wireguard-openbsd/lib/libc/arch/amd64/string/memmove.S, branch jd/queueboostsWireGuard implementation for the OpenBSD kernelhttps://git.zx2c4.com/wireguard-openbsd/atom/lib/libc/arch/amd64/string/memmove.S?h=jd%2Fqueueboosts2018-07-10T16:01:26ZIn asm.h ensure NENTRY uses the old-school nop-sled align, but change standard2018-07-10T16:01:26Zderaadtderaadt@openbsd.org2018-07-10T16:01:26Zurn:sha1:a5910b4fbcaa3dd48faff38aeadcb22c953af71e
ENTRY is a trapsled. Fix a few functions which fall-through into an ENTRY
macro. amd64 binaries now are free of double+-nop sequences (except for one
assember nit in aes-586.pl). Previous changes by guenther got us here.
ok mortimer kettenis
Add retguard macros for libc.2018-07-03T23:14:05Zmortimermortimer@openbsd.org2018-07-03T23:14:05Zurn:sha1:631951aab24dc7dce7719dfa2487185ea5ef120f
ok deraadt
clang doesn't propagate attributes like "asm labels" and "visibility(hidden)"2017-11-29T05:13:57Zguentherguenther@openbsd.org2017-11-29T05:13:57Zurn:sha1:ea6088e7d368d53c49ebfdf4520275cec2f78f5b
to builtins like mem{set,cpy,move} and __stack_smash_handler. So, when
building with clang, instead mark those as protected visibility to get rid
of the PLT relocations. We can't take the address of them then, but that's
ok: it's a build-time error not a run-time error.
ok kettenis@
Split the non-syscall ASM bits from SYS.h into DEFS.h and use that in the2015-11-14T21:53:03Zguentherguenther@openbsd.org2015-11-14T21:53:03Zurn:sha1:b4551fde27a9b4bffacf003cd41ed7a6c54b4036
non-syscall .S source
ok millert@ miod@
Add framework for resolving (pun intended) libc namespace issues, using2015-08-31T02:53:56Zguentherguenther@openbsd.org2015-08-31T02:53:56Zurn:sha1:9b9d2a55a62c8e82206c25f94fcc7f4e2765250e
wrapper .h files and asm labels to let internal calls resolve directly and
not be overridable or use the PLT. Then, apply that framework to most of
the functions in stdio.h, string.h, err.h, and wchar.h. Delete the
should-have-been-hidden-all-along _v?(err|warn)[cx]? symbols while here.
tests clean on i386, amd64, sparc64, powerpc, and mips64
naming feedback from kettenis@ and millert@
ok kettenis@
Sprinkle END() in some straightforward *.S files that have ENTRY(). The2015-05-29T07:21:09Zuebayasiuebayasi@openbsd.org2015-05-29T07:21:09Zurn:sha1:81af83dd3999903f087ff521acd72984b9808a05
resulting *.o have "FUNC" symbols with size set.
restructure libc/string + libc/arch/*/string coperation regarding2014-11-30T19:43:54Zderaadtderaadt@openbsd.org2014-11-30T19:43:54Zurn:sha1:5b859c19fe53bbea08f5c342e0a4470e99f883e1
(potentially) MD versions (function dependent, not filename dependent)
split out memcpy/memmove/bcopy and strchr/index/strrchr/rindex
Bring back amd64 .S versions
And the final touch: switch all architectures temporarily to MI
memcpy.c, which contains syslog + abort for overlapping copies. A nice
harsh undefined behaviour. We will clean the entire userland of the
remaining issues in this catagory, then switch to the optimised memcpy
which skips the memmove check.
I tried to cut this change into pieces, but testing each sub-step on
every architecture is too time consuming and mindnumbing.
ok miod
One of these optimized stubs creates some incredibly subtle damage,2014-11-20T21:51:02Zderaadtderaadt@openbsd.org2014-11-20T21:51:02Zurn:sha1:d1a57517ee8d16db5fa9ffc7f8a82afb73782cbe
causing as(1) to create a wrong nop-sled for text segment aligns.
Revert, until it is found and fixed.
Add amd64 assembler versions of some standard functions to libc. The2014-11-20T14:33:00Zreykreyk@openbsd.org2014-11-20T14:33:00Zurn:sha1:71cd5ab55175e4ba4e758f03a22a8b3c34a8d5a9
code is already used in the kernel and the files are unmodified copies
from src/sys/lib/libkern/arch/amd64/. Depending on the function, this
gives us some speedup in userland.
ok deraadt@, no objections from miod@