diff options
Diffstat (limited to 'crypto/aegis128-neon-inner.c')
-rw-r--r-- | crypto/aegis128-neon-inner.c | 161 |
1 files changed, 147 insertions, 14 deletions
diff --git a/crypto/aegis128-neon-inner.c b/crypto/aegis128-neon-inner.c index f05310ca22aa..b6a52a386b22 100644 --- a/crypto/aegis128-neon-inner.c +++ b/crypto/aegis128-neon-inner.c @@ -16,11 +16,11 @@ #define AEGIS_BLOCK_SIZE 16 #include <stddef.h> +#include "aegis-neon.h" extern int aegis128_have_aes_insn; void *memcpy(void *dest, const void *src, size_t n); -void *memset(void *s, int c, size_t n); struct aegis128_state { uint8x16_t v[5]; @@ -132,6 +132,36 @@ void preload_sbox(void) :: "r"(crypto_aes_sbox)); } +void crypto_aegis128_init_neon(void *state, const void *key, const void *iv) +{ + static const uint8_t const0[] = { + 0x00, 0x01, 0x01, 0x02, 0x03, 0x05, 0x08, 0x0d, + 0x15, 0x22, 0x37, 0x59, 0x90, 0xe9, 0x79, 0x62, + }; + static const uint8_t const1[] = { + 0xdb, 0x3d, 0x18, 0x55, 0x6d, 0xc2, 0x2f, 0xf1, + 0x20, 0x11, 0x31, 0x42, 0x73, 0xb5, 0x28, 0xdd, + }; + uint8x16_t k = vld1q_u8(key); + uint8x16_t kiv = k ^ vld1q_u8(iv); + struct aegis128_state st = {{ + kiv, + vld1q_u8(const1), + vld1q_u8(const0), + k ^ vld1q_u8(const0), + k ^ vld1q_u8(const1), + }}; + int i; + + preload_sbox(); + + for (i = 0; i < 5; i++) { + st = aegis128_update_neon(st, k); + st = aegis128_update_neon(st, kiv); + } + aegis128_save_state_neon(st, state); +} + void crypto_aegis128_update_neon(void *state, const void *msg) { struct aegis128_state st = aegis128_load_state_neon(state); @@ -143,10 +173,57 @@ void crypto_aegis128_update_neon(void *state, const void *msg) aegis128_save_state_neon(st, state); } +#ifdef CONFIG_ARM +/* + * AArch32 does not provide these intrinsics natively because it does not + * implement the underlying instructions. AArch32 only provides 64-bit + * wide vtbl.8/vtbx.8 instruction, so use those instead. + */ +static uint8x16_t vqtbl1q_u8(uint8x16_t a, uint8x16_t b) +{ + union { + uint8x16_t val; + uint8x8x2_t pair; + } __a = { a }; + + return vcombine_u8(vtbl2_u8(__a.pair, vget_low_u8(b)), + vtbl2_u8(__a.pair, vget_high_u8(b))); +} + +static uint8x16_t vqtbx1q_u8(uint8x16_t v, uint8x16_t a, uint8x16_t b) +{ + union { + uint8x16_t val; + uint8x8x2_t pair; + } __a = { a }; + + return vcombine_u8(vtbx2_u8(vget_low_u8(v), __a.pair, vget_low_u8(b)), + vtbx2_u8(vget_high_u8(v), __a.pair, vget_high_u8(b))); +} + +static int8_t vminvq_s8(int8x16_t v) +{ + int8x8_t s = vpmin_s8(vget_low_s8(v), vget_high_s8(v)); + + s = vpmin_s8(s, s); + s = vpmin_s8(s, s); + s = vpmin_s8(s, s); + + return vget_lane_s8(s, 0); +} +#endif + +static const uint8_t permute[] __aligned(64) = { + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, + -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, +}; + void crypto_aegis128_encrypt_chunk_neon(void *state, void *dst, const void *src, unsigned int size) { struct aegis128_state st = aegis128_load_state_neon(state); + const int short_input = size < AEGIS_BLOCK_SIZE; uint8x16_t msg; preload_sbox(); @@ -156,7 +233,8 @@ void crypto_aegis128_encrypt_chunk_neon(void *state, void *dst, const void *src, msg = vld1q_u8(src); st = aegis128_update_neon(st, msg); - vst1q_u8(dst, msg ^ s); + msg ^= s; + vst1q_u8(dst, msg); size -= AEGIS_BLOCK_SIZE; src += AEGIS_BLOCK_SIZE; @@ -165,13 +243,26 @@ void crypto_aegis128_encrypt_chunk_neon(void *state, void *dst, const void *src, if (size > 0) { uint8x16_t s = st.v[1] ^ (st.v[2] & st.v[3]) ^ st.v[4]; - uint8_t buf[AEGIS_BLOCK_SIZE] = {}; + uint8_t buf[AEGIS_BLOCK_SIZE]; + const void *in = src; + void *out = dst; + uint8x16_t m; - memcpy(buf, src, size); - msg = vld1q_u8(buf); - st = aegis128_update_neon(st, msg); - vst1q_u8(buf, msg ^ s); - memcpy(dst, buf, size); + if (__builtin_expect(short_input, 0)) + in = out = memcpy(buf + AEGIS_BLOCK_SIZE - size, src, size); + + m = vqtbl1q_u8(vld1q_u8(in + size - AEGIS_BLOCK_SIZE), + vld1q_u8(permute + 32 - size)); + + st = aegis128_update_neon(st, m); + + vst1q_u8(out + size - AEGIS_BLOCK_SIZE, + vqtbl1q_u8(m ^ s, vld1q_u8(permute + size))); + + if (__builtin_expect(short_input, 0)) + memcpy(dst, out, size); + else + vst1q_u8(out - AEGIS_BLOCK_SIZE, msg); } aegis128_save_state_neon(st, state); @@ -181,6 +272,7 @@ void crypto_aegis128_decrypt_chunk_neon(void *state, void *dst, const void *src, unsigned int size) { struct aegis128_state st = aegis128_load_state_neon(state); + const int short_input = size < AEGIS_BLOCK_SIZE; uint8x16_t msg; preload_sbox(); @@ -198,15 +290,56 @@ void crypto_aegis128_decrypt_chunk_neon(void *state, void *dst, const void *src, if (size > 0) { uint8x16_t s = st.v[1] ^ (st.v[2] & st.v[3]) ^ st.v[4]; uint8_t buf[AEGIS_BLOCK_SIZE]; + const void *in = src; + void *out = dst; + uint8x16_t m; - vst1q_u8(buf, s); - memcpy(buf, src, size); - msg = vld1q_u8(buf) ^ s; - vst1q_u8(buf, msg); - memcpy(dst, buf, size); + if (__builtin_expect(short_input, 0)) + in = out = memcpy(buf + AEGIS_BLOCK_SIZE - size, src, size); - st = aegis128_update_neon(st, msg); + m = s ^ vqtbx1q_u8(s, vld1q_u8(in + size - AEGIS_BLOCK_SIZE), + vld1q_u8(permute + 32 - size)); + + st = aegis128_update_neon(st, m); + + vst1q_u8(out + size - AEGIS_BLOCK_SIZE, + vqtbl1q_u8(m, vld1q_u8(permute + size))); + + if (__builtin_expect(short_input, 0)) + memcpy(dst, out, size); + else + vst1q_u8(out - AEGIS_BLOCK_SIZE, msg); } aegis128_save_state_neon(st, state); } + +int crypto_aegis128_final_neon(void *state, void *tag_xor, + unsigned int assoclen, + unsigned int cryptlen, + unsigned int authsize) +{ + struct aegis128_state st = aegis128_load_state_neon(state); + uint8x16_t v; + int i; + + preload_sbox(); + + v = st.v[3] ^ (uint8x16_t)vcombine_u64(vmov_n_u64(8ULL * assoclen), + vmov_n_u64(8ULL * cryptlen)); + + for (i = 0; i < 7; i++) + st = aegis128_update_neon(st, v); + + v = st.v[0] ^ st.v[1] ^ st.v[2] ^ st.v[3] ^ st.v[4]; + + if (authsize > 0) { + v = vqtbl1q_u8(~vceqq_u8(v, vld1q_u8(tag_xor)), + vld1q_u8(permute + authsize)); + + return vminvq_s8((int8x16_t)v); + } + + vst1q_u8(tag_xor, v); + return 0; +} |