAILED); + SDT_PROBE4(vfs, namei, lookup, return, error, ndp->ni_vp, true, + ndp); if (error != 0) { cache_fpl_cleanup_cnp(fpl.cnp); MPASS(fpl.dvp == NULL); diff --git a/sys/kern/vfs_inotify.c b/sys/kern/vfs_inotify.c index 41e73bb41a49..d3cd0d1f9832 100644 --- a/sys/kern/vfs_inotify.c +++ b/sys/kern/vfs_inotify.c @@ -371,7 +371,7 @@ inotify_unlink_watch_locked(struct inotify_softc *sc, struct inotify_watch *watc TAILQ_REMOVE(&vp->v_pollinfo->vpi_inotify, watch, vlink); if (TAILQ_EMPTY(&vp->v_pollinfo->vpi_inotify)) - vn_irflag_unset_locked(vp, VIRF_INOTIFY); + vn_irflag_unset(vp, VIRF_INOTIFY); } /* @@ -675,7 +675,8 @@ vn_inotify(struct vnode *vp, struct vnode *dvp, struct componentname *cnp, struct vattr va; int error; - error = VOP_GETATTR(vp, &va, cnp->cn_cred); + error = VOP_GETATTR(vp, &va, + cnp->cn_cred); if (error == 0 && va.va_nlink != 0) selfevent = 0; } @@ -760,9 +761,11 @@ vn_inotify_add_watch(struct vnode *vp, struct inotify_softc *sc, uint32_t mask, * directory if it's specified as a vnode. */ vrefact(vp); + VOP_UNLOCK(vp); NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, dp->d_name, vp); error = namei(&nd); + vn_lock(vp, LK_SHARED | LK_RETRY); if (error != 0) break; vn_irflag_set_cond(nd.ni_vp, VIRF_INOTIFY_PARENT); diff --git a/sys/kern/vfs_subr.c b/sys/kern/vfs_subr.c index 918b256e6c59..29774cf87393 100644 --- a/sys/kern/vfs_subr.c +++ b/sys/kern/vfs_subr.c @@ -6533,17 +6533,6 @@ vop_read_pgcache_post(void *ap, int rc) VFS_KNOTE_UNLOCKED(a->a_vp, NOTE_READ); } -void -vop_readdir_post(void *ap, int rc) -{ - struct vop_readdir_args *a = ap; - - if (!rc) { - VFS_KNOTE_LOCKED(a->a_vp, NOTE_READ); - INOTIFY(a->a_vp, IN_ACCESS); - } -} - static struct knlist fs_knlist; static void diff --git a/sys/kern/vfs_syscalls.c b/sys/kern/vfs_syscalls.c index d880733cbfe7..25d40a9806cb 100644 --- a/sys/kern/vfs_syscalls.c +++ b/sys/kern/vfs_syscalls.c @@ -2253,10 +2253,10 @@ kern_accessat(struct thread *td, int fd, const char *path, cred = td->td_ucred; if ((flag & AT_EACCESS) == 0 && ((cred->cr_uid != cred->cr_ruid || - cred->cr_rgid != cred->cr_groups[0]))) { + cred->cr_rgid != cred->cr_gid))) { usecred = crdup(cred); usecred->cr_uid = cred->cr_ruid; - usecred->cr_groups[0] = cred->cr_rgid; + usecred->cr_gid = cred->cr_rgid; td->td_ucred = usecred; } else usecred = cred; @@ -4314,10 +4314,6 @@ kern_getdirentries(struct thread *td, int fd, char *buf, size_t count, vp = fp->f_vnode; foffset = foffset_lock(fp, 0); unionread: - if (vp->v_type != VDIR) { - error = EINVAL; - goto fail; - } if (__predict_false((vp->v_vflag & VV_UNLINKED) != 0)) { error = ENOENT; goto fail; @@ -4330,6 +4326,19 @@ unionread: auio.uio_segflg = bufseg; auio.uio_td = td; vn_lock(vp, LK_SHARED | LK_RETRY); + /* + * We want to return ENOTDIR for anything that is not VDIR, but + * not for VBAD, and we can't check for VBAD while the vnode is + * unlocked. + */ + if (vp->v_type != VDIR) { + if (vp->v_type == VBAD) + error = EBADF; + else + error = ENOTDIR; + VOP_UNLOCK(vp); + goto fail; + } AUDIT_ARG_VNODE1(vp); loff = auio.uio_offset = foffset; #ifdef MAC diff --git a/sys/kern/vnode_if.src b/sys/kern/vnode_if.src index 38138a4af921..2e63215b2f97 100644 --- a/sys/kern/vnode_if.src +++ b/sys/kern/vnode_if.src @@ -242,8 +242,8 @@ vop_read_pgcache { %% write vp L L L -%! write pre VOP_WRITE_PRE -%! write post VOP_WRITE_POST +%! write pre vop_write_pre +%! write post vop_write_post vop_write { IN struct vnode *vp; @@ -380,6 +380,7 @@ vop_symlink { %% readdir vp L L L +%! readdir pre vop_readdir_pre %! readdir post vop_readdir_post vop_readdir { diff --git a/sys/modules/Makefile b/sys/modules/Makefile index 9922796f8a1d..99c9ec9dcd01 100644 --- a/sys/modules/Makefile +++ b/sys/modules/Makefile @@ -34,6 +34,7 @@ SUBDIR= \ alq \ ${_amd_ecc_inject} \ ${_amdgpio} \ + ${_amdsmu} \ ${_amdsbwd} \ ${_amdsmn} \ ${_amdtemp} \ @@ -326,6 +327,7 @@ SUBDIR= \ proto \ pseudofs \ ${_pst} \ + ${_pt} \ pty \ puc \ pwm \ @@ -771,6 +773,7 @@ _acpi= acpi _aesni= aesni .endif _amd_ecc_inject=amd_ecc_inject +_amdsmu= amdsmu _amdsbwd= amdsbwd _amdsmn= amdsmn _amdtemp= amdtemp @@ -842,6 +845,7 @@ _iwx= iwx _ixl= ixl _nvdimm= nvdimm _pms= pms +_pt= pt _qat= qat .if ${MK_SOURCELESS_UCODE} != "no" _qatfw= qatfw diff --git a/sys/modules/amdsmu/Makefile b/sys/modules/amdsmu/Makefile new file mode 100644 index 000000000000..752f57173d61 --- /dev/null +++ b/sys/modules/amdsmu/Makefile @@ -0,0 +1,14 @@ +# SPDX-License-Identifier: BSD-2-Clause +# +# Copyright (c) 2025 The FreeBSD Foundation +# +# This software was developed by Aymeric Wibo <obiwac@freebsd.org> +# under sponsorship from the FreeBSD Foundation. + +.PATH: ${SRCTOP}/sys/dev/amdsmu + +KMOD= amdsmu +SRCS= amdsmu.c +SRCS+= bus_if.h device_if.h pci_if.h + +.include <bsd.kmod.mk> diff --git a/sys/modules/efirt/Makefile b/sys/modules/efirt/Makefile index 4738996fd4e6..c46484465b68 100644 --- a/sys/modules/efirt/Makefile +++ b/sys/modules/efirt/Makefile @@ -9,7 +9,7 @@ SRCS+= device_if.h bus_if.h clock_if.h DPSRCS+= assym.inc .if ${MACHINE_CPUARCH} == "amd64" -SRCS+= opt_hwpmc_hooks.h opt_kstack_pages.h +SRCS+= opt_acpi.h opt_hwpmc_hooks.h opt_kstack_pages.h .endif efirt_support.o: efirt_support.S assym.inc diff --git a/sys/modules/ice/Makefile b/sys/modules/ice/Makefile index 91f20193d878..9f9c9f602cda 100644 --- a/sys/modules/ice/Makefile +++ b/sys/modules/ice/Makefile @@ -13,6 +13,7 @@ SRCS += opt_inet.h opt_inet6.h opt_rss.h opt_iflib.h SRCS += ice_lib.c ice_osdep.c ice_resmgr.c ice_strings.c SRCS += ice_iflib_recovery_txrx.c ice_iflib_txrx.c if_ice_iflib.c SRCS += ice_fw_logging.c ice_ddp_common.c +SRCS.PCI_IOV += pci_iov_if.h ice_iov.c ice_vf_mbx.c # RDMA Client interface # TODO: Is this the right way to compile this? diff --git a/sys/modules/pt/Makefile b/sys/modules/pt/Makefile new file mode 100644 index 000000000000..416b072face9 --- /dev/null +++ b/sys/modules/pt/Makefile @@ -0,0 +1,8 @@ + +.PATH: ${SRCTOP}/sys/amd64/pt + +KMOD= pt +SRCS= pt.c pt.h device_if.h bus_if.h +SRCS+= opt_hwpmc_hooks.h opt_kstack_pages.h + +.include <bsd.kmod.mk> diff --git a/sys/modules/qlnx/qlnxe/Makefile b/sys/modules/qlnx/qlnxe/Makefile index 3d8415cf0e57..2a44ae6ddde5 100644 --- a/sys/modules/qlnx/qlnxe/Makefile +++ b/sys/modules/qlnx/qlnxe/Makefile @@ -58,6 +58,7 @@ SRCS+=qlnx_rdma.c SRCS+=qlnx_ioctl.c SRCS+=qlnx_os.c +SRCS+=opt_inet.h SRCS+= ${LINUXKPI_GENSRCS} diff --git a/sys/net/ethernet.h b/sys/net/ethernet.h index cf4f75bd0b6c..01485cf26e06 100644 --- a/sys/net/ethernet.h +++ b/sys/net/ethernet.h @@ -62,6 +62,8 @@ struct ether_header { u_char ether_shost[ETHER_ADDR_LEN]; u_short ether_type; } __packed; +_Static_assert(sizeof(struct ether_header) == ETHER_HDR_LEN, + "size of struct ether_header is wrong"); /* * Structure of a 48-bit Ethernet address. @@ -69,6 +71,8 @@ struct ether_header { struct ether_addr { u_char octet[ETHER_ADDR_LEN]; } __packed; +_Static_assert(sizeof(struct ether_addr) == ETHER_ADDR_LEN, + "size of struct ether_addr is wrong"); #define ETHER_IS_MULTICAST(addr) (*(addr) & 0x01) /* is address mcast/bcast? */ #define ETHER_IS_IPV6_MULTICAST(addr) \ @@ -112,6 +116,8 @@ struct ether_vlan_header { uint16_t evl_tag; uint16_t evl_proto; } __packed; +_Static_assert(sizeof(struct ether_vlan_header) == ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN, + "size of struct ether_vlan_header is wrong"); #define EVL_VLID_MASK 0x0FFF #define EVL_PRI_MASK 0xE000 diff --git a/sys/net/if_bridge.c b/sys/net/if_bridge.c index 5b3ee740d75e..0a35fb4095fb 100644 --- a/sys/net/if_bridge.c +++ b/sys/net/if_bridge.c @@ -76,31 +76,34 @@ * heterogeneous bridges). */ -#include <sys/cdefs.h> #include "opt_inet.h" #include "opt_inet6.h" +#define EXTERR_CATEGORY EXTERR_CAT_BRIDGE + #include <sys/param.h> +#include <sys/ctype.h> /* string functions */ #include <sys/eventhandler.h> -#include <sys/mbuf.h> +#include <sys/exterrvar.h> +#include <sys/jail.h> +#include <sys/kernel.h> +#include <sys/lock.h> #include <sys/malloc.h> +#include <sys/mbuf.h> +#include <sys/module.h> +#include <sys/mutex.h> +#include <sys/priv.h> +#include <sys/proc.h> #include <sys/protosw.h> +#include <sys/random.h> #include <sys/systm.h> -#include <sys/jail.h> -#include <sys/time.h> #include <sys/socket.h> /* for net/if.h */ #include <sys/sockio.h> -#include <sys/ctype.h> /* string functions */ -#include <sys/kernel.h> -#include <sys/random.h> #include <sys/syslog.h> #include <sys/sysctl.h> +#include <sys/time.h> + #include <vm/uma.h> -#include <sys/module.h> -#include <sys/priv.h> -#include <sys/proc.h> -#include <sys/lock.h> -#include <sys/mutex.h> #include <net/bpf.h> #include <net/if.h> @@ -254,8 +257,8 @@ struct bridge_iflist { uint32_t bif_addrcnt; /* cur. # of addresses */ uint32_t bif_addrexceeded;/* # of address violations */ struct epoch_context bif_epoch_ctx; - ether_vlanid_t bif_untagged; /* untagged vlan id */ - ifbvlan_set_t bif_vlan_set; /* allowed tagged vlans */ + ether_vlanid_t bif_pvid; /* port vlan id */ + ifbvlan_set_t bif_vlan_set; /* if allowed tagged vlans */ }; /* @@ -404,7 +407,7 @@ static int bridge_ioctl_sma(struct bridge_softc *, void *); static int bridge_ioctl_sifprio(struct bridge_softc *, void *); static int bridge_ioctl_sifcost(struct bridge_softc *, void *); static int bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *); -static int bridge_ioctl_sifuntagged(struct bridge_softc *, void *); +static int bridge_ioctl_sifpvid(struct bridge_softc *, void *); static int bridge_ioctl_sifvlanset(struct bridge_softc *, void *); static int bridge_ioctl_gifvlanset(struct bridge_softc *, void *); static int bridge_ioctl_addspan(struct bridge_softc *, void *); @@ -625,7 +628,7 @@ static const struct bridge_control bridge_control_table[] = { { bridge_ioctl_sifmaxaddr, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER }, - { bridge_ioctl_sifuntagged, sizeof(struct ifbreq), + { bridge_ioctl_sifpvid, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER }, { bridge_ioctl_sifvlanset, sizeof(struct ifbif_vlan_req), @@ -986,31 +989,37 @@ bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) case SIOCGDRVSPEC: case SIOCSDRVSPEC: if (ifd->ifd_cmd >= bridge_control_table_size) { - error = EINVAL; + error = EXTERROR(EINVAL, "Invalid control command"); break; } bc = &bridge_control_table[ifd->ifd_cmd]; if (cmd == SIOCGDRVSPEC && (bc->bc_flags & BC_F_COPYOUT) == 0) { - error = EINVAL; + error = EXTERROR(EINVAL, + "Inappropriate ioctl for command " + "(expected SIOCSDRVSPEC)"); break; } else if (cmd == SIOCSDRVSPEC && (bc->bc_flags & BC_F_COPYOUT) != 0) { - error = EINVAL; + error = EXTERROR(EINVAL, + "Inappropriate ioctl for command " + "(expected SIOCGDRVSPEC)"); break; } if (bc->bc_flags & BC_F_SUSER) { error = priv_check(td, PRIV_NET_BRIDGE); - if (error) + if (error) { + EXTERROR(error, "PRIV_NET_BRIDGE required"); break; + } } if (ifd->ifd_len != bc->bc_argsize || ifd->ifd_len > sizeof(args)) { - error = EINVAL; + error = EXTERROR(EINVAL, "Invalid argument size"); break; } @@ -1062,7 +1071,8 @@ bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) oldmtu = sc->sc_ifp->if_mtu; if (ifr->ifr_mtu < IF_MINMTU) { - error = EINVAL; + error = EXTERROR(EINVAL, + "Requested MTU is lower than IF_MINMTU"); break; } if (CK_LIST_EMPTY(&sc->sc_iflist)) { @@ -1088,6 +1098,8 @@ bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) (*bif->bif_ifp->if_ioctl)(bif->bif_ifp, SIOCSIFMTU, (caddr_t)ifr); } + EXTERROR(error, + "Failed to set MTU on member interface"); } else { sc->sc_ifp->if_mtu = ifr->ifr_mtu; } @@ -1125,14 +1137,14 @@ bridge_mutecaps(struct bridge_softc *sc) mask = BRIDGE_IFCAPS_MASK; CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { - /* Every member must support it or its disabled */ + /* Every member must support it or it's disabled */ mask &= bif->bif_savedcaps; } CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) { enabled = bif->bif_ifp->if_capenable; enabled &= ~BRIDGE_IFCAPS_STRIP; - /* strip off mask bits and enable them again if allowed */ + /* Strip off mask bits and enable them again if allowed */ enabled &= ~BRIDGE_IFCAPS_MASK; enabled |= mask; bridge_set_ifcap(sc, bif, enabled); @@ -1282,7 +1294,7 @@ bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif, #endif break; } - /* reneable any interface capabilities */ + /* Re-enable any interface capabilities */ bridge_set_ifcap(sc, bif, bif->bif_savedcaps); } bstp_destroy(&bif->bif_stp); /* prepare to free */ @@ -1318,21 +1330,48 @@ bridge_ioctl_add(struct bridge_softc *sc, void *arg) ifs = ifunit(req->ifbr_ifsname); if (ifs == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "No such interface", + req->ifbr_ifsname)); if (ifs->if_ioctl == NULL) /* must be supported */ - return (EINVAL); + return (EXTERROR(EINVAL, "Interface must support ioctl(2)")); + + /* + * If the new interface is a vlan(4), it could be a bridge SVI. + * Don't allow such things to be added to bridges. + */ + if (ifs->if_type == IFT_L2VLAN) { + struct ifnet *parent; + struct epoch_tracker et; + bool is_bridge; + + /* + * Entering NET_EPOCH with BRIDGE_LOCK held, but this is okay + * since we don't sleep here. + */ + NET_EPOCH_ENTER(et); + parent = VLAN_TRUNKDEV(ifs); + is_bridge = (parent != NULL && parent->if_type == IFT_BRIDGE); + NET_EPOCH_EXIT(et); + + if (is_bridge) + return (EXTERROR(EINVAL, + "Bridge SVI cannot be added to a bridge")); + } /* If it's in the span list, it can't be a member. */ CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) if (ifs == bif->bif_ifp) - return (EBUSY); + return (EXTERROR(EBUSY, + "Span interface cannot be a member")); if (ifs->if_bridge) { struct bridge_iflist *sbif = ifs->if_bridge; if (sbif->bif_sc == sc) - return (EEXIST); + return (EXTERROR(EEXIST, + "Interface is already a member of this bridge")); - return (EBUSY); + return (EXTERROR(EBUSY, + "Interface is already a member of another bridge")); } switch (ifs->if_type) { @@ -1342,7 +1381,7 @@ bridge_ioctl_add(struct bridge_softc *sc, void *arg) /* permitted interface types */ break; default: - return (EINVAL); + return (EXTERROR(EINVAL, "Unsupported interface type")); } #ifdef INET6 @@ -1394,11 +1433,15 @@ bridge_ioctl_add(struct bridge_softc *sc, void *arg) CK_STAILQ_FOREACH(ifa, &ifs->if_addrhead, ifa_link) { #ifdef INET if (ifa->ifa_addr->sa_family == AF_INET) - return (EINVAL); + return (EXTERROR(EINVAL, + "Member interface may not have " + "an IPv4 address configured")); #endif #ifdef INET6 if (ifa->ifa_addr->sa_family == AF_INET6) - return (EINVAL); + return (EXTERROR(EINVAL, + "Member interface may not have " + "an IPv6 address configured")); #endif } } @@ -1420,7 +1463,8 @@ bridge_ioctl_add(struct bridge_softc *sc, void *arg) " new member %s\n", sc->sc_ifp->if_xname, ifr.ifr_mtu, ifs->if_xname); - return (EINVAL); + return (EXTERROR(EINVAL, + "Failed to set MTU on new member")); } } @@ -1482,7 +1526,7 @@ bridge_ioctl_del(struct bridge_softc *sc, void *arg) bif = bridge_lookup_member(sc, req->ifbr_ifsname); if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); bridge_delete_member(sc, bif, 0); @@ -1498,7 +1542,7 @@ bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg) bif = bridge_lookup_member(sc, req->ifbr_ifsname); if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); bp = &bif->bif_stp; req->ifbr_ifsflags = bif->bif_flags; @@ -1512,7 +1556,7 @@ bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg) req->ifbr_addrcnt = bif->bif_addrcnt; req->ifbr_addrmax = bif->bif_addrmax; req->ifbr_addrexceeded = bif->bif_addrexceeded; - req->ifbr_untagged = bif->bif_untagged; + req->ifbr_pvid = bif->bif_pvid; /* Copy STP state options as flags */ if (bp->bp_operedge) @@ -1541,12 +1585,12 @@ bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg) bif = bridge_lookup_member(sc, req->ifbr_ifsname); if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); bp = &bif->bif_stp; if (req->ifbr_ifsflags & IFBIF_SPAN) /* SPAN is readonly */ - return (EINVAL); + return (EXTERROR(EINVAL, "Span interface cannot be modified")); NET_EPOCH_ENTER(et); @@ -1555,7 +1599,8 @@ bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg) error = bstp_enable(&bif->bif_stp); if (error) { NET_EPOCH_EXIT(et); - return (error); + return (EXTERROR(error, + "Failed to enable STP")); } } } else { @@ -1724,7 +1769,7 @@ bridge_ioctl_saddr(struct bridge_softc *sc, void *arg) bif = bridge_lookup_member(sc, req->ifba_ifsname); if (bif == NULL) { NET_EPOCH_EXIT(et); - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); } /* bridge_rtupdate() may acquire the lock. */ @@ -1858,7 +1903,7 @@ bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg) bif = bridge_lookup_member(sc, req->ifbr_ifsname); if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority)); } @@ -1871,7 +1916,7 @@ bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg) bif = bridge_lookup_member(sc, req->ifbr_ifsname); if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost)); } @@ -1884,28 +1929,28 @@ bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg) bif = bridge_lookup_member(sc, req->ifbr_ifsname); if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); bif->bif_addrmax = req->ifbr_addrmax; return (0); } static int -bridge_ioctl_sifuntagged(struct bridge_softc *sc, void *arg) +bridge_ioctl_sifpvid(struct bridge_softc *sc, void *arg) { struct ifbreq *req = arg; struct bridge_iflist *bif; bif = bridge_lookup_member(sc, req->ifbr_ifsname); if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); - if (req->ifbr_untagged > DOT1Q_VID_MAX) - return (EINVAL); + if (req->ifbr_pvid > DOT1Q_VID_MAX) + return (EXTERROR(EINVAL, "Invalid VLAN ID")); - if (req->ifbr_untagged != DOT1Q_VID_NULL) + if (req->ifbr_pvid != DOT1Q_VID_NULL) bif->bif_flags |= IFBIF_VLANFILTER; - bif->bif_untagged = req->ifbr_untagged; + bif->bif_pvid = req->ifbr_pvid; return (0); } @@ -1917,12 +1962,12 @@ bridge_ioctl_sifvlanset(struct bridge_softc *sc, void *arg) bif = bridge_lookup_member(sc, req->bv_ifname); if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); /* Reject invalid VIDs. */ if (BRVLAN_TEST(&req->bv_set, DOT1Q_VID_NULL) || BRVLAN_TEST(&req->bv_set, DOT1Q_VID_RSVD_IMPL)) - return (EINVAL); + return (EXTERROR(EINVAL, "Invalid VLAN ID in set")); switch (req->bv_op) { /* Replace the existing vlan set with the new set */ @@ -1942,7 +1987,8 @@ bridge_ioctl_sifvlanset(struct bridge_softc *sc, void *arg) /* Invalid or unknown operation */ default: - return (EINVAL); + return (EXTERROR(EINVAL, + "Unsupported BRDGSIFVLANSET operation")); } /* @@ -1962,7 +2008,7 @@ bridge_ioctl_gifvlanset(struct bridge_softc *sc, void *arg) bif = bridge_lookup_member(sc, req->bv_ifname); if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a bridge member")); BIT_COPY(BRVLAN_SETSIZE, &bif->bif_vlan_set, &req->bv_set); return (0); @@ -1977,14 +2023,16 @@ bridge_ioctl_addspan(struct bridge_softc *sc, void *arg) ifs = ifunit(req->ifbr_ifsname); if (ifs == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "No such interface")); CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) if (ifs == bif->bif_ifp) - return (EBUSY); + return (EXTERROR(EBUSY, + "Interface is already a span port")); if (ifs->if_bridge != NULL) - return (EBUSY); + return (EXTERROR(EEXIST, + "Interface is already a bridge member")); switch (ifs->if_type) { case IFT_ETHER: @@ -1992,7 +2040,7 @@ bridge_ioctl_addspan(struct bridge_softc *sc, void *arg) case IFT_L2VLAN: break; default: - return (EINVAL); + return (EXTERROR(EINVAL, "Unsupported interface type")); } bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO); @@ -2016,14 +2064,14 @@ bridge_ioctl_delspan(struct bridge_softc *sc, void *arg) ifs = ifunit(req->ifbr_ifsname); if (ifs == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "No such interface")); CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) if (ifs == bif->bif_ifp) break; if (bif == NULL) - return (ENOENT); + return (EXTERROR(ENOENT, "Interface is not a span port")); bridge_delete_span(sc, bif); @@ -2278,8 +2326,8 @@ bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m, * the VLAN header. */ if ((bif->bif_flags & IFBIF_VLANFILTER) && - bif->bif_untagged != DOT1Q_VID_NULL && - VLANTAGOF(m) == bif->bif_untagged) { + bif->bif_pvid != DOT1Q_VID_NULL && + VLANTAGOF(m) == bif->bif_pvid) { m->m_flags &= ~M_VLANTAG; m->m_pkthdr.ether_vtag = 0; } @@ -3145,14 +3193,14 @@ bridge_vfilter_in(const struct bridge_iflist *sbif, struct mbuf *m) * The frame doesn't have a tag. If the interface does not * have an untagged vlan configured, drop the frame. */ - if (sbif->bif_untagged == DOT1Q_VID_NULL) + if (sbif->bif_pvid == DOT1Q_VID_NULL) return (false); /* * Otherwise, insert a new tag based on the interface's * untagged vlan id. */ - m->m_pkthdr.ether_vtag = sbif->bif_untagged; + m->m_pkthdr.ether_vtag = sbif->bif_pvid; m->m_flags |= M_VLANTAG; } else { /* @@ -3213,7 +3261,7 @@ bridge_vfilter_out(const struct bridge_iflist *dbif, const struct mbuf *m) * If the frame's vlan matches the interfaces's untagged vlan, * allow it. */ - if (vlan == dbif->bif_untagged) + if (vlan == dbif->bif_pvid) return (true); /* @@ -3244,10 +3292,11 @@ bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc); /* Check the source address is valid and not multicast. */ - if (ETHER_IS_MULTICAST(dst) || - (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 && - dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0) - return (EINVAL); + if (ETHER_IS_MULTICAST(dst)) + return (EXTERROR(EINVAL, "Multicast address not permitted")); + if (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 && + dst[3] == 0 && dst[4] == 0 && dst[5] == 0) + return (EXTERROR(EINVAL, "Zero address not permitted")); /* * A route for this destination might already exist. If so, @@ -3266,13 +3315,14 @@ bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, if (sc->sc_brtcnt >= sc->sc_brtmax) { sc->sc_brtexceeded++; BRIDGE_RT_UNLOCK(sc); - return (ENOSPC); + return (EXTERROR(ENOSPC, "Address table is full")); } /* Check per interface address limits (if enabled) */ if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) { bif->bif_addrexceeded++; BRIDGE_RT_UNLOCK(sc); - return (ENOSPC); + return (EXTERROR(ENOSPC, + "Interface address limit exceeded")); } /* @@ -3283,7 +3333,8 @@ bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO); if (brt == NULL) { BRIDGE_RT_UNLOCK(sc); - return (ENOMEM); + return (EXTERROR(ENOMEM, + "Cannot allocate address node")); } brt->brt_vnet = curvnet; @@ -3631,7 +3682,7 @@ bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt) do { dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr); if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan) - return (EEXIST); + return (EXTERROR(EEXIST, "Address already exists")); if (dir > 0) { CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash); goto out; diff --git a/sys/net/if_bridgevar.h b/sys/net/if_bridgevar.h index 97b63e3d4416..c458dcc152a0 100644 --- a/sys/net/if_bridgevar.h +++ b/sys/net/if_bridgevar.h @@ -124,7 +124,7 @@ #define BRDGSPROTO 28 /* set protocol (ifbrparam) */ #define BRDGSTXHC 29 /* set tx hold count (ifbrparam) */ #define BRDGSIFAMAX 30 /* set max interface addrs (ifbreq) */ -#define BRDGSIFUNTAGGED 31 /* set if untagged vlan */ +#define BRDGSIFPVID 31 /* set if PVID */ #define BRDGSIFVLANSET 32 /* set if vlan set */ #define BRDGGIFVLANSET 33 /* get if vlan set */ @@ -144,7 +144,7 @@ struct ifbreq { uint32_t ifbr_addrcnt; /* member if addr number */ uint32_t ifbr_addrmax; /* member if addr max */ uint32_t ifbr_addrexceeded; /* member if addr violations */ - ether_vlanid_t ifbr_untagged; /* member if untagged vlan */ + ether_vlanid_t ifbr_pvid; /* member if PVID */ uint8_t pad[32]; }; diff --git a/sys/net/if_ethersubr.c b/sys/net/if_ethersubr.c index 7be4dfac23e7..3ae0c01c0efc 100644 --- a/sys/net/if_ethersubr.c +++ b/sys/net/if_ethersubr.c @@ -92,11 +92,6 @@ #include <crypto/sha1.h> -#ifdef CTASSERT -CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2); -CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN); -#endif - VNET_DEFINE(pfil_head_t, link_pfil_head); /* Packet filter hooks */ /* netgraph node hooks for ng_ether(4) */ diff --git a/sys/net/if_gif.h b/sys/net/if_gif.h index 3c1846b8f82a..c6692d3dd6bc 100644 --- a/sys/net/if_gif.h +++ b/sys/net/if_gif.h @@ -120,7 +120,8 @@ int in6_gif_setopts(struct gif_softc *, u_int); #define GIFGOPTS _IOWR('i', 150, struct ifreq) #define GIFSOPTS _IOW('i', 151, struct ifreq) +#define GIF_NOCLAMP 0x0001 #define GIF_IGNORE_SOURCE 0x0002 -#define GIF_OPTMASK (GIF_IGNORE_SOURCE) +#define GIF_OPTMASK (GIF_NOCLAMP|GIF_IGNORE_SOURCE) #endif /* _NET_IF_GIF_H_ */ diff --git a/sys/net/if_lagg.c b/sys/net/if_lagg.c index 9867a718e148..5b52bfa80e3b 100644 --- a/sys/net/if_lagg.c +++ b/sys/net/if_lagg.c @@ -718,6 +718,7 @@ lagg_capabilities(struct lagg_softc *sc) sc->sc_ifp->if_capenable = ena; sc->sc_ifp->if_capenable2 = ena2; sc->sc_ifp->if_hwassist = hwa; + (void)if_hw_tsomax_update(sc->sc_ifp, &hw_tsomax); getmicrotime(&sc->sc_ifp->if_lastchange); if (sc->sc_ifflags & IFF_DEBUG) diff --git a/sys/net/if_ovpn.c b/sys/net/if_ovpn.c index 7bdbc565f4ca..fe3e7bbd7fff 100644 --- a/sys/net/if_ovpn.c +++ b/sys/net/if_ovpn.c @@ -34,11 +34,13 @@ #include <sys/epoch.h> #include <sys/file.h> #include <sys/filedesc.h> +#include <sys/jail.h> #include <sys/kernel.h> #include <sys/malloc.h> #include <sys/mbuf.h> #include <sys/module.h> #include <sys/nv.h> +#include <sys/osd.h> #include <sys/priv.h> #include <sys/protosw.h> #include <sys/rmlock.h> @@ -79,7 +81,6 @@ #include "if_ovpn.h" struct ovpn_kkey_dir { - int refcount; uint8_t key[32]; uint8_t keylen; uint8_t nonce[8]; @@ -132,6 +133,9 @@ struct ovpn_notification { /* Delete notification */ enum ovpn_del_reason del_reason; struct ovpn_peer_counters counters; + + /* Float notification */ + struct sockaddr_storage address; }; struct ovpn_softc; @@ -196,6 +200,10 @@ struct ovpn_softc { struct epoch_context epoch_ctx; }; +struct ovpn_mtag { + struct sockaddr_storage addr; +}; + static struct ovpn_kpeer *ovpn_find_peer(struct ovpn_softc *, uint32_t); static bool ovpn_udp_input(struct mbuf *, int, struct inpcb *, const struct sockaddr *, void *); @@ -205,7 +213,10 @@ static int ovpn_encap(struct ovpn_softc *, uint32_t, struct mbuf *); static int ovpn_get_af(struct mbuf *); static void ovpn_free_kkey_dir(struct ovpn_kkey_dir *); static bool ovpn_check_replay(struct ovpn_kkey_dir *, uint32_t); -static int ovpn_peer_compare(struct ovpn_kpeer *, struct ovpn_kpeer *); +static int ovpn_peer_compare(const struct ovpn_kpeer *, + const struct ovpn_kpeer *); +static bool ovpn_sockaddr_compare(const struct sockaddr *, + const struct sockaddr *); static RB_PROTOTYPE(ovpn_kpeers, ovpn_kpeer, tree, ovpn_peer_compare); static RB_GENERATE(ovpn_kpeers, ovpn_kpeer, tree, ovpn_peer_compare); @@ -278,11 +289,48 @@ SYSCTL_INT(_net_link_openvpn, OID_AUTO, netisr_queue, "Use netisr_queue() rather than netisr_dispatch()."); static int -ovpn_peer_compare(struct ovpn_kpeer *a, struct ovpn_kpeer *b) +ovpn_peer_compare(const struct ovpn_kpeer *a, const struct ovpn_kpeer *b) { return (a->peerid - b->peerid); } +static bool +ovpn_sockaddr_compare(const struct sockaddr *a, + const struct sockaddr *b) +{ + if (a->sa_family != b->sa_family) + return (false); + MPASS(a->sa_len == b->sa_len); + + switch (a->sa_family) { + case AF_INET: { + const struct sockaddr_in *a4, *b4; + + a4 = (const struct sockaddr_in *)a; + b4 = (const struct sockaddr_in *)b; + + if (a4->sin_port != b4->sin_port) + return (false); + + return (a4->sin_addr.s_addr == b4->sin_addr.s_addr); + } + case AF_INET6: { + const struct sockaddr_in6 *a6, *b6; + + a6 = (const struct sockaddr_in6 *)a; + b6 = (const struct sockaddr_in6 *)b; + + if (a6->sin6_port != b6->sin6_port) + return (false); + + return (memcmp(&a6->sin6_addr, &b6->sin6_addr, + sizeof(a6->sin6_addr)) == 0); + } + default: + panic("Unknown address family %d", a->sa_family); + } +} + static struct ovpn_kpeer * ovpn_find_peer(struct ovpn_softc *sc, uint32_t peerid) { @@ -304,15 +352,15 @@ ovpn_find_only_peer(struct ovpn_softc *sc) } static uint16_t -ovpn_get_port(struct sockaddr_storage *s) +ovpn_get_port(const struct sockaddr_storage *s) { switch (s->ss_family) { case AF_INET: { - struct sockaddr_in *in = (struct sockaddr_in *)s; + const struct sockaddr_in *in = (const struct sockaddr_in *)s; return (in->sin_port); } case AF_INET6: { - struct sockaddr_in6 *in6 = (struct sockaddr_in6 *)s; + const struct sockaddr_in6 *in6 = (const struct sockaddr_in6 *)s; return (in6->sin6_port); } default: @@ -320,6 +368,25 @@ ovpn_get_port(struct sockaddr_storage *s) } } +static void +ovpn_set_port(struct sockaddr_storage *s, unsigned short port) +{ + switch (s->ss_family) { + case AF_INET: { + struct sockaddr_in *in = (struct sockaddr_in *)s; + in->sin_port = port; + break; + } + case AF_INET6: { + struct sockaddr_in6 *in6 = (struct sockaddr_in6 *)s; + in6->sin6_port = port; + break; + } + default: + panic("Unsupported address family %d", s->ss_family); + } +} + static int ovpn_nvlist_to_sockaddr(const nvlist_t *nvl, struct sockaddr_storage *sa) { @@ -333,14 +400,16 @@ ovpn_nvlist_to_sockaddr(const nvlist_t *nvl, struct sockaddr_storage *sa) return (EINVAL); af = nvlist_get_number(nvl, "af"); - switch (af) { #ifdef INET case AF_INET: { struct sockaddr_in *in = (struct sockaddr_in *)sa; size_t len; const void *addr = nvlist_get_binary(nvl, "address", &len); + + memset(in, 0, sizeof(*in)); in->sin_family = af; + in->sin_len = sizeof(*in); if (len != sizeof(in->sin_addr)) return (EINVAL); @@ -354,7 +423,10 @@ ovpn_nvlist_to_sockaddr(const nvlist_t *nvl, struct sockaddr_storage *sa) struct sockaddr_in6 *in6 = (struct sockaddr_in6 *)sa; size_t len; const void *addr = nvlist_get_binary(nvl, "address", &len); + + memset(in6, 0, sizeof(*in6)); in6->sin6_family = af; + in6->sin6_len = sizeof(*in6); if (len != sizeof(in6->sin6_addr)) return (EINVAL); @@ -370,31 +442,42 @@ ovpn_nvlist_to_sockaddr(const nvlist_t *nvl, struct sockaddr_storage *sa) return (0); } -static bool -ovpn_has_peers(struct ovpn_softc *sc) +static int +ovpn_add_sockaddr(nvlist_t *parent, const char *name, const struct sockaddr *s) { - OVPN_ASSERT(sc); - - return (sc->peercount > 0); -} + nvlist_t *nvl; -static void -ovpn_rele_so(struct ovpn_softc *sc) -{ - bool has_peers; + nvl = nvlist_create(0); + if (nvl == NULL) + return (ENOMEM); - OVPN_WASSERT(sc); + nvlist_add_number(nvl, "af", s->sa_family); - if (sc->so == NULL) - return; + switch (s->sa_family) { + case AF_INET: { + const struct sockaddr_in *s4 = (const struct sockaddr_in *)s; - has_peers = ovpn_has_peers(sc); + nvlist_add_number(nvl, "port", s4->sin_port); + nvlist_add_binary(nvl, "address", &s4->sin_addr, + sizeof(s4->sin_addr)); + break; + } + case AF_INET6: { + const struct sockaddr_in6 *s6 = (const struct sockaddr_in6 *)s; - if (! has_peers) { - MPASS(sc->peercount == 0); - } else { - MPASS(sc->peercount > 0); + nvlist_add_number(nvl, "port", s6->sin6_port); + nvlist_add_binary(nvl, "address", &s6->sin6_addr, + sizeof(s6->sin6_addr)); + break; + } + default: + nvlist_destroy(nvl); + return (EINVAL); } + + nvlist_move_nvlist(parent, name, nvl); + + return (0); } static void @@ -449,6 +532,33 @@ ovpn_notify_key_rotation(struct ovpn_softc *sc, struct ovpn_kpeer *peer) } } +static int +ovpn_notify_float(struct ovpn_softc *sc, uint32_t peerid, + const struct sockaddr_storage *remote) +{ + struct ovpn_notification *n; + + n = malloc(sizeof(*n), M_OVPN, M_NOWAIT | M_ZERO); + if (n == NULL) + return (ENOMEM); + + n->peerid = peerid; + n->type = OVPN_NOTIF_FLOAT; + memcpy(&n->address, remote, sizeof(n->address)); + + if (buf_ring_enqueue(sc->notifring, n) != 0) { + free(n, M_OVPN); + return (ENOMEM); + } else if (sc->so != NULL) { + /* Wake up userspace */ + sc->so->so_error = EAGAIN; + sorwakeup(sc->so); + sowwakeup(sc->so); + } + + return (0); +} + static void ovpn_peer_release_ref(struct ovpn_kpeer *peer, bool locked) { @@ -485,8 +595,6 @@ ovpn_peer_release_ref(struct ovpn_kpeer *peer, bool locked) ovpn_free_kkey_dir(peer->keys[i].decrypt); } - ovpn_rele_so(sc); - callout_stop(&peer->ping_send); callout_stop(&peer->ping_rcv); uma_zfree_pcpu(pcpu_zone_4, peer->last_active); @@ -502,7 +610,7 @@ ovpn_new_peer(struct ifnet *ifp, const nvlist_t *nvl) #ifdef INET6 struct epoch_tracker et; #endif - struct sockaddr_storage remote; + struct sockaddr_storage local, remote; struct ovpn_kpeer *peer = NULL; struct file *fp = NULL; struct ovpn_softc *sc = ifp->if_softc; @@ -571,20 +679,37 @@ ovpn_new_peer(struct ifnet *ifp, const nvlist_t *nvl) callout_init_rm(&peer->ping_send, &sc->lock, CALLOUT_SHAREDLOCK); callout_init_rm(&peer->ping_rcv, &sc->lock, 0); - peer->local.ss_len = sizeof(peer->local); - ret = sosockaddr(so, (struct sockaddr *)&peer->local); - if (ret) + memset(&local, 0, sizeof(local)); + local.ss_len = sizeof(local); + ret = sosockaddr(so, (struct sockaddr *)&local); + if (ret != 0) goto error; + if (nvlist_exists_nvlist(nvl, "local")) { + struct sockaddr_storage local1; + + ret = ovpn_nvlist_to_sockaddr(nvlist_get_nvlist(nvl, "local"), + &local1); + if (ret != 0) + goto error; - if (ovpn_get_port(&peer->local) == 0) { + /* + * openvpn doesn't provide a port here when in multihome mode, + * just steal the one the socket is bound to. + */ + if (ovpn_get_port(&local1) == 0) + ovpn_set_port(&local1, ovpn_get_port(&local)); + memcpy(&local, &local1, sizeof(local1)); + } + if (ovpn_get_port(&local) == 0) { ret = EINVAL; goto error; } - if (peer->local.ss_family != remote.ss_family) { + if (local.ss_family != remote.ss_family) { ret = EINVAL; goto error; } + memcpy(&peer->local, &local, sizeof(local)); memcpy(&peer->remote, &remote, sizeof(remote)); #ifdef INET6 @@ -633,6 +758,7 @@ ovpn_new_peer(struct ifnet *ifp, const nvlist_t *nvl) * a new one. */ ret = udp_set_kernel_tunneling(sc->so, NULL, NULL, NULL); + MPASS(ret == 0); sorele(sc->so); sc->so = NULL; } @@ -1364,12 +1490,36 @@ opvn_get_pkt(struct ovpn_softc *sc, nvlist_t **onvl) } nvlist_add_number(nvl, "peerid", n->peerid); nvlist_add_number(nvl, "notification", n->type); - if (n->type == OVPN_NOTIF_DEL_PEER) { + switch (n->type) { + case OVPN_NOTIF_DEL_PEER: { nvlist_add_number(nvl, "del_reason", n->del_reason); /* No error handling, because we want to send the notification * even if we can't attach the counters. */ ovpn_notif_add_counters(nvl, n); + break; + } + case OVPN_NOTIF_FLOAT: { + int ret; + + ret = ovpn_add_sockaddr(nvl, "address", + (struct sockaddr *)&n->address); + + if (ret) { + /* + * Try to re-enqueue the notification. Maybe we'll + * have better luck next time. No error handling, + * because if we fail to re-enqueue there's nothing we can do. + */ + (void)ovpn_notify_float(sc, n->peerid, &n->address); + nvlist_destroy(nvl); + free(n, M_OVPN); + return (ret); + } + break; + } + default: + break; } free(n, M_OVPN); @@ -1525,6 +1675,7 @@ ovpn_finish_rx(struct ovpn_softc *sc, struct mbuf *m, struct rm_priotracker *_ovpn_lock_trackerp) { uint32_t af; + struct m_tag *mtag; OVPN_RASSERT(sc); NET_EPOCH_ASSERT(); @@ -1543,6 +1694,38 @@ ovpn_finish_rx(struct ovpn_softc *sc, struct mbuf *m, OVPN_RUNLOCK(sc); + /* Check if the peer changed to a new source address. */ + mtag = m_tag_find(m, PACKET_TAG_OVPN, NULL); + if (mtag != NULL) { + struct ovpn_mtag *ot = (struct ovpn_mtag *)(mtag + 1); + + OVPN_WLOCK(sc); + + /* + * Check the address against the peer's remote again, because we may race + * against ourselves (i.e. we may have tagged multiple packets to indicate we + * floated). + */ + if (ovpn_sockaddr_compare((struct sockaddr *)&ot->addr, + (struct sockaddr *)&peer->remote)) { + OVPN_WUNLOCK(sc); + goto skip_float; + } + + /* And notify userspace. */ + if (ovpn_notify_float(sc, peer->peerid, &ot->addr) == 0) { + /* + * Update the 'remote' for this peer, but only if + * we've actually enqueued the notification. + * Otherwise we can try again later. + */ + memcpy(&peer->remote, &ot->addr, sizeof(peer->remote)); + } + + OVPN_WUNLOCK(sc); + } + +skip_float: OVPN_COUNTER_ADD(sc, received_data_pkts, 1); OVPN_COUNTER_ADD(sc, tunnel_bytes_received, m->m_pkthdr.len); OVPN_PEER_COUNTER_ADD(peer, pkt_in, 1); @@ -2305,6 +2488,29 @@ ovpn_udp_input(struct mbuf *m, int off, struct inpcb *inp, return (true); } + /* + * If we got this from a different address than we expected tag the packet. + * We'll deal with notifiying userspace later, after we've decrypted and + * verified. + */ + if (! ovpn_sockaddr_compare((struct sockaddr *)&peer->remote, sa)) { + struct m_tag *mt; + struct ovpn_mtag *ot; + + MPASS(sa->sa_len <= sizeof(ot->addr)); + mt = m_tag_get(PACKET_TAG_OVPN, sizeof(*ot), M_NOWAIT); + /* + * If we fail to allocate here we'll just try again on the next + * packet. + */ + if (mt != NULL) { + ot = (struct ovpn_mtag *)(mt + 1); + memcpy(&ot->addr, sa, sa->sa_len); + + m_tag_prepend(m, mt); + } + } + if (key->decrypt->cipher == OVPN_CIPHER_ALG_NONE) { /* Now remove the outer headers */ m_adj_decap(m, sizeof(struct udphdr) + ohdrlen); @@ -2519,6 +2725,7 @@ ovpn_clone_destroy_cb(struct epoch_context *ctx) COUNTER_ARRAY_FREE(sc->counters, OVPN_COUNTER_SIZE); + rm_destroy(&sc->lock); if_free(sc->ifp); free(sc, M_OVPN); } @@ -2579,23 +2786,53 @@ vnet_ovpn_init(const void *unused __unused) VNET_SYSINIT(vnet_ovpn_init, SI_SUB_PSEUDO, SI_ORDER_ANY, vnet_ovpn_init, NULL); -static void -vnet_ovpn_uninit(const void *unused __unused) +static int +ovpn_prison_remove(void *obj, void *data __unused) { - if_clone_detach(V_ovpn_cloner); +#ifdef VIMAGE + struct prison *pr; + + pr = obj; + if (prison_owns_vnet(pr)) { + CURVNET_SET(pr->pr_vnet); + if (V_ovpn_cloner != NULL) { + ifc_detach_cloner(V_ovpn_cloner); + V_ovpn_cloner = NULL; + } + CURVNET_RESTORE(); + } +#endif + return (0); } -VNET_SYSUNINIT(vnet_ovpn_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY, - vnet_ovpn_uninit, NULL); static int ovpnmodevent(module_t mod, int type, void *data) { + static int ovpn_osd_jail_slot; + switch (type) { - case MOD_LOAD: - /* Done in vnet_ovpn_init() */ + case MOD_LOAD: { + /* + * Registration is handled in vnet_ovpn_init(), but cloned + * interfaces must be destroyed via PR_METHOD_REMOVE since they + * hold a reference to the prison via the UDP socket, which + * prevents the prison from being destroyed. + */ + osd_method_t methods[PR_MAXMETHOD] = { + [PR_METHOD_REMOVE] = ovpn_prison_remove, + }; + ovpn_osd_jail_slot = osd_jail_register(NULL, methods); break; + } case MOD_UNLOAD: - /* Done in vnet_ovpn_uninit() */ + if (ovpn_osd_jail_slot != 0) + osd_jail_deregister(ovpn_osd_jail_slot); + CURVNET_SET(vnet0); + if (V_ovpn_cloner != NULL) { + ifc_detach_cloner(V_ovpn_cloner); + V_ovpn_cloner = NULL; + } + CURVNET_RESTORE(); break; default: return (EOPNOTSUPP); diff --git a/sys/net/if_ovpn.h b/sys/net/if_ovpn.h index 2d6b8c1e7eff..2a24c35788a9 100644 --- a/sys/net/if_ovpn.h +++ b/sys/net/if_ovpn.h @@ -37,6 +37,7 @@ enum ovpn_notif_type { OVPN_NOTIF_DEL_PEER, OVPN_NOTIF_ROTATE_KEY, + OVPN_NOTIF_FLOAT, }; enum ovpn_del_reason { diff --git a/sys/net/if_tuntap.c b/sys/net/if_tuntap.c index 3bab04aa4d38..5e6f65c04b2f 100644 --- a/sys/net/if_tuntap.c +++ b/sys/net/if_tuntap.c @@ -74,6 +74,7 @@ #include <sys/malloc.h> #include <sys/random.h> #include <sys/ctype.h> +#include <sys/osd.h> #include <net/ethernet.h> #include <net/if.h> @@ -178,6 +179,7 @@ struct tuntap_softc { static struct mtx tunmtx; static eventhandler_tag arrival_tag; static eventhandler_tag clone_tag; +static int tuntap_osd_jail_slot; static const char tunname[] = "tun"; static const char tapname[] = "tap"; static const char vmnetname[] = "vmnet"; @@ -497,6 +499,10 @@ vmnet_clone_match(struct if_clone *ifc, const char *name) return (0); } +/* + * Create a clone via the ifnet cloning mechanism. Note that this is invoked + * indirectly by tunclone() below. + */ static int tun_clone_create(struct if_clone *ifc, char *name, size_t len, struct ifc_data *ifd, struct ifnet **ifpp) @@ -532,15 +538,19 @@ tun_clone_create(struct if_clone *ifc, char *name, size_t len, if (i != 0) i = tun_create_device(drv, unit, NULL, &dev, name); if (i == 0) { - dev_ref(dev); + struct tuntap_softc *tp; + tuncreate(dev); - struct tuntap_softc *tp = dev->si_drv1; + tp = dev->si_drv1; *ifpp = tp->tun_ifp; } return (i); } +/* + * Create a clone via devfs access. + */ static void tunclone(void *arg, struct ucred *cred, char *name, int namelen, struct cdev **dev) @@ -595,11 +605,12 @@ tunclone(void *arg, struct ucred *cred, char *name, int namelen, } i = tun_create_device(drv, u, cred, dev, name); - } - if (i == 0) { + } else { + /* Consumed by the dev_clone invoker. */ dev_ref(*dev); - if_clone_create(name, namelen, NULL); } + if (i == 0) + if_clone_create(name, namelen, NULL); out: CURVNET_RESTORE(); } @@ -670,16 +681,6 @@ VNET_SYSINIT(vnet_tun_init, SI_SUB_PROTO_IF, SI_ORDER_ANY, vnet_tun_init, NULL); static void -vnet_tun_uninit(const void *unused __unused) -{ - - for (u_int i = 0; i < NDRV; ++i) - if_clone_detach(V_tuntap_driver_cloners[i]); -} -VNET_SYSUNINIT(vnet_tun_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY, - vnet_tun_uninit, NULL); - -static void tun_uninit(const void *unused __unused) { struct tuntap_driver *drv; @@ -689,6 +690,16 @@ tun_uninit(const void *unused __unused) EVENTHANDLER_DEREGISTER(ifnet_arrival_event, arrival_tag); EVENTHANDLER_DEREGISTER(dev_clone, clone_tag); + CURVNET_SET(vnet0); + for (u_int i = 0; i < NDRV; i++) { + if_clone_detach(V_tuntap_driver_cloners[i]); + V_tuntap_driver_cloners[i] = NULL; + } + CURVNET_RESTORE(); + + if (tuntap_osd_jail_slot != 0) + osd_jail_deregister(tuntap_osd_jail_slot); + mtx_lock(&tunmtx); while ((tp = TAILQ_FIRST(&tunhead)) != NULL) { TAILQ_REMOVE(&tunhead, tp, tun_list); @@ -724,6 +735,30 @@ tuntap_driver_from_ifnet(const struct ifnet *ifp) return (NULL); } +/* + * Remove devices that were created by devfs cloning, as they hold references + * which prevent the prison from collapsing, in which state VNET sysuninits will + * not be invoked. + */ +static int +tuntap_prison_remove(void *obj, void *data __unused) +{ +#ifdef VIMAGE + struct prison *pr; + + pr = obj; + if (prison_owns_vnet(pr)) { + CURVNET_SET(pr->pr_vnet); + for (u_int i = 0; i < NDRV; i++) { + if_clone_detach(V_tuntap_driver_cloners[i]); + V_tuntap_driver_cloners[i] = NULL; + } + CURVNET_RESTORE(); + } +#endif + return (0); +} + static int tuntapmodevent(module_t mod, int type, void *data) { @@ -738,8 +773,12 @@ tuntapmodevent(module_t mod, int type, void *data) clone_setup(&drv->clones); drv->unrhdr = new_unrhdr(0, IF_MAXUNIT, &tunmtx); } + osd_method_t methods[PR_MAXMETHOD] = { + [PR_METHOD_REMOVE] = tuntap_prison_remove, + }; + tuntap_osd_jail_slot = osd_jail_register(NULL, methods); arrival_tag = EVENTHANDLER_REGISTER(ifnet_arrival_event, - tunrename, 0, 1000); + tunrename, 0, 1000); if (arrival_tag == NULL) return (ENOMEM); clone_tag = EVENTHANDLER_REGISTER(dev_clone, tunclone, 0, 1000); @@ -747,7 +786,7 @@ tuntapmodevent(module_t mod, int type, void *data) return (ENOMEM); break; case MOD_UNLOAD: - /* See tun_uninit, so it's done after the vnet_sysuninit() */ + /* See tun_uninit(). */ break; default: return EOPNOTSUPP; @@ -798,6 +837,8 @@ tun_create_device(struct tuntap_driver *drv, int unit, struct ucred *cr, args.mda_si_drv1 = tp; error = make_dev_s(&args, dev, "%s", name); if (error != 0) { + mtx_destroy(&tp->tun_mtx); + cv_destroy(&tp->tun_cv); free(tp, M_TUN); return (error); } @@ -914,7 +955,6 @@ tap_transmit(struct ifnet *ifp, struct mbuf *m) return (error); } -/* XXX: should return an error code so it can fail. */ static void tuncreate(struct cdev *dev) { diff --git a/sys/net/if_vlan.c b/sys/net/if_vlan.c index 22fcb7bf7c64..61000018e5a4 100644 --- a/sys/net/if_vlan.c +++ b/sys/net/if_vlan.c @@ -2336,6 +2336,18 @@ vlan_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) error = ENOENT; break; } + + /* + * If the ifp is in a bridge, do not allow setting the device + * to a bridge; this prevents having a bridge SVI as a bridge + * member (which is not permitted). + */ + if (ifp->if_bridge != NULL && p->if_type == IFT_BRIDGE) { + if_rele(p); + error = EINVAL; + break; + } + if (vlr.vlr_proto == 0) vlr.vlr_proto = ETHERTYPE_VLAN; oldmtu = ifp->if_mtu; diff --git a/sys/net/pfvar.h b/sys/net/pfvar.h index 1416f0c2cdbe..d55afe750869 100644 --- a/sys/net/pfvar.h +++ b/sys/net/pfvar.h @@ -331,6 +331,14 @@ MALLOC_DECLARE(M_PF_RULE_ITEM); SDT_PROVIDER_DECLARE(pf); SDT_PROBE_DECLARE(pf, , test, reason_set); +SDT_PROBE_DECLARE(pf, , log, log); + +#define DPFPRINTF(n, fmt, x...) \ + do { \ + SDT_PROBE2(pf, , log, log, (n), fmt); \ + if (V_pf_status.debug >= (n)) \ + printf(fmt "\n", ##x); \ + } while (0) struct pfi_dynaddr { TAILQ_ENTRY(pfi_dynaddr) entry; @@ -551,6 +559,9 @@ extern struct sx pf_end_lock; #endif /* PF_INET_INET6 */ #ifdef _KERNEL + +void unhandled_af(int) __dead2; + static void inline pf_addrcpy(struct pf_addr *dst, const struct pf_addr *src, sa_family_t af) { @@ -565,6 +576,8 @@ pf_addrcpy(struct pf_addr *dst, const struct pf_addr *src, sa_family_t af) memcpy(&dst->v6, &src->v6, sizeof(dst->v6)); break; #endif /* INET6 */ + default: + unhandled_af(af); } } #endif @@ -1365,7 +1378,6 @@ struct pf_kruleset { struct pf_krulequeue queues[2]; struct { struct pf_krulequeue *ptr; - struct pf_krule **ptr_array; u_int32_t rcount; u_int32_t ticket; int open; @@ -1672,6 +1684,9 @@ struct pf_pdesc { u_int32_t fragoff; /* fragment header offset */ u_int32_t jumbolen; /* length from v6 jumbo header */ u_int32_t badopts; /* v4 options or v6 routing headers */ +#define PF_OPT_OTHER 0x0001 +#define PF_OPT_JUMBO 0x0002 +#define PF_OPT_ROUTER_ALERT 0x0004 u_int16_t *ip_sum; u_int16_t flags; /* Let SCRUB trigger behavior in @@ -2300,7 +2315,6 @@ VNET_DECLARE(struct pf_krule *, pf_rulemarker); #define V_pf_rulemarker VNET(pf_rulemarker) #endif -void unhandled_af(int) __dead2; int pf_start(void); int pf_stop(void); void pf_initialize(void); @@ -2496,7 +2510,7 @@ int pfr_match_addr(struct pfr_ktable *, struct pf_addr *, sa_family_t); void pfr_update_stats(struct pfr_ktable *, struct pf_addr *, sa_family_t, u_int64_t, int, int, int); int pfr_pool_get(struct pfr_ktable *, int *, struct pf_addr *, sa_family_t, - pf_addr_filter_func_t); + pf_addr_filter_func_t, bool); void pfr_dynaddr_update(struct pfr_ktable *, struct pfi_dynaddr *); struct pfr_ktable * pfr_attach_table(struct pf_kruleset *, char *); @@ -2530,6 +2544,8 @@ int pfr_ina_rollback(struct pfr_table *, u_int32_t, int *, int); int pfr_ina_commit(struct pfr_table *, u_int32_t, int *, int *, int); int pfr_ina_define(struct pfr_table *, struct pfr_addr *, int, int *, int *, u_int32_t, int); +struct pfr_ktable + *pfr_ktable_select_active(struct pfr_ktable *); MALLOC_DECLARE(PFI_MTYPE); VNET_DECLARE(struct pfi_kkif *, pfi_all); @@ -2670,6 +2686,7 @@ int pf_ioctl_get_addrs(struct pf_nl_pooladdr *); int pf_ioctl_get_addr(struct pf_nl_pooladdr *); int pf_ioctl_get_rulesets(struct pfioc_ruleset *); int pf_ioctl_get_ruleset(struct pfioc_ruleset *); +int pf_ioctl_natlook(struct pfioc_natlook *); void pf_krule_free(struct pf_krule *); void pf_krule_clear_counters(struct pf_krule *); @@ -2707,7 +2724,6 @@ u_short pf_map_addr(u_int8_t, struct pf_krule *, u_short pf_map_addr_sn(u_int8_t, struct pf_krule *, struct pf_addr *, struct pf_addr *, struct pfi_kkif **nkif, struct pf_addr *, - struct pf_ksrc_node **, struct pf_srchash **, struct pf_kpool *, pf_sn_types_t); int pf_get_transaddr_af(struct pf_krule *, struct pf_pdesc *); diff --git a/sys/net80211/ieee80211_hostap.c b/sys/net80211/ieee80211_hostap.c index c5a478533313..9074878e17e4 100644 --- a/sys/net80211/ieee80211_hostap.c +++ b/sys/net80211/ieee80211_hostap.c @@ -2214,12 +2214,9 @@ hostap_recv_mgmt(struct ieee80211_node *ni, struct mbuf *m0, /* VHT */ if (IEEE80211_IS_CHAN_VHT(ni->ni_chan) && - vhtcap != NULL && - vhtinfo != NULL) { - /* XXX TODO; see below */ - net80211_vap_printf(vap, "%s: VHT TODO!\n", __func__); + vhtcap != NULL) { ieee80211_vht_node_init(ni); - ieee80211_vht_update_cap(ni, vhtcap, vhtinfo); + ieee80211_vht_update_cap(ni, vhtcap); } else if (ni->ni_flags & IEEE80211_NODE_VHT) ieee80211_vht_node_cleanup(ni); diff --git a/sys/net80211/ieee80211_ht.c b/sys/net80211/ieee80211_ht.c index 5ec80e3646b8..c28f124648a1 100644 --- a/sys/net80211/ieee80211_ht.c +++ b/sys/net80211/ieee80211_ht.c @@ -1952,6 +1952,11 @@ do { \ _RETURN_CHAN_BITS(0); /* + * TODO: should we bail out if there's no htinfo? + * Or just treat it as if we can't do the HT20/HT40 check? + */ + + /* * The original code was based on * 802.11ac-2013, Table 8-183x-VHT Operation Information subfields. * 802.11-2020, Table 9-274-VHT Operation Information subfields @@ -1962,8 +1967,12 @@ do { \ */ htinfo = (const struct ieee80211_ie_htinfo *)ni->ni_ies.htinfo_ie; - ht40 = ((htinfo->hi_byte1 & IEEE80211_HTINFO_TXWIDTH) == - IEEE80211_HTINFO_TXWIDTH_2040); + if (htinfo != NULL) + ht40 = ((htinfo->hi_byte1 & IEEE80211_HTINFO_TXWIDTH) == + IEEE80211_HTINFO_TXWIDTH_2040); + else + ht40 = false; + can_vht160 = can_vht80p80 = can_vht80 = false; /* 20 Mhz */ diff --git a/sys/net80211/ieee80211_vht.c b/sys/net80211/ieee80211_vht.c index e91977f1ef98..de0b691d4d2a 100644 --- a/sys/net80211/ieee80211_vht.c +++ b/sys/net80211/ieee80211_vht.c @@ -838,12 +838,10 @@ ieee80211_add_vhtinfo(uint8_t *frm, struct ieee80211_node *ni) } void -ieee80211_vht_update_cap(struct ieee80211_node *ni, const uint8_t *vhtcap_ie, - const uint8_t *vhtop_ie) +ieee80211_vht_update_cap(struct ieee80211_node *ni, const uint8_t *vhtcap_ie) { ieee80211_parse_vhtcap(ni, vhtcap_ie); - ieee80211_parse_vhtopmode(ni, vhtop_ie); } static struct ieee80211_channel * diff --git a/sys/net80211/ieee80211_vht.h b/sys/net80211/ieee80211_vht.h index 2964de63c343..a1529df4a85b 100644 --- a/sys/net80211/ieee80211_vht.h +++ b/sys/net80211/ieee80211_vht.h @@ -52,8 +52,7 @@ uint8_t * ieee80211_add_vhtinfo(uint8_t *frm, struct ieee80211_node *); uint8_t *ieee80211_add_vhtcap_ch(uint8_t *, struct ieee80211vap *, struct ieee80211_channel *); -void ieee80211_vht_update_cap(struct ieee80211_node *, - const uint8_t *, const uint8_t *); +void ieee80211_vht_update_cap(struct ieee80211_node *, const uint8_t *); struct ieee80211_channel * ieee80211_vht_adjust_channel(struct ieee80211com *, diff --git a/sys/netinet/icmp_var.h b/sys/netinet/icmp_var.h index b1f2b0ebf911..d6b75e482e35 100644 --- a/sys/netinet/icmp_var.h +++ b/sys/netinet/icmp_var.h @@ -104,11 +104,10 @@ extern int badport_bandlim(int); #define BANDLIM_ICMP_UNREACH 0 #define BANDLIM_ICMP_ECHO 1 #define BANDLIM_ICMP_TSTAMP 2 -#define BANDLIM_RST_CLOSEDPORT 3 /* No connection, and no listeners */ -#define BANDLIM_RST_OPENPORT 4 /* No connection, listener */ -#define BANDLIM_ICMP6_UNREACH 5 -#define BANDLIM_SCTP_OOTB 6 -#define BANDLIM_MAX 7 +#define BANDLIM_TCP_RST 3 +#define BANDLIM_ICMP6_UNREACH 4 +#define BANDLIM_SCTP_OOTB 5 +#define BANDLIM_MAX 6 #endif #endif diff --git a/sys/netinet/in_fib_dxr.c b/sys/netinet/in_fib_dxr.c index b889131b544b..538cd43a88a3 100644 --- a/sys/netinet/in_fib_dxr.c +++ b/sys/netinet/in_fib_dxr.c @@ -345,7 +345,7 @@ initheap(struct dxr_aux *da, uint32_t dst_u32, uint32_t chunk) struct heap_entry *fhp = &da->heap[0]; struct rtentry *rt; struct route_nhop_data rnd; - + da->heap_index = 0; da->dst.sin_addr.s_addr = htonl(dst_u32); rt = fib4_lookup_rt(da->fibnum, da->dst.sin_addr, 0, NHR_UNLOCKED, @@ -1143,7 +1143,7 @@ dxr_destroy(void *data) free(da, M_DXRAUX); } -static void +static void epoch_dxr_destroy(epoch_context_t ctx) { struct dxr *dxr = __containerof(ctx, struct dxr, epoch_ctx); @@ -1202,7 +1202,7 @@ dxr_dump_end(void *data, struct fib_dp *dp) static enum flm_op_result dxr_dump_rib_item(struct rtentry *rt, void *data) { - + return (FLM_SUCCESS); } diff --git a/sys/netinet/in_pcb.c b/sys/netinet/in_pcb.c index bccd4b84561a..dbe48242381d 100644 --- a/sys/netinet/in_pcb.c +++ b/sys/netinet/in_pcb.c @@ -1745,6 +1745,23 @@ in_pcbrele(struct inpcb *inp, const inp_lookup_t lock) } /* + * Dereference and rlock inp, for which the caller must own the + * reference. Returns true if inp no longer usable, false otherwise. + */ +bool +in_pcbrele_rlock(struct inpcb *inp) +{ + INP_RLOCK(inp); + if (in_pcbrele_rlocked(inp)) + return (true); + if ((inp->inp_flags & INP_FREED) != 0) { + INP_RUNLOCK(inp); + return (true); + } + return (false); +} + +/* * Unconditionally schedule an inpcb to be freed by decrementing its * reference count, which should occur only after the inpcb has been detached * from its socket. If another thread holds a temporary reference (acquired diff --git a/sys/netinet/in_pcb.h b/sys/netinet/in_pcb.h index 57cf15ca37fc..9e0618e87601 100644 --- a/sys/netinet/in_pcb.h +++ b/sys/netinet/in_pcb.h @@ -681,6 +681,7 @@ void in_pcbref(struct inpcb *); bool in_pcbrele(struct inpcb *, inp_lookup_t); bool in_pcbrele_rlocked(struct inpcb *); bool in_pcbrele_wlocked(struct inpcb *); +bool in_pcbrele_rlock(struct inpcb *inp); typedef bool inp_match_t(const struct inpcb *, void *); struct inpcb_iterator { diff --git a/sys/netinet/ip_icmp.c b/sys/netinet/ip_icmp.c index cb4b6df57c57..71b75d18efd0 100644 --- a/sys/netinet/ip_icmp.c +++ b/sys/netinet/ip_icmp.c @@ -1097,8 +1097,7 @@ static const char *icmp_rate_descrs[BANDLIM_MAX] = { [BANDLIM_ICMP_UNREACH] = "icmp unreach", [BANDLIM_ICMP_ECHO] = "icmp ping", [BANDLIM_ICMP_TSTAMP] = "icmp tstamp", - [BANDLIM_RST_CLOSEDPORT] = "closed port RST", - [BANDLIM_RST_OPENPORT] = "open port RST", + [BANDLIM_TCP_RST] = "tcp reset", [BANDLIM_ICMP6_UNREACH] = "icmp6 unreach", [BANDLIM_SCTP_OOTB] = "sctp ootb", }; diff --git a/sys/netinet/sctp_timer.c b/sys/netinet/sctp_timer.c index 66af716eea52..7d8cb965ab09 100644 --- a/sys/netinet/sctp_timer.c +++ b/sys/netinet/sctp_timer.c @@ -35,7 +35,6 @@ #define _IP_VHL #include <netinet/sctp_os.h> #include <netinet/sctp_pcb.h> - #include <netinet/sctp_var.h> #include <netinet/sctp_sysctl.h> #include <netinet/sctp_timer.h> diff --git a/sys/netinet/tcp_hpts.c b/sys/netinet/tcp_hpts.c index 91f8251589e4..b60cdf45af52 100644 --- a/sys/netinet/tcp_hpts.c +++ b/sys/netinet/tcp_hpts.c @@ -433,38 +433,40 @@ static void tcp_hpts_log(struct tcp_hpts_entry *hpts, struct tcpcb *tp, struct timeval *tv, int slots_to_run, int idx, bool from_callout) { - union tcp_log_stackspecific log; - /* - * Unused logs are - * 64 bit - delRate, rttProp, bw_inuse - * 16 bit - cwnd_gain - * 8 bit - bbr_state, bbr_substate, inhpts; - */ - memset(&log, 0, sizeof(log)); - log.u_bbr.flex1 = hpts->p_nxt_slot; - log.u_bbr.flex2 = hpts->p_cur_slot; - log.u_bbr.flex3 = hpts->p_prev_slot; - log.u_bbr.flex4 = idx; - log.u_bbr.flex5 = hpts->p_curtick; - log.u_bbr.flex6 = hpts->p_on_queue_cnt; - log.u_bbr.flex7 = hpts->p_cpu; - log.u_bbr.flex8 = (uint8_t)from_callout; - log.u_bbr.inflight = slots_to_run; - log.u_bbr.applimited = hpts->overidden_sleep; - log.u_bbr.delivered = hpts->saved_curtick; - log.u_bbr.timeStamp = tcp_tv_to_usectick(tv); - log.u_bbr.epoch = hpts->saved_curslot; - log.u_bbr.lt_epoch = hpts->saved_prev_slot; - log.u_bbr.pkts_out = hpts->p_delayed_by; - log.u_bbr.lost = hpts->p_hpts_sleep_time; - log.u_bbr.pacing_gain = hpts->p_cpu; - log.u_bbr.pkt_epoch = hpts->p_runningslot; - log.u_bbr.use_lt_bw = 1; - TCP_LOG_EVENTP(tp, NULL, - &tptosocket(tp)->so_rcv, - &tptosocket(tp)->so_snd, - BBR_LOG_HPTSDIAG, 0, - 0, &log, false, tv); + if (hpts_does_tp_logging && tcp_bblogging_on(tp)) { + union tcp_log_stackspecific log; + /* + * Unused logs are + * 64 bit - delRate, rttProp, bw_inuse + * 16 bit - cwnd_gain + * 8 bit - bbr_state, bbr_substate, inhpts; + */ + memset(&log, 0, sizeof(log)); + log.u_bbr.flex1 = hpts->p_nxt_slot; + log.u_bbr.flex2 = hpts->p_cur_slot; + log.u_bbr.flex3 = hpts->p_prev_slot; + log.u_bbr.flex4 = idx; + log.u_bbr.flex5 = hpts->p_curtick; + log.u_bbr.flex6 = hpts->p_on_queue_cnt; + log.u_bbr.flex7 = hpts->p_cpu; + log.u_bbr.flex8 = (uint8_t)from_callout; + log.u_bbr.inflight = slots_to_run; + log.u_bbr.applimited = hpts->overidden_sleep; + log.u_bbr.delivered = hpts->saved_curtick; + log.u_bbr.timeStamp = tcp_tv_to_usectick(tv); + log.u_bbr.epoch = hpts->saved_curslot; + log.u_bbr.lt_epoch = hpts->saved_prev_slot; + log.u_bbr.pkts_out = hpts->p_delayed_by; + log.u_bbr.lost = hpts->p_hpts_sleep_time; + log.u_bbr.pacing_gain = hpts->p_cpu; + log.u_bbr.pkt_epoch = hpts->p_runningslot; + log.u_bbr.use_lt_bw = 1; + TCP_LOG_EVENTP(tp, NULL, + &tptosocket(tp)->so_rcv, + &tptosocket(tp)->so_snd, + BBR_LOG_HPTSDIAG, 0, + 0, &log, false, tv); + } } static void @@ -1353,10 +1355,7 @@ again: } CURVNET_SET(inp->inp_vnet); /* Lets do any logging that we might want to */ - if (hpts_does_tp_logging && tcp_bblogging_on(tp)) { - tcp_hpts_log(hpts, tp, &tv, slots_to_run, i, - from_callout); - } + tcp_hpts_log(hpts, tp, &tv, slots_to_run, i, from_callout); if (tp->t_fb_ptr != NULL) { kern_prefetch(tp->t_fb_ptr, &did_prefetch); @@ -1487,7 +1486,7 @@ no_run: } void -__tcp_set_hpts(struct tcpcb *tp, int32_t line) +tcp_set_hpts(struct tcpcb *tp) { struct tcp_hpts_entry *hpts; int failed; diff --git a/sys/netinet/tcp_hpts.h b/sys/netinet/tcp_hpts.h index b097a2b98db9..f5856ed8e688 100644 --- a/sys/netinet/tcp_hpts.h +++ b/sys/netinet/tcp_hpts.h @@ -149,8 +149,7 @@ uint32_t tcp_hpts_insert_diag(struct tcpcb *tp, uint32_t slot, int32_t line, #define tcp_hpts_insert(inp, slot) \ tcp_hpts_insert_diag((inp), (slot), __LINE__, NULL) -void __tcp_set_hpts(struct tcpcb *tp, int32_t line); -#define tcp_set_hpts(a) __tcp_set_hpts(a, __LINE__) +void tcp_set_hpts(struct tcpcb *tp); void tcp_set_inp_to_drop(struct inpcb *inp, uint16_t reason); @@ -165,25 +164,25 @@ extern int32_t tcp_min_hptsi_time; * The following functions should also be available * to userspace as well. */ -static __inline uint32_t +static inline uint32_t tcp_tv_to_hptstick(const struct timeval *sv) { return ((sv->tv_sec * 100000) + (sv->tv_usec / HPTS_TICKS_PER_SLOT)); } -static __inline uint32_t +static inline uint32_t tcp_tv_to_usectick(const struct timeval *sv) { return ((uint32_t) ((sv->tv_sec * HPTS_USEC_IN_SEC) + sv->tv_usec)); } -static __inline uint32_t +static inline uint32_t tcp_tv_to_mssectick(const struct timeval *sv) { return ((uint32_t) ((sv->tv_sec * HPTS_MSEC_IN_SEC) + (sv->tv_usec/HPTS_USEC_IN_MSEC))); } -static __inline uint64_t +static inline uint64_t tcp_tv_to_lusectick(const struct timeval *sv) { return ((uint64_t)((sv->tv_sec * HPTS_USEC_IN_SEC) + sv->tv_usec)); @@ -199,7 +198,7 @@ get_hpts_min_sleep_time(void) return (tcp_min_hptsi_time + HPTS_TICKS_PER_SLOT); } -static __inline uint32_t +static inline uint32_t tcp_gethptstick(struct timeval *sv) { struct timeval tv; @@ -210,7 +209,7 @@ tcp_gethptstick(struct timeval *sv) return (tcp_tv_to_hptstick(sv)); } -static __inline uint64_t +static inline uint64_t tcp_get_u64_usecs(struct timeval *tv) { struct timeval tvd; @@ -221,7 +220,7 @@ tcp_get_u64_usecs(struct timeval *tv) return (tcp_tv_to_lusectick(tv)); } -static __inline uint32_t +static inline uint32_t tcp_get_usecs(struct timeval *tv) { struct timeval tvd; diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c index 7c032e13f37a..de428ae1af6f 100644 --- a/sys/netinet/tcp_input.c +++ b/sys/netinet/tcp_input.c @@ -621,6 +621,7 @@ tcp_input_with_port(struct mbuf **mp, int *offp, int proto, uint16_t port) #endif /* INET6 */ struct tcpopt to; /* options in this segment */ char *s = NULL; /* address and port logging */ + bool closed_port = false; /* segment is hitting a closed port */ NET_EPOCH_ASSERT(); @@ -907,7 +908,8 @@ findpcb: log(LOG_INFO, "%s; %s: Connection attempt " "to closed port\n", s, __func__); } - rstreason = BANDLIM_RST_CLOSEDPORT; + rstreason = BANDLIM_TCP_RST; + closed_port = true; goto dropwithreset; } INP_LOCK_ASSERT(inp); @@ -998,12 +1000,14 @@ findpcb: * down or it is in the CLOSED state. Either way we drop the * segment and send an appropriate response. */ - rstreason = BANDLIM_RST_CLOSEDPORT; + rstreason = BANDLIM_TCP_RST; + closed_port = true; goto dropwithreset; } if ((tp->t_port != port) && (tp->t_state > TCPS_LISTEN)) { - rstreason = BANDLIM_RST_CLOSEDPORT; + rstreason = BANDLIM_TCP_RST; + closed_port = true; goto dropwithreset; } @@ -1055,6 +1059,8 @@ findpcb: * socket appended to the listen queue in SYN_RECEIVED state. */ if ((thflags & (TH_RST|TH_ACK|TH_SYN)) == TH_ACK) { + int result; + /* * Parse the TCP options here because * syncookies need access to the reflected @@ -1064,8 +1070,8 @@ findpcb: /* * NB: syncache_expand() doesn't unlock inp. */ - rstreason = syncache_expand(&inc, &to, th, &so, m, port); - if (rstreason < 0) { + result = syncache_expand(&inc, &to, th, &so, m, port); + if (result < 0) { /* * A failing TCP MD5 signature comparison * must result in the segment being dropped @@ -1073,7 +1079,7 @@ findpcb: * to the sender. */ goto dropunlock; - } else if (rstreason == 0) { + } else if (result == 0) { /* * No syncache entry, or ACK was not for our * SYN/ACK. Do our protection against double @@ -1092,7 +1098,7 @@ findpcb: * of the failure cause. */ INP_WUNLOCK(inp); - rstreason = BANDLIM_RST_OPENPORT; + rstreason = BANDLIM_TCP_RST; lookupflag &= ~INPLOOKUP_WILDCARD; goto findpcb; } @@ -1183,7 +1189,7 @@ tfo_socket_result: s, __func__); syncache_badack(&inc, port); /* XXX: Not needed! */ TCPSTAT_INC(tcps_badsyn); - rstreason = BANDLIM_RST_OPENPORT; + rstreason = BANDLIM_TCP_RST; goto dropwithreset; } /* @@ -1259,7 +1265,7 @@ tfo_socket_result: "Connection attempt to deprecated " "IPv6 address rejected\n", s, __func__); - rstreason = BANDLIM_RST_OPENPORT; + rstreason = BANDLIM_TCP_RST; goto dropwithreset; } } @@ -1380,9 +1386,10 @@ dropwithreset: * When blackholing do not respond with a RST but * completely ignore the segment and drop it. */ - if (((rstreason == BANDLIM_RST_OPENPORT && V_blackhole == 3) || - (rstreason == BANDLIM_RST_CLOSEDPORT && - ((V_blackhole == 1 && (thflags & TH_SYN)) || V_blackhole > 1))) && + if (rstreason == BANDLIM_TCP_RST && + ((!closed_port && V_blackhole == 3) || + (closed_port && + ((V_blackhole == 1 && (thflags & TH_SYN)) || V_blackhole > 1))) && (V_blackhole_local || ( #ifdef INET6 isipv6 ? !in6_localip(&ip6->ip6_src) : @@ -1515,7 +1522,9 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, struct tcpopt to; int tfo_syn; u_int maxseg = 0; + bool no_data; + no_data = (tlen == 0); thflags = tcp_get_flags(th); tp->sackhint.last_sack_ack = 0; sack_changed = SACK_NOCHANGE; @@ -1754,7 +1763,7 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, tp->ts_recent = to.to_tsval; } - if (tlen == 0) { + if (no_data) { if (SEQ_GT(th->th_ack, tp->snd_una) && SEQ_LEQ(th->th_ack, tp->snd_max) && !IN_RECOVERY(tp->t_flags) && @@ -1963,7 +1972,7 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, if ((thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->snd_una) || SEQ_GT(th->th_ack, tp->snd_max))) { - rstreason = BANDLIM_RST_OPENPORT; + rstreason = BANDLIM_TCP_RST; tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); goto dropwithreset; } @@ -1976,7 +1985,7 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, * FIN, or a RST. */ if ((thflags & (TH_SYN|TH_ACK)) == (TH_SYN|TH_ACK)) { - rstreason = BANDLIM_RST_OPENPORT; + rstreason = BANDLIM_TCP_RST; tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); goto dropwithreset; } else if (thflags & TH_SYN) { @@ -2244,7 +2253,7 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, * for the "LAND" DoS attack. */ if (tp->t_state == TCPS_SYN_RECEIVED && SEQ_LT(th->th_seq, tp->irs)) { - rstreason = BANDLIM_RST_OPENPORT; + rstreason = BANDLIM_TCP_RST; tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); goto dropwithreset; } @@ -2557,7 +2566,7 @@ tcp_do_segment(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, if (SEQ_LEQ(th->th_ack, tp->snd_una)) { maxseg = tcp_maxseg(tp); - if (tlen == 0 && + if (no_data && (tiwin == tp->snd_wnd || (tp->t_flags & TF_SACK_PERMIT))) { /* @@ -3113,8 +3122,7 @@ step6: (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) || (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) { /* keep track of pure window updates */ - if (tlen == 0 && - tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) + if (no_data && tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd) TCPSTAT_INC(tcps_rcvwinupd); tp->snd_wnd = tiwin; tp->snd_wl1 = th->th_seq; @@ -3424,7 +3432,7 @@ dropafterack: if (tp->t_state == TCPS_SYN_RECEIVED && (thflags & TH_ACK) && (SEQ_GT(tp->snd_una, th->th_ack) || SEQ_GT(th->th_ack, tp->snd_max)) ) { - rstreason = BANDLIM_RST_OPENPORT; + rstreason = BANDLIM_TCP_RST; tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); goto dropwithreset; } diff --git a/sys/netinet/tcp_log_buf.c b/sys/netinet/tcp_log_buf.c index 75d693bc019b..e24790ece43d 100644 --- a/sys/netinet/tcp_log_buf.c +++ b/sys/netinet/tcp_log_buf.c @@ -2878,7 +2878,7 @@ tcp_log_sendfile(struct socket *so, off_t offset, size_t nbytes, int flags) /* double check log state now that we have the lock */ if (inp->inp_flags & INP_DROPPED) goto done; - if (tp->_t_logstate != TCP_LOG_STATE_OFF) { + if (tcp_bblogging_on(tp)) { struct timeval tv; tcp_log_eventspecific_t log; diff --git a/sys/netinet/tcp_log_buf.h b/sys/netinet/tcp_log_buf.h index fef32e16b2e4..f8c064b6a104 100644 --- a/sys/netinet/tcp_log_buf.h +++ b/sys/netinet/tcp_log_buf.h @@ -377,12 +377,12 @@ extern int32_t tcp_trace_point_count; /* * Returns true if any sort of BB logging is enabled, - * commonly used throughout the codebase. + * commonly used throughout the codebase. */ static inline int tcp_bblogging_on(struct tcpcb *tp) { - if (tp->_t_logstate <= TCP_LOG_STATE_OFF) + if (tp->_t_logstate <= TCP_LOG_STATE_OFF) return (0); if (tp->_t_logstate == TCP_LOG_VIA_BBPOINTS) return (0); @@ -427,7 +427,7 @@ tcp_set_bblog_state(struct tcpcb *tp, uint8_t ls, uint8_t bbpoint) } } -static inline uint32_t +static inline uint32_t tcp_get_bblog_state(struct tcpcb *tp) { return (tp->_t_logstate); @@ -539,12 +539,12 @@ struct tcpcb; NULL, NULL, 0, NULL); \ } while (0) #endif /* TCP_LOG_FORCEVERBOSE */ +/* Assumes/requires the caller has already checked tcp_bblogging_on(tp). */ #define TCP_LOG_EVENTP(tp, th, rxbuf, txbuf, eventid, errornum, len, stackinfo, th_hostorder, tv) \ do { \ - if (tcp_bblogging_on(tp)) \ - tcp_log_event(tp, th, rxbuf, txbuf, eventid, \ - errornum, len, stackinfo, th_hostorder, \ - NULL, NULL, 0, tv); \ + KASSERT(tcp_bblogging_on(tp), ("bblogging is off")); \ + tcp_log_event(tp, th, rxbuf, txbuf, eventid, errornum, len, \ + stackinfo, th_hostorder, NULL, NULL, 0, tv); \ } while (0) #ifdef TCP_BLACKBOX diff --git a/sys/netinet/tcp_lro.c b/sys/netinet/tcp_lro.c index 10afed17bf3b..7512679bd4e9 100644 --- a/sys/netinet/tcp_lro.c +++ b/sys/netinet/tcp_lro.c @@ -1301,9 +1301,9 @@ tcp_lro_rx_common(struct lro_ctrl *lc, struct mbuf *m, uint32_t csum, bool use_h return (TCP_LRO_CANNOT); #endif if (((m->m_pkthdr.csum_flags & (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)) != - ((CSUM_DATA_VALID | CSUM_PSEUDO_HDR))) || + ((CSUM_DATA_VALID | CSUM_PSEUDO_HDR))) || (m->m_pkthdr.csum_data != 0xffff)) { - /* + /* * The checksum either did not have hardware offload * or it was a bad checksum. We can't LRO such * a packet. @@ -1334,7 +1334,7 @@ tcp_lro_rx_common(struct lro_ctrl *lc, struct mbuf *m, uint32_t csum, bool use_h #endif /* If no hardware or arrival stamp on the packet add timestamp */ if ((m->m_flags & (M_TSTMP_LRO | M_TSTMP)) == 0) { - m->m_pkthdr.rcv_tstmp = bintime2ns(&lc->lro_last_queue_time); + m->m_pkthdr.rcv_tstmp = bintime2ns(&lc->lro_last_queue_time); m->m_flags |= M_TSTMP_LRO; } @@ -1429,9 +1429,9 @@ tcp_lro_rx(struct lro_ctrl *lc, struct mbuf *m, uint32_t csum) int error; if (((m->m_pkthdr.csum_flags & (CSUM_DATA_VALID | CSUM_PSEUDO_HDR)) != - ((CSUM_DATA_VALID | CSUM_PSEUDO_HDR))) || + ((CSUM_DATA_VALID | CSUM_PSEUDO_HDR))) || (m->m_pkthdr.csum_data != 0xffff)) { - /* + /* * The checksum either did not have hardware offload * or it was a bad checksum. We can't LRO such * a packet. @@ -1481,7 +1481,7 @@ tcp_lro_queue_mbuf(struct lro_ctrl *lc, struct mbuf *mb) ((mb->m_flags & M_TSTMP) == 0)) { /* Add in an LRO time since no hardware */ binuptime(&lc->lro_last_queue_time); - mb->m_pkthdr.rcv_tstmp = bintime2ns(&lc->lro_last_queue_time); + mb->m_pkthdr.rcv_tstmp = bintime2ns(&lc->lro_last_queue_time); mb->m_flags |= M_TSTMP_LRO; } diff --git a/sys/netinet/tcp_sack.c b/sys/netinet/tcp_sack.c index 90d789f0e224..4405098a8620 100644 --- a/sys/netinet/tcp_sack.c +++ b/sys/netinet/tcp_sack.c @@ -744,7 +744,7 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) while (cur != NULL) { if (!(sblkp >= sack_blocks)) { if (((loss_sblks >= tcprexmtthresh) || - (loss_thresh > (tcprexmtthresh-1)*tp->t_maxseg))) + (loss_thresh > (tcprexmtthresh-1)*tp->t_maxseg))) break; loss_thresh += loss_hiack - cur->end; loss_hiack = cur->start; diff --git a/sys/netinet/tcp_stacks/bbr.c b/sys/netinet/tcp_stacks/bbr.c index e2cfec5c9275..b232d3f08fe6 100644 --- a/sys/netinet/tcp_stacks/bbr.c +++ b/sys/netinet/tcp_stacks/bbr.c @@ -5126,8 +5126,8 @@ bbr_timeout_rxt(struct tcpcb *tp, struct tcp_bbr *bbr, uint32_t cts) tp->t_maxseg = tp->t_pmtud_saved_maxseg; if (tp->t_maxseg < V_tcp_mssdflt) { /* - * The MSS is so small we should not - * process incoming SACK's since we are + * The MSS is so small we should not + * process incoming SACK's since we are * subject to attack in such a case. */ tp->t_flags2 |= TF2_PROC_SACK_PROHIBIT; @@ -8763,7 +8763,7 @@ bbr_do_syn_sent(struct mbuf *m, struct tcphdr *th, struct socket *so, (SEQ_LEQ(th->th_ack, tp->iss) || SEQ_GT(th->th_ack, tp->snd_max))) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } if ((thflags & (TH_ACK | TH_RST)) == (TH_ACK | TH_RST)) { @@ -8965,7 +8965,7 @@ bbr_do_syn_recv(struct mbuf *m, struct tcphdr *th, struct socket *so, (SEQ_LEQ(th->th_ack, tp->snd_una) || SEQ_GT(th->th_ack, tp->snd_max))) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } if (tp->t_flags & TF_FASTOPEN) { @@ -8977,7 +8977,7 @@ bbr_do_syn_recv(struct mbuf *m, struct tcphdr *th, struct socket *so, */ if ((thflags & (TH_SYN | TH_ACK)) == (TH_SYN | TH_ACK)) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } else if (thflags & TH_SYN) { /* non-initial SYN is ignored */ @@ -9010,7 +9010,7 @@ bbr_do_syn_recv(struct mbuf *m, struct tcphdr *th, struct socket *so, */ if (SEQ_LT(th->th_seq, tp->irs)) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } if (ctf_drop_checks(to, m, th, tp, &tlen, &thflags, &drop_hdrlen, &ret_val)) { @@ -9288,7 +9288,7 @@ bbr_do_established(struct mbuf *m, struct tcphdr *th, struct socket *so, if (sbavail(&so->so_snd)) { if (ctf_progress_timeout_check(tp, true)) { bbr_log_progress_event(bbr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -9385,7 +9385,7 @@ bbr_do_close_wait(struct mbuf *m, struct tcphdr *th, struct socket *so, if (sbavail(&so->so_snd)) { if (ctf_progress_timeout_check(tp, true)) { bbr_log_progress_event(bbr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -9535,7 +9535,7 @@ bbr_do_fin_wait_1(struct mbuf *m, struct tcphdr *th, struct socket *so, if (sbavail(&so->so_snd)) { if (ctf_progress_timeout_check(tp, true)) { bbr_log_progress_event(bbr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -9637,7 +9637,7 @@ bbr_do_closing(struct mbuf *m, struct tcphdr *th, struct socket *so, if (sbavail(&so->so_snd)) { if (ctf_progress_timeout_check(tp, true)) { bbr_log_progress_event(bbr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -9739,7 +9739,7 @@ bbr_do_lastack(struct mbuf *m, struct tcphdr *th, struct socket *so, if (sbavail(&so->so_snd)) { if (ctf_progress_timeout_check(tp, true)) { bbr_log_progress_event(bbr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -9848,7 +9848,7 @@ bbr_do_fin_wait_2(struct mbuf *m, struct tcphdr *th, struct socket *so, if (sbavail(&so->so_snd)) { if (ctf_progress_timeout_check(tp, true)) { bbr_log_progress_event(bbr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -10141,7 +10141,7 @@ bbr_init(struct tcpcb *tp, void **ptr) * flags. */ bbr_stop_all_timers(tp, bbr); - /* + /* * Validate the timers are not in usec, if they are convert. * BBR should in theory move to USEC and get rid of a * lot of the TICKS_2 calls.. but for now we stay @@ -11510,7 +11510,7 @@ bbr_do_segment_nounlock(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, if ((tp->t_state == TCPS_SYN_SENT) && (thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->iss) || SEQ_GT(th->th_ack, tp->snd_max))) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } if (tiwin > bbr->r_ctl.rc_high_rwnd) @@ -11544,7 +11544,7 @@ bbr_do_segment_nounlock(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, bbr_check_bbr_for_state(bbr, cts, __LINE__, (bbr->r_ctl.rc_lost - lost)); if (nxt_pkt == 0) { if ((bbr->r_wanted_output != 0) || - (tp->t_flags & TF_ACKNOW)) { + (tp->t_flags & TF_ACKNOW)) { bbr->rc_output_starts_timer = 0; did_out = 1; @@ -13172,11 +13172,7 @@ send: mb, moff, &len, if_hw_tsomaxsegcount, if_hw_tsomaxsegsize, msb, - ((rsm == NULL) ? hw_tls : 0) -#ifdef NETFLIX_COPY_ARGS - , NULL, NULL -#endif - ); + ((rsm == NULL) ? hw_tls : 0)); if (len <= maxseg) { /* * Must have ran out of mbufs for the copy @@ -13806,8 +13802,8 @@ nomore: tp->t_maxseg = old_maxseg - 40; if (tp->t_maxseg < V_tcp_mssdflt) { /* - * The MSS is so small we should not - * process incoming SACK's since we are + * The MSS is so small we should not + * process incoming SACK's since we are * subject to attack in such a case. */ tp->t_flags2 |= TF2_PROC_SACK_PROHIBIT; diff --git a/sys/netinet/tcp_stacks/rack.c b/sys/netinet/tcp_stacks/rack.c index 8e05498863b9..940a4024bb73 100644 --- a/sys/netinet/tcp_stacks/rack.c +++ b/sys/netinet/tcp_stacks/rack.c @@ -40,7 +40,6 @@ #endif #include <sys/lock.h> #include <sys/malloc.h> -#include <sys/lock.h> #include <sys/mutex.h> #include <sys/mbuf.h> #include <sys/proc.h> /* for proc0 declaration */ @@ -198,7 +197,7 @@ static uint32_t rack_pcm_blast = 0; static uint32_t rack_pcm_is_enabled = 1; static uint8_t rack_ssthresh_rest_rto_rec = 0; /* Do we restore ssthresh when we have rec -> rto -> rec */ -static uint32_t rack_gp_gain_req = 1200; /* Amount percent wise required to gain to record a round has "gaining" */ +static uint32_t rack_gp_gain_req = 1200; /* Amount percent wise required to gain to record a round as "gaining" */ static uint32_t rack_rnd_cnt_req = 0x10005; /* Default number of rounds if we are below rack_gp_gain_req where we exit ss */ @@ -938,7 +937,7 @@ rack_init_sysctls(void) SYSCTL_ADD_U32(&rack_sysctl_ctx, SYSCTL_CHILDREN(rack_probertt), OID_AUTO, "time_between", CTLFLAG_RW, - & rack_time_between_probertt, 96000000, + &rack_time_between_probertt, 96000000, "How many useconds between the lowest rtt falling must past before we enter probertt"); SYSCTL_ADD_U32(&rack_sysctl_ctx, SYSCTL_CHILDREN(rack_probertt), @@ -3480,9 +3479,9 @@ static void rack_free(struct tcp_rack *rack, struct rack_sendmap *rsm) { if (rsm->r_flags & RACK_APP_LIMITED) { - if (rack->r_ctl.rc_app_limited_cnt > 0) { - rack->r_ctl.rc_app_limited_cnt--; - } + KASSERT((rack->r_ctl.rc_app_limited_cnt > 0), + ("app_cnt %u, rsm %p", rack->r_ctl.rc_app_limited_cnt, rsm)); + rack->r_ctl.rc_app_limited_cnt--; } if (rsm->r_limit_type) { /* currently there is only one limit type */ @@ -3554,8 +3553,7 @@ rack_get_measure_window(struct tcpcb *tp, struct tcp_rack *rack) * earlier. * * So lets calculate the BDP with the "known" b/w using - * the SRTT has our rtt and then multiply it by the - * goal. + * the SRTT as our rtt and then multiply it by the goal. */ bw = rack_get_bw(rack); srtt = (uint64_t)tp->t_srtt; @@ -5793,7 +5791,7 @@ rack_cong_signal(struct tcpcb *tp, uint32_t type, uint32_t ack, int line) tp->t_badrxtwin = 0; break; } - if ((CC_ALGO(tp)->cong_signal != NULL) && + if ((CC_ALGO(tp)->cong_signal != NULL) && (type != CC_RTO)){ tp->t_ccv.curack = ack; CC_ALGO(tp)->cong_signal(&tp->t_ccv, type); @@ -5904,7 +5902,7 @@ rack_calc_thresh_rack(struct tcp_rack *rack, uint32_t srtt, uint32_t cts, int li * * If reorder-fade is configured, then we track the last time we saw * re-ordering occur. If we reach the point where enough time as - * passed we no longer consider reordering has occuring. + * passed we no longer consider reordering as occurring. * * Or if reorder-face is 0, then once we see reordering we consider * the connection to alway be subject to reordering and just set lro @@ -7045,6 +7043,9 @@ rack_clone_rsm(struct tcp_rack *rack, struct rack_sendmap *nrsm, /* Push bit must go to the right edge as well */ if (rsm->r_flags & RACK_HAD_PUSH) rsm->r_flags &= ~RACK_HAD_PUSH; + /* Update the count if app limited */ + if (nrsm->r_flags & RACK_APP_LIMITED) + rack->r_ctl.rc_app_limited_cnt++; /* Clone over the state of the hw_tls flag */ nrsm->r_hw_tls = rsm->r_hw_tls; /* @@ -7096,7 +7097,7 @@ rack_merge_rsm(struct tcp_rack *rack, l_rsm->r_flags |= RACK_TLP; if (r_rsm->r_flags & RACK_RWND_COLLAPSED) l_rsm->r_flags |= RACK_RWND_COLLAPSED; - if ((r_rsm->r_flags & RACK_APP_LIMITED) && + if ((r_rsm->r_flags & RACK_APP_LIMITED) && ((l_rsm->r_flags & RACK_APP_LIMITED) == 0)) { /* * If both are app-limited then let the @@ -7887,8 +7888,8 @@ drop_it: tp->t_maxseg = tp->t_pmtud_saved_maxseg; if (tp->t_maxseg < V_tcp_mssdflt) { /* - * The MSS is so small we should not - * process incoming SACK's since we are + * The MSS is so small we should not + * process incoming SACK's since we are * subject to attack in such a case. */ tp->t_flags2 |= TF2_PROC_SACK_PROHIBIT; @@ -8137,7 +8138,7 @@ rack_update_rsm(struct tcpcb *tp, struct tcp_rack *rack, * remove the lost desgination and reduce the * bytes considered lost. */ - rsm->r_flags &= ~RACK_WAS_LOST; + rsm->r_flags &= ~RACK_WAS_LOST; KASSERT((rack->r_ctl.rc_considered_lost >= (rsm->r_end - rsm->r_start)), ("rsm:%p rack:%p rc_considered_lost goes negative", rsm, rack)); if (rack->r_ctl.rc_considered_lost >= (rsm->r_end - rsm->r_start)) @@ -8832,7 +8833,7 @@ rack_apply_updated_usrtt(struct tcp_rack *rack, uint32_t us_rtt, uint32_t us_cts val = rack_probertt_lower_within * rack_time_between_probertt; val /= 100; - if ((rack->in_probe_rtt == 0) && + if ((rack->in_probe_rtt == 0) && (rack->rc_skip_timely == 0) && ((us_cts - rack->r_ctl.rc_lower_rtt_us_cts) >= (rack_time_between_probertt - val))) { rack_enter_probertt(rack, us_cts); @@ -10369,7 +10370,7 @@ more: * and yet before retransmitting we get an ack * which can happen due to reordering. */ - rsm->r_flags &= ~RACK_WAS_LOST; + rsm->r_flags &= ~RACK_WAS_LOST; KASSERT((rack->r_ctl.rc_considered_lost >= (rsm->r_end - rsm->r_start)), ("rsm:%p rack:%p rc_considered_lost goes negative", rsm, rack)); if (rack->r_ctl.rc_considered_lost >= (rsm->r_end - rsm->r_start)) @@ -11065,7 +11066,7 @@ rack_strike_dupack(struct tcp_rack *rack, tcp_seq th_ack) * We need to skip anything already set * to be retransmitted. */ - if ((rsm->r_dupack >= DUP_ACK_THRESHOLD) || + if ((rsm->r_dupack >= DUP_ACK_THRESHOLD) || (rsm->r_flags & RACK_MUST_RXT)) { rsm = TAILQ_NEXT(rsm, r_tnext); continue; @@ -12875,7 +12876,7 @@ rack_do_syn_sent(struct mbuf *m, struct tcphdr *th, struct socket *so, (SEQ_LEQ(th->th_ack, tp->iss) || SEQ_GT(th->th_ack, tp->snd_max))) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } if ((thflags & (TH_ACK | TH_RST)) == (TH_ACK | TH_RST)) { @@ -13089,7 +13090,7 @@ rack_do_syn_recv(struct mbuf *m, struct tcphdr *th, struct socket *so, (SEQ_LEQ(th->th_ack, tp->snd_una) || SEQ_GT(th->th_ack, tp->snd_max))) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } if (tp->t_flags & TF_FASTOPEN) { @@ -13102,7 +13103,7 @@ rack_do_syn_recv(struct mbuf *m, struct tcphdr *th, struct socket *so, */ if ((thflags & (TH_SYN | TH_ACK)) == (TH_SYN | TH_ACK)) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } else if (thflags & TH_SYN) { /* non-initial SYN is ignored */ @@ -13136,7 +13137,7 @@ rack_do_syn_recv(struct mbuf *m, struct tcphdr *th, struct socket *so, */ if (SEQ_LT(th->th_seq, tp->irs)) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } if (ctf_drop_checks(to, m, th, tp, &tlen, &thflags, &drop_hdrlen, &ret_val)) { @@ -13399,7 +13400,7 @@ rack_do_established(struct mbuf *m, struct tcphdr *th, struct socket *so, if (sbavail(&so->so_snd)) { if (ctf_progress_timeout_check(tp, true)) { rack_log_progress_event(rack, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -13495,7 +13496,7 @@ rack_do_close_wait(struct mbuf *m, struct tcphdr *th, struct socket *so, if (ctf_progress_timeout_check(tp, true)) { rack_log_progress_event((struct tcp_rack *)tp->t_fb_ptr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -13645,7 +13646,7 @@ rack_do_fin_wait_1(struct mbuf *m, struct tcphdr *th, struct socket *so, if (ctf_progress_timeout_check(tp, true)) { rack_log_progress_event((struct tcp_rack *)tp->t_fb_ptr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -13746,7 +13747,7 @@ rack_do_closing(struct mbuf *m, struct tcphdr *th, struct socket *so, if (ctf_progress_timeout_check(tp, true)) { rack_log_progress_event((struct tcp_rack *)tp->t_fb_ptr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -13848,7 +13849,7 @@ rack_do_lastack(struct mbuf *m, struct tcphdr *th, struct socket *so, if (ctf_progress_timeout_check(tp, true)) { rack_log_progress_event((struct tcp_rack *)tp->t_fb_ptr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -13952,7 +13953,7 @@ rack_do_fin_wait_2(struct mbuf *m, struct tcphdr *th, struct socket *so, if (ctf_progress_timeout_check(tp, true)) { rack_log_progress_event((struct tcp_rack *)tp->t_fb_ptr, tp, tick, PROGRESS_DROP, __LINE__); - ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset_conn(m, tp, th, BANDLIM_TCP_RST, tlen); return (1); } } @@ -14637,9 +14638,6 @@ rack_init(struct tcpcb *tp, void **ptr) if (rack->r_ctl.pcm_s == NULL) { rack->r_ctl.pcm_i.cnt_alloc = 0; } -#ifdef NETFLIX_STATS - rack->r_ctl.side_chan_dis_mask = tcp_sidechannel_disable_mask; -#endif rack->r_ctl.rack_per_upper_bound_ss = (uint8_t)rack_per_upper_bound_ss; rack->r_ctl.rack_per_upper_bound_ca = (uint8_t)rack_per_upper_bound_ca; if (rack_enable_shared_cwnd) @@ -15563,7 +15561,7 @@ rack_log_pcm(struct tcp_rack *rack, uint8_t mod, uint32_t flex1, uint32_t flex2, if (tcp_bblogging_on(rack->rc_tp)) { union tcp_log_stackspecific log; struct timeval tv; - + (void)tcp_get_usecs(&tv); memset(&log, 0, sizeof(log)); log.u_bbr.timeStamp = tcp_tv_to_usectick(&tv); @@ -16655,7 +16653,7 @@ rack_do_segment_nounlock(struct tcpcb *tp, struct mbuf *m, struct tcphdr *th, if ((tp->t_state == TCPS_SYN_SENT) && (thflags & TH_ACK) && (SEQ_LEQ(th->th_ack, tp->iss) || SEQ_GT(th->th_ack, tp->snd_max))) { tcp_log_end_status(tp, TCP_EI_STATUS_RST_IN_FRONT); - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); #ifdef TCP_ACCOUNTING sched_unpin(); #endif @@ -16919,7 +16917,7 @@ do_output_now: } else if ((nxt_pkt == 0) && (tp->t_flags & TF_ACKNOW)) { goto do_output_now; } else if ((no_output == 1) && - (nxt_pkt == 0) && + (nxt_pkt == 0) && (tcp_in_hpts(rack->rc_tp) == 0)) { /* * We are not in hpts and we had a pacing timer up. Use @@ -17546,7 +17544,7 @@ rack_get_pacing_delay(struct tcp_rack *rack, struct tcpcb *tp, uint32_t len, str rack->r_ctl.rc_last_us_rtt, 88, __LINE__, NULL, gain); } - if (((bw_est == 0) || (rate_wanted == 0) || (rack->gp_ready == 0)) && + if (((bw_est == 0) || (rate_wanted == 0) || (rack->gp_ready == 0)) && (rack->use_fixed_rate == 0)) { /* * No way yet to make a b/w estimate or @@ -17986,7 +17984,7 @@ start_set: tp->gput_ack = tp->gput_seq + rack_get_measure_window(tp, rack); rack->r_ctl.rc_gp_cumack_ts = 0; if ((rack->r_ctl.cleared_app_ack == 1) && - (SEQ_GEQ(rack->r_ctl.cleared_app_ack, tp->gput_seq))) { + (SEQ_GEQ(tp->gput_seq, rack->r_ctl.cleared_app_ack_seq))) { /* * We just cleared an application limited period * so the next seq out needs to skip the first @@ -19914,7 +19912,7 @@ rack_output(struct tcpcb *tp) goto nomore; } else { /* Return == 0, if there is more we can send tot_len wise fall through and send */ - if (tot_len_this_send >= pace_max_seg) + if (tot_len_this_send >= pace_max_seg) return (ret); #ifdef TCP_ACCOUNTING /* We need to re-pin since fast_output un-pined */ @@ -20043,7 +20041,7 @@ again: rack->r_ctl.pcm_max_seg = ctf_fixed_maxseg(tp) * 10; } } - if ((rack->r_ctl.pcm_max_seg != 0) && (rack->pcm_needed == 1)) { + if ((rack->r_ctl.pcm_max_seg != 0) && (rack->pcm_needed == 1)) { uint32_t rw_avail, cwa; if (tp->snd_wnd > ctf_outstanding(tp)) @@ -21031,7 +21029,7 @@ just_return_nolock: } else log = 1; } - /* Mark the last packet has app limited */ + /* Mark the last packet as app limited */ rsm = tqhash_max(rack->r_ctl.tqh); if (rsm && ((rsm->r_flags & RACK_APP_LIMITED) == 0)) { if (rack->r_ctl.rc_app_limited_cnt == 0) @@ -21555,11 +21553,7 @@ send: m->m_next = tcp_m_copym( mb, moff, &len, if_hw_tsomaxsegcount, if_hw_tsomaxsegsize, msb, - ((rsm == NULL) ? hw_tls : 0) -#ifdef NETFLIX_COPY_ARGS - , &s_mb, &s_moff -#endif - ); + ((rsm == NULL) ? hw_tls : 0)); if (len <= (tp->t_maxseg - optlen)) { /* * Must have ran out of mbufs for the copy diff --git a/sys/netinet/tcp_stacks/rack_bbr_common.c b/sys/netinet/tcp_stacks/rack_bbr_common.c index da26b8cb1f9b..d1c4ba58bf55 100644 --- a/sys/netinet/tcp_stacks/rack_bbr_common.c +++ b/sys/netinet/tcp_stacks/rack_bbr_common.c @@ -672,7 +672,7 @@ ctf_do_dropafterack(struct mbuf *m, struct tcpcb *tp, struct tcphdr *th, int32_t (SEQ_GT(tp->snd_una, th->th_ack) || SEQ_GT(th->th_ack, tp->snd_max))) { *ret_val = 1; - ctf_do_dropwithreset(m, tp, th, BANDLIM_RST_OPENPORT, tlen); + ctf_do_dropwithreset(m, tp, th, BANDLIM_TCP_RST, tlen); return; } else *ret_val = 0; diff --git a/sys/netinet/tcp_stacks/rack_pcm.c b/sys/netinet/tcp_stacks/rack_pcm.c index b0e300847c4a..101e6826536c 100644 --- a/sys/netinet/tcp_stacks/rack_pcm.c +++ b/sys/netinet/tcp_stacks/rack_pcm.c @@ -172,7 +172,7 @@ rack_update_pcm_ack(struct tcp_rack *rack, int was_cumack, uint32_t start, uint3 goto skip_ack_accounting; } /* - * Record ACK data. + * Record ACK data. */ ack_arrival = tcp_tv_to_lusectick(&rack->r_ctl.act_rcv_time); if (SEQ_GT(end, rack->r_ctl.pcm_i.eseq)) { @@ -305,7 +305,7 @@ skip_ack_accounting: 0, &log, false, NULL, NULL, 0, &tv); } } - /* + /* * Here we need a lot to be added including: * 1) Some form of measurement, where if we think the measurement * is valid we iterate over the PCM data and come up with a path diff --git a/sys/netinet/tcp_stacks/sack_filter.c b/sys/netinet/tcp_stacks/sack_filter.c index fc9ee8454a1e..2b70548f3cc6 100644 --- a/sys/netinet/tcp_stacks/sack_filter.c +++ b/sys/netinet/tcp_stacks/sack_filter.c @@ -400,7 +400,7 @@ sack_filter_run(struct sack_filter *sf, struct sackblk *in, int numblks, tcp_seq break; } /* Copy it out to the outbound */ - memcpy(&in[at], &blkboard[i], sizeof(struct sackblk)); + memcpy(&in[at], &blkboard[i], sizeof(struct sackblk)); at++; room--; /* now lets add it to our sack-board */ @@ -588,7 +588,7 @@ sack_filter_blks(struct tcpcb *tp, struct sack_filter *sf, struct sackblk *in, i sf->sf_ack = th_ack; for(i=0, sf->sf_cur=0; i<numblks; i++) { - if ((in[i].end != tp->snd_max) && + if ((in[i].end != tp->snd_max) && ((in[i].end - in[i].start) < segmax)) { /* * We do not accept blocks less than a MSS minus all @@ -707,7 +707,7 @@ main(int argc, char **argv) out = stdout; memset(&tp, 0, sizeof(tp)); tp.t_maxseg = 1460; - + while ((i = getopt(argc, argv, "dIi:o:?hS:")) != -1) { switch (i) { case 'S': @@ -883,7 +883,7 @@ main(int argc, char **argv) } else { printf("can't open sack_setup.bin -- sorry no load\n"); } - + } else if (strncmp(buffer, "help", 4) == 0) { help: fprintf(out, "You can input:\n"); diff --git a/sys/netinet/tcp_stacks/sack_filter.h b/sys/netinet/tcp_stacks/sack_filter.h index b12fcf84567c..a1c0684a4359 100644 --- a/sys/netinet/tcp_stacks/sack_filter.h +++ b/sys/netinet/tcp_stacks/sack_filter.h @@ -42,7 +42,7 @@ * previously processed sack information. * * The second thing that the sack filter does is help protect against malicious - * attackers that are trying to attack any linked lists (or other data structures) + * attackers that are trying to attack any linked lists (or other data structures) * that are used in sack processing. Consider an attacker sending in sacks for * every other byte of data outstanding. This could in theory drastically split * up any scoreboard you are maintaining and make you search through a very large diff --git a/sys/netinet/tcp_subr.c b/sys/netinet/tcp_subr.c index cd42a67294a6..26e7e53d540c 100644 --- a/sys/netinet/tcp_subr.c +++ b/sys/netinet/tcp_subr.c @@ -2720,9 +2720,15 @@ tcp_ktlslist_locked(SYSCTL_HANDLER_ARGS, bool export_keys) ksr->snd_tag->sw->snd_tag_status_str != NULL) { sz = SND_TAG_STATUS_MAXLEN; - ksr->snd_tag->sw->snd_tag_status_str( + in_pcbref(inp); + INP_RUNLOCK(inp); + error = ksr->snd_tag->sw-> + snd_tag_status_str( ksr->snd_tag, NULL, &sz); - len += sz; + if (in_pcbrele_rlock(inp)) + return (EDEADLK); + if (error == 0) + len += sz; } } kss = so->so_snd.sb_tls_info; @@ -2739,9 +2745,15 @@ tcp_ktlslist_locked(SYSCTL_HANDLER_ARGS, bool export_keys) kss->snd_tag->sw->snd_tag_status_str != NULL) { sz = SND_TAG_STATUS_MAXLEN; - kss->snd_tag->sw->snd_tag_status_str( + in_pcbref(inp); + INP_RUNLOCK(inp); + error = kss->snd_tag->sw-> + snd_tag_status_str( kss->snd_tag, NULL, &sz); - len += sz; + if (in_pcbrele_rlock(inp)) + return (EDEADLK); + if (error == 0) + len += sz; } } if (p) { @@ -2811,9 +2823,16 @@ tcp_ktlslist_locked(SYSCTL_HANDLER_ARGS, bool export_keys) if (ksr->snd_tag != NULL && ksr->snd_tag->sw->snd_tag_status_str != NULL) { sz = SND_TAG_STATUS_MAXLEN; - ksr->snd_tag->sw->snd_tag_status_str( + in_pcbref(inp); + INP_RUNLOCK(inp); + error = ksr->snd_tag->sw->snd_tag_status_str( ksr->snd_tag, buf + len, &sz); - len += sz; + if (in_pcbrele_rlock(inp)) + return (EDEADLK); + if (error == 0) { + xktls->rcv.drv_st_len = sz; + len += sz; + } } } if (kss != NULL && kss->gen == xig.xig_gen) { @@ -2828,9 +2847,16 @@ tcp_ktlslist_locked(SYSCTL_HANDLER_ARGS, bool export_keys) if (kss->snd_tag != NULL && kss->snd_tag->sw->snd_tag_status_str != NULL) { sz = SND_TAG_STATUS_MAXLEN; - kss->snd_tag->sw->snd_tag_status_str( + in_pcbref(inp); + INP_RUNLOCK(inp); + error = kss->snd_tag->sw->snd_tag_status_str( kss->snd_tag, buf + len, &sz); - len += sz; + if (in_pcbrele_rlock(inp)) + return (EDEADLK); + if (error == 0) { + xktls->snd.drv_st_len = sz; + len += sz; + } } } len = roundup2(len, __alignof(*xktls)); @@ -2858,12 +2884,23 @@ tcp_ktlslist_locked(SYSCTL_HANDLER_ARGS, bool export_keys) static int tcp_ktlslist1(SYSCTL_HANDLER_ARGS, bool export_keys) { - int res; - - sx_xlock(&ktlslist_lock); - res = tcp_ktlslist_locked(oidp, arg1, arg2, req, export_keys); - sx_xunlock(&ktlslist_lock); - return (res); + int repeats, error; + + for (repeats = 0; repeats < 100; repeats++) { + if (sx_xlock_sig(&ktlslist_lock)) + return (EINTR); + error = tcp_ktlslist_locked(oidp, arg1, arg2, req, + export_keys); + sx_xunlock(&ktlslist_lock); + if (error != EDEADLK) + break; + if (sig_intr() != 0) { + error = EINTR; + break; + } + req->oldidx = 0; + } + return (error); } static int @@ -4500,7 +4537,7 @@ tcp_change_time_units(struct tcpcb *tp, int granularity) panic("Unknown granularity:%d tp:%p", granularity, tp); } -#endif +#endif } void diff --git a/sys/netinet/tcp_timer.c b/sys/netinet/tcp_timer.c index 32ce3001929c..3b9fe7a317b0 100644 --- a/sys/netinet/tcp_timer.c +++ b/sys/netinet/tcp_timer.c @@ -757,8 +757,8 @@ tcp_timer_rexmt(struct tcpcb *tp) tp->t_maxseg = tp->t_pmtud_saved_maxseg; if (tp->t_maxseg < V_tcp_mssdflt) { /* - * The MSS is so small we should not - * process incoming SACK's since we are + * The MSS is so small we should not + * process incoming SACK's since we are * subject to attack in such a case. */ tp->t_flags2 |= TF2_PROC_SACK_PROHIBIT; diff --git a/sys/netinet/tcp_usrreq.c b/sys/netinet/tcp_usrreq.c index 687b0d538666..98c934955121 100644 --- a/sys/netinet/tcp_usrreq.c +++ b/sys/netinet/tcp_usrreq.c @@ -164,7 +164,7 @@ tcp_usr_attach(struct socket *so, int proto, struct thread *td) goto out; so->so_rcv.sb_flags |= SB_AUTOSIZE; - so->so_snd.sb_flags |= SB_AUTOSIZE; + so->so_snd.sb_flags |= (SB_AUTOLOWAT | SB_AUTOSIZE); error = in_pcballoc(so, &V_tcbinfo); if (error) goto out; @@ -1768,9 +1768,9 @@ tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt) /* * Release the ref count the lookup * acquired. - */ + */ refcount_release(&blk->tfb_refcnt); - /* + /* * Now there is a chance that the * init() function mucked with some * things before it failed, such as @@ -1800,7 +1800,7 @@ tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt) * new one already. */ refcount_release(&tp->t_fb->tfb_refcnt); - /* + /* * Set in the new stack. */ tp->t_fb = blk; @@ -1934,7 +1934,7 @@ tcp_set_cc_mod(struct inpcb *inp, struct sockopt *sopt) CC_LIST_RUNLOCK(); return(ESRCH); } - /* + /* * With a reference the algorithm cannot be removed * so we hold a reference through the change process. */ diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h index 059b2aff689d..b90f65e83cb1 100644 --- a/sys/netinet/tcp_var.h +++ b/sys/netinet/tcp_var.h @@ -182,7 +182,7 @@ struct tcp_sendfile_track { * snd_una). When the response comes back indicating * that there was data (return value 1), then the caller * can build a sendmap entry based on the range and the - * times. The next query would then be done at the + * times. The next query would then be done at the * newly created sendmap_end. Repeated until sendmap_end == snd_max. * * Flags in sendmap_flags are defined below as well. @@ -197,7 +197,7 @@ struct tcp_sendfile_track { * The rack_times are a misc collection of information that * the old stack might possibly fill in. Of course its possible * that an old stack may not have a piece of information. If so - * then setting that value to zero is advised. Setting any + * then setting that value to zero is advised. Setting any * timestamp passed should only place a zero in it when it * is unfilled. This may mean that a time is off by a micro-second * but this is ok in the grand scheme of things. @@ -205,13 +205,13 @@ struct tcp_sendfile_track { * When switching stacks it is desireable to get as much information * from the old stack to the new stack as possible. Though not always * will the stack be compatible in the types of information. The - * init() function needs to take care when it begins changing + * init() function needs to take care when it begins changing * things such as inp_flags2 and the timer units to position these * changes at a point where it is unlikely they will fail after * making such changes. A stack optionally can have an "undo" - * function + * function * - * To transfer information to the old stack from the new in + * To transfer information to the old stack from the new in * respect to LRO and the inp_flags2, the new stack should set * the inp_flags2 to what it supports. The old stack in its * fini() function should call the tcp_handle_orphaned_packets() @@ -544,13 +544,13 @@ typedef enum { * do is: * a) Make sure that the inp_flags2 is setup correctly * for LRO. There are two flags that the previous - * stack may have set INP_MBUF_ACKCMP and + * stack may have set INP_MBUF_ACKCMP and * INP_SUPPORTS_MBUFQ. If the new stack does not * support these it *should* clear the flags. * b) Make sure that the timers are in the proper * granularity that the stack wants. The stack * should check the t_tmr_granularity field. Currently - * there are two values that it may hold + * there are two values that it may hold * TCP_TMR_GRANULARITY_TICKS and TCP_TMR_GRANULARITY_USEC. * Use the functions tcp_timer_convert(tp, granularity); * to move the timers to the correct format for your stack. @@ -558,14 +558,14 @@ typedef enum { * The new stack may also optionally query the tfb_chg_query * function if the old stack has one. The new stack may ask * for one of three entries and can also state to the old - * stack its support for the INP_MBUF_ACKCMP and + * stack its support for the INP_MBUF_ACKCMP and * INP_SUPPORTS_MBUFQ. This is important since if there are * queued ack's without that statement the old stack will * be forced to discard the queued acks. The requests that * can be made for information by the new stacks are: * * Note also that the tfb_tcp_fb_init() when called can - * determine if a query is needed by looking at the + * determine if a query is needed by looking at the * value passed in the ptr. The ptr is designed to be * set in with any allocated memory, but the address * of the condtion (ptr == &tp->t_fb_ptr) will be @@ -573,17 +573,17 @@ typedef enum { * setup of a tcb (which means no query would be needed). * If, however, the value is not t_fb_ptr, then the caller * is in the middle of a stack switch and is the new stack. - * A query would be appropriate (if the new stack support + * A query would be appropriate (if the new stack support * the query mechanism). * * TCP_QUERY_SENDMAP - Query of outstanding data. * TCP_QUERY_TIMERS_UP - Query about running timers. - * TCP_SUPPORTED_LRO - Declaration in req_param of - * the inp_flags2 supported by + * TCP_SUPPORTED_LRO - Declaration in req_param of + * the inp_flags2 supported by * the new stack. * TCP_QUERY_RACK_TIMES - Enquire about various timestamps * and states the old stack may be in. - * + * * tfb_tcp_fb_fini is changed to add a flag to tell * the old stack if the tcb is being destroyed or * not. A one in the flag means the TCB is being diff --git a/sys/netinet/udp_usrreq.c b/sys/netinet/udp_usrreq.c index dafbaf6dc672..42cfb919e263 100644 --- a/sys/netinet/udp_usrreq.c +++ b/sys/netinet/udp_usrreq.c @@ -243,7 +243,6 @@ udp_append(struct inpcb *inp, struct ip *ip, struct mbuf *n, int off, struct sockaddr_in6 udp_in6; #endif struct udpcb *up; - bool filtered; INP_LOCK_ASSERT(inp); @@ -252,13 +251,19 @@ udp_append(struct inpcb *inp, struct ip *ip, struct mbuf *n, int off, */ up = intoudpcb(inp); if (up->u_tun_func != NULL) { + bool filtered; + in_pcbref(inp); INP_RUNLOCK(inp); filtered = (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&udp_in[0], up->u_tun_ctx); INP_RLOCK(inp); - if (filtered) - return (in_pcbrele_rlocked(inp)); + if (in_pcbrele_rlocked(inp)) + return (1); + if (filtered) { + INP_RUNLOCK(inp); + return (1); + } } off += sizeof(struct udphdr); diff --git a/sys/netinet6/in6_gif.c b/sys/netinet6/in6_gif.c index d476829e8e3b..2bab1c57ce2a 100644 --- a/sys/netinet6/in6_gif.c +++ b/sys/netinet6/in6_gif.c @@ -194,6 +194,11 @@ in6_gif_setopts(struct gif_softc *sc, u_int options) sc->gif_options = options; in6_gif_attach(sc); } + + if ((options & GIF_NOCLAMP) != + (sc->gif_options & GIF_NOCLAMP)) { + sc->gif_options = options; + } return (0); } @@ -289,6 +294,7 @@ in6_gif_output(struct ifnet *ifp, struct mbuf *m, int proto, uint8_t ecn) { struct gif_softc *sc = ifp->if_softc; struct ip6_hdr *ip6; + u_long mtu; /* prepend new IP header */ NET_EPOCH_ASSERT(); @@ -304,11 +310,15 @@ in6_gif_output(struct ifnet *ifp, struct mbuf *m, int proto, uint8_t ecn) ip6->ip6_nxt = proto; ip6->ip6_hlim = V_ip6_gif_hlim; /* - * force fragmentation to minimum MTU, to avoid path MTU discovery. - * it is too painful to ask for resend of inner packet, to achieve - * path MTU discovery for encapsulated packets. + * Enforce fragmentation to minimum MTU, even if the interface MTU + * is larger, to avoid path MTU discovery when NOCLAMP is not + * set (default). IPv6 does not allow fragmentation on intermediate + * router nodes, so it is too painful to ask for resend of inner + * packet, to achieve path MTU discovery for encapsulated packets. */ - return (ip6_output(m, 0, NULL, IPV6_MINMTU, 0, NULL, NULL)); + mtu = ((sc->gif_options & GIF_NOCLAMP) == 0) ? IPV6_MINMTU : 0; + + return (ip6_output(m, 0, NULL, mtu, 0, NULL, NULL)); } static int diff --git a/sys/netinet6/mld6.c b/sys/netinet6/mld6.c index 06fe9e8820c9..a825658bd9ee 100644 --- a/sys/netinet6/mld6.c +++ b/sys/netinet6/mld6.c @@ -234,17 +234,20 @@ static SYSCTL_NODE(_net_inet6_mld, OID_AUTO, ifinfo, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mld_ifinfo, "Per-interface MLDv2 state"); -static int mld_v1enable = 1; -SYSCTL_INT(_net_inet6_mld, OID_AUTO, v1enable, CTLFLAG_RWTUN, - &mld_v1enable, 0, "Enable fallback to MLDv1"); +VNET_DEFINE_STATIC(bool, mld_v1enable) = true; +#define V_mld_v1enable VNET(mld_v1enable) +SYSCTL_BOOL(_net_inet6_mld, OID_AUTO, v1enable, CTLFLAG_VNET | CTLFLAG_RWTUN, + &VNET_NAME(mld_v1enable), 0, "Enable fallback to MLDv1"); -static int mld_v2enable = 1; -SYSCTL_INT(_net_inet6_mld, OID_AUTO, v2enable, CTLFLAG_RWTUN, - &mld_v2enable, 0, "Enable MLDv2"); +VNET_DEFINE_STATIC(bool, mld_v2enable) = true; +#define V_mld_v2enable VNET(mld_v2enable) +SYSCTL_BOOL(_net_inet6_mld, OID_AUTO, v2enable, CTLFLAG_VNET | CTLFLAG_RWTUN, + &VNET_NAME(mld_v2enable), 0, "Enable MLDv2"); -static int mld_use_allow = 1; -SYSCTL_INT(_net_inet6_mld, OID_AUTO, use_allow, CTLFLAG_RWTUN, - &mld_use_allow, 0, "Use ALLOW/BLOCK for RFC 4604 SSM joins/leaves"); +VNET_DEFINE_STATIC(bool, mld_use_allow) = true; +#define V_mld_use_allow VNET(mld_use_allow) +SYSCTL_BOOL(_net_inet6_mld, OID_AUTO, use_allow, CTLFLAG_VNET | CTLFLAG_RWTUN, + &VNET_NAME(mld_use_allow), 0, "Use ALLOW/BLOCK for RFC 4604 SSM joins/leaves"); /* * Packed Router Alert option structure declaration. @@ -481,7 +484,7 @@ mld_domifattach(struct ifnet *ifp) mbufq_init(&mli->mli_gq, MLD_MAX_RESPONSE_PACKETS); if ((ifp->if_flags & IFF_MULTICAST) == 0) mli->mli_flags |= MLIF_SILENT; - if (mld_use_allow) + if (V_mld_use_allow) mli->mli_flags |= MLIF_USEALLOW; MLD_LOCK(); @@ -614,7 +617,7 @@ mld_v1_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6, is_general_query = 0; - if (!mld_v1enable) { + if (!V_mld_v1enable) { CTR3(KTR_MLD, "ignore v1 query %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp)); @@ -790,7 +793,7 @@ mld_v2_input_query(struct ifnet *ifp, const struct ip6_hdr *ip6, NET_EPOCH_ASSERT(); - if (!mld_v2enable) { + if (!V_mld_v2enable) { CTR3(KTR_MLD, "ignore v2 query src %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &ip6->ip6_src), ifp, if_name(ifp)); @@ -1076,7 +1079,7 @@ mld_v1_input_report(struct ifnet *ifp, const struct ip6_hdr *ip6, NET_EPOCH_ASSERT(); - if (!mld_v1enable) { + if (!V_mld_v1enable) { CTR3(KTR_MLD, "ignore v1 report %s on ifp %p(%s)", ip6_sprintf(ip6tbuf, &mld->mld_addr), ifp, if_name(ifp)); diff --git a/sys/netinet6/raw_ip6.c b/sys/netinet6/raw_ip6.c index 0379ef7c789a..c90a1213bd66 100644 --- a/sys/netinet6/raw_ip6.c +++ b/sys/netinet6/raw_ip6.c @@ -765,8 +765,7 @@ rip6_bind(struct socket *so, struct sockaddr *nam, struct thread *td) } if (ifa != NULL && ((struct in6_ifaddr *)ifa)->ia6_flags & - (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY| - IN6_IFF_DETACHED|IN6_IFF_DEPRECATED)) { + (IN6_IFF_NOTREADY|IN6_IFF_DETACHED|IN6_IFF_DEPRECATED)) { NET_EPOCH_EXIT(et); return (EADDRNOTAVAIL); } diff --git a/sys/netinet6/scope6.c b/sys/netinet6/scope6.c index 0987ea7e99ad..08702a2e81ab 100644 --- a/sys/netinet6/scope6.c +++ b/sys/netinet6/scope6.c @@ -505,8 +505,23 @@ in6_set_unicast_scopeid(struct in6_addr *in6, uint32_t scopeid) struct ifnet* in6_getlinkifnet(uint32_t zoneid) { + struct ifnet *ifp; - return (ifnet_byindex((u_short)zoneid)); + ifp = ifnet_byindex((u_short)zoneid); + + if (ifp == NULL) + return (NULL); + + /* An interface might not be IPv6 capable. */ + if (ifp->if_afdata[AF_INET6] == NULL) { + log(LOG_NOTICE, + "%s: embedded scope points to an interface without " + "IPv6: %s%%%d.\n", __func__, + if_name(ifp), zoneid); + return (NULL); + } + + return (ifp); } /* diff --git a/sys/netinet6/udp6_usrreq.c b/sys/netinet6/udp6_usrreq.c index 304effa26e01..b3ed16fda713 100644 --- a/sys/netinet6/udp6_usrreq.c +++ b/sys/netinet6/udp6_usrreq.c @@ -142,7 +142,6 @@ udp6_append(struct inpcb *inp, struct mbuf *n, int off, struct socket *so; struct mbuf *opts = NULL, *tmp_opts; struct udpcb *up; - bool filtered; INP_LOCK_ASSERT(inp); @@ -151,13 +150,19 @@ udp6_append(struct inpcb *inp, struct mbuf *n, int off, */ up = intoudpcb(inp); if (up->u_tun_func != NULL) { + bool filtered; + in_pcbref(inp); INP_RUNLOCK(inp); filtered = (*up->u_tun_func)(n, off, inp, (struct sockaddr *)&fromsa[0], up->u_tun_ctx); INP_RLOCK(inp); - if (filtered) - return (in_pcbrele_rlocked(inp)); + if (in_pcbrele_rlocked(inp)) + return (1); + if (filtered) { + INP_RUNLOCK(inp); + return (1); + } } off += sizeof(struct udphdr); diff --git a/sys/netipsec/ipsec.c b/sys/netipsec/ipsec.c index 6bacc68b7441..92d0201b398a 100644 --- a/sys/netipsec/ipsec.c +++ b/sys/netipsec/ipsec.c @@ -636,8 +636,10 @@ ipsec4_in_reject1(const struct mbuf *m, struct ip *ip1, struct inpcb *inp) #ifdef IPSEC_OFFLOAD tag = ipsec_accel_input_tag_lookup(m); - if (tag != NULL) - return (0); + if (tag != NULL) { + tag->tag.m_tag_id = PACKET_TAG_IPSEC_IN_DONE; + __DECONST(struct mbuf *, m)->m_flags |= M_DECRYPTED; + } #endif if (ip1 == NULL) { diff --git a/sys/netipsec/ipsec_offload.c b/sys/netipsec/ipsec_offload.c index 467d5ded1d7a..8a09d5f37b4a 100644 --- a/sys/netipsec/ipsec_offload.c +++ b/sys/netipsec/ipsec_offload.c @@ -94,6 +94,7 @@ struct ifp_handle_sav { size_t hdr_ext_size; uint64_t cnt_octets; uint64_t cnt_allocs; + struct xform_history xfh; }; #define IFP_HS_HANDLED 0x00000001 @@ -159,6 +160,8 @@ static void ipsec_accel_drv_sa_lifetime_update_impl(struct secasvar *sav, static int ipsec_accel_drv_sa_lifetime_fetch_impl(struct secasvar *sav, if_t ifp, u_int drv_spi, uint64_t *octets, uint64_t *allocs); static void ipsec_accel_ifdetach_event(void *arg, struct ifnet *ifp); +static bool ipsec_accel_fill_xh_impl(if_t ifp, uint32_t drv_spi, + struct xform_history *xh); static void ipsec_accel_init(void *arg) @@ -185,6 +188,7 @@ ipsec_accel_init(void *arg) ipsec_accel_drv_sa_lifetime_update_impl; ipsec_accel_drv_sa_lifetime_fetch_p = ipsec_accel_drv_sa_lifetime_fetch_impl; + ipsec_accel_fill_xh_p = ipsec_accel_fill_xh_impl; pctrie_init(&drv_spi_pctrie); ipsec_accel_ifdetach_event_tag = EVENTHANDLER_REGISTER( ifnet_departure_event, ipsec_accel_ifdetach_event, NULL, @@ -209,6 +213,7 @@ ipsec_accel_fini(void *arg) ipsec_accel_on_ifdown_p = NULL; ipsec_accel_drv_sa_lifetime_update_p = NULL; ipsec_accel_drv_sa_lifetime_fetch_p = NULL; + ipsec_accel_fill_xh_p = NULL; ipsec_accel_sync_imp(); clean_unrhdr(drv_spi_unr); /* avoid panic, should go later */ clear_unrhdr(drv_spi_unr); @@ -412,6 +417,10 @@ ipsec_accel_handle_sav(struct secasvar *sav, struct ifnet *ifp, ihs->ifdata = priv; ihs->flags = flags; ihs->hdr_ext_size = esp_hdrsiz(sav); + memcpy(&ihs->xfh.dst, &sav->sah->saidx.dst, sizeof(ihs->xfh.dst)); + ihs->xfh.spi = sav->spi; + ihs->xfh.proto = sav->sah->saidx.proto; + ihs->xfh.mode = sav->sah->saidx.mode; mtx_lock(&ipsec_accel_sav_tmp); CK_LIST_FOREACH(i, &sav->accel_ifps, sav_link) { if (i->ifp == ifp) { @@ -1162,4 +1171,20 @@ ipsec_accel_key_setaccelif_impl(struct secasvar *sav) return (m); } +static bool +ipsec_accel_fill_xh_impl(if_t ifp, uint32_t drv_spi, struct xform_history *xh) +{ + struct ifp_handle_sav *i; + + if (drv_spi < IPSEC_ACCEL_DRV_SPI_MIN || + drv_spi > IPSEC_ACCEL_DRV_SPI_MAX) + return (false); + + i = DRVSPI_SA_PCTRIE_LOOKUP(&drv_spi_pctrie, drv_spi); + if (i == NULL) + return (false); + memcpy(xh, &i->xfh, sizeof(*xh)); + return (true); +} + #endif /* IPSEC_OFFLOAD */ diff --git a/sys/netipsec/ipsec_offload.h b/sys/netipsec/ipsec_offload.h index 904fe6252396..ae60eaa8ae78 100644 --- a/sys/netipsec/ipsec_offload.h +++ b/sys/netipsec/ipsec_offload.h @@ -30,6 +30,7 @@ #include <sys/errno.h> #include <net/if.h> #include <net/if_var.h> +#include <netipsec/xform.h> struct secpolicy; struct secasvar; @@ -42,6 +43,7 @@ struct ipsec_accel_out_tag { struct ipsec_accel_in_tag { struct m_tag tag; + struct xform_history xh; /* Must be first to mimic IPSEC_IN_DONE */ uint16_t drv_spi; }; @@ -66,6 +68,8 @@ extern void (*ipsec_accel_drv_sa_lifetime_update_p)(struct secasvar *sav, if_t ifp, u_int drv_spi, uint64_t octets, uint64_t allocs); extern int (*ipsec_accel_drv_sa_lifetime_fetch_p)(struct secasvar *sav, if_t ifp, u_int drv_spi, uint64_t *octets, uint64_t *allocs); +extern bool (*ipsec_accel_fill_xh_p)(if_t ifp, uint32_t drv_spi, + struct xform_history *xh); #ifdef IPSEC_OFFLOAD /* @@ -158,6 +162,16 @@ ipsec_accel_key_setaccelif(struct secasvar *sav) return (NULL); } +static inline bool +ipsec_accel_fill_xh(if_t ifp, uint32_t drv_spi, struct xform_history *xh) +{ + bool (*p)(if_t ifp, uint32_t drv_spi, struct xform_history *xh); + + p = atomic_load_ptr(&ipsec_accel_fill_xh_p); + if (p != NULL) + return (p(ifp, drv_spi, xh)); + return (false); +} #else #define ipsec_accel_sa_newkey(a) @@ -168,6 +182,7 @@ ipsec_accel_key_setaccelif(struct secasvar *sav) #define ipsec_accel_sync() #define ipsec_accel_is_accel_sav(a) #define ipsec_accel_key_setaccelif(a) +#define ipsec_accel_fill_xh(a, b, c) (false) #endif void ipsec_accel_forget_sav_impl(struct secasvar *sav); @@ -180,6 +195,7 @@ bool ipsec_accel_output(struct ifnet *ifp, struct mbuf *m, struct inpcb *inp, struct secpolicy *sp, struct secasvar *sav, int af, int mtu, int *hwassist); void ipsec_accel_forget_sav(struct secasvar *sav); +struct xform_history; #else #define ipsec_accel_input(a, b, c) (ENXIO) #define ipsec_accel_output(a, b, c, d, e, f, g, h) ({ \ diff --git a/sys/netipsec/key.c b/sys/netipsec/key.c index ae67d83c6d13..4ba1b49c24f0 100644 --- a/sys/netipsec/key.c +++ b/sys/netipsec/key.c @@ -114,6 +114,8 @@ void (*ipsec_accel_drv_sa_lifetime_update_p)(struct secasvar *sav, if_t ifp, u_int drv_spi, uint64_t octets, uint64_t allocs); int (*ipsec_accel_drv_sa_lifetime_fetch_p)(struct secasvar *sav, if_t ifp, u_int drv_spi, uint64_t *octets, uint64_t *allocs); +bool (*ipsec_accel_fill_xh_p)(if_t ifp, uint32_t drv_spi, + struct xform_history *xh); #endif #define FULLMASK 0xff diff --git a/sys/netlink/netlink_message_parser.h b/sys/netlink/netlink_message_parser.h index 8492ecb3021b..720317ed74f3 100644 --- a/sys/netlink/netlink_message_parser.h +++ b/sys/netlink/netlink_message_parser.h @@ -209,7 +209,8 @@ int nlattr_get_nested(struct nlattr *nla, struct nl_pstate *npt, int nlattr_get_nested_ptr(struct nlattr *nla, struct nl_pstate *npt, const void *arg, void *target); -bool nlmsg_report_err_msg(struct nl_pstate *npt, const char *fmt, ...); +bool nlmsg_report_err_msg(struct nl_pstate *npt, const char *fmt, ...) + __printflike(2, 3); #define NLMSG_REPORT_ERR_MSG(_npt, _fmt, ...) { \ nlmsg_report_err_msg(_npt, _fmt, ## __VA_ARGS__); \ diff --git a/sys/netpfil/ipfilter/netinet/ip_fil_freebsd.c b/sys/netpfil/ipfilter/netinet/ip_fil_freebsd.c index 04850549db98..6eb6cf2a7a47 100644 --- a/sys/netpfil/ipfilter/netinet/ip_fil_freebsd.c +++ b/sys/netpfil/ipfilter/netinet/ip_fil_freebsd.c @@ -463,13 +463,14 @@ ipf_send_ip(fr_info_t *fin, mb_t *m) int ipf_send_icmp_err(int type, fr_info_t *fin, int dst) { - int err, hlen, xtra, iclen, ohlen, avail, code; + int err, hlen, xtra, iclen, ohlen, avail; struct in_addr dst4; struct icmp *icmp; struct mbuf *m; i6addr_t dst6; void *ifp; #ifdef USE_INET6 + int code; ip6_t *ip6; #endif ip_t *ip, *ip2; @@ -477,8 +478,8 @@ ipf_send_icmp_err(int type, fr_info_t *fin, int dst) if ((type < 0) || (type >= ICMP_MAXTYPE)) return (-1); - code = fin->fin_icode; #ifdef USE_INET6 + code = fin->fin_icode; /* See NetBSD ip_fil_netbsd.c r1.4: */ if ((code < 0) || (code >= sizeof(icmptoicmp6unreach)/sizeof(int))) return (-1); diff --git a/sys/netpfil/ipfw/ip_fw2.c b/sys/netpfil/ipfw/ip_fw2.c index 923633d76df7..c129c8c49921 100644 --- a/sys/netpfil/ipfw/ip_fw2.c +++ b/sys/netpfil/ipfw/ip_fw2.c @@ -196,7 +196,7 @@ SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, "Firewall"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, - "Only do a single pass through ipfw when using dummynet(4)"); + "Only do a single pass through ipfw when using dummynet(4), ipfw_nat or other divert(4)-like interfaces"); SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, "Rule number auto-increment step"); diff --git a/sys/netpfil/pf/if_pflog.c b/sys/netpfil/pf/if_pflog.c index 6a87ea2471cb..cb96d2fcc44c 100644 --- a/sys/netpfil/pf/if_pflog.c +++ b/sys/netpfil/pf/if_pflog.c @@ -284,12 +284,12 @@ pflog_packet(uint8_t action, u_int8_t reason, * state lock, since this leads to unsafe LOR. * These conditions are very very rare, however. */ - if (trigger->log & PF_LOG_SOCKET_LOOKUP && !pd->lookup.done && lookupsafe) + if (trigger->log & PF_LOG_USER && !pd->lookup.done && lookupsafe) pd->lookup.done = pf_socket_lookup(pd); - if (pd->lookup.done > 0) + if (trigger->log & PF_LOG_USER && pd->lookup.done > 0) hdr.uid = pd->lookup.uid; else - hdr.uid = UID_MAX; + hdr.uid = -1; hdr.pid = NO_PID; hdr.rule_uid = rm->cuid; hdr.rule_pid = rm->cpid; diff --git a/sys/netpfil/pf/if_pfsync.c b/sys/netpfil/pf/if_pfsync.c index fdedb9424117..ee10a997c977 100644 --- a/sys/netpfil/pf/if_pfsync.c +++ b/sys/netpfil/pf/if_pfsync.c @@ -110,8 +110,6 @@ #include <netpfil/pf/pfsync_nv.h> -#define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x - struct pfsync_bucket; struct pfsync_softc; @@ -532,6 +530,7 @@ pfsync_state_import(union pfsync_state_union *sp, int flags, int msg_version) struct pf_kpooladdr *rpool_first; int error; uint8_t rt = 0; + int n = 0; PF_RULES_RASSERT(); @@ -557,10 +556,12 @@ pfsync_state_import(union pfsync_state_union *sp, int flags, int msg_version) */ if (sp->pfs_1301.rule != htonl(-1) && sp->pfs_1301.anchor == htonl(-1) && (flags & (PFSYNC_SI_IOCTL | PFSYNC_SI_CKSUM)) && ntohl(sp->pfs_1301.rule) < - pf_main_ruleset.rules[PF_RULESET_FILTER].active.rcount) - r = pf_main_ruleset.rules[ - PF_RULESET_FILTER].active.ptr_array[ntohl(sp->pfs_1301.rule)]; - else + pf_main_ruleset.rules[PF_RULESET_FILTER].active.rcount) { + TAILQ_FOREACH(r, pf_main_ruleset.rules[ + PF_RULESET_FILTER].active.ptr, entries) + if (ntohl(sp->pfs_1301.rule) == n++) + break; + } else r = &V_pf_default_rule; /* @@ -594,9 +595,9 @@ pfsync_state_import(union pfsync_state_union *sp, int flags, int msg_version) if ((rpool_first == NULL) || (TAILQ_NEXT(rpool_first, entries) != NULL)) { DPFPRINTF(PF_DEBUG_MISC, - ("%s: can't recover routing information " - "because of empty or bad redirection pool\n", - __func__)); + "%s: can't recover routing information " + "because of empty or bad redirection pool", + __func__); return ((flags & PFSYNC_SI_IOCTL) ? EINVAL : 0); } rt = r->rt; @@ -607,8 +608,8 @@ pfsync_state_import(union pfsync_state_union *sp, int flags, int msg_version) * give up on recovering. */ DPFPRINTF(PF_DEBUG_MISC, - ("%s: can't recover routing information " - "because of different ruleset\n", __func__)); + "%s: can't recover routing information " + "because of different ruleset", __func__); return ((flags & PFSYNC_SI_IOCTL) ? EINVAL : 0); } break; @@ -621,8 +622,8 @@ pfsync_state_import(union pfsync_state_union *sp, int flags, int msg_version) rt_kif = pfi_kkif_find(sp->pfs_1400.rt_ifname); if (rt_kif == NULL) { DPFPRINTF(PF_DEBUG_MISC, - ("%s: unknown route interface: %s\n", - __func__, sp->pfs_1400.rt_ifname)); + "%s: unknown route interface: %s", + __func__, sp->pfs_1400.rt_ifname); return ((flags & PFSYNC_SI_IOCTL) ? EINVAL : 0); } rt = sp->pfs_1400.rt; @@ -763,6 +764,10 @@ pfsync_state_import(union pfsync_state_union *sp, int flags, int msg_version) __func__, msg_version); } + if (! (st->act.rtableid == -1 || + (st->act.rtableid >= 0 && st->act.rtableid < rt_numfibs))) + goto cleanup; + st->id = sp->pfs_1301.id; st->creatorid = sp->pfs_1301.creatorid; pf_state_peer_ntoh(&sp->pfs_1301.src, &st->src); @@ -1083,7 +1088,7 @@ pfsync_in_ins(struct mbuf *m, int offset, int count, int flags, int action) msg_version = PFSYNC_MSG_VERSION_1400; break; default: - V_pfsyncstats.pfsyncs_badact++; + V_pfsyncstats.pfsyncs_badver++; return (-1); } @@ -1110,9 +1115,8 @@ pfsync_in_ins(struct mbuf *m, int offset, int count, int flags, int action) continue; } - if (pfsync_state_import(sp, flags, msg_version) == ENOMEM) - /* Drop out, but process the rest of the actions. */ - break; + if (pfsync_state_import(sp, flags, msg_version) != 0) + V_pfsyncstats.pfsyncs_badact++; } return (total_len); diff --git a/sys/netpfil/pf/pf.c b/sys/netpfil/pf/pf.c index 127b29320acb..79c298c18b46 100644 --- a/sys/netpfil/pf/pf.c +++ b/sys/netpfil/pf/pf.c @@ -119,8 +119,6 @@ #include <machine/in_cksum.h> #include <security/mac/mac_framework.h> -#define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x - SDT_PROVIDER_DEFINE(pf); SDT_PROBE_DEFINE2(pf, , test, reason_set, "int", "int"); SDT_PROBE_DEFINE4(pf, ip, test, done, "int", "int", "struct pf_krule *", @@ -161,6 +159,7 @@ SDT_PROBE_DEFINE2(pf, eth, test_rule, match, "int", "struct pf_keth_rule *"); SDT_PROBE_DEFINE2(pf, eth, test_rule, final_match, "int", "struct pf_keth_rule *"); SDT_PROBE_DEFINE2(pf, purge, state, rowcount, "int", "size_t"); +SDT_PROBE_DEFINE2(pf, , log, log, "int", "const char *"); /* * Global variables @@ -375,6 +374,8 @@ static u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t, int, u_int16_t); static int pf_check_proto_cksum(struct mbuf *, int, int, u_int8_t, sa_family_t); +static int pf_walk_option(struct pf_pdesc *, struct ip *, + int, int, u_short *); static int pf_walk_header(struct pf_pdesc *, struct ip *, u_short *); #ifdef INET6 static int pf_walk_option6(struct pf_pdesc *, struct ip6_hdr *, @@ -4579,7 +4580,7 @@ pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) static int pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) { - if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) + if (u == -1 && op != PF_OP_EQ && op != PF_OP_NE) return (0); return (pf_match(op, a1, a2, u)); } @@ -4587,7 +4588,7 @@ pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) static int pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) { - if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) + if (g == -1 && op != PF_OP_EQ && op != PF_OP_NE) return (0); return (pf_match(op, a1, a2, g)); } @@ -4615,8 +4616,8 @@ pf_match_rcvif(struct mbuf *m, struct pf_krule *r) if (kif == NULL) { DPFPRINTF(PF_DEBUG_URGENT, - ("%s: kif == NULL, @%d via %s\n", __func__, r->nr, - r->rcv_ifname)); + "%s: kif == NULL, @%d via %s", __func__, r->nr, + r->rcv_ifname); return (0); } @@ -4914,8 +4915,8 @@ pf_socket_lookup(struct pf_pdesc *pd) struct inpcbinfo *pi; struct inpcb *inp; - pd->lookup.uid = UID_MAX; - pd->lookup.gid = GID_MAX; + pd->lookup.uid = -1; + pd->lookup.gid = -1; switch (pd->proto) { case IPPROTO_TCP: @@ -4975,7 +4976,7 @@ pf_socket_lookup(struct pf_pdesc *pd) } INP_RLOCK_ASSERT(inp); pd->lookup.uid = inp->inp_cred->cr_uid; - pd->lookup.gid = inp->inp_cred->cr_groups[0]; + pd->lookup.gid = inp->inp_cred->cr_gid; INP_RUNLOCK(inp); return (1); @@ -5242,8 +5243,8 @@ pf_test_eth_rule(int dir, struct pfi_kkif *kif, struct mbuf **m0) if (__predict_false(m->m_len < sizeof(struct ether_header)) && (m = *m0 = m_pullup(*m0, sizeof(struct ether_header))) == NULL) { DPFPRINTF(PF_DEBUG_URGENT, - ("%s: m_len < sizeof(struct ether_header)" - ", pullup failed\n", __func__)); + "%s: m_len < sizeof(struct ether_header)" + ", pullup failed", __func__); return (PF_DROP); } e = mtod(m, struct ether_header *); @@ -5759,7 +5760,7 @@ pf_test_rule(struct pf_krule **rm, struct pf_kstate **sm, if (inp != NULL) { INP_LOCK_ASSERT(inp); pd->lookup.uid = inp->inp_cred->cr_uid; - pd->lookup.gid = inp->inp_cred->cr_groups[0]; + pd->lookup.gid = inp->inp_cred->cr_gid; pd->lookup.done = 1; } @@ -5901,18 +5902,17 @@ pf_test_rule(struct pf_krule **rm, struct pf_kstate **sm, M_SETFIB(pd->m, pd->act.rtableid); if (r->rt) { - struct pf_ksrc_node *sn = NULL; - struct pf_srchash *snh = NULL; /* * Set act.rt here instead of in pf_rule_to_actions() because * it is applied only from the last pass rule. */ pd->act.rt = r->rt; - /* Don't use REASON_SET, pf_map_addr increases the reason counters */ - ctx.reason = pf_map_addr_sn(pd->af, r, pd->src, &pd->act.rt_addr, - &pd->act.rt_kif, NULL, &sn, &snh, &(r->route), PF_SN_ROUTE); - if (ctx.reason != 0) + if ((transerror = pf_map_addr_sn(pd->af, r, pd->src, + &pd->act.rt_addr, &pd->act.rt_kif, NULL, &(r->route), + PF_SN_ROUTE)) != PFRES_MATCH) { + REASON_SET(&ctx.reason, transerror); goto cleanup; + } } if (pd->virtual_proto != PF_VPROTO_FRAGMENT && @@ -6056,9 +6056,16 @@ pf_create_state(struct pf_krule *r, struct pf_test_ctx *ctx, /* src node for translation rule */ if (ctx->nr != NULL) { KASSERT(ctx->nat_pool != NULL, ("%s: nat_pool is NULL", __func__)); + /* + * The NAT addresses are chosen during ruleset parsing. + * The new afto code stores post-nat addresses in nsaddr. + * The old nat code (also used for new nat-to rules) creates + * state keys and stores addresses in them. + */ if ((ctx->nat_pool->opts & PF_POOL_STICKYADDR) && (sn_reason = pf_insert_src_node(sns, snhs, ctx->nr, - &ctx->sk->addr[pd->sidx], pd->af, &ctx->nk->addr[1], NULL, + ctx->sk ? &(ctx->sk->addr[pd->sidx]) : pd->src, pd->af, + ctx->nk ? &(ctx->nk->addr[1]) : &(pd->nsaddr), NULL, PF_SN_NAT)) != 0 ) { REASON_SET(&ctx->reason, sn_reason); goto csfailed; @@ -6162,8 +6169,8 @@ pf_create_state(struct pf_krule *r, struct pf_test_ctx *ctx, &s->src, &s->dst, &ctx->rewrite)) { /* This really shouldn't happen!!! */ DPFPRINTF(PF_DEBUG_URGENT, - ("%s: tcp normalize failed on first " - "pkt\n", __func__)); + "%s: tcp normalize failed on first " + "pkt", __func__); goto csfailed; } } else if (pd->proto == IPPROTO_SCTP) { @@ -6213,7 +6220,7 @@ pf_create_state(struct pf_krule *r, struct pf_test_ctx *ctx, if (ctx->tag > 0) s->tag = ctx->tag; if (pd->proto == IPPROTO_TCP && (tcp_get_flags(th) & (TH_SYN|TH_ACK)) == - TH_SYN && r->keep_state == PF_STATE_SYNPROXY) { + TH_SYN && r->keep_state == PF_STATE_SYNPROXY && pd->dir == PF_IN) { pf_set_protostate(s, PF_PEER_SRC, PF_TCPS_PROXY_SRC); pf_undo_nat(ctx->nr, pd, bip_sum); s->src.seqhi = arc4random(); @@ -7392,7 +7399,7 @@ pf_sctp_multihome_delayed(struct pf_pdesc *pd, struct pfi_kkif *kif, { struct pf_sctp_multihome_job *j, *tmp; struct pf_sctp_source *i; - int ret __unused; + int ret; struct pf_kstate *sm = NULL; struct pf_krule *ra = NULL; struct pf_krule *r = &V_pf_default_rule; @@ -7959,8 +7966,8 @@ pf_test_state_icmp(struct pf_kstate **state, struct pf_pdesc *pd, if (!pf_pull_hdr(pd->m, ipoff2, &h2, sizeof(h2), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: ICMP error message too short " - "(ip)\n")); + "pf: ICMP error message too short " + "(ip)"); return (PF_DROP); } /* @@ -7990,8 +7997,8 @@ pf_test_state_icmp(struct pf_kstate **state, struct pf_pdesc *pd, if (!pf_pull_hdr(pd->m, ipoff2, &h2_6, sizeof(h2_6), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: ICMP error message too short " - "(ip6)\n")); + "pf: ICMP error message too short " + "(ip6)"); return (PF_DROP); } pd2.off = ipoff2; @@ -8043,8 +8050,8 @@ pf_test_state_icmp(struct pf_kstate **state, struct pf_pdesc *pd, if (!pf_pull_hdr(pd->m, pd2.off, th, 8, NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: ICMP error message too short " - "(tcp)\n")); + "pf: ICMP error message too short " + "(tcp)"); return (PF_DROP); } pd2.pcksum = &pd2.hdr.tcp.th_sum; @@ -8238,8 +8245,8 @@ pf_test_state_icmp(struct pf_kstate **state, struct pf_pdesc *pd, if (!pf_pull_hdr(pd->m, pd2.off, uh, sizeof(*uh), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: ICMP error message too short " - "(udp)\n")); + "pf: ICMP error message too short " + "(udp)"); return (PF_DROP); } pd2.pcksum = &pd2.hdr.udp.uh_sum; @@ -8370,8 +8377,8 @@ pf_test_state_icmp(struct pf_kstate **state, struct pf_pdesc *pd, if (! pf_pull_hdr(pd->m, pd2.off, sh, sizeof(*sh), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: ICMP error message too short " - "(sctp)\n")); + "pf: ICMP error message too short " + "(sctp)"); return (PF_DROP); } pd2.pcksum = &pd2.sctp_dummy_sum; @@ -8401,8 +8408,8 @@ pf_test_state_icmp(struct pf_kstate **state, struct pf_pdesc *pd, if (src->scrub->pfss_v_tag != sh->v_tag) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: ICMP error message has incorrect " - "SCTP v_tag\n")); + "pf: ICMP error message has incorrect " + "SCTP v_tag"); return (PF_DROP); } @@ -8525,8 +8532,8 @@ pf_test_state_icmp(struct pf_kstate **state, struct pf_pdesc *pd, if (!pf_pull_hdr(pd->m, pd2.off, iih, ICMP_MINLEN, NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: ICMP error message too short i" - "(icmp)\n")); + "pf: ICMP error message too short i" + "(icmp)"); return (PF_DROP); } pd2.pcksum = &pd2.hdr.icmp.icmp_cksum; @@ -8645,8 +8652,8 @@ pf_test_state_icmp(struct pf_kstate **state, struct pf_pdesc *pd, if (!pf_pull_hdr(pd->m, pd2.off, iih, sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: ICMP error message too short " - "(icmp6)\n")); + "pf: ICMP error message too short " + "(icmp6)"); return (PF_DROP); } pd2.pcksum = &pd2.hdr.icmp6.icmp6_cksum; @@ -9062,6 +9069,9 @@ pf_route(struct pf_krule *r, struct ifnet *oifp, goto bad; } + if (r->rt == PF_DUPTO) + skip_test = true; + if (pd->dir == PF_IN && !skip_test) { if (pf_test(AF_INET, PF_OUT, PFIL_FWD, ifp, &m0, inp, &pd->act) != PF_PASS) { @@ -9073,7 +9083,7 @@ pf_route(struct pf_krule *r, struct ifnet *oifp, } if (m0->m_len < sizeof(struct ip)) { DPFPRINTF(PF_DEBUG_URGENT, - ("%s: m0->m_len < sizeof(struct ip)\n", __func__)); + "%s: m0->m_len < sizeof(struct ip)", __func__); SDT_PROBE1(pf, ip, route_to, drop, __LINE__); goto bad; } @@ -9364,6 +9374,9 @@ pf_route6(struct pf_krule *r, struct ifnet *oifp, goto bad; } + if (r->rt == PF_DUPTO) + skip_test = true; + if (pd->dir == PF_IN && !skip_test) { if (pf_test(AF_INET6, PF_OUT, PFIL_FWD | PF_PFIL_NOREFRAGMENT, ifp, &m0, inp, &pd->act) != PF_PASS) { @@ -9375,8 +9388,8 @@ pf_route6(struct pf_krule *r, struct ifnet *oifp, } if (m0->m_len < sizeof(struct ip6_hdr)) { DPFPRINTF(PF_DEBUG_URGENT, - ("%s: m0->m_len < sizeof(struct ip6_hdr)\n", - __func__)); + "%s: m0->m_len < sizeof(struct ip6_hdr)", + __func__); SDT_PROBE1(pf, ip6, route_to, drop, __LINE__); goto bad; } @@ -9671,7 +9684,7 @@ pf_test_eth(int dir, int pflags, struct ifnet *ifp, struct mbuf **m0, if (kif == NULL) { DPFPRINTF(PF_DEBUG_URGENT, - ("%s: kif == NULL, if_xname %s\n", __func__, ifp->if_xname)); + "%s: kif == NULL, if_xname %s", __func__, ifp->if_xname); return (PF_DROP); } if (kif->pfik_flags & PFI_IFLAG_SKIP) @@ -9786,6 +9799,62 @@ pf_dummynet_route(struct pf_pdesc *pd, struct pf_kstate *s, } static int +pf_walk_option(struct pf_pdesc *pd, struct ip *h, int off, int end, + u_short *reason) +{ + uint8_t type, length, opts[15 * 4 - sizeof(struct ip)]; + + /* IP header in payload of ICMP packet may be too short */ + if (pd->m->m_pkthdr.len < end) { + DPFPRINTF(PF_DEBUG_MISC, "IP option too short"); + REASON_SET(reason, PFRES_SHORT); + return (PF_DROP); + } + + MPASS(end - off <= sizeof(opts)); + m_copydata(pd->m, off, end - off, opts); + end -= off; + off = 0; + + while (off < end) { + type = opts[off]; + if (type == IPOPT_EOL) + break; + if (type == IPOPT_NOP) { + off++; + continue; + } + if (off + 2 > end) { + DPFPRINTF(PF_DEBUG_MISC, "IP length opt"); + REASON_SET(reason, PFRES_IPOPTIONS); + return (PF_DROP); + } + length = opts[off + 1]; + if (length < 2) { + DPFPRINTF(PF_DEBUG_MISC, "IP short opt"); + REASON_SET(reason, PFRES_IPOPTIONS); + return (PF_DROP); + } + if (off + length > end) { + DPFPRINTF(PF_DEBUG_MISC, "IP long opt"); + REASON_SET(reason, PFRES_IPOPTIONS); + return (PF_DROP); + } + switch (type) { + case IPOPT_RA: + pd->badopts |= PF_OPT_ROUTER_ALERT; + break; + default: + pd->badopts |= PF_OPT_OTHER; + break; + } + off += length; + } + + return (PF_PASS); +} + +static int pf_walk_header(struct pf_pdesc *pd, struct ip *h, u_short *reason) { struct ah ext; @@ -9797,11 +9866,28 @@ pf_walk_header(struct pf_pdesc *pd, struct ip *h, u_short *reason) REASON_SET(reason, PFRES_SHORT); return (PF_DROP); } - if (hlen != sizeof(struct ip)) - pd->badopts++; + if (hlen != sizeof(struct ip)) { + if (pf_walk_option(pd, h, pd->off + sizeof(struct ip), + pd->off + hlen, reason) != PF_PASS) + return (PF_DROP); + /* header options which contain only padding is fishy */ + if (pd->badopts == 0) + pd->badopts |= PF_OPT_OTHER; + } end = pd->off + ntohs(h->ip_len); pd->off += hlen; pd->proto = h->ip_p; + /* IGMP packets have router alert options, allow them */ + if (pd->proto == IPPROTO_IGMP) { + /* According to RFC 1112 ttl must be set to 1. */ + if ((h->ip_ttl != 1) || + !IN_MULTICAST(ntohl(h->ip_dst.s_addr))) { + DPFPRINTF(PF_DEBUG_MISC, "Invalid IGMP"); + REASON_SET(reason, PFRES_IPOPTIONS); + return (PF_DROP); + } + pd->badopts &= ~PF_OPT_ROUTER_ALERT; + } /* stop walking over non initial fragments */ if ((h->ip_off & htons(IP_OFFMASK)) != 0) return (PF_PASS); @@ -9814,7 +9900,7 @@ pf_walk_header(struct pf_pdesc *pd, struct ip *h, u_short *reason) return (PF_PASS); if (!pf_pull_hdr(pd->m, pd->off, &ext, sizeof(ext), NULL, reason, AF_INET)) { - DPFPRINTF(PF_DEBUG_MISC, ("IP short exthdr")); + DPFPRINTF(PF_DEBUG_MISC, "IP short exthdr"); return (PF_DROP); } pd->off += (ext.ah_len + 2) * 4; @@ -9824,7 +9910,7 @@ pf_walk_header(struct pf_pdesc *pd, struct ip *h, u_short *reason) return (PF_PASS); } } - DPFPRINTF(PF_DEBUG_MISC, ("IPv4 nested authentication header limit")); + DPFPRINTF(PF_DEBUG_MISC, "IPv4 nested authentication header limit"); REASON_SET(reason, PFRES_IPOPTIONS); return (PF_DROP); } @@ -9840,7 +9926,7 @@ pf_walk_option6(struct pf_pdesc *pd, struct ip6_hdr *h, int off, int end, while (off < end) { if (!pf_pull_hdr(pd->m, off, &opt.ip6o_type, sizeof(opt.ip6o_type), NULL, reason, AF_INET6)) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 short opt type")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 short opt type"); return (PF_DROP); } if (opt.ip6o_type == IP6OPT_PAD1) { @@ -9849,41 +9935,48 @@ pf_walk_option6(struct pf_pdesc *pd, struct ip6_hdr *h, int off, int end, } if (!pf_pull_hdr(pd->m, off, &opt, sizeof(opt), NULL, reason, AF_INET6)) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 short opt")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 short opt"); return (PF_DROP); } if (off + sizeof(opt) + opt.ip6o_len > end) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 long opt")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 long opt"); REASON_SET(reason, PFRES_IPOPTIONS); return (PF_DROP); } switch (opt.ip6o_type) { + case IP6OPT_PADN: + break; case IP6OPT_JUMBO: + pd->badopts |= PF_OPT_JUMBO; if (pd->jumbolen != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 multiple jumbo")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 multiple jumbo"); REASON_SET(reason, PFRES_IPOPTIONS); return (PF_DROP); } if (ntohs(h->ip6_plen) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 bad jumbo plen")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 bad jumbo plen"); REASON_SET(reason, PFRES_IPOPTIONS); return (PF_DROP); } if (!pf_pull_hdr(pd->m, off, &jumbo, sizeof(jumbo), NULL, reason, AF_INET6)) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 short jumbo")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 short jumbo"); return (PF_DROP); } memcpy(&pd->jumbolen, jumbo.ip6oj_jumbo_len, sizeof(pd->jumbolen)); pd->jumbolen = ntohl(pd->jumbolen); if (pd->jumbolen < IPV6_MAXPACKET) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 short jumbolen")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 short jumbolen"); REASON_SET(reason, PFRES_IPOPTIONS); return (PF_DROP); } break; + case IP6OPT_ROUTER_ALERT: + pd->badopts |= PF_OPT_ROUTER_ALERT; + break; default: + pd->badopts |= PF_OPT_OTHER; break; } off += sizeof(opt) + opt.ip6o_len; @@ -9897,6 +9990,7 @@ pf_walk_header6(struct pf_pdesc *pd, struct ip6_hdr *h, u_short *reason) { struct ip6_frag frag; struct ip6_ext ext; + struct icmp6_hdr icmp6; struct ip6_rthdr rthdr; uint32_t end; int hdr_cnt, fraghdr_cnt = 0, rthdr_cnt = 0; @@ -9908,27 +10002,40 @@ pf_walk_header6(struct pf_pdesc *pd, struct ip6_hdr *h, u_short *reason) for (hdr_cnt = 0; hdr_cnt < PF_HDR_LIMIT; hdr_cnt++) { switch (pd->proto) { case IPPROTO_ROUTING: - case IPPROTO_HOPOPTS: case IPPROTO_DSTOPTS: - pd->badopts++; + pd->badopts |= PF_OPT_OTHER; + break; + case IPPROTO_HOPOPTS: + if (!pf_pull_hdr(pd->m, pd->off, &ext, sizeof(ext), + NULL, reason, AF_INET6)) { + DPFPRINTF(PF_DEBUG_MISC, "IPv6 short exthdr"); + return (PF_DROP); + } + if (pf_walk_option6(pd, h, pd->off + sizeof(ext), + pd->off + (ext.ip6e_len + 1) * 8, + reason) != PF_PASS) + return (PF_DROP); + /* option header which contains only padding is fishy */ + if (pd->badopts == 0) + pd->badopts |= PF_OPT_OTHER; break; } switch (pd->proto) { case IPPROTO_FRAGMENT: if (fraghdr_cnt++) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 multiple fragment")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 multiple fragment"); REASON_SET(reason, PFRES_FRAG); return (PF_DROP); } /* jumbo payload packets cannot be fragmented */ if (pd->jumbolen != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 fragmented jumbo")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 fragmented jumbo"); REASON_SET(reason, PFRES_FRAG); return (PF_DROP); } if (!pf_pull_hdr(pd->m, pd->off, &frag, sizeof(frag), NULL, reason, AF_INET6)) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 short fragment")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 short fragment"); return (PF_DROP); } /* stop walking over non initial fragments */ @@ -9944,7 +10051,7 @@ pf_walk_header6(struct pf_pdesc *pd, struct ip6_hdr *h, u_short *reason) break; case IPPROTO_ROUTING: if (rthdr_cnt++) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 multiple rthdr")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 multiple rthdr"); REASON_SET(reason, PFRES_IPOPTIONS); return (PF_DROP); } @@ -9956,11 +10063,11 @@ pf_walk_header6(struct pf_pdesc *pd, struct ip6_hdr *h, u_short *reason) } if (!pf_pull_hdr(pd->m, pd->off, &rthdr, sizeof(rthdr), NULL, reason, AF_INET6)) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 short rthdr")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 short rthdr"); return (PF_DROP); } if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 rthdr0")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 rthdr0"); REASON_SET(reason, PFRES_IPOPTIONS); return (PF_DROP); } @@ -9968,7 +10075,7 @@ pf_walk_header6(struct pf_pdesc *pd, struct ip6_hdr *h, u_short *reason) case IPPROTO_HOPOPTS: /* RFC2460 4.1: Hop-by-Hop only after IPv6 header */ if (pd->proto == IPPROTO_HOPOPTS && hdr_cnt > 0) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 hopopts not first")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 hopopts not first"); REASON_SET(reason, PFRES_IPOPTIONS); return (PF_DROP); } @@ -9977,7 +10084,7 @@ pf_walk_header6(struct pf_pdesc *pd, struct ip6_hdr *h, u_short *reason) case IPPROTO_DSTOPTS: if (!pf_pull_hdr(pd->m, pd->off, &ext, sizeof(ext), NULL, reason, AF_INET6)) { - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 short exthdr")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 short exthdr"); return (PF_DROP); } /* fragments may be short */ @@ -9989,18 +10096,11 @@ pf_walk_header6(struct pf_pdesc *pd, struct ip6_hdr *h, u_short *reason) /* reassembly needs the ext header before the frag */ if (pd->fragoff == 0) pd->extoff = pd->off; - if (pd->proto == IPPROTO_HOPOPTS && pd->fragoff == 0) { - if (pf_walk_option6(pd, h, - pd->off + sizeof(ext), - pd->off + (ext.ip6e_len + 1) * 8, reason) - != PF_PASS) - return (PF_DROP); - if (ntohs(h->ip6_plen) == 0 && pd->jumbolen != 0) { - DPFPRINTF(PF_DEBUG_MISC, - ("IPv6 missing jumbo")); - REASON_SET(reason, PFRES_IPOPTIONS); - return (PF_DROP); - } + if (pd->proto == IPPROTO_HOPOPTS && pd->fragoff == 0 && + ntohs(h->ip6_plen) == 0 && pd->jumbolen != 0) { + DPFPRINTF(PF_DEBUG_MISC, "IPv6 missing jumbo"); + REASON_SET(reason, PFRES_IPOPTIONS); + return (PF_DROP); } if (pd->proto == IPPROTO_AH) pd->off += (ext.ip6e_len + 2) * 4; @@ -10008,10 +10108,45 @@ pf_walk_header6(struct pf_pdesc *pd, struct ip6_hdr *h, u_short *reason) pd->off += (ext.ip6e_len + 1) * 8; pd->proto = ext.ip6e_nxt; break; + case IPPROTO_ICMPV6: + /* fragments may be short, ignore inner header then */ + if (pd->fragoff != 0 && end < pd->off + sizeof(icmp6)) { + pd->off = pd->fragoff; + pd->proto = IPPROTO_FRAGMENT; + return (PF_PASS); + } + if (!pf_pull_hdr(pd->m, pd->off, &icmp6, sizeof(icmp6), + NULL, reason, AF_INET6)) { + DPFPRINTF(PF_DEBUG_MISC, + "IPv6 short icmp6hdr"); + return (PF_DROP); + } + /* ICMP multicast packets have router alert options */ + switch (icmp6.icmp6_type) { + case MLD_LISTENER_QUERY: + case MLD_LISTENER_REPORT: + case MLD_LISTENER_DONE: + case MLDV2_LISTENER_REPORT: + /* + * According to RFC 2710 all MLD messages are + * sent with hop-limit (ttl) set to 1, and link + * local source address. If either one is + * missing then MLD message is invalid and + * should be discarded. + */ + if ((h->ip6_hlim != 1) || + !IN6_IS_ADDR_LINKLOCAL(&h->ip6_src)) { + DPFPRINTF(PF_DEBUG_MISC, "Invalid MLD"); + REASON_SET(reason, PFRES_IPOPTIONS); + return (PF_DROP); + } + pd->badopts &= ~PF_OPT_ROUTER_ALERT; + break; + } + return (PF_PASS); case IPPROTO_TCP: case IPPROTO_UDP: case IPPROTO_SCTP: - case IPPROTO_ICMPV6: /* fragments may be short, ignore inner header then */ if (pd->fragoff != 0 && end < pd->off + (pd->proto == IPPROTO_TCP ? sizeof(struct tcphdr) : @@ -10026,7 +10161,7 @@ pf_walk_header6(struct pf_pdesc *pd, struct ip6_hdr *h, u_short *reason) return (PF_PASS); } } - DPFPRINTF(PF_DEBUG_MISC, ("IPv6 nested extension header limit")); + DPFPRINTF(PF_DEBUG_MISC, "IPv6 nested extension header limit"); REASON_SET(reason, PFRES_IPOPTIONS); return (PF_DROP); } @@ -10052,6 +10187,8 @@ pf_setup_pdesc(sa_family_t af, int dir, struct pf_pdesc *pd, struct mbuf **m0, pd->didx = (dir == PF_IN) ? 1 : 0; pd->af = pd->naf = af; + PF_RULES_ASSERT(); + TAILQ_INIT(&pd->sctp_multihome_jobs); if (default_actions != NULL) memcpy(&pd->act, default_actions, sizeof(pd->act)); @@ -10069,8 +10206,15 @@ pf_setup_pdesc(sa_family_t af, int dir, struct pf_pdesc *pd, struct mbuf **m0, if (__predict_false((*m0)->m_len < sizeof(struct ip)) && (pd->m = *m0 = m_pullup(*m0, sizeof(struct ip))) == NULL) { DPFPRINTF(PF_DEBUG_URGENT, - ("%s: m_len < sizeof(struct ip), pullup failed\n", - __func__)); + "%s: m_len < sizeof(struct ip), pullup failed", + __func__); + *action = PF_DROP; + REASON_SET(reason, PFRES_SHORT); + return (-1); + } + + h = mtod(pd->m, struct ip *); + if (pd->m->m_pkthdr.len < ntohs(h->ip_len)) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); @@ -10083,13 +10227,7 @@ pf_setup_pdesc(sa_family_t af, int dir, struct pf_pdesc *pd, struct mbuf **m0, return (-1); } *m0 = pd->m; - h = mtod(pd->m, struct ip *); - if (pd->m->m_pkthdr.len < ntohs(h->ip_len)) { - *action = PF_DROP; - REASON_SET(reason, PFRES_SHORT); - return (-1); - } if (pf_walk_header(pd, h, reason) != PF_PASS) { *action = PF_DROP; @@ -10119,14 +10257,29 @@ pf_setup_pdesc(sa_family_t af, int dir, struct pf_pdesc *pd, struct mbuf **m0, if (__predict_false((*m0)->m_len < sizeof(struct ip6_hdr)) && (pd->m = *m0 = m_pullup(*m0, sizeof(struct ip6_hdr))) == NULL) { DPFPRINTF(PF_DEBUG_URGENT, - ("%s: m_len < sizeof(struct ip6_hdr)" - ", pullup failed\n", __func__)); + "%s: m_len < sizeof(struct ip6_hdr)" + ", pullup failed", __func__); *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } h = mtod(pd->m, struct ip6_hdr *); + if (pd->m->m_pkthdr.len < + sizeof(struct ip6_hdr) + ntohs(h->ip6_plen)) { + *action = PF_DROP; + REASON_SET(reason, PFRES_SHORT); + return (-1); + } + + /* + * we do not support jumbogram. if we keep going, zero ip6_plen + * will do something bad, so drop the packet for now. + */ + if (htons(h->ip6_plen) == 0) { + *action = PF_DROP; + return (-1); + } if (pf_walk_header6(pd, h, reason) != PF_PASS) { *action = PF_DROP; @@ -10147,15 +10300,6 @@ pf_setup_pdesc(sa_family_t af, int dir, struct pf_pdesc *pd, struct mbuf **m0, pd->virtual_proto = (pd->fragoff != 0) ? PF_VPROTO_FRAGMENT : pd->proto; - /* - * we do not support jumbogram. if we keep going, zero ip6_plen - * will do something bad, so drop the packet for now. - */ - if (htons(h->ip6_plen) == 0) { - *action = PF_DROP; - return (-1); - } - /* We do IP header normalization and packet reassembly here */ if (pf_normalize_ip6(pd->fragoff, reason, pd) != PF_PASS) { @@ -10465,35 +10609,30 @@ pf_test(sa_family_t af, int dir, int pflags, struct ifnet *ifp, struct mbuf **m0 PF_RULES_RLOCK_TRACKER; KASSERT(dir == PF_IN || dir == PF_OUT, ("%s: bad direction %d\n", __func__, dir)); M_ASSERTPKTHDR(*m0); + NET_EPOCH_ASSERT(); if (!V_pf_status.running) return (PF_PASS); - PF_RULES_RLOCK(); - kif = (struct pfi_kkif *)ifp->if_pf_kif; if (__predict_false(kif == NULL)) { DPFPRINTF(PF_DEBUG_URGENT, - ("%s: kif == NULL, if_xname %s\n", - __func__, ifp->if_xname)); - PF_RULES_RUNLOCK(); + "%s: kif == NULL, if_xname %s", + __func__, ifp->if_xname); return (PF_DROP); } if (kif->pfik_flags & PFI_IFLAG_SKIP) { - PF_RULES_RUNLOCK(); return (PF_PASS); } if ((*m0)->m_flags & M_SKIP_FIREWALL) { - PF_RULES_RUNLOCK(); return (PF_PASS); } if (__predict_false(! M_WRITABLE(*m0))) { *m0 = m_unshare(*m0, M_NOWAIT); if (*m0 == NULL) { - PF_RULES_RUNLOCK(); return (PF_DROP); } } @@ -10506,12 +10645,10 @@ pf_test(sa_family_t af, int dir, int pflags, struct ifnet *ifp, struct mbuf **m0 ifp = ifnet_byindexgen(pd.pf_mtag->if_index, pd.pf_mtag->if_idxgen); if (ifp == NULL || ifp->if_flags & IFF_DYING) { - PF_RULES_RUNLOCK(); m_freem(*m0); *m0 = NULL; return (PF_PASS); } - PF_RULES_RUNLOCK(); (ifp->if_output)(ifp, *m0, sintosa(&pd.pf_mtag->dst), NULL); *m0 = NULL; return (PF_PASS); @@ -10526,11 +10663,12 @@ pf_test(sa_family_t af, int dir, int pflags, struct ifnet *ifp, struct mbuf **m0 /* But only once. We may see the packet multiple times (e.g. * PFIL_IN/PFIL_OUT). */ pf_dummynet_flag_remove(pd.m, pd.pf_mtag); - PF_RULES_RUNLOCK(); return (PF_PASS); } + PF_RULES_RLOCK(); + if (pf_setup_pdesc(af, dir, &pd, m0, &action, &reason, kif, default_actions) == -1) { if (action != PF_PASS) @@ -10685,14 +10823,14 @@ pf_test(sa_family_t af, int dir, int pflags, struct ifnet *ifp, struct mbuf **m0 action = PF_DROP; REASON_SET(&reason, PFRES_NORM); DPFPRINTF(PF_DEBUG_MISC, - ("dropping IPv6 packet with ICMPv4 payload")); + "dropping IPv6 packet with ICMPv4 payload"); break; } if (pd.virtual_proto == IPPROTO_ICMPV6 && af != AF_INET6) { action = PF_DROP; REASON_SET(&reason, PFRES_NORM); DPFPRINTF(PF_DEBUG_MISC, - ("pf: dropping IPv4 packet with ICMPv6 payload\n")); + "pf: dropping IPv4 packet with ICMPv6 payload"); break; } action = pf_test_state_icmp(&s, &pd, &reason); @@ -10718,12 +10856,12 @@ done: if (s) memcpy(&pd.act, &s->act, sizeof(s->act)); - if (action == PF_PASS && pd.badopts && !pd.act.allow_opts) { + if (action == PF_PASS && pd.badopts != 0 && !pd.act.allow_opts) { action = PF_DROP; REASON_SET(&reason, PFRES_IPOPTIONS); pd.act.log = PF_LOG_FORCE; DPFPRINTF(PF_DEBUG_MISC, - ("pf: dropping packet with dangerous headers\n")); + "pf: dropping packet with dangerous headers"); } if (pd.act.max_pkt_size && pd.act.max_pkt_size && @@ -10732,7 +10870,7 @@ done: REASON_SET(&reason, PFRES_NORM); pd.act.log = PF_LOG_FORCE; DPFPRINTF(PF_DEBUG_MISC, - ("pf: dropping overly long packet\n")); + "pf: dropping overly long packet"); } if (s) { @@ -10764,7 +10902,7 @@ done: REASON_SET(&reason, PFRES_MEMORY); pd.act.log = PF_LOG_FORCE; DPFPRINTF(PF_DEBUG_MISC, - ("pf: failed to allocate 802.1q mtag\n")); + "pf: failed to allocate 802.1q mtag"); } } @@ -10821,7 +10959,7 @@ done: REASON_SET(&reason, PFRES_MEMORY); pd.act.log = PF_LOG_FORCE; DPFPRINTF(PF_DEBUG_MISC, - ("pf: failed to allocate tag\n")); + "pf: failed to allocate tag"); } else { pd.pf_mtag->flags |= PF_MTAG_FLAG_FASTFWD_OURS_PRESENT; @@ -10838,7 +10976,7 @@ done: REASON_SET(&reason, PFRES_MEMORY); pd.act.log = PF_LOG_FORCE; DPFPRINTF(PF_DEBUG_MISC, - ("pf: failed to allocate divert tag\n")); + "pf: failed to allocate divert tag"); } } /* XXX: Anybody working on it?! */ diff --git a/sys/netpfil/pf/pf.h b/sys/netpfil/pf/pf.h index 2009d2907985..cfff58064922 100644 --- a/sys/netpfil/pf/pf.h +++ b/sys/netpfil/pf/pf.h @@ -140,7 +140,7 @@ enum { PF_ADDR_ADDRMASK, PF_ADDR_NOROUTE, PF_ADDR_DYNIFTL, #define PF_LOG 0x01 #define PF_LOG_ALL 0x02 -#define PF_LOG_SOCKET_LOOKUP 0x04 +#define PF_LOG_USER 0x04 #define PF_LOG_FORCE 0x08 #define PF_LOG_MATCHES 0x10 @@ -490,6 +490,7 @@ struct pf_osfp_ioctl { #define PF_ANCHOR_NAME_SIZE 64 #define PF_ANCHOR_MAXPATH (MAXPATHLEN - PF_ANCHOR_NAME_SIZE - 1) +#define PF_OPTIMIZER_TABLE_PFX "__automatic_" struct pf_rule { struct pf_rule_addr src; diff --git a/sys/netpfil/pf/pf_ioctl.c b/sys/netpfil/pf/pf_ioctl.c index 357b2be194a5..9abc07c36788 100644 --- a/sys/netpfil/pf/pf_ioctl.c +++ b/sys/netpfil/pf/pf_ioctl.c @@ -217,8 +217,6 @@ static u_int16_t tagname2tag(struct pf_tagset *, const char *); static u_int16_t pf_tagname2tag(const char *); static void tag_unref(struct pf_tagset *, u_int16_t); -#define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x - struct cdev *pf_dev; /* @@ -1274,7 +1272,9 @@ pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr) PF_MD5_UPD(pfr, addr.iflags); break; case PF_ADDR_TABLE: - PF_MD5_UPD(pfr, addr.v.tblname); + if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX, + strlen(PF_OPTIMIZER_TABLE_PFX))) + PF_MD5_UPD(pfr, addr.v.tblname); break; case PF_ADDR_ADDRMASK: /* XXX ignore af? */ @@ -1357,7 +1357,7 @@ static int pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) { struct pf_kruleset *rs; - struct pf_krule *rule, **old_array, *old_rule; + struct pf_krule *rule, *old_rule; struct pf_krulequeue *old_rules; struct pf_krule_global *old_tree; int error; @@ -1382,13 +1382,10 @@ pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) /* Swap rules, keep the old. */ old_rules = rs->rules[rs_num].active.ptr; old_rcount = rs->rules[rs_num].active.rcount; - old_array = rs->rules[rs_num].active.ptr_array; old_tree = rs->rules[rs_num].active.tree; rs->rules[rs_num].active.ptr = rs->rules[rs_num].inactive.ptr; - rs->rules[rs_num].active.ptr_array = - rs->rules[rs_num].inactive.ptr_array; rs->rules[rs_num].active.tree = rs->rules[rs_num].inactive.tree; rs->rules[rs_num].active.rcount = @@ -1418,7 +1415,6 @@ pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) } rs->rules[rs_num].inactive.ptr = old_rules; - rs->rules[rs_num].inactive.ptr_array = old_array; rs->rules[rs_num].inactive.tree = NULL; /* important for pf_ioctl_addrule */ rs->rules[rs_num].inactive.rcount = old_rcount; @@ -1431,9 +1427,6 @@ pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor) while ((rule = TAILQ_FIRST(old_rules)) != NULL) pf_unlink_rule_locked(old_rules, rule); PF_UNLNKDRULES_UNLOCK(); - if (rs->rules[rs_num].inactive.ptr_array) - free(rs->rules[rs_num].inactive.ptr_array, M_TEMP); - rs->rules[rs_num].inactive.ptr_array = NULL; rs->rules[rs_num].inactive.rcount = 0; rs->rules[rs_num].inactive.open = 0; pf_remove_if_empty_kruleset(rs); @@ -1456,24 +1449,11 @@ pf_setup_pfsync_matching(struct pf_kruleset *rs) if (rs_cnt == PF_RULESET_SCRUB) continue; - if (rs->rules[rs_cnt].inactive.ptr_array) - free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP); - rs->rules[rs_cnt].inactive.ptr_array = NULL; - if (rs->rules[rs_cnt].inactive.rcount) { - rs->rules[rs_cnt].inactive.ptr_array = - mallocarray(rs->rules[rs_cnt].inactive.rcount, - sizeof(struct pf_rule **), - M_TEMP, M_NOWAIT); - - if (!rs->rules[rs_cnt].inactive.ptr_array) - return (ENOMEM); - } - - TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, - entries) { - pf_hash_rule_rolling(&ctx, rule); - (rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule; + TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr, + entries) { + pf_hash_rule_rolling(&ctx, rule); + } } } @@ -2059,6 +2039,47 @@ pf_ioctl_getrules(struct pfioc_rule *pr) return (0); } +static int +pf_rule_checkaf(struct pf_krule *r) +{ + switch (r->af) { + case 0: + if (r->rule_flag & PFRULE_AFTO) + return (EPFNOSUPPORT); + break; + case AF_INET: + if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET6) + return (EPFNOSUPPORT); + break; +#ifdef INET6 + case AF_INET6: + if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET) + return (EPFNOSUPPORT); + break; +#endif /* INET6 */ + default: + return (EPFNOSUPPORT); + } + + if ((r->rule_flag & PFRULE_AFTO) == 0 && r->naf != 0) + return (EPFNOSUPPORT); + + return (0); +} + +static int +pf_validate_range(uint8_t op, uint16_t port[2]) +{ + uint16_t a = ntohs(port[0]); + uint16_t b = ntohs(port[1]); + + if ((op == PF_OP_RRG && a > b) || /* 34:12, i.e. none */ + (op == PF_OP_IRG && a >= b) || /* 34><12, i.e. none */ + (op == PF_OP_XRG && a > b)) /* 34<>22, i.e. all */ + return 1; + return 0; +} + int pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, uint32_t pool_ticket, const char *anchor, const char *anchor_call, @@ -2071,12 +2092,18 @@ pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, int rs_num; int error = 0; - if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) { - error = EINVAL; - goto errout_unlocked; - } +#define ERROUT(x) ERROUT_FUNCTION(errout, x) +#define ERROUT_UNLOCKED(x) ERROUT_FUNCTION(errout_unlocked, x) + + if ((rule->return_icmp >> 8) > ICMP_MAXTYPE) + ERROUT_UNLOCKED(EINVAL); -#define ERROUT(x) ERROUT_FUNCTION(errout, x) + if ((error = pf_rule_checkaf(rule))) + ERROUT_UNLOCKED(error); + if (pf_validate_range(rule->src.port_op, rule->src.port)) + ERROUT_UNLOCKED(EINVAL); + if (pf_validate_range(rule->dst.port_op, rule->dst.port)) + ERROUT_UNLOCKED(EINVAL); if (rule->ifname[0]) kif = pf_kkif_create(M_WAITOK); @@ -2113,14 +2140,14 @@ pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, ERROUT(EINVAL); if (ticket != ruleset->rules[rs_num].inactive.ticket) { DPFPRINTF(PF_DEBUG_MISC, - ("ticket: %d != [%d]%d\n", ticket, rs_num, - ruleset->rules[rs_num].inactive.ticket)); + "ticket: %d != [%d]%d", ticket, rs_num, + ruleset->rules[rs_num].inactive.ticket); ERROUT(EBUSY); } if (pool_ticket != V_ticket_pabuf) { DPFPRINTF(PF_DEBUG_MISC, - ("pool_ticket: %d != %d\n", pool_ticket, - V_ticket_pabuf)); + "pool_ticket: %d != %d", pool_ticket, + V_ticket_pabuf); ERROUT(EBUSY); } /* @@ -2155,51 +2182,51 @@ pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, rule->rcv_kif = NULL; if (rule->rtableid > 0 && rule->rtableid >= rt_numfibs) - error = EBUSY; + ERROUT(EBUSY); #ifdef ALTQ /* set queue IDs */ if (rule->qname[0] != 0) { if ((rule->qid = pf_qname2qid(rule->qname)) == 0) - error = EBUSY; + ERROUT(EBUSY); else if (rule->pqname[0] != 0) { if ((rule->pqid = pf_qname2qid(rule->pqname)) == 0) - error = EBUSY; + ERROUT(EBUSY); } else rule->pqid = rule->qid; } #endif if (rule->tagname[0]) if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0) - error = EBUSY; + ERROUT(EBUSY); if (rule->match_tagname[0]) if ((rule->match_tag = pf_tagname2tag(rule->match_tagname)) == 0) - error = EBUSY; + ERROUT(EBUSY); if (rule->rt && !rule->direction) - error = EINVAL; + ERROUT(EINVAL); if (!rule->log) rule->logif = 0; if (! pf_init_threshold(&rule->pktrate, rule->pktrate.limit, rule->pktrate.seconds)) - error = ENOMEM; + ERROUT(ENOMEM); if (pf_addr_setup(ruleset, &rule->src.addr, rule->af)) - error = ENOMEM; + ERROUT(ENOMEM); if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af)) - error = ENOMEM; + ERROUT(ENOMEM); if (pf_kanchor_setup(rule, ruleset, anchor_call)) - error = EINVAL; + ERROUT(EINVAL); if (rule->scrub_flags & PFSTATE_SETPRIO && (rule->set_prio[0] > PF_PRIO_MAX || rule->set_prio[1] > PF_PRIO_MAX)) - error = EINVAL; + ERROUT(EINVAL); for (int i = 0; i < 3; i++) { TAILQ_FOREACH(pa, &V_pf_pabuf[i], entries) if (pa->addr.type == PF_ADDR_TABLE) { pa->addr.p.tbl = pfr_attach_table(ruleset, pa->addr.v.tblname); if (pa->addr.p.tbl == NULL) - error = ENOMEM; + ERROUT(ENOMEM); } } @@ -2207,7 +2234,7 @@ pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, if (rule->overload_tblname[0]) { if ((rule->overload_tbl = pfr_attach_table(ruleset, rule->overload_tblname)) == NULL) - error = EINVAL; + ERROUT(EINVAL); else rule->overload_tbl->pfrkt_flags |= PFR_TFLAG_ACTIVE; @@ -2230,23 +2257,19 @@ pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, if (((rule->action == PF_NAT) || (rule->action == PF_RDR) || (rule->action == PF_BINAT)) && rule->anchor == NULL && TAILQ_FIRST(&rule->rdr.list) == NULL) { - error = EINVAL; + ERROUT(EINVAL); } if (rule->rt > PF_NOPFROUTE && (TAILQ_FIRST(&rule->route.list) == NULL)) { - error = EINVAL; + ERROUT(EINVAL); } if (rule->action == PF_PASS && (rule->rdr.opts & PF_POOL_STICKYADDR || rule->nat.opts & PF_POOL_STICKYADDR) && !rule->keep_state) { - error = EINVAL; + ERROUT(EINVAL); } - if (error) { - pf_free_rule(rule); - rule = NULL; - ERROUT(error); - } + MPASS(error == 0); rule->nat.cur = TAILQ_FIRST(&rule->nat.list); rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list); @@ -2270,6 +2293,7 @@ pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, return (0); #undef ERROUT +#undef ERROUT_UNLOCKED errout: PF_RULES_WUNLOCK(); PF_CONFIG_UNLOCK(); @@ -2443,7 +2467,7 @@ pf_start(void) V_pf_status.since = time_uptime; new_unrhdr64(&V_pf_stateid, time_second); - DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n")); + DPFPRINTF(PF_DEBUG_MISC, "pf: started"); } sx_xunlock(&V_pf_ioctl_lock); @@ -2463,7 +2487,7 @@ pf_stop(void) dehook_pf(); dehook_pf_eth(); V_pf_status.since = time_uptime; - DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n")); + DPFPRINTF(PF_DEBUG_MISC, "pf: stopped"); } sx_xunlock(&V_pf_ioctl_lock); @@ -2699,7 +2723,7 @@ pf_ioctl_get_addr(struct pf_nl_pooladdr *pp) PF_RULES_RLOCK_TRACKER; - pp->anchor[sizeof(pp->anchor) - 1] = 0; + pp->anchor[sizeof(pp->anchor) - 1] = '\0'; PF_RULES_RLOCK(); pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action, @@ -2732,7 +2756,7 @@ pf_ioctl_get_rulesets(struct pfioc_ruleset *pr) PF_RULES_RLOCK_TRACKER; - pr->path[sizeof(pr->path) - 1] = 0; + pr->path[sizeof(pr->path) - 1] = '\0'; PF_RULES_RLOCK(); if ((ruleset = pf_find_kruleset(pr->path)) == NULL) { @@ -2771,7 +2795,7 @@ pf_ioctl_get_ruleset(struct pfioc_ruleset *pr) return (ENOENT); } - pr->name[0] = 0; + pr->name[0] = '\0'; if (ruleset == &pf_main_ruleset) { /* XXX kludge for pf_main_ruleset */ RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) @@ -2796,6 +2820,78 @@ pf_ioctl_get_ruleset(struct pfioc_ruleset *pr) return (error); } +int +pf_ioctl_natlook(struct pfioc_natlook *pnl) +{ + struct pf_state_key *sk; + struct pf_kstate *state; + struct pf_state_key_cmp key; + int m = 0, direction = pnl->direction; + int sidx, didx; + + /* NATLOOK src and dst are reversed, so reverse sidx/didx */ + sidx = (direction == PF_IN) ? 1 : 0; + didx = (direction == PF_IN) ? 0 : 1; + + if (!pnl->proto || + PF_AZERO(&pnl->saddr, pnl->af) || + PF_AZERO(&pnl->daddr, pnl->af) || + ((pnl->proto == IPPROTO_TCP || + pnl->proto == IPPROTO_UDP) && + (!pnl->dport || !pnl->sport))) + return (EINVAL); + + switch (pnl->direction) { + case PF_IN: + case PF_OUT: + case PF_INOUT: + break; + default: + return (EINVAL); + } + + switch (pnl->af) { +#ifdef INET + case AF_INET: + break; +#endif /* INET */ +#ifdef INET6 + case AF_INET6: + break; +#endif /* INET6 */ + default: + return (EAFNOSUPPORT); + } + + bzero(&key, sizeof(key)); + key.af = pnl->af; + key.proto = pnl->proto; + pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af); + key.port[sidx] = pnl->sport; + pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af); + key.port[didx] = pnl->dport; + + state = pf_find_state_all(&key, direction, &m); + if (state == NULL) + return (ENOENT); + + if (m > 1) { + PF_STATE_UNLOCK(state); + return (E2BIG); /* more than one state */ + } + + sk = state->key[sidx]; + pf_addrcpy(&pnl->rsaddr, + &sk->addr[sidx], sk->af); + pnl->rsport = sk->port[sidx]; + pf_addrcpy(&pnl->rdaddr, + &sk->addr[didx], sk->af); + pnl->rdport = sk->port[didx]; + PF_STATE_UNLOCK(state); + + return (0); +} + static int pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td) { @@ -3166,9 +3262,9 @@ DIOCGETETHRULE_error: if (nvlist_get_number(nvl, "ticket") != ruleset->inactive.ticket) { DPFPRINTF(PF_DEBUG_MISC, - ("ticket: %d != %d\n", + "ticket: %d != %d", (u_int32_t)nvlist_get_number(nvl, "ticket"), - ruleset->inactive.ticket)); + ruleset->inactive.ticket); ERROUT(EBUSY); } @@ -3499,10 +3595,10 @@ DIOCADDRULENV_error: error = pf_rule_to_krule(&pr->rule, rule); if (error != 0) { pf_krule_free(rule); - break; + goto fail; } - pr->anchor[sizeof(pr->anchor) - 1] = 0; + pr->anchor[sizeof(pr->anchor) - 1] = '\0'; /* Frees rule on error */ error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket, @@ -3514,7 +3610,7 @@ DIOCADDRULENV_error: case DIOCGETRULES: { struct pfioc_rule *pr = (struct pfioc_rule *)addr; - pr->anchor[sizeof(pr->anchor) - 1] = 0; + pr->anchor[sizeof(pr->anchor) - 1] = '\0'; error = pf_ioctl_getrules(pr); @@ -3653,16 +3749,16 @@ DIOCGETRULENV_error: u_int32_t nr = 0; int rs_num; - pcr->anchor[sizeof(pcr->anchor) - 1] = 0; + pcr->anchor[sizeof(pcr->anchor) - 1] = '\0'; if (pcr->action < PF_CHANGE_ADD_HEAD || pcr->action > PF_CHANGE_GET_TICKET) { error = EINVAL; - break; + goto fail; } if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) { error = EINVAL; - break; + goto fail; } if (pcr->action != PF_CHANGE_REMOVE) { @@ -3670,9 +3766,13 @@ DIOCGETRULENV_error: error = pf_rule_to_krule(&pcr->rule, newrule); if (error != 0) { pf_krule_free(newrule); - break; + goto fail; } + if ((error = pf_rule_checkaf(newrule))) { + pf_krule_free(newrule); + goto fail; + } if (newrule->ifname[0]) kif = pf_kkif_create(M_WAITOK); pf_counter_u64_init(&newrule->evaluations, M_WAITOK); @@ -3820,7 +3920,7 @@ DIOCGETRULENV_error: pf_free_rule(newrule); PF_RULES_WUNLOCK(); PF_CONFIG_UNLOCK(); - break; + goto fail; } newrule->nat.cur = TAILQ_FIRST(&newrule->nat.list); @@ -3847,7 +3947,7 @@ DIOCGETRULENV_error: PF_RULES_WUNLOCK(); PF_CONFIG_UNLOCK(); error = EINVAL; - break; + goto fail; } } @@ -3865,7 +3965,7 @@ DIOCGETRULENV_error: PF_RULES_WUNLOCK(); PF_CONFIG_UNLOCK(); error = EEXIST; - break; + goto fail; } if (oldrule == NULL) @@ -3921,7 +4021,7 @@ DIOCCHANGERULE_error: if (sp->timeout >= PFTM_MAX) { error = EINVAL; - break; + goto fail; } if (V_pfsync_state_import_ptr != NULL) { PF_RULES_RLOCK(); @@ -3941,7 +4041,7 @@ DIOCCHANGERULE_error: s = pf_find_state_byid(ps->state.id, ps->state.creatorid); if (s == NULL) { error = ENOENT; - break; + goto fail; } pfsync_state_export((union pfsync_state_union*)&ps->state, @@ -4020,7 +4120,7 @@ DIOCGETSTATES_retry: error = copyout(pstore, out, sizeof(struct pfsync_state_1301) * count); if (error) - break; + goto fail; out = ps->ps_states + nr; } DIOCGETSTATES_full: @@ -4040,7 +4140,7 @@ DIOCGETSTATES_full: if (ps->ps_req_version > PF_STATE_VERSION) { error = ENOTSUP; - break; + goto fail; } if (ps->ps_len <= 0) { @@ -4098,7 +4198,7 @@ DIOCGETSTATESV2_retry: error = copyout(pstore, out, sizeof(struct pf_state_export) * count); if (error) - break; + goto fail; out = ps->ps_states + nr; } DIOCGETSTATESV2_full: @@ -4133,51 +4233,8 @@ DIOCGETSTATESV2_full: case DIOCNATLOOK: { struct pfioc_natlook *pnl = (struct pfioc_natlook *)addr; - struct pf_state_key *sk; - struct pf_kstate *state; - struct pf_state_key_cmp key; - int m = 0, direction = pnl->direction; - int sidx, didx; - - /* NATLOOK src and dst are reversed, so reverse sidx/didx */ - sidx = (direction == PF_IN) ? 1 : 0; - didx = (direction == PF_IN) ? 0 : 1; - - if (!pnl->proto || - PF_AZERO(&pnl->saddr, pnl->af) || - PF_AZERO(&pnl->daddr, pnl->af) || - ((pnl->proto == IPPROTO_TCP || - pnl->proto == IPPROTO_UDP) && - (!pnl->dport || !pnl->sport))) - error = EINVAL; - else { - bzero(&key, sizeof(key)); - key.af = pnl->af; - key.proto = pnl->proto; - pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af); - key.port[sidx] = pnl->sport; - pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af); - key.port[didx] = pnl->dport; - - state = pf_find_state_all(&key, direction, &m); - if (state == NULL) { - error = ENOENT; - } else { - if (m > 1) { - PF_STATE_UNLOCK(state); - error = E2BIG; /* more than one state */ - } else { - sk = state->key[sidx]; - pf_addrcpy(&pnl->rsaddr, - &sk->addr[sidx], sk->af); - pnl->rsport = sk->port[sidx]; - pf_addrcpy(&pnl->rdaddr, - &sk->addr[didx], sk->af); - pnl->rdport = sk->port[didx]; - PF_STATE_UNLOCK(state); - } - } - } + + error = pf_ioctl_natlook(pnl); break; } @@ -4247,12 +4304,12 @@ DIOCGETSTATESV2_full: if (psp->ifname[0] == '\0') { error = EINVAL; - break; + goto fail; } error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ); if (error != 0) - break; + goto fail; ifp = ifunit(ps.ifname); if (ifp != NULL) { psp->baudrate32 = @@ -4281,7 +4338,7 @@ DIOCGETSTATESV2_full: if (error == 0) V_pf_altq_running = 1; PF_RULES_WUNLOCK(); - DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n")); + DPFPRINTF(PF_DEBUG_MISC, "altq: started"); break; } @@ -4300,7 +4357,7 @@ DIOCGETSTATESV2_full: if (error == 0) V_pf_altq_running = 0; PF_RULES_WUNLOCK(); - DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n")); + DPFPRINTF(PF_DEBUG_MISC, "altq: stopped"); break; } @@ -4313,7 +4370,7 @@ DIOCGETSTATESV2_full: altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO); error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd)); if (error) - break; + goto fail; altq->local_flags = 0; PF_RULES_WLOCK(); @@ -4321,7 +4378,7 @@ DIOCGETSTATESV2_full: PF_RULES_WUNLOCK(); free(altq, M_PFALTQ); error = EBUSY; - break; + goto fail; } /* @@ -4333,7 +4390,7 @@ DIOCGETSTATESV2_full: PF_RULES_WUNLOCK(); error = EBUSY; free(altq, M_PFALTQ); - break; + goto fail; } altq->altq_disc = NULL; TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) { @@ -4353,7 +4410,7 @@ DIOCGETSTATESV2_full: if (error) { PF_RULES_WUNLOCK(); free(altq, M_PFALTQ); - break; + goto fail; } if (altq->qname[0] != 0) @@ -4391,13 +4448,13 @@ DIOCGETSTATESV2_full: if (pa->ticket != V_ticket_altqs_active) { PF_RULES_RUNLOCK(); error = EBUSY; - break; + goto fail; } altq = pf_altq_get_nth_active(pa->nr); if (altq == NULL) { PF_RULES_RUNLOCK(); error = EBUSY; - break; + goto fail; } pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd)); PF_RULES_RUNLOCK(); @@ -4421,20 +4478,20 @@ DIOCGETSTATESV2_full: if (pq->ticket != V_ticket_altqs_active) { PF_RULES_RUNLOCK(); error = EBUSY; - break; + goto fail; } nbytes = pq->nbytes; altq = pf_altq_get_nth_active(pq->nr); if (altq == NULL) { PF_RULES_RUNLOCK(); error = EBUSY; - break; + goto fail; } if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) { PF_RULES_RUNLOCK(); error = ENXIO; - break; + goto fail; } PF_RULES_RUNLOCK(); if (cmd == DIOCGETQSTATSV0) @@ -4498,35 +4555,35 @@ DIOCGETSTATESV2_full: struct pf_kruleset *ruleset; struct pfi_kkif *kif = NULL; - pca->anchor[sizeof(pca->anchor) - 1] = 0; + pca->anchor[sizeof(pca->anchor) - 1] = '\0'; if (pca->action < PF_CHANGE_ADD_HEAD || pca->action > PF_CHANGE_REMOVE) { error = EINVAL; - break; + goto fail; } if (pca->addr.addr.type != PF_ADDR_ADDRMASK && pca->addr.addr.type != PF_ADDR_DYNIFTL && pca->addr.addr.type != PF_ADDR_TABLE) { error = EINVAL; - break; + goto fail; } if (pca->addr.addr.p.dyn != NULL) { error = EINVAL; - break; + goto fail; } if (pca->action != PF_CHANGE_REMOVE) { #ifndef INET if (pca->af == AF_INET) { error = EAFNOSUPPORT; - break; + goto fail; } #endif /* INET */ #ifndef INET6 if (pca->af == AF_INET6) { error = EAFNOSUPPORT; - break; + goto fail; } #endif /* INET6 */ newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK); @@ -4629,7 +4686,7 @@ DIOCCHANGEADDR_error: case DIOCGETRULESETS: { struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; - pr->path[sizeof(pr->path) - 1] = 0; + pr->path[sizeof(pr->path) - 1] = '\0'; error = pf_ioctl_get_rulesets(pr); break; @@ -4638,7 +4695,7 @@ DIOCCHANGEADDR_error: case DIOCGETRULESET: { struct pfioc_ruleset *pr = (struct pfioc_ruleset *)addr; - pr->path[sizeof(pr->path) - 1] = 0; + pr->path[sizeof(pr->path) - 1] = '\0'; error = pf_ioctl_get_ruleset(pr); break; @@ -4649,7 +4706,7 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != 0) { error = ENODEV; - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel, @@ -4665,13 +4722,13 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { error = ENOMEM; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_table); @@ -4680,7 +4737,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfrts, totlen); if (error) { free(pfrts, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_add_tables(pfrts, io->pfrio_size, @@ -4697,13 +4754,13 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) { error = ENOMEM; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_table); @@ -4712,7 +4769,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfrts, totlen); if (error) { free(pfrts, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_del_tables(pfrts, io->pfrio_size, @@ -4730,14 +4787,14 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; - break; + goto fail; } PF_RULES_RLOCK(); n = pfr_table_count(&io->pfrio_table, io->pfrio_flags); if (n < 0) { PF_RULES_RUNLOCK(); error = EINVAL; - break; + goto fail; } io->pfrio_size = min(io->pfrio_size, n); @@ -4748,7 +4805,7 @@ DIOCCHANGEADDR_error: if (pfrts == NULL) { error = ENOMEM; PF_RULES_RUNLOCK(); - break; + goto fail; } error = pfr_get_tables(&io->pfrio_table, pfrts, &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); @@ -4767,7 +4824,7 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_tstats)) { error = ENODEV; - break; + goto fail; } PF_TABLE_STATS_LOCK(); PF_RULES_RLOCK(); @@ -4776,7 +4833,7 @@ DIOCCHANGEADDR_error: PF_RULES_RUNLOCK(); PF_TABLE_STATS_UNLOCK(); error = EINVAL; - break; + goto fail; } io->pfrio_size = min(io->pfrio_size, n); @@ -4787,7 +4844,7 @@ DIOCCHANGEADDR_error: error = ENOMEM; PF_RULES_RUNLOCK(); PF_TABLE_STATS_UNLOCK(); - break; + goto fail; } error = pfr_get_tstats(&io->pfrio_table, pfrtstats, &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL); @@ -4806,7 +4863,7 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || @@ -4815,7 +4872,7 @@ DIOCCHANGEADDR_error: * size, so we didn't fail on overly large requests. * Keep doing so. */ io->pfrio_size = pf_ioctl_maxcount; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_table); @@ -4824,7 +4881,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfrts, totlen); if (error) { free(pfrts, M_TEMP); - break; + goto fail; } PF_TABLE_STATS_LOCK(); @@ -4845,7 +4902,7 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_table)) { error = ENODEV; - break; + goto fail; } PF_RULES_RLOCK(); @@ -4853,7 +4910,7 @@ DIOCCHANGEADDR_error: if (n < 0) { PF_RULES_RUNLOCK(); error = EINVAL; - break; + goto fail; } io->pfrio_size = min(io->pfrio_size, n); @@ -4865,7 +4922,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfrts, totlen); if (error) { free(pfrts, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_set_tflags(pfrts, io->pfrio_size, @@ -4881,7 +4938,7 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != 0) { error = ENODEV; - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel, @@ -4897,13 +4954,13 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { error = EINVAL; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), @@ -4911,7 +4968,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_add_addrs(&io->pfrio_table, pfras, @@ -4931,13 +4988,13 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { error = EINVAL; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), @@ -4945,7 +5002,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_del_addrs(&io->pfrio_table, pfras, @@ -4965,17 +5022,17 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size2 < 0) { error = EINVAL; - break; + goto fail; } count = max(io->pfrio_size, io->pfrio_size2); if (count > pf_ioctl_maxcount || WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) { error = EINVAL; - break; + goto fail; } totlen = count * sizeof(struct pfr_addr); pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP, @@ -4983,7 +5040,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_set_addrs(&io->pfrio_table, pfras, @@ -5004,13 +5061,13 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { error = EINVAL; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), @@ -5032,13 +5089,13 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_astats)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) { error = EINVAL; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_astats); pfrastats = mallocarray(io->pfrio_size, @@ -5060,13 +5117,13 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { error = EINVAL; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), @@ -5074,7 +5131,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_clr_astats(&io->pfrio_table, pfras, @@ -5094,13 +5151,13 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { error = EINVAL; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), @@ -5108,7 +5165,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); - break; + goto fail; } PF_RULES_RLOCK(); error = pfr_tst_addrs(&io->pfrio_table, pfras, @@ -5128,13 +5185,13 @@ DIOCCHANGEADDR_error: if (io->pfrio_esize != sizeof(struct pfr_addr)) { error = ENODEV; - break; + goto fail; } if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) { error = EINVAL; - break; + goto fail; } totlen = io->pfrio_size * sizeof(struct pfr_addr); pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr), @@ -5142,7 +5199,7 @@ DIOCCHANGEADDR_error: error = copyin(io->pfrio_buffer, pfras, totlen); if (error) { free(pfras, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); error = pfr_ina_define(&io->pfrio_table, pfras, @@ -5177,13 +5234,13 @@ DIOCCHANGEADDR_error: if (io->esize != sizeof(*ioe)) { error = ENODEV; - break; + goto fail; } if (io->size < 0 || io->size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { error = EINVAL; - break; + goto fail; } totlen = sizeof(struct pfioc_trans_e) * io->size; ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), @@ -5191,7 +5248,7 @@ DIOCCHANGEADDR_error: error = copyin(io->array, ioes, totlen); if (error) { free(ioes, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { @@ -5258,13 +5315,13 @@ DIOCCHANGEADDR_error: if (io->esize != sizeof(*ioe)) { error = ENODEV; - break; + goto fail; } if (io->size < 0 || io->size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { error = EINVAL; - break; + goto fail; } totlen = sizeof(struct pfioc_trans_e) * io->size; ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e), @@ -5272,7 +5329,7 @@ DIOCCHANGEADDR_error: error = copyin(io->array, ioes, totlen); if (error) { free(ioes, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { @@ -5341,14 +5398,14 @@ DIOCCHANGEADDR_error: if (io->esize != sizeof(*ioe)) { error = ENODEV; - break; + goto fail; } if (io->size < 0 || io->size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) { error = EINVAL; - break; + goto fail; } totlen = sizeof(struct pfioc_trans_e) * io->size; @@ -5357,12 +5414,12 @@ DIOCCHANGEADDR_error: error = copyin(io->array, ioes, totlen); if (error) { free(ioes, M_TEMP); - break; + goto fail; } PF_RULES_WLOCK(); /* First makes sure everything will succeed. */ for (i = 0, ioe = ioes; i < io->size; i++, ioe++) { - ioe->anchor[sizeof(ioe->anchor) - 1] = 0; + ioe->anchor[sizeof(ioe->anchor) - 1] = '\0'; switch (ioe->rs_num) { case PF_RULESET_ETH: ers = pf_find_keth_ruleset(ioe->anchor); @@ -5498,7 +5555,7 @@ DIOCCHANGEADDR_error: if (psn->psn_len == 0) { psn->psn_len = sizeof(struct pf_src_node) * nr; - break; + goto fail; } nr = 0; @@ -5523,7 +5580,7 @@ DIOCCHANGEADDR_error: sizeof(struct pf_src_node) * nr); if (error) { free(pstore, M_TEMP); - break; + goto fail; } psn->psn_len = sizeof(struct pf_src_node) * nr; free(pstore, M_TEMP); @@ -5579,14 +5636,14 @@ DIOCCHANGEADDR_error: if (io->pfiio_esize != sizeof(struct pfi_kif)) { error = ENODEV; - break; + goto fail; } if (io->pfiio_size < 0 || io->pfiio_size > pf_ioctl_maxcount || WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) { error = EINVAL; - break; + goto fail; } io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0'; @@ -6398,9 +6455,9 @@ shutdown_pf(void) for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) { if ((error = pf_begin_rules(&t[rs_num], rs_num, anchor->path)) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("%s: " - "anchor.path=%s rs_num=%d\n", - __func__, anchor->path, rs_num)); + DPFPRINTF(PF_DEBUG_MISC, "%s: " + "anchor.path=%s rs_num=%d", + __func__, anchor->path, rs_num); goto error; /* XXX: rollback? */ } } @@ -6422,9 +6479,9 @@ shutdown_pf(void) eth_anchor->refcnt = 1; if ((error = pf_begin_eth(&t[0], eth_anchor->path)) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("%s: eth " - "anchor.path=%s\n", __func__, - eth_anchor->path)); + DPFPRINTF(PF_DEBUG_MISC, "%s: eth " + "anchor.path=%s", __func__, + eth_anchor->path); goto error; } error = pf_commit_eth(t[0], eth_anchor->path); @@ -6433,27 +6490,27 @@ shutdown_pf(void) if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("%s: SCRUB\n", __func__)); + DPFPRINTF(PF_DEBUG_MISC, "%s: SCRUB", __func__); break; } if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("%s: FILTER\n", __func__)); + DPFPRINTF(PF_DEBUG_MISC, "%s: FILTER", __func__); break; /* XXX: rollback? */ } if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn)) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("%s: NAT\n", __func__)); + DPFPRINTF(PF_DEBUG_MISC, "%s: NAT", __func__); break; /* XXX: rollback? */ } if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn)) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("%s: BINAT\n", __func__)); + DPFPRINTF(PF_DEBUG_MISC, "%s: BINAT", __func__); break; /* XXX: rollback? */ } if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn)) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("%s: RDR\n", __func__)); + DPFPRINTF(PF_DEBUG_MISC, "%s: RDR", __func__); break; /* XXX: rollback? */ } @@ -6472,7 +6529,7 @@ shutdown_pf(void) break; if ((error = pf_begin_eth(&t[0], &nn)) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("%s: eth\n", __func__)); + DPFPRINTF(PF_DEBUG_MISC, "%s: eth", __func__); break; } error = pf_commit_eth(t[0], &nn); @@ -6480,7 +6537,7 @@ shutdown_pf(void) #ifdef ALTQ if ((error = pf_begin_altq(&t[0])) != 0) { - DPFPRINTF(PF_DEBUG_MISC, ("%s: ALTQ\n", __func__)); + DPFPRINTF(PF_DEBUG_MISC, "%s: ALTQ", __func__); break; } pf_commit_altq(t[0]); diff --git a/sys/netpfil/pf/pf_lb.c b/sys/netpfil/pf/pf_lb.c index 308d76c46e5b..ea0d6facf695 100644 --- a/sys/netpfil/pf/pf_lb.c +++ b/sys/netpfil/pf/pf_lb.c @@ -71,8 +71,6 @@ #define V_pf_rdr_srcport_rewrite_tries VNET(pf_rdr_srcport_rewrite_tries) VNET_DEFINE_STATIC(int, pf_rdr_srcport_rewrite_tries) = 16; -#define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x - static uint64_t pf_hash(struct pf_addr *, struct pf_addr *, struct pf_poolhashkey *, sa_family_t); struct pf_krule *pf_match_translation(int, struct pf_test_ctx *); @@ -80,7 +78,6 @@ static enum pf_test_status pf_step_into_translation_anchor(int, struct pf_test_c struct pf_krule *); static int pf_get_sport(struct pf_pdesc *, struct pf_krule *, struct pf_addr *, uint16_t *, uint16_t, uint16_t, - struct pf_ksrc_node **, struct pf_srchash **, struct pf_kpool *, struct pf_udp_mapping **, pf_sn_types_t); static bool pf_islinklocal(const sa_family_t, const struct pf_addr *); @@ -291,10 +288,8 @@ pf_match_translation(int rs_num, struct pf_test_ctx *ctx) } static int -pf_get_sport(struct pf_pdesc *pd, struct pf_krule *r, - struct pf_addr *naddr, uint16_t *nport, uint16_t low, - uint16_t high, struct pf_ksrc_node **sn, - struct pf_srchash **sh, struct pf_kpool *rpool, +pf_get_sport(struct pf_pdesc *pd, struct pf_krule *r, struct pf_addr *naddr, + uint16_t *nport, uint16_t low, uint16_t high, struct pf_kpool *rpool, struct pf_udp_mapping **udp_mapping, pf_sn_types_t sn_type) { struct pf_state_key_cmp key; @@ -322,19 +317,24 @@ pf_get_sport(struct pf_pdesc *pd, struct pf_krule *r, pf_addrcpy(&udp_source.addr, &pd->nsaddr, pd->af); udp_source.port = pd->nsport; if (udp_mapping) { + struct pf_ksrc_node *sn = NULL; + struct pf_srchash *sh = NULL; *udp_mapping = pf_udp_mapping_find(&udp_source); if (*udp_mapping) { pf_addrcpy(naddr, &(*udp_mapping)->endpoints[1].addr, pd->af); *nport = (*udp_mapping)->endpoints[1].port; - /* Try to find a src_node as per pf_map_addr(). */ - if (*sn == NULL && rpool->opts & PF_POOL_STICKYADDR && + /* + * Try to find a src_node as per pf_map_addr(). + * XXX: Why? This code seems to do nothing. + */ + if (rpool->opts & PF_POOL_STICKYADDR && (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) - *sn = pf_find_src_node(&pd->nsaddr, r, - pd->af, sh, sn_type, false); - if (*sn != NULL) - PF_SRC_NODE_UNLOCK(*sn); + sn = pf_find_src_node(&pd->nsaddr, r, + pd->af, &sh, sn_type, false); + if (sn != NULL) + PF_SRC_NODE_UNLOCK(sn); return (0); } else { *udp_mapping = pf_udp_mapping_create(pd->af, &pd->nsaddr, @@ -346,7 +346,7 @@ pf_get_sport(struct pf_pdesc *pd, struct pf_krule *r, } if (pf_map_addr_sn(pd->naf, r, &pd->nsaddr, naddr, NULL, &init_addr, - sn, sh, rpool, sn_type)) + rpool, sn_type)) goto failed; if (pd->proto == IPPROTO_ICMP) { @@ -470,9 +470,8 @@ pf_get_sport(struct pf_pdesc *pd, struct pf_krule *r, * pick a different source address since we're out * of free port choices for the current one. */ - (*sn) = NULL; if (pf_map_addr_sn(pd->naf, r, &pd->nsaddr, naddr, NULL, - &init_addr, sn, sh, rpool, sn_type)) + &init_addr, rpool, sn_type)) return (1); break; case PF_POOL_NONE: @@ -503,7 +502,6 @@ pf_islinklocal(const sa_family_t af, const struct pf_addr *addr) static int pf_get_mape_sport(struct pf_pdesc *pd, struct pf_krule *r, struct pf_addr *naddr, uint16_t *nport, - struct pf_ksrc_node **sn, struct pf_srchash **sh, struct pf_udp_mapping **udp_mapping, struct pf_kpool *rpool) { uint16_t psmask, low, highmask; @@ -523,16 +521,14 @@ pf_get_mape_sport(struct pf_pdesc *pd, struct pf_krule *r, for (i = cut; i <= ahigh; i++) { low = (i << ashift) | psmask; - if (!pf_get_sport(pd, r, - naddr, nport, low, low | highmask, sn, sh, rpool, - udp_mapping, PF_SN_NAT)) + if (!pf_get_sport(pd, r, naddr, nport, low, low | highmask, + rpool, udp_mapping, PF_SN_NAT)) return (0); } for (i = cut - 1; i > 0; i--) { low = (i << ashift) | psmask; - if (!pf_get_sport(pd, r, - naddr, nport, low, low | highmask, sn, sh, rpool, - udp_mapping, PF_SN_NAT)) + if (!pf_get_sport(pd, r, naddr, nport, low, low | highmask, + rpool, udp_mapping, PF_SN_NAT)) return (0); } return (1); @@ -545,6 +541,7 @@ pf_map_addr(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, { u_short reason = PFRES_MATCH; struct pf_addr *raddr = NULL, *rmask = NULL; + struct pfr_ktable *kt; uint64_t hashidx; int cnt; @@ -600,29 +597,25 @@ pf_map_addr(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, pf_poolmask(naddr, raddr, rmask, saddr, af); break; case PF_POOL_RANDOM: - if (rpool->cur->addr.type == PF_ADDR_TABLE) { - cnt = rpool->cur->addr.p.tbl->pfrkt_cnt; - if (cnt == 0) - rpool->tblidx = 0; + if (rpool->cur->addr.type == PF_ADDR_TABLE || + rpool->cur->addr.type == PF_ADDR_DYNIFTL) { + if (rpool->cur->addr.type == PF_ADDR_TABLE) + kt = rpool->cur->addr.p.tbl; else - rpool->tblidx = (int)arc4random_uniform(cnt); - memset(&rpool->counter, 0, sizeof(rpool->counter)); - if (pfr_pool_get(rpool->cur->addr.p.tbl, - &rpool->tblidx, &rpool->counter, af, NULL)) { + kt = rpool->cur->addr.p.dyn->pfid_kt; + kt = pfr_ktable_select_active(kt); + if (kt == NULL) { reason = PFRES_MAPFAILED; goto done_pool_mtx; /* unsupported */ } - pf_addrcpy(naddr, &rpool->counter, af); - } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { - cnt = rpool->cur->addr.p.dyn->pfid_kt->pfrkt_cnt; + cnt = kt->pfrkt_cnt; if (cnt == 0) rpool->tblidx = 0; else rpool->tblidx = (int)arc4random_uniform(cnt); memset(&rpool->counter, 0, sizeof(rpool->counter)); - if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, - &rpool->tblidx, &rpool->counter, af, - pf_islinklocal)) { + if (pfr_pool_get(kt, &rpool->tblidx, &rpool->counter, + af, pf_islinklocal, false)) { reason = PFRES_MAPFAILED; goto done_pool_mtx; /* unsupported */ } @@ -671,29 +664,25 @@ pf_map_addr(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, hashidx = pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af); - if (rpool->cur->addr.type == PF_ADDR_TABLE) { - cnt = rpool->cur->addr.p.tbl->pfrkt_cnt; - if (cnt == 0) - rpool->tblidx = 0; + if (rpool->cur->addr.type == PF_ADDR_TABLE || + rpool->cur->addr.type == PF_ADDR_DYNIFTL) { + if (rpool->cur->addr.type == PF_ADDR_TABLE) + kt = rpool->cur->addr.p.tbl; else - rpool->tblidx = (int)(hashidx % cnt); - memset(&rpool->counter, 0, sizeof(rpool->counter)); - if (pfr_pool_get(rpool->cur->addr.p.tbl, - &rpool->tblidx, &rpool->counter, af, NULL)) { + kt = rpool->cur->addr.p.dyn->pfid_kt; + kt = pfr_ktable_select_active(kt); + if (kt == NULL) { reason = PFRES_MAPFAILED; goto done_pool_mtx; /* unsupported */ } - pf_addrcpy(naddr, &rpool->counter, af); - } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { - cnt = rpool->cur->addr.p.dyn->pfid_kt->pfrkt_cnt; + cnt = kt->pfrkt_cnt; if (cnt == 0) rpool->tblidx = 0; else rpool->tblidx = (int)(hashidx % cnt); memset(&rpool->counter, 0, sizeof(rpool->counter)); - if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, - &rpool->tblidx, &rpool->counter, af, - pf_islinklocal)) { + if (pfr_pool_get(kt, &rpool->tblidx, &rpool->counter, + af, pf_islinklocal, false)) { reason = PFRES_MAPFAILED; goto done_pool_mtx; /* unsupported */ } @@ -710,11 +699,12 @@ pf_map_addr(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, if (rpool->cur->addr.type == PF_ADDR_TABLE) { if (!pfr_pool_get(rpool->cur->addr.p.tbl, - &rpool->tblidx, &rpool->counter, af, NULL)) + &rpool->tblidx, &rpool->counter, af, NULL, true)) goto get_addr; } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, - &rpool->tblidx, &rpool->counter, af, pf_islinklocal)) + &rpool->tblidx, &rpool->counter, af, pf_islinklocal, + true)) goto get_addr; } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af)) goto get_addr; @@ -724,9 +714,10 @@ pf_map_addr(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, rpool->cur = TAILQ_FIRST(&rpool->list); else rpool->cur = TAILQ_NEXT(rpool->cur, entries); + rpool->tblidx = -1; if (rpool->cur->addr.type == PF_ADDR_TABLE) { if (pfr_pool_get(rpool->cur->addr.p.tbl, - &rpool->tblidx, &rpool->counter, af, NULL)) { + &rpool->tblidx, &rpool->counter, af, NULL, true)) { /* table contains no address of type 'af' */ if (rpool->cur != acur) goto try_next; @@ -734,9 +725,9 @@ pf_map_addr(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, goto done_pool_mtx; } } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { - rpool->tblidx = -1; if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, - &rpool->tblidx, &rpool->counter, af, pf_islinklocal)) { + &rpool->tblidx, &rpool->counter, af, pf_islinklocal, + true)) { /* table contains no address of type 'af' */ if (rpool->cur != acur) goto try_next; @@ -764,48 +755,41 @@ pf_map_addr(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, done_pool_mtx: mtx_unlock(&rpool->mtx); - if (reason) { - counter_u64_add(V_pf_status.counters[reason], 1); - } - return (reason); } u_short pf_map_addr_sn(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, struct pf_addr *naddr, struct pfi_kkif **nkif, struct pf_addr *init_addr, - struct pf_ksrc_node **sn, struct pf_srchash **sh, struct pf_kpool *rpool, - pf_sn_types_t sn_type) + struct pf_kpool *rpool, pf_sn_types_t sn_type) { + struct pf_ksrc_node *sn = NULL; + struct pf_srchash *sh = NULL; u_short reason = 0; - KASSERT(*sn == NULL, ("*sn not NULL")); - /* * If this is a sticky-address rule, try to find an existing src_node. - * Request the sh to be unlocked if sn was not found, as we never - * insert a new sn when parsing the ruleset. */ if (rpool->opts & PF_POOL_STICKYADDR && (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) - *sn = pf_find_src_node(saddr, r, af, sh, sn_type, false); + sn = pf_find_src_node(saddr, r, af, &sh, sn_type, false); - if (*sn != NULL) { - PF_SRC_NODE_LOCK_ASSERT(*sn); + if (sn != NULL) { + PF_SRC_NODE_LOCK_ASSERT(sn); /* If the supplied address is the same as the current one we've * been asked before, so tell the caller that there's no other * address to be had. */ - if (PF_AEQ(naddr, &(*sn)->raddr, af)) { + if (PF_AEQ(naddr, &(sn->raddr), af)) { reason = PFRES_MAPFAILED; goto done; } - pf_addrcpy(naddr, &(*sn)->raddr, af); + pf_addrcpy(naddr, &(sn->raddr), af); if (nkif) - *nkif = (*sn)->rkif; + *nkif = sn->rkif; if (V_pf_status.debug >= PF_DEBUG_NOISY) { - printf("pf_map_addr: src tracking maps "); + printf("%s: src tracking maps ", __func__); pf_print_host(saddr, 0, af); printf(" to "); pf_print_host(naddr, 0, af); @@ -820,14 +804,16 @@ pf_map_addr_sn(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, * Source node has not been found. Find a new address and store it * in variables given by the caller. */ - if (pf_map_addr(af, r, saddr, naddr, nkif, init_addr, rpool) != 0) { - /* pf_map_addr() sets reason counters on its own */ + if ((reason = pf_map_addr(af, r, saddr, naddr, nkif, init_addr, + rpool)) != 0) { + if (V_pf_status.debug >= PF_DEBUG_MISC) + printf("%s: pf_map_addr has failed\n", __func__); goto done; } if (V_pf_status.debug >= PF_DEBUG_NOISY && (rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) { - printf("pf_map_addr: selected address "); + printf("%s: selected address ", __func__); pf_print_host(naddr, 0, af); if (nkif) printf("@%s", (*nkif)->pfik_name); @@ -835,12 +821,8 @@ pf_map_addr_sn(sa_family_t af, struct pf_krule *r, struct pf_addr *saddr, } done: - if ((*sn) != NULL) - PF_SRC_NODE_UNLOCK(*sn); - - if (reason) { - counter_u64_add(V_pf_status.counters[reason], 1); - } + if (sn != NULL) + PF_SRC_NODE_UNLOCK(sn); return (reason); } @@ -890,8 +872,6 @@ pf_get_transaddr(struct pf_test_ctx *ctx, struct pf_krule *r, { struct pf_pdesc *pd = ctx->pd; struct pf_addr *naddr; - struct pf_ksrc_node *sn = NULL; - struct pf_srchash *sh = NULL; uint16_t *nportp; uint16_t low, high; u_short reason; @@ -919,22 +899,22 @@ pf_get_transaddr(struct pf_test_ctx *ctx, struct pf_krule *r, high = rpool->proxy_port[1]; } if (rpool->mape.offset > 0) { - if (pf_get_mape_sport(pd, r, naddr, nportp, &sn, - &sh, &ctx->udp_mapping, rpool)) { + if (pf_get_mape_sport(pd, r, naddr, nportp, + &ctx->udp_mapping, rpool)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: MAP-E port allocation (%u/%u/%u)" - " failed\n", + "pf: MAP-E port allocation (%u/%u/%u)" + " failed", rpool->mape.offset, rpool->mape.psidlen, - rpool->mape.psid)); + rpool->mape.psid); reason = PFRES_MAPFAILED; goto notrans; } - } else if (pf_get_sport(pd, r, naddr, nportp, low, high, &sn, - &sh, rpool, &ctx->udp_mapping, PF_SN_NAT)) { + } else if (pf_get_sport(pd, r, naddr, nportp, low, high, + rpool, &ctx->udp_mapping, PF_SN_NAT)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: NAT proxy port allocation (%u-%u) failed\n", - rpool->proxy_port[0], rpool->proxy_port[1])); + "pf: NAT proxy port allocation (%u-%u) failed", + rpool->proxy_port[0], rpool->proxy_port[1]); reason = PFRES_MAPFAILED; goto notrans; } @@ -1017,7 +997,7 @@ pf_get_transaddr(struct pf_test_ctx *ctx, struct pf_krule *r, uint16_t cut, low, high, nport; reason = pf_map_addr_sn(pd->af, r, &pd->nsaddr, naddr, NULL, - NULL, &sn, &sh, rpool, PF_SN_NAT); + NULL, rpool, PF_SN_NAT); if (reason != 0) goto notrans; if ((rpool->opts & PF_POOL_TYPEMASK) == PF_POOL_BITMASK) @@ -1030,10 +1010,13 @@ pf_get_transaddr(struct pf_test_ctx *ctx, struct pf_krule *r, if (rpool->proxy_port[1]) { uint32_t tmp_nport; + uint16_t div; - tmp_nport = ((ntohs(pd->ndport) - ntohs(r->dst.port[0])) % - (rpool->proxy_port[1] - rpool->proxy_port[0] + - 1)) + rpool->proxy_port[0]; + div = r->rdr.proxy_port[1] - r->rdr.proxy_port[0] + 1; + div = (div == 0) ? 1 : div; + + tmp_nport = ((ntohs(pd->ndport) - ntohs(r->dst.port[0])) % div) + + rpool->proxy_port[0]; /* Wrap around if necessary. */ if (tmp_nport > 65535) @@ -1100,13 +1083,13 @@ pf_get_transaddr(struct pf_test_ctx *ctx, struct pf_krule *r, * the state may be reused if the TCP state is terminal. */ DPFPRINTF(PF_DEBUG_MISC, - ("pf: RDR source port allocation failed\n")); + "pf: RDR source port allocation failed"); break; out: DPFPRINTF(PF_DEBUG_MISC, - ("pf: RDR source port allocation %u->%u\n", - ntohs(pd->nsport), ntohs(ctx->nk->port[0]))); + "pf: RDR source port allocation %u->%u", + ntohs(pd->nsport), ntohs(ctx->nk->port[0])); break; } default: @@ -1134,8 +1117,6 @@ pf_get_transaddr_af(struct pf_krule *r, struct pf_pdesc *pd) struct pf_addr ndaddr, nsaddr, naddr; u_int16_t nport = 0; int prefixlen = 96; - struct pf_srchash *sh = NULL; - struct pf_ksrc_node *sns = NULL; bzero(&nsaddr, sizeof(nsaddr)); bzero(&ndaddr, sizeof(ndaddr)); @@ -1154,12 +1135,11 @@ pf_get_transaddr_af(struct pf_krule *r, struct pf_pdesc *pd) panic("pf_get_transaddr_af: no nat pool for source address"); /* get source address and port */ - if (pf_get_sport(pd, r, &nsaddr, &nport, - r->nat.proxy_port[0], r->nat.proxy_port[1], &sns, &sh, &r->nat, - NULL, PF_SN_NAT)) { + if (pf_get_sport(pd, r, &nsaddr, &nport, r->nat.proxy_port[0], + r->nat.proxy_port[1], &r->nat, NULL, PF_SN_NAT)) { DPFPRINTF(PF_DEBUG_MISC, - ("pf: af-to NAT proxy port allocation (%u-%u) failed", - r->nat.proxy_port[0], r->nat.proxy_port[1])); + "pf: af-to NAT proxy port allocation (%u-%u) failed", + r->nat.proxy_port[0], r->nat.proxy_port[1]); return (-1); } @@ -1182,7 +1162,7 @@ pf_get_transaddr_af(struct pf_krule *r, struct pf_pdesc *pd) /* get the destination address and port */ if (! TAILQ_EMPTY(&r->rdr.list)) { if (pf_map_addr_sn(pd->naf, r, &nsaddr, &naddr, NULL, NULL, - &sns, NULL, &r->rdr, PF_SN_NAT)) + &r->rdr, PF_SN_NAT)) return (-1); if (r->rdr.proxy_port[0]) pd->ndport = htons(r->rdr.proxy_port[0]); diff --git a/sys/netpfil/pf/pf_nl.c b/sys/netpfil/pf/pf_nl.c index d5d6dc70255e..73933c022ca2 100644 --- a/sys/netpfil/pf/pf_nl.c +++ b/sys/netpfil/pf/pf_nl.c @@ -1256,23 +1256,13 @@ pf_handle_clear_status(struct nlmsghdr *hdr, struct nl_pstate *npt) return (0); } -struct pf_nl_natlook { - sa_family_t af; - uint8_t direction; - uint8_t proto; - struct pf_addr src; - struct pf_addr dst; - uint16_t sport; - uint16_t dport; -}; - -#define _OUT(_field) offsetof(struct pf_nl_natlook, _field) +#define _OUT(_field) offsetof(struct pfioc_natlook, _field) static const struct nlattr_parser nla_p_natlook[] = { { .type = PF_NL_AF, .off = _OUT(af), .cb = nlattr_get_uint8 }, { .type = PF_NL_DIRECTION, .off = _OUT(direction), .cb = nlattr_get_uint8 }, { .type = PF_NL_PROTO, .off = _OUT(proto), .cb = nlattr_get_uint8 }, - { .type = PF_NL_SRC_ADDR, .off = _OUT(src), .cb = nlattr_get_in6_addr }, - { .type = PF_NL_DST_ADDR, .off = _OUT(dst), .cb = nlattr_get_in6_addr }, + { .type = PF_NL_SRC_ADDR, .off = _OUT(saddr), .cb = nlattr_get_in6_addr }, + { .type = PF_NL_DST_ADDR, .off = _OUT(daddr), .cb = nlattr_get_in6_addr }, { .type = PF_NL_SRC_PORT, .off = _OUT(sport), .cb = nlattr_get_uint16 }, { .type = PF_NL_DST_PORT, .off = _OUT(dport), .cb = nlattr_get_uint16 }, }; @@ -1282,63 +1272,31 @@ NL_DECLARE_PARSER(natlook_parser, struct genlmsghdr, nlf_p_empty, nla_p_natlook) static int pf_handle_natlook(struct nlmsghdr *hdr, struct nl_pstate *npt) { - struct pf_nl_natlook attrs = {}; - struct pf_state_key_cmp key = {}; + struct pfioc_natlook attrs = {}; struct nl_writer *nw = npt->nw; - struct pf_state_key *sk; - struct pf_kstate *state; struct genlmsghdr *ghdr_new; - int error, m = 0; - int sidx, didx; + int error; error = nl_parse_nlmsg(hdr, &natlook_parser, npt, &attrs); if (error != 0) return (error); - if (attrs.proto == 0 || - PF_AZERO(&attrs.src, attrs.af) || - PF_AZERO(&attrs.dst, attrs.af) || - ((attrs.proto == IPPROTO_TCP || attrs.proto == IPPROTO_UDP) && - (attrs.sport == 0 || attrs.dport == 0))) - return (EINVAL); - - /* NATLOOK src and dst are reversed, so reverse sidx/didx */ - sidx = (attrs.direction == PF_IN) ? 1 : 0; - didx = (attrs.direction == PF_IN) ? 0 : 1; - - key.af = attrs.af; - key.proto = attrs.proto; - pf_addrcpy(&key.addr[sidx], &attrs.src, attrs.af); - key.port[sidx] = attrs.sport; - pf_addrcpy(&key.addr[didx], &attrs.dst, attrs.af); - key.port[didx] = attrs.dport; - - state = pf_find_state_all(&key, attrs.direction, &m); - if (state == NULL) - return (ENOENT); - if (m > 1) { - PF_STATE_UNLOCK(state); - return (E2BIG); - } + error = pf_ioctl_natlook(&attrs); + if (error != 0) + return (error); - if (!nlmsg_reply(nw, hdr, sizeof(struct genlmsghdr))) { - PF_STATE_UNLOCK(state); + if (!nlmsg_reply(nw, hdr, sizeof(struct genlmsghdr))) return (ENOMEM); - } ghdr_new = nlmsg_reserve_object(nw, struct genlmsghdr); ghdr_new->cmd = PFNL_CMD_NATLOOK; ghdr_new->version = 0; ghdr_new->reserved = 0; - sk = state->key[sidx]; - - nlattr_add_in6_addr(nw, PF_NL_SRC_ADDR, &sk->addr[sidx].v6); - nlattr_add_in6_addr(nw, PF_NL_DST_ADDR, &sk->addr[didx].v6); - nlattr_add_u16(nw, PF_NL_SRC_PORT, sk->port[sidx]); - nlattr_add_u16(nw, PF_NL_DST_PORT, sk->port[didx]); - - PF_STATE_UNLOCK(state); + nlattr_add_in6_addr(nw, PF_NL_SRC_ADDR, &attrs.rsaddr.v6); + nlattr_add_in6_addr(nw, PF_NL_DST_ADDR, &attrs.rdaddr.v6); + nlattr_add_u16(nw, PF_NL_SRC_PORT, attrs.rsport); + nlattr_add_u16(nw, PF_NL_DST_PORT, attrs.rdport); if (!nlmsg_end(nw)) { nlmsg_abort(nw); diff --git a/sys/netpfil/pf/pf_norm.c b/sys/netpfil/pf/pf_norm.c index 369292ca365e..8cea9557633c 100644 --- a/sys/netpfil/pf/pf_norm.c +++ b/sys/netpfil/pf/pf_norm.c @@ -160,13 +160,6 @@ static int pf_reassemble6(struct mbuf **, struct ip6_frag *, uint16_t, uint16_t, u_short *); #endif /* INET6 */ -#define DPFPRINTF(x) do { \ - if (V_pf_status.debug >= PF_DEBUG_MISC) { \ - printf("%s: ", __func__); \ - printf x ; \ - } \ -} while(0) - #ifdef INET static void pf_ip2key(struct ip *ip, struct pf_frnode *key) @@ -262,7 +255,8 @@ pf_purge_fragments(uint32_t expire) if (frag->fr_timeout > expire) break; - DPFPRINTF(("expiring %d(%p)\n", frag->fr_id, frag)); + DPFPRINTF(PF_DEBUG_MISC, "expiring %d(%p)", + frag->fr_id, frag); pf_free_fragment(frag); } @@ -281,7 +275,7 @@ pf_flush_fragments(void) PF_FRAG_ASSERT(); goal = uma_zone_get_cur(V_pf_frent_z) * 9 / 10; - DPFPRINTF(("trying to free %d frag entriess\n", goal)); + DPFPRINTF(PF_DEBUG_MISC, "trying to free %d frag entriess", goal); while (goal < uma_zone_get_cur(V_pf_frent_z)) { frag = TAILQ_LAST(&V_pf_fragqueue, pf_fragqueue); if (frag) @@ -573,26 +567,30 @@ pf_fillup_fragment(struct pf_frnode *key, uint32_t id, /* No empty fragments. */ if (frent->fe_len == 0) { - DPFPRINTF(("bad fragment: len 0\n")); + DPFPRINTF(PF_DEBUG_MISC, "bad fragment: len 0"); goto bad_fragment; } /* All fragments are 8 byte aligned. */ if (frent->fe_mff && (frent->fe_len & 0x7)) { - DPFPRINTF(("bad fragment: mff and len %d\n", frent->fe_len)); + DPFPRINTF(PF_DEBUG_MISC, "bad fragment: mff and len %d", + frent->fe_len); goto bad_fragment; } /* Respect maximum length, IP_MAXPACKET == IPV6_MAXPACKET. */ if (frent->fe_off + frent->fe_len > IP_MAXPACKET) { - DPFPRINTF(("bad fragment: max packet %d\n", - frent->fe_off + frent->fe_len)); + DPFPRINTF(PF_DEBUG_MISC, "bad fragment: max packet %d", + frent->fe_off + frent->fe_len); goto bad_fragment; } - DPFPRINTF((key->fn_af == AF_INET ? - "reass frag %d @ %d-%d\n" : "reass frag %#08x @ %d-%d\n", - id, frent->fe_off, frent->fe_off + frent->fe_len)); + if (key->fn_af == AF_INET) + DPFPRINTF(PF_DEBUG_MISC, "reass frag %d @ %d-%d\n", + id, frent->fe_off, frent->fe_off + frent->fe_len); + else + DPFPRINTF(PF_DEBUG_MISC, "reass frag %#08x @ %d-%d", + id, frent->fe_off, frent->fe_off + frent->fe_len); /* Fully buffer all of the fragments in this fragment queue. */ frag = pf_find_fragment(key, id); @@ -690,10 +688,10 @@ pf_fillup_fragment(struct pf_frnode *key, uint32_t id, precut = prev->fe_off + prev->fe_len - frent->fe_off; if (precut >= frent->fe_len) { - DPFPRINTF(("new frag overlapped\n")); + DPFPRINTF(PF_DEBUG_MISC, "new frag overlapped"); goto drop_fragment; } - DPFPRINTF(("frag head overlap %d\n", precut)); + DPFPRINTF(PF_DEBUG_MISC, "frag head overlap %d", precut); m_adj(frent->fe_m, precut); frent->fe_off += precut; frent->fe_len -= precut; @@ -705,7 +703,8 @@ pf_fillup_fragment(struct pf_frnode *key, uint32_t id, aftercut = frent->fe_off + frent->fe_len - after->fe_off; if (aftercut < after->fe_len) { - DPFPRINTF(("frag tail overlap %d", aftercut)); + DPFPRINTF(PF_DEBUG_MISC, "frag tail overlap %d", + aftercut); m_adj(after->fe_m, aftercut); /* Fragment may switch queue as fe_off changes */ pf_frent_remove(frag, after); @@ -713,7 +712,8 @@ pf_fillup_fragment(struct pf_frnode *key, uint32_t id, after->fe_len -= aftercut; /* Insert into correct queue */ if (pf_frent_insert(frag, after, prev)) { - DPFPRINTF(("fragment requeue limit exceeded")); + DPFPRINTF(PF_DEBUG_MISC, + "fragment requeue limit exceeded"); m_freem(after->fe_m); uma_zfree(V_pf_frent_z, after); /* There is not way to recover */ @@ -723,7 +723,7 @@ pf_fillup_fragment(struct pf_frnode *key, uint32_t id, } /* This fragment is completely overlapped, lose it. */ - DPFPRINTF(("old frag overlapped\n")); + DPFPRINTF(PF_DEBUG_MISC, "old frag overlapped"); next = TAILQ_NEXT(after, fr_next); pf_frent_remove(frag, after); m_freem(after->fe_m); @@ -732,7 +732,7 @@ pf_fillup_fragment(struct pf_frnode *key, uint32_t id, /* If part of the queue gets too long, there is not way to recover. */ if (pf_frent_insert(frag, frent, prev)) { - DPFPRINTF(("fragment queue limit exceeded\n")); + DPFPRINTF(PF_DEBUG_MISC, "fragment queue limit exceeded"); goto bad_fragment; } @@ -748,7 +748,7 @@ free_fragment: * fragment, the entire datagram (and any constituent fragments) MUST * be silently discarded. */ - DPFPRINTF(("flush overlapping fragments\n")); + DPFPRINTF(PF_DEBUG_MISC, "flush overlapping fragments"); pf_free_fragment(frag); bad_fragment: @@ -826,7 +826,8 @@ pf_reassemble(struct mbuf **m0, u_short *reason) m = *m0 = NULL; if (frag->fr_holes) { - DPFPRINTF(("frag %d, holes %d\n", frag->fr_id, frag->fr_holes)); + DPFPRINTF(PF_DEBUG_MISC, "frag %d, holes %d", + frag->fr_id, frag->fr_holes); return (PF_PASS); /* drop because *m0 is NULL, no error */ } @@ -872,14 +873,14 @@ pf_reassemble(struct mbuf **m0, u_short *reason) ip->ip_off &= ~(IP_MF|IP_OFFMASK); if (hdrlen + total > IP_MAXPACKET) { - DPFPRINTF(("drop: too big: %d\n", total)); + DPFPRINTF(PF_DEBUG_MISC, "drop: too big: %d", total); ip->ip_len = 0; REASON_SET(reason, PFRES_SHORT); /* PF_DROP requires a valid mbuf *m0 in pf_test() */ return (PF_DROP); } - DPFPRINTF(("complete: %p(%d)\n", m, ntohs(ip->ip_len))); + DPFPRINTF(PF_DEBUG_MISC, "complete: %p(%d)", m, ntohs(ip->ip_len)); return (PF_PASS); } #endif /* INET */ @@ -931,8 +932,8 @@ pf_reassemble6(struct mbuf **m0, struct ip6_frag *fraghdr, m = *m0 = NULL; if (frag->fr_holes) { - DPFPRINTF(("frag %d, holes %d\n", frag->fr_id, - frag->fr_holes)); + DPFPRINTF(PF_DEBUG_MISC, "frag %d, holes %d", frag->fr_id, + frag->fr_holes); PF_FRAG_UNLOCK(); return (PF_PASS); /* Drop because *m0 is NULL, no error. */ } @@ -993,14 +994,15 @@ pf_reassemble6(struct mbuf **m0, struct ip6_frag *fraghdr, ip6->ip6_nxt = proto; if (hdrlen - sizeof(struct ip6_hdr) + total > IPV6_MAXPACKET) { - DPFPRINTF(("drop: too big: %d\n", total)); + DPFPRINTF(PF_DEBUG_MISC, "drop: too big: %d", total); ip6->ip6_plen = 0; REASON_SET(reason, PFRES_SHORT); /* PF_DROP requires a valid mbuf *m0 in pf_test6(). */ return (PF_DROP); } - DPFPRINTF(("complete: %p(%d)\n", m, ntohs(ip6->ip6_plen))); + DPFPRINTF(PF_DEBUG_MISC, "complete: %p(%d)", m, + ntohs(ip6->ip6_plen)); return (PF_PASS); fail: @@ -1090,7 +1092,7 @@ pf_refragment6(struct ifnet *ifp, struct mbuf **m0, struct m_tag *mtag, action = PF_PASS; } else { /* Drop expects an mbuf to free. */ - DPFPRINTF(("refragment error %d\n", error)); + DPFPRINTF(PF_DEBUG_MISC, "refragment error %d", error); action = PF_DROP; } for (; m; m = t) { @@ -1230,7 +1232,7 @@ pf_normalize_ip(u_short *reason, struct pf_pdesc *pd) * no-df above, fine. Otherwise drop it. */ if (h->ip_off & htons(IP_DF)) { - DPFPRINTF(("IP_DF\n")); + DPFPRINTF(PF_DEBUG_MISC, "IP_DF"); goto bad; } @@ -1238,13 +1240,13 @@ pf_normalize_ip(u_short *reason, struct pf_pdesc *pd) /* All fragments are 8 byte aligned */ if (mff && (ip_len & 0x7)) { - DPFPRINTF(("mff and %d\n", ip_len)); + DPFPRINTF(PF_DEBUG_MISC, "mff and %d", ip_len); goto bad; } /* Respect maximum length */ if (fragoff + ip_len > IP_MAXPACKET) { - DPFPRINTF(("max packet %d\n", fragoff + ip_len)); + DPFPRINTF(PF_DEBUG_MISC, "max packet %d", fragoff + ip_len); goto bad; } @@ -1256,7 +1258,8 @@ pf_normalize_ip(u_short *reason, struct pf_pdesc *pd) /* Fully buffer all of the fragments * Might return a completely reassembled mbuf, or NULL */ PF_FRAG_LOCK(); - DPFPRINTF(("reass frag %d @ %d-%d\n", h->ip_id, fragoff, max)); + DPFPRINTF(PF_DEBUG_MISC, "reass frag %d @ %d-%d", + h->ip_id, fragoff, max); verdict = pf_reassemble(&pd->m, reason); PF_FRAG_UNLOCK(); @@ -1282,7 +1285,7 @@ pf_normalize_ip(u_short *reason, struct pf_pdesc *pd) return (PF_PASS); bad: - DPFPRINTF(("dropping bad fragment\n")); + DPFPRINTF(PF_DEBUG_MISC, "dropping bad fragment"); REASON_SET(reason, PFRES_FRAG); drop: if (r != NULL && r->log) @@ -1711,7 +1714,7 @@ pf_normalize_tcp_stateful(struct pf_pdesc *pd, (uptime.tv_sec - src->scrub->pfss_last.tv_sec > TS_MAX_IDLE || time_uptime - (state->creation / 1000) > TS_MAX_CONN)) { if (V_pf_status.debug >= PF_DEBUG_MISC) { - DPFPRINTF(("src idled out of PAWS\n")); + DPFPRINTF(PF_DEBUG_MISC, "src idled out of PAWS"); pf_print_state(state); printf("\n"); } @@ -1721,7 +1724,7 @@ pf_normalize_tcp_stateful(struct pf_pdesc *pd, if (dst->scrub && (dst->scrub->pfss_flags & PFSS_PAWS) && uptime.tv_sec - dst->scrub->pfss_last.tv_sec > TS_MAX_IDLE) { if (V_pf_status.debug >= PF_DEBUG_MISC) { - DPFPRINTF(("dst idled out of PAWS\n")); + DPFPRINTF(PF_DEBUG_MISC, "dst idled out of PAWS"); pf_print_state(state); printf("\n"); } @@ -1826,22 +1829,22 @@ pf_normalize_tcp_stateful(struct pf_pdesc *pd, * an old timestamp. */ - DPFPRINTF(("Timestamp failed %c%c%c%c\n", + DPFPRINTF(PF_DEBUG_MISC, "Timestamp failed %c%c%c%c", SEQ_LT(tsval, dst->scrub->pfss_tsecr) ? '0' : ' ', SEQ_GT(tsval, src->scrub->pfss_tsval + tsval_from_last) ? '1' : ' ', SEQ_GT(tsecr, dst->scrub->pfss_tsval) ? '2' : ' ', - SEQ_LT(tsecr, dst->scrub->pfss_tsval0)? '3' : ' ')); - DPFPRINTF((" tsval: %u tsecr: %u +ticks: %u " - "idle: %jus %lums\n", + SEQ_LT(tsecr, dst->scrub->pfss_tsval0)? '3' : ' '); + DPFPRINTF(PF_DEBUG_MISC, " tsval: %u tsecr: %u +ticks: " + "%u idle: %jus %lums", tsval, tsecr, tsval_from_last, (uintmax_t)delta_ts.tv_sec, - delta_ts.tv_usec / 1000)); - DPFPRINTF((" src->tsval: %u tsecr: %u\n", - src->scrub->pfss_tsval, src->scrub->pfss_tsecr)); - DPFPRINTF((" dst->tsval: %u tsecr: %u tsval0: %u" - "\n", dst->scrub->pfss_tsval, - dst->scrub->pfss_tsecr, dst->scrub->pfss_tsval0)); + delta_ts.tv_usec / 1000); + DPFPRINTF(PF_DEBUG_MISC, " src->tsval: %u tsecr: %u", + src->scrub->pfss_tsval, src->scrub->pfss_tsecr); + DPFPRINTF(PF_DEBUG_MISC, " dst->tsval: %u tsecr: %u " + "tsval0: %u", dst->scrub->pfss_tsval, + dst->scrub->pfss_tsecr, dst->scrub->pfss_tsval0); if (V_pf_status.debug >= PF_DEBUG_MISC) { pf_print_state(state); pf_print_flags(tcp_get_flags(th)); @@ -1891,8 +1894,8 @@ pf_normalize_tcp_stateful(struct pf_pdesc *pd, * stack changed its RFC1323 behavior?!?! */ if (V_pf_status.debug >= PF_DEBUG_MISC) { - DPFPRINTF(("Did not receive expected RFC1323 " - "timestamp\n")); + DPFPRINTF(PF_DEBUG_MISC, "Did not receive expected " + "RFC1323 timestamp"); pf_print_state(state); pf_print_flags(tcp_get_flags(th)); printf("\n"); @@ -1919,9 +1922,9 @@ pf_normalize_tcp_stateful(struct pf_pdesc *pd, if (V_pf_status.debug >= PF_DEBUG_MISC && dst->scrub && (dst->scrub->pfss_flags & PFSS_TIMESTAMP)) { /* Don't warn if other host rejected RFC1323 */ - DPFPRINTF(("Broken RFC1323 stack did not " - "timestamp data packet. Disabled PAWS " - "security.\n")); + DPFPRINTF(PF_DEBUG_MISC, "Broken RFC1323 stack did " + "not timestamp data packet. Disabled PAWS " + "security."); pf_print_state(state); pf_print_flags(tcp_get_flags(th)); printf("\n"); diff --git a/sys/netpfil/pf/pf_osfp.c b/sys/netpfil/pf/pf_osfp.c index 3e00cc7c80a2..150626c5f3fb 100644 --- a/sys/netpfil/pf/pf_osfp.c +++ b/sys/netpfil/pf/pf_osfp.c @@ -40,9 +40,6 @@ #endif static MALLOC_DEFINE(M_PFOSFP, "pf_osfp", "pf(4) operating system fingerprints"); -#define DPFPRINTF(format, x...) \ - if (V_pf_status.debug >= PF_DEBUG_NOISY) \ - printf(format , ##x) SLIST_HEAD(pf_osfp_list, pf_os_fingerprint); VNET_DEFINE_STATIC(struct pf_osfp_list, pf_osfp_list) = @@ -189,8 +186,8 @@ pf_osfp_fingerprint_hdr(const struct ip *ip, const struct ip6_hdr *ip6, const st optlen = MAX(optlen, 1); /* paranoia */ } - DPFPRINTF("fingerprinted %s:%d %d:%d:%d:%d:%llx (%d) " - "(TS=%s,M=%s%d,W=%s%d)\n", + DPFPRINTF(PF_DEBUG_NOISY, "fingerprinted %s:%d %d:%d:%d:%d:%llx (%d) " + "(TS=%s,M=%s%d,W=%s%d)", srcname, ntohs(tcp->th_sport), fp.fp_wsize, fp.fp_ttl, (fp.fp_flags & PF_OSFP_DF) != 0, fp.fp_psize, (long long int)fp.fp_tcpopts, fp.fp_optcnt, @@ -219,7 +216,7 @@ pf_osfp_match(struct pf_osfp_enlist *list, pf_osfp_t os) if (os == PF_OSFP_ANY) return (1); if (list == NULL) { - DPFPRINTF("osfp no match against %x\n", os); + DPFPRINTF(PF_DEBUG_NOISY, "osfp no match against %x", os); return (os == PF_OSFP_UNKNOWN); } PF_OSFP_UNPACK(os, os_class, os_version, os_subtype); @@ -228,13 +225,13 @@ pf_osfp_match(struct pf_osfp_enlist *list, pf_osfp_t os) if ((os_class == PF_OSFP_ANY || en_class == os_class) && (os_version == PF_OSFP_ANY || en_version == os_version) && (os_subtype == PF_OSFP_ANY || en_subtype == os_subtype)) { - DPFPRINTF("osfp matched %s %s %s %x==%x\n", + DPFPRINTF(PF_DEBUG_NOISY, "osfp matched %s %s %s %x==%x", entry->fp_class_nm, entry->fp_version_nm, entry->fp_subtype_nm, os, entry->fp_os); return (1); } } - DPFPRINTF("fingerprint 0x%x didn't match\n", os); + DPFPRINTF(PF_DEBUG_NOISY, "fingerprint 0x%x didn't match", os); return (0); } @@ -275,8 +272,8 @@ pf_osfp_add(struct pf_osfp_ioctl *fpioc) fpadd.fp_ttl = fpioc->fp_ttl; #if 0 /* XXX RYAN wants to fix logging */ - DPFPRINTF("adding osfp %s %s %s = %s%d:%d:%d:%s%d:0x%llx %d " - "(TS=%s,M=%s%d,W=%s%d) %x\n", + DPFPRINTF(PF_DEBUG_NOISY, "adding osfp %s %s %s =" + " %s%d:%d:%d:%s%d:0x%llx %d (TS=%s,M=%s%d,W=%s%d) %x", fpioc->fp_os.fp_class_nm, fpioc->fp_os.fp_version_nm, fpioc->fp_os.fp_subtype_nm, (fpadd.fp_flags & PF_OSFP_WSIZE_MOD) ? "%" : diff --git a/sys/netpfil/pf/pf_ruleset.c b/sys/netpfil/pf/pf_ruleset.c index 2e5165a9900c..43b51f2933f4 100644 --- a/sys/netpfil/pf/pf_ruleset.c +++ b/sys/netpfil/pf/pf_ruleset.c @@ -59,9 +59,6 @@ #error "Kernel only file. Please use sbin/pfctl/pf_ruleset.c instead." #endif -#define DPFPRINTF(format, x...) \ - if (V_pf_status.debug >= PF_DEBUG_NOISY) \ - printf(format , ##x) #define rs_malloc(x) malloc(x, M_TEMP, M_NOWAIT|M_ZERO) #define rs_free(x) free(x, M_TEMP) @@ -386,7 +383,8 @@ pf_kanchor_setup(struct pf_krule *r, const struct pf_kruleset *s, strlcpy(path, s->anchor->path, MAXPATHLEN); while (name[0] == '.' && name[1] == '.' && name[2] == '/') { if (!path[0]) { - DPFPRINTF("%s: .. beyond root\n", __func__); + DPFPRINTF(PF_DEBUG_NOISY, "%s: .. beyond root", + __func__); rs_free(path); return (1); } @@ -408,7 +406,7 @@ pf_kanchor_setup(struct pf_krule *r, const struct pf_kruleset *s, ruleset = pf_find_or_create_kruleset(path); rs_free(path); if (ruleset == NULL || ruleset == &pf_main_ruleset) { - DPFPRINTF("%s: ruleset\n", __func__); + DPFPRINTF(PF_DEBUG_NOISY, "%s: ruleset", __func__); return (1); } r->anchor = ruleset->anchor; @@ -690,7 +688,8 @@ pf_keth_anchor_setup(struct pf_keth_rule *r, const struct pf_keth_ruleset *s, strlcpy(path, s->anchor->path, MAXPATHLEN); while (name[0] == '.' && name[1] == '.' && name[2] == '/') { if (!path[0]) { - DPFPRINTF("%s: .. beyond root\n", __func__); + DPFPRINTF(PF_DEBUG_NOISY, "%s: .. beyond root", + __func__); rs_free(path); return (1); } @@ -712,7 +711,7 @@ pf_keth_anchor_setup(struct pf_keth_rule *r, const struct pf_keth_ruleset *s, ruleset = pf_find_or_create_keth_ruleset(path); rs_free(path); if (ruleset == NULL || ruleset->anchor == NULL) { - DPFPRINTF("%s: ruleset\n", __func__); + DPFPRINTF(PF_DEBUG_NOISY, "%s: ruleset", __func__); return (1); } r->anchor = ruleset->anchor; diff --git a/sys/netpfil/pf/pf_syncookies.c b/sys/netpfil/pf/pf_syncookies.c index 66757fa4b756..4a935bc65767 100644 --- a/sys/netpfil/pf/pf_syncookies.c +++ b/sys/netpfil/pf/pf_syncookies.c @@ -88,8 +88,6 @@ #include <net/pfvar.h> #include <netpfil/pf/pf_nv.h> -#define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x - union pf_syncookie { uint8_t cookie; struct { @@ -281,7 +279,7 @@ pf_synflood_check(struct pf_pdesc *pd) pf_syncookie_rotate, curvnet); V_pf_status.syncookies_active = true; DPFPRINTF(LOG_WARNING, - ("synflood detected, enabling syncookies\n")); + "synflood detected, enabling syncookies"); // XXXTODO V_pf_status.lcounters[LCNT_SYNFLOODS]++; } @@ -367,7 +365,7 @@ pf_syncookie_rotate(void *arg) V_pf_status.syncookies_mode == PF_SYNCOOKIES_NEVER) ) { V_pf_status.syncookies_active = false; - DPFPRINTF(PF_DEBUG_MISC, ("syncookies disabled\n")); + DPFPRINTF(PF_DEBUG_MISC, "syncookies disabled"); } /* nothing in flight any more? delete keys and return */ diff --git a/sys/netpfil/pf/pf_table.c b/sys/netpfil/pf/pf_table.c index 43e4366845a2..ecc185f89ad7 100644 --- a/sys/netpfil/pf/pf_table.c +++ b/sys/netpfil/pf/pf_table.c @@ -49,8 +49,6 @@ #include <net/vnet.h> #include <net/pfvar.h> -#define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x - #define ACCEPT_FLAGS(flags, oklist) \ do { \ if ((flags & ~(oklist)) & \ @@ -819,10 +817,10 @@ pfr_create_kentry(struct pfr_addr *ad, bool counters) static void pfr_destroy_kentries(struct pfr_kentryworkq *workq) { - struct pfr_kentry *p, *q; + struct pfr_kentry *p; - for (p = SLIST_FIRST(workq); p != NULL; p = q) { - q = SLIST_NEXT(p, pfrke_workq); + while ((p = SLIST_FIRST(workq)) != NULL) { + SLIST_REMOVE_HEAD(workq, pfrke_workq); pfr_destroy_kentry(p); } } @@ -1680,8 +1678,7 @@ pfr_ina_commit(struct pfr_table *trs, u_int32_t ticket, int *nadd, } if (!(flags & PFR_FLAG_DUMMY)) { - for (p = SLIST_FIRST(&workq); p != NULL; p = q) { - q = SLIST_NEXT(p, pfrkt_workq); + SLIST_FOREACH_SAFE(p, &workq, pfrkt_workq, q) { pfr_commit_ktable(p, tzero); } rs->topen = 0; @@ -1710,7 +1707,7 @@ pfr_commit_ktable(struct pfr_ktable *kt, time_t tzero) } else if (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) { /* kt might contain addresses */ struct pfr_kentryworkq addrq, addq, changeq, delq, garbageq; - struct pfr_kentry *p, *q, *next; + struct pfr_kentry *p, *q; struct pfr_addr ad; pfr_enqueue_addrs(shadow, &addrq, NULL, 0); @@ -1720,7 +1717,8 @@ pfr_commit_ktable(struct pfr_ktable *kt, time_t tzero) SLIST_INIT(&delq); SLIST_INIT(&garbageq); pfr_clean_node_mask(shadow, &addrq); - SLIST_FOREACH_SAFE(p, &addrq, pfrke_workq, next) { + while ((p = SLIST_FIRST(&addrq)) != NULL) { + SLIST_REMOVE_HEAD(&addrq, pfrke_workq); pfr_copyout_addr(&ad, p); q = pfr_lookup_addr(kt, &ad, 1); if (q != NULL) { @@ -1864,8 +1862,7 @@ pfr_setflags_ktables(struct pfr_ktableworkq *workq) { struct pfr_ktable *p, *q; - for (p = SLIST_FIRST(workq); p; p = q) { - q = SLIST_NEXT(p, pfrkt_workq); + SLIST_FOREACH_SAFE(p, workq, pfrkt_workq, q) { pfr_setflags_ktable(p, p->pfrkt_nflags); } } @@ -2015,10 +2012,10 @@ pfr_create_ktable(struct pfr_table *tbl, time_t tzero, int attachruleset) static void pfr_destroy_ktables(struct pfr_ktableworkq *workq, int flushaddr) { - struct pfr_ktable *p, *q; + struct pfr_ktable *p; - for (p = SLIST_FIRST(workq); p; p = q) { - q = SLIST_NEXT(p, pfrkt_workq); + while ((p = SLIST_FIRST(workq)) != NULL) { + SLIST_REMOVE_HEAD(workq, pfrkt_workq); pfr_destroy_ktable(p, flushaddr); } } @@ -2074,17 +2071,16 @@ pfr_lookup_table(struct pfr_table *tbl) (struct pfr_ktable *)tbl)); } -int -pfr_match_addr(struct pfr_ktable *kt, struct pf_addr *a, sa_family_t af) +static struct pfr_kentry * +pfr_kentry_byaddr(struct pfr_ktable *kt, struct pf_addr *a, sa_family_t af, + int exact) { struct pfr_kentry *ke = NULL; - int match; PF_RULES_RASSERT(); - if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) - kt = kt->pfrkt_root; - if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE)) + kt = pfr_ktable_select_active(kt); + if (kt == NULL) return (0); switch (af) { @@ -2121,11 +2117,26 @@ pfr_match_addr(struct pfr_ktable *kt, struct pf_addr *a, sa_family_t af) default: unhandled_af(af); } + if (exact && ke && KENTRY_NETWORK(ke)) + ke = NULL; + + return (ke); +} + +int +pfr_match_addr(struct pfr_ktable *kt, struct pf_addr *a, sa_family_t af) +{ + struct pfr_kentry *ke = NULL; + int match; + + ke = pfr_kentry_byaddr(kt, a, af, 0); + match = (ke && !ke->pfrke_not); if (match) pfr_kstate_counter_add(&kt->pfrkt_match, 1); else pfr_kstate_counter_add(&kt->pfrkt_nomatch, 1); + return (match); } @@ -2135,9 +2146,8 @@ pfr_update_stats(struct pfr_ktable *kt, struct pf_addr *a, sa_family_t af, { struct pfr_kentry *ke = NULL; - if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) - kt = kt->pfrkt_root; - if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE)) + kt = pfr_ktable_select_active(kt); + if (kt == NULL) return; switch (af) { @@ -2177,7 +2187,7 @@ pfr_update_stats(struct pfr_ktable *kt, struct pf_addr *a, sa_family_t af, if ((ke == NULL || ke->pfrke_not) != notrule) { if (op_pass != PFR_OP_PASS) DPFPRINTF(PF_DEBUG_URGENT, - ("pfr_update_stats: assertion failed.\n")); + "pfr_update_stats: assertion failed."); op_pass = PFR_OP_XPASS; } pfr_kstate_counter_add(&kt->pfrkt_packets[dir_out][op_pass], 1); @@ -2281,7 +2291,7 @@ pfr_detach_table(struct pfr_ktable *kt) int pfr_pool_get(struct pfr_ktable *kt, int *pidx, struct pf_addr *counter, - sa_family_t af, pf_addr_filter_func_t filter) + sa_family_t af, pf_addr_filter_func_t filter, bool loop_once) { struct pf_addr *addr, cur, mask, umask_addr; union sockaddr_union uaddr, umask; @@ -2306,9 +2316,8 @@ pfr_pool_get(struct pfr_ktable *kt, int *pidx, struct pf_addr *counter, unhandled_af(af); } - if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) - kt = kt->pfrkt_root; - if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE)) + kt = pfr_ktable_select_active(kt); + if (kt == NULL) return (-1); idx = *pidx; @@ -2327,7 +2336,7 @@ _next_block: ke = pfr_kentry_byidx(kt, idx, af); if (ke == NULL) { /* we don't have this idx, try looping */ - if (loop || (ke = pfr_kentry_byidx(kt, 0, af)) == NULL) { + if ((loop || loop_once) || (ke = pfr_kentry_byidx(kt, 0, af)) == NULL) { pfr_kstate_counter_add(&kt->pfrkt_nomatch, 1); return (1); } @@ -2455,3 +2464,14 @@ pfr_dynaddr_update(struct pfr_ktable *kt, struct pfi_dynaddr *dyn) unhandled_af(dyn->pfid_af); } } + +struct pfr_ktable * +pfr_ktable_select_active(struct pfr_ktable *kt) +{ + if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL) + kt = kt->pfrkt_root; + if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE)) + return (NULL); + + return (kt); +} diff --git a/sys/netsmb/smb_conn.c b/sys/netsmb/smb_conn.c index 259635e2d8d5..ab6cd130a057 100644 --- a/sys/netsmb/smb_conn.c +++ b/sys/netsmb/smb_conn.c @@ -422,7 +422,7 @@ smb_vc_create(struct smb_vcspec *vcspec, if (uid == SMBM_ANY_OWNER) uid = realuid; if (gid == SMBM_ANY_GROUP) - gid = cred->cr_groups[0]; + gid = cred->cr_gid; vcp->vc_uid = uid; vcp->vc_grp = gid; @@ -765,7 +765,7 @@ smb_share_create(struct smb_vc *vcp, struct smb_sharespec *shspec, if (uid == SMBM_ANY_OWNER) uid = realuid; if (gid == SMBM_ANY_GROUP) - gid = cred->cr_groups[0]; + gid = cred->cr_gid; ssp = smb_zmalloc(sizeof(*ssp), M_SMBCONN, M_WAITOK); smb_co_init(SSTOCP(ssp), SMBL_SHARE, "smbss ilock", "smbss"); ssp->obj.co_free = smb_share_free; diff --git a/sys/powerpc/aim/mmu_oea.c b/sys/powerpc/aim/mmu_oea.c index 7746b668265d..ae17b3289593 100644 --- a/sys/powerpc/aim/mmu_oea.c +++ b/sys/powerpc/aim/mmu_oea.c @@ -1469,6 +1469,9 @@ moea_page_set_memattr(vm_page_t m, vm_memattr_t ma) pmap_t pmap; u_int lo; + if (m->md.mdpg_cache_attrs == ma) + return; + if ((m->oflags & VPO_UNMANAGED) != 0) { m->md.mdpg_cache_attrs = ma; return; diff --git a/sys/powerpc/aim/mmu_oea64.c b/sys/powerpc/aim/mmu_oea64.c index 79cea408bb5f..796b1719b8ba 100644 --- a/sys/powerpc/aim/mmu_oea64.c +++ b/sys/powerpc/aim/mmu_oea64.c @@ -2134,6 +2134,9 @@ moea64_page_set_memattr(vm_page_t m, vm_memattr_t ma) CTR3(KTR_PMAP, "%s: pa=%#jx, ma=%#x", __func__, (uintmax_t)VM_PAGE_TO_PHYS(m), ma); + if (m->md.mdpg_cache_attrs == ma) + return; + if ((m->oflags & VPO_UNMANAGED) != 0) { m->md.mdpg_cache_attrs = ma; return; diff --git a/sys/powerpc/aim/mmu_radix.c b/sys/powerpc/aim/mmu_radix.c index 45f7bef8bcc9..a12142fc2d7b 100644 --- a/sys/powerpc/aim/mmu_radix.c +++ b/sys/powerpc/aim/mmu_radix.c @@ -5937,6 +5937,10 @@ mmu_radix_page_set_memattr(vm_page_t m, vm_memattr_t ma) { CTR3(KTR_PMAP, "%s(%p, %#x)", __func__, m, ma); + + if (m->md.mdpg_cache_attrs == ma) + return; + m->md.mdpg_cache_attrs = ma; /* diff --git a/sys/powerpc/include/pcb.h b/sys/powerpc/include/pcb.h index 050ada6b0f64..0230cf78aba7 100644 --- a/sys/powerpc/include/pcb.h +++ b/sys/powerpc/include/pcb.h @@ -66,16 +66,8 @@ struct pcb { #define PCB_VECREGS 0x200 /* Process had Altivec registers initialized */ struct fpu { union { -#if _BYTE_ORDER == _BIG_ENDIAN - double fpr; - uint32_t vsr[4]; -#else uint32_t vsr[4]; - struct { - double padding; - double fpr; - }; -#endif + double fpr; } fpr[32]; double fpscr; /* FPSCR stored as double for easier access */ } pcb_fpu; /* Floating point processor */ diff --git a/sys/powerpc/include/ucontext.h b/sys/powerpc/include/ucontext.h index d35c6c773fe0..dc87edd578bc 100644 --- a/sys/powerpc/include/ucontext.h +++ b/sys/powerpc/include/ucontext.h @@ -41,6 +41,7 @@ typedef struct __mcontext { int mc_flags; #define _MC_FP_VALID 0x01 #define _MC_AV_VALID 0x02 +#define _MC_VS_VALID 0x04 int mc_onstack; /* saved onstack flag */ int mc_len; /* sizeof(__mcontext) */ __uint64_t mc_avec[32*2]; /* vector register file */ @@ -56,6 +57,7 @@ typedef struct __mcontext32 { int mc_flags; #define _MC_FP_VALID 0x01 #define _MC_AV_VALID 0x02 +#define _MC_VS_VALID 0x04 int mc_onstack; /* saved onstack flag */ int mc_len; /* sizeof(__mcontext) */ uint64_t mc_avec[32*2]; /* vector register file */ diff --git a/sys/powerpc/powerpc/exec_machdep.c b/sys/powerpc/powerpc/exec_machdep.c index 1893d79f29a8..8a33d0f589a7 100644 --- a/sys/powerpc/powerpc/exec_machdep.c +++ b/sys/powerpc/powerpc/exec_machdep.c @@ -214,10 +214,10 @@ sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask) sfpsize = sizeof(sf); #ifdef __powerpc64__ /* - * 64-bit PPC defines a 288 byte scratch region - * below the stack. + * 64-bit PPC defines a 512 byte red zone below + * the existing stack (ELF ABI v2 §2.2.2.4) */ - rndfsize = 288 + roundup(sizeof(sf), 48); + rndfsize = 512 + roundup(sizeof(sf), 48); #else rndfsize = roundup(sizeof(sf), 16); #endif @@ -349,13 +349,6 @@ sys_sigreturn(struct thread *td, struct sigreturn_args *uap) if (error != 0) return (error); - /* - * Save FPU state if needed. User may have changed it on - * signal handler - */ - if (uc.uc_mcontext.mc_srr1 & PSL_FP) - save_fpu(td); - kern_sigprocmask(td, SIG_SETMASK, &uc.uc_sigmask, NULL, 0); CTR3(KTR_SIG, "sigreturn: return td=%p pc=%#x sp=%#x", @@ -432,6 +425,7 @@ grab_mcontext(struct thread *td, mcontext_t *mcp, int flags) } if (pcb->pcb_flags & PCB_VSX) { + mcp->mc_flags |= _MC_VS_VALID; for (i = 0; i < 32; i++) memcpy(&mcp->mc_vsxfpreg[i], &pcb->pcb_fpu.fpr[i].vsr[2], sizeof(double)); @@ -481,6 +475,7 @@ set_mcontext(struct thread *td, mcontext_t *mcp) struct pcb *pcb; struct trapframe *tf; register_t tls; + register_t msr; int i; pcb = td->td_pcb; @@ -531,6 +526,22 @@ set_mcontext(struct thread *td, mcontext_t *mcp) tf->srr1 &= ~(PSL_FP | PSL_VSX | PSL_VEC); pcb->pcb_flags &= ~(PCB_FPU | PCB_VSX | PCB_VEC); + /* + * Ensure the FPU is also disabled in hardware. + * + * Without this, it's possible for the register reload to fail if we + * don't switch to a FPU disabled context before resuming the original + * thread. Specifically, if the FPU/VSX unavailable exception is never + * hit, then whatever data is still in the FP/VSX registers when + * sigresume is callled will used by the resumed thread, instead of the + * previously saved data from the mcontext. + */ + critical_enter(); + msr = mfmsr() & ~(PSL_FP | PSL_VSX | PSL_VEC); + isync(); + mtmsr(msr); + critical_exit(); + if (mcp->mc_flags & _MC_FP_VALID) { /* enable_fpu() will happen lazily on a fault */ pcb->pcb_flags |= PCB_FPREGS; @@ -539,8 +550,12 @@ set_mcontext(struct thread *td, mcontext_t *mcp) for (i = 0; i < 32; i++) { memcpy(&pcb->pcb_fpu.fpr[i].fpr, &mcp->mc_fpreg[i], sizeof(double)); - memcpy(&pcb->pcb_fpu.fpr[i].vsr[2], - &mcp->mc_vsxfpreg[i], sizeof(double)); + } + if (mcp->mc_flags & _MC_VS_VALID) { + for (i = 0; i < 32; i++) { + memcpy(&pcb->pcb_fpu.fpr[i].vsr[2], + &mcp->mc_vsxfpreg[i], sizeof(double)); + } } } diff --git a/sys/powerpc/powerpc/fpu.c b/sys/powerpc/powerpc/fpu.c index 0eaff2ea4932..cc8f22f7dda3 100644 --- a/sys/powerpc/powerpc/fpu.c +++ b/sys/powerpc/powerpc/fpu.c @@ -64,8 +64,19 @@ save_fpu_int(struct thread *td) * Save the floating-point registers and FPSCR to the PCB */ if (pcb->pcb_flags & PCB_VSX) { - #define SFP(n) __asm ("stxvw4x " #n ", 0,%0" \ +#if _BYTE_ORDER == _BIG_ENDIAN + #define SFP(n) __asm("stxvw4x " #n ", 0,%0" \ :: "b"(&pcb->pcb_fpu.fpr[n])); +#else + /* + * stxvw2x will swap words within the FP double word on LE systems, + * leading to corruption if VSX is used to store state and FP is + * subsequently used to restore state. + * Use stxvd2x instead. + */ + #define SFP(n) __asm("stxvd2x " #n ", 0,%0" \ + :: "b"(&pcb->pcb_fpu.fpr[n])); +#endif SFP(0); SFP(1); SFP(2); SFP(3); SFP(4); SFP(5); SFP(6); SFP(7); SFP(8); SFP(9); SFP(10); SFP(11); @@ -76,7 +87,7 @@ save_fpu_int(struct thread *td) SFP(28); SFP(29); SFP(30); SFP(31); #undef SFP } else { - #define SFP(n) __asm ("stfd " #n ", 0(%0)" \ + #define SFP(n) __asm("stfd " #n ", 0(%0)" \ :: "b"(&pcb->pcb_fpu.fpr[n].fpr)); SFP(0); SFP(1); SFP(2); SFP(3); SFP(4); SFP(5); SFP(6); SFP(7); @@ -149,8 +160,19 @@ enable_fpu(struct thread *td) :: "b"(&pcb->pcb_fpu.fpscr)); if (pcb->pcb_flags & PCB_VSX) { - #define LFP(n) __asm ("lxvw4x " #n ", 0,%0" \ +#if _BYTE_ORDER == _BIG_ENDIAN + #define LFP(n) __asm("lxvw4x " #n ", 0,%0" \ + :: "b"(&pcb->pcb_fpu.fpr[n])); +#else + /* + * lxvw4x will swap words within the FP double word on LE systems, + * leading to corruption if FP is used to store state and VSX is + * subsequently used to restore state. + * Use lxvd2x instead. + */ + #define LFP(n) __asm("lxvd2x " #n ", 0,%0" \ :: "b"(&pcb->pcb_fpu.fpr[n])); +#endif LFP(0); LFP(1); LFP(2); LFP(3); LFP(4); LFP(5); LFP(6); LFP(7); LFP(8); LFP(9); LFP(10); LFP(11); @@ -161,7 +183,7 @@ enable_fpu(struct thread *td) LFP(28); LFP(29); LFP(30); LFP(31); #undef LFP } else { - #define LFP(n) __asm ("lfd " #n ", 0(%0)" \ + #define LFP(n) __asm("lfd " #n ", 0(%0)" \ :: "b"(&pcb->pcb_fpu.fpr[n].fpr)); LFP(0); LFP(1); LFP(2); LFP(3); LFP(4); LFP(5); LFP(6); LFP(7); diff --git a/sys/riscv/allwinner/files.allwinner b/sys/riscv/allwinner/files.allwinner index 73fa9660e2d2..7a4ff6b9c62e 100644 --- a/sys/riscv/allwinner/files.allwinner +++ b/sys/riscv/allwinner/files.allwinner @@ -1,5 +1,6 @@ arm/allwinner/aw_gpio.c optional gpio aw_gpio fdt +arm/allwinner/aw_mmc.c optional mmc aw_mmc fdt | mmccam aw_mmc fdt arm/allwinner/aw_rtc.c optional aw_rtc fdt arm/allwinner/aw_syscon.c optional syscon arm/allwinner/aw_sid.c optional aw_sid nvmem diff --git a/sys/riscv/conf/std.allwinner b/sys/riscv/conf/std.allwinner index 2b1e0d4e09dc..34fe195b01ba 100644 --- a/sys/riscv/conf/std.allwinner +++ b/sys/riscv/conf/std.allwinner @@ -7,6 +7,7 @@ options SOC_ALLWINNER_D1 device aw_ccu # Allwinner clock controller device aw_gpio # Allwinner GPIO controller +device aw_mmc # Allwinner SD/MMC controller device aw_rtc # Allwinner Real-time Clock device aw_sid # Allwinner Secure ID EFUSE device aw_timer # Allwinner Timer diff --git a/sys/riscv/include/vmm_dev.h b/sys/riscv/include/vmm_dev.h index 856ff0778b95..4d30d5a1c35b 100644 --- a/sys/riscv/include/vmm_dev.h +++ b/sys/riscv/include/vmm_dev.h @@ -34,6 +34,8 @@ #ifndef _VMM_DEV_H_ #define _VMM_DEV_H_ +#include <sys/domainset.h> + #include <machine/vmm.h> struct vm_memmap { @@ -56,6 +58,9 @@ struct vm_memseg { int segid; size_t len; char name[VM_MAX_SUFFIXLEN + 1]; + domainset_t *ds_mask; + size_t ds_mask_size; + int ds_policy; }; struct vm_register { diff --git a/sys/riscv/riscv/pmap.c b/sys/riscv/riscv/pmap.c index 5d15bd671285..26efaecc64d1 100644 --- a/sys/riscv/riscv/pmap.c +++ b/sys/riscv/riscv/pmap.c @@ -4838,6 +4838,8 @@ pmap_unmapbios(void *p, vm_size_t size) void pmap_page_set_memattr(vm_page_t m, vm_memattr_t ma) { + if (m->md.pv_memattr == ma) + return; m->md.pv_memattr = ma; diff --git a/sys/rpc/authunix_prot.c b/sys/rpc/authunix_prot.c index 91fb96f44397..7b531946488a 100644 --- a/sys/rpc/authunix_prot.c +++ b/sys/rpc/authunix_prot.c @@ -93,9 +93,10 @@ xdr_authunix_parms(XDR *xdrs, uint32_t *time, struct xucred *cred) if (!xdr_uint32_t(xdrs, &cred->cr_uid)) return (FALSE); - if (!xdr_uint32_t(xdrs, &cred->cr_groups[0])) + if (!xdr_uint32_t(xdrs, &cred->cr_gid)) return (FALSE); + /* XXXKE Fix this is cr_gid gets separated out. */ if (xdrs->x_op == XDR_ENCODE) { ngroups = cred->cr_ngroups - 1; if (ngroups > NGRPS) @@ -105,7 +106,7 @@ xdr_authunix_parms(XDR *xdrs, uint32_t *time, struct xucred *cred) if (!xdr_uint32_t(xdrs, &ngroups)) return (FALSE); for (i = 0; i < ngroups; i++) { - if (i + 1 < ngroups_max + 1) { + if (i < ngroups_max) { if (!xdr_uint32_t(xdrs, &cred->cr_groups[i + 1])) return (FALSE); } else { @@ -115,7 +116,7 @@ xdr_authunix_parms(XDR *xdrs, uint32_t *time, struct xucred *cred) } if (xdrs->x_op == XDR_DECODE) { - if (ngroups + 1 > ngroups_max + 1) + if (ngroups > ngroups_max) cred->cr_ngroups = ngroups_max + 1; else cred->cr_ngroups = ngroups + 1; diff --git a/sys/rpc/clnt_rc.c b/sys/rpc/clnt_rc.c index 9e87af578885..44b63e38a8e6 100644 --- a/sys/rpc/clnt_rc.c +++ b/sys/rpc/clnt_rc.c @@ -198,6 +198,12 @@ clnt_reconnect_connect(CLIENT *cl) newclient = clnt_vc_create(so, (struct sockaddr *) &rc->rc_addr, rc->rc_prog, rc->rc_vers, rc->rc_sendsz, rc->rc_recvsz, rc->rc_intr); + /* + * CLSET_FD_CLOSE must be done now, in case rpctls_connect() + * fails just below. + */ + if (newclient != NULL) + CLNT_CONTROL(newclient, CLSET_FD_CLOSE, 0); if (rc->rc_tls && newclient != NULL) { CURVNET_SET(so->so_vnet); stat = rpctls_connect(newclient, rc->rc_tlscertname, so, @@ -236,7 +242,6 @@ clnt_reconnect_connect(CLIENT *cl) goto out; } - CLNT_CONTROL(newclient, CLSET_FD_CLOSE, 0); CLNT_CONTROL(newclient, CLSET_CONNECT, &one); CLNT_CONTROL(newclient, CLSET_TIMEOUT, &rc->rc_timeout); CLNT_CONTROL(newclient, CLSET_RETRY_TIMEOUT, &rc->rc_retry); diff --git a/sys/rpc/rpcsec_gss/rpcsec_gss.c b/sys/rpc/rpcsec_gss/rpcsec_gss.c index 62c71937a185..983dd251f81f 100644 --- a/sys/rpc/rpcsec_gss/rpcsec_gss.c +++ b/sys/rpc/rpcsec_gss/rpcsec_gss.c @@ -67,6 +67,7 @@ #include <sys/param.h> #include <sys/systm.h> #include <sys/hash.h> +#include <sys/jail.h> #include <sys/kernel.h> #include <sys/kobj.h> #include <sys/lock.h> @@ -772,6 +773,17 @@ rpc_gss_init(AUTH *auth, rpc_gss_options_ret_t *options_ret) gd->gd_cred.gc_seq = 0; /* + * XXX Threads from inside jails can get here via calls + * to clnt_vc_call()->AUTH_REFRESH()->rpc_gss_refresh() + * but the NFS mount is always done outside of the + * jails in vnet0. Since the thread credentials won't + * necessarily have cr_prison == vnet0 and this function + * has no access to the socket, using vnet0 seems the + * only option. This is broken if NFS mounts are enabled + * within vnet prisons. + */ + KGSS_CURVNET_SET_QUIET(vnet0); + /* * For KerberosV, if there is a client principal name, that implies * that this is a host based initiator credential in the default * keytab file. For this case, it is necessary to do a @@ -994,12 +1006,14 @@ out: gss_delete_sec_context(&min_stat, &gd->gd_ctx, GSS_C_NO_BUFFER); } + KGSS_CURVNET_RESTORE(); mtx_lock(&gd->gd_lock); gd->gd_state = RPCSEC_GSS_START; wakeup(gd); mtx_unlock(&gd->gd_lock); return (FALSE); } + KGSS_CURVNET_RESTORE(); mtx_lock(&gd->gd_lock); gd->gd_state = RPCSEC_GSS_ESTABLISHED; diff --git a/sys/rpc/rpcsec_tls/rpctls_impl.c b/sys/rpc/rpcsec_tls/rpctls_impl.c index 93fe283e65fd..51fe270b13d9 100644 --- a/sys/rpc/rpcsec_tls/rpctls_impl.c +++ b/sys/rpc/rpcsec_tls/rpctls_impl.c @@ -240,6 +240,14 @@ rpctls_rpc_failed(struct upsock *ups, struct socket *so) * failed to do the handshake. */ mtx_unlock(&rpctls_lock); + /* + * Do a shutdown on the socket, since the daemon is + * probably stuck in SSL_accept() or SSL_connect() trying to + * read the socket. Do not soclose() the socket, since the + * daemon will close() the socket after SSL_accept() + * returns an error. + */ + soshutdown(so, SHUT_RD); } } diff --git a/sys/rpc/svc_auth_unix.c b/sys/rpc/svc_auth_unix.c index 5d6402a05006..b10ef33be704 100644 --- a/sys/rpc/svc_auth_unix.c +++ b/sys/rpc/svc_auth_unix.c @@ -83,12 +83,13 @@ _svcauth_unix(struct svc_req *rqst, struct rpc_msg *msg) str_len = RNDUP(str_len); buf += str_len / sizeof (int32_t); xcr->cr_uid = IXDR_GET_UINT32(buf); - xcr->cr_groups[0] = IXDR_GET_UINT32(buf); + xcr->cr_gid = IXDR_GET_UINT32(buf); gid_len = (size_t)IXDR_GET_UINT32(buf); if (gid_len > NGRPS) { stat = AUTH_BADCRED; goto done; } + /* XXXKE Fix this if cr_gid gets separated out. */ for (i = 0; i < gid_len; i++) { if (i + 1 < XU_NGROUPS) xcr->cr_groups[i + 1] = IXDR_GET_INT32(buf); diff --git a/sys/security/audit/audit.c b/sys/security/audit/audit.c index 05928f1c33e8..7ec50d990d4e 100644 --- a/sys/security/audit/audit.c +++ b/sys/security/audit/audit.c @@ -279,7 +279,7 @@ audit_record_ctor(void *mem, int size, void *arg, int flags) cru2x(cred, &ar->k_ar.ar_subj_cred); ar->k_ar.ar_subj_ruid = cred->cr_ruid; ar->k_ar.ar_subj_rgid = cred->cr_rgid; - ar->k_ar.ar_subj_egid = cred->cr_groups[0]; + ar->k_ar.ar_subj_egid = cred->cr_gid; ar->k_ar.ar_subj_auid = cred->cr_audit.ai_auid; ar->k_ar.ar_subj_asid = cred->cr_audit.ai_asid; ar->k_ar.ar_subj_pid = td->td_proc->p_pid; diff --git a/sys/security/audit/audit_arg.c b/sys/security/audit/audit_arg.c index c667d3968817..3ea645373dbe 100644 --- a/sys/security/audit/audit_arg.c +++ b/sys/security/audit/audit_arg.c @@ -408,7 +408,7 @@ audit_arg_process(struct proc *p) cred = p->p_ucred; ar->k_ar.ar_arg_auid = cred->cr_audit.ai_auid; ar->k_ar.ar_arg_euid = cred->cr_uid; - ar->k_ar.ar_arg_egid = cred->cr_groups[0]; + ar->k_ar.ar_arg_egid = cred->cr_gid; ar->k_ar.ar_arg_ruid = cred->cr_ruid; ar->k_ar.ar_arg_rgid = cred->cr_rgid; ar->k_ar.ar_arg_asid = cred->cr_audit.ai_asid; diff --git a/sys/sys/compressor.h b/sys/sys/compressor.h index cad9080b46ff..e59eeabec2cd 100644 --- a/sys/sys/compressor.h +++ b/sys/sys/compressor.h @@ -42,6 +42,7 @@ struct compressor; bool compressor_avail(int format); struct compressor *compressor_init(compressor_cb_t cb, int format, size_t maxiosize, int level, void *arg); +int compressor_format(const struct compressor *stream); void compressor_reset(struct compressor *stream); int compressor_write(struct compressor *stream, void *data, size_t len); diff --git a/sys/sys/domainset.h b/sys/sys/domainset.h index f98b175e9bc8..f3dc92ec6383 100644 --- a/sys/sys/domainset.h +++ b/sys/sys/domainset.h @@ -113,6 +113,20 @@ void domainset_zero(void); * returned value will not match the key pointer. */ struct domainset *domainset_create(const struct domainset *); + +/* + * Remove empty domains from a given domainset. + * Returns 'false' if the domainset consists entirely of empty domains. + */ +bool domainset_empty_vm(struct domainset *domain); + +/* + * Validate and populate a domainset structure according to the specified + * policy and mask. + */ +int domainset_populate(struct domainset *domain, const domainset_t *mask, int policy, + size_t mask_size); + #ifdef _SYS_SYSCTL_H_ int sysctl_handle_domainset(SYSCTL_HANDLER_ARGS); #endif diff --git a/sys/sys/efi.h b/sys/sys/efi.h index 95a433a950db..89c8b15519de 100644 --- a/sys/sys/efi.h +++ b/sys/sys/efi.h @@ -42,6 +42,8 @@ {0xb122a263,0x3661,0x4f68,{0x99,0x29,0x78,0xf8,0xb0,0xd6,0x21,0x80}} #define EFI_PROPERTIES_TABLE \ {0x880aaca3,0x4adc,0x4a04,{0x90,0x79,0xb7,0x47,0x34,0x08,0x25,0xe5}} +#define EFI_MEMORY_ATTRIBUTES_TABLE \ + {0xdcfa911d,0x26eb,0x469f,{0xa2,0x20,0x38,0xb7,0xdc,0x46,0x12,0x20}} #define LINUX_EFI_MEMRESERVE_TABLE \ {0x888eb0c6,0x8ede,0x4ff5,{0xa8,0xf0,0x9a,0xee,0x5c,0xb9,0x77,0xc2}} @@ -166,6 +168,22 @@ struct efi_prop_table { uint64_t memory_protection_attribute; }; +struct efi_memory_descriptor { + uint32_t type; + caddr_t phy_addr; + caddr_t virt_addr; + uint64_t pages; + uint64_t attrs; +}; + +struct efi_memory_attribute_table { + uint32_t version; + uint32_t num_ents; + uint32_t descriptor_size; + uint32_t flags; + struct efi_memory_descriptor tables[]; +}; + #ifdef _KERNEL #ifdef EFIABI_ATTR diff --git a/sys/sys/elf_common.h b/sys/sys/elf_common.h index 87460aae2dd4..efda38279848 100644 --- a/sys/sys/elf_common.h +++ b/sys/sys/elf_common.h @@ -306,7 +306,7 @@ typedef struct { and MPRC of Peking University */ #define EM_AARCH64 183 /* AArch64 (64-bit ARM) */ #define EM_RISCV 243 /* RISC-V */ -#define EM_LOONGARCH 258 /* Loongson LoongArch */ +#define EM_LOONGARCH 258 /* Loongson LoongArch */ /* Non-standard or deprecated. */ #define EM_486 6 /* Intel i486. */ @@ -392,15 +392,15 @@ typedef struct { */ /* LoongArch Base ABI Modifiers */ -#define EF_LOONGARCH_ABI_SOFT_FLOAT 0x00000001 -#define EF_LOONGARCH_ABI_SINGLE_FLOAT 0x00000002 -#define EF_LOONGARCH_ABI_DOUBLE_FLOAT 0x00000003 -#define EF_LOONGARCH_ABI_MODIFIER_MASK 0x00000007 +#define EF_LOONGARCH_ABI_SOFT_FLOAT 0x00000001 +#define EF_LOONGARCH_ABI_SINGLE_FLOAT 0x00000002 +#define EF_LOONGARCH_ABI_DOUBLE_FLOAT 0x00000003 +#define EF_LOONGARCH_ABI_MODIFIER_MASK 0x00000007 /* LoongArch Object file ABI versions */ -#define EF_LOONGARCH_OBJABI_V0 0x00000000 -#define EF_LOONGARCH_OBJABI_V1 0x00000040 -#define EF_LOONGARCH_OBJABI_MASK 0x000000C0 +#define EF_LOONGARCH_OBJABI_V0 0x00000000 +#define EF_LOONGARCH_OBJABI_V1 0x00000040 +#define EF_LOONGARCH_OBJABI_MASK 0x000000C0 #define EF_SPARC_EXT_MASK 0x00ffff00 #define EF_SPARC_32PLUS 0x00000100 @@ -470,12 +470,12 @@ typedef struct { #define SHT_HIOS 0x6fffffff /* Last of OS specific semantics */ #define SHT_LOPROC 0x70000000 /* reserved range for processor */ #define SHT_X86_64_UNWIND 0x70000001 /* unwind information */ -#define SHT_AMD64_UNWIND SHT_X86_64_UNWIND +#define SHT_AMD64_UNWIND SHT_X86_64_UNWIND #define SHT_ARM_EXIDX 0x70000001 /* Exception index table. */ -#define SHT_ARM_PREEMPTMAP 0x70000002 /* BPABI DLL dynamic linking +#define SHT_ARM_PREEMPTMAP 0x70000002 /* BPABI DLL dynamic linking pre-emption map. */ -#define SHT_ARM_ATTRIBUTES 0x70000003 /* Object file compatibility +#define SHT_ARM_ATTRIBUTES 0x70000003 /* Object file compatibility attributes. */ #define SHT_ARM_DEBUGOVERLAY 0x70000004 /* See DBGOVL for details. */ #define SHT_ARM_OVERLAYSECTION 0x70000005 /* See DBGOVL for details. */ @@ -791,7 +791,7 @@ typedef struct { #define DF_1_NODELETE 0x00000008 /* Set the RTLD_NODELETE for object */ #define DF_1_LOADFLTR 0x00000010 /* Immediate loading of filtees */ #define DF_1_INITFIRST 0x00000020 /* Initialize DSO first at runtime */ -#define DF_1_NOOPEN 0x00000040 /* Do not allow loading on dlopen() */ +#define DF_1_NOOPEN 0x00000040 /* Do not allow loading on dlopen() */ #define DF_1_ORIGIN 0x00000080 /* Process $ORIGIN */ #define DF_1_INTERPOSE 0x00000400 /* Interpose all objects but main */ #define DF_1_NODEFLIB 0x00000800 /* Do not search default paths */ @@ -908,7 +908,7 @@ typedef struct { #define STV_ELIMINATE 0x6 /* Architecture specific data - st_other */ -#define STO_AARCH64_VARIANT_PCS 0x80 +#define STO_AARCH64_VARIANT_PCS 0x80 /* Special symbol table indexes. */ #define STN_UNDEF 0 /* Undefined symbol index. */ @@ -1084,11 +1084,11 @@ typedef struct { #define R_AARCH64_COPY 1024 /* Copy data from shared object */ #define R_AARCH64_GLOB_DAT 1025 /* Set GOT entry to data address */ #define R_AARCH64_JUMP_SLOT 1026 /* Set GOT entry to code address */ -#define R_AARCH64_RELATIVE 1027 /* Add load address of shared object */ +#define R_AARCH64_RELATIVE 1027 /* Add load address of shared object */ #define R_AARCH64_TLS_DTPREL64 1028 #define R_AARCH64_TLS_DTPMOD64 1029 -#define R_AARCH64_TLS_TPREL64 1030 -#define R_AARCH64_TLSDESC 1031 /* Identify the TLS descriptor */ +#define R_AARCH64_TLS_TPREL64 1030 +#define R_AARCH64_TLSDESC 1031 /* Identify the TLS descriptor */ #define R_AARCH64_IRELATIVE 1032 #define R_ARM_NONE 0 /* No relocation. */ @@ -1231,8 +1231,8 @@ typedef struct { #define R_MIPS_GOT_HI16 22 /* GOT HI 16 bit */ #define R_MIPS_GOT_LO16 23 /* GOT LO 16 bit */ #define R_MIPS_SUB 24 -#define R_MIPS_CALLHI16 30 /* upper 16 bit GOT entry for function */ -#define R_MIPS_CALLLO16 31 /* lower 16 bit GOT entry for function */ +#define R_MIPS_CALLHI16 30 /* upper 16 bit GOT entry for function */ +#define R_MIPS_CALLLO16 31 /* lower 16 bit GOT entry for function */ #define R_MIPS_JALR 37 #define R_MIPS_TLS_GD 42 #define R_MIPS_COPY 126 @@ -1352,7 +1352,6 @@ typedef struct { * RISC-V relocation types. */ -/* Relocation types used by the dynamic linker. */ #define R_RISCV_NONE 0 #define R_RISCV_32 1 #define R_RISCV_64 2 @@ -1365,8 +1364,7 @@ typedef struct { #define R_RISCV_TLS_DTPREL64 9 #define R_RISCV_TLS_TPREL32 10 #define R_RISCV_TLS_TPREL64 11 - -/* Relocation types not used by the dynamic linker. */ +#define R_RISCV_TLSDESC 12 #define R_RISCV_BRANCH 16 #define R_RISCV_JAL 17 #define R_RISCV_CALL 18 @@ -1392,10 +1390,10 @@ typedef struct { #define R_RISCV_SUB16 38 #define R_RISCV_SUB32 39 #define R_RISCV_SUB64 40 +#define R_RISCV_GOT32_PCREL 41 #define R_RISCV_ALIGN 43 #define R_RISCV_RVC_BRANCH 44 #define R_RISCV_RVC_JUMP 45 -#define R_RISCV_RVC_LUI 46 #define R_RISCV_RELAX 51 #define R_RISCV_SUB6 52 #define R_RISCV_SET6 53 @@ -1404,6 +1402,14 @@ typedef struct { #define R_RISCV_SET32 56 #define R_RISCV_32_PCREL 57 #define R_RISCV_IRELATIVE 58 +#define R_RISCV_PLT32 59 +#define R_RISCV_SET_ULEB128 60 +#define R_RISCV_SUB_ULEB128 61 +#define R_RISCV_TLSDESC_HI20 62 +#define R_RISCV_TLSDESC_LOAD_LO12 63 +#define R_RISCV_TLSDESC_ADD_LO12 64 +#define R_RISCV_TLSDESC_CALL 65 +#define R_RISCV_VENDOR 191 /* * Loongson LoongArch relocation types. @@ -1413,101 +1419,101 @@ typedef struct { */ /* Relocation types used by the dynamic linker */ -#define R_LARCH_NONE 0 -#define R_LARCH_32 1 -#define R_LARCH_64 2 -#define R_LARCH_RELATIVE 3 -#define R_LARCH_COPY 4 -#define R_LARCH_JUMP_SLOT 5 -#define R_LARCH_TLS_DTPMOD32 6 -#define R_LARCH_TLS_DTPMOD64 7 -#define R_LARCH_TLS_DTPREL32 8 -#define R_LARCH_TLS_DTPREL64 9 -#define R_LARCH_TLS_TPREL32 10 -#define R_LARCH_TLS_TPREL64 11 -#define R_LARCH_IRELATIVE 12 -#define R_LARCH_MARK_LA 20 -#define R_LARCH_MARK_PCREL 21 -#define R_LARCH_SOP_PUSH_PCREL 22 -#define R_LARCH_SOP_PUSH_ABSOLUTE 23 -#define R_LARCH_SOP_PUSH_DUP 24 -#define R_LARCH_SOP_PUSH_GPREL 25 -#define R_LARCH_SOP_PUSH_TLS_TPREL 26 -#define R_LARCH_SOP_PUSH_TLS_GOT 27 -#define R_LARCH_SOP_PUSH_TLS_GD 28 -#define R_LARCH_SOP_PUSH_PLT_PCREL 29 -#define R_LARCH_SOP_ASSERT 30 -#define R_LARCH_SOP_NOT 31 -#define R_LARCH_SOP_SUB 32 -#define R_LARCH_SOP_SL 33 -#define R_LARCH_SOP_SR 34 -#define R_LARCH_SOP_ADD 35 -#define R_LARCH_SOP_AND 36 -#define R_LARCH_SOP_IF_ELSE 37 -#define R_LARCH_SOP_POP_32_S_10_5 38 -#define R_LARCH_SOP_POP_32_U_10_12 39 -#define R_LARCH_SOP_POP_32_S_10_12 40 -#define R_LARCH_SOP_POP_32_S_10_16 41 -#define R_LARCH_SOP_POP_32_S_10_16_S2 42 -#define R_LARCH_SOP_POP_32_S_5_20 43 -#define R_LARCH_SOP_POP_32_S_0_5_10_16_S2 44 -#define R_LARCH_SOP_POP_32_S_0_10_10_16_S2 45 -#define R_LARCH_SOP_POP_32_U 46 -#define R_LARCH_ADD8 47 -#define R_LARCH_ADD16 48 -#define R_LARCH_ADD24 49 -#define R_LARCH_ADD32 50 -#define R_LARCH_ADD64 51 -#define R_LARCH_SUB8 52 -#define R_LARCH_SUB16 53 -#define R_LARCH_SUB24 54 -#define R_LARCH_SUB32 55 -#define R_LARCH_SUB64 56 -#define R_LARCH_GNU_VTINHERIT 57 -#define R_LARCH_GNU_VTENTRY 58 +#define R_LARCH_NONE 0 +#define R_LARCH_32 1 +#define R_LARCH_64 2 +#define R_LARCH_RELATIVE 3 +#define R_LARCH_COPY 4 +#define R_LARCH_JUMP_SLOT 5 +#define R_LARCH_TLS_DTPMOD32 6 +#define R_LARCH_TLS_DTPMOD64 7 +#define R_LARCH_TLS_DTPREL32 8 +#define R_LARCH_TLS_DTPREL64 9 +#define R_LARCH_TLS_TPREL32 10 +#define R_LARCH_TLS_TPREL64 11 +#define R_LARCH_IRELATIVE 12 +#define R_LARCH_MARK_LA 20 +#define R_LARCH_MARK_PCREL 21 +#define R_LARCH_SOP_PUSH_PCREL 22 +#define R_LARCH_SOP_PUSH_ABSOLUTE 23 +#define R_LARCH_SOP_PUSH_DUP 24 +#define R_LARCH_SOP_PUSH_GPREL 25 +#define R_LARCH_SOP_PUSH_TLS_TPREL 26 +#define R_LARCH_SOP_PUSH_TLS_GOT 27 +#define R_LARCH_SOP_PUSH_TLS_GD 28 +#define R_LARCH_SOP_PUSH_PLT_PCREL 29 +#define R_LARCH_SOP_ASSERT 30 +#define R_LARCH_SOP_NOT 31 +#define R_LARCH_SOP_SUB 32 +#define R_LARCH_SOP_SL 33 +#define R_LARCH_SOP_SR 34 +#define R_LARCH_SOP_ADD 35 +#define R_LARCH_SOP_AND 36 +#define R_LARCH_SOP_IF_ELSE 37 +#define R_LARCH_SOP_POP_32_S_10_5 38 +#define R_LARCH_SOP_POP_32_U_10_12 39 +#define R_LARCH_SOP_POP_32_S_10_12 40 +#define R_LARCH_SOP_POP_32_S_10_16 41 +#define R_LARCH_SOP_POP_32_S_10_16_S2 42 +#define R_LARCH_SOP_POP_32_S_5_20 43 +#define R_LARCH_SOP_POP_32_S_0_5_10_16_S2 44 +#define R_LARCH_SOP_POP_32_S_0_10_10_16_S2 45 +#define R_LARCH_SOP_POP_32_U 46 +#define R_LARCH_ADD8 47 +#define R_LARCH_ADD16 48 +#define R_LARCH_ADD24 49 +#define R_LARCH_ADD32 50 +#define R_LARCH_ADD64 51 +#define R_LARCH_SUB8 52 +#define R_LARCH_SUB16 53 +#define R_LARCH_SUB24 54 +#define R_LARCH_SUB32 55 +#define R_LARCH_SUB64 56 +#define R_LARCH_GNU_VTINHERIT 57 +#define R_LARCH_GNU_VTENTRY 58 /* * Relocs whose processing do not require a stack machine. * * Spec addition: https://github.com/loongson/LoongArch-Documentation/pull/57 */ -#define R_LARCH_B16 64 -#define R_LARCH_B21 65 -#define R_LARCH_B26 66 -#define R_LARCH_ABS_HI20 67 -#define R_LARCH_ABS_LO12 68 -#define R_LARCH_ABS64_LO20 69 -#define R_LARCH_ABS64_HI12 70 -#define R_LARCH_PCALA_HI20 71 -#define R_LARCH_PCALA_LO12 72 -#define R_LARCH_PCALA64_LO20 73 -#define R_LARCH_PCALA64_HI12 74 -#define R_LARCH_GOT_PC_HI20 75 -#define R_LARCH_GOT_PC_LO12 76 -#define R_LARCH_GOT64_PC_LO20 77 -#define R_LARCH_GOT64_PC_HI12 78 -#define R_LARCH_GOT_HI20 79 -#define R_LARCH_GOT_LO12 80 -#define R_LARCH_GOT64_LO20 81 -#define R_LARCH_GOT64_HI12 82 -#define R_LARCH_TLS_LE_HI20 83 -#define R_LARCH_TLS_LE_LO12 84 -#define R_LARCH_TLS_LE64_LO20 85 -#define R_LARCH_TLS_LE64_HI12 86 -#define R_LARCH_TLS_IE_PC_HI20 87 -#define R_LARCH_TLS_IE_PC_LO12 88 -#define R_LARCH_TLS_IE64_PC_LO20 89 -#define R_LARCH_TLS_IE64_PC_HI12 90 -#define R_LARCH_TLS_IE_HI20 91 -#define R_LARCH_TLS_IE_LO12 92 -#define R_LARCH_TLS_IE64_LO20 93 -#define R_LARCH_TLS_IE64_HI12 94 -#define R_LARCH_TLS_LD_PC_HI20 95 -#define R_LARCH_TLS_LD_HI20 96 -#define R_LARCH_TLS_GD_PC_HI20 97 -#define R_LARCH_TLS_GD_HI20 98 -#define R_LARCH_32_PCREL 99 -#define R_LARCH_RELAX 100 +#define R_LARCH_B16 64 +#define R_LARCH_B21 65 +#define R_LARCH_B26 66 +#define R_LARCH_ABS_HI20 67 +#define R_LARCH_ABS_LO12 68 +#define R_LARCH_ABS64_LO20 69 +#define R_LARCH_ABS64_HI12 70 +#define R_LARCH_PCALA_HI20 71 +#define R_LARCH_PCALA_LO12 72 +#define R_LARCH_PCALA64_LO20 73 +#define R_LARCH_PCALA64_HI12 74 +#define R_LARCH_GOT_PC_HI20 75 +#define R_LARCH_GOT_PC_LO12 76 +#define R_LARCH_GOT64_PC_LO20 77 +#define R_LARCH_GOT64_PC_HI12 78 +#define R_LARCH_GOT_HI20 79 +#define R_LARCH_GOT_LO12 80 +#define R_LARCH_GOT64_LO20 81 +#define R_LARCH_GOT64_HI12 82 +#define R_LARCH_TLS_LE_HI20 83 +#define R_LARCH_TLS_LE_LO12 84 +#define R_LARCH_TLS_LE64_LO20 85 +#define R_LARCH_TLS_LE64_HI12 86 +#define R_LARCH_TLS_IE_PC_HI20 87 +#define R_LARCH_TLS_IE_PC_LO12 88 +#define R_LARCH_TLS_IE64_PC_LO20 89 +#define R_LARCH_TLS_IE64_PC_HI12 90 +#define R_LARCH_TLS_IE_HI20 91 +#define R_LARCH_TLS_IE_LO12 92 +#define R_LARCH_TLS_IE64_LO20 93 +#define R_LARCH_TLS_IE64_HI12 94 +#define R_LARCH_TLS_LD_PC_HI20 95 +#define R_LARCH_TLS_LD_HI20 96 +#define R_LARCH_TLS_GD_PC_HI20 97 +#define R_LARCH_TLS_GD_HI20 98 +#define R_LARCH_32_PCREL 99 +#define R_LARCH_RELAX 100 /* * Relocs added in ELF for the LoongArchâ„¢ Architecture v20230519, part of the @@ -1520,13 +1526,13 @@ typedef struct { * in psABI v2.20 because they were proved not necessary to be exposed outside * of the linker. */ -#define R_LARCH_ALIGN 102 -#define R_LARCH_PCREL20_S2 103 -#define R_LARCH_ADD6 105 -#define R_LARCH_SUB6 106 -#define R_LARCH_ADD_ULEB128 107 -#define R_LARCH_SUB_ULEB128 108 -#define R_LARCH_64_PCREL 109 +#define R_LARCH_ALIGN 102 +#define R_LARCH_PCREL20_S2 103 +#define R_LARCH_ADD6 105 +#define R_LARCH_SUB6 106 +#define R_LARCH_ADD_ULEB128 107 +#define R_LARCH_SUB_ULEB128 108 +#define R_LARCH_64_PCREL 109 /* * Relocs added in ELF for the LoongArchâ„¢ Architecture v20231102, part of the @@ -1534,7 +1540,7 @@ typedef struct { * * Spec addition: https://github.com/loongson/la-abi-specs/pull/4 */ -#define R_LARCH_CALL36 110 +#define R_LARCH_CALL36 110 /* * Relocs added in ELF for the LoongArchâ„¢ Architecture v20231219, part of the @@ -1542,24 +1548,24 @@ typedef struct { * * Spec addition: https://github.com/loongson/la-abi-specs/pull/5 */ -#define R_LARCH_TLS_DESC32 13 -#define R_LARCH_TLS_DESC64 14 -#define R_LARCH_TLS_DESC_PC_HI20 111 -#define R_LARCH_TLS_DESC_PC_LO12 112 -#define R_LARCH_TLS_DESC64_PC_LO20 113 -#define R_LARCH_TLS_DESC64_PC_HI12 114 -#define R_LARCH_TLS_DESC_HI20 115 -#define R_LARCH_TLS_DESC_LO12 116 -#define R_LARCH_TLS_DESC64_LO20 117 -#define R_LARCH_TLS_DESC64_HI12 118 -#define R_LARCH_TLS_DESC_LD 119 -#define R_LARCH_TLS_DESC_CALL 120 -#define R_LARCH_TLS_LE_HI20_R 121 -#define R_LARCH_TLS_LE_ADD_R 122 -#define R_LARCH_TLS_LE_LO12_R 123 -#define R_LARCH_TLS_LD_PCREL20_S2 124 -#define R_LARCH_TLS_GD_PCREL20_S2 125 -#define R_LARCH_TLS_DESC_PCREL20_S2 126 +#define R_LARCH_TLS_DESC32 13 +#define R_LARCH_TLS_DESC64 14 +#define R_LARCH_TLS_DESC_PC_HI20 111 +#define R_LARCH_TLS_DESC_PC_LO12 112 +#define R_LARCH_TLS_DESC64_PC_LO20 113 +#define R_LARCH_TLS_DESC64_PC_HI12 114 +#define R_LARCH_TLS_DESC_HI20 115 +#define R_LARCH_TLS_DESC_LO12 116 +#define R_LARCH_TLS_DESC64_LO20 117 +#define R_LARCH_TLS_DESC64_HI12 118 +#define R_LARCH_TLS_DESC_LD 119 +#define R_LARCH_TLS_DESC_CALL 120 +#define R_LARCH_TLS_LE_HI20_R 121 +#define R_LARCH_TLS_LE_ADD_R 122 +#define R_LARCH_TLS_LE_LO12_R 123 +#define R_LARCH_TLS_LD_PCREL20_S2 124 +#define R_LARCH_TLS_GD_PCREL20_S2 125 +#define R_LARCH_TLS_DESC_PCREL20_S2 126 #define R_SPARC_NONE 0 #define R_SPARC_8 1 diff --git a/sys/sys/exec.h b/sys/sys/exec.h index 4bf114a7c698..580a5372c4db 100644 --- a/sys/sys/exec.h +++ b/sys/sys/exec.h @@ -57,16 +57,6 @@ struct ps_strings { unsigned int ps_nenvstr; /* the number of environment strings */ }; -/* Coredump output parameters. */ -struct coredump_params { - off_t offset; - struct ucred *active_cred; - struct ucred *file_cred; - struct thread *td; - struct vnode *vp; - struct compressor *comp; -}; - struct image_params; struct execsw { @@ -105,16 +95,6 @@ int exec_unregister(const struct execsw *); enum uio_seg; -#define CORE_BUF_SIZE (16 * 1024) - -int core_write(struct coredump_params *, const void *, size_t, off_t, - enum uio_seg, size_t *); -int core_output(char *, size_t, off_t, struct coredump_params *, void *); -int sbuf_drain_core_output(void *, const char *, int); - -extern int coredump_pack_fileinfo; -extern int coredump_pack_vmmapinfo; - /* * note: name##_mod cannot be const storage because the * linker_file_sysinit() function modifies _file in the diff --git a/sys/sys/exterr_cat.h b/sys/sys/exterr_cat.h index cab94ac511a5..80cff53b3576 100644 --- a/sys/sys/exterr_cat.h +++ b/sys/sys/exterr_cat.h @@ -18,6 +18,8 @@ #define EXTERR_CAT_FUSE 4 #define EXTERR_CAT_INOTIFY 5 #define EXTERR_CAT_GENIO 6 +#define EXTERR_CAT_BRIDGE 7 +#define EXTERR_CAT_SWAP 8 #endif diff --git a/sys/sys/exterrvar.h b/sys/sys/exterrvar.h index 15557c614f88..7bf1d264ff5e 100644 --- a/sys/sys/exterrvar.h +++ b/sys/sys/exterrvar.h @@ -21,6 +21,7 @@ #define EXTERRCTL_ENABLE 1 #define EXTERRCTL_DISABLE 2 +#define EXTERRCTL_UD 3 #define EXTERRCTLF_FORCE 0x00000001 diff --git a/sys/sys/imgact_elf.h b/sys/sys/imgact_elf.h index c9444e5aec41..2845a9dbc1e2 100644 --- a/sys/sys/imgact_elf.h +++ b/sys/sys/imgact_elf.h @@ -45,6 +45,7 @@ {(pos)->a_type = (id); (pos)->a_un.a_ptr = (ptr); (pos)++;} #endif +struct coredump_writer; struct image_params; struct thread; struct vnode; @@ -114,7 +115,7 @@ bool __elfN(brand_inuse)(Elf_Brandinfo *entry); int __elfN(insert_brand_entry)(Elf_Brandinfo *entry); int __elfN(remove_brand_entry)(Elf_Brandinfo *entry); int __elfN(freebsd_fixup)(uintptr_t *, struct image_params *); -int __elfN(coredump)(struct thread *, struct vnode *, off_t, int); +int __elfN(coredump)(struct thread *, struct coredump_writer *, off_t, int); size_t __elfN(populate_note)(int, void *, void *, size_t, void **); int __elfN(freebsd_copyout_auxargs)(struct image_params *, uintptr_t); void __elfN(puthdr)(struct thread *, void *, size_t, int, size_t, int); diff --git a/sys/sys/inotify.h b/sys/sys/inotify.h index 65dc5dba43f3..d1f23d5898bb 100644 --- a/sys/sys/inotify.h +++ b/sys/sys/inotify.h @@ -107,11 +107,18 @@ void vn_inotify_revoke(struct vnode *); } while (0) /* Log an inotify event using a specific name for the vnode. */ -#define INOTIFY_NAME(vp, dvp, cnp, ev) do { \ +#define INOTIFY_NAME_LOCK(vp, dvp, cnp, ev, lock) do { \ if (__predict_false((vn_irflag_read(vp) & VIRF_INOTIFY) != 0 || \ - (vn_irflag_read(dvp) & VIRF_INOTIFY) != 0)) \ + (vn_irflag_read(dvp) & VIRF_INOTIFY) != 0)) { \ + if (lock) \ + vn_lock((vp), LK_SHARED | LK_RETRY); \ VOP_INOTIFY((vp), (dvp), (cnp), (ev), 0); \ + if (lock) \ + VOP_UNLOCK(vp); \ + } \ } while (0) +#define INOTIFY_NAME(vp, dvp, cnp, ev) \ + INOTIFY_NAME_LOCK((vp), (dvp), (cnp), (ev), false) extern __uint32_t inotify_rename_cookie; @@ -126,7 +133,8 @@ extern __uint32_t inotify_rename_cookie; VOP_INOTIFY((vp), (tdvp), (tcnp), IN_MOVED_TO, cookie); \ } \ if ((tvp) != NULL) \ - INOTIFY_NAME((tvp), (tdvp), (tcnp), _IN_MOVE_DELETE); \ + INOTIFY_NAME_LOCK((tvp), (tdvp), (tcnp), \ + _IN_MOVE_DELETE, true); \ } while (0) #define INOTIFY_REVOKE(vp) do { \ diff --git a/sys/sys/jail.h b/sys/sys/jail.h index 08caa9f49270..24c420e2c976 100644 --- a/sys/sys/jail.h +++ b/sys/sys/jail.h @@ -435,7 +435,7 @@ void prison0_init(void); bool prison_allow(struct ucred *, unsigned); int prison_check(struct ucred *cred1, struct ucred *cred2); bool prison_check_nfsd(struct ucred *cred); -bool prison_owns_vnet(struct ucred *); +bool prison_owns_vnet(struct prison *pr); int prison_canseemount(struct ucred *cred, struct mount *mp); void prison_enforce_statfs(struct ucred *cred, struct mount *mp, struct statfs *sp); diff --git a/sys/sys/mbuf.h b/sys/sys/mbuf.h index c75094aea450..304bd019c9fc 100644 --- a/sys/sys/mbuf.h +++ b/sys/sys/mbuf.h @@ -1391,6 +1391,7 @@ extern bool mb_use_ext_pgs; /* Use ext_pgs for sendfile */ #define PACKET_TAG_PF_REASSEMBLED 31 #define PACKET_TAG_IPSEC_ACCEL_OUT 32 /* IPSEC accel out */ #define PACKET_TAG_IPSEC_ACCEL_IN 33 /* IPSEC accel in */ +#define PACKET_TAG_OVPN 34 /* if_ovpn */ /* Specific cookies and tags. */ diff --git a/sys/sys/param.h b/sys/sys/param.h index af116d6e3f7a..33d61e8a1619 100644 --- a/sys/sys/param.h +++ b/sys/sys/param.h @@ -74,7 +74,7 @@ * cannot include sys/param.h and should only be updated here. */ #undef __FreeBSD_version -#define __FreeBSD_version 1500051 +#define __FreeBSD_version 1500055 /* * __FreeBSD_kernel__ indicates that this system uses the kernel of FreeBSD, diff --git a/sys/sys/random.h b/sys/sys/random.h index 254ba9451d0a..5abf762cd200 100644 --- a/sys/sys/random.h +++ b/sys/sys/random.h @@ -85,7 +85,8 @@ enum random_entropy_source { RANDOM_FS_ATIME, RANDOM_UMA, /* Special!! UMA/SLAB Allocator */ RANDOM_CALLOUT, - RANDOM_ENVIRONMENTAL_END = RANDOM_CALLOUT, + RANDOM_RANDOMDEV, + RANDOM_ENVIRONMENTAL_END = RANDOM_RANDOMDEV, /* Fast hardware random-number sources from here on. */ RANDOM_PURE_START, RANDOM_PURE_OCTEON = RANDOM_PURE_START, diff --git a/sys/sys/signalvar.h b/sys/sys/signalvar.h index 23e8426b26ee..8f181b7beee6 100644 --- a/sys/sys/signalvar.h +++ b/sys/sys/signalvar.h @@ -403,6 +403,7 @@ int sigev_findtd(struct proc *p, struct sigevent *sigev, struct thread **); void sigfastblock_clear(struct thread *td); void sigfastblock_fetch(struct thread *td); int sig_intr(void); +bool sig_do_core(int); void siginit(struct proc *p); void signotify(struct thread *td); void sigqueue_delete(struct sigqueue *queue, int sig); diff --git a/sys/sys/sockbuf.h b/sys/sys/sockbuf.h index 7f6234ade6f4..b4593f38f592 100644 --- a/sys/sys/sockbuf.h +++ b/sys/sys/sockbuf.h @@ -40,7 +40,7 @@ #define SB_SEL 0x08 /* someone is selecting */ #define SB_ASYNC 0x10 /* ASYNC I/O, need signals */ #define SB_UPCALL 0x20 /* someone wants an upcall */ -/* was SB_NOINTR 0x40 */ +#define SB_AUTOLOWAT 0x40 /* sendfile(2) may autotune sb_lowat */ #define SB_AIO 0x80 /* AIO operations queued */ #define SB_KNOTE 0x100 /* kernel note attached */ #define SB_NOCOALESCE 0x200 /* don't coalesce new data into existing mbufs */ @@ -210,8 +210,6 @@ typedef enum { SO_RCV, SO_SND } sb_which; * Socket buffer private mbuf(9) flags. */ #define M_NOTREADY M_PROTO1 /* m_data not populated yet */ -#define M_BLOCKED M_PROTO2 /* M_NOTREADY in front of m */ -#define M_NOTAVAIL (M_NOTREADY | M_BLOCKED) void sbappend(struct sockbuf *sb, struct mbuf *m, int flags); void sbappend_locked(struct sockbuf *sb, struct mbuf *m, int flags); diff --git a/sys/sys/syscallsubr.h b/sys/sys/syscallsubr.h index fd183ffbc7a4..8237165b84ce 100644 --- a/sys/sys/syscallsubr.h +++ b/sys/sys/syscallsubr.h @@ -60,6 +60,7 @@ struct rusage; struct sched_param; struct sembuf; union semun; +struct shmfd; struct sockaddr; struct spacectl_range; struct stat; @@ -337,7 +338,7 @@ int kern_shm_open(struct thread *td, const char *userpath, int flags, mode_t mode, struct filecaps *fcaps); int kern_shm_open2(struct thread *td, const char *path, int flags, mode_t mode, int shmflags, struct filecaps *fcaps, - const char *name); + const char *name, struct shmfd *shmfd); int kern_shmat(struct thread *td, int shmid, const void *shmaddr, int shmflg); int kern_shmctl(struct thread *td, int shmid, int cmd, void *buf, diff --git a/sys/sys/sysent.h b/sys/sys/sysent.h index 4ddfc8516053..1714fa5a7416 100644 --- a/sys/sys/sysent.h +++ b/sys/sys/sysent.h @@ -90,6 +90,7 @@ struct sysent { /* system call table */ #define SY_THR_STATIC_KLD SY_THR_STATIC #endif +struct coredump_writer; struct image_params; struct proc; struct __sigset; @@ -108,7 +109,8 @@ struct sysentvec { int *sv_szsigcode; /* size of sigtramp code */ int sv_sigcodeoff; char *sv_name; /* name of binary type */ - int (*sv_coredump)(struct thread *, struct vnode *, off_t, int); + int (*sv_coredump)(struct thread *, struct coredump_writer *, + off_t, int); /* function to dump core, or NULL */ int sv_elf_core_osabi; const char *sv_elf_core_abi_vendor; diff --git a/sys/sys/ucoredump.h b/sys/sys/ucoredump.h new file mode 100644 index 000000000000..0a51ee7f50c8 --- /dev/null +++ b/sys/sys/ucoredump.h @@ -0,0 +1,99 @@ +/* + * + * Copyright (c) 2015 Mark Johnston <markj@FreeBSD.org> + * Copyright (c) 2025 Kyle Evans <kevans@FreeBSD.org> + * + * SPDX-License-Identifier: BSD-2-Clause + * + */ + +#ifndef _SYS_UCOREDUMP_H_ +#define _SYS_UCOREDUMP_H_ + +#ifdef _KERNEL + +#include <sys/_uio.h> +#include <sys/blockcount.h> +#include <sys/queue.h> + +/* Coredump output parameters. */ +struct coredump_params; +struct coredump_writer; +struct thread; +struct ucred; + +typedef int coredump_init_fn(const struct coredump_writer *, + const struct coredump_params *); +typedef int coredump_write_fn(const struct coredump_writer *, const void *, size_t, + off_t, enum uio_seg, struct ucred *, size_t *, struct thread *); +typedef int coredump_extend_fn(const struct coredump_writer *, off_t, + struct ucred *); + +struct coredump_vnode_ctx { + struct vnode *vp; + struct ucred *fcred; +}; + +coredump_write_fn core_vn_write; +coredump_extend_fn core_vn_extend; + +struct coredump_writer { + void *ctx; + coredump_init_fn *init_fn; + coredump_write_fn *write_fn; + coredump_extend_fn *extend_fn; +}; + +struct coredump_params { + off_t offset; + struct ucred *active_cred; + struct thread *td; + const struct coredump_writer *cdw; + struct compressor *comp; +}; + +#define CORE_BUF_SIZE (16 * 1024) + +int core_write(struct coredump_params *, const void *, size_t, off_t, + enum uio_seg, size_t *); +int core_output(char *, size_t, off_t, struct coredump_params *, void *); +int sbuf_drain_core_output(void *, const char *, int); + +extern int coredump_pack_fileinfo; +extern int coredump_pack_vmmapinfo; + +extern int compress_user_cores; +extern int compress_user_cores_level; + +typedef int coredumper_probe_fn(struct thread *); + +/* + * Some arbitrary values for coredumper probes to return. The highest priority + * we can find wins. It's somewhat expected that a coredumper may want to bid + * differently based on the process in question. Note that probe functions will + * be called with the proc lock held, so they must not sleep. + */ +#define COREDUMPER_NOMATCH (-1) /* Decline to touch it */ +#define COREDUMPER_GENERIC (0) /* I handle coredumps */ +#define COREDUMPER_SPECIAL (50) /* Special handler */ +#define COREDUMPER_HIGH_PRIORITY (100) /* High-priority handler */ + +/* + * The handle functions will be called with the proc lock held, and should + * return with the proc lock dropped. + */ +typedef int coredumper_handle_fn(struct thread *, off_t); + +struct coredumper { + SLIST_ENTRY(coredumper) cd_entry; + const char *cd_name; + coredumper_probe_fn *cd_probe; + coredumper_handle_fn *cd_handle; + blockcount_t cd_refcount; +}; + +void coredumper_register(struct coredumper *); +void coredumper_unregister(struct coredumper *); + +#endif /* _KERNEL */ +#endif /* _SYS_UCOREDUMP_H_ */ diff --git a/sys/sys/unistd.h b/sys/sys/unistd.h index c291c1dc2b95..85ed93fd359d 100644 --- a/sys/sys/unistd.h +++ b/sys/sys/unistd.h @@ -156,6 +156,8 @@ #define _PC_DEALLOC_PRESENT 65 #define _PC_NAMEDATTR_ENABLED 66 #define _PC_HAS_NAMEDATTR 67 +#define _PC_XATTR_ENABLED _PC_NAMEDATTR_ENABLED /* Solaris Compatible */ +#define _PC_XATTR_EXISTS _PC_HAS_NAMEDATTR /* Solaris Compatible */ #define _PC_HAS_HIDDENSYSTEM 68 #endif diff --git a/sys/sys/vnode.h b/sys/sys/vnode.h index 2c6947103c94..a416fddcddc3 100644 --- a/sys/sys/vnode.h +++ b/sys/sys/vnode.h @@ -939,7 +939,6 @@ void vop_mknod_post(void *a, int rc); void vop_open_post(void *a, int rc); void vop_read_post(void *a, int rc); void vop_read_pgcache_post(void *ap, int rc); -void vop_readdir_post(void *a, int rc); void vop_reclaim_post(void *a, int rc); void vop_remove_pre(void *a); void vop_remove_post(void *a, int rc); @@ -1015,7 +1014,36 @@ void vop_rename_fail(struct vop_rename_args *ap); _error; \ }) -#define VOP_WRITE_PRE(ap) \ +#ifdef INVARIANTS +#define vop_readdir_pre_assert(ap) \ + ssize_t nresid, oresid; \ + \ + oresid = (ap)->a_uio->uio_resid; + +#define vop_readdir_post_assert(ap, ret) \ + nresid = (ap)->a_uio->uio_resid; \ + if ((ret) == 0 && (ap)->a_eofflag != NULL) { \ + VNASSERT(oresid == 0 || nresid != oresid || \ + *(ap)->a_eofflag == 1, \ + (ap)->a_vp, ("VOP_READDIR: eofflag not set")); \ + } +#else +#define vop_readdir_pre_assert(ap) +#define vop_readdir_post_assert(ap, ret) +#endif + +#define vop_readdir_pre(ap) do { \ + vop_readdir_pre_assert(ap) + +#define vop_readdir_post(ap, ret) \ + vop_readdir_post_assert(ap, ret); \ + if ((ret) == 0) { \ + VFS_KNOTE_LOCKED((ap)->a_vp, NOTE_READ); \ + INOTIFY((ap)->a_vp, IN_ACCESS); \ + } \ +} while (0) + +#define vop_write_pre(ap) \ struct vattr va; \ int error; \ off_t osize, ooffset, noffset; \ @@ -1029,7 +1057,7 @@ void vop_rename_fail(struct vop_rename_args *ap); osize = (off_t)va.va_size; \ } -#define VOP_WRITE_POST(ap, ret) \ +#define vop_write_post(ap, ret) \ noffset = (ap)->a_uio->uio_offset; \ if (noffset > ooffset) { \ if (!VN_KNLIST_EMPTY((ap)->a_vp)) { \ diff --git a/sys/ufs/ufs/ufs_vnops.c b/sys/ufs/ufs/ufs_vnops.c index 53fac4b0665e..b7453db9013c 100644 --- a/sys/ufs/ufs/ufs_vnops.c +++ b/sys/ufs/ufs/ufs_vnops.c @@ -1268,7 +1268,8 @@ ufs_rename( struct inode *fip, *tip, *tdp, *fdp; struct direct newdir; off_t endoff; - int doingdirectory, newparent; + int doingdirectory; + u_int newparent; int error = 0; struct mount *mp; ino_t ino; @@ -1475,7 +1476,7 @@ relock: * the user must have write permission in the source so * as to be able to change "..". */ - if (doingdirectory && newparent) { + if (doingdirectory && newparent != 0) { error = VOP_ACCESS(fvp, VWRITE, tcnp->cn_cred, curthread); if (error) goto unlockout; @@ -1538,7 +1539,7 @@ relock: if (tip == NULL) { if (ITODEV(tdp) != ITODEV(fip)) panic("ufs_rename: EXDEV"); - if (doingdirectory && newparent) { + if (doingdirectory && newparent != 0) { /* * Account for ".." in new directory. * When source and destination have the same @@ -1631,7 +1632,7 @@ relock: goto bad; } if (doingdirectory) { - if (!newparent) { + if (newparent == 0) { tdp->i_effnlink--; if (DOINGSOFTDEP(tdvp)) softdep_change_linkcnt(tdp); @@ -1641,11 +1642,11 @@ relock: softdep_change_linkcnt(tip); } error = ufs_dirrewrite(tdp, tip, fip->i_number, - IFTODT(fip->i_mode), - (doingdirectory && newparent) ? newparent : doingdirectory); + IFTODT(fip->i_mode), (doingdirectory && newparent != 0) ? + newparent : doingdirectory); if (error) { if (doingdirectory) { - if (!newparent) { + if (newparent == 0) { tdp->i_effnlink++; if (DOINGSOFTDEP(tdvp)) softdep_change_linkcnt(tdp); @@ -1668,7 +1669,7 @@ relock: * disk, so when running with that code we avoid doing * them now. */ - if (!newparent) { + if (newparent == 0) { tdp->i_nlink--; DIP_SET_NLINK(tdp, tdp->i_nlink); UFS_INODE_SET_FLAG(tdp, IN_CHANGE); @@ -1697,7 +1698,7 @@ relock: * parent directory must be decremented * and ".." set to point to the new parent. */ - if (doingdirectory && newparent) { + if (doingdirectory && newparent != 0) { /* * Set the directory depth based on its new parent. */ @@ -2064,9 +2065,13 @@ ufs_mkdir( */ ucred.cr_ref = 1; ucred.cr_uid = ip->i_uid; + + /* + * XXXKE Fix this is cr_gid gets separated out + */ ucred.cr_ngroups = 1; ucred.cr_groups = &ucred_group; - ucred.cr_groups[0] = dp->i_gid; + ucred.cr_gid = ucred_group = dp->i_gid; ucp = &ucred; } #endif @@ -2823,9 +2828,13 @@ ufs_makeinode(int mode, struct vnode *dvp, struct vnode **vpp, */ ucred.cr_ref = 1; ucred.cr_uid = ip->i_uid; + + /* + * XXXKE Fix this is cr_gid gets separated out + */ ucred.cr_ngroups = 1; ucred.cr_groups = &ucred_group; - ucred.cr_groups[0] = pdir->i_gid; + ucred.cr_gid = ucred_group = pdir->i_gid; ucp = &ucred; #endif } else { diff --git a/sys/vm/swap_pager.c b/sys/vm/swap_pager.c index 86b75a2d7989..327cac661044 100644 --- a/sys/vm/swap_pager.c +++ b/sys/vm/swap_pager.c @@ -65,9 +65,9 @@ * from: Utah $Hdr: swap_pager.c 1.4 91/04/30$ */ -#include <sys/cdefs.h> #include "opt_vm.h" +#define EXTERR_CATEGORY EXTERR_CAT_SWAP #include <sys/param.h> #include <sys/bio.h> #include <sys/blist.h> @@ -76,6 +76,7 @@ #include <sys/disk.h> #include <sys/disklabel.h> #include <sys/eventhandler.h> +#include <sys/exterrvar.h> #include <sys/fcntl.h> #include <sys/limits.h> #include <sys/lock.h> @@ -384,8 +385,8 @@ swap_release_by_cred(vm_ooffset_t decr, struct ucred *cred) #endif } -static int swap_pager_full = 2; /* swap space exhaustion (task killing) */ -static int swap_pager_almost_full = 1; /* swap space exhaustion (w/hysteresis)*/ +static bool swap_pager_full = true; /* swap space exhaustion (task killing) */ +static bool swap_pager_almost_full = true; /* swap space exhaustion (w/hysteresis) */ static struct mtx swbuf_mtx; /* to sync nsw_wcount_async */ static int nsw_wcount_async; /* limit async write buffers */ static int nsw_wcount_async_max;/* assigned maximum */ @@ -642,14 +643,14 @@ swp_sizecheck(void) { if (swap_pager_avail < nswap_lowat) { - if (swap_pager_almost_full == 0) { + if (!swap_pager_almost_full) { printf("swap_pager: out of swap space\n"); - swap_pager_almost_full = 1; + swap_pager_almost_full = true; } } else { - swap_pager_full = 0; + swap_pager_full = false; if (swap_pager_avail > nswap_hiwat) - swap_pager_almost_full = 0; + swap_pager_almost_full = false; } } @@ -958,11 +959,10 @@ swp_pager_getswapspace(int *io_npages) swp_sizecheck(); swdevhd = TAILQ_NEXT(sp, sw_list); } else { - if (swap_pager_full != 2) { + if (!swap_pager_full) { printf("swp_pager_getswapspace(%d): failed\n", *io_npages); - swap_pager_full = 2; - swap_pager_almost_full = 1; + swap_pager_full = swap_pager_almost_full = true; } swdevhd = NULL; } @@ -2687,7 +2687,7 @@ swapon_check_swzone(void) } } -static void +static int swaponsomething(struct vnode *vp, void *id, u_long nblks, sw_strategy_t *strategy, sw_close_t *close, dev_t dev, int flags) { @@ -2702,6 +2702,8 @@ swaponsomething(struct vnode *vp, void *id, u_long nblks, */ nblks &= ~(ctodb(1) - 1); nblks = dbtoc(nblks); + if (nblks == 0) + return (EXTERROR(EINVAL, "swap device too small")); sp = malloc(sizeof *sp, M_VMPGDATA, M_WAITOK | M_ZERO); sp->sw_blist = blist_create(nblks, M_WAITOK); @@ -2743,6 +2745,8 @@ swaponsomething(struct vnode *vp, void *id, u_long nblks, swp_sizecheck(); mtx_unlock(&sw_dev_mtx); EVENTHANDLER_INVOKE(swapon, sp); + + return (0); } /* @@ -2863,10 +2867,8 @@ swapoff_one(struct swdevt *sp, struct ucred *cred, u_int flags) sp->sw_id = NULL; TAILQ_REMOVE(&swtailq, sp, sw_list); nswapdev--; - if (nswapdev == 0) { - swap_pager_full = 2; - swap_pager_almost_full = 1; - } + if (nswapdev == 0) + swap_pager_full = swap_pager_almost_full = true; if (swdevhd == sp) swdevhd = NULL; mtx_unlock(&sw_dev_mtx); @@ -3289,10 +3291,10 @@ swapongeom_locked(struct cdev *dev, struct vnode *vp) return (error); } nblks = pp->mediasize / DEV_BSIZE; - swaponsomething(vp, cp, nblks, swapgeom_strategy, + error = swaponsomething(vp, cp, nblks, swapgeom_strategy, swapgeom_close, dev2udev(dev), (pp->flags & G_PF_ACCEPT_UNMAPPED) != 0 ? SW_UNMAPPED : 0); - return (0); + return (error); } static int @@ -3381,9 +3383,9 @@ swaponvp(struct thread *td, struct vnode *vp, u_long nblks) if (error != 0) return (error); - swaponsomething(vp, vp, nblks, swapdev_strategy, swapdev_close, + error = swaponsomething(vp, vp, nblks, swapdev_strategy, swapdev_close, NODEV, 0); - return (0); + return (error); } static int diff --git a/sys/vm/vm_domainset.c b/sys/vm/vm_domainset.c index 7b8bf4c77663..b44bdb96b0d4 100644 --- a/sys/vm/vm_domainset.c +++ b/sys/vm/vm_domainset.c @@ -131,8 +131,7 @@ static void vm_domainset_iter_next(struct vm_domainset_iter *di, int *domain) { - KASSERT(di->di_n > 0, - ("vm_domainset_iter_first: Invalid n %d", di->di_n)); + KASSERT(di->di_n > 0, ("%s: Invalid n %d", __func__, di->di_n)); switch (di->di_policy) { case DOMAINSET_POLICY_FIRSTTOUCH: /* @@ -149,11 +148,10 @@ vm_domainset_iter_next(struct vm_domainset_iter *di, int *domain) vm_domainset_iter_prefer(di, domain); break; default: - panic("vm_domainset_iter_first: Unknown policy %d", - di->di_policy); + panic("%s: Unknown policy %d", __func__, di->di_policy); } KASSERT(*domain < vm_ndomains, - ("vm_domainset_iter_next: Invalid domain %d", *domain)); + ("%s: Invalid domain %d", __func__, *domain)); } static void @@ -189,13 +187,11 @@ vm_domainset_iter_first(struct vm_domainset_iter *di, int *domain) di->di_n = di->di_domain->ds_cnt; break; default: - panic("vm_domainset_iter_first: Unknown policy %d", - di->di_policy); + panic("%s: Unknown policy %d", __func__, di->di_policy); } - KASSERT(di->di_n > 0, - ("vm_domainset_iter_first: Invalid n %d", di->di_n)); + KASSERT(di->di_n > 0, ("%s: Invalid n %d", __func__, di->di_n)); KASSERT(*domain < vm_ndomains, - ("vm_domainset_iter_first: Invalid domain %d", *domain)); + ("%s: Invalid domain %d", __func__, *domain)); } void diff --git a/sys/vm/vm_kern.c b/sys/vm/vm_kern.c index 875c22d27628..e7d7b6726d2c 100644 --- a/sys/vm/vm_kern.c +++ b/sys/vm/vm_kern.c @@ -110,11 +110,18 @@ u_int exec_map_entry_size; u_int exec_map_entries; SYSCTL_ULONG(_vm, OID_AUTO, min_kernel_address, CTLFLAG_RD, - SYSCTL_NULL_ULONG_PTR, VM_MIN_KERNEL_ADDRESS, "Min kernel address"); +#if defined(__amd64__) + &kva_layout.km_low, 0, +#else + SYSCTL_NULL_ULONG_PTR, VM_MIN_KERNEL_ADDRESS, +#endif + "Min kernel address"); SYSCTL_ULONG(_vm, OID_AUTO, max_kernel_address, CTLFLAG_RD, #if defined(__arm__) &vm_max_kernel_address, 0, +#elif defined(__amd64__) + &kva_layout.km_high, 0, #else SYSCTL_NULL_ULONG_PTR, VM_MAX_KERNEL_ADDRESS, #endif diff --git a/sys/vm/vm_page.c b/sys/vm/vm_page.c index bbae55895c2c..b239a6ffb4ce 100644 --- a/sys/vm/vm_page.c +++ b/sys/vm/vm_page.c @@ -396,7 +396,7 @@ vm_page_blacklist_load(char **list, char **end) } *list = ptr; if (ptr != NULL) - *end = ptr + len; + *end = ptr + len - 1; else *end = NULL; return; diff --git a/sys/vm/vm_pagequeue.h b/sys/vm/vm_pagequeue.h index cbbd27389662..9bd3b389fb60 100644 --- a/sys/vm/vm_pagequeue.h +++ b/sys/vm/vm_pagequeue.h @@ -260,9 +260,9 @@ struct vm_domain { u_int vmd_inactive_shortage; /* Per-thread shortage. */ blockcount_t vmd_inactive_running; /* Number of inactive threads. */ blockcount_t vmd_inactive_starting; /* Number of threads started. */ - volatile u_int vmd_addl_shortage; /* Shortage accumulator. */ - volatile u_int vmd_inactive_freed; /* Successful inactive frees. */ - volatile u_int vmd_inactive_us; /* Microseconds for above. */ + u_int vmd_addl_shortage; /* (a) Shortage accumulator. */ + u_int vmd_inactive_freed; /* (a) Successful inactive frees. */ + u_int vmd_inactive_us; /* (a) Microseconds for above. */ u_int vmd_inactive_pps; /* Exponential decay frees/second. */ int vmd_oom_seq; int vmd_last_active_scan; |