| Message ID | 20221219154118.955831880@infradead.org (mailing list archive) |
|---|---|
| State | Not Applicable |
| Delegated to: | Herbert Xu |
| Headers | show |
| Series | Introduce cmpxchg128() -- aka. the demise of cmpxchg_double() | expand |
On Mon, Dec 19, 2022 at 04:35:27PM +0100, Peter Zijlstra wrote: > Even though x86 is firmly little endian, use be128 because le128 is in > fact the wrong way around :/ The actual code is already using be128 in > ghash_setkey() so this shouldn't be more confusing. > > This frees up the u128 name for a real u128 type. > > Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> This patch doesn't make sense. The x86 ghash code is definitely storing the key as a little endian value, not big endian. The reason be128 shows up in ghash_setkey() is because the code is doing a byteswap from the original key bytes. Also, this patch causes 'sparse' warnings. Can you consider https://lore.kernel.org/linux-crypto/20221220054042.188537-1-ebiggers@kernel.org/T/#u instead? - Eric
--- a/arch/x86/crypto/ghash-clmulni-intel_asm.S +++ b/arch/x86/crypto/ghash-clmulni-intel_asm.S @@ -88,7 +88,7 @@ SYM_FUNC_START_LOCAL(__clmul_gf128mul_bl RET SYM_FUNC_END(__clmul_gf128mul_ble) -/* void clmul_ghash_mul(char *dst, const u128 *shash) */ +/* void clmul_ghash_mul(char *dst, const le128 *shash) */ SYM_FUNC_START(clmul_ghash_mul) FRAME_BEGIN movups (%rdi), DATA @@ -104,7 +104,7 @@ SYM_FUNC_END(clmul_ghash_mul) /* * void clmul_ghash_update(char *dst, const char *src, unsigned int srclen, - * const u128 *shash); + * const le128 *shash); */ SYM_FUNC_START(clmul_ghash_update) FRAME_BEGIN --- a/arch/x86/crypto/ghash-clmulni-intel_glue.c +++ b/arch/x86/crypto/ghash-clmulni-intel_glue.c @@ -23,17 +23,17 @@ #define GHASH_BLOCK_SIZE 16 #define GHASH_DIGEST_SIZE 16 -void clmul_ghash_mul(char *dst, const u128 *shash); +void clmul_ghash_mul(char *dst, const be128 *shash); void clmul_ghash_update(char *dst, const char *src, unsigned int srclen, - const u128 *shash); + const be128 *shash); struct ghash_async_ctx { struct cryptd_ahash *cryptd_tfm; }; struct ghash_ctx { - u128 shash; + be128 shash; }; struct ghash_desc_ctx {
Even though x86 is firmly little endian, use be128 because le128 is in fact the wrong way around :/ The actual code is already using be128 in ghash_setkey() so this shouldn't be more confusing. This frees up the u128 name for a real u128 type. Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> --- arch/x86/crypto/ghash-clmulni-intel_asm.S | 4 ++-- arch/x86/crypto/ghash-clmulni-intel_glue.c | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-)