diff options
Diffstat (limited to 'lib/libvmmapi')
| -rw-r--r-- | lib/libvmmapi/Makefile | 15 | ||||
| -rw-r--r-- | lib/libvmmapi/Makefile.depend | 18 | ||||
| -rw-r--r-- | lib/libvmmapi/aarch64/Makefile.inc | 1 | ||||
| -rw-r--r-- | lib/libvmmapi/aarch64/vmmapi_machdep.c | 127 | ||||
| -rw-r--r-- | lib/libvmmapi/amd64/Makefile.inc | 5 | ||||
| -rw-r--r-- | lib/libvmmapi/amd64/vmmapi_freebsd_machdep.c | 343 | ||||
| -rw-r--r-- | lib/libvmmapi/amd64/vmmapi_machdep.c | 607 | ||||
| -rw-r--r-- | lib/libvmmapi/internal.h | 75 | ||||
| -rw-r--r-- | lib/libvmmapi/ppt.c | 144 | ||||
| -rw-r--r-- | lib/libvmmapi/riscv/Makefile.inc | 1 | ||||
| -rw-r--r-- | lib/libvmmapi/riscv/vmmapi_machdep.c | 118 | ||||
| -rw-r--r-- | lib/libvmmapi/vmmapi.c | 1319 | ||||
| -rw-r--r-- | lib/libvmmapi/vmmapi.h | 301 |
13 files changed, 3074 insertions, 0 deletions
diff --git a/lib/libvmmapi/Makefile b/lib/libvmmapi/Makefile new file mode 100644 index 000000000000..6dd0deeaa9c0 --- /dev/null +++ b/lib/libvmmapi/Makefile @@ -0,0 +1,15 @@ +PACKAGE=lib${LIB} +LIB= vmmapi +SHLIB_MAJOR= 7 +SRCS= vmmapi.c +INCS= vmmapi.h + +CFLAGS+= -I${.CURDIR} + +.PATH: ${.CURDIR}/${MACHINE_CPUARCH} + +.include "${MACHINE_CPUARCH}/Makefile.inc" + +LIBADD= util + +.include <bsd.lib.mk> diff --git a/lib/libvmmapi/Makefile.depend b/lib/libvmmapi/Makefile.depend new file mode 100644 index 000000000000..ad56045f44d3 --- /dev/null +++ b/lib/libvmmapi/Makefile.depend @@ -0,0 +1,18 @@ +# Autogenerated - do NOT edit! + +DIRDEPS = \ + include \ + include/xlocale \ + lib/${CSU_DIR} \ + lib/libc \ + lib/libcapsicum \ + lib/libcasper/libcasper \ + lib/libcompiler_rt \ + lib/libutil \ + + +.include <dirdeps.mk> + +.if ${DEP_RELDIR} == ${_DEP_RELDIR} +# local dependencies - needed for -jN in clean tree +.endif diff --git a/lib/libvmmapi/aarch64/Makefile.inc b/lib/libvmmapi/aarch64/Makefile.inc new file mode 100644 index 000000000000..663ea0ab90a3 --- /dev/null +++ b/lib/libvmmapi/aarch64/Makefile.inc @@ -0,0 +1 @@ +SRCS+= vmmapi_machdep.c diff --git a/lib/libvmmapi/aarch64/vmmapi_machdep.c b/lib/libvmmapi/aarch64/vmmapi_machdep.c new file mode 100644 index 000000000000..fb2556af3be2 --- /dev/null +++ b/lib/libvmmapi/aarch64/vmmapi_machdep.c @@ -0,0 +1,127 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2011 NetApp, Inc. + * 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 NETAPP, INC ``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 NETAPP, INC 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/types.h> +#include <sys/ioctl.h> + +#include <machine/vmm.h> +#include <machine/vmm_dev.h> +#include <machine/vmm_snapshot.h> + +#include <assert.h> +#include <string.h> + +#include "vmmapi.h" +#include "internal.h" + +const char *vm_capstrmap[] = { + [VM_CAP_MAX] = NULL, +}; + +#define VM_MD_IOCTLS \ + VM_GET_VGIC_VERSION, \ + VM_ATTACH_VGIC, \ + VM_ASSERT_IRQ, \ + VM_DEASSERT_IRQ, \ + VM_RAISE_MSI + +const cap_ioctl_t vm_ioctl_cmds[] = { + VM_COMMON_IOCTLS, + VM_MD_IOCTLS, +}; +size_t vm_ioctl_ncmds = nitems(vm_ioctl_cmds); + +int +vm_attach_vgic(struct vmctx *ctx, uint64_t dist_start, size_t dist_size, + uint64_t redist_start, size_t redist_size) +{ + struct vm_vgic_descr vgic; + int error; + + bzero(&vgic, sizeof(vgic)); + error = ioctl(ctx->fd, VM_GET_VGIC_VERSION, &vgic.ver); + if (error != 0) + return (error); + assert(vgic.ver.version == 3); + vgic.v3_regs.dist_start = dist_start; + vgic.v3_regs.dist_size = dist_size; + vgic.v3_regs.redist_start = redist_start; + vgic.v3_regs.redist_size = redist_size; + + return (ioctl(ctx->fd, VM_ATTACH_VGIC, &vgic)); +} + +int +vm_assert_irq(struct vmctx *ctx, uint32_t irq) +{ + struct vm_irq vi; + + bzero(&vi, sizeof(vi)); + vi.irq = irq; + + return (ioctl(ctx->fd, VM_ASSERT_IRQ, &vi)); +} + +int +vm_deassert_irq(struct vmctx *ctx, uint32_t irq) +{ + struct vm_irq vi; + + bzero(&vi, sizeof(vi)); + vi.irq = irq; + + return (ioctl(ctx->fd, VM_DEASSERT_IRQ, &vi)); +} + +int +vm_raise_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg, + int bus, int slot, int func) +{ + struct vm_msi vmsi; + + bzero(&vmsi, sizeof(vmsi)); + vmsi.addr = addr; + vmsi.msg = msg; + vmsi.bus = bus; + vmsi.slot = slot; + vmsi.func = func; + + return (ioctl(ctx->fd, VM_RAISE_MSI, &vmsi)); +} + +int +vm_inject_exception(struct vcpu *vcpu, uint64_t esr, uint64_t far) +{ + struct vm_exception vmexc; + + bzero(&vmexc, sizeof(vmexc)); + vmexc.esr = esr; + vmexc.far = far; + + return (vcpu_ioctl(vcpu, VM_INJECT_EXCEPTION, &vmexc)); +} diff --git a/lib/libvmmapi/amd64/Makefile.inc b/lib/libvmmapi/amd64/Makefile.inc new file mode 100644 index 000000000000..a3da271433a1 --- /dev/null +++ b/lib/libvmmapi/amd64/Makefile.inc @@ -0,0 +1,5 @@ +SRCS+= ppt.c \ + vmmapi_machdep.c \ + vmmapi_freebsd_machdep.c + +CFLAGS+= -DWITH_VMMAPI_SNAPSHOT diff --git a/lib/libvmmapi/amd64/vmmapi_freebsd_machdep.c b/lib/libvmmapi/amd64/vmmapi_freebsd_machdep.c new file mode 100644 index 000000000000..a1d39ac3795e --- /dev/null +++ b/lib/libvmmapi/amd64/vmmapi_freebsd_machdep.c @@ -0,0 +1,343 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2011 NetApp, Inc. + * 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 NETAPP, INC ``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 NETAPP, INC 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/types.h> + +#include <machine/specialreg.h> +#include <machine/segments.h> +#include <machine/vmm.h> + +#include <errno.h> +#include <string.h> + +#include "vmmapi.h" +#include "internal.h" + +#define I386_TSS_SIZE 104 + +#define DESC_PRESENT 0x00000080 +#define DESC_LONGMODE 0x00002000 +#define DESC_DEF32 0x00004000 +#define DESC_GRAN 0x00008000 +#define DESC_UNUSABLE 0x00010000 + +#define GUEST_NULL_SEL 0 +#define GUEST_CODE_SEL 1 +#define GUEST_DATA_SEL 2 +#define GUEST_TSS_SEL 3 +#define GUEST_GDTR_LIMIT64 (3 * 8 - 1) + +static struct segment_descriptor i386_gdt[] = { + {}, /* NULL */ + { .sd_lolimit = 0xffff, .sd_type = SDT_MEMER, /* CODE */ + .sd_p = 1, .sd_hilimit = 0xf, .sd_def32 = 1, .sd_gran = 1 }, + { .sd_lolimit = 0xffff, .sd_type = SDT_MEMRW, /* DATA */ + .sd_p = 1, .sd_hilimit = 0xf, .sd_def32 = 1, .sd_gran = 1 }, + { .sd_lolimit = I386_TSS_SIZE - 1, /* TSS */ + .sd_type = SDT_SYS386TSS, .sd_p = 1 } +}; + +/* + * Setup the 'vcpu' register set such that it will begin execution at + * 'eip' in flat mode. + */ +int +vm_setup_freebsd_registers_i386(struct vcpu *vcpu, uint32_t eip, + uint32_t gdtbase, uint32_t esp) +{ + uint64_t cr0, rflags, desc_base; + uint32_t desc_access, desc_limit, tssbase; + uint16_t gsel; + struct segment_descriptor *gdt; + int error, tmp; + + /* A 32-bit guest requires unrestricted mode. */ + error = vm_get_capability(vcpu, VM_CAP_UNRESTRICTED_GUEST, &tmp); + if (error) + goto done; + error = vm_set_capability(vcpu, VM_CAP_UNRESTRICTED_GUEST, 1); + if (error) + goto done; + + cr0 = CR0_PE | CR0_NE; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CR0, cr0)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CR4, 0)) != 0) + goto done; + + /* + * Forcing EFER to 0 causes bhyve to clear the "IA-32e guest + * mode" entry control. + */ + if ((error = vm_set_register(vcpu, VM_REG_GUEST_EFER, 0))) + goto done; + + gdt = vm_map_gpa(vcpu->ctx, gdtbase, 0x1000); + if (gdt == NULL) + return (EFAULT); + memcpy(gdt, i386_gdt, sizeof(i386_gdt)); + desc_base = gdtbase; + desc_limit = sizeof(i386_gdt) - 1; + error = vm_set_desc(vcpu, VM_REG_GUEST_GDTR, + desc_base, desc_limit, 0); + if (error != 0) + goto done; + + /* Place the TSS one page above the GDT. */ + tssbase = gdtbase + 0x1000; + gdt[3].sd_lobase = tssbase; + + rflags = 0x2; + error = vm_set_register(vcpu, VM_REG_GUEST_RFLAGS, rflags); + if (error) + goto done; + + desc_base = 0; + desc_limit = 0xffffffff; + desc_access = DESC_GRAN | DESC_DEF32 | DESC_PRESENT | SDT_MEMERA; + error = vm_set_desc(vcpu, VM_REG_GUEST_CS, + desc_base, desc_limit, desc_access); + + desc_access = DESC_GRAN | DESC_DEF32 | DESC_PRESENT | SDT_MEMRWA; + error = vm_set_desc(vcpu, VM_REG_GUEST_DS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_ES, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_FS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_GS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_SS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + desc_base = tssbase; + desc_limit = I386_TSS_SIZE - 1; + desc_access = DESC_PRESENT | SDT_SYS386BSY; + error = vm_set_desc(vcpu, VM_REG_GUEST_TR, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + + error = vm_set_desc(vcpu, VM_REG_GUEST_LDTR, 0, 0, + DESC_UNUSABLE); + if (error) + goto done; + + gsel = GSEL(GUEST_CODE_SEL, SEL_KPL); + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CS, gsel)) != 0) + goto done; + + gsel = GSEL(GUEST_DATA_SEL, SEL_KPL); + if ((error = vm_set_register(vcpu, VM_REG_GUEST_DS, gsel)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_ES, gsel)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_FS, gsel)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_GS, gsel)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_SS, gsel)) != 0) + goto done; + + gsel = GSEL(GUEST_TSS_SEL, SEL_KPL); + if ((error = vm_set_register(vcpu, VM_REG_GUEST_TR, gsel)) != 0) + goto done; + + /* LDTR is pointing to the null selector */ + if ((error = vm_set_register(vcpu, VM_REG_GUEST_LDTR, 0)) != 0) + goto done; + + /* entry point */ + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RIP, eip)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RSP, esp)) != 0) + goto done; + + error = 0; +done: + return (error); +} + +void +vm_setup_freebsd_gdt(uint64_t *gdtr) +{ + gdtr[GUEST_NULL_SEL] = 0; + gdtr[GUEST_CODE_SEL] = 0x0020980000000000; + gdtr[GUEST_DATA_SEL] = 0x0000900000000000; +} + +/* + * Setup the 'vcpu' register set such that it will begin execution at + * 'rip' in long mode. + */ +int +vm_setup_freebsd_registers(struct vcpu *vcpu, + uint64_t rip, uint64_t cr3, uint64_t gdtbase, + uint64_t rsp) +{ + int error; + uint64_t cr0, cr4, efer, rflags, desc_base; + uint32_t desc_access, desc_limit; + uint16_t gsel; + + cr0 = CR0_PE | CR0_PG | CR0_NE; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CR0, cr0)) != 0) + goto done; + + cr4 = CR4_PAE; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CR4, cr4)) != 0) + goto done; + + efer = EFER_LME | EFER_LMA; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_EFER, efer))) + goto done; + + rflags = 0x2; + error = vm_set_register(vcpu, VM_REG_GUEST_RFLAGS, rflags); + if (error) + goto done; + + desc_base = 0; + desc_limit = 0; + desc_access = 0x0000209B; + error = vm_set_desc(vcpu, VM_REG_GUEST_CS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + desc_access = 0x00000093; + error = vm_set_desc(vcpu, VM_REG_GUEST_DS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_ES, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_FS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_GS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_SS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + /* + * XXX TR is pointing to null selector even though we set the + * TSS segment to be usable with a base address and limit of 0. + */ + desc_access = 0x0000008b; + error = vm_set_desc(vcpu, VM_REG_GUEST_TR, 0, 0, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_LDTR, 0, 0, + DESC_UNUSABLE); + if (error) + goto done; + + gsel = GSEL(GUEST_CODE_SEL, SEL_KPL); + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CS, gsel)) != 0) + goto done; + + gsel = GSEL(GUEST_DATA_SEL, SEL_KPL); + if ((error = vm_set_register(vcpu, VM_REG_GUEST_DS, gsel)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_ES, gsel)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_FS, gsel)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_GS, gsel)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_SS, gsel)) != 0) + goto done; + + /* XXX TR is pointing to the null selector */ + if ((error = vm_set_register(vcpu, VM_REG_GUEST_TR, 0)) != 0) + goto done; + + /* LDTR is pointing to the null selector */ + if ((error = vm_set_register(vcpu, VM_REG_GUEST_LDTR, 0)) != 0) + goto done; + + /* entry point */ + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RIP, rip)) != 0) + goto done; + + /* page table base */ + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CR3, cr3)) != 0) + goto done; + + desc_base = gdtbase; + desc_limit = GUEST_GDTR_LIMIT64; + error = vm_set_desc(vcpu, VM_REG_GUEST_GDTR, + desc_base, desc_limit, 0); + if (error != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RSP, rsp)) != 0) + goto done; + + error = 0; +done: + return (error); +} diff --git a/lib/libvmmapi/amd64/vmmapi_machdep.c b/lib/libvmmapi/amd64/vmmapi_machdep.c new file mode 100644 index 000000000000..806c31abf426 --- /dev/null +++ b/lib/libvmmapi/amd64/vmmapi_machdep.c @@ -0,0 +1,607 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2011 NetApp, Inc. + * 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 NETAPP, INC ``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 NETAPP, INC 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/types.h> +#include <sys/ioctl.h> + +#include <machine/specialreg.h> +#include <machine/vmm.h> +#include <machine/vmm_dev.h> +#include <machine/vmm_snapshot.h> + +#include <string.h> + +#include "vmmapi.h" +#include "internal.h" + +const char *vm_capstrmap[] = { + [VM_CAP_HALT_EXIT] = "hlt_exit", + [VM_CAP_MTRAP_EXIT] = "mtrap_exit", + [VM_CAP_PAUSE_EXIT] = "pause_exit", + [VM_CAP_UNRESTRICTED_GUEST] = "unrestricted_guest", + [VM_CAP_ENABLE_INVPCID] = "enable_invpcid", + [VM_CAP_BPT_EXIT] = "bpt_exit", + [VM_CAP_RDPID] = "rdpid", + [VM_CAP_RDTSCP] = "rdtscp", + [VM_CAP_IPI_EXIT] = "ipi_exit", + [VM_CAP_MASK_HWINTR] = "mask_hwintr", + [VM_CAP_RFLAGS_TF] = "rflags_tf", + [VM_CAP_MAX] = NULL, +}; + +#define VM_MD_IOCTLS \ + VM_SET_SEGMENT_DESCRIPTOR, \ + VM_GET_SEGMENT_DESCRIPTOR, \ + VM_SET_KERNEMU_DEV, \ + VM_GET_KERNEMU_DEV, \ + VM_LAPIC_IRQ, \ + VM_LAPIC_LOCAL_IRQ, \ + VM_LAPIC_MSI, \ + VM_IOAPIC_ASSERT_IRQ, \ + VM_IOAPIC_DEASSERT_IRQ, \ + VM_IOAPIC_PULSE_IRQ, \ + VM_IOAPIC_PINCOUNT, \ + VM_ISA_ASSERT_IRQ, \ + VM_ISA_DEASSERT_IRQ, \ + VM_ISA_PULSE_IRQ, \ + VM_ISA_SET_IRQ_TRIGGER, \ + VM_INJECT_NMI, \ + VM_SET_X2APIC_STATE, \ + VM_GET_X2APIC_STATE, \ + VM_GET_HPET_CAPABILITIES, \ + VM_RTC_WRITE, \ + VM_RTC_READ, \ + VM_RTC_SETTIME, \ + VM_RTC_GETTIME, \ + VM_GET_GPA_PMAP, \ + VM_GLA2GPA, \ + VM_SET_INTINFO, \ + VM_GET_INTINFO, \ + VM_RESTART_INSTRUCTION, \ + VM_SNAPSHOT_REQ, \ + VM_RESTORE_TIME + +const cap_ioctl_t vm_ioctl_cmds[] = { + VM_COMMON_IOCTLS, + VM_PPT_IOCTLS, + VM_MD_IOCTLS, +}; +size_t vm_ioctl_ncmds = nitems(vm_ioctl_cmds); + +int +vm_set_desc(struct vcpu *vcpu, int reg, + uint64_t base, uint32_t limit, uint32_t access) +{ + int error; + struct vm_seg_desc vmsegdesc; + + bzero(&vmsegdesc, sizeof(vmsegdesc)); + vmsegdesc.regnum = reg; + vmsegdesc.desc.base = base; + vmsegdesc.desc.limit = limit; + vmsegdesc.desc.access = access; + + error = vcpu_ioctl(vcpu, VM_SET_SEGMENT_DESCRIPTOR, &vmsegdesc); + return (error); +} + +int +vm_get_desc(struct vcpu *vcpu, int reg, uint64_t *base, uint32_t *limit, + uint32_t *access) +{ + int error; + struct vm_seg_desc vmsegdesc; + + bzero(&vmsegdesc, sizeof(vmsegdesc)); + vmsegdesc.regnum = reg; + + error = vcpu_ioctl(vcpu, VM_GET_SEGMENT_DESCRIPTOR, &vmsegdesc); + if (error == 0) { + *base = vmsegdesc.desc.base; + *limit = vmsegdesc.desc.limit; + *access = vmsegdesc.desc.access; + } + return (error); +} + +int +vm_get_seg_desc(struct vcpu *vcpu, int reg, struct seg_desc *seg_desc) +{ + int error; + + error = vm_get_desc(vcpu, reg, &seg_desc->base, &seg_desc->limit, + &seg_desc->access); + return (error); +} + +int +vm_lapic_irq(struct vcpu *vcpu, int vector) +{ + struct vm_lapic_irq vmirq; + + bzero(&vmirq, sizeof(vmirq)); + vmirq.vector = vector; + + return (vcpu_ioctl(vcpu, VM_LAPIC_IRQ, &vmirq)); +} + +int +vm_lapic_local_irq(struct vcpu *vcpu, int vector) +{ + struct vm_lapic_irq vmirq; + + bzero(&vmirq, sizeof(vmirq)); + vmirq.vector = vector; + + return (vcpu_ioctl(vcpu, VM_LAPIC_LOCAL_IRQ, &vmirq)); +} + +int +vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg) +{ + struct vm_lapic_msi vmmsi; + + bzero(&vmmsi, sizeof(vmmsi)); + vmmsi.addr = addr; + vmmsi.msg = msg; + + return (ioctl(ctx->fd, VM_LAPIC_MSI, &vmmsi)); +} + +int +vm_raise_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg, + int bus __unused, int slot __unused, int func __unused) +{ + return (vm_lapic_msi(ctx, addr, msg)); +} + +int +vm_apicid2vcpu(struct vmctx *ctx __unused, int apicid) +{ + /* + * The apic id associated with the 'vcpu' has the same numerical value + * as the 'vcpu' itself. + */ + return (apicid); +} + +int +vm_ioapic_assert_irq(struct vmctx *ctx, int irq) +{ + struct vm_ioapic_irq ioapic_irq; + + bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); + ioapic_irq.irq = irq; + + return (ioctl(ctx->fd, VM_IOAPIC_ASSERT_IRQ, &ioapic_irq)); +} + +int +vm_ioapic_deassert_irq(struct vmctx *ctx, int irq) +{ + struct vm_ioapic_irq ioapic_irq; + + bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); + ioapic_irq.irq = irq; + + return (ioctl(ctx->fd, VM_IOAPIC_DEASSERT_IRQ, &ioapic_irq)); +} + +int +vm_ioapic_pulse_irq(struct vmctx *ctx, int irq) +{ + struct vm_ioapic_irq ioapic_irq; + + bzero(&ioapic_irq, sizeof(struct vm_ioapic_irq)); + ioapic_irq.irq = irq; + + return (ioctl(ctx->fd, VM_IOAPIC_PULSE_IRQ, &ioapic_irq)); +} + +int +vm_ioapic_pincount(struct vmctx *ctx, int *pincount) +{ + + return (ioctl(ctx->fd, VM_IOAPIC_PINCOUNT, pincount)); +} + +int +vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) +{ + struct vm_isa_irq isa_irq; + + bzero(&isa_irq, sizeof(struct vm_isa_irq)); + isa_irq.atpic_irq = atpic_irq; + isa_irq.ioapic_irq = ioapic_irq; + + return (ioctl(ctx->fd, VM_ISA_ASSERT_IRQ, &isa_irq)); +} + +int +vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) +{ + struct vm_isa_irq isa_irq; + + bzero(&isa_irq, sizeof(struct vm_isa_irq)); + isa_irq.atpic_irq = atpic_irq; + isa_irq.ioapic_irq = ioapic_irq; + + return (ioctl(ctx->fd, VM_ISA_DEASSERT_IRQ, &isa_irq)); +} + +int +vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq) +{ + struct vm_isa_irq isa_irq; + + bzero(&isa_irq, sizeof(struct vm_isa_irq)); + isa_irq.atpic_irq = atpic_irq; + isa_irq.ioapic_irq = ioapic_irq; + + return (ioctl(ctx->fd, VM_ISA_PULSE_IRQ, &isa_irq)); +} + +int +vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq, + enum vm_intr_trigger trigger) +{ + struct vm_isa_irq_trigger isa_irq_trigger; + + bzero(&isa_irq_trigger, sizeof(struct vm_isa_irq_trigger)); + isa_irq_trigger.atpic_irq = atpic_irq; + isa_irq_trigger.trigger = trigger; + + return (ioctl(ctx->fd, VM_ISA_SET_IRQ_TRIGGER, &isa_irq_trigger)); +} + +int +vm_inject_nmi(struct vcpu *vcpu) +{ + struct vm_nmi vmnmi; + + bzero(&vmnmi, sizeof(vmnmi)); + + return (vcpu_ioctl(vcpu, VM_INJECT_NMI, &vmnmi)); +} + +int +vm_inject_exception(struct vcpu *vcpu, int vector, int errcode_valid, + uint32_t errcode, int restart_instruction) +{ + struct vm_exception exc; + + exc.vector = vector; + exc.error_code = errcode; + exc.error_code_valid = errcode_valid; + exc.restart_instruction = restart_instruction; + + return (vcpu_ioctl(vcpu, VM_INJECT_EXCEPTION, &exc)); +} + +int +vm_readwrite_kernemu_device(struct vcpu *vcpu, vm_paddr_t gpa, + bool write, int size, uint64_t *value) +{ + struct vm_readwrite_kernemu_device irp = { + .access_width = fls(size) - 1, + .gpa = gpa, + .value = write ? *value : ~0ul, + }; + long cmd = (write ? VM_SET_KERNEMU_DEV : VM_GET_KERNEMU_DEV); + int rc; + + rc = vcpu_ioctl(vcpu, cmd, &irp); + if (rc == 0 && !write) + *value = irp.value; + return (rc); +} + +int +vm_get_x2apic_state(struct vcpu *vcpu, enum x2apic_state *state) +{ + int error; + struct vm_x2apic x2apic; + + bzero(&x2apic, sizeof(x2apic)); + + error = vcpu_ioctl(vcpu, VM_GET_X2APIC_STATE, &x2apic); + *state = x2apic.state; + return (error); +} + +int +vm_set_x2apic_state(struct vcpu *vcpu, enum x2apic_state state) +{ + int error; + struct vm_x2apic x2apic; + + bzero(&x2apic, sizeof(x2apic)); + x2apic.state = state; + + error = vcpu_ioctl(vcpu, VM_SET_X2APIC_STATE, &x2apic); + + return (error); +} + +int +vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities) +{ + int error; + struct vm_hpet_cap cap; + + bzero(&cap, sizeof(struct vm_hpet_cap)); + error = ioctl(ctx->fd, VM_GET_HPET_CAPABILITIES, &cap); + if (capabilities != NULL) + *capabilities = cap.capabilities; + return (error); +} + +int +vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value) +{ + struct vm_rtc_data rtcdata; + int error; + + bzero(&rtcdata, sizeof(struct vm_rtc_data)); + rtcdata.offset = offset; + rtcdata.value = value; + error = ioctl(ctx->fd, VM_RTC_WRITE, &rtcdata); + return (error); +} + +int +vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval) +{ + struct vm_rtc_data rtcdata; + int error; + + bzero(&rtcdata, sizeof(struct vm_rtc_data)); + rtcdata.offset = offset; + error = ioctl(ctx->fd, VM_RTC_READ, &rtcdata); + if (error == 0) + *retval = rtcdata.value; + return (error); +} + +int +vm_rtc_settime(struct vmctx *ctx, time_t secs) +{ + struct vm_rtc_time rtctime; + int error; + + bzero(&rtctime, sizeof(struct vm_rtc_time)); + rtctime.secs = secs; + error = ioctl(ctx->fd, VM_RTC_SETTIME, &rtctime); + return (error); +} + +int +vm_rtc_gettime(struct vmctx *ctx, time_t *secs) +{ + struct vm_rtc_time rtctime; + int error; + + bzero(&rtctime, sizeof(struct vm_rtc_time)); + error = ioctl(ctx->fd, VM_RTC_GETTIME, &rtctime); + if (error == 0) + *secs = rtctime.secs; + return (error); +} + +/* + * From Intel Vol 3a: + * Table 9-1. IA-32 Processor States Following Power-up, Reset or INIT + */ +int +vcpu_reset(struct vcpu *vcpu) +{ + int error; + uint64_t rflags, rip, cr0, cr4, zero, desc_base, rdx; + uint32_t desc_access, desc_limit; + uint16_t sel; + + zero = 0; + + rflags = 0x2; + error = vm_set_register(vcpu, VM_REG_GUEST_RFLAGS, rflags); + if (error) + goto done; + + rip = 0xfff0; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RIP, rip)) != 0) + goto done; + + /* + * According to Intels Software Developer Manual CR0 should be + * initialized with CR0_ET | CR0_NW | CR0_CD but that crashes some + * guests like Windows. + */ + cr0 = CR0_NE; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CR0, cr0)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CR2, zero)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CR3, zero)) != 0) + goto done; + + cr4 = 0; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CR4, cr4)) != 0) + goto done; + + /* + * CS: present, r/w, accessed, 16-bit, byte granularity, usable + */ + desc_base = 0xffff0000; + desc_limit = 0xffff; + desc_access = 0x0093; + error = vm_set_desc(vcpu, VM_REG_GUEST_CS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + sel = 0xf000; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_CS, sel)) != 0) + goto done; + + /* + * SS,DS,ES,FS,GS: present, r/w, accessed, 16-bit, byte granularity + */ + desc_base = 0; + desc_limit = 0xffff; + desc_access = 0x0093; + error = vm_set_desc(vcpu, VM_REG_GUEST_SS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_DS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_ES, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_FS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_GS, + desc_base, desc_limit, desc_access); + if (error) + goto done; + + sel = 0; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_SS, sel)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_DS, sel)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_ES, sel)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_FS, sel)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_GS, sel)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_EFER, zero)) != 0) + goto done; + + /* General purpose registers */ + rdx = 0xf00; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RAX, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RBX, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RCX, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RDX, rdx)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RSI, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RDI, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RBP, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_RSP, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_R8, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_R9, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_R10, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_R11, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_R12, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_R13, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_R14, zero)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_R15, zero)) != 0) + goto done; + + /* GDTR, IDTR */ + desc_base = 0; + desc_limit = 0xffff; + desc_access = 0; + error = vm_set_desc(vcpu, VM_REG_GUEST_GDTR, + desc_base, desc_limit, desc_access); + if (error != 0) + goto done; + + error = vm_set_desc(vcpu, VM_REG_GUEST_IDTR, + desc_base, desc_limit, desc_access); + if (error != 0) + goto done; + + /* TR */ + desc_base = 0; + desc_limit = 0xffff; + desc_access = 0x0000008b; + error = vm_set_desc(vcpu, VM_REG_GUEST_TR, 0, 0, desc_access); + if (error) + goto done; + + sel = 0; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_TR, sel)) != 0) + goto done; + + /* LDTR */ + desc_base = 0; + desc_limit = 0xffff; + desc_access = 0x00000082; + error = vm_set_desc(vcpu, VM_REG_GUEST_LDTR, desc_base, + desc_limit, desc_access); + if (error) + goto done; + + sel = 0; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_LDTR, 0)) != 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_DR6, + 0xffff0ff0)) != 0) + goto done; + if ((error = vm_set_register(vcpu, VM_REG_GUEST_DR7, 0x400)) != + 0) + goto done; + + if ((error = vm_set_register(vcpu, VM_REG_GUEST_INTR_SHADOW, + zero)) != 0) + goto done; + + error = 0; +done: + return (error); +} diff --git a/lib/libvmmapi/internal.h b/lib/libvmmapi/internal.h new file mode 100644 index 000000000000..4afe1cab3460 --- /dev/null +++ b/lib/libvmmapi/internal.h @@ -0,0 +1,75 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2022 John Baldwin <jhb@FreeBSD.org> + */ + +#ifndef __VMMAPI_INTERNAL_H__ +#define __VMMAPI_INTERNAL_H__ + +#include <sys/types.h> +#include <dev/vmm/vmm_mem.h> + +struct vmctx { + int fd; /* device file descriptor */ + int ctlfd; /* vmm control descriptor */ + struct { + vm_paddr_t base; + vm_size_t size; + } memsegs[VM_MAX_MEMSEGS]; + size_t lowmem_size; + size_t highmem_size; + int memflags; + char *baseaddr; + char *name; +}; + +struct vcpu { + struct vmctx *ctx; + int vcpuid; +}; + +int vcpu_ioctl(struct vcpu *vcpu, u_long cmd, void *arg); + +extern const char *vm_capstrmap[]; + +#define VM_COMMON_IOCTLS \ + VM_RUN, \ + VM_SUSPEND, \ + VM_REINIT, \ + VM_ALLOC_MEMSEG, \ + VM_GET_MEMSEG, \ + VM_MMAP_MEMSEG, \ + VM_MMAP_MEMSEG, \ + VM_MMAP_GETNEXT, \ + VM_MUNMAP_MEMSEG, \ + VM_SET_REGISTER, \ + VM_GET_REGISTER, \ + VM_SET_REGISTER_SET, \ + VM_GET_REGISTER_SET, \ + VM_INJECT_EXCEPTION, \ + VM_SET_CAPABILITY, \ + VM_GET_CAPABILITY, \ + VM_STATS, \ + VM_STAT_DESC, \ + VM_GLA2GPA_NOFAULT, \ + VM_ACTIVATE_CPU, \ + VM_GET_CPUS, \ + VM_SUSPEND_CPU, \ + VM_RESUME_CPU, \ + VM_SET_TOPOLOGY, \ + VM_GET_TOPOLOGY + +#define VM_PPT_IOCTLS \ + VM_BIND_PPTDEV, \ + VM_UNBIND_PPTDEV, \ + VM_MAP_PPTDEV_MMIO, \ + VM_PPTDEV_MSI, \ + VM_PPTDEV_MSIX, \ + VM_UNMAP_PPTDEV_MMIO, \ + VM_PPTDEV_DISABLE_MSIX + +extern const cap_ioctl_t vm_ioctl_cmds[]; +extern size_t vm_ioctl_ncmds; + +#endif /* !__VMMAPI_INTERNAL_H__ */ diff --git a/lib/libvmmapi/ppt.c b/lib/libvmmapi/ppt.c new file mode 100644 index 000000000000..fd49f8eed168 --- /dev/null +++ b/lib/libvmmapi/ppt.c @@ -0,0 +1,144 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2011 NetApp, Inc. + * 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 NETAPP, INC ``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 NETAPP, INC 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/types.h> +#include <sys/ioctl.h> + +#include <machine/vmm.h> + +#include <string.h> + +#include "vmmapi.h" +#include "internal.h" + +int +vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func) +{ + struct vm_pptdev pptdev; + + bzero(&pptdev, sizeof(pptdev)); + pptdev.bus = bus; + pptdev.slot = slot; + pptdev.func = func; + + return (ioctl(ctx->fd, VM_BIND_PPTDEV, &pptdev)); +} + +int +vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func) +{ + struct vm_pptdev pptdev; + + bzero(&pptdev, sizeof(pptdev)); + pptdev.bus = bus; + pptdev.slot = slot; + pptdev.func = func; + + return (ioctl(ctx->fd, VM_UNBIND_PPTDEV, &pptdev)); +} + +int +vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func, + vm_paddr_t gpa, size_t len, vm_paddr_t hpa) +{ + struct vm_pptdev_mmio pptmmio; + + bzero(&pptmmio, sizeof(pptmmio)); + pptmmio.bus = bus; + pptmmio.slot = slot; + pptmmio.func = func; + pptmmio.gpa = gpa; + pptmmio.len = len; + pptmmio.hpa = hpa; + + return (ioctl(ctx->fd, VM_MAP_PPTDEV_MMIO, &pptmmio)); +} + +int +vm_unmap_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func, + vm_paddr_t gpa, size_t len) +{ + struct vm_pptdev_mmio pptmmio; + + bzero(&pptmmio, sizeof(pptmmio)); + pptmmio.bus = bus; + pptmmio.slot = slot; + pptmmio.func = func; + pptmmio.gpa = gpa; + pptmmio.len = len; + + return (ioctl(ctx->fd, VM_UNMAP_PPTDEV_MMIO, &pptmmio)); +} + +int +vm_setup_pptdev_msi(struct vmctx *ctx, int bus, int slot, int func, + uint64_t addr, uint64_t msg, int numvec) +{ + struct vm_pptdev_msi pptmsi; + + bzero(&pptmsi, sizeof(pptmsi)); + pptmsi.bus = bus; + pptmsi.slot = slot; + pptmsi.func = func; + pptmsi.msg = msg; + pptmsi.addr = addr; + pptmsi.numvec = numvec; + + return (ioctl(ctx->fd, VM_PPTDEV_MSI, &pptmsi)); +} + +int +vm_setup_pptdev_msix(struct vmctx *ctx, int bus, int slot, int func, + int idx, uint64_t addr, uint64_t msg, uint32_t vector_control) +{ + struct vm_pptdev_msix pptmsix; + + bzero(&pptmsix, sizeof(pptmsix)); + pptmsix.bus = bus; + pptmsix.slot = slot; + pptmsix.func = func; + pptmsix.idx = idx; + pptmsix.msg = msg; + pptmsix.addr = addr; + pptmsix.vector_control = vector_control; + + return (ioctl(ctx->fd, VM_PPTDEV_MSIX, &pptmsix)); +} + +int +vm_disable_pptdev_msix(struct vmctx *ctx, int bus, int slot, int func) +{ + struct vm_pptdev ppt; + + bzero(&ppt, sizeof(ppt)); + ppt.bus = bus; + ppt.slot = slot; + ppt.func = func; + + return (ioctl(ctx->fd, VM_PPTDEV_DISABLE_MSIX, &ppt)); +} diff --git a/lib/libvmmapi/riscv/Makefile.inc b/lib/libvmmapi/riscv/Makefile.inc new file mode 100644 index 000000000000..663ea0ab90a3 --- /dev/null +++ b/lib/libvmmapi/riscv/Makefile.inc @@ -0,0 +1 @@ +SRCS+= vmmapi_machdep.c diff --git a/lib/libvmmapi/riscv/vmmapi_machdep.c b/lib/libvmmapi/riscv/vmmapi_machdep.c new file mode 100644 index 000000000000..4da2fb909f61 --- /dev/null +++ b/lib/libvmmapi/riscv/vmmapi_machdep.c @@ -0,0 +1,118 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2011 NetApp, Inc. + * 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 NETAPP, INC ``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 NETAPP, INC 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/types.h> +#include <sys/ioctl.h> + +#include <machine/vmm.h> +#include <machine/vmm_dev.h> +#include <machine/vmm_snapshot.h> + +#include <assert.h> +#include <string.h> + +#include "vmmapi.h" +#include "internal.h" + +const char *vm_capstrmap[] = { + [VM_CAP_SSTC] = "sstc", + [VM_CAP_MAX] = NULL, +}; + +#define VM_MD_IOCTLS \ + VM_ATTACH_APLIC, \ + VM_ASSERT_IRQ, \ + VM_DEASSERT_IRQ, \ + VM_RAISE_MSI + +const cap_ioctl_t vm_ioctl_cmds[] = { + VM_COMMON_IOCTLS, + VM_MD_IOCTLS, +}; +size_t vm_ioctl_ncmds = nitems(vm_ioctl_cmds); + +int +vm_attach_aplic(struct vmctx *ctx, uint64_t mem_start, size_t mem_size) +{ + struct vm_aplic_descr aplic; + + bzero(&aplic, sizeof(aplic)); + aplic.mem_start = mem_start; + aplic.mem_size = mem_size; + + return (ioctl(ctx->fd, VM_ATTACH_APLIC, &aplic)); +} + +int +vm_assert_irq(struct vmctx *ctx, uint32_t irq) +{ + struct vm_irq vi; + + bzero(&vi, sizeof(vi)); + vi.irq = irq; + + return (ioctl(ctx->fd, VM_ASSERT_IRQ, &vi)); +} + +int +vm_deassert_irq(struct vmctx *ctx, uint32_t irq) +{ + struct vm_irq vi; + + bzero(&vi, sizeof(vi)); + vi.irq = irq; + + return (ioctl(ctx->fd, VM_DEASSERT_IRQ, &vi)); +} + +int +vm_raise_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg, + int bus, int slot, int func) +{ + struct vm_msi vmsi; + + bzero(&vmsi, sizeof(vmsi)); + vmsi.addr = addr; + vmsi.msg = msg; + vmsi.bus = bus; + vmsi.slot = slot; + vmsi.func = func; + + return (ioctl(ctx->fd, VM_RAISE_MSI, &vmsi)); +} + +int +vm_inject_exception(struct vcpu *vcpu, uint64_t scause) +{ + struct vm_exception vmexc; + + bzero(&vmexc, sizeof(vmexc)); + vmexc.scause = scause; + + return (vcpu_ioctl(vcpu, VM_INJECT_EXCEPTION, &vmexc)); +} diff --git a/lib/libvmmapi/vmmapi.c b/lib/libvmmapi/vmmapi.c new file mode 100644 index 000000000000..77f0f8f5c581 --- /dev/null +++ b/lib/libvmmapi/vmmapi.c @@ -0,0 +1,1319 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2011 NetApp, Inc. + * 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 NETAPP, INC ``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 NETAPP, INC 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/capsicum.h> +#include <sys/cpuset.h> +#include <sys/domainset.h> +#include <sys/sysctl.h> +#include <sys/ioctl.h> +#include <sys/mman.h> +#include <sys/linker.h> +#include <sys/module.h> +#include <sys/_iovec.h> + +#include <capsicum_helpers.h> +#include <err.h> +#include <errno.h> +#include <stdbool.h> +#include <stdio.h> +#include <stdlib.h> +#include <assert.h> +#include <string.h> +#include <fcntl.h> +#include <unistd.h> + +#include <libutil.h> + +#include <vm/vm.h> +#include <machine/vmm.h> +#ifdef WITH_VMMAPI_SNAPSHOT +#include <machine/vmm_snapshot.h> +#endif + +#include <dev/vmm/vmm_dev.h> + +#include "vmmapi.h" +#include "internal.h" + +#define MB (1024 * 1024UL) +#define GB (1024 * 1024 * 1024UL) + +#ifdef __amd64__ +#define VM_LOWMEM_LIMIT (3 * GB) +#else +#define VM_LOWMEM_LIMIT 0 +#endif +#define VM_HIGHMEM_BASE (4 * GB) + +/* + * Size of the guard region before and after the virtual address space + * mapping the guest physical memory. This must be a multiple of the + * superpage size for performance reasons. + */ +#define VM_MMAP_GUARD_SIZE (4 * MB) + +#define PROT_RW (PROT_READ | PROT_WRITE) +#define PROT_ALL (PROT_READ | PROT_WRITE | PROT_EXEC) + +static int +vm_device_open(const char *name) +{ + char devpath[PATH_MAX]; + + assert(strlen(name) <= VM_MAX_NAMELEN); + (void)snprintf(devpath, sizeof(devpath), "/dev/vmm/%s", name); + return (open(devpath, O_RDWR)); +} + +static int +vm_ctl_open(void) +{ + if (modfind("vmm") < 0) + (void)kldload("vmm"); + return (open("/dev/vmmctl", O_RDWR, 0)); +} + +static int +vm_ctl_create(const char *name, int ctlfd) +{ + struct vmmctl_vm_create vmc; + + memset(&vmc, 0, sizeof(vmc)); + if (strlcpy(vmc.name, name, sizeof(vmc.name)) >= sizeof(vmc.name)) { + errno = ENAMETOOLONG; + return (-1); + } + return (ioctl(ctlfd, VMMCTL_VM_CREATE, &vmc)); +} + +int +vm_create(const char *name) +{ + int error, fd; + + fd = vm_ctl_open(); + if (fd < 0) + return (-1); + + error = vm_ctl_create(name, fd); + if (error != 0) { + error = errno; + (void)close(fd); + errno = error; + return (-1); + } + (void)close(fd); + return (0); +} + +struct vmctx * +vm_open(const char *name) +{ + return (vm_openf(name, 0)); +} + +struct vmctx * +vm_openf(const char *name, int flags) +{ + struct vmctx *vm; + int saved_errno; + bool created; + + created = false; + + vm = malloc(sizeof(struct vmctx) + strlen(name) + 1); + assert(vm != NULL); + + vm->fd = vm->ctlfd = -1; + vm->memflags = 0; + vm->name = (char *)(vm + 1); + strcpy(vm->name, name); + memset(vm->memsegs, 0, sizeof(vm->memsegs)); + + if ((vm->ctlfd = vm_ctl_open()) < 0) + goto err; + + vm->fd = vm_device_open(vm->name); + if (vm->fd < 0 && errno == ENOENT) { + if (flags & VMMAPI_OPEN_CREATE) { + if (vm_ctl_create(vm->name, vm->ctlfd) != 0) + goto err; + vm->fd = vm_device_open(vm->name); + created = true; + } + } + if (vm->fd < 0) + goto err; + + if (!created && (flags & VMMAPI_OPEN_REINIT) != 0 && vm_reinit(vm) != 0) + goto err; + + return (vm); +err: + saved_errno = errno; + if (created) + vm_destroy(vm); + else + vm_close(vm); + errno = saved_errno; + return (NULL); +} + +void +vm_close(struct vmctx *vm) +{ + assert(vm != NULL); + + if (vm->fd >= 0) + (void)close(vm->fd); + if (vm->ctlfd >= 0) + (void)close(vm->ctlfd); + free(vm); +} + +void +vm_destroy(struct vmctx *vm) +{ + struct vmmctl_vm_destroy vmd; + + memset(&vmd, 0, sizeof(vmd)); + (void)strlcpy(vmd.name, vm->name, sizeof(vmd.name)); + if (ioctl(vm->ctlfd, VMMCTL_VM_DESTROY, &vmd) != 0) + warn("ioctl(VMMCTL_VM_DESTROY)"); + + vm_close(vm); +} + +struct vcpu * +vm_vcpu_open(struct vmctx *ctx, int vcpuid) +{ + struct vcpu *vcpu; + + vcpu = malloc(sizeof(*vcpu)); + vcpu->ctx = ctx; + vcpu->vcpuid = vcpuid; + return (vcpu); +} + +void +vm_vcpu_close(struct vcpu *vcpu) +{ + free(vcpu); +} + +int +vcpu_id(struct vcpu *vcpu) +{ + return (vcpu->vcpuid); +} + +int +vm_parse_memsize(const char *opt, size_t *ret_memsize) +{ + char *endptr; + size_t optval; + int error; + + optval = strtoul(opt, &endptr, 0); + if (*opt != '\0' && *endptr == '\0') { + /* + * For the sake of backward compatibility if the memory size + * specified on the command line is less than a megabyte then + * it is interpreted as being in units of MB. + */ + if (optval < MB) + optval *= MB; + *ret_memsize = optval; + error = 0; + } else + error = expand_number(opt, ret_memsize); + + return (error); +} + +uint32_t +vm_get_lowmem_limit(struct vmctx *ctx __unused) +{ + + return (VM_LOWMEM_LIMIT); +} + +void +vm_set_memflags(struct vmctx *ctx, int flags) +{ + + ctx->memflags = flags; +} + +int +vm_get_memflags(struct vmctx *ctx) +{ + + return (ctx->memflags); +} + +/* + * Map segment 'segid' starting at 'off' into guest address range [gpa,gpa+len). + */ +int +vm_mmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, int segid, vm_ooffset_t off, + size_t len, int prot) +{ + struct vm_memmap memmap; + int error, flags; + + memmap.gpa = gpa; + memmap.segid = segid; + memmap.segoff = off; + memmap.len = len; + memmap.prot = prot; + memmap.flags = 0; + + if (ctx->memflags & VM_MEM_F_WIRED) + memmap.flags |= VM_MEMMAP_F_WIRED; + + /* + * If this mapping already exists then don't create it again. This + * is the common case for SYSMEM mappings created by bhyveload(8). + */ + error = vm_mmap_getnext(ctx, &gpa, &segid, &off, &len, &prot, &flags); + if (error == 0 && gpa == memmap.gpa) { + if (segid != memmap.segid || off != memmap.segoff || + prot != memmap.prot || flags != memmap.flags) { + errno = EEXIST; + return (-1); + } else { + return (0); + } + } + + error = ioctl(ctx->fd, VM_MMAP_MEMSEG, &memmap); + return (error); +} + +int +vm_get_guestmem_from_ctx(struct vmctx *ctx, char **guest_baseaddr, + size_t *lowmem_size, size_t *highmem_size) +{ + + *guest_baseaddr = ctx->baseaddr; + *lowmem_size = ctx->lowmem_size; + *highmem_size = ctx->highmem_size; + return (0); +} + +int +vm_munmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, size_t len) +{ + struct vm_munmap munmap; + int error; + + munmap.gpa = gpa; + munmap.len = len; + + error = ioctl(ctx->fd, VM_MUNMAP_MEMSEG, &munmap); + return (error); +} + +int +vm_mmap_getnext(struct vmctx *ctx, vm_paddr_t *gpa, int *segid, + vm_ooffset_t *segoff, size_t *len, int *prot, int *flags) +{ + struct vm_memmap memmap; + int error; + + bzero(&memmap, sizeof(struct vm_memmap)); + memmap.gpa = *gpa; + error = ioctl(ctx->fd, VM_MMAP_GETNEXT, &memmap); + if (error == 0) { + *gpa = memmap.gpa; + *segid = memmap.segid; + *segoff = memmap.segoff; + *len = memmap.len; + *prot = memmap.prot; + *flags = memmap.flags; + } + return (error); +} + +/* + * Return 0 if the segments are identical and non-zero otherwise. + * + * This is slightly complicated by the fact that only device memory segments + * are named. + */ +static int +cmpseg(size_t len, const char *str, size_t len2, const char *str2) +{ + + if (len == len2) { + if ((!str && !str2) || (str && str2 && !strcmp(str, str2))) + return (0); + } + return (-1); +} + +static int +vm_alloc_memseg(struct vmctx *ctx, int segid, size_t len, const char *name, + int ds_policy, domainset_t *ds_mask, size_t ds_size) +{ + struct vm_memseg memseg; + size_t n; + int error; + + /* + * If the memory segment has already been created then just return. + * This is the usual case for the SYSMEM segment created by userspace + * loaders like bhyveload(8). + */ + error = vm_get_memseg(ctx, segid, &memseg.len, memseg.name, + sizeof(memseg.name)); + if (error) + return (error); + + if (memseg.len != 0) { + if (cmpseg(len, name, memseg.len, VM_MEMSEG_NAME(&memseg))) { + errno = EINVAL; + return (-1); + } else { + return (0); + } + } + + bzero(&memseg, sizeof(struct vm_memseg)); + memseg.segid = segid; + memseg.len = len; + if (ds_mask == NULL) { + memseg.ds_policy = DOMAINSET_POLICY_INVALID; + } else { + memseg.ds_policy = ds_policy; + memseg.ds_mask = ds_mask; + memseg.ds_mask_size = ds_size; + } + if (name != NULL) { + n = strlcpy(memseg.name, name, sizeof(memseg.name)); + if (n >= sizeof(memseg.name)) { + errno = ENAMETOOLONG; + return (-1); + } + } + + error = ioctl(ctx->fd, VM_ALLOC_MEMSEG, &memseg); + return (error); +} + +int +vm_get_memseg(struct vmctx *ctx, int segid, size_t *lenp, char *namebuf, + size_t bufsize) +{ + struct vm_memseg memseg; + size_t n; + int error; + + bzero(&memseg, sizeof(memseg)); + memseg.segid = segid; + error = ioctl(ctx->fd, VM_GET_MEMSEG, &memseg); + if (error == 0) { + *lenp = memseg.len; + n = strlcpy(namebuf, memseg.name, bufsize); + if (n >= bufsize) { + errno = ENAMETOOLONG; + error = -1; + } + } + return (error); +} + +static int +map_memory_segment(struct vmctx *ctx, int segid, vm_paddr_t gpa, size_t len, + size_t segoff, char *base) +{ + char *ptr; + int error, flags; + + /* Map 'len' bytes starting at 'gpa' in the guest address space */ + error = vm_mmap_memseg(ctx, gpa, segid, segoff, len, PROT_ALL); + if (error) + return (error); + + flags = MAP_SHARED | MAP_FIXED; + if ((ctx->memflags & VM_MEM_F_INCORE) == 0) + flags |= MAP_NOCORE; + + /* mmap into the process address space on the host */ + ptr = mmap(base + gpa, len, PROT_RW, flags, ctx->fd, gpa); + if (ptr == MAP_FAILED) + return (-1); + + return (0); +} + +/* + * Allocates and maps virtual machine memory segments according + * to the NUMA topology specified by the 'doms' array. + * + * The domains are laid out sequentially in the guest's physical address space. + * The [VM_LOWMEM_LIMIT, VM_HIGHMEM_BASE) address range is skipped and + * left unmapped. + */ +int +vm_setup_memory_domains(struct vmctx *ctx, enum vm_mmap_style vms, + struct vm_mem_domain *doms, int ndoms) +{ + size_t low_len, len, totalsize; + struct vm_mem_domain *dom; + struct vm_memseg memseg; + char *baseaddr, *ptr; + int error, i, segid; + vm_paddr_t gpa; + + /* Sanity checks. */ + assert(vms == VM_MMAP_ALL); + if (doms == NULL || ndoms <= 0 || ndoms > VM_MAXMEMDOM) { + errno = EINVAL; + return (-1); + } + + /* Calculate total memory size. */ + totalsize = 0; + for (i = 0; i < ndoms; i++) + totalsize += doms[i].size; + + if (totalsize > VM_LOWMEM_LIMIT) + totalsize = VM_HIGHMEM_BASE + (totalsize - VM_LOWMEM_LIMIT); + + /* + * Stake out a contiguous region covering the guest physical memory + * and the adjoining guard regions. + */ + len = VM_MMAP_GUARD_SIZE + totalsize + VM_MMAP_GUARD_SIZE; + ptr = mmap(NULL, len, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1, 0); + if (ptr == MAP_FAILED) + return (-1); + baseaddr = ptr + VM_MMAP_GUARD_SIZE; + + /* + * Allocate and map memory segments for the virtual machine. + */ + gpa = VM_LOWMEM_LIMIT > 0 ? 0 : VM_HIGHMEM_BASE; + ctx->lowmem_size = 0; + ctx->highmem_size = 0; + for (i = 0; i < ndoms; i++) { + segid = VM_SYSMEM + i; + dom = &doms[i]; + + /* + * Check if the memory segment already exists. + * If 'ndoms' is greater than one, refuse to proceed if the + * memseg already exists. If only one domain was requested, use + * the existing segment to preserve the behaviour of the previous + * implementation. + * + * Splitting existing memory segments is tedious and + * error-prone, which is why we don't support NUMA + * domains for bhyveload(8)-loaded VMs. + */ + error = vm_get_memseg(ctx, segid, &len, memseg.name, + sizeof(memseg.name)); + if (error == 0 && len != 0) { + if (ndoms != 1) { + errno = EEXIST; + return (-1); + } else + doms[0].size = len; + } else { + error = vm_alloc_memseg(ctx, segid, dom->size, NULL, + dom->ds_policy, dom->ds_mask, dom->ds_size); + if (error) + return (error); + } + + /* + * If a domain is split by VM_LOWMEM_LIMIT then break + * its segment mapping into two parts, one below VM_LOWMEM_LIMIT + * and one above VM_HIGHMEM_BASE. + */ + if (gpa <= VM_LOWMEM_LIMIT && + gpa + dom->size > VM_LOWMEM_LIMIT) { + low_len = VM_LOWMEM_LIMIT - gpa; + error = map_memory_segment(ctx, segid, gpa, low_len, 0, + baseaddr); + if (error) + return (error); + ctx->lowmem_size = VM_LOWMEM_LIMIT; + /* Map the remainder. */ + gpa = VM_HIGHMEM_BASE; + len = dom->size - low_len; + error = map_memory_segment(ctx, segid, gpa, len, + low_len, baseaddr); + if (error) + return (error); + } else { + len = dom->size; + error = map_memory_segment(ctx, segid, gpa, len, 0, + baseaddr); + if (error) + return (error); + } + if (gpa <= VM_LOWMEM_LIMIT) + ctx->lowmem_size += len; + else + ctx->highmem_size += len; + gpa += len; + } + ctx->baseaddr = baseaddr; + + return (0); +} + +int +vm_setup_memory(struct vmctx *ctx, size_t memsize, enum vm_mmap_style vms) +{ + struct vm_mem_domain dom0; + + memset(&dom0, 0, sizeof(dom0)); + dom0.ds_policy = DOMAINSET_POLICY_INVALID; + dom0.size = memsize; + + return (vm_setup_memory_domains(ctx, vms, &dom0, 1)); +} + +/* + * Returns a non-NULL pointer if [gaddr, gaddr+len) is entirely contained in + * the lowmem or highmem regions. + * + * In particular return NULL if [gaddr, gaddr+len) falls in guest MMIO region. + * The instruction emulation code depends on this behavior. + */ +void * +vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len) +{ + vm_size_t lowsize, highsize; + + lowsize = ctx->lowmem_size; + if (lowsize > 0) { + if (gaddr < lowsize && len <= lowsize && gaddr + len <= lowsize) + return (ctx->baseaddr + gaddr); + } + + highsize = ctx->highmem_size; + if (highsize > 0 && gaddr >= VM_HIGHMEM_BASE) { + if (gaddr < VM_HIGHMEM_BASE + highsize && len <= highsize && + gaddr + len <= VM_HIGHMEM_BASE + highsize) + return (ctx->baseaddr + gaddr); + } + + return (NULL); +} + +vm_paddr_t +vm_rev_map_gpa(struct vmctx *ctx, void *addr) +{ + vm_paddr_t offaddr; + vm_size_t lowsize, highsize; + + offaddr = (char *)addr - ctx->baseaddr; + + lowsize = ctx->lowmem_size; + if (lowsize > 0) + if (offaddr <= lowsize) + return (offaddr); + + highsize = ctx->highmem_size; + if (highsize > 0) + if (offaddr >= VM_HIGHMEM_BASE && + offaddr < VM_HIGHMEM_BASE + highsize) + return (offaddr); + + return ((vm_paddr_t)-1); +} + +const char * +vm_get_name(struct vmctx *ctx) +{ + + return (ctx->name); +} + +size_t +vm_get_lowmem_size(struct vmctx *ctx) +{ + return (ctx->lowmem_size); +} + +vm_paddr_t +vm_get_highmem_base(struct vmctx *ctx __unused) +{ + + return (VM_HIGHMEM_BASE); +} + +size_t +vm_get_highmem_size(struct vmctx *ctx) +{ + return (ctx->highmem_size); +} + +void * +vm_create_devmem(struct vmctx *ctx, int segid, const char *name, size_t len) +{ + char pathname[MAXPATHLEN]; + size_t len2; + char *base, *ptr; + int fd, error, flags; + + fd = -1; + ptr = MAP_FAILED; + if (name == NULL || strlen(name) == 0) { + errno = EINVAL; + goto done; + } + + error = vm_alloc_memseg(ctx, segid, len, name, 0, NULL, 0); + if (error) + goto done; + + strlcpy(pathname, "/dev/vmm.io/", sizeof(pathname)); + strlcat(pathname, ctx->name, sizeof(pathname)); + strlcat(pathname, ".", sizeof(pathname)); + strlcat(pathname, name, sizeof(pathname)); + + fd = open(pathname, O_RDWR); + if (fd < 0) + goto done; + + /* + * Stake out a contiguous region covering the device memory and the + * adjoining guard regions. + */ + len2 = VM_MMAP_GUARD_SIZE + len + VM_MMAP_GUARD_SIZE; + base = mmap(NULL, len2, PROT_NONE, MAP_GUARD | MAP_ALIGNED_SUPER, -1, + 0); + if (base == MAP_FAILED) + goto done; + + flags = MAP_SHARED | MAP_FIXED; + if ((ctx->memflags & VM_MEM_F_INCORE) == 0) + flags |= MAP_NOCORE; + + /* mmap the devmem region in the host address space */ + ptr = mmap(base + VM_MMAP_GUARD_SIZE, len, PROT_RW, flags, fd, 0); +done: + if (fd >= 0) + close(fd); + return (ptr); +} + +int +vcpu_ioctl(struct vcpu *vcpu, u_long cmd, void *arg) +{ + /* + * XXX: fragile, handle with care + * Assumes that the first field of the ioctl data + * is the vcpuid. + */ + *(int *)arg = vcpu->vcpuid; + return (ioctl(vcpu->ctx->fd, cmd, arg)); +} + +int +vm_set_register(struct vcpu *vcpu, int reg, uint64_t val) +{ + int error; + struct vm_register vmreg; + + bzero(&vmreg, sizeof(vmreg)); + vmreg.regnum = reg; + vmreg.regval = val; + + error = vcpu_ioctl(vcpu, VM_SET_REGISTER, &vmreg); + return (error); +} + +int +vm_get_register(struct vcpu *vcpu, int reg, uint64_t *ret_val) +{ + int error; + struct vm_register vmreg; + + bzero(&vmreg, sizeof(vmreg)); + vmreg.regnum = reg; + + error = vcpu_ioctl(vcpu, VM_GET_REGISTER, &vmreg); + *ret_val = vmreg.regval; + return (error); +} + +int +vm_set_register_set(struct vcpu *vcpu, unsigned int count, + const int *regnums, uint64_t *regvals) +{ + int error; + struct vm_register_set vmregset; + + bzero(&vmregset, sizeof(vmregset)); + vmregset.count = count; + vmregset.regnums = regnums; + vmregset.regvals = regvals; + + error = vcpu_ioctl(vcpu, VM_SET_REGISTER_SET, &vmregset); + return (error); +} + +int +vm_get_register_set(struct vcpu *vcpu, unsigned int count, + const int *regnums, uint64_t *regvals) +{ + int error; + struct vm_register_set vmregset; + + bzero(&vmregset, sizeof(vmregset)); + vmregset.count = count; + vmregset.regnums = regnums; + vmregset.regvals = regvals; + + error = vcpu_ioctl(vcpu, VM_GET_REGISTER_SET, &vmregset); + return (error); +} + +int +vm_run(struct vcpu *vcpu, struct vm_run *vmrun) +{ + return (vcpu_ioctl(vcpu, VM_RUN, vmrun)); +} + +int +vm_suspend(struct vmctx *ctx, enum vm_suspend_how how) +{ + struct vm_suspend vmsuspend; + + bzero(&vmsuspend, sizeof(vmsuspend)); + vmsuspend.how = how; + return (ioctl(ctx->fd, VM_SUSPEND, &vmsuspend)); +} + +int +vm_reinit(struct vmctx *ctx) +{ + + return (ioctl(ctx->fd, VM_REINIT, 0)); +} + +int +vm_capability_name2type(const char *capname) +{ + int i; + + for (i = 0; i < VM_CAP_MAX; i++) { + if (vm_capstrmap[i] != NULL && + strcmp(vm_capstrmap[i], capname) == 0) + return (i); + } + + return (-1); +} + +const char * +vm_capability_type2name(int type) +{ + if (type >= 0 && type < VM_CAP_MAX) + return (vm_capstrmap[type]); + + return (NULL); +} + +int +vm_get_capability(struct vcpu *vcpu, enum vm_cap_type cap, int *retval) +{ + int error; + struct vm_capability vmcap; + + bzero(&vmcap, sizeof(vmcap)); + vmcap.captype = cap; + + error = vcpu_ioctl(vcpu, VM_GET_CAPABILITY, &vmcap); + *retval = vmcap.capval; + return (error); +} + +int +vm_set_capability(struct vcpu *vcpu, enum vm_cap_type cap, int val) +{ + struct vm_capability vmcap; + + bzero(&vmcap, sizeof(vmcap)); + vmcap.captype = cap; + vmcap.capval = val; + + return (vcpu_ioctl(vcpu, VM_SET_CAPABILITY, &vmcap)); +} + +uint64_t * +vm_get_stats(struct vcpu *vcpu, struct timeval *ret_tv, + int *ret_entries) +{ + static _Thread_local uint64_t *stats_buf; + static _Thread_local u_int stats_count; + uint64_t *new_stats; + struct vm_stats vmstats; + u_int count, index; + bool have_stats; + + have_stats = false; + count = 0; + for (index = 0;; index += nitems(vmstats.statbuf)) { + vmstats.index = index; + if (vcpu_ioctl(vcpu, VM_STATS, &vmstats) != 0) + break; + if (stats_count < index + vmstats.num_entries) { + new_stats = realloc(stats_buf, + (index + vmstats.num_entries) * sizeof(uint64_t)); + if (new_stats == NULL) { + errno = ENOMEM; + return (NULL); + } + stats_count = index + vmstats.num_entries; + stats_buf = new_stats; + } + memcpy(stats_buf + index, vmstats.statbuf, + vmstats.num_entries * sizeof(uint64_t)); + count += vmstats.num_entries; + have_stats = true; + + if (vmstats.num_entries != nitems(vmstats.statbuf)) + break; + } + if (have_stats) { + if (ret_entries) + *ret_entries = count; + if (ret_tv) + *ret_tv = vmstats.tv; + return (stats_buf); + } else + return (NULL); +} + +const char * +vm_get_stat_desc(struct vmctx *ctx, int index) +{ + static struct vm_stat_desc statdesc; + + statdesc.index = index; + if (ioctl(ctx->fd, VM_STAT_DESC, &statdesc) == 0) + return (statdesc.desc); + else + return (NULL); +} + +#ifdef __amd64__ +int +vm_get_gpa_pmap(struct vmctx *ctx, uint64_t gpa, uint64_t *pte, int *num) +{ + int error, i; + struct vm_gpa_pte gpapte; + + bzero(&gpapte, sizeof(gpapte)); + gpapte.gpa = gpa; + + error = ioctl(ctx->fd, VM_GET_GPA_PMAP, &gpapte); + + if (error == 0) { + *num = gpapte.ptenum; + for (i = 0; i < gpapte.ptenum; i++) + pte[i] = gpapte.pte[i]; + } + + return (error); +} + +int +vm_gla2gpa(struct vcpu *vcpu, struct vm_guest_paging *paging, + uint64_t gla, int prot, uint64_t *gpa, int *fault) +{ + struct vm_gla2gpa gg; + int error; + + bzero(&gg, sizeof(struct vm_gla2gpa)); + gg.prot = prot; + gg.gla = gla; + gg.paging = *paging; + + error = vcpu_ioctl(vcpu, VM_GLA2GPA, &gg); + if (error == 0) { + *fault = gg.fault; + *gpa = gg.gpa; + } + return (error); +} +#endif + +int +vm_gla2gpa_nofault(struct vcpu *vcpu, struct vm_guest_paging *paging, + uint64_t gla, int prot, uint64_t *gpa, int *fault) +{ + struct vm_gla2gpa gg; + int error; + + bzero(&gg, sizeof(struct vm_gla2gpa)); + gg.prot = prot; + gg.gla = gla; + gg.paging = *paging; + + error = vcpu_ioctl(vcpu, VM_GLA2GPA_NOFAULT, &gg); + if (error == 0) { + *fault = gg.fault; + *gpa = gg.gpa; + } + return (error); +} + +#ifndef min +#define min(a,b) (((a) < (b)) ? (a) : (b)) +#endif + +#ifdef __amd64__ +int +vm_copy_setup(struct vcpu *vcpu, struct vm_guest_paging *paging, + uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt, + int *fault) +{ + void *va; + uint64_t gpa, off; + int error, i, n; + + for (i = 0; i < iovcnt; i++) { + iov[i].iov_base = 0; + iov[i].iov_len = 0; + } + + while (len) { + assert(iovcnt > 0); + error = vm_gla2gpa(vcpu, paging, gla, prot, &gpa, fault); + if (error || *fault) + return (error); + + off = gpa & PAGE_MASK; + n = MIN(len, PAGE_SIZE - off); + + va = vm_map_gpa(vcpu->ctx, gpa, n); + if (va == NULL) + return (EFAULT); + + iov->iov_base = va; + iov->iov_len = n; + iov++; + iovcnt--; + + gla += n; + len -= n; + } + return (0); +} +#endif + +void +vm_copy_teardown(struct iovec *iov __unused, int iovcnt __unused) +{ + /* + * Intentionally empty. This is used by the instruction + * emulation code shared with the kernel. The in-kernel + * version of this is non-empty. + */ +} + +void +vm_copyin(struct iovec *iov, void *vp, size_t len) +{ + const char *src; + char *dst; + size_t n; + + dst = vp; + while (len) { + assert(iov->iov_len); + n = min(len, iov->iov_len); + src = iov->iov_base; + bcopy(src, dst, n); + + iov++; + dst += n; + len -= n; + } +} + +void +vm_copyout(const void *vp, struct iovec *iov, size_t len) +{ + const char *src; + char *dst; + size_t n; + + src = vp; + while (len) { + assert(iov->iov_len); + n = min(len, iov->iov_len); + dst = iov->iov_base; + bcopy(src, dst, n); + + iov++; + src += n; + len -= n; + } +} + +static int +vm_get_cpus(struct vmctx *ctx, int which, cpuset_t *cpus) +{ + struct vm_cpuset vm_cpuset; + int error; + + bzero(&vm_cpuset, sizeof(struct vm_cpuset)); + vm_cpuset.which = which; + vm_cpuset.cpusetsize = sizeof(cpuset_t); + vm_cpuset.cpus = cpus; + + error = ioctl(ctx->fd, VM_GET_CPUS, &vm_cpuset); + return (error); +} + +int +vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus) +{ + + return (vm_get_cpus(ctx, VM_ACTIVE_CPUS, cpus)); +} + +int +vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus) +{ + + return (vm_get_cpus(ctx, VM_SUSPENDED_CPUS, cpus)); +} + +int +vm_debug_cpus(struct vmctx *ctx, cpuset_t *cpus) +{ + + return (vm_get_cpus(ctx, VM_DEBUG_CPUS, cpus)); +} + +int +vm_activate_cpu(struct vcpu *vcpu) +{ + struct vm_activate_cpu ac; + int error; + + bzero(&ac, sizeof(struct vm_activate_cpu)); + error = vcpu_ioctl(vcpu, VM_ACTIVATE_CPU, &ac); + return (error); +} + +int +vm_suspend_all_cpus(struct vmctx *ctx) +{ + struct vm_activate_cpu ac; + int error; + + bzero(&ac, sizeof(struct vm_activate_cpu)); + ac.vcpuid = -1; + error = ioctl(ctx->fd, VM_SUSPEND_CPU, &ac); + return (error); +} + +int +vm_suspend_cpu(struct vcpu *vcpu) +{ + struct vm_activate_cpu ac; + int error; + + bzero(&ac, sizeof(struct vm_activate_cpu)); + error = vcpu_ioctl(vcpu, VM_SUSPEND_CPU, &ac); + return (error); +} + +int +vm_resume_cpu(struct vcpu *vcpu) +{ + struct vm_activate_cpu ac; + int error; + + bzero(&ac, sizeof(struct vm_activate_cpu)); + error = vcpu_ioctl(vcpu, VM_RESUME_CPU, &ac); + return (error); +} + +int +vm_resume_all_cpus(struct vmctx *ctx) +{ + struct vm_activate_cpu ac; + int error; + + bzero(&ac, sizeof(struct vm_activate_cpu)); + ac.vcpuid = -1; + error = ioctl(ctx->fd, VM_RESUME_CPU, &ac); + return (error); +} + +#ifdef __amd64__ +int +vm_get_intinfo(struct vcpu *vcpu, uint64_t *info1, uint64_t *info2) +{ + struct vm_intinfo vmii; + int error; + + bzero(&vmii, sizeof(struct vm_intinfo)); + error = vcpu_ioctl(vcpu, VM_GET_INTINFO, &vmii); + if (error == 0) { + *info1 = vmii.info1; + *info2 = vmii.info2; + } + return (error); +} + +int +vm_set_intinfo(struct vcpu *vcpu, uint64_t info1) +{ + struct vm_intinfo vmii; + int error; + + bzero(&vmii, sizeof(struct vm_intinfo)); + vmii.info1 = info1; + error = vcpu_ioctl(vcpu, VM_SET_INTINFO, &vmii); + return (error); +} +#endif + +#ifdef WITH_VMMAPI_SNAPSHOT +int +vm_restart_instruction(struct vcpu *vcpu) +{ + int arg; + + return (vcpu_ioctl(vcpu, VM_RESTART_INSTRUCTION, &arg)); +} + +int +vm_snapshot_req(struct vmctx *ctx, struct vm_snapshot_meta *meta) +{ + + if (ioctl(ctx->fd, VM_SNAPSHOT_REQ, meta) == -1) { +#ifdef SNAPSHOT_DEBUG + fprintf(stderr, "%s: snapshot failed for %s: %d\r\n", + __func__, meta->dev_name, errno); +#endif + return (-1); + } + return (0); +} + +int +vm_restore_time(struct vmctx *ctx) +{ + int dummy; + + dummy = 0; + return (ioctl(ctx->fd, VM_RESTORE_TIME, &dummy)); +} +#endif + +int +vm_set_topology(struct vmctx *ctx, + uint16_t sockets, uint16_t cores, uint16_t threads, uint16_t maxcpus) +{ + struct vm_cpu_topology topology; + + bzero(&topology, sizeof (struct vm_cpu_topology)); + topology.sockets = sockets; + topology.cores = cores; + topology.threads = threads; + topology.maxcpus = maxcpus; + return (ioctl(ctx->fd, VM_SET_TOPOLOGY, &topology)); +} + +int +vm_get_topology(struct vmctx *ctx, + uint16_t *sockets, uint16_t *cores, uint16_t *threads, uint16_t *maxcpus) +{ + struct vm_cpu_topology topology; + int error; + + bzero(&topology, sizeof (struct vm_cpu_topology)); + error = ioctl(ctx->fd, VM_GET_TOPOLOGY, &topology); + if (error == 0) { + *sockets = topology.sockets; + *cores = topology.cores; + *threads = topology.threads; + *maxcpus = topology.maxcpus; + } + return (error); +} + +int +vm_limit_rights(struct vmctx *ctx) +{ + cap_rights_t rights; + + cap_rights_init(&rights, CAP_IOCTL, CAP_MMAP_RW); + if (caph_rights_limit(ctx->fd, &rights) != 0) + return (-1); + if (caph_ioctls_limit(ctx->fd, vm_ioctl_cmds, vm_ioctl_ncmds) != 0) + return (-1); + return (0); +} + +/* + * Avoid using in new code. Operations on the fd should be wrapped here so that + * capability rights can be kept in sync. + */ +int +vm_get_device_fd(struct vmctx *ctx) +{ + + return (ctx->fd); +} + +/* Legacy interface, do not use. */ +const cap_ioctl_t * +vm_get_ioctls(size_t *len) +{ + cap_ioctl_t *cmds; + size_t sz; + + if (len == NULL) { + sz = vm_ioctl_ncmds * sizeof(vm_ioctl_cmds[0]); + cmds = malloc(sz); + if (cmds == NULL) + return (NULL); + bcopy(vm_ioctl_cmds, cmds, sz); + return (cmds); + } + + *len = vm_ioctl_ncmds; + return (NULL); +} diff --git a/lib/libvmmapi/vmmapi.h b/lib/libvmmapi/vmmapi.h new file mode 100644 index 000000000000..b637c45d1eff --- /dev/null +++ b/lib/libvmmapi/vmmapi.h @@ -0,0 +1,301 @@ +/*- + * SPDX-License-Identifier: BSD-2-Clause + * + * Copyright (c) 2011 NetApp, Inc. + * 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 NETAPP, INC ``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 NETAPP, INC 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 _VMMAPI_H_ +#define _VMMAPI_H_ + +#include <sys/param.h> +#include <sys/cpuset.h> +#include <machine/vmm.h> +#include <machine/vmm_dev.h> + +#include <stdbool.h> + +/* + * API version for out-of-tree consumers like grub-bhyve for making compile + * time decisions. + */ +#define VMMAPI_VERSION 0300 /* 2 digit major followed by 2 digit minor */ + +struct iovec; +struct vcpu; +struct vmctx; +struct vm_snapshot_meta; +enum x2apic_state; + +/* + * Different styles of mapping the memory assigned to a VM into the address + * space of the controlling process. + */ +enum vm_mmap_style { + VM_MMAP_NONE, /* no mapping */ + VM_MMAP_ALL, /* fully and statically mapped */ + VM_MMAP_SPARSE, /* mappings created on-demand */ +}; + +/* + * 'flags' value passed to 'vm_set_memflags()'. + */ +#define VM_MEM_F_INCORE 0x01 /* include guest memory in core file */ +#define VM_MEM_F_WIRED 0x02 /* guest memory is wired */ + +/* Memory size and allocation policy for a single NUMA domain. */ +struct vm_mem_domain { + size_t size; + int ds_policy; + domainset_t *ds_mask; + size_t ds_size; +}; + +__BEGIN_DECLS +/* + * Get the length and name of the memory segment identified by 'segid'. + * Note that system memory segments are identified with a nul name. + * + * Returns 0 on success and non-zero otherwise. + */ +int vm_get_memseg(struct vmctx *ctx, int ident, size_t *lenp, char *name, + size_t namesiz); + +/* + * Iterate over the guest address space. This function finds an address range + * that starts at an address >= *gpa. + * + * Returns 0 if the next address range was found and non-zero otherwise. + */ +int vm_mmap_getnext(struct vmctx *ctx, vm_paddr_t *gpa, int *segid, + vm_ooffset_t *segoff, size_t *len, int *prot, int *flags); + +int vm_get_guestmem_from_ctx(struct vmctx *ctx, char **guest_baseaddr, + size_t *lowmem_size, size_t *highmem_size); + +/* + * Create a device memory segment identified by 'segid'. + * + * Returns a pointer to the memory segment on success and MAP_FAILED otherwise. + */ +void *vm_create_devmem(struct vmctx *ctx, int segid, const char *name, + size_t len); + +/* + * Map the memory segment identified by 'segid' into the guest address space + * at [gpa,gpa+len) with protection 'prot'. + */ +int vm_mmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, int segid, + vm_ooffset_t segoff, size_t len, int prot); + +int vm_munmap_memseg(struct vmctx *ctx, vm_paddr_t gpa, size_t len); + +int vm_create(const char *name); +struct vmctx *vm_open(const char *name); +#define VMMAPI_OPEN_CREATE 0x01 /* create if the VM does not exist */ +#define VMMAPI_OPEN_REINIT 0x02 /* reinitialize the VM if it exists */ +struct vmctx *vm_openf(const char *name, int flags); +void vm_close(struct vmctx *ctx); +void vm_destroy(struct vmctx *ctx); +int vm_limit_rights(struct vmctx *ctx); +struct vcpu *vm_vcpu_open(struct vmctx *ctx, int vcpuid); +void vm_vcpu_close(struct vcpu *vcpu); +int vcpu_id(struct vcpu *vcpu); +int vm_parse_memsize(const char *optarg, size_t *memsize); +int vm_setup_memory(struct vmctx *ctx, size_t len, enum vm_mmap_style s); +int vm_setup_memory_domains(struct vmctx *ctx, enum vm_mmap_style s, + struct vm_mem_domain *doms, int ndoms); +void *vm_map_gpa(struct vmctx *ctx, vm_paddr_t gaddr, size_t len); +/* inverse operation to vm_map_gpa - extract guest address from host pointer */ +vm_paddr_t vm_rev_map_gpa(struct vmctx *ctx, void *addr); +#ifdef __amd64__ +int vm_get_gpa_pmap(struct vmctx *, uint64_t gpa, uint64_t *pte, int *num); +int vm_gla2gpa(struct vcpu *vcpu, struct vm_guest_paging *paging, + uint64_t gla, int prot, uint64_t *gpa, int *fault); +#endif +int vm_gla2gpa_nofault(struct vcpu *vcpu, + struct vm_guest_paging *paging, uint64_t gla, int prot, + uint64_t *gpa, int *fault); +uint32_t vm_get_lowmem_limit(struct vmctx *ctx); +void vm_set_memflags(struct vmctx *ctx, int flags); +int vm_get_memflags(struct vmctx *ctx); +const char *vm_get_name(struct vmctx *ctx); +size_t vm_get_lowmem_size(struct vmctx *ctx); +vm_paddr_t vm_get_highmem_base(struct vmctx *ctx); +size_t vm_get_highmem_size(struct vmctx *ctx); +#ifdef __amd64__ +int vm_set_desc(struct vcpu *vcpu, int reg, + uint64_t base, uint32_t limit, uint32_t access); +int vm_get_desc(struct vcpu *vcpu, int reg, + uint64_t *base, uint32_t *limit, uint32_t *access); +int vm_get_seg_desc(struct vcpu *vcpu, int reg, struct seg_desc *seg_desc); +#endif +int vm_set_register(struct vcpu *vcpu, int reg, uint64_t val); +int vm_get_register(struct vcpu *vcpu, int reg, uint64_t *retval); +int vm_set_register_set(struct vcpu *vcpu, unsigned int count, + const int *regnums, uint64_t *regvals); +int vm_get_register_set(struct vcpu *vcpu, unsigned int count, + const int *regnums, uint64_t *regvals); +int vm_run(struct vcpu *vcpu, struct vm_run *vmrun); +int vm_suspend(struct vmctx *ctx, enum vm_suspend_how how); +int vm_reinit(struct vmctx *ctx); +int vm_raise_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg, + int bus, int slot, int func); +#if defined(__aarch64__) +int vm_attach_vgic(struct vmctx *ctx, uint64_t dist_start, size_t dist_size, + uint64_t redist_start, size_t redist_size); +int vm_inject_exception(struct vcpu *vcpu, uint64_t esr, uint64_t far); +#elif defined(__riscv) +int vm_attach_aplic(struct vmctx *ctx, uint64_t mem_start, size_t mem_size); +int vm_inject_exception(struct vcpu *vcpu, uint64_t scause); +#endif +#if defined(__aarch64__) || defined(__riscv) +int vm_assert_irq(struct vmctx *ctx, uint32_t irq); +int vm_deassert_irq(struct vmctx *ctx, uint32_t irq); +#endif +#ifdef __amd64__ +int vm_apicid2vcpu(struct vmctx *ctx, int apicid); +int vm_inject_exception(struct vcpu *vcpu, int vector, + int errcode_valid, uint32_t errcode, int restart_instruction); +int vm_lapic_irq(struct vcpu *vcpu, int vector); +int vm_lapic_local_irq(struct vcpu *vcpu, int vector); +int vm_lapic_msi(struct vmctx *ctx, uint64_t addr, uint64_t msg); +int vm_ioapic_assert_irq(struct vmctx *ctx, int irq); +int vm_ioapic_deassert_irq(struct vmctx *ctx, int irq); +int vm_ioapic_pulse_irq(struct vmctx *ctx, int irq); +int vm_ioapic_pincount(struct vmctx *ctx, int *pincount); +int vm_isa_assert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq); +int vm_isa_deassert_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq); +int vm_isa_pulse_irq(struct vmctx *ctx, int atpic_irq, int ioapic_irq); +int vm_isa_set_irq_trigger(struct vmctx *ctx, int atpic_irq, + enum vm_intr_trigger trigger); +int vm_inject_nmi(struct vcpu *vcpu); +int vm_readwrite_kernemu_device(struct vcpu *vcpu, + vm_paddr_t gpa, bool write, int size, uint64_t *value); +#endif +int vm_capability_name2type(const char *capname); +const char *vm_capability_type2name(int type); +int vm_get_capability(struct vcpu *vcpu, enum vm_cap_type cap, + int *retval); +int vm_set_capability(struct vcpu *vcpu, enum vm_cap_type cap, + int val); +int vm_assign_pptdev(struct vmctx *ctx, int bus, int slot, int func); +int vm_unassign_pptdev(struct vmctx *ctx, int bus, int slot, int func); +int vm_map_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func, + vm_paddr_t gpa, size_t len, vm_paddr_t hpa); +int vm_unmap_pptdev_mmio(struct vmctx *ctx, int bus, int slot, int func, + vm_paddr_t gpa, size_t len); +int vm_setup_pptdev_msi(struct vmctx *ctx, int bus, int slot, + int func, uint64_t addr, uint64_t msg, int numvec); +int vm_setup_pptdev_msix(struct vmctx *ctx, int bus, int slot, + int func, int idx, uint64_t addr, uint64_t msg, + uint32_t vector_control); +int vm_disable_pptdev_msix(struct vmctx *ctx, int bus, int slot, int func); + +int vm_get_intinfo(struct vcpu *vcpu, uint64_t *i1, uint64_t *i2); +int vm_set_intinfo(struct vcpu *vcpu, uint64_t exit_intinfo); + +/* + * Return a pointer to the statistics buffer. Note that this is not MT-safe. + */ +uint64_t *vm_get_stats(struct vcpu *vcpu, struct timeval *ret_tv, + int *ret_entries); +const char *vm_get_stat_desc(struct vmctx *ctx, int index); + +#ifdef __amd64__ +int vm_get_x2apic_state(struct vcpu *vcpu, enum x2apic_state *s); +int vm_set_x2apic_state(struct vcpu *vcpu, enum x2apic_state s); + +int vm_get_hpet_capabilities(struct vmctx *ctx, uint32_t *capabilities); + +/* + * Translate the GLA range [gla,gla+len) into GPA segments in 'iov'. + * The 'iovcnt' should be big enough to accommodate all GPA segments. + * + * retval fault Interpretation + * 0 0 Success + * 0 1 An exception was injected into the guest + * EFAULT N/A Error + */ +int vm_copy_setup(struct vcpu *vcpu, struct vm_guest_paging *pg, + uint64_t gla, size_t len, int prot, struct iovec *iov, int iovcnt, + int *fault); +#endif +void vm_copyin(struct iovec *guest_iov, void *host_dst, size_t len); +void vm_copyout(const void *host_src, struct iovec *guest_iov, size_t len); +void vm_copy_teardown(struct iovec *iov, int iovcnt); + +#ifdef __amd64__ +/* RTC */ +int vm_rtc_write(struct vmctx *ctx, int offset, uint8_t value); +int vm_rtc_read(struct vmctx *ctx, int offset, uint8_t *retval); +int vm_rtc_settime(struct vmctx *ctx, time_t secs); +int vm_rtc_gettime(struct vmctx *ctx, time_t *secs); +#endif + +/* Reset vcpu register state */ +int vcpu_reset(struct vcpu *vcpu); + +int vm_active_cpus(struct vmctx *ctx, cpuset_t *cpus); +int vm_suspended_cpus(struct vmctx *ctx, cpuset_t *cpus); +int vm_debug_cpus(struct vmctx *ctx, cpuset_t *cpus); +int vm_activate_cpu(struct vcpu *vcpu); +int vm_suspend_all_cpus(struct vmctx *ctx); +int vm_suspend_cpu(struct vcpu *vcpu); +int vm_resume_all_cpus(struct vmctx *ctx); +int vm_resume_cpu(struct vcpu *vcpu); +int vm_restart_instruction(struct vcpu *vcpu); + +/* CPU topology */ +int vm_set_topology(struct vmctx *ctx, uint16_t sockets, uint16_t cores, + uint16_t threads, uint16_t maxcpus); +int vm_get_topology(struct vmctx *ctx, uint16_t *sockets, uint16_t *cores, + uint16_t *threads, uint16_t *maxcpus); + +/* + * FreeBSD specific APIs + */ +int vm_setup_freebsd_registers(struct vcpu *vcpu, + uint64_t rip, uint64_t cr3, uint64_t gdtbase, + uint64_t rsp); +int vm_setup_freebsd_registers_i386(struct vcpu *vcpu, + uint32_t eip, uint32_t gdtbase, + uint32_t esp); +void vm_setup_freebsd_gdt(uint64_t *gdtr); + +/* + * Save and restore + */ +int vm_snapshot_req(struct vmctx *ctx, struct vm_snapshot_meta *meta); +int vm_restore_time(struct vmctx *ctx); + +/* + * Deprecated interfaces, do not use them in new code. + */ +int vm_get_device_fd(struct vmctx *ctx); +const cap_ioctl_t *vm_get_ioctls(size_t *len); + +__END_DECLS + +#endif /* _VMMAPI_H_ */ |