FreeBSD source tree https://cgit-dev.freebsd.org/src/atom?h=releng%2F12.2 2020-08-24T12:59:55Z 2020-08-24T12:59:55Z Emmanuel Vadot manu@FreeBSD.org 2020-08-24T12:59:55Z urn:sha1:b87c9716d25fff3a883933c07950c72f19bc3f86 r361247: linuxkpi: Add irq_work.h Since handlers are call in a thread context we can simply use a workqueue to emulate those functions. The DRM code was patched to do that already, having it in linuxkpi allows us to not patch the upstream code. Sponsored-by: The FreeBSD Foundation Reviewed by: hselasky Differential Revision: https://reviews.freebsd.org/D24859 r361343: linuxkpi: Add rcu_work functions The rcu_work function helps to queue some work after waiting for a grace period. This is needed by DRM drivers. Sponsored-by: The FreeBSD Foundation Reviewed by: hselasky Differential Revision: https://reviews.freebsd.org/D24942 r361418: libkern: Add arc4random_uniform This variant get a random number up to the limit passed as the argument. This is simply a copy of the libc version. Sponsored-by: The FreeBSD Foundation Reviewed by: cem, hselasky (previous version) Differential Revision: https://reviews.freebsd.org/D24962 r361419: linuxkpi: Add prandom_u32_max This is just a wrapper around arc4random_uniform Needed by DRM v5.3 Sponsored-by: The FreeBSD Foundation Reviewed by: cem, hselasky Differential Revision: https://reviews.freebsd.org/D24961 r361422: bbr: Use arc4random_uniform from libkern. This unbreak LINT build Reported by: jenkins, melifaro r361449: linuxkpi: Add __same_type and __must_be_array macros The same_type macro simply wraps around builtin_types_compatible_p which exist for both GCC and CLANG, which returns 1 if both types are the same. The __must_be_array macros returns 1 if the argument is an array. This is needed for DRM v5.3 Sponsored-by: The FreeBSD Foundation Reviewed by: hselasky Differential Revision: https://reviews.freebsd.org/D24953 2019-05-25T19:59:59Z Johannes Lundberg johalun@FreeBSD.org 2019-05-25T19:59:59Z urn:sha1:d5a3bd4b8b3d148e63fb69e1a50504d7cb739592 Add non-sleepable strdup variant strdup_flags debugfs expects to do non-sleepable allocations Reviewed by: hps@ Sponsored by: iX Systems Differential Revision: https://reviews.freebsd.org/D19259 2018-06-21T17:35:13Z Conrad Meyer cem@FreeBSD.org 2018-06-21T17:35:13Z urn:sha1:e38b830780d47a9014e09015c4acdb1a3f6d3416 No functional change. Please remember to update libkern copies of libc functions when you update libc. Sponsored by: Dell EMC Isilon 2018-06-04T18:47:14Z Conrad Meyer cem@FreeBSD.org 2018-06-04T18:47:14Z urn:sha1:3d825c42cad66b2c7f537c1d487cbf7cbb06a937 Writing one union member and reading another is technically illegal C, although we do it in many places in the tree. Use the __DECONST macro instead, which is (technically) a valid C construct. Trivial style(9) cleanups to touched lines while here. Sponsored by: Dell EMC Isilon 2018-06-02T18:03:35Z Mateusz Guzik mjg@FreeBSD.org 2018-06-02T18:03:35Z urn:sha1:ba96f37758412151885d50838d2031623fc2b918 Some of the routines were using artificially limited builtin already, drop the explicit limit. The use of builtins allows quite often allows the compiler to elide the call or most zeroing to begin with. For instance, if the target object is 32 bytes in size and gets zeroed + has 16 bytes initialized, the compiler can just add code to zero out the rest. Note not all the primites have asm variants and some of the existing ones are not optimized. Maintaines are strongly encourage to take a look (regardless of this change). 2018-06-02T17:57:09Z Mateusz Guzik mjg@FreeBSD.org 2018-06-02T17:57:09Z urn:sha1:97e8984893905ab824659c12dcc6b8ffb44ac0ed 1. Remove special-casing of 0 as it just results in an extra function call. This is clearly pessimal. 2. Drop the inline stuff. For the most part it is much better served with __builtin_memset (coming later). 3. Move the declaration to systm.h to match other funcs. Archs are encouraged to implement the variant for their own platform so that this implementation can be dropped. 2018-06-02T04:25:09Z Bruce Evans bde@FreeBSD.org 2018-06-02T04:25:09Z urn:sha1:dbe3061729e3b1b0a56370d529b4a3f90543a0c3 kernel profiling remains broken). memmove() was broken using ALTENTRY(). ALTENTRY() is only different from ENTRY() in the profiling case, and its use in that case was sort of backwards. The backwardness magically turned memmove() into memcpy() instead of completely breaking it. Only the high resolution parts of profiling itself were broken. Use ordinary ENTRY() for memmove(). Turn bcopy() into a tail call to memmove() to reduce complications. This gives slightly different pessimizations and profiling lossage. The pessimizations are minimized by not using a frame pointer() for bcopy(). Calls to profiling functions from exception trampolines were not relocated. This caused crashes on the first exception. Fix this using function pointers. Addresses of exception handlers in trampolines were not relocated. This caused unknown offsets in the profiling data. Relocate by abusing setidt_disp as for pmc although this is slower than necessary and requires namespace pollution. pmc seems to be missing some relocations. Stack traces and lots of other things in debuggers need similar relocations. Most user addresses were misclassified as unknown kernel addresses and then ignored. Treat all unknown addresses as user. Now only user addresses in the kernel text range are significantly misclassified (as known kernel addresses). The ibrs functions didn't preserve enough registers. This is the only recent breakage on amd64. Although these functions are written in asm, in the profiling case they call profiling functions which are mostly for the C ABI, so they only have to save call-used registers. They also have to save arg and return registers in some cases and actually save them in all cases to reduce complications. They end up saving all registers except %ecx on i386 and %r10 and %r11 on amd64. Saving these is only needed for 1 caller on each of amd64 and i386. Save them there. This is slightly simpler. Remove saving %ecx in handle_ibrs_exit on i386. Both handle_ibrs_entry and handle_ibrs_exit use %ecx, but only the latter needed to or did save it. But saving it there doesn't work for the profiling case. amd64 has more automatic saving of the most common scratch registers %rax, %rcx and %rdx (its complications for %r10 are from unusual use of %r10 by SYSCALL). Thus profiling of handle_ibrs_exit_rs() was not broken, and I didn't simplify the saving by moving the saving of these registers from it to the caller. 2018-05-24T23:20:10Z Warner Losh imp@FreeBSD.org 2018-05-24T23:20:10Z urn:sha1:c6f7141fc0fc9449499ca75763602d851f2e8d46 Shortly, we'll be moving to defining bzero and memset in terms of __builting_memset. To do that, we can't have macro calls to bzero in the fallback impelmentation of memset. Normal calls to bzero are fine. All 4 architectures that use this have their own copies of bzero, so there's no mutual recursion issue between memset and bcopy. 2018-05-24T21:11:38Z Warner Losh imp@FreeBSD.org 2018-05-24T21:11:38Z urn:sha1:509b94387ab8011705b31e805cdb57e89e7ad72b Differential Revision: https://reviews.freebsd.org/D15374 2018-05-10T06:41:08Z Xin LI delphij@FreeBSD.org 2018-05-10T06:41:08Z urn:sha1:b6f7731dba533ce4a5991c118e590b526a4c1208 See r333391 for the rationale. MFC after: 1 week