| Summary: | app-crypt/veracrypt should filter -mno-aes flags (fails to compile, error: inlining failed in call to ‘always_inline’ ‘_mm_aesenc_si128’: target specific option mismatch) | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Oleh <moonlapse81> |
| Component: | Current packages | Assignee: | Göktürk Yüksek <gokturk> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | ago, sam, soap |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| URL: | https://github.com/veracrypt/VeraCrypt/issues/892 | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
I'm not the maintainer, so I'm just offering an opinion, etc. I've told everyone who will listen to not set -mno-flags for exactly this reason. I think if the user specified -march=native instead of -march=ivybridge they would get what they want (no AES) without explicitly disabling anything. I suppose if they're building binpkgs that won't work though. Maybe cflag filtering is the best solution. Users can do whatever they want in CFLAGS (because it's not forbidden). It's a remote report for such exact case where -mno-aes is set by someone. Due to fact veracrypt build system has no ability to control AES enable/disable at build time that can be feed into portage, filtering -mno-aes would be the workaround for such CFLAGS custom choices. (In reply to Oleg from comment #0) > As in summary, veracrypt has AES cpu instruction detection by brief look at > Cpu.h in sources. However, there is no support for enable/disable for AES > relying on detection. In turn, users, who manually adjust their CFLAGS and > set -mno-aes due to their cpu's such as celeron or pentium with cut-off > instructions will face failure. Ebuild should filter "-mno-aes" and it > should be god for those. > > Reproducible: Always > > Steps to Reproduce: > 1. CFLAGS="$CFLAGS -mno-aes" emerge veracrypt > 2. > 3. > Actual Results: > x86_64-pc-linux-gnu-gcc -MMD -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE > -D_LARGE_FILES > -I/var/tmp/portage/app-crypt/veracrypt-1.22/work/VeraCrypt-VeraCrypt_1.22/ > src > -I/var/tmp/portage/app-crypt/veracrypt-1.22/work/VeraCrypt-VeraCrypt_1.22/ > src/Crypto > -I/var/tmp/portage/app-crypt/veracrypt-1.22/work/VeraCrypt-VeraCrypt_1.22/ > src/PKCS11 -O2 -fno-strict-aliasing -D TC_ARCH_X64 -DTC_UNIX -DTC_LINUX > -fdata-sections -ffunction-sections -fpie -Wall -msse2 -maes -mssse3 -mssse3 > -msse4.1 -march=ivybridge -O2 -pipe -mno-aes -mno-rdrnd -c ../Crypto/cpu.c > -o ../Crypto/cpu.o > In file included from ../Crypto/cpu.h:175:0, > from ../Crypto/cpu.c:3: > ../Crypto/cpu.c: In function ‘TryAESNI’: > /usr/lib/gcc/x86_64-pc-linux-gnu/7.3.1/include/wmmintrin.h:61:1: error: > inlining failed in call to always_inline ‘_mm_aesenc_si128’: target specific > option mismatch > _mm_aesenc_si128 (__m128i __X, __m128i __Y) > ^~~~~~~~~~~~~~~~ > ../Crypto/cpu.c:245:14: note: called from here > ciphered = _mm_aesenc_si128(block, subkey); > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > In file included from ../Crypto/cpu.h:175:0, > from ../Crypto/cpu.c:3: > /usr/lib/gcc/x86_64-pc-linux-gnu/7.3.1/include/wmmintrin.h:61:1: error: > inlining failed in call to always_inline ‘_mm_aesenc_si128’: target specific > option mismatch > _mm_aesenc_si128 (__m128i __X, __m128i __Y) > ^~~~~~~~~~~~~~~~ > ../Crypto/cpu.c:245:14: note: called from here > ciphered = _mm_aesenc_si128(block, subkey); Can you try adding '-DCRYPTOPP_DISABLE_AESNI' to your CFLAGS and see if it works? Compiling sha512-x64-nayuki.S
x86_64-pc-linux-gnu-gcc -MMD -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_LARGE_FILES -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/Crypto -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/PKCS11 -O2 -fno-strict-aliasing -D TC_ARCH_X64 -DTC_UNIX -DTC_LINUX -fdata-sections -ffunction-sections -fpie -Wall -msse2 -maes -mssse3 -mssse3 -msse4.1 -DCRYPTOPP_DISABLE_AESNI -c ../Crypto/sha512-x64-nayuki.S -o ../Crypto/sha512-x64-nayuki.o
Assembling sha256_avx1_x64.asm
yasm -D __GNUC__ -D __YASM__ -f elf64 -D __BITS__=64 -o ../Crypto/sha256_avx1_x64.o ../Crypto/sha256_avx1_x64.asm
Assembling sha256_avx2_x64.asm
yasm -D __GNUC__ -D __YASM__ -f elf64 -D __BITS__=64 -o ../Crypto/sha256_avx2_x64.o ../Crypto/sha256_avx2_x64.asm
Assembling sha256_sse4_x64.asm
yasm -D __GNUC__ -D __YASM__ -f elf64 -D __BITS__=64 -o ../Crypto/sha256_sse4_x64.o ../Crypto/sha256_sse4_x64.asm
Assembling sha512_avx1_x64.asm
yasm -D __GNUC__ -D __YASM__ -f elf64 -D __BITS__=64 -o ../Crypto/sha512_avx1_x64.o ../Crypto/sha512_avx1_x64.asm
Assembling sha512_avx2_x64.asm
yasm -D __GNUC__ -D __YASM__ -f elf64 -D __BITS__=64 -o ../Crypto/sha512_avx2_x64.o ../Crypto/sha512_avx2_x64.asm
Assembling sha512_sse4_x64.asm
yasm -D __GNUC__ -D __YASM__ -f elf64 -D __BITS__=64 -o ../Crypto/sha512_sse4_x64.o ../Crypto/sha512_sse4_x64.asm
Compiling Aeskey.c
x86_64-pc-linux-gnu-gcc -MMD -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_LARGE_FILES -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/Crypto -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/PKCS11 -O2 -fno-strict-aliasing -D TC_ARCH_X64 -DTC_UNIX -DTC_LINUX -fdata-sections -ffunction-sections -fpie -Wall -msse2 -maes -mssse3 -mssse3 -msse4.1 -DCRYPTOPP_DISABLE_AESNI -c ../Crypto/Aeskey.c -o ../Crypto/Aeskey.o
Compiling Aestab.c
x86_64-pc-linux-gnu-gcc -MMD -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_LARGE_FILES -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/Crypto -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/PKCS11 -O2 -fno-strict-aliasing -D TC_ARCH_X64 -DTC_UNIX -DTC_LINUX -fdata-sections -ffunction-sections -fpie -Wall -msse2 -maes -mssse3 -mssse3 -msse4.1 -DCRYPTOPP_DISABLE_AESNI -c ../Crypto/Aestab.c -o ../Crypto/Aestab.o
Compiling cpu.c
x86_64-pc-linux-gnu-gcc -MMD -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_LARGE_FILES -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/Crypto -I/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/PKCS11 -O2 -fno-strict-aliasing -D TC_ARCH_X64 -DTC_UNIX -DTC_LINUX -fdata-sections -ffunction-sections -fpie -Wall -msse2 -maes -mssse3 -mssse3 -msse4.1 -DCRYPTOPP_DISABLE_AESNI -c ../Crypto/cpu.c -o ../Crypto/cpu.o
^[[01m^[[K../Crypto/cpu.c:^[[m^[[K In function ‘^[[01m^[[KTryAESNI^[[m^[[K’:
^[[01m^[[K../Crypto/cpu.c:245:14:^[[m^[[K ^[[01;35m^[[Kwarning: ^[[m^[[Kimplicit declaration of function ‘^[[01m^[[K_mm_aesenc_si128^[[m^[[K’; did you mean ‘^[[01m^[[K_mm_and_si128^[[m^[[K’? [^[[01;35m^[[K-Wimplicit-function-declaration^[[m^[[K]
ciphered = ^[[01;35m^[[K_mm_aesenc_si128^[[m^[[K(block, subkey);
^[[01;35m^[[K^~~~~~~~~~~~~~~~^[[m^[[K
^[[32m^[[K_mm_and_si128^[[m^[[K
^[[01m^[[K../Crypto/cpu.c:245:12:^[[m^[[K ^[[01;31m^[[Kerror: ^[[m^[[Kincompatible types when assigning to type ‘^[[01m^[[K__m128i {aka __vector(2) long long int}^[[m^[[K’ from type ‘^[[01m^[[Kint^[[m^[[K’
ciphered ^[[01;31m^[[K=^[[m^[[K _mm_aesenc_si128(block, subkey);
^[[01;31m^[[K^^[[m^[[K
make[1]: *** [/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/Build/Include/Makefile.inc:21: ../Crypto/cpu.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make[1]: Leaving directory '/var/tmp/portage/app-crypt/veracrypt-1.23/work/VeraCrypt-VeraCrypt_1.23/src/Volume'
make: *** [Makefile:391: all] Error 2
*** Bug 806133 has been marked as a duplicate of this bug. *** I did go through the source code again and I couldn't see a clear way to honor '-mno-aes'. I asked upstream if they support it. Please try 1.25.9 (https://gitweb.gentoo.org/repo/gentoo.git/commit/app-crypt/veracrypt/veracrypt-1.25.9.ebuild?id=0e6ddb7525688d55595ad746c47adb88f98c1b0a) and re-open if the issue still exists. |
As in summary, veracrypt has AES cpu instruction detection by brief look at Cpu.h in sources. However, there is no support for enable/disable for AES relying on detection. In turn, users, who manually adjust their CFLAGS and set -mno-aes due to their cpu's such as celeron or pentium with cut-off instructions will face failure. Ebuild should filter "-mno-aes" and it should be god for those. Reproducible: Always Steps to Reproduce: 1. CFLAGS="$CFLAGS -mno-aes" emerge veracrypt 2. 3. Actual Results: x86_64-pc-linux-gnu-gcc -MMD -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_LARGE_FILES -I/var/tmp/portage/app-crypt/veracrypt-1.22/work/VeraCrypt-VeraCrypt_1.22/src -I/var/tmp/portage/app-crypt/veracrypt-1.22/work/VeraCrypt-VeraCrypt_1.22/src/Crypto -I/var/tmp/portage/app-crypt/veracrypt-1.22/work/VeraCrypt-VeraCrypt_1.22/src/PKCS11 -O2 -fno-strict-aliasing -D TC_ARCH_X64 -DTC_UNIX -DTC_LINUX -fdata-sections -ffunction-sections -fpie -Wall -msse2 -maes -mssse3 -mssse3 -msse4.1 -march=ivybridge -O2 -pipe -mno-aes -mno-rdrnd -c ../Crypto/cpu.c -o ../Crypto/cpu.o In file included from ../Crypto/cpu.h:175:0, from ../Crypto/cpu.c:3: ../Crypto/cpu.c: In function ‘TryAESNI’: /usr/lib/gcc/x86_64-pc-linux-gnu/7.3.1/include/wmmintrin.h:61:1: error: inlining failed in call to always_inline ‘_mm_aesenc_si128’: target specific option mismatch _mm_aesenc_si128 (__m128i __X, __m128i __Y) ^~~~~~~~~~~~~~~~ ../Crypto/cpu.c:245:14: note: called from here ciphered = _mm_aesenc_si128(block, subkey); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In file included from ../Crypto/cpu.h:175:0, from ../Crypto/cpu.c:3: /usr/lib/gcc/x86_64-pc-linux-gnu/7.3.1/include/wmmintrin.h:61:1: error: inlining failed in call to always_inline ‘_mm_aesenc_si128’: target specific option mismatch _mm_aesenc_si128 (__m128i __X, __m128i __Y) ^~~~~~~~~~~~~~~~ ../Crypto/cpu.c:245:14: note: called from here ciphered = _mm_aesenc_si128(block, subkey);