kernel.git/arch/x86/crypto, branch linux-4.14.y

crypto: aesni - add compatibility with IAS

2020-08-21T07:48:13Z

[ Upstream commit 44069737ac9625a0f02f0f7f5ab96aae4cd819bc ] Clang's integrated assembler complains "invalid reassignment of non-absolute variable 'var_ddq_add'" while assembling arch/x86/crypto/aes_ctrby8_avx-x86_64.S. It was because var_ddq_add was reassigned with non-absolute values several times, which IAS did not support. We can avoid the reassignment by replacing the uses of var_ddq_add with its definitions accordingly to have compatilibility with IAS. Link: https://github.com/ClangBuiltLinux/linux/issues/1008 Reported-by: Sedat Dilek Reported-by: Fangrui Song Tested-by: Sedat Dilek # build+boot Linux v5.7.5; clang v11.0.0-git Signed-off-by: Jian Cai Signed-off-by: Herbert Xu Signed-off-by: Sasha Levin

crypto: x86/crct10dif-pcl - fix use via crypto_shash_digest()

2019-05-21T16:50:15Z

commit dec3d0b1071a0f3194e66a83d26ecf4aa8c5910e upstream. The ->digest() method of crct10dif-pclmul reads the current CRC value from the shash_desc context. But this value is uninitialized, causing crypto_shash_digest() to compute the wrong result. Fix it. Probably this wasn't noticed before because lib/crc-t10dif.c only uses crypto_shash_update(), not crypto_shash_digest(). Likewise, crypto_shash_digest() is not yet tested by the crypto self-tests because those only test the ahash API which only uses shash init/update/final. Fixes: 0b95a7f85718 ("crypto: crct10dif - Glue code to cast accelerated CRCT10DIF assembly as a crypto transform") Cc: # v3.11+ Cc: Tim Chen Signed-off-by: Eric Biggers Signed-off-by: Herbert Xu Signed-off-by: Greg Kroah-Hartman

crypto: x86/poly1305 - fix overflow during partial reduction

2019-04-27T07:35:37Z

commit 678cce4019d746da6c680c48ba9e6d417803e127 upstream. The x86_64 implementation of Poly1305 produces the wrong result on some inputs because poly1305_4block_avx2() incorrectly assumes that when partially reducing the accumulator, the bits carried from limb 'd4' to limb 'h0' fit in a 32-bit integer. This is true for poly1305-generic which processes only one block at a time. However, it's not true for the AVX2 implementation, which processes 4 blocks at a time and therefore can produce intermediate limbs about 4x larger. Fix it by making the relevant calculations use 64-bit arithmetic rather than 32-bit. Note that most of the carries already used 64-bit arithmetic, but the d4 -> h0 carry was different for some reason. To be safe I also made the same change to the corresponding SSE2 code, though that only operates on 1 or 2 blocks at a time. I don't think it's really needed for poly1305_block_sse2(), but it doesn't hurt because it's already x86_64 code. It *might* be needed for poly1305_2block_sse2(), but overflows aren't easy to reproduce there. This bug was originally detected by my patches that improve testmgr to fuzz algorithms against their generic implementation. But also add a test vector which reproduces it directly (in the AVX2 case). Fixes: b1ccc8f4b631 ("crypto: poly1305 - Add a four block AVX2 variant for x86_64") Fixes: c70f4abef07a ("crypto: poly1305 - Add a SSE2 SIMD variant for x86_64") Cc: # v4.3+ Cc: Martin Willi Cc: Jason A. Donenfeld Signed-off-by: Eric Biggers Reviewed-by: Martin Willi Signed-off-by: Herbert Xu Signed-off-by: Greg Kroah-Hartman

crypto: x86/sha256-mb - fix digest copy in sha256_mb_mgr_get_comp_job_avx2()

2018-08-17T19:01:10Z

commit af839b4e546613aed1fbd64def73956aa98631e7 upstream. There is a copy-paste error where sha256_mb_mgr_get_comp_job_avx2() copies the SHA-256 digest state from sha256_mb_mgr::args::digest to job_sha256::result_digest. Consequently, the sha256_mb algorithm sometimes calculates the wrong digest. Fix it. Reproducer using AF_ALG: #include #include #include #include #include #include static const __u8 expected[32] = "\xad\x7f\xac\xb2\x58\x6f\xc6\xe9\x66\xc0\x04\xd7\xd1\xd1\x6b\x02" "\x4f\x58\x05\xff\x7c\xb4\x7c\x7a\x85\xda\xbd\x8b\x48\x89\x2c\xa7"; int main() { int fd; struct sockaddr_alg addr = { .salg_type = "hash", .salg_name = "sha256_mb", }; __u8 data[4096] = { 0 }; __u8 digest[32]; int ret; int i; fd = socket(AF_ALG, SOCK_SEQPACKET, 0); bind(fd, (void *)&addr, sizeof(addr)); fork(); fd = accept(fd, 0, 0); do { ret = write(fd, data, 4096); assert(ret == 4096); ret = read(fd, digest, 32); assert(ret == 32); } while (memcmp(digest, expected, 32) == 0); printf("wrong digest: "); for (i = 0; i < 32; i++) printf("%02x", digest[i]); printf("\n"); } Output was: wrong digest: ad7facb2000000000000000000000000ffffffef7cb47c7a85dabd8b48892ca7 Fixes: 172b1d6b5a93 ("crypto: sha256-mb - fix ctx pointer and digest copy") Cc: # v4.8+ Signed-off-by: Eric Biggers Signed-off-by: Herbert Xu Signed-off-by: Greg Kroah-Hartman

crypto: x86/salsa20 - remove x86 salsa20 implementations

2018-07-17T09:39:31Z

commit b7b73cd5d74694ed59abcdb4974dacb4ff8b2a2a upstream. The x86 assembly implementations of Salsa20 use the frame base pointer register (%ebp or %rbp), which breaks frame pointer convention and breaks stack traces when unwinding from an interrupt in the crypto code. Recent (v4.10+) kernels will warn about this, e.g. WARNING: kernel stack regs at 00000000a8291e69 in syzkaller047086:4677 has bad 'bp' value 000000001077994c [...] But after looking into it, I believe there's very little reason to still retain the x86 Salsa20 code. First, these are *not* vectorized (SSE2/SSSE3/AVX2) implementations, which would be needed to get anywhere close to the best Salsa20 performance on any remotely modern x86 processor; they're just regular x86 assembly. Second, it's still unclear that anyone is actually using the kernel's Salsa20 at all, especially given that now ChaCha20 is supported too, and with much more efficient SSSE3 and AVX2 implementations. Finally, in benchmarks I did on both Intel and AMD processors with both gcc 8.1.0 and gcc 4.9.4, the x86_64 salsa20-asm is actually slightly *slower* than salsa20-generic (~3% slower on Skylake, ~10% slower on Zen), while the i686 salsa20-asm is only slightly faster than salsa20-generic (~15% faster on Skylake, ~20% faster on Zen). The gcc version made little difference. So, the x86_64 salsa20-asm is pretty clearly useless. That leaves just the i686 salsa20-asm, which based on my tests provides a 15-20% speed boost. But that's without updating the code to not use %ebp. And given the maintenance cost, the small speed difference vs. salsa20-generic, the fact that few people still use i686 kernels, the doubt that anyone is even using the kernel's Salsa20 at all, and the fact that a SSE2 implementation would almost certainly be much faster on any remotely modern x86 processor yet no one has cared enough to add one yet, I don't think it's worthwhile to keep. Thus, just remove both the x86_64 and i686 salsa20-asm implementations. Reported-by: syzbot+ffa3a158337bbc01ff09@syzkaller.appspotmail.com Signed-off-by: Eric Biggers Signed-off-by: Herbert Xu Signed-off-by: Greg Kroah-Hartman

crypto: x86/cast5-avx - fix ECB encryption when long sg follows short one

2018-04-08T12:26:32Z

commit 8f461b1e02ed546fbd0f11611138da67fd85a30f upstream. With ecb-cast5-avx, if a 128+ byte scatterlist element followed a shorter one, then the algorithm accidentally encrypted/decrypted only 8 bytes instead of the expected 128 bytes. Fix it by setting the encryption/decryption 'fn' correctly. Fixes: c12ab20b162c ("crypto: cast5/avx - avoid using temporary stack buffers") Cc: # v3.8+ Signed-off-by: Eric Biggers Signed-off-by: Herbert Xu Signed-off-by: Greg Kroah-Hartman

crypto: x86/twofish-3way - Fix %rbp usage

2018-02-25T10:07:49Z

commit d8c7fe9f2a486a6e5f0d5229ca43807af5ab22c6 upstream. Using %rbp as a temporary register breaks frame pointer convention and breaks stack traces when unwinding from an interrupt in the crypto code. In twofish-3way, we can't simply replace %rbp with another register because there are none available. Instead, we use the stack to hold the values that %rbp, %r11, and %r12 were holding previously. Each of these values represents the half of the output from the previous Feistel round that is being passed on unchanged to the following round. They are only used once per round, when they are exchanged with %rax, %rbx, and %rcx. As a result, we free up 3 registers (one per block) and can reassign them so that %rbp is not used, and additionally %r14 and %r15 are not used so they do not need to be saved/restored. There may be a small overhead caused by replacing 'xchg REG, REG' with the needed sequence 'mov MEM, REG; mov REG, MEM; mov REG, REG' once per round. But, counterintuitively, when I tested "ctr-twofish-3way" on a Haswell processor, the new version was actually about 2% faster. (Perhaps 'xchg' is not as well optimized as plain moves.) Reported-by: syzbot Signed-off-by: Eric Biggers Reviewed-by: Josh Poimboeuf Signed-off-by: Herbert Xu Signed-off-by: Greg Kroah-Hartman

crypto: sha512-mb - initialize pending lengths correctly

2018-02-16T19:23:02Z

commit eff84b379089cd8b4e83599639c1f5f6e34ef7bf upstream. The SHA-512 multibuffer code keeps track of the number of blocks pending in each lane. The minimum of these values is used to identify the next lane that will be completed. Unused lanes are set to a large number (0xFFFFFFFF) so that they don't affect this calculation. However, it was forgotten to set the lengths to this value in the initial state, where all lanes are unused. As a result it was possible for sha512_mb_mgr_get_comp_job_avx2() to select an unused lane, causing a NULL pointer dereference. Specifically this could happen in the case where ->update() was passed fewer than SHA512_BLOCK_SIZE bytes of data, so it then called sha_complete_job() without having actually submitted any blocks to the multi-buffer code. This hit a NULL pointer dereference if another task happened to have submitted blocks concurrently to the same CPU and the flush timer had not yet expired. Fix this by initializing sha512_mb_mgr->lens correctly. As usual, this bug was found by syzkaller. Fixes: 45691e2d9b18 ("crypto: sha512-mb - submit/flush routines for AVX2") Reported-by: syzbot Signed-off-by: Eric Biggers Signed-off-by: Herbert Xu Signed-off-by: Greg Kroah-Hartman

crypto: hash - annotate algorithms taking optional key

2018-02-16T19:23:00Z

commit a208fa8f33031b9e0aba44c7d1b7e68eb0cbd29e upstream. We need to consistently enforce that keyed hashes cannot be used without setting the key. To do this we need a reliable way to determine whether a given hash algorithm is keyed or not. AF_ALG currently does this by checking for the presence of a ->setkey() method. However, this is actually slightly broken because the CRC-32 algorithms implement ->setkey() but can also be used without a key. (The CRC-32 "key" is not actually a cryptographic key but rather represents the initial state. If not overridden, then a default initial state is used.) Prepare to fix this by introducing a flag CRYPTO_ALG_OPTIONAL_KEY which indicates that the algorithm has a ->setkey() method, but it is not required to be called. Then set it on all the CRC-32 algorithms. The same also applies to the Adler-32 implementation in Lustre. Also, the cryptd and mcryptd templates have to pass through the flag from their underlying algorithm. Signed-off-by: Eric Biggers Signed-off-by: Herbert Xu Signed-off-by: Greg Kroah-Hartman

crypto: poly1305 - remove ->setkey() method

2018-02-16T19:23:00Z

commit a16e772e664b9a261424107784804cffc8894977 upstream. Since Poly1305 requires a nonce per invocation, the Linux kernel implementations of Poly1305 don't use the crypto API's keying mechanism and instead expect the key and nonce as the first 32 bytes of the data. But ->setkey() is still defined as a stub returning an error code. This prevents Poly1305 from being used through AF_ALG and will also break it completely once we start enforcing that all crypto API users (not just AF_ALG) call ->setkey() if present. Fix it by removing crypto_poly1305_setkey(), leaving ->setkey as NULL. Signed-off-by: Eric Biggers Signed-off-by: Herbert Xu Signed-off-by: Greg Kroah-Hartman