Squeeze out another 5% of performance by minimizing the number
of invocations of kernel_neon_begin()/kernel_neon_end() on the
common path, which also allows some reloads of the key schedule
to be optimized away.
The resulting code runs at 2.3 cycles per byte on a Cortex-A53.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>
---
Raw numbers after the patch.
arch/arm64/crypto/ghash-ce-core.S | 9 ++-
arch/arm64/crypto/ghash-ce-glue.c | 81 +++++++++++---------
2 files changed, 49 insertions(+), 41 deletions(-)
diff --git a/arch/arm64/crypto/ghash-ce-core.S b/arch/arm64/crypto/ghash-ce-core.S
index f7281e7a592f..913e49932ae6 100644
--- a/arch/arm64/crypto/ghash-ce-core.S
+++ b/arch/arm64/crypto/ghash-ce-core.S
@@ -1,7 +1,7 @@
/*
* Accelerated GHASH implementation with ARMv8 PMULL instructions.
*
- * Copyright (C) 2014 - 2017 Linaro Ltd. <ard.biesheuvel@xxxxxxxxxx>
+ * Copyright (C) 2014 - 2018 Linaro Ltd. <ard.biesheuvel@xxxxxxxxxx>
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 as published
@@ -332,8 +332,6 @@ ENDPROC(pmull_ghash_update_p8)
ld1 {XL.2d}, [x1]
ldr x8, [x5, #8] // load lower counter
- load_round_keys w7, x6
-
movi MASK.16b, #0xe1
trn1 SHASH2.2d, SHASH.2d, HH.2d
trn2 T1.2d, SHASH.2d, HH.2d
@@ -346,6 +344,8 @@ CPU_LE( rev x8, x8 )
ld1 {KS0.16b-KS1.16b}, [x10]
.endif
+ cbnz x6, 4f
+
0: ld1 {INP0.16b-INP1.16b}, [x3], #32
rev x9, x8
@@ -471,6 +471,9 @@ CPU_LE( rev x8, x8 )
enc_round KS0, v20
enc_round KS1, v20
b 1b
+
+4: load_round_keys w7, x6
+ b 0b
.endm
/*
diff --git a/arch/arm64/crypto/ghash-ce-glue.c b/arch/arm64/crypto/ghash-ce-glue.c
index c41ac62c90e9..88e3d93fa7c7 100644
--- a/arch/arm64/crypto/ghash-ce-glue.c
+++ b/arch/arm64/crypto/ghash-ce-glue.c
@@ -1,7 +1,7 @@
/*
* Accelerated GHASH implementation with ARMv8 PMULL instructions.
*
- * Copyright (C) 2014 - 2017 Linaro Ltd. <ard.biesheuvel@xxxxxxxxxx>
+ * Copyright (C) 2014 - 2018 Linaro Ltd. <ard.biesheuvel@xxxxxxxxxx>
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 as published
@@ -374,37 +374,39 @@ static int gcm_encrypt(struct aead_request *req)
memcpy(iv, req->iv, GCM_IV_SIZE);
put_unaligned_be32(1, iv + GCM_IV_SIZE);
- if (likely(may_use_simd())) {
- kernel_neon_begin();
+ err = skcipher_walk_aead_encrypt(&walk, req, false);
+ if (likely(may_use_simd() && walk.total >= 2 * AES_BLOCK_SIZE)) {
+ u32 const *rk = NULL;
+
+ kernel_neon_begin();
pmull_gcm_encrypt_block(tag, iv, ctx->aes_key.key_enc, nrounds);
put_unaligned_be32(2, iv + GCM_IV_SIZE);
pmull_gcm_encrypt_block(ks, iv, NULL, nrounds);
put_unaligned_be32(3, iv + GCM_IV_SIZE);
pmull_gcm_encrypt_block(ks + AES_BLOCK_SIZE, iv, NULL, nrounds);
put_unaligned_be32(4, iv + GCM_IV_SIZE);
- kernel_neon_end();
-
- err = skcipher_walk_aead_encrypt(&walk, req, false);
- while (walk.nbytes >= 2 * AES_BLOCK_SIZE) {
+ do {
int blocks = walk.nbytes / (2 * AES_BLOCK_SIZE) * 2;
- kernel_neon_begin();
+ if (rk)
+ kernel_neon_begin();
+
pmull_gcm_encrypt(blocks, dg, walk.dst.virt.addr,
walk.src.virt.addr, ctx->h2, iv,
- ctx->aes_key.key_enc, nrounds, ks);
+ rk, nrounds, ks);
kernel_neon_end();
err = skcipher_walk_done(&walk,
walk.nbytes % (2 * AES_BLOCK_SIZE));
- }
+
+ rk = ctx->aes_key.key_enc;
+ } while (walk.nbytes >= 2 * AES_BLOCK_SIZE);
} else {
__aes_arm64_encrypt(ctx->aes_key.key_enc, tag, iv, nrounds);
put_unaligned_be32(2, iv + GCM_IV_SIZE);
- err = skcipher_walk_aead_encrypt(&walk, req, false);
-
while (walk.nbytes >= AES_BLOCK_SIZE) {
int blocks = walk.nbytes / AES_BLOCK_SIZE;
u8 *dst = walk.dst.virt.addr;
@@ -486,50 +488,53 @@ static int gcm_decrypt(struct aead_request *req)
memcpy(iv, req->iv, GCM_IV_SIZE);
put_unaligned_be32(1, iv + GCM_IV_SIZE);
- if (likely(may_use_simd())) {
+ err = skcipher_walk_aead_decrypt(&walk, req, false);
+
+ if (likely(may_use_simd() && walk.total >= 2 * AES_BLOCK_SIZE)) {
+ u32 const *rk = NULL;
+
kernel_neon_begin();
pmull_gcm_encrypt_block(tag, iv, ctx->aes_key.key_enc, nrounds);
put_unaligned_be32(2, iv + GCM_IV_SIZE);
- kernel_neon_end();
- err = skcipher_walk_aead_decrypt(&walk, req, false);
-
- while (walk.nbytes >= 2 * AES_BLOCK_SIZE) {
+ do {
int blocks = walk.nbytes / (2 * AES_BLOCK_SIZE) * 2;
+ int rem = walk.total - blocks * AES_BLOCK_SIZE;
+
+ if (rk)
+ kernel_neon_begin();
- kernel_neon_begin();
pmull_gcm_decrypt(blocks, dg, walk.dst.virt.addr,
walk.src.virt.addr, ctx->h2, iv,
- ctx->aes_key.key_enc, nrounds);
- kernel_neon_end();
+ rk, nrounds);
- err = skcipher_walk_done(&walk,
- walk.nbytes % (2 * AES_BLOCK_SIZE));
- }
+ /* check if this is the final iteration of the loop */
+ if (rem < (2 * AES_BLOCK_SIZE)) {
+ u8 *iv2 = iv + AES_BLOCK_SIZE;
- if (walk.nbytes) {
- u8 *iv2 = iv + AES_BLOCK_SIZE;
+ if (rem > AES_BLOCK_SIZE) {
+ memcpy(iv2, iv, AES_BLOCK_SIZE);
+ crypto_inc(iv2, AES_BLOCK_SIZE);
+ }
- if (walk.nbytes > AES_BLOCK_SIZE) {
- memcpy(iv2, iv, AES_BLOCK_SIZE);
- crypto_inc(iv2, AES_BLOCK_SIZE);
- }
+ pmull_gcm_encrypt_block(iv, iv, NULL, nrounds);
- kernel_neon_begin();
- pmull_gcm_encrypt_block(iv, iv, ctx->aes_key.key_enc,
- nrounds);
+ if (rem > AES_BLOCK_SIZE)
+ pmull_gcm_encrypt_block(iv2, iv2, NULL,
+ nrounds);
+ }
- if (walk.nbytes > AES_BLOCK_SIZE)
- pmull_gcm_encrypt_block(iv2, iv2, NULL,
- nrounds);
kernel_neon_end();
- }
+
+ err = skcipher_walk_done(&walk,
+ walk.nbytes % (2 * AES_BLOCK_SIZE));
+
+ rk = ctx->aes_key.key_enc;
+ } while (walk.nbytes >= 2 * AES_BLOCK_SIZE);
} else {
__aes_arm64_encrypt(ctx->aes_key.key_enc, tag, iv, nrounds);
put_unaligned_be32(2, iv + GCM_IV_SIZE);
- err = skcipher_walk_aead_decrypt(&walk, req, false);
-
while (walk.nbytes >= AES_BLOCK_SIZE) {
int blocks = walk.nbytes / AES_BLOCK_SIZE;
u8 *dst = walk.dst.virt.addr;
--
2.18.0
testing speed of gcm(aes) (gcm-aes-ce) encryption
test 0 (128 bit key, 16 byte blocks): 365343 operations in 1 seconds ( 5845488 bytes)
test 1 (128 bit key, 64 byte blocks): 504620 operations in 1 seconds ( 32295680 bytes)
test 2 (128 bit key, 256 byte blocks): 418881 operations in 1 seconds (107233536 bytes)
test 3 (128 bit key, 512 byte blocks): 343166 operations in 1 seconds (175700992 bytes)
test 4 (128 bit key, 1024 byte blocks): 252229 operations in 1 seconds (258282496 bytes)
test 5 (128 bit key, 2048 byte blocks): 164862 operations in 1 seconds (337637376 bytes)
test 6 (128 bit key, 4096 byte blocks): 98274 operations in 1 seconds (402530304 bytes)
test 7 (128 bit key, 8192 byte blocks): 52530 operations in 1 seconds (430325760 bytes)
test 8 (192 bit key, 16 byte blocks): 343221 operations in 1 seconds ( 5491536 bytes)
test 9 (192 bit key, 64 byte blocks): 495929 operations in 1 seconds ( 31739456 bytes)
test 10 (192 bit key, 256 byte blocks): 404755 operations in 1 seconds (103617280 bytes)
test 11 (192 bit key, 512 byte blocks): 326728 operations in 1 seconds (167284736 bytes)
test 12 (192 bit key, 1024 byte blocks): 235987 operations in 1 seconds (241650688 bytes)
test 13 (192 bit key, 2048 byte blocks): 151724 operations in 1 seconds (310730752 bytes)
test 14 (192 bit key, 4096 byte blocks): 89285 operations in 1 seconds (365711360 bytes)
test 15 (192 bit key, 8192 byte blocks): 47432 operations in 1 seconds (388562944 bytes)
test 16 (256 bit key, 16 byte blocks): 323574 operations in 1 seconds ( 5177184 bytes)
test 17 (256 bit key, 64 byte blocks): 489854 operations in 1 seconds ( 31350656 bytes)
test 18 (256 bit key, 256 byte blocks): 396979 operations in 1 seconds (101626624 bytes)
test 19 (256 bit key, 512 byte blocks): 317923 operations in 1 seconds (162776576 bytes)
test 20 (256 bit key, 1024 byte blocks): 211440 operations in 1 seconds (216514560 bytes)
test 21 (256 bit key, 2048 byte blocks): 145407 operations in 1 seconds (297793536 bytes)
test 22 (256 bit key, 4096 byte blocks): 85050 operations in 1 seconds (348364800 bytes)
test 23 (256 bit key, 8192 byte blocks): 45068 operations in 1 seconds (369197056 bytes)
