openbsd/src/lib/libcrypto/aes/asm, branch libressl-v2.9.2 A mirror of https://github.com/libressl/openbsd.git https://git.lua4.win/openbsd/atom?h=libressl-v2.9.2 2018-01-07T12:35:52+00:00 On OpenBSD/armv7 we deliberately trap unaligned access. Unfortunately 2018-01-07T12:35:52+00:00 kettenis 2018-01-07T12:35:52+00:00 urn:sha1:849b82a11d72ca02a63517d93855de2ddddff11b the assembly code in libcrypto assumes unaligned access is allowed for ARMv7. Make these paths conditional on __STRICT_ALIGNMENT not being defined and define __STRICT_ALIGNMENT in arm_arch.h for OpenBSD. ok tom@ http://repzret.org/p/repzret/ 2017-12-11T05:48:01+00:00 deraadt 2017-12-11T05:48:01+00:00 urn:sha1:06a79a412573b1755b84374f8fdb1970ca59f1b6 My read of this: Long time ago (Think Conan, not dinasaurs) during the race to make speedier processors, a cpu vendor built a pipeline with a bad stall, and proposed a tremendously hasky workaround. A wizard adopted this into his perl scroll, and failed to reflect later when no compiler adopted the practice. This relic remains at the tail end of some functions in OpenSSL as ".byte 0xf3,0xc3". Banish it straight to hell. ok mlarkin, others also stared blankly Replace all uses of magic numbers when operating on OPENSSL_ia32_P[] by 2016-11-04T17:30:30+00:00 miod 2016-11-04T17:30:30+00:00 urn:sha1:1a12fc8399638223feca8f853e2ac2cc22eeb471 meaningful constants in a private header file, so that reviewers can actually get a chance to figure out what the code is attempting to do without knowing all cpuid bits. While there, turn it from an array of two 32-bit ints into a properly aligned 64-bit int. Use of OPENSSL_ia32_P is now restricted to the assembler parts. C code will now always use OPENSSL_cpu_caps() and check for the proper bits in the whole 64-bit word it returns. i386 tests and ok jsing@ Less S390. 2016-09-04T14:31:29+00:00 jsing 2016-09-04T14:31:29+00:00 urn:sha1:d613aa0d6ef93fa8587113d3e866da6c582bdd0d ok deraadt@ Less IA64. 2016-09-04T14:06:46+00:00 jsing 2016-09-04T14:06:46+00:00 urn:sha1:08b60d422d5acc1a98c47af6c53680ea9cc45bc2 ok deraadt@ Remove the logic responsible for outputting most AES-NI instructions as 2015-07-19T02:26:32+00:00 miod 2015-07-19T02:26:32+00:00 urn:sha1:3a1c21ea05417cdafece6fcfc145ab3f83910695 raw byte sequences. The toolchains have had some time to update and assemble the instructions correctly (except for p{ins,ext}rd which are not supported yet by as(1) under OpenBSD, but will be fixed shortly). Inspired by a discussion between tedu@ and John-Mark Gurney. Verified to still work on Mac OS X and average Linux distros by bcook@ Replace `.byte 0x48,0x83,0xEC,0x08' with `sub \$8,%rsp' which is exactly the 2015-07-19T02:21:41+00:00 miod 2015-07-19T02:21:41+00:00 urn:sha1:fd6d9bfbcd2deccd5544fc90cd1f003d1653b2cd same four bytes, unobfuscated. extenstion -> extension 2015-07-17T20:27:36+00:00 miod 2015-07-17T20:27:36+00:00 urn:sha1:696727d83ea6699d5845e715c4db0142ff1bcda2 Make sure to load absolute symbol address with `dla' instead of `la' when 2014-12-07T15:55:58+00:00 miod 2014-12-07T15:55:58+00:00 urn:sha1:afd8fbfad682973b2970228a5a1fd7e5a19f0cb4 generating code for 64-bit mips userland. Remove OPENSSL_FIPSCANISTER mentions. 2014-12-07T15:37:27+00:00 miod 2014-12-07T15:37:27+00:00 urn:sha1:7791c532168af515a1d14ad730289593f6fd18a9