* [PATCH] crypto: arm64/aegis128 - use explicit vector load for permute vectors
@ 2019-08-19 14:15 Ard Biesheuvel
2019-08-19 15:44 ` Nathan Chancellor
2019-08-30 8:14 ` Herbert Xu
0 siblings, 2 replies; 3+ messages in thread
From: Ard Biesheuvel @ 2019-08-19 14:15 UTC (permalink / raw)
To: linux-crypto; +Cc: herbert, natechancellor, Ard Biesheuvel
When building the new aegis128 NEON code in big endian mode, Clang
complains about the const uint8x16_t permute vectors in the following
way:
crypto/aegis128-neon-inner.c:58:40: warning: vector initializers are not
compatible with NEON intrinsics in big endian mode
[-Wnonportable-vector-initialization]
static const uint8x16_t shift_rows = {
^
crypto/aegis128-neon-inner.c:58:40: note: consider using vld1q_u8() to
initialize a vector from memory, or vcombine_u8(vcreate_u8(), vcreate_u8())
to initialize from integer constants
Since the same issue applies to the uint8x16x4_t loads of the AES Sbox,
update those references as well. However, since GCC does not implement
the vld1q_u8_x4() intrinsic, switch from IS_ENABLED() to a preprocessor
conditional to conditionally include this code.
Reported-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
crypto/aegis128-neon-inner.c | 38 ++++++++++----------
1 file changed, 19 insertions(+), 19 deletions(-)
diff --git a/crypto/aegis128-neon-inner.c b/crypto/aegis128-neon-inner.c
index ed55568afd1b..f05310ca22aa 100644
--- a/crypto/aegis128-neon-inner.c
+++ b/crypto/aegis128-neon-inner.c
@@ -26,7 +26,7 @@ struct aegis128_state {
uint8x16_t v[5];
};
-extern const uint8x16x4_t crypto_aes_sbox[];
+extern const uint8_t crypto_aes_sbox[];
static struct aegis128_state aegis128_load_state_neon(const void *state)
{
@@ -55,39 +55,39 @@ uint8x16_t aegis_aes_round(uint8x16_t w)
#ifdef CONFIG_ARM64
if (!__builtin_expect(aegis128_have_aes_insn, 1)) {
- static const uint8x16_t shift_rows = {
+ static const uint8_t shift_rows[] = {
0x0, 0x5, 0xa, 0xf, 0x4, 0x9, 0xe, 0x3,
0x8, 0xd, 0x2, 0x7, 0xc, 0x1, 0x6, 0xb,
};
- static const uint8x16_t ror32by8 = {
+ static const uint8_t ror32by8[] = {
0x1, 0x2, 0x3, 0x0, 0x5, 0x6, 0x7, 0x4,
0x9, 0xa, 0xb, 0x8, 0xd, 0xe, 0xf, 0xc,
};
uint8x16_t v;
// shift rows
- w = vqtbl1q_u8(w, shift_rows);
+ w = vqtbl1q_u8(w, vld1q_u8(shift_rows));
// sub bytes
- if (!IS_ENABLED(CONFIG_CC_IS_GCC)) {
- v = vqtbl4q_u8(crypto_aes_sbox[0], w);
- v = vqtbx4q_u8(v, crypto_aes_sbox[1], w - 0x40);
- v = vqtbx4q_u8(v, crypto_aes_sbox[2], w - 0x80);
- v = vqtbx4q_u8(v, crypto_aes_sbox[3], w - 0xc0);
- } else {
- asm("tbl %0.16b, {v16.16b-v19.16b}, %1.16b" : "=w"(v) : "w"(w));
- w -= 0x40;
- asm("tbx %0.16b, {v20.16b-v23.16b}, %1.16b" : "+w"(v) : "w"(w));
- w -= 0x40;
- asm("tbx %0.16b, {v24.16b-v27.16b}, %1.16b" : "+w"(v) : "w"(w));
- w -= 0x40;
- asm("tbx %0.16b, {v28.16b-v31.16b}, %1.16b" : "+w"(v) : "w"(w));
- }
+#ifndef CONFIG_CC_IS_GCC
+ v = vqtbl4q_u8(vld1q_u8_x4(crypto_aes_sbox), w);
+ v = vqtbx4q_u8(v, vld1q_u8_x4(crypto_aes_sbox + 0x40), w - 0x40);
+ v = vqtbx4q_u8(v, vld1q_u8_x4(crypto_aes_sbox + 0x80), w - 0x80);
+ v = vqtbx4q_u8(v, vld1q_u8_x4(crypto_aes_sbox + 0xc0), w - 0xc0);
+#else
+ asm("tbl %0.16b, {v16.16b-v19.16b}, %1.16b" : "=w"(v) : "w"(w));
+ w -= 0x40;
+ asm("tbx %0.16b, {v20.16b-v23.16b}, %1.16b" : "+w"(v) : "w"(w));
+ w -= 0x40;
+ asm("tbx %0.16b, {v24.16b-v27.16b}, %1.16b" : "+w"(v) : "w"(w));
+ w -= 0x40;
+ asm("tbx %0.16b, {v28.16b-v31.16b}, %1.16b" : "+w"(v) : "w"(w));
+#endif
// mix columns
w = (v << 1) ^ (uint8x16_t)(((int8x16_t)v >> 7) & 0x1b);
w ^= (uint8x16_t)vrev32q_u16((uint16x8_t)v);
- w ^= vqtbl1q_u8(v ^ w, ror32by8);
+ w ^= vqtbl1q_u8(v ^ w, vld1q_u8(ror32by8));
return w;
}
--
2.17.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] crypto: arm64/aegis128 - use explicit vector load for permute vectors
2019-08-19 14:15 [PATCH] crypto: arm64/aegis128 - use explicit vector load for permute vectors Ard Biesheuvel
@ 2019-08-19 15:44 ` Nathan Chancellor
2019-08-30 8:14 ` Herbert Xu
1 sibling, 0 replies; 3+ messages in thread
From: Nathan Chancellor @ 2019-08-19 15:44 UTC (permalink / raw)
To: Ard Biesheuvel; +Cc: linux-crypto, herbert
On Mon, Aug 19, 2019 at 05:15:00PM +0300, Ard Biesheuvel wrote:
> When building the new aegis128 NEON code in big endian mode, Clang
> complains about the const uint8x16_t permute vectors in the following
> way:
>
> crypto/aegis128-neon-inner.c:58:40: warning: vector initializers are not
> compatible with NEON intrinsics in big endian mode
> [-Wnonportable-vector-initialization]
> static const uint8x16_t shift_rows = {
> ^
> crypto/aegis128-neon-inner.c:58:40: note: consider using vld1q_u8() to
> initialize a vector from memory, or vcombine_u8(vcreate_u8(), vcreate_u8())
> to initialize from integer constants
>
> Since the same issue applies to the uint8x16x4_t loads of the AES Sbox,
> update those references as well. However, since GCC does not implement
> the vld1q_u8_x4() intrinsic, switch from IS_ENABLED() to a preprocessor
> conditional to conditionally include this code.
>
> Reported-by: Nathan Chancellor <natechancellor@gmail.com>
> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
I am not familiar enough with vectors and such to confidently give a
review but I can say this fixes the warning and doesn't introduce any
new ones. Thank you for the fix!
Tested-by: Nathan Chancellor <natechancellor@gmail.com>
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] crypto: arm64/aegis128 - use explicit vector load for permute vectors
2019-08-19 14:15 [PATCH] crypto: arm64/aegis128 - use explicit vector load for permute vectors Ard Biesheuvel
2019-08-19 15:44 ` Nathan Chancellor
@ 2019-08-30 8:14 ` Herbert Xu
1 sibling, 0 replies; 3+ messages in thread
From: Herbert Xu @ 2019-08-30 8:14 UTC (permalink / raw)
To: Ard Biesheuvel; +Cc: linux-crypto, natechancellor
On Mon, Aug 19, 2019 at 05:15:00PM +0300, Ard Biesheuvel wrote:
> When building the new aegis128 NEON code in big endian mode, Clang
> complains about the const uint8x16_t permute vectors in the following
> way:
>
> crypto/aegis128-neon-inner.c:58:40: warning: vector initializers are not
> compatible with NEON intrinsics in big endian mode
> [-Wnonportable-vector-initialization]
> static const uint8x16_t shift_rows = {
> ^
> crypto/aegis128-neon-inner.c:58:40: note: consider using vld1q_u8() to
> initialize a vector from memory, or vcombine_u8(vcreate_u8(), vcreate_u8())
> to initialize from integer constants
>
> Since the same issue applies to the uint8x16x4_t loads of the AES Sbox,
> update those references as well. However, since GCC does not implement
> the vld1q_u8_x4() intrinsic, switch from IS_ENABLED() to a preprocessor
> conditional to conditionally include this code.
>
> Reported-by: Nathan Chancellor <natechancellor@gmail.com>
> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
> ---
> crypto/aegis128-neon-inner.c | 38 ++++++++++----------
> 1 file changed, 19 insertions(+), 19 deletions(-)
Patch applied. Thanks.
--
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2019-08-30 8:14 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-08-19 14:15 [PATCH] crypto: arm64/aegis128 - use explicit vector load for permute vectors Ard Biesheuvel
2019-08-19 15:44 ` Nathan Chancellor
2019-08-30 8:14 ` Herbert Xu
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.