diff options
| author |   Jason A. Donenfeld <Jason@zx2c4.com> | 2018-09-27 01:49:19 +0200 |
|---|---|---|
| committer | Jason A. Donenfeld <Jason@zx2c4.com> | 2018-10-02 03:41:49 +0200 |
| commit | 0c81ba3fbc6b5da55c0a4349d603c5a533ecd63b (patch) | |
| tree | 37939646450d5eaf71376391cef8bb7087a0671d /src/crypto/zinc/chacha20/chacha20-x86_64-glue.h | |
| parent | chacha20: test multiple page span in selftest (diff) | |
| download | wireguard-monolithic-historical-0c81ba3fbc6b5da55c0a4349d603c5a533ecd63b.tar.xz wireguard-monolithic-historical-0c81ba3fbc6b5da55c0a4349d603c5a533ecd63b.zip | |
chacha20: feed fpu functions PAGE_SIZE at a time
Suggested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Diffstat (limited to 'src/crypto/zinc/chacha20/chacha20-x86_64-glue.h')
| -rw-r--r-- | src/crypto/zinc/chacha20/chacha20-x86_64-glue.h | 40 |
1 files changed, 27 insertions, 13 deletions
diff --git a/src/crypto/zinc/chacha20/chacha20-x86_64-glue.h b/src/crypto/zinc/chacha20/chacha20-x86_64-glue.h index 03075c9..2939f1e 100644 --- a/src/crypto/zinc/chacha20/chacha20-x86_64-glue.h +++ b/src/crypto/zinc/chacha20/chacha20-x86_64-glue.h @@ -51,26 +51,40 @@ static void __init chacha20_fpu_init(void) } static inline bool chacha20_arch(struct chacha20_ctx *state, u8 *dst, - const u8 *src, const size_t len, + const u8 *src, size_t len, simd_context_t *simd_context) { + /* SIMD disables preemption, so relax after processing each page. */ + BUILD_BUG_ON(PAGE_SIZE < CHACHA20_BLOCK_SIZE || + PAGE_SIZE % CHACHA20_BLOCK_SIZE); + if (!IS_ENABLED(CONFIG_AS_SSSE3) || !chacha20_use_ssse3 || len <= CHACHA20_BLOCK_SIZE || !simd_use(simd_context)) return false; - if (IS_ENABLED(CONFIG_AS_AVX512) && chacha20_use_avx512 && - len >= CHACHA20_BLOCK_SIZE * 8) - chacha20_avx512(dst, src, len, state->key, state->counter); - else if (IS_ENABLED(CONFIG_AS_AVX512) && chacha20_use_avx512vl && - len >= CHACHA20_BLOCK_SIZE * 4) - chacha20_avx512vl(dst, src, len, state->key, state->counter); - else if (IS_ENABLED(CONFIG_AS_AVX2) && chacha20_use_avx2 && - len >= CHACHA20_BLOCK_SIZE * 4) - chacha20_avx2(dst, src, len, state->key, state->counter); - else - chacha20_ssse3(dst, src, len, state->key, state->counter); + for (;;) { + const size_t bytes = min_t(size_t, len, PAGE_SIZE); + + if (IS_ENABLED(CONFIG_AS_AVX512) && chacha20_use_avx512 && + len >= CHACHA20_BLOCK_SIZE * 8) + chacha20_avx512(dst, src, bytes, state->key, state->counter); + else if (IS_ENABLED(CONFIG_AS_AVX512) && chacha20_use_avx512vl && + len >= CHACHA20_BLOCK_SIZE * 4) + chacha20_avx512vl(dst, src, bytes, state->key, state->counter); + else if (IS_ENABLED(CONFIG_AS_AVX2) && chacha20_use_avx2 && + len >= CHACHA20_BLOCK_SIZE * 4) + chacha20_avx2(dst, src, bytes, state->key, state->counter); + else + chacha20_ssse3(dst, src, bytes, state->key, state->counter); + state->counter[0] += (bytes + 63) / 64; + len -= bytes; + if (!len) + break; + dst += bytes; + src += bytes; + simd_relax(simd_context); + } - state->counter[0] += (len + 63) / 64; return true; } |