* [PATCH v2 1/7] crypto: arm64/chacha20 - implement NEON version based on SSE3 code
2017-01-11 16:41 [PATCH v2 0/7] crypto: ARM/arm64 - AES and ChaCha20 updates for v4.11 Ard Biesheuvel
@ 2017-01-11 16:41 ` Ard Biesheuvel
2017-01-11 16:41 ` [PATCH v2 2/7] crypto: arm/chacha20 " Ard Biesheuvel
` (5 subsequent siblings)
6 siblings, 0 replies; 10+ messages in thread
From: Ard Biesheuvel @ 2017-01-11 16:41 UTC (permalink / raw)
To: linux-crypto; +Cc: herbert, linux-arm-kernel, Ard Biesheuvel
This is a straight port to arm64/NEON of the x86 SSE3 implementation
of the ChaCha20 stream cipher. It uses the new skcipher walksize
attribute to process the input in strides of 4x the block size.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
arch/arm64/crypto/Kconfig | 6 +
arch/arm64/crypto/Makefile | 3 +
arch/arm64/crypto/chacha20-neon-core.S | 450 ++++++++++++++++++++
arch/arm64/crypto/chacha20-neon-glue.c | 127 ++++++
4 files changed, 586 insertions(+)
diff --git a/arch/arm64/crypto/Kconfig b/arch/arm64/crypto/Kconfig
index 450a85df041a..0bf0f531f539 100644
--- a/arch/arm64/crypto/Kconfig
+++ b/arch/arm64/crypto/Kconfig
@@ -72,4 +72,10 @@ config CRYPTO_CRC32_ARM64
depends on ARM64
select CRYPTO_HASH
+config CRYPTO_CHACHA20_NEON
+ tristate "NEON accelerated ChaCha20 symmetric cipher"
+ depends on KERNEL_MODE_NEON
+ select CRYPTO_BLKCIPHER
+ select CRYPTO_CHACHA20
+
endif
diff --git a/arch/arm64/crypto/Makefile b/arch/arm64/crypto/Makefile
index aa8888d7b744..9d2826c5fccf 100644
--- a/arch/arm64/crypto/Makefile
+++ b/arch/arm64/crypto/Makefile
@@ -41,6 +41,9 @@ sha256-arm64-y := sha256-glue.o sha256-core.o
obj-$(CONFIG_CRYPTO_SHA512_ARM64) += sha512-arm64.o
sha512-arm64-y := sha512-glue.o sha512-core.o
+obj-$(CONFIG_CRYPTO_CHACHA20_NEON) += chacha20-neon.o
+chacha20-neon-y := chacha20-neon-core.o chacha20-neon-glue.o
+
AFLAGS_aes-ce.o := -DINTERLEAVE=4
AFLAGS_aes-neon.o := -DINTERLEAVE=4
diff --git a/arch/arm64/crypto/chacha20-neon-core.S b/arch/arm64/crypto/chacha20-neon-core.S
new file mode 100644
index 000000000000..13c85e272c2a
--- /dev/null
+++ b/arch/arm64/crypto/chacha20-neon-core.S
@@ -0,0 +1,450 @@
+/*
+ * ChaCha20 256-bit cipher algorithm, RFC7539, arm64 NEON functions
+ *
+ * Copyright (C) 2016 Linaro, Ltd. <ard.biesheuvel@linaro.org>
+ *
+ * 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 by the Free Software Foundation.
+ *
+ * Based on:
+ * ChaCha20 256-bit cipher algorithm, RFC7539, x64 SSSE3 functions
+ *
+ * Copyright (C) 2015 Martin Willi
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <linux/linkage.h>
+
+ .text
+ .align 6
+
+ENTRY(chacha20_block_xor_neon)
+ // x0: Input state matrix, s
+ // x1: 1 data block output, o
+ // x2: 1 data block input, i
+
+ //
+ // This function encrypts one ChaCha20 block by loading the state matrix
+ // in four NEON registers. It performs matrix operation on four words in
+ // parallel, but requires shuffling to rearrange the words after each
+ // round.
+ //
+
+ // x0..3 = s0..3
+ adr x3, ROT8
+ ld1 {v0.4s-v3.4s}, [x0]
+ ld1 {v8.4s-v11.4s}, [x0]
+ ld1 {v12.4s}, [x3]
+
+ mov x3, #10
+
+.Ldoubleround:
+ // x0 += x1, x3 = rotl32(x3 ^ x0, 16)
+ add v0.4s, v0.4s, v1.4s
+ eor v3.16b, v3.16b, v0.16b
+ rev32 v3.8h, v3.8h
+
+ // x2 += x3, x1 = rotl32(x1 ^ x2, 12)
+ add v2.4s, v2.4s, v3.4s
+ eor v4.16b, v1.16b, v2.16b
+ shl v1.4s, v4.4s, #12
+ sri v1.4s, v4.4s, #20
+
+ // x0 += x1, x3 = rotl32(x3 ^ x0, 8)
+ add v0.4s, v0.4s, v1.4s
+ eor v3.16b, v3.16b, v0.16b
+ tbl v3.16b, {v3.16b}, v12.16b
+
+ // x2 += x3, x1 = rotl32(x1 ^ x2, 7)
+ add v2.4s, v2.4s, v3.4s
+ eor v4.16b, v1.16b, v2.16b
+ shl v1.4s, v4.4s, #7
+ sri v1.4s, v4.4s, #25
+
+ // x1 = shuffle32(x1, MASK(0, 3, 2, 1))
+ ext v1.16b, v1.16b, v1.16b, #4
+ // x2 = shuffle32(x2, MASK(1, 0, 3, 2))
+ ext v2.16b, v2.16b, v2.16b, #8
+ // x3 = shuffle32(x3, MASK(2, 1, 0, 3))
+ ext v3.16b, v3.16b, v3.16b, #12
+
+ // x0 += x1, x3 = rotl32(x3 ^ x0, 16)
+ add v0.4s, v0.4s, v1.4s
+ eor v3.16b, v3.16b, v0.16b
+ rev32 v3.8h, v3.8h
+
+ // x2 += x3, x1 = rotl32(x1 ^ x2, 12)
+ add v2.4s, v2.4s, v3.4s
+ eor v4.16b, v1.16b, v2.16b
+ shl v1.4s, v4.4s, #12
+ sri v1.4s, v4.4s, #20
+
+ // x0 += x1, x3 = rotl32(x3 ^ x0, 8)
+ add v0.4s, v0.4s, v1.4s
+ eor v3.16b, v3.16b, v0.16b
+ tbl v3.16b, {v3.16b}, v12.16b
+
+ // x2 += x3, x1 = rotl32(x1 ^ x2, 7)
+ add v2.4s, v2.4s, v3.4s
+ eor v4.16b, v1.16b, v2.16b
+ shl v1.4s, v4.4s, #7
+ sri v1.4s, v4.4s, #25
+
+ // x1 = shuffle32(x1, MASK(2, 1, 0, 3))
+ ext v1.16b, v1.16b, v1.16b, #12
+ // x2 = shuffle32(x2, MASK(1, 0, 3, 2))
+ ext v2.16b, v2.16b, v2.16b, #8
+ // x3 = shuffle32(x3, MASK(0, 3, 2, 1))
+ ext v3.16b, v3.16b, v3.16b, #4
+
+ subs x3, x3, #1
+ b.ne .Ldoubleround
+
+ ld1 {v4.16b-v7.16b}, [x2]
+
+ // o0 = i0 ^ (x0 + s0)
+ add v0.4s, v0.4s, v8.4s
+ eor v0.16b, v0.16b, v4.16b
+
+ // o1 = i1 ^ (x1 + s1)
+ add v1.4s, v1.4s, v9.4s
+ eor v1.16b, v1.16b, v5.16b
+
+ // o2 = i2 ^ (x2 + s2)
+ add v2.4s, v2.4s, v10.4s
+ eor v2.16b, v2.16b, v6.16b
+
+ // o3 = i3 ^ (x3 + s3)
+ add v3.4s, v3.4s, v11.4s
+ eor v3.16b, v3.16b, v7.16b
+
+ st1 {v0.16b-v3.16b}, [x1]
+
+ ret
+ENDPROC(chacha20_block_xor_neon)
+
+ .align 6
+ENTRY(chacha20_4block_xor_neon)
+ // x0: Input state matrix, s
+ // x1: 4 data blocks output, o
+ // x2: 4 data blocks input, i
+
+ //
+ // This function encrypts four consecutive ChaCha20 blocks by loading
+ // the state matrix in NEON registers four times. The algorithm performs
+ // each operation on the corresponding word of each state matrix, hence
+ // requires no word shuffling. For final XORing step we transpose the
+ // matrix by interleaving 32- and then 64-bit words, which allows us to
+ // do XOR in NEON registers.
+ //
+ adr x3, CTRINC // ... and ROT8
+ ld1 {v30.4s-v31.4s}, [x3]
+
+ // x0..15[0-3] = s0..3[0..3]
+ mov x4, x0
+ ld4r { v0.4s- v3.4s}, [x4], #16
+ ld4r { v4.4s- v7.4s}, [x4], #16
+ ld4r { v8.4s-v11.4s}, [x4], #16
+ ld4r {v12.4s-v15.4s}, [x4]
+
+ // x12 += counter values 0-3
+ add v12.4s, v12.4s, v30.4s
+
+ mov x3, #10
+
+.Ldoubleround4:
+ // x0 += x4, x12 = rotl32(x12 ^ x0, 16)
+ // x1 += x5, x13 = rotl32(x13 ^ x1, 16)
+ // x2 += x6, x14 = rotl32(x14 ^ x2, 16)
+ // x3 += x7, x15 = rotl32(x15 ^ x3, 16)
+ add v0.4s, v0.4s, v4.4s
+ add v1.4s, v1.4s, v5.4s
+ add v2.4s, v2.4s, v6.4s
+ add v3.4s, v3.4s, v7.4s
+
+ eor v12.16b, v12.16b, v0.16b
+ eor v13.16b, v13.16b, v1.16b
+ eor v14.16b, v14.16b, v2.16b
+ eor v15.16b, v15.16b, v3.16b
+
+ rev32 v12.8h, v12.8h
+ rev32 v13.8h, v13.8h
+ rev32 v14.8h, v14.8h
+ rev32 v15.8h, v15.8h
+
+ // x8 += x12, x4 = rotl32(x4 ^ x8, 12)
+ // x9 += x13, x5 = rotl32(x5 ^ x9, 12)
+ // x10 += x14, x6 = rotl32(x6 ^ x10, 12)
+ // x11 += x15, x7 = rotl32(x7 ^ x11, 12)
+ add v8.4s, v8.4s, v12.4s
+ add v9.4s, v9.4s, v13.4s
+ add v10.4s, v10.4s, v14.4s
+ add v11.4s, v11.4s, v15.4s
+
+ eor v16.16b, v4.16b, v8.16b
+ eor v17.16b, v5.16b, v9.16b
+ eor v18.16b, v6.16b, v10.16b
+ eor v19.16b, v7.16b, v11.16b
+
+ shl v4.4s, v16.4s, #12
+ shl v5.4s, v17.4s, #12
+ shl v6.4s, v18.4s, #12
+ shl v7.4s, v19.4s, #12
+
+ sri v4.4s, v16.4s, #20
+ sri v5.4s, v17.4s, #20
+ sri v6.4s, v18.4s, #20
+ sri v7.4s, v19.4s, #20
+
+ // x0 += x4, x12 = rotl32(x12 ^ x0, 8)
+ // x1 += x5, x13 = rotl32(x13 ^ x1, 8)
+ // x2 += x6, x14 = rotl32(x14 ^ x2, 8)
+ // x3 += x7, x15 = rotl32(x15 ^ x3, 8)
+ add v0.4s, v0.4s, v4.4s
+ add v1.4s, v1.4s, v5.4s
+ add v2.4s, v2.4s, v6.4s
+ add v3.4s, v3.4s, v7.4s
+
+ eor v12.16b, v12.16b, v0.16b
+ eor v13.16b, v13.16b, v1.16b
+ eor v14.16b, v14.16b, v2.16b
+ eor v15.16b, v15.16b, v3.16b
+
+ tbl v12.16b, {v12.16b}, v31.16b
+ tbl v13.16b, {v13.16b}, v31.16b
+ tbl v14.16b, {v14.16b}, v31.16b
+ tbl v15.16b, {v15.16b}, v31.16b
+
+ // x8 += x12, x4 = rotl32(x4 ^ x8, 7)
+ // x9 += x13, x5 = rotl32(x5 ^ x9, 7)
+ // x10 += x14, x6 = rotl32(x6 ^ x10, 7)
+ // x11 += x15, x7 = rotl32(x7 ^ x11, 7)
+ add v8.4s, v8.4s, v12.4s
+ add v9.4s, v9.4s, v13.4s
+ add v10.4s, v10.4s, v14.4s
+ add v11.4s, v11.4s, v15.4s
+
+ eor v16.16b, v4.16b, v8.16b
+ eor v17.16b, v5.16b, v9.16b
+ eor v18.16b, v6.16b, v10.16b
+ eor v19.16b, v7.16b, v11.16b
+
+ shl v4.4s, v16.4s, #7
+ shl v5.4s, v17.4s, #7
+ shl v6.4s, v18.4s, #7
+ shl v7.4s, v19.4s, #7
+
+ sri v4.4s, v16.4s, #25
+ sri v5.4s, v17.4s, #25
+ sri v6.4s, v18.4s, #25
+ sri v7.4s, v19.4s, #25
+
+ // x0 += x5, x15 = rotl32(x15 ^ x0, 16)
+ // x1 += x6, x12 = rotl32(x12 ^ x1, 16)
+ // x2 += x7, x13 = rotl32(x13 ^ x2, 16)
+ // x3 += x4, x14 = rotl32(x14 ^ x3, 16)
+ add v0.4s, v0.4s, v5.4s
+ add v1.4s, v1.4s, v6.4s
+ add v2.4s, v2.4s, v7.4s
+ add v3.4s, v3.4s, v4.4s
+
+ eor v15.16b, v15.16b, v0.16b
+ eor v12.16b, v12.16b, v1.16b
+ eor v13.16b, v13.16b, v2.16b
+ eor v14.16b, v14.16b, v3.16b
+
+ rev32 v15.8h, v15.8h
+ rev32 v12.8h, v12.8h
+ rev32 v13.8h, v13.8h
+ rev32 v14.8h, v14.8h
+
+ // x10 += x15, x5 = rotl32(x5 ^ x10, 12)
+ // x11 += x12, x6 = rotl32(x6 ^ x11, 12)
+ // x8 += x13, x7 = rotl32(x7 ^ x8, 12)
+ // x9 += x14, x4 = rotl32(x4 ^ x9, 12)
+ add v10.4s, v10.4s, v15.4s
+ add v11.4s, v11.4s, v12.4s
+ add v8.4s, v8.4s, v13.4s
+ add v9.4s, v9.4s, v14.4s
+
+ eor v16.16b, v5.16b, v10.16b
+ eor v17.16b, v6.16b, v11.16b
+ eor v18.16b, v7.16b, v8.16b
+ eor v19.16b, v4.16b, v9.16b
+
+ shl v5.4s, v16.4s, #12
+ shl v6.4s, v17.4s, #12
+ shl v7.4s, v18.4s, #12
+ shl v4.4s, v19.4s, #12
+
+ sri v5.4s, v16.4s, #20
+ sri v6.4s, v17.4s, #20
+ sri v7.4s, v18.4s, #20
+ sri v4.4s, v19.4s, #20
+
+ // x0 += x5, x15 = rotl32(x15 ^ x0, 8)
+ // x1 += x6, x12 = rotl32(x12 ^ x1, 8)
+ // x2 += x7, x13 = rotl32(x13 ^ x2, 8)
+ // x3 += x4, x14 = rotl32(x14 ^ x3, 8)
+ add v0.4s, v0.4s, v5.4s
+ add v1.4s, v1.4s, v6.4s
+ add v2.4s, v2.4s, v7.4s
+ add v3.4s, v3.4s, v4.4s
+
+ eor v15.16b, v15.16b, v0.16b
+ eor v12.16b, v12.16b, v1.16b
+ eor v13.16b, v13.16b, v2.16b
+ eor v14.16b, v14.16b, v3.16b
+
+ tbl v15.16b, {v15.16b}, v31.16b
+ tbl v12.16b, {v12.16b}, v31.16b
+ tbl v13.16b, {v13.16b}, v31.16b
+ tbl v14.16b, {v14.16b}, v31.16b
+
+ // x10 += x15, x5 = rotl32(x5 ^ x10, 7)
+ // x11 += x12, x6 = rotl32(x6 ^ x11, 7)
+ // x8 += x13, x7 = rotl32(x7 ^ x8, 7)
+ // x9 += x14, x4 = rotl32(x4 ^ x9, 7)
+ add v10.4s, v10.4s, v15.4s
+ add v11.4s, v11.4s, v12.4s
+ add v8.4s, v8.4s, v13.4s
+ add v9.4s, v9.4s, v14.4s
+
+ eor v16.16b, v5.16b, v10.16b
+ eor v17.16b, v6.16b, v11.16b
+ eor v18.16b, v7.16b, v8.16b
+ eor v19.16b, v4.16b, v9.16b
+
+ shl v5.4s, v16.4s, #7
+ shl v6.4s, v17.4s, #7
+ shl v7.4s, v18.4s, #7
+ shl v4.4s, v19.4s, #7
+
+ sri v5.4s, v16.4s, #25
+ sri v6.4s, v17.4s, #25
+ sri v7.4s, v18.4s, #25
+ sri v4.4s, v19.4s, #25
+
+ subs x3, x3, #1
+ b.ne .Ldoubleround4
+
+ ld4r {v16.4s-v19.4s}, [x0], #16
+ ld4r {v20.4s-v23.4s}, [x0], #16
+
+ // x12 += counter values 0-3
+ add v12.4s, v12.4s, v30.4s
+
+ // x0[0-3] += s0[0]
+ // x1[0-3] += s0[1]
+ // x2[0-3] += s0[2]
+ // x3[0-3] += s0[3]
+ add v0.4s, v0.4s, v16.4s
+ add v1.4s, v1.4s, v17.4s
+ add v2.4s, v2.4s, v18.4s
+ add v3.4s, v3.4s, v19.4s
+
+ ld4r {v24.4s-v27.4s}, [x0], #16
+ ld4r {v28.4s-v31.4s}, [x0]
+
+ // x4[0-3] += s1[0]
+ // x5[0-3] += s1[1]
+ // x6[0-3] += s1[2]
+ // x7[0-3] += s1[3]
+ add v4.4s, v4.4s, v20.4s
+ add v5.4s, v5.4s, v21.4s
+ add v6.4s, v6.4s, v22.4s
+ add v7.4s, v7.4s, v23.4s
+
+ // x8[0-3] += s2[0]
+ // x9[0-3] += s2[1]
+ // x10[0-3] += s2[2]
+ // x11[0-3] += s2[3]
+ add v8.4s, v8.4s, v24.4s
+ add v9.4s, v9.4s, v25.4s
+ add v10.4s, v10.4s, v26.4s
+ add v11.4s, v11.4s, v27.4s
+
+ // x12[0-3] += s3[0]
+ // x13[0-3] += s3[1]
+ // x14[0-3] += s3[2]
+ // x15[0-3] += s3[3]
+ add v12.4s, v12.4s, v28.4s
+ add v13.4s, v13.4s, v29.4s
+ add v14.4s, v14.4s, v30.4s
+ add v15.4s, v15.4s, v31.4s
+
+ // interleave 32-bit words in state n, n+1
+ zip1 v16.4s, v0.4s, v1.4s
+ zip2 v17.4s, v0.4s, v1.4s
+ zip1 v18.4s, v2.4s, v3.4s
+ zip2 v19.4s, v2.4s, v3.4s
+ zip1 v20.4s, v4.4s, v5.4s
+ zip2 v21.4s, v4.4s, v5.4s
+ zip1 v22.4s, v6.4s, v7.4s
+ zip2 v23.4s, v6.4s, v7.4s
+ zip1 v24.4s, v8.4s, v9.4s
+ zip2 v25.4s, v8.4s, v9.4s
+ zip1 v26.4s, v10.4s, v11.4s
+ zip2 v27.4s, v10.4s, v11.4s
+ zip1 v28.4s, v12.4s, v13.4s
+ zip2 v29.4s, v12.4s, v13.4s
+ zip1 v30.4s, v14.4s, v15.4s
+ zip2 v31.4s, v14.4s, v15.4s
+
+ // interleave 64-bit words in state n, n+2
+ zip1 v0.2d, v16.2d, v18.2d
+ zip2 v4.2d, v16.2d, v18.2d
+ zip1 v8.2d, v17.2d, v19.2d
+ zip2 v12.2d, v17.2d, v19.2d
+ ld1 {v16.16b-v19.16b}, [x2], #64
+
+ zip1 v1.2d, v20.2d, v22.2d
+ zip2 v5.2d, v20.2d, v22.2d
+ zip1 v9.2d, v21.2d, v23.2d
+ zip2 v13.2d, v21.2d, v23.2d
+ ld1 {v20.16b-v23.16b}, [x2], #64
+
+ zip1 v2.2d, v24.2d, v26.2d
+ zip2 v6.2d, v24.2d, v26.2d
+ zip1 v10.2d, v25.2d, v27.2d
+ zip2 v14.2d, v25.2d, v27.2d
+ ld1 {v24.16b-v27.16b}, [x2], #64
+
+ zip1 v3.2d, v28.2d, v30.2d
+ zip2 v7.2d, v28.2d, v30.2d
+ zip1 v11.2d, v29.2d, v31.2d
+ zip2 v15.2d, v29.2d, v31.2d
+ ld1 {v28.16b-v31.16b}, [x2]
+
+ // xor with corresponding input, write to output
+ eor v16.16b, v16.16b, v0.16b
+ eor v17.16b, v17.16b, v1.16b
+ eor v18.16b, v18.16b, v2.16b
+ eor v19.16b, v19.16b, v3.16b
+ eor v20.16b, v20.16b, v4.16b
+ eor v21.16b, v21.16b, v5.16b
+ st1 {v16.16b-v19.16b}, [x1], #64
+ eor v22.16b, v22.16b, v6.16b
+ eor v23.16b, v23.16b, v7.16b
+ eor v24.16b, v24.16b, v8.16b
+ eor v25.16b, v25.16b, v9.16b
+ st1 {v20.16b-v23.16b}, [x1], #64
+ eor v26.16b, v26.16b, v10.16b
+ eor v27.16b, v27.16b, v11.16b
+ eor v28.16b, v28.16b, v12.16b
+ st1 {v24.16b-v27.16b}, [x1], #64
+ eor v29.16b, v29.16b, v13.16b
+ eor v30.16b, v30.16b, v14.16b
+ eor v31.16b, v31.16b, v15.16b
+ st1 {v28.16b-v31.16b}, [x1]
+
+ ret
+ENDPROC(chacha20_4block_xor_neon)
+
+CTRINC: .word 0, 1, 2, 3
+ROT8: .word 0x02010003, 0x06050407, 0x0a09080b, 0x0e0d0c0f
diff --git a/arch/arm64/crypto/chacha20-neon-glue.c b/arch/arm64/crypto/chacha20-neon-glue.c
new file mode 100644
index 000000000000..a7f2337d46cf
--- /dev/null
+++ b/arch/arm64/crypto/chacha20-neon-glue.c
@@ -0,0 +1,127 @@
+/*
+ * ChaCha20 256-bit cipher algorithm, RFC7539, arm64 NEON functions
+ *
+ * Copyright (C) 2016 Linaro, Ltd. <ard.biesheuvel@linaro.org>
+ *
+ * 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 by the Free Software Foundation.
+ *
+ * Based on:
+ * ChaCha20 256-bit cipher algorithm, RFC7539, SIMD glue code
+ *
+ * Copyright (C) 2015 Martin Willi
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <crypto/algapi.h>
+#include <crypto/chacha20.h>
+#include <crypto/internal/skcipher.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+
+#include <asm/hwcap.h>
+#include <asm/neon.h>
+
+asmlinkage void chacha20_block_xor_neon(u32 *state, u8 *dst, const u8 *src);
+asmlinkage void chacha20_4block_xor_neon(u32 *state, u8 *dst, const u8 *src);
+
+static void chacha20_doneon(u32 *state, u8 *dst, const u8 *src,
+ unsigned int bytes)
+{
+ u8 buf[CHACHA20_BLOCK_SIZE];
+
+ while (bytes >= CHACHA20_BLOCK_SIZE * 4) {
+ chacha20_4block_xor_neon(state, dst, src);
+ bytes -= CHACHA20_BLOCK_SIZE * 4;
+ src += CHACHA20_BLOCK_SIZE * 4;
+ dst += CHACHA20_BLOCK_SIZE * 4;
+ state[12] += 4;
+ }
+ while (bytes >= CHACHA20_BLOCK_SIZE) {
+ chacha20_block_xor_neon(state, dst, src);
+ bytes -= CHACHA20_BLOCK_SIZE;
+ src += CHACHA20_BLOCK_SIZE;
+ dst += CHACHA20_BLOCK_SIZE;
+ state[12]++;
+ }
+ if (bytes) {
+ memcpy(buf, src, bytes);
+ chacha20_block_xor_neon(state, buf, buf);
+ memcpy(dst, buf, bytes);
+ }
+}
+
+static int chacha20_neon(struct skcipher_request *req)
+{
+ struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+ struct chacha20_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct skcipher_walk walk;
+ u32 state[16];
+ int err;
+
+ if (req->cryptlen <= CHACHA20_BLOCK_SIZE)
+ return crypto_chacha20_crypt(req);
+
+ err = skcipher_walk_virt(&walk, req, true);
+
+ crypto_chacha20_init(state, ctx, walk.iv);
+
+ kernel_neon_begin();
+ while (walk.nbytes > 0) {
+ unsigned int nbytes = walk.nbytes;
+
+ if (nbytes < walk.total)
+ nbytes = round_down(nbytes, walk.stride);
+
+ chacha20_doneon(state, walk.dst.virt.addr, walk.src.virt.addr,
+ nbytes);
+ err = skcipher_walk_done(&walk, walk.nbytes - nbytes);
+ }
+ kernel_neon_end();
+
+ return err;
+}
+
+static struct skcipher_alg alg = {
+ .base.cra_name = "chacha20",
+ .base.cra_driver_name = "chacha20-neon",
+ .base.cra_priority = 300,
+ .base.cra_blocksize = 1,
+ .base.cra_ctxsize = sizeof(struct chacha20_ctx),
+ .base.cra_alignmask = 1,
+ .base.cra_module = THIS_MODULE,
+
+ .min_keysize = CHACHA20_KEY_SIZE,
+ .max_keysize = CHACHA20_KEY_SIZE,
+ .ivsize = CHACHA20_IV_SIZE,
+ .chunksize = CHACHA20_BLOCK_SIZE,
+ .walksize = 4 * CHACHA20_BLOCK_SIZE,
+ .setkey = crypto_chacha20_setkey,
+ .encrypt = chacha20_neon,
+ .decrypt = chacha20_neon,
+};
+
+static int __init chacha20_simd_mod_init(void)
+{
+ if (!(elf_hwcap & HWCAP_ASIMD))
+ return -ENODEV;
+
+ return crypto_register_skcipher(&alg);
+}
+
+static void __exit chacha20_simd_mod_fini(void)
+{
+ crypto_unregister_skcipher(&alg);
+}
+
+module_init(chacha20_simd_mod_init);
+module_exit(chacha20_simd_mod_fini);
+
+MODULE_AUTHOR("Ard Biesheuvel <ard.biesheuvel@linaro.org>");
+MODULE_LICENSE("GPL v2");
+MODULE_ALIAS_CRYPTO("chacha20");
--
2.7.4
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH v2 2/7] crypto: arm/chacha20 - implement NEON version based on SSE3 code
2017-01-11 16:41 [PATCH v2 0/7] crypto: ARM/arm64 - AES and ChaCha20 updates for v4.11 Ard Biesheuvel
2017-01-11 16:41 ` [PATCH v2 1/7] crypto: arm64/chacha20 - implement NEON version based on SSE3 code Ard Biesheuvel
@ 2017-01-11 16:41 ` Ard Biesheuvel
2017-01-11 16:41 ` [PATCH v2 3/7] crypto: arm64/aes-blk - expose AES-CTR as synchronous cipher as well Ard Biesheuvel
` (4 subsequent siblings)
6 siblings, 0 replies; 10+ messages in thread
From: Ard Biesheuvel @ 2017-01-11 16:41 UTC (permalink / raw)
To: linux-crypto; +Cc: herbert, linux-arm-kernel, Ard Biesheuvel
This is a straight port to ARM/NEON of the x86 SSE3 implementation
of the ChaCha20 stream cipher. It uses the new skcipher walksize
attribute to process the input in strides of 4x the block size.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
arch/arm/crypto/Kconfig | 6 +
arch/arm/crypto/Makefile | 2 +
arch/arm/crypto/chacha20-neon-core.S | 524 ++++++++++++++++++++
arch/arm/crypto/chacha20-neon-glue.c | 128 +++++
4 files changed, 660 insertions(+)
diff --git a/arch/arm/crypto/Kconfig b/arch/arm/crypto/Kconfig
index 13f1b4c289d4..2f3339f015d3 100644
--- a/arch/arm/crypto/Kconfig
+++ b/arch/arm/crypto/Kconfig
@@ -130,4 +130,10 @@ config CRYPTO_CRC32_ARM_CE
depends on KERNEL_MODE_NEON && CRC32
select CRYPTO_HASH
+config CRYPTO_CHACHA20_NEON
+ tristate "NEON accelerated ChaCha20 symmetric cipher"
+ depends on KERNEL_MODE_NEON
+ select CRYPTO_BLKCIPHER
+ select CRYPTO_CHACHA20
+
endif
diff --git a/arch/arm/crypto/Makefile b/arch/arm/crypto/Makefile
index b578a1820ab1..8d74e55eacd4 100644
--- a/arch/arm/crypto/Makefile
+++ b/arch/arm/crypto/Makefile
@@ -8,6 +8,7 @@ obj-$(CONFIG_CRYPTO_SHA1_ARM) += sha1-arm.o
obj-$(CONFIG_CRYPTO_SHA1_ARM_NEON) += sha1-arm-neon.o
obj-$(CONFIG_CRYPTO_SHA256_ARM) += sha256-arm.o
obj-$(CONFIG_CRYPTO_SHA512_ARM) += sha512-arm.o
+obj-$(CONFIG_CRYPTO_CHACHA20_NEON) += chacha20-neon.o
ce-obj-$(CONFIG_CRYPTO_AES_ARM_CE) += aes-arm-ce.o
ce-obj-$(CONFIG_CRYPTO_SHA1_ARM_CE) += sha1-arm-ce.o
@@ -40,6 +41,7 @@ aes-arm-ce-y := aes-ce-core.o aes-ce-glue.o
ghash-arm-ce-y := ghash-ce-core.o ghash-ce-glue.o
crct10dif-arm-ce-y := crct10dif-ce-core.o crct10dif-ce-glue.o
crc32-arm-ce-y:= crc32-ce-core.o crc32-ce-glue.o
+chacha20-neon-y := chacha20-neon-core.o chacha20-neon-glue.o
quiet_cmd_perl = PERL $@
cmd_perl = $(PERL) $(<) > $(@)
diff --git a/arch/arm/crypto/chacha20-neon-core.S b/arch/arm/crypto/chacha20-neon-core.S
new file mode 100644
index 000000000000..ff1d337bdb4a
--- /dev/null
+++ b/arch/arm/crypto/chacha20-neon-core.S
@@ -0,0 +1,524 @@
+/*
+ * ChaCha20 256-bit cipher algorithm, RFC7539, ARM NEON functions
+ *
+ * Copyright (C) 2016 Linaro, Ltd. <ard.biesheuvel@linaro.org>
+ *
+ * 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 by the Free Software Foundation.
+ *
+ * Based on:
+ * ChaCha20 256-bit cipher algorithm, RFC7539, x64 SSE3 functions
+ *
+ * Copyright (C) 2015 Martin Willi
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <linux/linkage.h>
+
+ .text
+ .fpu neon
+ .align 5
+
+ENTRY(chacha20_block_xor_neon)
+ // r0: Input state matrix, s
+ // r1: 1 data block output, o
+ // r2: 1 data block input, i
+
+ //
+ // This function encrypts one ChaCha20 block by loading the state matrix
+ // in four NEON registers. It performs matrix operation on four words in
+ // parallel, but requireds shuffling to rearrange the words after each
+ // round.
+ //
+
+ // x0..3 = s0..3
+ add ip, r0, #0x20
+ vld1.32 {q0-q1}, [r0]
+ vld1.32 {q2-q3}, [ip]
+
+ vmov q8, q0
+ vmov q9, q1
+ vmov q10, q2
+ vmov q11, q3
+
+ mov r3, #10
+
+.Ldoubleround:
+ // x0 += x1, x3 = rotl32(x3 ^ x0, 16)
+ vadd.i32 q0, q0, q1
+ veor q4, q3, q0
+ vshl.u32 q3, q4, #16
+ vsri.u32 q3, q4, #16
+
+ // x2 += x3, x1 = rotl32(x1 ^ x2, 12)
+ vadd.i32 q2, q2, q3
+ veor q4, q1, q2
+ vshl.u32 q1, q4, #12
+ vsri.u32 q1, q4, #20
+
+ // x0 += x1, x3 = rotl32(x3 ^ x0, 8)
+ vadd.i32 q0, q0, q1
+ veor q4, q3, q0
+ vshl.u32 q3, q4, #8
+ vsri.u32 q3, q4, #24
+
+ // x2 += x3, x1 = rotl32(x1 ^ x2, 7)
+ vadd.i32 q2, q2, q3
+ veor q4, q1, q2
+ vshl.u32 q1, q4, #7
+ vsri.u32 q1, q4, #25
+
+ // x1 = shuffle32(x1, MASK(0, 3, 2, 1))
+ vext.8 q1, q1, q1, #4
+ // x2 = shuffle32(x2, MASK(1, 0, 3, 2))
+ vext.8 q2, q2, q2, #8
+ // x3 = shuffle32(x3, MASK(2, 1, 0, 3))
+ vext.8 q3, q3, q3, #12
+
+ // x0 += x1, x3 = rotl32(x3 ^ x0, 16)
+ vadd.i32 q0, q0, q1
+ veor q4, q3, q0
+ vshl.u32 q3, q4, #16
+ vsri.u32 q3, q4, #16
+
+ // x2 += x3, x1 = rotl32(x1 ^ x2, 12)
+ vadd.i32 q2, q2, q3
+ veor q4, q1, q2
+ vshl.u32 q1, q4, #12
+ vsri.u32 q1, q4, #20
+
+ // x0 += x1, x3 = rotl32(x3 ^ x0, 8)
+ vadd.i32 q0, q0, q1
+ veor q4, q3, q0
+ vshl.u32 q3, q4, #8
+ vsri.u32 q3, q4, #24
+
+ // x2 += x3, x1 = rotl32(x1 ^ x2, 7)
+ vadd.i32 q2, q2, q3
+ veor q4, q1, q2
+ vshl.u32 q1, q4, #7
+ vsri.u32 q1, q4, #25
+
+ // x1 = shuffle32(x1, MASK(2, 1, 0, 3))
+ vext.8 q1, q1, q1, #12
+ // x2 = shuffle32(x2, MASK(1, 0, 3, 2))
+ vext.8 q2, q2, q2, #8
+ // x3 = shuffle32(x3, MASK(0, 3, 2, 1))
+ vext.8 q3, q3, q3, #4
+
+ subs r3, r3, #1
+ bne .Ldoubleround
+
+ add ip, r2, #0x20
+ vld1.8 {q4-q5}, [r2]
+ vld1.8 {q6-q7}, [ip]
+
+ // o0 = i0 ^ (x0 + s0)
+ vadd.i32 q0, q0, q8
+ veor q0, q0, q4
+
+ // o1 = i1 ^ (x1 + s1)
+ vadd.i32 q1, q1, q9
+ veor q1, q1, q5
+
+ // o2 = i2 ^ (x2 + s2)
+ vadd.i32 q2, q2, q10
+ veor q2, q2, q6
+
+ // o3 = i3 ^ (x3 + s3)
+ vadd.i32 q3, q3, q11
+ veor q3, q3, q7
+
+ add ip, r1, #0x20
+ vst1.8 {q0-q1}, [r1]
+ vst1.8 {q2-q3}, [ip]
+
+ bx lr
+ENDPROC(chacha20_block_xor_neon)
+
+ .align 5
+ENTRY(chacha20_4block_xor_neon)
+ push {r4-r6, lr}
+ mov ip, sp // preserve the stack pointer
+ sub r3, sp, #0x20 // allocate a 32 byte buffer
+ bic r3, r3, #0x1f // aligned to 32 bytes
+ mov sp, r3
+
+ // r0: Input state matrix, s
+ // r1: 4 data blocks output, o
+ // r2: 4 data blocks input, i
+
+ //
+ // This function encrypts four consecutive ChaCha20 blocks by loading
+ // the state matrix in NEON registers four times. The algorithm performs
+ // each operation on the corresponding word of each state matrix, hence
+ // requires no word shuffling. For final XORing step we transpose the
+ // matrix by interleaving 32- and then 64-bit words, which allows us to
+ // do XOR in NEON registers.
+ //
+
+ // x0..15[0-3] = s0..3[0..3]
+ add r3, r0, #0x20
+ vld1.32 {q0-q1}, [r0]
+ vld1.32 {q2-q3}, [r3]
+
+ adr r3, CTRINC
+ vdup.32 q15, d7[1]
+ vdup.32 q14, d7[0]
+ vld1.32 {q11}, [r3, :128]
+ vdup.32 q13, d6[1]
+ vdup.32 q12, d6[0]
+ vadd.i32 q12, q12, q11 // x12 += counter values 0-3
+ vdup.32 q11, d5[1]
+ vdup.32 q10, d5[0]
+ vdup.32 q9, d4[1]
+ vdup.32 q8, d4[0]
+ vdup.32 q7, d3[1]
+ vdup.32 q6, d3[0]
+ vdup.32 q5, d2[1]
+ vdup.32 q4, d2[0]
+ vdup.32 q3, d1[1]
+ vdup.32 q2, d1[0]
+ vdup.32 q1, d0[1]
+ vdup.32 q0, d0[0]
+
+ mov r3, #10
+
+.Ldoubleround4:
+ // x0 += x4, x12 = rotl32(x12 ^ x0, 16)
+ // x1 += x5, x13 = rotl32(x13 ^ x1, 16)
+ // x2 += x6, x14 = rotl32(x14 ^ x2, 16)
+ // x3 += x7, x15 = rotl32(x15 ^ x3, 16)
+ vadd.i32 q0, q0, q4
+ vadd.i32 q1, q1, q5
+ vadd.i32 q2, q2, q6
+ vadd.i32 q3, q3, q7
+
+ veor q12, q12, q0
+ veor q13, q13, q1
+ veor q14, q14, q2
+ veor q15, q15, q3
+
+ vrev32.16 q12, q12
+ vrev32.16 q13, q13
+ vrev32.16 q14, q14
+ vrev32.16 q15, q15
+
+ // x8 += x12, x4 = rotl32(x4 ^ x8, 12)
+ // x9 += x13, x5 = rotl32(x5 ^ x9, 12)
+ // x10 += x14, x6 = rotl32(x6 ^ x10, 12)
+ // x11 += x15, x7 = rotl32(x7 ^ x11, 12)
+ vadd.i32 q8, q8, q12
+ vadd.i32 q9, q9, q13
+ vadd.i32 q10, q10, q14
+ vadd.i32 q11, q11, q15
+
+ vst1.32 {q8-q9}, [sp, :256]
+
+ veor q8, q4, q8
+ veor q9, q5, q9
+ vshl.u32 q4, q8, #12
+ vshl.u32 q5, q9, #12
+ vsri.u32 q4, q8, #20
+ vsri.u32 q5, q9, #20
+
+ veor q8, q6, q10
+ veor q9, q7, q11
+ vshl.u32 q6, q8, #12
+ vshl.u32 q7, q9, #12
+ vsri.u32 q6, q8, #20
+ vsri.u32 q7, q9, #20
+
+ // x0 += x4, x12 = rotl32(x12 ^ x0, 8)
+ // x1 += x5, x13 = rotl32(x13 ^ x1, 8)
+ // x2 += x6, x14 = rotl32(x14 ^ x2, 8)
+ // x3 += x7, x15 = rotl32(x15 ^ x3, 8)
+ vadd.i32 q0, q0, q4
+ vadd.i32 q1, q1, q5
+ vadd.i32 q2, q2, q6
+ vadd.i32 q3, q3, q7
+
+ veor q8, q12, q0
+ veor q9, q13, q1
+ vshl.u32 q12, q8, #8
+ vshl.u32 q13, q9, #8
+ vsri.u32 q12, q8, #24
+ vsri.u32 q13, q9, #24
+
+ veor q8, q14, q2
+ veor q9, q15, q3
+ vshl.u32 q14, q8, #8
+ vshl.u32 q15, q9, #8
+ vsri.u32 q14, q8, #24
+ vsri.u32 q15, q9, #24
+
+ vld1.32 {q8-q9}, [sp, :256]
+
+ // x8 += x12, x4 = rotl32(x4 ^ x8, 7)
+ // x9 += x13, x5 = rotl32(x5 ^ x9, 7)
+ // x10 += x14, x6 = rotl32(x6 ^ x10, 7)
+ // x11 += x15, x7 = rotl32(x7 ^ x11, 7)
+ vadd.i32 q8, q8, q12
+ vadd.i32 q9, q9, q13
+ vadd.i32 q10, q10, q14
+ vadd.i32 q11, q11, q15
+
+ vst1.32 {q8-q9}, [sp, :256]
+
+ veor q8, q4, q8
+ veor q9, q5, q9
+ vshl.u32 q4, q8, #7
+ vshl.u32 q5, q9, #7
+ vsri.u32 q4, q8, #25
+ vsri.u32 q5, q9, #25
+
+ veor q8, q6, q10
+ veor q9, q7, q11
+ vshl.u32 q6, q8, #7
+ vshl.u32 q7, q9, #7
+ vsri.u32 q6, q8, #25
+ vsri.u32 q7, q9, #25
+
+ vld1.32 {q8-q9}, [sp, :256]
+
+ // x0 += x5, x15 = rotl32(x15 ^ x0, 16)
+ // x1 += x6, x12 = rotl32(x12 ^ x1, 16)
+ // x2 += x7, x13 = rotl32(x13 ^ x2, 16)
+ // x3 += x4, x14 = rotl32(x14 ^ x3, 16)
+ vadd.i32 q0, q0, q5
+ vadd.i32 q1, q1, q6
+ vadd.i32 q2, q2, q7
+ vadd.i32 q3, q3, q4
+
+ veor q15, q15, q0
+ veor q12, q12, q1
+ veor q13, q13, q2
+ veor q14, q14, q3
+
+ vrev32.16 q15, q15
+ vrev32.16 q12, q12
+ vrev32.16 q13, q13
+ vrev32.16 q14, q14
+
+ // x10 += x15, x5 = rotl32(x5 ^ x10, 12)
+ // x11 += x12, x6 = rotl32(x6 ^ x11, 12)
+ // x8 += x13, x7 = rotl32(x7 ^ x8, 12)
+ // x9 += x14, x4 = rotl32(x4 ^ x9, 12)
+ vadd.i32 q10, q10, q15
+ vadd.i32 q11, q11, q12
+ vadd.i32 q8, q8, q13
+ vadd.i32 q9, q9, q14
+
+ vst1.32 {q8-q9}, [sp, :256]
+
+ veor q8, q7, q8
+ veor q9, q4, q9
+ vshl.u32 q7, q8, #12
+ vshl.u32 q4, q9, #12
+ vsri.u32 q7, q8, #20
+ vsri.u32 q4, q9, #20
+
+ veor q8, q5, q10
+ veor q9, q6, q11
+ vshl.u32 q5, q8, #12
+ vshl.u32 q6, q9, #12
+ vsri.u32 q5, q8, #20
+ vsri.u32 q6, q9, #20
+
+ // x0 += x5, x15 = rotl32(x15 ^ x0, 8)
+ // x1 += x6, x12 = rotl32(x12 ^ x1, 8)
+ // x2 += x7, x13 = rotl32(x13 ^ x2, 8)
+ // x3 += x4, x14 = rotl32(x14 ^ x3, 8)
+ vadd.i32 q0, q0, q5
+ vadd.i32 q1, q1, q6
+ vadd.i32 q2, q2, q7
+ vadd.i32 q3, q3, q4
+
+ veor q8, q15, q0
+ veor q9, q12, q1
+ vshl.u32 q15, q8, #8
+ vshl.u32 q12, q9, #8
+ vsri.u32 q15, q8, #24
+ vsri.u32 q12, q9, #24
+
+ veor q8, q13, q2
+ veor q9, q14, q3
+ vshl.u32 q13, q8, #8
+ vshl.u32 q14, q9, #8
+ vsri.u32 q13, q8, #24
+ vsri.u32 q14, q9, #24
+
+ vld1.32 {q8-q9}, [sp, :256]
+
+ // x10 += x15, x5 = rotl32(x5 ^ x10, 7)
+ // x11 += x12, x6 = rotl32(x6 ^ x11, 7)
+ // x8 += x13, x7 = rotl32(x7 ^ x8, 7)
+ // x9 += x14, x4 = rotl32(x4 ^ x9, 7)
+ vadd.i32 q10, q10, q15
+ vadd.i32 q11, q11, q12
+ vadd.i32 q8, q8, q13
+ vadd.i32 q9, q9, q14
+
+ vst1.32 {q8-q9}, [sp, :256]
+
+ veor q8, q7, q8
+ veor q9, q4, q9
+ vshl.u32 q7, q8, #7
+ vshl.u32 q4, q9, #7
+ vsri.u32 q7, q8, #25
+ vsri.u32 q4, q9, #25
+
+ veor q8, q5, q10
+ veor q9, q6, q11
+ vshl.u32 q5, q8, #7
+ vshl.u32 q6, q9, #7
+ vsri.u32 q5, q8, #25
+ vsri.u32 q6, q9, #25
+
+ subs r3, r3, #1
+ beq 0f
+
+ vld1.32 {q8-q9}, [sp, :256]
+ b .Ldoubleround4
+
+ // x0[0-3] += s0[0]
+ // x1[0-3] += s0[1]
+ // x2[0-3] += s0[2]
+ // x3[0-3] += s0[3]
+0: ldmia r0!, {r3-r6}
+ vdup.32 q8, r3
+ vdup.32 q9, r4
+ vadd.i32 q0, q0, q8
+ vadd.i32 q1, q1, q9
+ vdup.32 q8, r5
+ vdup.32 q9, r6
+ vadd.i32 q2, q2, q8
+ vadd.i32 q3, q3, q9
+
+ // x4[0-3] += s1[0]
+ // x5[0-3] += s1[1]
+ // x6[0-3] += s1[2]
+ // x7[0-3] += s1[3]
+ ldmia r0!, {r3-r6}
+ vdup.32 q8, r3
+ vdup.32 q9, r4
+ vadd.i32 q4, q4, q8
+ vadd.i32 q5, q5, q9
+ vdup.32 q8, r5
+ vdup.32 q9, r6
+ vadd.i32 q6, q6, q8
+ vadd.i32 q7, q7, q9
+
+ // interleave 32-bit words in state n, n+1
+ vzip.32 q0, q1
+ vzip.32 q2, q3
+ vzip.32 q4, q5
+ vzip.32 q6, q7
+
+ // interleave 64-bit words in state n, n+2
+ vswp d1, d4
+ vswp d3, d6
+ vswp d9, d12
+ vswp d11, d14
+
+ // xor with corresponding input, write to output
+ vld1.8 {q8-q9}, [r2]!
+ veor q8, q8, q0
+ veor q9, q9, q4
+ vst1.8 {q8-q9}, [r1]!
+
+ vld1.32 {q8-q9}, [sp, :256]
+
+ // x8[0-3] += s2[0]
+ // x9[0-3] += s2[1]
+ // x10[0-3] += s2[2]
+ // x11[0-3] += s2[3]
+ ldmia r0!, {r3-r6}
+ vdup.32 q0, r3
+ vdup.32 q4, r4
+ vadd.i32 q8, q8, q0
+ vadd.i32 q9, q9, q4
+ vdup.32 q0, r5
+ vdup.32 q4, r6
+ vadd.i32 q10, q10, q0
+ vadd.i32 q11, q11, q4
+
+ // x12[0-3] += s3[0]
+ // x13[0-3] += s3[1]
+ // x14[0-3] += s3[2]
+ // x15[0-3] += s3[3]
+ ldmia r0!, {r3-r6}
+ vdup.32 q0, r3
+ vdup.32 q4, r4
+ adr r3, CTRINC
+ vadd.i32 q12, q12, q0
+ vld1.32 {q0}, [r3, :128]
+ vadd.i32 q13, q13, q4
+ vadd.i32 q12, q12, q0 // x12 += counter values 0-3
+
+ vdup.32 q0, r5
+ vdup.32 q4, r6
+ vadd.i32 q14, q14, q0
+ vadd.i32 q15, q15, q4
+
+ // interleave 32-bit words in state n, n+1
+ vzip.32 q8, q9
+ vzip.32 q10, q11
+ vzip.32 q12, q13
+ vzip.32 q14, q15
+
+ // interleave 64-bit words in state n, n+2
+ vswp d17, d20
+ vswp d19, d22
+ vswp d25, d28
+ vswp d27, d30
+
+ vmov q4, q1
+
+ vld1.8 {q0-q1}, [r2]!
+ veor q0, q0, q8
+ veor q1, q1, q12
+ vst1.8 {q0-q1}, [r1]!
+
+ vld1.8 {q0-q1}, [r2]!
+ veor q0, q0, q2
+ veor q1, q1, q6
+ vst1.8 {q0-q1}, [r1]!
+
+ vld1.8 {q0-q1}, [r2]!
+ veor q0, q0, q10
+ veor q1, q1, q14
+ vst1.8 {q0-q1}, [r1]!
+
+ vld1.8 {q0-q1}, [r2]!
+ veor q0, q0, q4
+ veor q1, q1, q5
+ vst1.8 {q0-q1}, [r1]!
+
+ vld1.8 {q0-q1}, [r2]!
+ veor q0, q0, q9
+ veor q1, q1, q13
+ vst1.8 {q0-q1}, [r1]!
+
+ vld1.8 {q0-q1}, [r2]!
+ veor q0, q0, q3
+ veor q1, q1, q7
+ vst1.8 {q0-q1}, [r1]!
+
+ vld1.8 {q0-q1}, [r2]
+ veor q0, q0, q11
+ veor q1, q1, q15
+ vst1.8 {q0-q1}, [r1]
+
+ mov sp, ip
+ pop {r4-r6, pc}
+ENDPROC(chacha20_4block_xor_neon)
+
+ .align 4
+CTRINC: .word 0, 1, 2, 3
+
diff --git a/arch/arm/crypto/chacha20-neon-glue.c b/arch/arm/crypto/chacha20-neon-glue.c
new file mode 100644
index 000000000000..592f75ae4fa1
--- /dev/null
+++ b/arch/arm/crypto/chacha20-neon-glue.c
@@ -0,0 +1,128 @@
+/*
+ * ChaCha20 256-bit cipher algorithm, RFC7539, ARM NEON functions
+ *
+ * Copyright (C) 2016 Linaro, Ltd. <ard.biesheuvel@linaro.org>
+ *
+ * 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 by the Free Software Foundation.
+ *
+ * Based on:
+ * ChaCha20 256-bit cipher algorithm, RFC7539, SIMD glue code
+ *
+ * Copyright (C) 2015 Martin Willi
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <crypto/algapi.h>
+#include <crypto/chacha20.h>
+#include <crypto/internal/skcipher.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+
+#include <asm/hwcap.h>
+#include <asm/neon.h>
+#include <asm/simd.h>
+
+asmlinkage void chacha20_block_xor_neon(u32 *state, u8 *dst, const u8 *src);
+asmlinkage void chacha20_4block_xor_neon(u32 *state, u8 *dst, const u8 *src);
+
+static void chacha20_doneon(u32 *state, u8 *dst, const u8 *src,
+ unsigned int bytes)
+{
+ u8 buf[CHACHA20_BLOCK_SIZE];
+
+ while (bytes >= CHACHA20_BLOCK_SIZE * 4) {
+ chacha20_4block_xor_neon(state, dst, src);
+ bytes -= CHACHA20_BLOCK_SIZE * 4;
+ src += CHACHA20_BLOCK_SIZE * 4;
+ dst += CHACHA20_BLOCK_SIZE * 4;
+ state[12] += 4;
+ }
+ while (bytes >= CHACHA20_BLOCK_SIZE) {
+ chacha20_block_xor_neon(state, dst, src);
+ bytes -= CHACHA20_BLOCK_SIZE;
+ src += CHACHA20_BLOCK_SIZE;
+ dst += CHACHA20_BLOCK_SIZE;
+ state[12]++;
+ }
+ if (bytes) {
+ memcpy(buf, src, bytes);
+ chacha20_block_xor_neon(state, buf, buf);
+ memcpy(dst, buf, bytes);
+ }
+}
+
+static int chacha20_neon(struct skcipher_request *req)
+{
+ struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+ struct chacha20_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct skcipher_walk walk;
+ u32 state[16];
+ int err;
+
+ if (req->cryptlen <= CHACHA20_BLOCK_SIZE || !may_use_simd())
+ return crypto_chacha20_crypt(req);
+
+ err = skcipher_walk_virt(&walk, req, true);
+
+ crypto_chacha20_init(state, ctx, walk.iv);
+
+ kernel_neon_begin();
+ while (walk.nbytes > 0) {
+ unsigned int nbytes = walk.nbytes;
+
+ if (nbytes < walk.total)
+ nbytes = round_down(nbytes, walk.stride);
+
+ chacha20_doneon(state, walk.dst.virt.addr, walk.src.virt.addr,
+ nbytes);
+ err = skcipher_walk_done(&walk, walk.nbytes - nbytes);
+ }
+ kernel_neon_end();
+
+ return err;
+}
+
+static struct skcipher_alg alg = {
+ .base.cra_name = "chacha20",
+ .base.cra_driver_name = "chacha20-neon",
+ .base.cra_priority = 300,
+ .base.cra_blocksize = 1,
+ .base.cra_ctxsize = sizeof(struct chacha20_ctx),
+ .base.cra_alignmask = 1,
+ .base.cra_module = THIS_MODULE,
+
+ .min_keysize = CHACHA20_KEY_SIZE,
+ .max_keysize = CHACHA20_KEY_SIZE,
+ .ivsize = CHACHA20_IV_SIZE,
+ .chunksize = CHACHA20_BLOCK_SIZE,
+ .walksize = 4 * CHACHA20_BLOCK_SIZE,
+ .setkey = crypto_chacha20_setkey,
+ .encrypt = chacha20_neon,
+ .decrypt = chacha20_neon,
+};
+
+static int __init chacha20_simd_mod_init(void)
+{
+ if (!(elf_hwcap & HWCAP_NEON))
+ return -ENODEV;
+
+ return crypto_register_skcipher(&alg);
+}
+
+static void __exit chacha20_simd_mod_fini(void)
+{
+ crypto_unregister_skcipher(&alg);
+}
+
+module_init(chacha20_simd_mod_init);
+module_exit(chacha20_simd_mod_fini);
+
+MODULE_AUTHOR("Ard Biesheuvel <ard.biesheuvel@linaro.org>");
+MODULE_LICENSE("GPL v2");
+MODULE_ALIAS_CRYPTO("chacha20");
--
2.7.4
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH v2 7/7] crypto: arm64/aes - reimplement bit-sliced ARM/NEON implementation for arm64
2017-01-11 16:41 [PATCH v2 0/7] crypto: ARM/arm64 - AES and ChaCha20 updates for v4.11 Ard Biesheuvel
` (4 preceding siblings ...)
2017-01-11 16:41 ` [PATCH v2 5/7] crypto: arm/aes - replace scalar AES cipher Ard Biesheuvel
@ 2017-01-11 16:41 ` Ard Biesheuvel
2017-01-12 16:45 ` [PATCH v2 0/7] crypto: ARM/arm64 - AES and ChaCha20 updates for v4.11 Herbert Xu
6 siblings, 0 replies; 10+ messages in thread
From: Ard Biesheuvel @ 2017-01-11 16:41 UTC (permalink / raw)
To: linux-crypto; +Cc: herbert, linux-arm-kernel, Ard Biesheuvel
This is a reimplementation of the NEON version of the bit-sliced AES
algorithm. This code is heavily based on Andy Polyakov's OpenSSL version
for ARM, which is also available in the kernel. This is an alternative for
the existing NEON implementation for arm64 authored by me, which suffers
from poor performance due to its reliance on the pathologically slow four
register variant of the tbl/tbx NEON instruction.
This version is about ~30% (*) faster than the generic C code, but only in
cases where the input can be 8x interleaved (this is a fundamental property
of bit slicing). For this reason, only the chaining modes ECB, XTS and CTR
are implemented. (The significance of ECB is that it could potentially be
used by other chaining modes)
* Measured on Cortex-A57. Note that this is still an order of magnitude
slower than the implementations that use the dedicated AES instructions
introduced in ARMv8, but those are part of an optional extension, and so
it is good to have a fallback.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
arch/arm64/crypto/Kconfig | 7 +
arch/arm64/crypto/Makefile | 3 +
arch/arm64/crypto/aes-neonbs-core.S | 963 ++++++++++++++++++++
arch/arm64/crypto/aes-neonbs-glue.c | 420 +++++++++
4 files changed, 1393 insertions(+)
diff --git a/arch/arm64/crypto/Kconfig b/arch/arm64/crypto/Kconfig
index 0826f8e599a6..5de75c3dcbd4 100644
--- a/arch/arm64/crypto/Kconfig
+++ b/arch/arm64/crypto/Kconfig
@@ -82,4 +82,11 @@ config CRYPTO_CHACHA20_NEON
select CRYPTO_BLKCIPHER
select CRYPTO_CHACHA20
+config CRYPTO_AES_ARM64_BS
+ tristate "AES in ECB/CBC/CTR/XTS modes using bit-sliced NEON algorithm"
+ depends on KERNEL_MODE_NEON
+ select CRYPTO_BLKCIPHER
+ select CRYPTO_AES_ARM64
+ select CRYPTO_SIMD
+
endif
diff --git a/arch/arm64/crypto/Makefile b/arch/arm64/crypto/Makefile
index a893507629eb..d1ae1b9cbe70 100644
--- a/arch/arm64/crypto/Makefile
+++ b/arch/arm64/crypto/Makefile
@@ -47,6 +47,9 @@ chacha20-neon-y := chacha20-neon-core.o chacha20-neon-glue.o
obj-$(CONFIG_CRYPTO_AES_ARM64) += aes-arm64.o
aes-arm64-y := aes-cipher-core.o aes-cipher-glue.o
+obj-$(CONFIG_CRYPTO_AES_ARM64_BS) += aes-neon-bs.o
+aes-neon-bs-y := aes-neonbs-core.o aes-neonbs-glue.o
+
AFLAGS_aes-ce.o := -DINTERLEAVE=4
AFLAGS_aes-neon.o := -DINTERLEAVE=4
diff --git a/arch/arm64/crypto/aes-neonbs-core.S b/arch/arm64/crypto/aes-neonbs-core.S
new file mode 100644
index 000000000000..8d0cdaa2768d
--- /dev/null
+++ b/arch/arm64/crypto/aes-neonbs-core.S
@@ -0,0 +1,963 @@
+/*
+ * Bit sliced AES using NEON instructions
+ *
+ * Copyright (C) 2016 Linaro Ltd <ard.biesheuvel@linaro.org>
+ *
+ * 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 by the Free Software Foundation.
+ */
+
+/*
+ * The algorithm implemented here is described in detail by the paper
+ * 'Faster and Timing-Attack Resistant AES-GCM' by Emilia Kaesper and
+ * Peter Schwabe (https://eprint.iacr.org/2009/129.pdf)
+ *
+ * This implementation is based primarily on the OpenSSL implementation
+ * for 32-bit ARM written by Andy Polyakov <appro@openssl.org>
+ */
+
+#include <linux/linkage.h>
+#include <asm/assembler.h>
+
+ .text
+
+ rounds .req x11
+ bskey .req x12
+
+ .macro in_bs_ch, b0, b1, b2, b3, b4, b5, b6, b7
+ eor \b2, \b2, \b1
+ eor \b5, \b5, \b6
+ eor \b3, \b3, \b0
+ eor \b6, \b6, \b2
+ eor \b5, \b5, \b0
+ eor \b6, \b6, \b3
+ eor \b3, \b3, \b7
+ eor \b7, \b7, \b5
+ eor \b3, \b3, \b4
+ eor \b4, \b4, \b5
+ eor \b2, \b2, \b7
+ eor \b3, \b3, \b1
+ eor \b1, \b1, \b5
+ .endm
+
+ .macro out_bs_ch, b0, b1, b2, b3, b4, b5, b6, b7
+ eor \b0, \b0, \b6
+ eor \b1, \b1, \b4
+ eor \b4, \b4, \b6
+ eor \b2, \b2, \b0
+ eor \b6, \b6, \b1
+ eor \b1, \b1, \b5
+ eor \b5, \b5, \b3
+ eor \b3, \b3, \b7
+ eor \b7, \b7, \b5
+ eor \b2, \b2, \b5
+ eor \b4, \b4, \b7
+ .endm
+
+ .macro inv_in_bs_ch, b6, b1, b2, b4, b7, b0, b3, b5
+ eor \b1, \b1, \b7
+ eor \b4, \b4, \b7
+ eor \b7, \b7, \b5
+ eor \b1, \b1, \b3
+ eor \b2, \b2, \b5
+ eor \b3, \b3, \b7
+ eor \b6, \b6, \b1
+ eor \b2, \b2, \b0
+ eor \b5, \b5, \b3
+ eor \b4, \b4, \b6
+ eor \b0, \b0, \b6
+ eor \b1, \b1, \b4
+ .endm
+
+ .macro inv_out_bs_ch, b6, b5, b0, b3, b7, b1, b4, b2
+ eor \b1, \b1, \b5
+ eor \b2, \b2, \b7
+ eor \b3, \b3, \b1
+ eor \b4, \b4, \b5
+ eor \b7, \b7, \b5
+ eor \b3, \b3, \b4
+ eor \b5, \b5, \b0
+ eor \b3, \b3, \b7
+ eor \b6, \b6, \b2
+ eor \b2, \b2, \b1
+ eor \b6, \b6, \b3
+ eor \b3, \b3, \b0
+ eor \b5, \b5, \b6
+ .endm
+
+ .macro mul_gf4, x0, x1, y0, y1, t0, t1
+ eor \t0, \y0, \y1
+ and \t0, \t0, \x0
+ eor \x0, \x0, \x1
+ and \t1, \x1, \y0
+ and \x0, \x0, \y1
+ eor \x1, \t1, \t0
+ eor \x0, \x0, \t1
+ .endm
+
+ .macro mul_gf4_n_gf4, x0, x1, y0, y1, t0, x2, x3, y2, y3, t1
+ eor \t0, \y0, \y1
+ eor \t1, \y2, \y3
+ and \t0, \t0, \x0
+ and \t1, \t1, \x2
+ eor \x0, \x0, \x1
+ eor \x2, \x2, \x3
+ and \x1, \x1, \y0
+ and \x3, \x3, \y2
+ and \x0, \x0, \y1
+ and \x2, \x2, \y3
+ eor \x1, \x1, \x0
+ eor \x2, \x2, \x3
+ eor \x0, \x0, \t0
+ eor \x3, \x3, \t1
+ .endm
+
+ .macro mul_gf16_2, x0, x1, x2, x3, x4, x5, x6, x7, \
+ y0, y1, y2, y3, t0, t1, t2, t3
+ eor \t0, \x0, \x2
+ eor \t1, \x1, \x3
+ mul_gf4 \x0, \x1, \y0, \y1, \t2, \t3
+ eor \y0, \y0, \y2
+ eor \y1, \y1, \y3
+ mul_gf4_n_gf4 \t0, \t1, \y0, \y1, \t3, \x2, \x3, \y2, \y3, \t2
+ eor \x0, \x0, \t0
+ eor \x2, \x2, \t0
+ eor \x1, \x1, \t1
+ eor \x3, \x3, \t1
+ eor \t0, \x4, \x6
+ eor \t1, \x5, \x7
+ mul_gf4_n_gf4 \t0, \t1, \y0, \y1, \t3, \x6, \x7, \y2, \y3, \t2
+ eor \y0, \y0, \y2
+ eor \y1, \y1, \y3
+ mul_gf4 \x4, \x5, \y0, \y1, \t2, \t3
+ eor \x4, \x4, \t0
+ eor \x6, \x6, \t0
+ eor \x5, \x5, \t1
+ eor \x7, \x7, \t1
+ .endm
+
+ .macro inv_gf256, x0, x1, x2, x3, x4, x5, x6, x7, \
+ t0, t1, t2, t3, s0, s1, s2, s3
+ eor \t3, \x4, \x6
+ eor \t0, \x5, \x7
+ eor \t1, \x1, \x3
+ eor \s1, \x7, \x6
+ eor \s0, \x0, \x2
+ eor \s3, \t3, \t0
+ orr \t2, \t0, \t1
+ and \s2, \t3, \s0
+ orr \t3, \t3, \s0
+ eor \s0, \s0, \t1
+ and \t0, \t0, \t1
+ eor \t1, \x3, \x2
+ and \s3, \s3, \s0
+ and \s1, \s1, \t1
+ eor \t1, \x4, \x5
+ eor \s0, \x1, \x0
+ eor \t3, \t3, \s1
+ eor \t2, \t2, \s1
+ and \s1, \t1, \s0
+ orr \t1, \t1, \s0
+ eor \t3, \t3, \s3
+ eor \t0, \t0, \s1
+ eor \t2, \t2, \s2
+ eor \t1, \t1, \s3
+ eor \t0, \t0, \s2
+ and \s0, \x7, \x3
+ eor \t1, \t1, \s2
+ and \s1, \x6, \x2
+ and \s2, \x5, \x1
+ orr \s3, \x4, \x0
+ eor \t3, \t3, \s0
+ eor \t1, \t1, \s2
+ eor \s0, \t0, \s3
+ eor \t2, \t2, \s1
+ and \s2, \t3, \t1
+ eor \s1, \t2, \s2
+ eor \s3, \s0, \s2
+ bsl \s1, \t1, \s0
+ not \t0, \s0
+ bsl \s0, \s1, \s3
+ bsl \t0, \s1, \s3
+ bsl \s3, \t3, \t2
+ eor \t3, \t3, \t2
+ and \s2, \s0, \s3
+ eor \t1, \t1, \t0
+ eor \s2, \s2, \t3
+ mul_gf16_2 \x0, \x1, \x2, \x3, \x4, \x5, \x6, \x7, \
+ \s3, \s2, \s1, \t1, \s0, \t0, \t2, \t3
+ .endm
+
+ .macro sbox, b0, b1, b2, b3, b4, b5, b6, b7, \
+ t0, t1, t2, t3, s0, s1, s2, s3
+ in_bs_ch \b0\().16b, \b1\().16b, \b2\().16b, \b3\().16b, \
+ \b4\().16b, \b5\().16b, \b6\().16b, \b7\().16b
+ inv_gf256 \b6\().16b, \b5\().16b, \b0\().16b, \b3\().16b, \
+ \b7\().16b, \b1\().16b, \b4\().16b, \b2\().16b, \
+ \t0\().16b, \t1\().16b, \t2\().16b, \t3\().16b, \
+ \s0\().16b, \s1\().16b, \s2\().16b, \s3\().16b
+ out_bs_ch \b7\().16b, \b1\().16b, \b4\().16b, \b2\().16b, \
+ \b6\().16b, \b5\().16b, \b0\().16b, \b3\().16b
+ .endm
+
+ .macro inv_sbox, b0, b1, b2, b3, b4, b5, b6, b7, \
+ t0, t1, t2, t3, s0, s1, s2, s3
+ inv_in_bs_ch \b0\().16b, \b1\().16b, \b2\().16b, \b3\().16b, \
+ \b4\().16b, \b5\().16b, \b6\().16b, \b7\().16b
+ inv_gf256 \b5\().16b, \b1\().16b, \b2\().16b, \b6\().16b, \
+ \b3\().16b, \b7\().16b, \b0\().16b, \b4\().16b, \
+ \t0\().16b, \t1\().16b, \t2\().16b, \t3\().16b, \
+ \s0\().16b, \s1\().16b, \s2\().16b, \s3\().16b
+ inv_out_bs_ch \b3\().16b, \b7\().16b, \b0\().16b, \b4\().16b, \
+ \b5\().16b, \b1\().16b, \b2\().16b, \b6\().16b
+ .endm
+
+ .macro enc_next_rk
+ ldp q16, q17, [bskey], #128
+ ldp q18, q19, [bskey, #-96]
+ ldp q20, q21, [bskey, #-64]
+ ldp q22, q23, [bskey, #-32]
+ .endm
+
+ .macro dec_next_rk
+ ldp q16, q17, [bskey, #-128]!
+ ldp q18, q19, [bskey, #32]
+ ldp q20, q21, [bskey, #64]
+ ldp q22, q23, [bskey, #96]
+ .endm
+
+ .macro add_round_key, x0, x1, x2, x3, x4, x5, x6, x7
+ eor \x0\().16b, \x0\().16b, v16.16b
+ eor \x1\().16b, \x1\().16b, v17.16b
+ eor \x2\().16b, \x2\().16b, v18.16b
+ eor \x3\().16b, \x3\().16b, v19.16b
+ eor \x4\().16b, \x4\().16b, v20.16b
+ eor \x5\().16b, \x5\().16b, v21.16b
+ eor \x6\().16b, \x6\().16b, v22.16b
+ eor \x7\().16b, \x7\().16b, v23.16b
+ .endm
+
+ .macro shift_rows, x0, x1, x2, x3, x4, x5, x6, x7, mask
+ tbl \x0\().16b, {\x0\().16b}, \mask\().16b
+ tbl \x1\().16b, {\x1\().16b}, \mask\().16b
+ tbl \x2\().16b, {\x2\().16b}, \mask\().16b
+ tbl \x3\().16b, {\x3\().16b}, \mask\().16b
+ tbl \x4\().16b, {\x4\().16b}, \mask\().16b
+ tbl \x5\().16b, {\x5\().16b}, \mask\().16b
+ tbl \x6\().16b, {\x6\().16b}, \mask\().16b
+ tbl \x7\().16b, {\x7\().16b}, \mask\().16b
+ .endm
+
+ .macro mix_cols, x0, x1, x2, x3, x4, x5, x6, x7, \
+ t0, t1, t2, t3, t4, t5, t6, t7, inv
+ ext \t0\().16b, \x0\().16b, \x0\().16b, #12
+ ext \t1\().16b, \x1\().16b, \x1\().16b, #12
+ eor \x0\().16b, \x0\().16b, \t0\().16b
+ ext \t2\().16b, \x2\().16b, \x2\().16b, #12
+ eor \x1\().16b, \x1\().16b, \t1\().16b
+ ext \t3\().16b, \x3\().16b, \x3\().16b, #12
+ eor \x2\().16b, \x2\().16b, \t2\().16b
+ ext \t4\().16b, \x4\().16b, \x4\().16b, #12
+ eor \x3\().16b, \x3\().16b, \t3\().16b
+ ext \t5\().16b, \x5\().16b, \x5\().16b, #12
+ eor \x4\().16b, \x4\().16b, \t4\().16b
+ ext \t6\().16b, \x6\().16b, \x6\().16b, #12
+ eor \x5\().16b, \x5\().16b, \t5\().16b
+ ext \t7\().16b, \x7\().16b, \x7\().16b, #12
+ eor \x6\().16b, \x6\().16b, \t6\().16b
+ eor \t1\().16b, \t1\().16b, \x0\().16b
+ eor \x7\().16b, \x7\().16b, \t7\().16b
+ ext \x0\().16b, \x0\().16b, \x0\().16b, #8
+ eor \t2\().16b, \t2\().16b, \x1\().16b
+ eor \t0\().16b, \t0\().16b, \x7\().16b
+ eor \t1\().16b, \t1\().16b, \x7\().16b
+ ext \x1\().16b, \x1\().16b, \x1\().16b, #8
+ eor \t5\().16b, \t5\().16b, \x4\().16b
+ eor \x0\().16b, \x0\().16b, \t0\().16b
+ eor \t6\().16b, \t6\().16b, \x5\().16b
+ eor \x1\().16b, \x1\().16b, \t1\().16b
+ ext \t0\().16b, \x4\().16b, \x4\().16b, #8
+ eor \t4\().16b, \t4\().16b, \x3\().16b
+ ext \t1\().16b, \x5\().16b, \x5\().16b, #8
+ eor \t7\().16b, \t7\().16b, \x6\().16b
+ ext \x4\().16b, \x3\().16b, \x3\().16b, #8
+ eor \t3\().16b, \t3\().16b, \x2\().16b
+ ext \x5\().16b, \x7\().16b, \x7\().16b, #8
+ eor \t4\().16b, \t4\().16b, \x7\().16b
+ ext \x3\().16b, \x6\().16b, \x6\().16b, #8
+ eor \t3\().16b, \t3\().16b, \x7\().16b
+ ext \x6\().16b, \x2\().16b, \x2\().16b, #8
+ eor \x7\().16b, \t1\().16b, \t5\().16b
+ .ifb \inv
+ eor \x2\().16b, \t0\().16b, \t4\().16b
+ eor \x4\().16b, \x4\().16b, \t3\().16b
+ eor \x5\().16b, \x5\().16b, \t7\().16b
+ eor \x3\().16b, \x3\().16b, \t6\().16b
+ eor \x6\().16b, \x6\().16b, \t2\().16b
+ .else
+ eor \t3\().16b, \t3\().16b, \x4\().16b
+ eor \x5\().16b, \x5\().16b, \t7\().16b
+ eor \x2\().16b, \x3\().16b, \t6\().16b
+ eor \x3\().16b, \t0\().16b, \t4\().16b
+ eor \x4\().16b, \x6\().16b, \t2\().16b
+ mov \x6\().16b, \t3\().16b
+ .endif
+ .endm
+
+ .macro inv_mix_cols, x0, x1, x2, x3, x4, x5, x6, x7, \
+ t0, t1, t2, t3, t4, t5, t6, t7
+ ext \t0\().16b, \x0\().16b, \x0\().16b, #8
+ ext \t6\().16b, \x6\().16b, \x6\().16b, #8
+ ext \t7\().16b, \x7\().16b, \x7\().16b, #8
+ eor \t0\().16b, \t0\().16b, \x0\().16b
+ ext \t1\().16b, \x1\().16b, \x1\().16b, #8
+ eor \t6\().16b, \t6\().16b, \x6\().16b
+ ext \t2\().16b, \x2\().16b, \x2\().16b, #8
+ eor \t7\().16b, \t7\().16b, \x7\().16b
+ ext \t3\().16b, \x3\().16b, \x3\().16b, #8
+ eor \t1\().16b, \t1\().16b, \x1\().16b
+ ext \t4\().16b, \x4\().16b, \x4\().16b, #8
+ eor \t2\().16b, \t2\().16b, \x2\().16b
+ ext \t5\().16b, \x5\().16b, \x5\().16b, #8
+ eor \t3\().16b, \t3\().16b, \x3\().16b
+ eor \t4\().16b, \t4\().16b, \x4\().16b
+ eor \t5\().16b, \t5\().16b, \x5\().16b
+ eor \x0\().16b, \x0\().16b, \t6\().16b
+ eor \x1\().16b, \x1\().16b, \t6\().16b
+ eor \x2\().16b, \x2\().16b, \t0\().16b
+ eor \x4\().16b, \x4\().16b, \t2\().16b
+ eor \x3\().16b, \x3\().16b, \t1\().16b
+ eor \x1\().16b, \x1\().16b, \t7\().16b
+ eor \x2\().16b, \x2\().16b, \t7\().16b
+ eor \x4\().16b, \x4\().16b, \t6\().16b
+ eor \x5\().16b, \x5\().16b, \t3\().16b
+ eor \x3\().16b, \x3\().16b, \t6\().16b
+ eor \x6\().16b, \x6\().16b, \t4\().16b
+ eor \x4\().16b, \x4\().16b, \t7\().16b
+ eor \x5\().16b, \x5\().16b, \t7\().16b
+ eor \x7\().16b, \x7\().16b, \t5\().16b
+ mix_cols \x0, \x1, \x2, \x3, \x4, \x5, \x6, \x7, \
+ \t0, \t1, \t2, \t3, \t4, \t5, \t6, \t7, 1
+ .endm
+
+ .macro swapmove_2x, a0, b0, a1, b1, n, mask, t0, t1
+ ushr \t0\().2d, \b0\().2d, #\n
+ ushr \t1\().2d, \b1\().2d, #\n
+ eor \t0\().16b, \t0\().16b, \a0\().16b
+ eor \t1\().16b, \t1\().16b, \a1\().16b
+ and \t0\().16b, \t0\().16b, \mask\().16b
+ and \t1\().16b, \t1\().16b, \mask\().16b
+ eor \a0\().16b, \a0\().16b, \t0\().16b
+ shl \t0\().2d, \t0\().2d, #\n
+ eor \a1\().16b, \a1\().16b, \t1\().16b
+ shl \t1\().2d, \t1\().2d, #\n
+ eor \b0\().16b, \b0\().16b, \t0\().16b
+ eor \b1\().16b, \b1\().16b, \t1\().16b
+ .endm
+
+ .macro bitslice, x7, x6, x5, x4, x3, x2, x1, x0, t0, t1, t2, t3
+ movi \t0\().16b, #0x55
+ movi \t1\().16b, #0x33
+ swapmove_2x \x0, \x1, \x2, \x3, 1, \t0, \t2, \t3
+ swapmove_2x \x4, \x5, \x6, \x7, 1, \t0, \t2, \t3
+ movi \t0\().16b, #0x0f
+ swapmove_2x \x0, \x2, \x1, \x3, 2, \t1, \t2, \t3
+ swapmove_2x \x4, \x6, \x5, \x7, 2, \t1, \t2, \t3
+ swapmove_2x \x0, \x4, \x1, \x5, 4, \t0, \t2, \t3
+ swapmove_2x \x2, \x6, \x3, \x7, 4, \t0, \t2, \t3
+ .endm
+
+
+ .align 6
+M0: .octa 0x0004080c0105090d02060a0e03070b0f
+
+M0SR: .octa 0x0004080c05090d010a0e02060f03070b
+SR: .octa 0x0f0e0d0c0a09080b0504070600030201
+SRM0: .octa 0x01060b0c0207080d0304090e00050a0f
+
+M0ISR: .octa 0x0004080c0d0105090a0e0206070b0f03
+ISR: .octa 0x0f0e0d0c080b0a090504070602010003
+ISRM0: .octa 0x0306090c00070a0d01040b0e0205080f
+
+ /*
+ * void aesbs_convert_key(u8 out[], u32 const rk[], int rounds)
+ */
+ENTRY(aesbs_convert_key)
+ ld1 {v7.4s}, [x1], #16 // load round 0 key
+ ld1 {v17.4s}, [x1], #16 // load round 1 key
+
+ movi v8.16b, #0x01 // bit masks
+ movi v9.16b, #0x02
+ movi v10.16b, #0x04
+ movi v11.16b, #0x08
+ movi v12.16b, #0x10
+ movi v13.16b, #0x20
+ movi v14.16b, #0x40
+ movi v15.16b, #0x80
+ ldr q16, M0
+
+ sub x2, x2, #1
+ str q7, [x0], #16 // save round 0 key
+
+.Lkey_loop:
+ tbl v7.16b ,{v17.16b}, v16.16b
+ ld1 {v17.4s}, [x1], #16 // load next round key
+
+ cmtst v0.16b, v7.16b, v8.16b
+ cmtst v1.16b, v7.16b, v9.16b
+ cmtst v2.16b, v7.16b, v10.16b
+ cmtst v3.16b, v7.16b, v11.16b
+ cmtst v4.16b, v7.16b, v12.16b
+ cmtst v5.16b, v7.16b, v13.16b
+ cmtst v6.16b, v7.16b, v14.16b
+ cmtst v7.16b, v7.16b, v15.16b
+ not v0.16b, v0.16b
+ not v1.16b, v1.16b
+ not v5.16b, v5.16b
+ not v6.16b, v6.16b
+
+ subs x2, x2, #1
+ stp q0, q1, [x0], #128
+ stp q2, q3, [x0, #-96]
+ stp q4, q5, [x0, #-64]
+ stp q6, q7, [x0, #-32]
+ b.ne .Lkey_loop
+
+ movi v7.16b, #0x63 // compose .L63
+ eor v17.16b, v17.16b, v7.16b
+ str q17, [x0]
+ ret
+ENDPROC(aesbs_convert_key)
+
+ .align 4
+aesbs_encrypt8:
+ ldr q9, [bskey], #16 // round 0 key
+ ldr q8, M0SR
+ ldr q24, SR
+
+ eor v10.16b, v0.16b, v9.16b // xor with round0 key
+ eor v11.16b, v1.16b, v9.16b
+ tbl v0.16b, {v10.16b}, v8.16b
+ eor v12.16b, v2.16b, v9.16b
+ tbl v1.16b, {v11.16b}, v8.16b
+ eor v13.16b, v3.16b, v9.16b
+ tbl v2.16b, {v12.16b}, v8.16b
+ eor v14.16b, v4.16b, v9.16b
+ tbl v3.16b, {v13.16b}, v8.16b
+ eor v15.16b, v5.16b, v9.16b
+ tbl v4.16b, {v14.16b}, v8.16b
+ eor v10.16b, v6.16b, v9.16b
+ tbl v5.16b, {v15.16b}, v8.16b
+ eor v11.16b, v7.16b, v9.16b
+ tbl v6.16b, {v10.16b}, v8.16b
+ tbl v7.16b, {v11.16b}, v8.16b
+
+ bitslice v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11
+
+ sub rounds, rounds, #1
+ b .Lenc_sbox
+
+.Lenc_loop:
+ shift_rows v0, v1, v2, v3, v4, v5, v6, v7, v24
+.Lenc_sbox:
+ sbox v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, \
+ v13, v14, v15
+ subs rounds, rounds, #1
+ b.cc .Lenc_done
+
+ enc_next_rk
+
+ mix_cols v0, v1, v4, v6, v3, v7, v2, v5, v8, v9, v10, v11, v12, \
+ v13, v14, v15
+
+ add_round_key v0, v1, v2, v3, v4, v5, v6, v7
+
+ b.ne .Lenc_loop
+ ldr q24, SRM0
+ b .Lenc_loop
+
+.Lenc_done:
+ ldr q12, [bskey] // last round key
+
+ bitslice v0, v1, v4, v6, v3, v7, v2, v5, v8, v9, v10, v11
+
+ eor v0.16b, v0.16b, v12.16b
+ eor v1.16b, v1.16b, v12.16b
+ eor v4.16b, v4.16b, v12.16b
+ eor v6.16b, v6.16b, v12.16b
+ eor v3.16b, v3.16b, v12.16b
+ eor v7.16b, v7.16b, v12.16b
+ eor v2.16b, v2.16b, v12.16b
+ eor v5.16b, v5.16b, v12.16b
+ ret
+ENDPROC(aesbs_encrypt8)
+
+ .align 4
+aesbs_decrypt8:
+ lsl x9, rounds, #7
+ add bskey, bskey, x9
+
+ ldr q9, [bskey, #-112]! // round 0 key
+ ldr q8, M0ISR
+ ldr q24, ISR
+
+ eor v10.16b, v0.16b, v9.16b // xor with round0 key
+ eor v11.16b, v1.16b, v9.16b
+ tbl v0.16b, {v10.16b}, v8.16b
+ eor v12.16b, v2.16b, v9.16b
+ tbl v1.16b, {v11.16b}, v8.16b
+ eor v13.16b, v3.16b, v9.16b
+ tbl v2.16b, {v12.16b}, v8.16b
+ eor v14.16b, v4.16b, v9.16b
+ tbl v3.16b, {v13.16b}, v8.16b
+ eor v15.16b, v5.16b, v9.16b
+ tbl v4.16b, {v14.16b}, v8.16b
+ eor v10.16b, v6.16b, v9.16b
+ tbl v5.16b, {v15.16b}, v8.16b
+ eor v11.16b, v7.16b, v9.16b
+ tbl v6.16b, {v10.16b}, v8.16b
+ tbl v7.16b, {v11.16b}, v8.16b
+
+ bitslice v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11
+
+ sub rounds, rounds, #1
+ b .Ldec_sbox
+
+.Ldec_loop:
+ shift_rows v0, v1, v2, v3, v4, v5, v6, v7, v24
+.Ldec_sbox:
+ inv_sbox v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, \
+ v13, v14, v15
+ subs rounds, rounds, #1
+ b.cc .Ldec_done
+
+ dec_next_rk
+
+ add_round_key v0, v1, v6, v4, v2, v7, v3, v5
+
+ inv_mix_cols v0, v1, v6, v4, v2, v7, v3, v5, v8, v9, v10, v11, v12, \
+ v13, v14, v15
+
+ b.ne .Ldec_loop
+ ldr q24, ISRM0
+ b .Ldec_loop
+.Ldec_done:
+ ldr q12, [bskey, #-16] // last round key
+
+ bitslice v0, v1, v6, v4, v2, v7, v3, v5, v8, v9, v10, v11
+
+ eor v0.16b, v0.16b, v12.16b
+ eor v1.16b, v1.16b, v12.16b
+ eor v6.16b, v6.16b, v12.16b
+ eor v4.16b, v4.16b, v12.16b
+ eor v2.16b, v2.16b, v12.16b
+ eor v7.16b, v7.16b, v12.16b
+ eor v3.16b, v3.16b, v12.16b
+ eor v5.16b, v5.16b, v12.16b
+ ret
+ENDPROC(aesbs_decrypt8)
+
+ /*
+ * aesbs_ecb_encrypt(u8 out[], u8 const in[], u8 const rk[], int rounds,
+ * int blocks)
+ * aesbs_ecb_decrypt(u8 out[], u8 const in[], u8 const rk[], int rounds,
+ * int blocks)
+ */
+ .macro __ecb_crypt, do8, o0, o1, o2, o3, o4, o5, o6, o7
+ stp x29, x30, [sp, #-16]!
+ mov x29, sp
+
+99: mov x5, #1
+ lsl x5, x5, x4
+ subs w4, w4, #8
+ csel x4, x4, xzr, pl
+ csel x5, x5, xzr, mi
+
+ ld1 {v0.16b}, [x1], #16
+ tbnz x5, #1, 0f
+ ld1 {v1.16b}, [x1], #16
+ tbnz x5, #2, 0f
+ ld1 {v2.16b}, [x1], #16
+ tbnz x5, #3, 0f
+ ld1 {v3.16b}, [x1], #16
+ tbnz x5, #4, 0f
+ ld1 {v4.16b}, [x1], #16
+ tbnz x5, #5, 0f
+ ld1 {v5.16b}, [x1], #16
+ tbnz x5, #6, 0f
+ ld1 {v6.16b}, [x1], #16
+ tbnz x5, #7, 0f
+ ld1 {v7.16b}, [x1], #16
+
+0: mov bskey, x2
+ mov rounds, x3
+ bl \do8
+
+ st1 {\o0\().16b}, [x0], #16
+ tbnz x5, #1, 1f
+ st1 {\o1\().16b}, [x0], #16
+ tbnz x5, #2, 1f
+ st1 {\o2\().16b}, [x0], #16
+ tbnz x5, #3, 1f
+ st1 {\o3\().16b}, [x0], #16
+ tbnz x5, #4, 1f
+ st1 {\o4\().16b}, [x0], #16
+ tbnz x5, #5, 1f
+ st1 {\o5\().16b}, [x0], #16
+ tbnz x5, #6, 1f
+ st1 {\o6\().16b}, [x0], #16
+ tbnz x5, #7, 1f
+ st1 {\o7\().16b}, [x0], #16
+
+ cbnz x4, 99b
+
+1: ldp x29, x30, [sp], #16
+ ret
+ .endm
+
+ .align 4
+ENTRY(aesbs_ecb_encrypt)
+ __ecb_crypt aesbs_encrypt8, v0, v1, v4, v6, v3, v7, v2, v5
+ENDPROC(aesbs_ecb_encrypt)
+
+ .align 4
+ENTRY(aesbs_ecb_decrypt)
+ __ecb_crypt aesbs_decrypt8, v0, v1, v6, v4, v2, v7, v3, v5
+ENDPROC(aesbs_ecb_decrypt)
+
+ /*
+ * aesbs_cbc_decrypt(u8 out[], u8 const in[], u8 const rk[], int rounds,
+ * int blocks, u8 iv[])
+ */
+ .align 4
+ENTRY(aesbs_cbc_decrypt)
+ stp x29, x30, [sp, #-16]!
+ mov x29, sp
+
+99: mov x6, #1
+ lsl x6, x6, x4
+ subs w4, w4, #8
+ csel x4, x4, xzr, pl
+ csel x6, x6, xzr, mi
+
+ ld1 {v0.16b}, [x1], #16
+ mov v25.16b, v0.16b
+ tbnz x6, #1, 0f
+ ld1 {v1.16b}, [x1], #16
+ mov v26.16b, v1.16b
+ tbnz x6, #2, 0f
+ ld1 {v2.16b}, [x1], #16
+ mov v27.16b, v2.16b
+ tbnz x6, #3, 0f
+ ld1 {v3.16b}, [x1], #16
+ mov v28.16b, v3.16b
+ tbnz x6, #4, 0f
+ ld1 {v4.16b}, [x1], #16
+ mov v29.16b, v4.16b
+ tbnz x6, #5, 0f
+ ld1 {v5.16b}, [x1], #16
+ mov v30.16b, v5.16b
+ tbnz x6, #6, 0f
+ ld1 {v6.16b}, [x1], #16
+ mov v31.16b, v6.16b
+ tbnz x6, #7, 0f
+ ld1 {v7.16b}, [x1]
+
+0: mov bskey, x2
+ mov rounds, x3
+ bl aesbs_decrypt8
+
+ ld1 {v24.16b}, [x5] // load IV
+
+ eor v1.16b, v1.16b, v25.16b
+ eor v6.16b, v6.16b, v26.16b
+ eor v4.16b, v4.16b, v27.16b
+ eor v2.16b, v2.16b, v28.16b
+ eor v7.16b, v7.16b, v29.16b
+ eor v0.16b, v0.16b, v24.16b
+ eor v3.16b, v3.16b, v30.16b
+ eor v5.16b, v5.16b, v31.16b
+
+ st1 {v0.16b}, [x0], #16
+ mov v24.16b, v25.16b
+ tbnz x6, #1, 1f
+ st1 {v1.16b}, [x0], #16
+ mov v24.16b, v26.16b
+ tbnz x6, #2, 1f
+ st1 {v6.16b}, [x0], #16
+ mov v24.16b, v27.16b
+ tbnz x6, #3, 1f
+ st1 {v4.16b}, [x0], #16
+ mov v24.16b, v28.16b
+ tbnz x6, #4, 1f
+ st1 {v2.16b}, [x0], #16
+ mov v24.16b, v29.16b
+ tbnz x6, #5, 1f
+ st1 {v7.16b}, [x0], #16
+ mov v24.16b, v30.16b
+ tbnz x6, #6, 1f
+ st1 {v3.16b}, [x0], #16
+ mov v24.16b, v31.16b
+ tbnz x6, #7, 1f
+ ld1 {v24.16b}, [x1], #16
+ st1 {v5.16b}, [x0], #16
+1: st1 {v24.16b}, [x5] // store IV
+
+ cbnz x4, 99b
+
+ ldp x29, x30, [sp], #16
+ ret
+ENDPROC(aesbs_cbc_decrypt)
+
+ .macro next_tweak, out, in, const, tmp
+ sshr \tmp\().2d, \in\().2d, #63
+ and \tmp\().16b, \tmp\().16b, \const\().16b
+ add \out\().2d, \in\().2d, \in\().2d
+ ext \tmp\().16b, \tmp\().16b, \tmp\().16b, #8
+ eor \out\().16b, \out\().16b, \tmp\().16b
+ .endm
+
+ .align 4
+.Lxts_mul_x:
+CPU_LE( .quad 1, 0x87 )
+CPU_BE( .quad 0x87, 1 )
+
+ /*
+ * aesbs_xts_encrypt(u8 out[], u8 const in[], u8 const rk[], int rounds,
+ * int blocks, u8 iv[])
+ * aesbs_xts_decrypt(u8 out[], u8 const in[], u8 const rk[], int rounds,
+ * int blocks, u8 iv[])
+ */
+__xts_crypt8:
+ mov x6, #1
+ lsl x6, x6, x4
+ subs w4, w4, #8
+ csel x4, x4, xzr, pl
+ csel x6, x6, xzr, mi
+
+ ld1 {v0.16b}, [x1], #16
+ next_tweak v26, v25, v30, v31
+ eor v0.16b, v0.16b, v25.16b
+ tbnz x6, #1, 0f
+
+ ld1 {v1.16b}, [x1], #16
+ next_tweak v27, v26, v30, v31
+ eor v1.16b, v1.16b, v26.16b
+ tbnz x6, #2, 0f
+
+ ld1 {v2.16b}, [x1], #16
+ next_tweak v28, v27, v30, v31
+ eor v2.16b, v2.16b, v27.16b
+ tbnz x6, #3, 0f
+
+ ld1 {v3.16b}, [x1], #16
+ next_tweak v29, v28, v30, v31
+ eor v3.16b, v3.16b, v28.16b
+ tbnz x6, #4, 0f
+
+ ld1 {v4.16b}, [x1], #16
+ str q29, [sp, #16]
+ eor v4.16b, v4.16b, v29.16b
+ next_tweak v29, v29, v30, v31
+ tbnz x6, #5, 0f
+
+ ld1 {v5.16b}, [x1], #16
+ str q29, [sp, #32]
+ eor v5.16b, v5.16b, v29.16b
+ next_tweak v29, v29, v30, v31
+ tbnz x6, #6, 0f
+
+ ld1 {v6.16b}, [x1], #16
+ str q29, [sp, #48]
+ eor v6.16b, v6.16b, v29.16b
+ next_tweak v29, v29, v30, v31
+ tbnz x6, #7, 0f
+
+ ld1 {v7.16b}, [x1], #16
+ str q29, [sp, #64]
+ eor v7.16b, v7.16b, v29.16b
+ next_tweak v29, v29, v30, v31
+
+0: mov bskey, x2
+ mov rounds, x3
+ br x7
+ENDPROC(__xts_crypt8)
+
+ .macro __xts_crypt, do8, o0, o1, o2, o3, o4, o5, o6, o7
+ stp x29, x30, [sp, #-80]!
+ mov x29, sp
+
+ ldr q30, .Lxts_mul_x
+ ld1 {v25.16b}, [x5]
+
+99: adr x7, \do8
+ bl __xts_crypt8
+
+ ldp q16, q17, [sp, #16]
+ ldp q18, q19, [sp, #48]
+
+ eor \o0\().16b, \o0\().16b, v25.16b
+ eor \o1\().16b, \o1\().16b, v26.16b
+ eor \o2\().16b, \o2\().16b, v27.16b
+ eor \o3\().16b, \o3\().16b, v28.16b
+
+ st1 {\o0\().16b}, [x0], #16
+ mov v25.16b, v26.16b
+ tbnz x6, #1, 1f
+ st1 {\o1\().16b}, [x0], #16
+ mov v25.16b, v27.16b
+ tbnz x6, #2, 1f
+ st1 {\o2\().16b}, [x0], #16
+ mov v25.16b, v28.16b
+ tbnz x6, #3, 1f
+ st1 {\o3\().16b}, [x0], #16
+ mov v25.16b, v29.16b
+ tbnz x6, #4, 1f
+
+ eor \o4\().16b, \o4\().16b, v16.16b
+ eor \o5\().16b, \o5\().16b, v17.16b
+ eor \o6\().16b, \o6\().16b, v18.16b
+ eor \o7\().16b, \o7\().16b, v19.16b
+
+ st1 {\o4\().16b}, [x0], #16
+ tbnz x6, #5, 1f
+ st1 {\o5\().16b}, [x0], #16
+ tbnz x6, #6, 1f
+ st1 {\o6\().16b}, [x0], #16
+ tbnz x6, #7, 1f
+ st1 {\o7\().16b}, [x0], #16
+
+ cbnz x4, 99b
+
+1: st1 {v25.16b}, [x5]
+ ldp x29, x30, [sp], #80
+ ret
+ .endm
+
+ENTRY(aesbs_xts_encrypt)
+ __xts_crypt aesbs_encrypt8, v0, v1, v4, v6, v3, v7, v2, v5
+ENDPROC(aesbs_xts_encrypt)
+
+ENTRY(aesbs_xts_decrypt)
+ __xts_crypt aesbs_decrypt8, v0, v1, v6, v4, v2, v7, v3, v5
+ENDPROC(aesbs_xts_decrypt)
+
+ .macro next_ctr, v
+ mov \v\().d[1], x8
+ adds x8, x8, #1
+ mov \v\().d[0], x7
+ adc x7, x7, xzr
+ rev64 \v\().16b, \v\().16b
+ .endm
+
+ /*
+ * aesbs_ctr_encrypt(u8 out[], u8 const in[], u8 const rk[],
+ * int rounds, int blocks, u8 iv[], bool final)
+ */
+ENTRY(aesbs_ctr_encrypt)
+ stp x29, x30, [sp, #-16]!
+ mov x29, sp
+
+ add x4, x4, x6 // do one extra block if final
+
+ ldp x7, x8, [x5]
+ ld1 {v0.16b}, [x5]
+CPU_LE( rev x7, x7 )
+CPU_LE( rev x8, x8 )
+ adds x8, x8, #1
+ adc x7, x7, xzr
+
+99: mov x9, #1
+ lsl x9, x9, x4
+ subs w4, w4, #8
+ csel x4, x4, xzr, pl
+ csel x9, x9, xzr, le
+
+ next_ctr v1
+ next_ctr v2
+ next_ctr v3
+ next_ctr v4
+ next_ctr v5
+ next_ctr v6
+ next_ctr v7
+
+0: mov bskey, x2
+ mov rounds, x3
+ bl aesbs_encrypt8
+
+ lsr x9, x9, x6 // disregard the extra block
+ tbnz x9, #0, 0f
+
+ ld1 {v8.16b}, [x1], #16
+ eor v0.16b, v0.16b, v8.16b
+ st1 {v0.16b}, [x0], #16
+ tbnz x9, #1, 1f
+
+ ld1 {v9.16b}, [x1], #16
+ eor v1.16b, v1.16b, v9.16b
+ st1 {v1.16b}, [x0], #16
+ tbnz x9, #2, 2f
+
+ ld1 {v10.16b}, [x1], #16
+ eor v4.16b, v4.16b, v10.16b
+ st1 {v4.16b}, [x0], #16
+ tbnz x9, #3, 3f
+
+ ld1 {v11.16b}, [x1], #16
+ eor v6.16b, v6.16b, v11.16b
+ st1 {v6.16b}, [x0], #16
+ tbnz x9, #4, 4f
+
+ ld1 {v12.16b}, [x1], #16
+ eor v3.16b, v3.16b, v12.16b
+ st1 {v3.16b}, [x0], #16
+ tbnz x9, #5, 5f
+
+ ld1 {v13.16b}, [x1], #16
+ eor v7.16b, v7.16b, v13.16b
+ st1 {v7.16b}, [x0], #16
+ tbnz x9, #6, 6f
+
+ ld1 {v14.16b}, [x1], #16
+ eor v2.16b, v2.16b, v14.16b
+ st1 {v2.16b}, [x0], #16
+ tbnz x9, #7, 7f
+
+ ld1 {v15.16b}, [x1], #16
+ eor v5.16b, v5.16b, v15.16b
+ st1 {v5.16b}, [x0], #16
+
+ next_ctr v0
+ cbnz x4, 99b
+
+0: st1 {v0.16b}, [x5]
+8: ldp x29, x30, [sp], #16
+ ret
+
+ /*
+ * If we are handling the tail of the input (x6 == 1), return the
+ * final keystream block back to the caller via the IV buffer.
+ */
+1: cbz x6, 8b
+ st1 {v1.16b}, [x5]
+ b 8b
+2: cbz x6, 8b
+ st1 {v4.16b}, [x5]
+ b 8b
+3: cbz x6, 8b
+ st1 {v6.16b}, [x5]
+ b 8b
+4: cbz x6, 8b
+ st1 {v3.16b}, [x5]
+ b 8b
+5: cbz x6, 8b
+ st1 {v7.16b}, [x5]
+ b 8b
+6: cbz x6, 8b
+ st1 {v2.16b}, [x5]
+ b 8b
+7: cbz x6, 8b
+ st1 {v5.16b}, [x5]
+ b 8b
+ENDPROC(aesbs_ctr_encrypt)
diff --git a/arch/arm64/crypto/aes-neonbs-glue.c b/arch/arm64/crypto/aes-neonbs-glue.c
new file mode 100644
index 000000000000..323dd76ae5f0
--- /dev/null
+++ b/arch/arm64/crypto/aes-neonbs-glue.c
@@ -0,0 +1,420 @@
+/*
+ * Bit sliced AES using NEON instructions
+ *
+ * Copyright (C) 2016 Linaro Ltd <ard.biesheuvel@linaro.org>
+ *
+ * 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 by the Free Software Foundation.
+ */
+
+#include <asm/neon.h>
+#include <crypto/aes.h>
+#include <crypto/cbc.h>
+#include <crypto/internal/simd.h>
+#include <crypto/internal/skcipher.h>
+#include <crypto/xts.h>
+#include <linux/module.h>
+
+MODULE_AUTHOR("Ard Biesheuvel <ard.biesheuvel@linaro.org>");
+MODULE_LICENSE("GPL v2");
+
+MODULE_ALIAS_CRYPTO("ecb(aes)");
+MODULE_ALIAS_CRYPTO("cbc(aes)");
+MODULE_ALIAS_CRYPTO("ctr(aes)");
+MODULE_ALIAS_CRYPTO("xts(aes)");
+
+asmlinkage void aesbs_convert_key(u8 out[], u32 const rk[], int rounds);
+
+asmlinkage void aesbs_ecb_encrypt(u8 out[], u8 const in[], u8 const rk[],
+ int rounds, int blocks);
+asmlinkage void aesbs_ecb_decrypt(u8 out[], u8 const in[], u8 const rk[],
+ int rounds, int blocks);
+
+asmlinkage void aesbs_cbc_decrypt(u8 out[], u8 const in[], u8 const rk[],
+ int rounds, int blocks, u8 iv[]);
+
+asmlinkage void aesbs_ctr_encrypt(u8 out[], u8 const in[], u8 const rk[],
+ int rounds, int blocks, u8 iv[], bool final);
+
+asmlinkage void aesbs_xts_encrypt(u8 out[], u8 const in[], u8 const rk[],
+ int rounds, int blocks, u8 iv[]);
+asmlinkage void aesbs_xts_decrypt(u8 out[], u8 const in[], u8 const rk[],
+ int rounds, int blocks, u8 iv[]);
+
+asmlinkage void __aes_arm64_encrypt(u32 *rk, u8 *out, const u8 *in, int rounds);
+
+struct aesbs_ctx {
+ u8 rk[13 * (8 * AES_BLOCK_SIZE) + 32];
+ int rounds;
+} __aligned(AES_BLOCK_SIZE);
+
+struct aesbs_cbc_ctx {
+ struct aesbs_ctx key;
+ u32 enc[AES_MAX_KEYLENGTH_U32];
+};
+
+struct aesbs_xts_ctx {
+ struct aesbs_ctx key;
+ u32 twkey[AES_MAX_KEYLENGTH_U32];
+};
+
+static int aesbs_setkey(struct crypto_skcipher *tfm, const u8 *in_key,
+ unsigned int key_len)
+{
+ struct aesbs_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct crypto_aes_ctx rk;
+ int err;
+
+ err = crypto_aes_expand_key(&rk, in_key, key_len);
+ if (err)
+ return err;
+
+ ctx->rounds = 6 + key_len / 4;
+
+ kernel_neon_begin();
+ aesbs_convert_key(ctx->rk, rk.key_enc, ctx->rounds);
+ kernel_neon_end();
+
+ return 0;
+}
+
+static int __ecb_crypt(struct skcipher_request *req,
+ void (*fn)(u8 out[], u8 const in[], u8 const rk[],
+ int rounds, int blocks))
+{
+ struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+ struct aesbs_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct skcipher_walk walk;
+ int err;
+
+ err = skcipher_walk_virt(&walk, req, true);
+
+ kernel_neon_begin();
+ while (walk.nbytes >= AES_BLOCK_SIZE) {
+ unsigned int blocks = walk.nbytes / AES_BLOCK_SIZE;
+
+ if (walk.nbytes < walk.total)
+ blocks = round_down(blocks,
+ walk.stride / AES_BLOCK_SIZE);
+
+ fn(walk.dst.virt.addr, walk.src.virt.addr, ctx->rk,
+ ctx->rounds, blocks);
+ err = skcipher_walk_done(&walk,
+ walk.nbytes - blocks * AES_BLOCK_SIZE);
+ }
+ kernel_neon_end();
+
+ return err;
+}
+
+static int ecb_encrypt(struct skcipher_request *req)
+{
+ return __ecb_crypt(req, aesbs_ecb_encrypt);
+}
+
+static int ecb_decrypt(struct skcipher_request *req)
+{
+ return __ecb_crypt(req, aesbs_ecb_decrypt);
+}
+
+static int aesbs_cbc_setkey(struct crypto_skcipher *tfm, const u8 *in_key,
+ unsigned int key_len)
+{
+ struct aesbs_cbc_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct crypto_aes_ctx rk;
+ int err;
+
+ err = crypto_aes_expand_key(&rk, in_key, key_len);
+ if (err)
+ return err;
+
+ ctx->key.rounds = 6 + key_len / 4;
+
+ memcpy(ctx->enc, rk.key_enc, sizeof(ctx->enc));
+
+ kernel_neon_begin();
+ aesbs_convert_key(ctx->key.rk, rk.key_enc, ctx->key.rounds);
+ kernel_neon_end();
+
+ return 0;
+}
+
+static void cbc_encrypt_one(struct crypto_skcipher *tfm, const u8 *src, u8 *dst)
+{
+ struct aesbs_cbc_ctx *ctx = crypto_skcipher_ctx(tfm);
+
+ __aes_arm64_encrypt(ctx->enc, dst, src, ctx->key.rounds);
+}
+
+static int cbc_encrypt(struct skcipher_request *req)
+{
+ return crypto_cbc_encrypt_walk(req, cbc_encrypt_one);
+}
+
+static int cbc_decrypt(struct skcipher_request *req)
+{
+ struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+ struct aesbs_cbc_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct skcipher_walk walk;
+ int err;
+
+ err = skcipher_walk_virt(&walk, req, true);
+
+ kernel_neon_begin();
+ while (walk.nbytes >= AES_BLOCK_SIZE) {
+ unsigned int blocks = walk.nbytes / AES_BLOCK_SIZE;
+
+ if (walk.nbytes < walk.total)
+ blocks = round_down(blocks,
+ walk.stride / AES_BLOCK_SIZE);
+
+ aesbs_cbc_decrypt(walk.dst.virt.addr, walk.src.virt.addr,
+ ctx->key.rk, ctx->key.rounds, blocks,
+ walk.iv);
+ err = skcipher_walk_done(&walk,
+ walk.nbytes - blocks * AES_BLOCK_SIZE);
+ }
+ kernel_neon_end();
+
+ return err;
+}
+
+static int ctr_encrypt(struct skcipher_request *req)
+{
+ struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+ struct aesbs_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct skcipher_walk walk;
+ int err;
+
+ err = skcipher_walk_virt(&walk, req, true);
+
+ kernel_neon_begin();
+ while (walk.nbytes > 0) {
+ unsigned int blocks = walk.nbytes / AES_BLOCK_SIZE;
+ bool final = (walk.total % AES_BLOCK_SIZE) != 0;
+
+ if (walk.nbytes < walk.total) {
+ blocks = round_down(blocks,
+ walk.stride / AES_BLOCK_SIZE);
+ final = false;
+ }
+
+ aesbs_ctr_encrypt(walk.dst.virt.addr, walk.src.virt.addr,
+ ctx->rk, ctx->rounds, blocks, walk.iv, final);
+
+ if (final) {
+ u8 *dst = walk.dst.virt.addr + blocks * AES_BLOCK_SIZE;
+ u8 *src = walk.src.virt.addr + blocks * AES_BLOCK_SIZE;
+
+ if (dst != src)
+ memcpy(dst, src, walk.total % AES_BLOCK_SIZE);
+ crypto_xor(dst, walk.iv, walk.total % AES_BLOCK_SIZE);
+
+ err = skcipher_walk_done(&walk, 0);
+ break;
+ }
+ err = skcipher_walk_done(&walk,
+ walk.nbytes - blocks * AES_BLOCK_SIZE);
+ }
+ kernel_neon_end();
+
+ return err;
+}
+
+static int aesbs_xts_setkey(struct crypto_skcipher *tfm, const u8 *in_key,
+ unsigned int key_len)
+{
+ struct aesbs_xts_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct crypto_aes_ctx rk;
+ int err;
+
+ err = xts_verify_key(tfm, in_key, key_len);
+ if (err)
+ return err;
+
+ key_len /= 2;
+ err = crypto_aes_expand_key(&rk, in_key + key_len, key_len);
+ if (err)
+ return err;
+
+ memcpy(ctx->twkey, rk.key_enc, sizeof(ctx->twkey));
+
+ return aesbs_setkey(tfm, in_key, key_len);
+}
+
+static int __xts_crypt(struct skcipher_request *req,
+ void (*fn)(u8 out[], u8 const in[], u8 const rk[],
+ int rounds, int blocks, u8 iv[]))
+{
+ struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+ struct aesbs_xts_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct skcipher_walk walk;
+ int err;
+
+ err = skcipher_walk_virt(&walk, req, true);
+
+ __aes_arm64_encrypt(ctx->twkey, walk.iv, walk.iv, ctx->key.rounds);
+
+ kernel_neon_begin();
+ while (walk.nbytes >= AES_BLOCK_SIZE) {
+ unsigned int blocks = walk.nbytes / AES_BLOCK_SIZE;
+
+ if (walk.nbytes < walk.total)
+ blocks = round_down(blocks,
+ walk.stride / AES_BLOCK_SIZE);
+
+ fn(walk.dst.virt.addr, walk.src.virt.addr, ctx->key.rk,
+ ctx->key.rounds, blocks, walk.iv);
+ err = skcipher_walk_done(&walk,
+ walk.nbytes - blocks * AES_BLOCK_SIZE);
+ }
+ kernel_neon_end();
+
+ return err;
+}
+
+static int xts_encrypt(struct skcipher_request *req)
+{
+ return __xts_crypt(req, aesbs_xts_encrypt);
+}
+
+static int xts_decrypt(struct skcipher_request *req)
+{
+ return __xts_crypt(req, aesbs_xts_decrypt);
+}
+
+static struct skcipher_alg aes_algs[] = { {
+ .base.cra_name = "__ecb(aes)",
+ .base.cra_driver_name = "__ecb-aes-neonbs",
+ .base.cra_priority = 250,
+ .base.cra_blocksize = AES_BLOCK_SIZE,
+ .base.cra_ctxsize = sizeof(struct aesbs_ctx),
+ .base.cra_module = THIS_MODULE,
+ .base.cra_flags = CRYPTO_ALG_INTERNAL,
+
+ .min_keysize = AES_MIN_KEY_SIZE,
+ .max_keysize = AES_MAX_KEY_SIZE,
+ .walksize = 8 * AES_BLOCK_SIZE,
+ .setkey = aesbs_setkey,
+ .encrypt = ecb_encrypt,
+ .decrypt = ecb_decrypt,
+}, {
+ .base.cra_name = "__cbc(aes)",
+ .base.cra_driver_name = "__cbc-aes-neonbs",
+ .base.cra_priority = 250,
+ .base.cra_blocksize = AES_BLOCK_SIZE,
+ .base.cra_ctxsize = sizeof(struct aesbs_cbc_ctx),
+ .base.cra_module = THIS_MODULE,
+ .base.cra_flags = CRYPTO_ALG_INTERNAL,
+
+ .min_keysize = AES_MIN_KEY_SIZE,
+ .max_keysize = AES_MAX_KEY_SIZE,
+ .walksize = 8 * AES_BLOCK_SIZE,
+ .ivsize = AES_BLOCK_SIZE,
+ .setkey = aesbs_cbc_setkey,
+ .encrypt = cbc_encrypt,
+ .decrypt = cbc_decrypt,
+}, {
+ .base.cra_name = "__ctr(aes)",
+ .base.cra_driver_name = "__ctr-aes-neonbs",
+ .base.cra_priority = 250,
+ .base.cra_blocksize = 1,
+ .base.cra_ctxsize = sizeof(struct aesbs_ctx),
+ .base.cra_module = THIS_MODULE,
+ .base.cra_flags = CRYPTO_ALG_INTERNAL,
+
+ .min_keysize = AES_MIN_KEY_SIZE,
+ .max_keysize = AES_MAX_KEY_SIZE,
+ .chunksize = AES_BLOCK_SIZE,
+ .walksize = 8 * AES_BLOCK_SIZE,
+ .ivsize = AES_BLOCK_SIZE,
+ .setkey = aesbs_setkey,
+ .encrypt = ctr_encrypt,
+ .decrypt = ctr_encrypt,
+}, {
+ .base.cra_name = "ctr(aes)",
+ .base.cra_driver_name = "ctr-aes-neonbs",
+ .base.cra_priority = 250 - 1,
+ .base.cra_blocksize = 1,
+ .base.cra_ctxsize = sizeof(struct aesbs_ctx),
+ .base.cra_module = THIS_MODULE,
+
+ .min_keysize = AES_MIN_KEY_SIZE,
+ .max_keysize = AES_MAX_KEY_SIZE,
+ .chunksize = AES_BLOCK_SIZE,
+ .walksize = 8 * AES_BLOCK_SIZE,
+ .ivsize = AES_BLOCK_SIZE,
+ .setkey = aesbs_setkey,
+ .encrypt = ctr_encrypt,
+ .decrypt = ctr_encrypt,
+}, {
+ .base.cra_name = "__xts(aes)",
+ .base.cra_driver_name = "__xts-aes-neonbs",
+ .base.cra_priority = 250,
+ .base.cra_blocksize = AES_BLOCK_SIZE,
+ .base.cra_ctxsize = sizeof(struct aesbs_xts_ctx),
+ .base.cra_module = THIS_MODULE,
+ .base.cra_flags = CRYPTO_ALG_INTERNAL,
+
+ .min_keysize = 2 * AES_MIN_KEY_SIZE,
+ .max_keysize = 2 * AES_MAX_KEY_SIZE,
+ .walksize = 8 * AES_BLOCK_SIZE,
+ .ivsize = AES_BLOCK_SIZE,
+ .setkey = aesbs_xts_setkey,
+ .encrypt = xts_encrypt,
+ .decrypt = xts_decrypt,
+} };
+
+static struct simd_skcipher_alg *aes_simd_algs[ARRAY_SIZE(aes_algs)];
+
+static void aes_exit(void)
+{
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(aes_simd_algs); i++)
+ if (aes_simd_algs[i])
+ simd_skcipher_free(aes_simd_algs[i]);
+
+ crypto_unregister_skciphers(aes_algs, ARRAY_SIZE(aes_algs));
+}
+
+static int __init aes_init(void)
+{
+ struct simd_skcipher_alg *simd;
+ const char *basename;
+ const char *algname;
+ const char *drvname;
+ int err;
+ int i;
+
+ if (!(elf_hwcap & HWCAP_ASIMD))
+ return -ENODEV;
+
+ err = crypto_register_skciphers(aes_algs, ARRAY_SIZE(aes_algs));
+ if (err)
+ return err;
+
+ for (i = 0; i < ARRAY_SIZE(aes_algs); i++) {
+ if (!(aes_algs[i].base.cra_flags & CRYPTO_ALG_INTERNAL))
+ continue;
+
+ algname = aes_algs[i].base.cra_name + 2;
+ drvname = aes_algs[i].base.cra_driver_name + 2;
+ basename = aes_algs[i].base.cra_driver_name;
+ simd = simd_skcipher_create_compat(algname, drvname, basename);
+ err = PTR_ERR(simd);
+ if (IS_ERR(simd))
+ goto unregister_simds;
+
+ aes_simd_algs[i] = simd;
+ }
+ return 0;
+
+unregister_simds:
+ aes_exit();
+ return err;
+}
+
+module_init(aes_init);
+module_exit(aes_exit);
--
2.7.4
^ permalink raw reply related [flat|nested] 10+ messages in thread