From: Ard Biesheuvel <ard.biesheuvel@linaro.org>
To: linux-crypto@vger.kernel.org
Cc: herbert@gondor.apana.org.au, ebiggers@kernel.org,
Ard Biesheuvel <ard.biesheuvel@linaro.org>
Subject: [PATCH v2 03/17] crypto: arm/aes-ce - switch to 4x interleave
Date: Tue, 3 Sep 2019 09:43:25 -0700 [thread overview]
Message-ID: <20190903164339.27984-4-ard.biesheuvel@linaro.org> (raw)
In-Reply-To: <20190903164339.27984-1-ard.biesheuvel@linaro.org>
When the ARM AES instruction based crypto driver was introduced, there
were no known implementations that could benefit from a 4-way interleave,
and so a 3-way interleave was used instead. Since we have sufficient
space in the SIMD register file, let's switch to a 4-way interleave to
align with the 64-bit driver, and to ensure that we can reach optimum
performance when running under emulation on high end 64-bit cores.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
arch/arm/crypto/aes-ce-core.S | 263 +++++++++++---------
1 file changed, 144 insertions(+), 119 deletions(-)
diff --git a/arch/arm/crypto/aes-ce-core.S b/arch/arm/crypto/aes-ce-core.S
index 1e0d45183590..a3ca4ac2d7bb 100644
--- a/arch/arm/crypto/aes-ce-core.S
+++ b/arch/arm/crypto/aes-ce-core.S
@@ -44,46 +44,56 @@
veor q0, q0, \key3
.endm
- .macro enc_dround_3x, key1, key2
+ .macro enc_dround_4x, key1, key2
enc_round q0, \key1
enc_round q1, \key1
enc_round q2, \key1
+ enc_round q3, \key1
enc_round q0, \key2
enc_round q1, \key2
enc_round q2, \key2
+ enc_round q3, \key2
.endm
- .macro dec_dround_3x, key1, key2
+ .macro dec_dround_4x, key1, key2
dec_round q0, \key1
dec_round q1, \key1
dec_round q2, \key1
+ dec_round q3, \key1
dec_round q0, \key2
dec_round q1, \key2
dec_round q2, \key2
+ dec_round q3, \key2
.endm
- .macro enc_fround_3x, key1, key2, key3
+ .macro enc_fround_4x, key1, key2, key3
enc_round q0, \key1
enc_round q1, \key1
enc_round q2, \key1
+ enc_round q3, \key1
aese.8 q0, \key2
aese.8 q1, \key2
aese.8 q2, \key2
+ aese.8 q3, \key2
veor q0, q0, \key3
veor q1, q1, \key3
veor q2, q2, \key3
+ veor q3, q3, \key3
.endm
- .macro dec_fround_3x, key1, key2, key3
+ .macro dec_fround_4x, key1, key2, key3
dec_round q0, \key1
dec_round q1, \key1
dec_round q2, \key1
+ dec_round q3, \key1
aesd.8 q0, \key2
aesd.8 q1, \key2
aesd.8 q2, \key2
+ aesd.8 q3, \key2
veor q0, q0, \key3
veor q1, q1, \key3
veor q2, q2, \key3
+ veor q3, q3, \key3
.endm
.macro do_block, dround, fround
@@ -114,8 +124,9 @@
* transforms. These should preserve all registers except q0 - q2 and ip
* Arguments:
* q0 : first in/output block
- * q1 : second in/output block (_3x version only)
- * q2 : third in/output block (_3x version only)
+ * q1 : second in/output block (_4x version only)
+ * q2 : third in/output block (_4x version only)
+ * q3 : fourth in/output block (_4x version only)
* q8 : first round key
* q9 : secound round key
* q14 : final round key
@@ -136,16 +147,16 @@ aes_decrypt:
ENDPROC(aes_decrypt)
.align 6
-aes_encrypt_3x:
+aes_encrypt_4x:
add ip, r2, #32 @ 3rd round key
- do_block enc_dround_3x, enc_fround_3x
-ENDPROC(aes_encrypt_3x)
+ do_block enc_dround_4x, enc_fround_4x
+ENDPROC(aes_encrypt_4x)
.align 6
-aes_decrypt_3x:
+aes_decrypt_4x:
add ip, r2, #32 @ 3rd round key
- do_block dec_dround_3x, dec_fround_3x
-ENDPROC(aes_decrypt_3x)
+ do_block dec_dround_4x, dec_fround_4x
+ENDPROC(aes_decrypt_4x)
.macro prepare_key, rk, rounds
add ip, \rk, \rounds, lsl #4
@@ -163,17 +174,17 @@ ENTRY(ce_aes_ecb_encrypt)
push {r4, lr}
ldr r4, [sp, #8]
prepare_key r2, r3
-.Lecbencloop3x:
- subs r4, r4, #3
+.Lecbencloop4x:
+ subs r4, r4, #4
bmi .Lecbenc1x
vld1.8 {q0-q1}, [r1]!
- vld1.8 {q2}, [r1]!
- bl aes_encrypt_3x
+ vld1.8 {q2-q3}, [r1]!
+ bl aes_encrypt_4x
vst1.8 {q0-q1}, [r0]!
- vst1.8 {q2}, [r0]!
- b .Lecbencloop3x
+ vst1.8 {q2-q3}, [r0]!
+ b .Lecbencloop4x
.Lecbenc1x:
- adds r4, r4, #3
+ adds r4, r4, #4
beq .Lecbencout
.Lecbencloop:
vld1.8 {q0}, [r1]!
@@ -189,17 +200,17 @@ ENTRY(ce_aes_ecb_decrypt)
push {r4, lr}
ldr r4, [sp, #8]
prepare_key r2, r3
-.Lecbdecloop3x:
- subs r4, r4, #3
+.Lecbdecloop4x:
+ subs r4, r4, #4
bmi .Lecbdec1x
vld1.8 {q0-q1}, [r1]!
- vld1.8 {q2}, [r1]!
- bl aes_decrypt_3x
+ vld1.8 {q2-q3}, [r1]!
+ bl aes_decrypt_4x
vst1.8 {q0-q1}, [r0]!
- vst1.8 {q2}, [r0]!
- b .Lecbdecloop3x
+ vst1.8 {q2-q3}, [r0]!
+ b .Lecbdecloop4x
.Lecbdec1x:
- adds r4, r4, #3
+ adds r4, r4, #4
beq .Lecbdecout
.Lecbdecloop:
vld1.8 {q0}, [r1]!
@@ -236,38 +247,40 @@ ENDPROC(ce_aes_cbc_encrypt)
ENTRY(ce_aes_cbc_decrypt)
push {r4-r6, lr}
ldrd r4, r5, [sp, #16]
- vld1.8 {q6}, [r5] @ keep iv in q6
+ vld1.8 {q15}, [r5] @ keep iv in q15
prepare_key r2, r3
-.Lcbcdecloop3x:
- subs r4, r4, #3
+.Lcbcdecloop4x:
+ subs r4, r4, #4
bmi .Lcbcdec1x
vld1.8 {q0-q1}, [r1]!
- vld1.8 {q2}, [r1]!
- vmov q3, q0
- vmov q4, q1
- vmov q5, q2
- bl aes_decrypt_3x
- veor q0, q0, q6
- veor q1, q1, q3
- veor q2, q2, q4
- vmov q6, q5
+ vld1.8 {q2-q3}, [r1]!
+ vmov q4, q0
+ vmov q5, q1
+ vmov q6, q2
+ vmov q7, q3
+ bl aes_decrypt_4x
+ veor q0, q0, q15
+ veor q1, q1, q4
+ veor q2, q2, q5
+ veor q3, q3, q6
+ vmov q15, q7
vst1.8 {q0-q1}, [r0]!
- vst1.8 {q2}, [r0]!
- b .Lcbcdecloop3x
+ vst1.8 {q2-q3}, [r0]!
+ b .Lcbcdecloop4x
.Lcbcdec1x:
- adds r4, r4, #3
+ adds r4, r4, #4
beq .Lcbcdecout
- vmov q15, q14 @ preserve last round key
+ vmov q6, q14 @ preserve last round key
.Lcbcdecloop:
vld1.8 {q0}, [r1]! @ get next ct block
veor q14, q15, q6 @ combine prev ct with last key
- vmov q6, q0
+ vmov q15, q0
bl aes_decrypt
vst1.8 {q0}, [r0]!
subs r4, r4, #1
bne .Lcbcdecloop
.Lcbcdecout:
- vst1.8 {q6}, [r5] @ keep iv in q6
+ vst1.8 {q15}, [r5] @ keep iv in q15
pop {r4-r6, pc}
ENDPROC(ce_aes_cbc_decrypt)
@@ -278,46 +291,52 @@ ENDPROC(ce_aes_cbc_decrypt)
ENTRY(ce_aes_ctr_encrypt)
push {r4-r6, lr}
ldrd r4, r5, [sp, #16]
- vld1.8 {q6}, [r5] @ load ctr
+ vld1.8 {q7}, [r5] @ load ctr
prepare_key r2, r3
- vmov r6, s27 @ keep swabbed ctr in r6
+ vmov r6, s31 @ keep swabbed ctr in r6
rev r6, r6
cmn r6, r4 @ 32 bit overflow?
bcs .Lctrloop
-.Lctrloop3x:
- subs r4, r4, #3
+.Lctrloop4x:
+ subs r4, r4, #4
bmi .Lctr1x
add r6, r6, #1
- vmov q0, q6
- vmov q1, q6
+ vmov q0, q7
+ vmov q1, q7
rev ip, r6
add r6, r6, #1
- vmov q2, q6
+ vmov q2, q7
vmov s7, ip
rev ip, r6
add r6, r6, #1
+ vmov q3, q7
vmov s11, ip
- vld1.8 {q3-q4}, [r1]!
- vld1.8 {q5}, [r1]!
- bl aes_encrypt_3x
- veor q0, q0, q3
- veor q1, q1, q4
- veor q2, q2, q5
+ rev ip, r6
+ add r6, r6, #1
+ vmov s15, ip
+ vld1.8 {q4-q5}, [r1]!
+ vld1.8 {q6}, [r1]!
+ vld1.8 {q15}, [r1]!
+ bl aes_encrypt_4x
+ veor q0, q0, q4
+ veor q1, q1, q5
+ veor q2, q2, q6
+ veor q3, q3, q15
rev ip, r6
vst1.8 {q0-q1}, [r0]!
- vst1.8 {q2}, [r0]!
- vmov s27, ip
- b .Lctrloop3x
+ vst1.8 {q2-q3}, [r0]!
+ vmov s31, ip
+ b .Lctrloop4x
.Lctr1x:
- adds r4, r4, #3
+ adds r4, r4, #4
beq .Lctrout
.Lctrloop:
- vmov q0, q6
+ vmov q0, q7
bl aes_encrypt
adds r6, r6, #1 @ increment BE ctr
rev ip, r6
- vmov s27, ip
+ vmov s31, ip
bcs .Lctrcarry
.Lctrcarrydone:
@@ -329,7 +348,7 @@ ENTRY(ce_aes_ctr_encrypt)
bne .Lctrloop
.Lctrout:
- vst1.8 {q6}, [r5] @ return next CTR value
+ vst1.8 {q7}, [r5] @ return next CTR value
pop {r4-r6, pc}
.Lctrtailblock:
@@ -337,7 +356,7 @@ ENTRY(ce_aes_ctr_encrypt)
b .Lctrout
.Lctrcarry:
- .irp sreg, s26, s25, s24
+ .irp sreg, s30, s29, s28
vmov ip, \sreg @ load next word of ctr
rev ip, ip @ ... to handle the carry
adds ip, ip, #1
@@ -368,8 +387,8 @@ ENDPROC(ce_aes_ctr_encrypt)
.quad 1, 0x87
ce_aes_xts_init:
- vldr d14, .Lxts_mul_x
- vldr d15, .Lxts_mul_x + 8
+ vldr d30, .Lxts_mul_x
+ vldr d31, .Lxts_mul_x + 8
ldrd r4, r5, [sp, #16] @ load args
ldr r6, [sp, #28]
@@ -390,48 +409,51 @@ ENTRY(ce_aes_xts_encrypt)
bl ce_aes_xts_init @ run shared prologue
prepare_key r2, r3
- vmov q3, q0
+ vmov q4, q0
teq r6, #0 @ start of a block?
- bne .Lxtsenc3x
+ bne .Lxtsenc4x
-.Lxtsencloop3x:
- next_tweak q3, q3, q7, q6
-.Lxtsenc3x:
- subs r4, r4, #3
+.Lxtsencloop4x:
+ next_tweak q4, q4, q15, q10
+.Lxtsenc4x:
+ subs r4, r4, #4
bmi .Lxtsenc1x
- vld1.8 {q0-q1}, [r1]! @ get 3 pt blocks
- vld1.8 {q2}, [r1]!
- next_tweak q4, q3, q7, q6
- veor q0, q0, q3
- next_tweak q5, q4, q7, q6
- veor q1, q1, q4
- veor q2, q2, q5
- bl aes_encrypt_3x
- veor q0, q0, q3
- veor q1, q1, q4
- veor q2, q2, q5
- vst1.8 {q0-q1}, [r0]! @ write 3 ct blocks
- vst1.8 {q2}, [r0]!
- vmov q3, q5
+ vld1.8 {q0-q1}, [r1]! @ get 4 pt blocks
+ vld1.8 {q2-q3}, [r1]!
+ next_tweak q5, q4, q15, q10
+ veor q0, q0, q4
+ next_tweak q6, q5, q15, q10
+ veor q1, q1, q5
+ next_tweak q7, q6, q15, q10
+ veor q2, q2, q6
+ veor q3, q3, q7
+ bl aes_encrypt_4x
+ veor q0, q0, q4
+ veor q1, q1, q5
+ veor q2, q2, q6
+ veor q3, q3, q7
+ vst1.8 {q0-q1}, [r0]! @ write 4 ct blocks
+ vst1.8 {q2-q3}, [r0]!
+ vmov q4, q7
teq r4, #0
beq .Lxtsencout
- b .Lxtsencloop3x
+ b .Lxtsencloop4x
.Lxtsenc1x:
- adds r4, r4, #3
+ adds r4, r4, #4
beq .Lxtsencout
.Lxtsencloop:
vld1.8 {q0}, [r1]!
- veor q0, q0, q3
+ veor q0, q0, q4
bl aes_encrypt
- veor q0, q0, q3
+ veor q0, q0, q4
vst1.8 {q0}, [r0]!
subs r4, r4, #1
beq .Lxtsencout
- next_tweak q3, q3, q7, q6
+ next_tweak q4, q4, q15, q6
b .Lxtsencloop
.Lxtsencout:
- vst1.8 {q3}, [r5]
+ vst1.8 {q4}, [r5]
pop {r4-r6, pc}
ENDPROC(ce_aes_xts_encrypt)
@@ -441,49 +463,52 @@ ENTRY(ce_aes_xts_decrypt)
bl ce_aes_xts_init @ run shared prologue
prepare_key r2, r3
- vmov q3, q0
+ vmov q4, q0
teq r6, #0 @ start of a block?
- bne .Lxtsdec3x
+ bne .Lxtsdec4x
-.Lxtsdecloop3x:
- next_tweak q3, q3, q7, q6
-.Lxtsdec3x:
- subs r4, r4, #3
+.Lxtsdecloop4x:
+ next_tweak q4, q4, q15, q10
+.Lxtsdec4x:
+ subs r4, r4, #4
bmi .Lxtsdec1x
- vld1.8 {q0-q1}, [r1]! @ get 3 ct blocks
- vld1.8 {q2}, [r1]!
- next_tweak q4, q3, q7, q6
- veor q0, q0, q3
- next_tweak q5, q4, q7, q6
- veor q1, q1, q4
- veor q2, q2, q5
- bl aes_decrypt_3x
- veor q0, q0, q3
- veor q1, q1, q4
- veor q2, q2, q5
- vst1.8 {q0-q1}, [r0]! @ write 3 pt blocks
- vst1.8 {q2}, [r0]!
- vmov q3, q5
+ vld1.8 {q0-q1}, [r1]! @ get 4 ct blocks
+ vld1.8 {q2-q3}, [r1]!
+ next_tweak q5, q4, q15, q10
+ veor q0, q0, q4
+ next_tweak q6, q5, q15, q10
+ veor q1, q1, q5
+ next_tweak q7, q6, q15, q10
+ veor q2, q2, q6
+ veor q3, q3, q7
+ bl aes_decrypt_4x
+ veor q0, q0, q4
+ veor q1, q1, q5
+ veor q2, q2, q6
+ veor q3, q3, q7
+ vst1.8 {q0-q1}, [r0]! @ write 4 pt blocks
+ vst1.8 {q2-q3}, [r0]!
+ vmov q4, q7
teq r4, #0
beq .Lxtsdecout
- b .Lxtsdecloop3x
+ b .Lxtsdecloop4x
.Lxtsdec1x:
- adds r4, r4, #3
+ adds r4, r4, #4
beq .Lxtsdecout
.Lxtsdecloop:
vld1.8 {q0}, [r1]!
- veor q0, q0, q3
+ veor q0, q0, q4
add ip, r2, #32 @ 3rd round key
bl aes_decrypt
- veor q0, q0, q3
+ veor q0, q0, q4
vst1.8 {q0}, [r0]!
subs r4, r4, #1
beq .Lxtsdecout
- next_tweak q3, q3, q7, q6
+ next_tweak q4, q4, q15, q6
b .Lxtsdecloop
.Lxtsdecout:
- vst1.8 {q3}, [r5]
+ vst1.8 {q4}, [r5]
pop {r4-r6, pc}
ENDPROC(ce_aes_xts_decrypt)
--
2.17.1
next prev parent reply other threads:[~2019-09-03 16:43 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-09-03 16:43 [PATCH v2 00/17] crypto: arm/aes - XTS ciphertext stealing and other updates Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 01/17] crypto: arm/aes - fix round key prototypes Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 02/17] crypto: arm/aes-ce - yield the SIMD unit between scatterwalk steps Ard Biesheuvel
2019-09-03 16:43 ` Ard Biesheuvel [this message]
2019-09-03 16:43 ` [PATCH v2 04/17] crypto: arm/aes-ce - replace tweak mask literal with composition Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 05/17] crypto: arm/aes-neonbs " Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 06/17] crypto: arm64/aes-neonbs " Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 07/17] crypto: arm64/aes-neon - limit exposed routines if faster driver is enabled Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 08/17] crypto: skcipher - add the ability to abort a skcipher walk Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 09/17] crypto: arm64/aes-cts-cbc-ce - performance tweak Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 10/17] crypto: arm64/aes-cts-cbc - move request context data to the stack Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 11/17] crypto: arm64/aes - implement support for XTS ciphertext stealing Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 12/17] crypto: arm64/aes-neonbs - implement ciphertext stealing for XTS Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 13/17] crypto: arm/aes-ce " Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 14/17] crypto: arm/aes-neonbs " Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 15/17] crypto: arm/aes-ce - implement ciphertext stealing for CBC Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 16/17] crypto: testmgr - add test vectors for XTS ciphertext stealing Ard Biesheuvel
2019-09-03 16:43 ` [PATCH v2 17/17] crypto: testmgr - Add additional AES-XTS vectors for covering CTS Ard Biesheuvel
2019-09-09 7:52 ` [PATCH v2 00/17] crypto: arm/aes - XTS ciphertext stealing and other updates Herbert Xu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20190903164339.27984-4-ard.biesheuvel@linaro.org \
--to=ard.biesheuvel@linaro.org \
--cc=ebiggers@kernel.org \
--cc=herbert@gondor.apana.org.au \
--cc=linux-crypto@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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.