diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2013-07-05 12:12:33 -0700 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2013-07-05 12:12:33 -0700 |
commit | b2c311075db578f1433d9b303698491bfa21279a (patch) | |
tree | 41d5f1b5ad6f45be7211f524328de81f7e9754be /arch/x86 | |
parent | 45175476ae2dbebc860d5cf486f2916044343513 (diff) | |
parent | 02c0241b600e4ab8a732c89749e252165145d60c (diff) | |
download | talos-op-linux-b2c311075db578f1433d9b303698491bfa21279a.tar.gz talos-op-linux-b2c311075db578f1433d9b303698491bfa21279a.zip |
Merge git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Pull crypto update from Herbert Xu:
- Do not idle omap device between crypto operations in one session.
- Added sha224/sha384 shims for SSSE3.
- More optimisations for camellia-aesni-avx2.
- Removed defunct blowfish/twofish AVX2 implementations.
- Added unaligned buffer self-tests.
- Added PCLMULQDQ optimisation for CRCT10DIF.
- Added support for Freescale's DCP co-processor
- Misc fixes.
* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (44 commits)
crypto: testmgr - test hash implementations with unaligned buffers
crypto: testmgr - test AEADs with unaligned buffers
crypto: testmgr - test skciphers with unaligned buffers
crypto: testmgr - check that entries in alg_test_descs are in correct order
Revert "crypto: twofish - add AVX2/x86_64 assembler implementation of twofish cipher"
Revert "crypto: blowfish - add AVX2/x86_64 implementation of blowfish cipher"
crypto: camellia-aesni-avx2 - tune assembly code for more performance
hwrng: bcm2835 - fix MODULE_LICENSE tag
hwrng: nomadik - use clk_prepare_enable()
crypto: picoxcell - replace strict_strtoul() with kstrtoul()
crypto: dcp - Staticize local symbols
crypto: dcp - Use NULL instead of 0
crypto: dcp - Use devm_* APIs
crypto: dcp - Remove redundant platform_set_drvdata()
hwrng: use platform_{get,set}_drvdata()
crypto: omap-aes - Don't idle/start AES device between Encrypt operations
crypto: crct10dif - Use PTR_RET
crypto: ux500 - Cocci spatch "resource_size.spatch"
crypto: sha256_ssse3 - add sha224 support
crypto: sha512_ssse3 - add sha384 support
...
Diffstat (limited to 'arch/x86')
-rw-r--r-- | arch/x86/crypto/Makefile | 8 | ||||
-rw-r--r-- | arch/x86/crypto/blowfish-avx2-asm_64.S | 449 | ||||
-rw-r--r-- | arch/x86/crypto/blowfish_avx2_glue.c | 585 | ||||
-rw-r--r-- | arch/x86/crypto/blowfish_glue.c | 32 | ||||
-rw-r--r-- | arch/x86/crypto/camellia-aesni-avx2-asm_64.S | 160 | ||||
-rw-r--r-- | arch/x86/crypto/crct10dif-pcl-asm_64.S | 643 | ||||
-rw-r--r-- | arch/x86/crypto/crct10dif-pclmul_glue.c | 151 | ||||
-rw-r--r-- | arch/x86/crypto/sha256_ssse3_glue.c | 57 | ||||
-rw-r--r-- | arch/x86/crypto/sha512_ssse3_glue.c | 58 | ||||
-rw-r--r-- | arch/x86/crypto/twofish-avx2-asm_64.S | 600 | ||||
-rw-r--r-- | arch/x86/crypto/twofish_avx2_glue.c | 584 | ||||
-rw-r--r-- | arch/x86/crypto/twofish_avx_glue.c | 14 | ||||
-rw-r--r-- | arch/x86/include/asm/crypto/blowfish.h | 43 | ||||
-rw-r--r-- | arch/x86/include/asm/crypto/twofish.h | 18 |
14 files changed, 1016 insertions, 2386 deletions
diff --git a/arch/x86/crypto/Makefile b/arch/x86/crypto/Makefile index a3a0ed80f17c..7d6ba9db1be9 100644 --- a/arch/x86/crypto/Makefile +++ b/arch/x86/crypto/Makefile @@ -3,8 +3,6 @@ # avx_supported := $(call as-instr,vpxor %xmm0$(comma)%xmm0$(comma)%xmm0,yes,no) -avx2_supported := $(call as-instr,vpgatherdd %ymm0$(comma)(%eax$(comma)%ymm1\ - $(comma)4)$(comma)%ymm2,yes,no) obj-$(CONFIG_CRYPTO_ABLK_HELPER_X86) += ablk_helper.o obj-$(CONFIG_CRYPTO_GLUE_HELPER_X86) += glue_helper.o @@ -29,6 +27,7 @@ obj-$(CONFIG_CRYPTO_SHA1_SSSE3) += sha1-ssse3.o obj-$(CONFIG_CRYPTO_CRC32_PCLMUL) += crc32-pclmul.o obj-$(CONFIG_CRYPTO_SHA256_SSSE3) += sha256-ssse3.o obj-$(CONFIG_CRYPTO_SHA512_SSSE3) += sha512-ssse3.o +obj-$(CONFIG_CRYPTO_CRCT10DIF_PCLMUL) += crct10dif-pclmul.o # These modules require assembler to support AVX. ifeq ($(avx_supported),yes) @@ -42,10 +41,8 @@ endif # These modules require assembler to support AVX2. ifeq ($(avx2_supported),yes) - obj-$(CONFIG_CRYPTO_BLOWFISH_AVX2_X86_64) += blowfish-avx2.o obj-$(CONFIG_CRYPTO_CAMELLIA_AESNI_AVX2_X86_64) += camellia-aesni-avx2.o obj-$(CONFIG_CRYPTO_SERPENT_AVX2_X86_64) += serpent-avx2.o - obj-$(CONFIG_CRYPTO_TWOFISH_AVX2_X86_64) += twofish-avx2.o endif aes-i586-y := aes-i586-asm_32.o aes_glue.o @@ -73,10 +70,8 @@ ifeq ($(avx_supported),yes) endif ifeq ($(avx2_supported),yes) - blowfish-avx2-y := blowfish-avx2-asm_64.o blowfish_avx2_glue.o camellia-aesni-avx2-y := camellia-aesni-avx2-asm_64.o camellia_aesni_avx2_glue.o serpent-avx2-y := serpent-avx2-asm_64.o serpent_avx2_glue.o - twofish-avx2-y := twofish-avx2-asm_64.o twofish_avx2_glue.o endif aesni-intel-y := aesni-intel_asm.o aesni-intel_glue.o fpu.o @@ -87,3 +82,4 @@ crc32c-intel-$(CONFIG_64BIT) += crc32c-pcl-intel-asm_64.o crc32-pclmul-y := crc32-pclmul_asm.o crc32-pclmul_glue.o sha256-ssse3-y := sha256-ssse3-asm.o sha256-avx-asm.o sha256-avx2-asm.o sha256_ssse3_glue.o sha512-ssse3-y := sha512-ssse3-asm.o sha512-avx-asm.o sha512-avx2-asm.o sha512_ssse3_glue.o +crct10dif-pclmul-y := crct10dif-pcl-asm_64.o crct10dif-pclmul_glue.o diff --git a/arch/x86/crypto/blowfish-avx2-asm_64.S b/arch/x86/crypto/blowfish-avx2-asm_64.S deleted file mode 100644 index 784452e0d05d..000000000000 --- a/arch/x86/crypto/blowfish-avx2-asm_64.S +++ /dev/null @@ -1,449 +0,0 @@ -/* - * x86_64/AVX2 assembler optimized version of Blowfish - * - * Copyright © 2012-2013 Jussi Kivilinna <jussi.kivilinna@iki.fi> - * - * 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> - -.file "blowfish-avx2-asm_64.S" - -.data -.align 32 - -.Lprefetch_mask: -.long 0*64 -.long 1*64 -.long 2*64 -.long 3*64 -.long 4*64 -.long 5*64 -.long 6*64 -.long 7*64 - -.Lbswap32_mask: -.long 0x00010203 -.long 0x04050607 -.long 0x08090a0b -.long 0x0c0d0e0f - -.Lbswap128_mask: - .byte 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0 -.Lbswap_iv_mask: - .byte 7, 6, 5, 4, 3, 2, 1, 0, 7, 6, 5, 4, 3, 2, 1, 0 - -.text -/* structure of crypto context */ -#define p 0 -#define s0 ((16 + 2) * 4) -#define s1 ((16 + 2 + (1 * 256)) * 4) -#define s2 ((16 + 2 + (2 * 256)) * 4) -#define s3 ((16 + 2 + (3 * 256)) * 4) - -/* register macros */ -#define CTX %rdi -#define RIO %rdx - -#define RS0 %rax -#define RS1 %r8 -#define RS2 %r9 -#define RS3 %r10 - -#define RLOOP %r11 -#define RLOOPd %r11d - -#define RXr0 %ymm8 -#define RXr1 %ymm9 -#define RXr2 %ymm10 -#define RXr3 %ymm11 -#define RXl0 %ymm12 -#define RXl1 %ymm13 -#define RXl2 %ymm14 -#define RXl3 %ymm15 - -/* temp regs */ -#define RT0 %ymm0 -#define RT0x %xmm0 -#define RT1 %ymm1 -#define RT1x %xmm1 -#define RIDX0 %ymm2 -#define RIDX1 %ymm3 -#define RIDX1x %xmm3 -#define RIDX2 %ymm4 -#define RIDX3 %ymm5 - -/* vpgatherdd mask and '-1' */ -#define RNOT %ymm6 - -/* byte mask, (-1 >> 24) */ -#define RBYTE %ymm7 - -/*********************************************************************** - * 32-way AVX2 blowfish - ***********************************************************************/ -#define F(xl, xr) \ - vpsrld $24, xl, RIDX0; \ - vpsrld $16, xl, RIDX1; \ - vpsrld $8, xl, RIDX2; \ - vpand RBYTE, RIDX1, RIDX1; \ - vpand RBYTE, RIDX2, RIDX2; \ - vpand RBYTE, xl, RIDX3; \ - \ - vpgatherdd RNOT, (RS0, RIDX0, 4), RT0; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - vpcmpeqd RIDX0, RIDX0, RIDX0; \ - \ - vpgatherdd RNOT, (RS1, RIDX1, 4), RT1; \ - vpcmpeqd RIDX1, RIDX1, RIDX1; \ - vpaddd RT0, RT1, RT0; \ - \ - vpgatherdd RIDX0, (RS2, RIDX2, 4), RT1; \ - vpxor RT0, RT1, RT0; \ - \ - vpgatherdd RIDX1, (RS3, RIDX3, 4), RT1; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - vpaddd RT0, RT1, RT0; \ - \ - vpxor RT0, xr, xr; - -#define add_roundkey(xl, nmem) \ - vpbroadcastd nmem, RT0; \ - vpxor RT0, xl ## 0, xl ## 0; \ - vpxor RT0, xl ## 1, xl ## 1; \ - vpxor RT0, xl ## 2, xl ## 2; \ - vpxor RT0, xl ## 3, xl ## 3; - -#define round_enc() \ - add_roundkey(RXr, p(CTX,RLOOP,4)); \ - F(RXl0, RXr0); \ - F(RXl1, RXr1); \ - F(RXl2, RXr2); \ - F(RXl3, RXr3); \ - \ - add_roundkey(RXl, p+4(CTX,RLOOP,4)); \ - F(RXr0, RXl0); \ - F(RXr1, RXl1); \ - F(RXr2, RXl2); \ - F(RXr3, RXl3); - -#define round_dec() \ - add_roundkey(RXr, p+4*2(CTX,RLOOP,4)); \ - F(RXl0, RXr0); \ - F(RXl1, RXr1); \ - F(RXl2, RXr2); \ - F(RXl3, RXr3); \ - \ - add_roundkey(RXl, p+4(CTX,RLOOP,4)); \ - F(RXr0, RXl0); \ - F(RXr1, RXl1); \ - F(RXr2, RXl2); \ - F(RXr3, RXl3); - -#define init_round_constants() \ - vpcmpeqd RNOT, RNOT, RNOT; \ - leaq s0(CTX), RS0; \ - leaq s1(CTX), RS1; \ - leaq s2(CTX), RS2; \ - leaq s3(CTX), RS3; \ - vpsrld $24, RNOT, RBYTE; - -#define transpose_2x2(x0, x1, t0) \ - vpunpckldq x0, x1, t0; \ - vpunpckhdq x0, x1, x1; \ - \ - vpunpcklqdq t0, x1, x0; \ - vpunpckhqdq t0, x1, x1; - -#define read_block(xl, xr) \ - vbroadcasti128 .Lbswap32_mask, RT1; \ - \ - vpshufb RT1, xl ## 0, xl ## 0; \ - vpshufb RT1, xr ## 0, xr ## 0; \ - vpshufb RT1, xl ## 1, xl ## 1; \ - vpshufb RT1, xr ## 1, xr ## 1; \ - vpshufb RT1, xl ## 2, xl ## 2; \ - vpshufb RT1, xr ## 2, xr ## 2; \ - vpshufb RT1, xl ## 3, xl ## 3; \ - vpshufb RT1, xr ## 3, xr ## 3; \ - \ - transpose_2x2(xl ## 0, xr ## 0, RT0); \ - transpose_2x2(xl ## 1, xr ## 1, RT0); \ - transpose_2x2(xl ## 2, xr ## 2, RT0); \ - transpose_2x2(xl ## 3, xr ## 3, RT0); - -#define write_block(xl, xr) \ - vbroadcasti128 .Lbswap32_mask, RT1; \ - \ - transpose_2x2(xl ## 0, xr ## 0, RT0); \ - transpose_2x2(xl ## 1, xr ## 1, RT0); \ - transpose_2x2(xl ## 2, xr ## 2, RT0); \ - transpose_2x2(xl ## 3, xr ## 3, RT0); \ - \ - vpshufb RT1, xl ## 0, xl ## 0; \ - vpshufb RT1, xr ## 0, xr ## 0; \ - vpshufb RT1, xl ## 1, xl ## 1; \ - vpshufb RT1, xr ## 1, xr ## 1; \ - vpshufb RT1, xl ## 2, xl ## 2; \ - vpshufb RT1, xr ## 2, xr ## 2; \ - vpshufb RT1, xl ## 3, xl ## 3; \ - vpshufb RT1, xr ## 3, xr ## 3; - -.align 8 -__blowfish_enc_blk32: - /* input: - * %rdi: ctx, CTX - * RXl0..4, RXr0..4: plaintext - * output: - * RXl0..4, RXr0..4: ciphertext (RXl <=> RXr swapped) - */ - init_round_constants(); - - read_block(RXl, RXr); - - movl $1, RLOOPd; - add_roundkey(RXl, p+4*(0)(CTX)); - -.align 4 -.L__enc_loop: - round_enc(); - - leal 2(RLOOPd), RLOOPd; - cmpl $17, RLOOPd; - jne .L__enc_loop; - - add_roundkey(RXr, p+4*(17)(CTX)); - - write_block(RXl, RXr); - - ret; -ENDPROC(__blowfish_enc_blk32) - -.align 8 -__blowfish_dec_blk32: - /* input: - * %rdi: ctx, CTX - * RXl0..4, RXr0..4: ciphertext - * output: - * RXl0..4, RXr0..4: plaintext (RXl <=> RXr swapped) - */ - init_round_constants(); - - read_block(RXl, RXr); - - movl $14, RLOOPd; - add_roundkey(RXl, p+4*(17)(CTX)); - -.align 4 -.L__dec_loop: - round_dec(); - - addl $-2, RLOOPd; - jns .L__dec_loop; - - add_roundkey(RXr, p+4*(0)(CTX)); - - write_block(RXl, RXr); - - ret; -ENDPROC(__blowfish_dec_blk32) - -ENTRY(blowfish_ecb_enc_32way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst - * %rdx: src - */ - - vzeroupper; - - vmovdqu 0*32(%rdx), RXl0; - vmovdqu 1*32(%rdx), RXr0; - vmovdqu 2*32(%rdx), RXl1; - vmovdqu 3*32(%rdx), RXr1; - vmovdqu 4*32(%rdx), RXl2; - vmovdqu 5*32(%rdx), RXr2; - vmovdqu 6*32(%rdx), RXl3; - vmovdqu 7*32(%rdx), RXr3; - - call __blowfish_enc_blk32; - - vmovdqu RXr0, 0*32(%rsi); - vmovdqu RXl0, 1*32(%rsi); - vmovdqu RXr1, 2*32(%rsi); - vmovdqu RXl1, 3*32(%rsi); - vmovdqu RXr2, 4*32(%rsi); - vmovdqu RXl2, 5*32(%rsi); - vmovdqu RXr3, 6*32(%rsi); - vmovdqu RXl3, 7*32(%rsi); - - vzeroupper; - - ret; -ENDPROC(blowfish_ecb_enc_32way) - -ENTRY(blowfish_ecb_dec_32way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst - * %rdx: src - */ - - vzeroupper; - - vmovdqu 0*32(%rdx), RXl0; - vmovdqu 1*32(%rdx), RXr0; - vmovdqu 2*32(%rdx), RXl1; - vmovdqu 3*32(%rdx), RXr1; - vmovdqu 4*32(%rdx), RXl2; - vmovdqu 5*32(%rdx), RXr2; - vmovdqu 6*32(%rdx), RXl3; - vmovdqu 7*32(%rdx), RXr3; - - call __blowfish_dec_blk32; - - vmovdqu RXr0, 0*32(%rsi); - vmovdqu RXl0, 1*32(%rsi); - vmovdqu RXr1, 2*32(%rsi); - vmovdqu RXl1, 3*32(%rsi); - vmovdqu RXr2, 4*32(%rsi); - vmovdqu RXl2, 5*32(%rsi); - vmovdqu RXr3, 6*32(%rsi); - vmovdqu RXl3, 7*32(%rsi); - - vzeroupper; - - ret; -ENDPROC(blowfish_ecb_dec_32way) - -ENTRY(blowfish_cbc_dec_32way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst - * %rdx: src - */ - - vzeroupper; - - vmovdqu 0*32(%rdx), RXl0; - vmovdqu 1*32(%rdx), RXr0; - vmovdqu 2*32(%rdx), RXl1; - vmovdqu 3*32(%rdx), RXr1; - vmovdqu 4*32(%rdx), RXl2; - vmovdqu 5*32(%rdx), RXr2; - vmovdqu 6*32(%rdx), RXl3; - vmovdqu 7*32(%rdx), RXr3; - - call __blowfish_dec_blk32; - - /* xor with src */ - vmovq (%rdx), RT0x; - vpshufd $0x4f, RT0x, RT0x; - vinserti128 $1, 8(%rdx), RT0, RT0; - vpxor RT0, RXr0, RXr0; - vpxor 0*32+24(%rdx), RXl0, RXl0; - vpxor 1*32+24(%rdx), RXr1, RXr1; - vpxor 2*32+24(%rdx), RXl1, RXl1; - vpxor 3*32+24(%rdx), RXr2, RXr2; - vpxor 4*32+24(%rdx), RXl2, RXl2; - vpxor 5*32+24(%rdx), RXr3, RXr3; - vpxor 6*32+24(%rdx), RXl3, RXl3; - - vmovdqu RXr0, (0*32)(%rsi); - vmovdqu RXl0, (1*32)(%rsi); - vmovdqu RXr1, (2*32)(%rsi); - vmovdqu RXl1, (3*32)(%rsi); - vmovdqu RXr2, (4*32)(%rsi); - vmovdqu RXl2, (5*32)(%rsi); - vmovdqu RXr3, (6*32)(%rsi); - vmovdqu RXl3, (7*32)(%rsi); - - vzeroupper; - - ret; -ENDPROC(blowfish_cbc_dec_32way) - -ENTRY(blowfish_ctr_32way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst - * %rdx: src - * %rcx: iv (big endian, 64bit) - */ - - vzeroupper; - - vpcmpeqd RT0, RT0, RT0; - vpsrldq $8, RT0, RT0; /* a: -1, b: 0, c: -1, d: 0 */ - - vpcmpeqd RT1x, RT1x, RT1x; - vpaddq RT1x, RT1x, RT1x; /* a: -2, b: -2 */ - vpxor RIDX0, RIDX0, RIDX0; - vinserti128 $1, RT1x, RIDX0, RIDX0; /* a: 0, b: 0, c: -2, d: -2 */ - - vpaddq RIDX0, RT0, RT0; /* a: -1, b: 0, c: -3, d: -2 */ - - vpcmpeqd RT1, RT1, RT1; - vpaddq RT1, RT1, RT1; /* a: -2, b: -2, c: -2, d: -2 */ - vpaddq RT1, RT1, RIDX2; /* a: -4, b: -4, c: -4, d: -4 */ - - vbroadcasti128 .Lbswap_iv_mask, RIDX0; - vbroadcasti128 .Lbswap128_mask, RIDX1; - - /* load IV and byteswap */ - vmovq (%rcx), RT1x; - vinserti128 $1, RT1x, RT1, RT1; /* a: BE, b: 0, c: BE, d: 0 */ - vpshufb RIDX0, RT1, RT1; /* a: LE, b: LE, c: LE, d: LE */ - - /* construct IVs */ - vpsubq RT0, RT1, RT1; /* a: le1, b: le0, c: le3, d: le2 */ - vpshufb RIDX1, RT1, RXl0; /* a: be0, b: be1, c: be2, d: be3 */ - vpsubq RIDX2, RT1, RT1; /* le5, le4, le7, le6 */ - vpshufb RIDX1, RT1, RXr0; /* be4, be5, be6, be7 */ - vpsubq RIDX2, RT1, RT1; - vpshufb RIDX1, RT1, RXl1; - vpsubq RIDX2, RT1, RT1; - vpshufb RIDX1, RT1, RXr1; - vpsubq RIDX2, RT1, RT1; - vpshufb RIDX1, RT1, RXl2; - vpsubq RIDX2, RT1, RT1; - vpshufb RIDX1, RT1, RXr2; - vpsubq RIDX2, RT1, RT1; - vpshufb RIDX1, RT1, RXl3; - vpsubq RIDX2, RT1, RT1; - vpshufb RIDX1, RT1, RXr3; - - /* store last IV */ - vpsubq RIDX2, RT1, RT1; /* a: le33, b: le32, ... */ - vpshufb RIDX1x, RT1x, RT1x; /* a: be32, ... */ - vmovq RT1x, (%rcx); - - call __blowfish_enc_blk32; - - /* dst = src ^ iv */ - vpxor 0*32(%rdx), RXr0, RXr0; - vpxor 1*32(%rdx), RXl0, RXl0; - vpxor 2*32(%rdx), RXr1, RXr1; - vpxor 3*32(%rdx), RXl1, RXl1; - vpxor 4*32(%rdx), RXr2, RXr2; - vpxor 5*32(%rdx), RXl2, RXl2; - vpxor 6*32(%rdx), RXr3, RXr3; - vpxor 7*32(%rdx), RXl3, RXl3; - vmovdqu RXr0, (0*32)(%rsi); - vmovdqu RXl0, (1*32)(%rsi); - vmovdqu RXr1, (2*32)(%rsi); - vmovdqu RXl1, (3*32)(%rsi); - vmovdqu RXr2, (4*32)(%rsi); - vmovdqu RXl2, (5*32)(%rsi); - vmovdqu RXr3, (6*32)(%rsi); - vmovdqu RXl3, (7*32)(%rsi); - - vzeroupper; - - ret; -ENDPROC(blowfish_ctr_32way) diff --git a/arch/x86/crypto/blowfish_avx2_glue.c b/arch/x86/crypto/blowfish_avx2_glue.c deleted file mode 100644 index 4417e9aea78d..000000000000 --- a/arch/x86/crypto/blowfish_avx2_glue.c +++ /dev/null @@ -1,585 +0,0 @@ -/* - * Glue Code for x86_64/AVX2 assembler optimized version of Blowfish - * - * Copyright © 2012-2013 Jussi Kivilinna <jussi.kivilinna@mbnet.fi> - * - * CBC & ECB parts based on code (crypto/cbc.c,ecb.c) by: - * Copyright (c) 2006 Herbert Xu <herbert@gondor.apana.org.au> - * CTR part based on code (crypto/ctr.c) by: - * (C) Copyright IBM Corp. 2007 - Joy Latten <latten@us.ibm.com> - * - * 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. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - */ - -#include <linux/module.h> -#include <linux/types.h> -#include <linux/crypto.h> -#include <linux/err.h> -#include <crypto/algapi.h> -#include <crypto/blowfish.h> -#include <crypto/cryptd.h> -#include <crypto/ctr.h> -#include <asm/i387.h> -#include <asm/xcr.h> -#include <asm/xsave.h> -#include <asm/crypto/blowfish.h> -#include <asm/crypto/ablk_helper.h> -#include <crypto/scatterwalk.h> - -#define BF_AVX2_PARALLEL_BLOCKS 32 - -/* 32-way AVX2 parallel cipher functions */ -asmlinkage void blowfish_ecb_enc_32way(struct bf_ctx *ctx, u8 *dst, - const u8 *src); -asmlinkage void blowfish_ecb_dec_32way(struct bf_ctx *ctx, u8 *dst, - const u8 *src); -asmlinkage void blowfish_cbc_dec_32way(struct bf_ctx *ctx, u8 *dst, - const u8 *src); -asmlinkage void blowfish_ctr_32way(struct bf_ctx *ctx, u8 *dst, const u8 *src, - __be64 *iv); - -static inline bool bf_fpu_begin(bool fpu_enabled, unsigned int nbytes) -{ - if (fpu_enabled) - return true; - - /* FPU is only used when chunk to be processed is large enough, so - * do not enable FPU until it is necessary. - */ - if (nbytes < BF_BLOCK_SIZE * BF_AVX2_PARALLEL_BLOCKS) - return false; - - kernel_fpu_begin(); - return true; -} - -static inline void bf_fpu_end(bool fpu_enabled) -{ - if (fpu_enabled) - kernel_fpu_end(); -} - -static int ecb_crypt(struct blkcipher_desc *desc, struct blkcipher_walk *walk, - bool enc) -{ - bool fpu_enabled = false; - struct bf_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); - const unsigned int bsize = BF_BLOCK_SIZE; - unsigned int nbytes; - int err; - - err = blkcipher_walk_virt(desc, walk); - desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; - - while ((nbytes = walk->nbytes)) { - u8 *wsrc = walk->src.virt.addr; - u8 *wdst = walk->dst.virt.addr; - - fpu_enabled = bf_fpu_begin(fpu_enabled, nbytes); - - /* Process multi-block AVX2 batch */ - if (nbytes >= bsize * BF_AVX2_PARALLEL_BLOCKS) { - do { - if (enc) - blowfish_ecb_enc_32way(ctx, wdst, wsrc); - else - blowfish_ecb_dec_32way(ctx, wdst, wsrc); - - wsrc += bsize * BF_AVX2_PARALLEL_BLOCKS; - wdst += bsize * BF_AVX2_PARALLEL_BLOCKS; - nbytes -= bsize * BF_AVX2_PARALLEL_BLOCKS; - } while (nbytes >= bsize * BF_AVX2_PARALLEL_BLOCKS); - - if (nbytes < bsize) - goto done; - } - - /* Process multi-block batch */ - if (nbytes >= bsize * BF_PARALLEL_BLOCKS) { - do { - if (enc) - blowfish_enc_blk_4way(ctx, wdst, wsrc); - else - blowfish_dec_blk_4way(ctx, wdst, wsrc); - - wsrc += bsize * BF_PARALLEL_BLOCKS; - wdst += bsize * BF_PARALLEL_BLOCKS; - nbytes -= bsize * BF_PARALLEL_BLOCKS; - } while (nbytes >= bsize * BF_PARALLEL_BLOCKS); - - if (nbytes < bsize) - goto done; - } - - /* Handle leftovers */ - do { - if (enc) - blowfish_enc_blk(ctx, wdst, wsrc); - else - blowfish_dec_blk(ctx, wdst, wsrc); - - wsrc += bsize; - wdst += bsize; - nbytes -= bsize; - } while (nbytes >= bsize); - -done: - err = blkcipher_walk_done(desc, walk, nbytes); - } - - bf_fpu_end(fpu_enabled); - return err; -} - -static int ecb_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - struct blkcipher_walk walk; - - blkcipher_walk_init(&walk, dst, src, nbytes); - return ecb_crypt(desc, &walk, true); -} - -static int ecb_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - struct blkcipher_walk walk; - - blkcipher_walk_init(&walk, dst, src, nbytes); - return ecb_crypt(desc, &walk, false); -} - -static unsigned int __cbc_encrypt(struct blkcipher_desc *desc, - struct blkcipher_walk *walk) -{ - struct bf_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); - unsigned int bsize = BF_BLOCK_SIZE; - unsigned int nbytes = walk->nbytes; - u64 *src = (u64 *)walk->src.virt.addr; - u64 *dst = (u64 *)walk->dst.virt.addr; - u64 *iv = (u64 *)walk->iv; - - do { - *dst = *src ^ *iv; - blowfish_enc_blk(ctx, (u8 *)dst, (u8 *)dst); - iv = dst; - - src += 1; - dst += 1; - nbytes -= bsize; - } while (nbytes >= bsize); - - *(u64 *)walk->iv = *iv; - return nbytes; -} - -static int cbc_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - struct blkcipher_walk walk; - int err; - - blkcipher_walk_init(&walk, dst, src, nbytes); - err = blkcipher_walk_virt(desc, &walk); - - while ((nbytes = walk.nbytes)) { - nbytes = __cbc_encrypt(desc, &walk); - err = blkcipher_walk_done(desc, &walk, nbytes); - } - - return err; -} - -static unsigned int __cbc_decrypt(struct blkcipher_desc *desc, - struct blkcipher_walk *walk) -{ - struct bf_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); - const unsigned int bsize = BF_BLOCK_SIZE; - unsigned int nbytes = walk->nbytes; - u64 *src = (u64 *)walk->src.virt.addr; - u64 *dst = (u64 *)walk->dst.virt.addr; - u64 last_iv; - int i; - - /* Start of the last block. */ - src += nbytes / bsize - 1; - dst += nbytes / bsize - 1; - - last_iv = *src; - - /* Process multi-block AVX2 batch */ - if (nbytes >= bsize * BF_AVX2_PARALLEL_BLOCKS) { - do { - nbytes -= bsize * (BF_AVX2_PARALLEL_BLOCKS - 1); - src -= BF_AVX2_PARALLEL_BLOCKS - 1; - dst -= BF_AVX2_PARALLEL_BLOCKS - 1; - - blowfish_cbc_dec_32way(ctx, (u8 *)dst, (u8 *)src); - - nbytes -= bsize; - if (nbytes < bsize) - goto done; - - *dst ^= *(src - 1); - src -= 1; - dst -= 1; - } while (nbytes >= bsize * BF_AVX2_PARALLEL_BLOCKS); - - if (nbytes < bsize) - goto done; - } - - /* Process multi-block batch */ - if (nbytes >= bsize * BF_PARALLEL_BLOCKS) { - u64 ivs[BF_PARALLEL_BLOCKS - 1]; - - do { - nbytes -= bsize * (BF_PARALLEL_BLOCKS - 1); - src -= BF_PARALLEL_BLOCKS - 1; - dst -= BF_PARALLEL_BLOCKS - 1; - - for (i = 0; i < BF_PARALLEL_BLOCKS - 1; i++) - ivs[i] = src[i]; - - blowfish_dec_blk_4way(ctx, (u8 *)dst, (u8 *)src); - - for (i = 0; i < BF_PARALLEL_BLOCKS - 1; i++) - dst[i + 1] ^= ivs[i]; - - nbytes -= bsize; - if (nbytes < bsize) - goto done; - - *dst ^= *(src - 1); - src -= 1; - dst -= 1; - } while (nbytes >= bsize * BF_PARALLEL_BLOCKS); - - if (nbytes < bsize) - goto done; - } - - /* Handle leftovers */ - for (;;) { - blowfish_dec_blk(ctx, (u8 *)dst, (u8 *)src); - - nbytes -= bsize; - if (nbytes < bsize) - break; - - *dst ^= *(src - 1); - src -= 1; - dst -= 1; - } - -done: - *dst ^= *(u64 *)walk->iv; - *(u64 *)walk->iv = last_iv; - - return nbytes; -} - -static int cbc_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - bool fpu_enabled = false; - struct blkcipher_walk walk; - int err; - - blkcipher_walk_init(&walk, dst, src, nbytes); - err = blkcipher_walk_virt(desc, &walk); - desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; - - while ((nbytes = walk.nbytes)) { - fpu_enabled = bf_fpu_begin(fpu_enabled, nbytes); - nbytes = __cbc_decrypt(desc, &walk); - err = blkcipher_walk_done(desc, &walk, nbytes); - } - - bf_fpu_end(fpu_enabled); - return err; -} - -static void ctr_crypt_final(struct blkcipher_desc *desc, - struct blkcipher_walk *walk) -{ - struct bf_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); - u8 *ctrblk = walk->iv; - u8 keystream[BF_BLOCK_SIZE]; - u8 *src = walk->src.virt.addr; - u8 *dst = walk->dst.virt.addr; - unsigned int nbytes = walk->nbytes; - - blowfish_enc_blk(ctx, keystream, ctrblk); - crypto_xor(keystream, src, nbytes); - memcpy(dst, keystream, nbytes); - - crypto_inc(ctrblk, BF_BLOCK_SIZE); -} - -static unsigned int __ctr_crypt(struct blkcipher_desc *desc, - struct blkcipher_walk *walk) -{ - struct bf_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); - unsigned int bsize = BF_BLOCK_SIZE; - unsigned int nbytes = walk->nbytes; - u64 *src = (u64 *)walk->src.virt.addr; - u64 *dst = (u64 *)walk->dst.virt.addr; - int i; - - /* Process multi-block AVX2 batch */ - if (nbytes >= bsize * BF_AVX2_PARALLEL_BLOCKS) { - do { - blowfish_ctr_32way(ctx, (u8 *)dst, (u8 *)src, - (__be64 *)walk->iv); - - src += BF_AVX2_PARALLEL_BLOCKS; - dst += BF_AVX2_PARALLEL_BLOCKS; - nbytes -= bsize * BF_AVX2_PARALLEL_BLOCKS; - } while (nbytes >= bsize * BF_AVX2_PARALLEL_BLOCKS); - - if (nbytes < bsize) - goto done; - } - - /* Process four block batch */ - if (nbytes >= bsize * BF_PARALLEL_BLOCKS) { - __be64 ctrblocks[BF_PARALLEL_BLOCKS]; - u64 ctrblk = be64_to_cpu(*(__be64 *)walk->iv); - - do { - /* create ctrblks for parallel encrypt */ - for (i = 0; i < BF_PARALLEL_BLOCKS; i++) { - if (dst != src) - dst[i] = src[i]; - - ctrblocks[i] = cpu_to_be64(ctrblk++); - } - - blowfish_enc_blk_xor_4way(ctx, (u8 *)dst, - (u8 *)ctrblocks); - - src += BF_PARALLEL_BLOCKS; - dst += BF_PARALLEL_BLOCKS; - nbytes -= bsize * BF_PARALLEL_BLOCKS; - } while (nbytes >= bsize * BF_PARALLEL_BLOCKS); - - *(__be64 *)walk->iv = cpu_to_be64(ctrblk); - - if (nbytes < bsize) - goto done; - } - - /* Handle leftovers */ - do { - u64 ctrblk; - - if (dst != src) - *dst = *src; - - ctrblk = *(u64 *)walk->iv; - be64_add_cpu((__be64 *)walk->iv, 1); - - blowfish_enc_blk_xor(ctx, (u8 *)dst, (u8 *)&ctrblk); - - src += 1; - dst += 1; - } while ((nbytes -= bsize) >= bsize); - -done: - return nbytes; -} - -static int ctr_crypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - bool fpu_enabled = false; - struct blkcipher_walk walk; - int err; - - blkcipher_walk_init(&walk, dst, src, nbytes); - err = blkcipher_walk_virt_block(desc, &walk, BF_BLOCK_SIZE); - desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; - - while ((nbytes = walk.nbytes) >= BF_BLOCK_SIZE) { - fpu_enabled = bf_fpu_begin(fpu_enabled, nbytes); - nbytes = __ctr_crypt(desc, &walk); - err = blkcipher_walk_done(desc, &walk, nbytes); - } - - bf_fpu_end(fpu_enabled); - - if (walk.nbytes) { - ctr_crypt_final(desc, &walk); - err = blkcipher_walk_done(desc, &walk, 0); - } - - return err; -} - -static struct crypto_alg bf_algs[6] = { { - .cra_name = "__ecb-blowfish-avx2", - .cra_driver_name = "__driver-ecb-blowfish-avx2", - .cra_priority = 0, - .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, - .cra_blocksize = BF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct bf_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_blkcipher_type, - .cra_module = THIS_MODULE, - .cra_u = { - .blkcipher = { - .min_keysize = BF_MIN_KEY_SIZE, - .max_keysize = BF_MAX_KEY_SIZE, - .setkey = blowfish_setkey, - .encrypt = ecb_encrypt, - .decrypt = ecb_decrypt, - }, - }, -}, { - .cra_name = "__cbc-blowfish-avx2", - .cra_driver_name = "__driver-cbc-blowfish-avx2", - .cra_priority = 0, - .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, - .cra_blocksize = BF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct bf_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_blkcipher_type, - .cra_module = THIS_MODULE, - .cra_u = { - .blkcipher = { - .min_keysize = BF_MIN_KEY_SIZE, - .max_keysize = BF_MAX_KEY_SIZE, - .setkey = blowfish_setkey, - .encrypt = cbc_encrypt, - .decrypt = cbc_decrypt, - }, - }, -}, { - .cra_name = "__ctr-blowfish-avx2", - .cra_driver_name = "__driver-ctr-blowfish-avx2", - .cra_priority = 0, - .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, - .cra_blocksize = 1, - .cra_ctxsize = sizeof(struct bf_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_blkcipher_type, - .cra_module = THIS_MODULE, - .cra_u = { - .blkcipher = { - .min_keysize = BF_MIN_KEY_SIZE, - .max_keysize = BF_MAX_KEY_SIZE, - .ivsize = BF_BLOCK_SIZE, - .setkey = blowfish_setkey, - .encrypt = ctr_crypt, - .decrypt = ctr_crypt, - }, - }, -}, { - .cra_name = "ecb(blowfish)", - .cra_driver_name = "ecb-blowfish-avx2", - .cra_priority = 400, - .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, - .cra_blocksize = BF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct async_helper_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_ablkcipher_type, - .cra_module = THIS_MODULE, - .cra_init = ablk_init, - .cra_exit = ablk_exit, - .cra_u = { - .ablkcipher = { - .min_keysize = BF_MIN_KEY_SIZE, - .max_keysize = BF_MAX_KEY_SIZE, - .setkey = ablk_set_key, - .encrypt = ablk_encrypt, - .decrypt = ablk_decrypt, - }, - }, -}, { - .cra_name = "cbc(blowfish)", - .cra_driver_name = "cbc-blowfish-avx2", - .cra_priority = 400, - .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, - .cra_blocksize = BF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct async_helper_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_ablkcipher_type, - .cra_module = THIS_MODULE, - .cra_init = ablk_init, - .cra_exit = ablk_exit, - .cra_u = { - .ablkcipher = { - .min_keysize = BF_MIN_KEY_SIZE, - .max_keysize = BF_MAX_KEY_SIZE, - .ivsize = BF_BLOCK_SIZE, - .setkey = ablk_set_key, - .encrypt = __ablk_encrypt, - .decrypt = ablk_decrypt, - }, - }, -}, { - .cra_name = "ctr(blowfish)", - .cra_driver_name = "ctr-blowfish-avx2", - .cra_priority = 400, - .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, - .cra_blocksize = 1, - .cra_ctxsize = sizeof(struct async_helper_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_ablkcipher_type, - .cra_module = THIS_MODULE, - .cra_init = ablk_init, - .cra_exit = ablk_exit, - .cra_u = { - .ablkcipher = { - .min_keysize = BF_MIN_KEY_SIZE, - .max_keysize = BF_MAX_KEY_SIZE, - .ivsize = BF_BLOCK_SIZE, - .setkey = ablk_set_key, - .encrypt = ablk_encrypt, - .decrypt = ablk_encrypt, - .geniv = "chainiv", - }, - }, -} }; - - -static int __init init(void) -{ - u64 xcr0; - - if (!cpu_has_avx2 || !cpu_has_osxsave) { - pr_info("AVX2 instructions are not detected.\n"); - return -ENODEV; - } - - xcr0 = xgetbv(XCR_XFEATURE_ENABLED_MASK); - if ((xcr0 & (XSTATE_SSE | XSTATE_YMM)) != (XSTATE_SSE | XSTATE_YMM)) { - pr_info("AVX detected but unusable.\n"); - return -ENODEV; - } - - return crypto_register_algs(bf_algs, ARRAY_SIZE(bf_algs)); -} - -static void __exit fini(void) -{ - crypto_unregister_algs(bf_algs, ARRAY_SIZE(bf_algs)); -} - -module_init(init); -module_exit(fini); - -MODULE_LICENSE("GPL"); -MODULE_DESCRIPTION("Blowfish Cipher Algorithm, AVX2 optimized"); -MODULE_ALIAS("blowfish"); -MODULE_ALIAS("blowfish-asm"); diff --git a/arch/x86/crypto/blowfish_glue.c b/arch/x86/crypto/blowfish_glue.c index 3548d76dbaa9..50ec333b70e6 100644 --- a/arch/x86/crypto/blowfish_glue.c +++ b/arch/x86/crypto/blowfish_glue.c @@ -1,7 +1,7 @@ /* * Glue Code for assembler optimized version of Blowfish * - * Copyright © 2011-2013 Jussi Kivilinna <jussi.kivilinna@iki.fi> + * Copyright (c) 2011 Jussi Kivilinna <jussi.kivilinna@mbnet.fi> * * CBC & ECB parts based on code (crypto/cbc.c,ecb.c) by: * Copyright (c) 2006 Herbert Xu <herbert@gondor.apana.org.au> @@ -32,24 +32,40 @@ #include <linux/module.h> #include <linux/types.h> #include <crypto/algapi.h> -#include <asm/crypto/blowfish.h> /* regular block cipher functions */ asmlinkage void __blowfish_enc_blk(struct bf_ctx *ctx, u8 *dst, const u8 *src, bool xor); -EXPORT_SYMBOL_GPL(__blowfish_enc_blk); - asmlinkage void blowfish_dec_blk(struct bf_ctx *ctx, u8 *dst, const u8 *src); -EXPORT_SYMBOL_GPL(blowfish_dec_blk); /* 4-way parallel cipher functions */ asmlinkage void __blowfish_enc_blk_4way(struct bf_ctx *ctx, u8 *dst, const u8 *src, bool xor); -EXPORT_SYMBOL_GPL(__blowfish_enc_blk_4way); - asmlinkage void blowfish_dec_blk_4way(struct bf_ctx *ctx, u8 *dst, const u8 *src); -EXPORT_SYMBOL_GPL(blowfish_dec_blk_4way); + +static inline void blowfish_enc_blk(struct bf_ctx *ctx, u8 *dst, const u8 *src) +{ + __blowfish_enc_blk(ctx, dst, src, false); +} + +static inline void blowfish_enc_blk_xor(struct bf_ctx *ctx, u8 *dst, + const u8 *src) +{ + __blowfish_enc_blk(ctx, dst, src, true); +} + +static inline void blowfish_enc_blk_4way(struct bf_ctx *ctx, u8 *dst, + const u8 *src) +{ + __blowfish_enc_blk_4way(ctx, dst, src, false); +} + +static inline void blowfish_enc_blk_xor_4way(struct bf_ctx *ctx, u8 *dst, + const u8 *src) +{ + __blowfish_enc_blk_4way(ctx, dst, src, true); +} static void blowfish_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src) { diff --git a/arch/x86/crypto/camellia-aesni-avx2-asm_64.S b/arch/x86/crypto/camellia-aesni-avx2-asm_64.S index 91a1878fcc3e..0e0b8863a34b 100644 --- a/arch/x86/crypto/camellia-aesni-avx2-asm_64.S +++ b/arch/x86/crypto/camellia-aesni-avx2-asm_64.S @@ -51,16 +51,6 @@ #define ymm14_x xmm14 #define ymm15_x xmm15 -/* - * AES-NI instructions do not support ymmX registers, so we need splitting and - * merging. - */ -#define vaesenclast256(zero, yreg, tmp) \ - vextracti128 $1, yreg, tmp##_x; \ - vaesenclast zero##_x, yreg##_x, yreg##_x; \ - vaesenclast zero##_x, tmp##_x, tmp##_x; \ - vinserti128 $1, tmp##_x, yreg, yreg; - /********************************************************************** 32-way camellia **********************************************************************/ @@ -79,46 +69,70 @@ * S-function with AES subbytes \ */ \ vbroadcasti128 .Linv_shift_row, t4; \ - vpbroadcastb .L0f0f0f0f, t7; \ - vbroadcasti128 .Lpre_tf_lo_s1, t0; \ - vbroadcasti128 .Lpre_tf_hi_s1, t1; \ + vpbroadcastd .L0f0f0f0f, t7; \ + vbroadcasti128 .Lpre_tf_lo_s1, t5; \ + vbroadcasti128 .Lpre_tf_hi_s1, t6; \ + vbroadcasti128 .Lpre_tf_lo_s4, t2; \ + vbroadcasti128 .Lpre_tf_hi_s4, t3; \ \ /* AES inverse shift rows */ \ vpshufb t4, x0, x0; \ vpshufb t4, x7, x7; \ - vpshufb t4, x1, x1; \ - vpshufb t4, x4, x4; \ - vpshufb t4, x2, x2; \ - vpshufb t4, x5, x5; \ vpshufb t4, x3, x3; \ vpshufb t4, x6, x6; \ + vpshufb t4, x2, x2; \ + vpshufb t4, x5, x5; \ + vpshufb t4, x1, x1; \ + vpshufb t4, x4, x4; \ \ /* prefilter sboxes 1, 2 and 3 */ \ - vbroadcasti128 .Lpre_tf_lo_s4, t2; \ - vbroadcasti128 .Lpre_tf_hi_s4, t3; \ - filter_8bit(x0, t0, t1, t7, t6); \ - filter_8bit(x7, t0, t1, t7, t6); \ - filter_8bit(x1, t0, t1, t7, t6); \ - filter_8bit(x4, t0, t1, t7, t6); \ - filter_8bit(x2, t0, t1, t7, t6); \ - filter_8bit(x5, t0, t1, t7, t6); \ - \ /* prefilter sbox 4 */ \ + filter_8bit(x0, t5, t6, t7, t4); \ + filter_8bit(x7, t5, t6, t7, t4); \ + vextracti128 $1, x0, t0##_x; \ + vextracti128 $1, x7, t1##_x; \ + filter_8bit(x3, t2, t3, t7, t4); \ + filter_8bit(x6, t2, t3, t7, t4); \ + vextracti128 $1, x3, t3##_x; \ + vextracti128 $1, x6, t2##_x; \ + filter_8bit(x2, t5, t6, t7, t4); \ + filter_8bit(x5, t5, t6, t7, t4); \ + filter_8bit(x1, t5, t6, t7, t4); \ + filter_8bit(x4, t5, t6, t7, t4); \ + \ vpxor t4##_x, t4##_x, t4##_x; \ - filter_8bit(x3, t2, t3, t7, t6); \ - filter_8bit(x6, t2, t3, t7, t6); \ \ /* AES subbytes + AES shift rows */ \ + vextracti128 $1, x2, t6##_x; \ + vextracti128 $1, x5, t5##_x; \ + vaesenclast t4##_x, x0##_x, x0##_x; \ + vaesenclast t4##_x, t0##_x, t0##_x; \ + vinserti128 $1, t0##_x, x0, x0; \ + vaesenclast t4##_x, x7##_x, x7##_x; \ + vaesenclast t4##_x, t1##_x, t1##_x; \ + vinserti128 $1, t1##_x, x7, x7; \ + vaesenclast t4##_x, x3##_x, x3##_x; \ + vaesenclast t4##_x, t3##_x, t3##_x; \ + vinserti128 $1, t3##_x, x3, x3; \ + vaesenclast t4##_x, x6##_x, x6##_x; \ + vaesenclast t4##_x, t2##_x, t2##_x; \ + vinserti128 $1, t2##_x, x6, x6; \ + vextracti128 $1, x1, t3##_x; \ + vextracti128 $1, x4, t2##_x; \ vbroadcasti128 .Lpost_tf_lo_s1, t0; \ vbroadcasti128 .Lpost_tf_hi_s1, t1; \ - vaesenclast256(t4, x0, t5); \ - vaesenclast256(t4, x7, t5); \ - vaesenclast256(t4, x1, t5); \ - vaesenclast256(t4, x4, t5); \ - vaesenclast256(t4, x2, t5); \ - vaesenclast256(t4, x5, t5); \ - vaesenclast256(t4, x3, t5); \ - vaesenclast256(t4, x6, t5); \ + vaesenclast t4##_x, x2##_x, x2##_x; \ + vaesenclast t4##_x, t6##_x, t6##_x; \ + vinserti128 $1, t6##_x, x2, x2; \ + vaesenclast t4##_x, x5##_x, x5##_x; \ + vaesenclast t4##_x, t5##_x, t5##_x; \ + vinserti128 $1, t5##_x, x5, x5; \ + vaesenclast t4##_x, x1##_x, x1##_x; \ + vaesenclast t4##_x, t3##_x, t3##_x; \ + vinserti128 $1, t3##_x, x1, x1; \ + vaesenclast t4##_x, x4##_x, x4##_x; \ + vaesenclast t4##_x, t2##_x, t2##_x; \ + vinserti128 $1, t2##_x, x4, x4; \ \ /* postfilter sboxes 1 and 4 */ \ vbroadcasti128 .Lpost_tf_lo_s3, t2; \ @@ -139,22 +153,12 @@ /* postfilter sbox 2 */ \ filter_8bit(x1, t4, t5, t7, t2); \ filter_8bit(x4, t4, t5, t7, t2); \ + vpxor t7, t7, t7; \ \ vpsrldq $1, t0, t1; \ vpsrldq $2, t0, t2; \ + vpshufb t7, t1, t1; \ vpsrldq $3, t0, t3; \ - vpsrldq $4, t0, t4; \ - vpsrldq $5, t0, t5; \ - vpsrldq $6, t0, t6; \ - vpsrldq $7, t0, t7; \ - vpbroadcastb t0##_x, t0; \ - vpbroadcastb t1##_x, t1; \ - vpbroadcastb t2##_x, t2; \ - vpbroadcastb t3##_x, t3; \ - vpbroadcastb t4##_x, t4; \ - vpbroadcastb t6##_x, t6; \ - vpbroadcastb t5##_x, t5; \ - vpbroadcastb t7##_x, t7; \ \ /* P-function */ \ vpxor x5, x0, x0; \ @@ -162,11 +166,21 @@ vpxor x7, x2, x2; \ vpxor x4, x3, x3; \ \ + vpshufb t7, t2, t2; \ + vpsrldq $4, t0, t4; \ + vpshufb t7, t3, t3; \ + vpsrldq $5, t0, t5; \ + vpshufb t7, t4, t4; \ + \ vpxor x2, x4, x4; \ vpxor x3, x5, x5; \ vpxor x0, x6, x6; \ vpxor x1, x7, x7; \ \ + vpsrldq $6, t0, t6; \ + vpshufb t7, t5, t5; \ + vpshufb t7, t6, t6; \ + \ vpxor x7, x0, x0; \ vpxor x4, x1, x1; \ vpxor x5, x2, x2; \ @@ -179,12 +193,16 @@ \ /* Add key material and result to CD (x becomes new CD) */ \ \ - vpxor t7, x0, x0; \ - vpxor 4 * 32(mem_cd), x0, x0; \ - \ vpxor t6, x1, x1; \ vpxor 5 * 32(mem_cd), x1, x1; \ \ + vpsrldq $7, t0, t6; \ + vpshufb t7, t0, t0; \ + vpshufb t7, t6, t7; \ + \ + vpxor t7, x0, x0; \ + vpxor 4 * 32(mem_cd), x0, x0; \ + \ vpxor t5, x2, x2; \ vpxor 6 * 32(mem_cd), x2, x2; \ \ @@ -204,7 +222,7 @@ vpxor 3 * 32(mem_cd), x7, x7; /* - * Size optimization... with inlined roundsm16 binary would be over 5 times + * Size optimization... with inlined roundsm32 binary would be over 5 times * larger and would only marginally faster. */ .align 8 @@ -324,13 +342,13 @@ ENDPROC(roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab) */ \ vpbroadcastd kll, t0; /* only lowest 32-bit used */ \ vpxor tt0, tt0, tt0; \ - vpbroadcastb t0##_x, t3; \ + vpshufb tt0, t0, t3; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t2; \ + vpshufb tt0, t0, t2; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t1; \ + vpshufb tt0, t0, t1; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t0; \ + vpshufb tt0, t0, t0; \ \ vpand l0, t0, t0; \ vpand l1, t1, t1; \ @@ -340,6 +358,7 @@ ENDPROC(roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab) rol32_1_32(t3, t2, t1, t0, tt1, tt2, tt3, tt0); \ \ vpxor l4, t0, l4; \ + vpbroadcastd krr, t0; /* only lowest 32-bit used */ \ vmovdqu l4, 4 * 32(l); \ vpxor l5, t1, l5; \ vmovdqu l5, 5 * 32(l); \ @@ -354,14 +373,13 @@ ENDPROC(roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab) * rl ^= t2; \ */ \ \ - vpbroadcastd krr, t0; /* only lowest 32-bit used */ \ - vpbroadcastb t0##_x, t3; \ + vpshufb tt0, t0, t3; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t2; \ + vpshufb tt0, t0, t2; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t1; \ + vpshufb tt0, t0, t1; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t0; \ + vpshufb tt0, t0, t0; \ \ vpor 4 * 32(r), t0, t0; \ vpor 5 * 32(r), t1, t1; \ @@ -373,6 +391,7 @@ ENDPROC(roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab) vpxor 2 * 32(r), t2, t2; \ vpxor 3 * 32(r), t3, t3; \ vmovdqu t0, 0 * 32(r); \ + vpbroadcastd krl, t0; /* only lowest 32-bit used */ \ vmovdqu t1, 1 * 32(r); \ vmovdqu t2, 2 * 32(r); \ vmovdqu t3, 3 * 32(r); \ @@ -382,14 +401,13 @@ ENDPROC(roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab) * t2 &= rl; \ * rr ^= rol32(t2, 1); \ */ \ - vpbroadcastd krl, t0; /* only lowest 32-bit used */ \ - vpbroadcastb t0##_x, t3; \ + vpshufb tt0, t0, t3; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t2; \ + vpshufb tt0, t0, t2; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t1; \ + vpshufb tt0, t0, t1; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t0; \ + vpshufb tt0, t0, t0; \ \ vpand 0 * 32(r), t0, t0; \ vpand 1 * 32(r), t1, t1; \ @@ -403,6 +421,7 @@ ENDPROC(roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab) vpxor 6 * 32(r), t2, t2; \ vpxor 7 * 32(r), t3, t3; \ vmovdqu t0, 4 * 32(r); \ + vpbroadcastd klr, t0; /* only lowest 32-bit used */ \ vmovdqu t1, 5 * 32(r); \ vmovdqu t2, 6 * 32(r); \ vmovdqu t3, 7 * 32(r); \ @@ -413,14 +432,13 @@ ENDPROC(roundsm32_x4_x5_x6_x7_x0_x1_x2_x3_y4_y5_y6_y7_y0_y1_y2_y3_ab) * ll ^= t0; \ */ \ \ - vpbroadcastd klr, t0; /* only lowest 32-bit used */ \ - vpbroadcastb t0##_x, t3; \ + vpshufb tt0, t0, t3; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t2; \ + vpshufb tt0, t0, t2; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t1; \ + vpshufb tt0, t0, t1; \ vpsrldq $1, t0, t0; \ - vpbroadcastb t0##_x, t0; \ + vpshufb tt0, t0, t0; \ \ vpor l4, t0, t0; \ vpor l5, t1, t1; \ diff --git a/arch/x86/crypto/crct10dif-pcl-asm_64.S b/arch/x86/crypto/crct10dif-pcl-asm_64.S new file mode 100644 index 000000000000..35e97569d05f --- /dev/null +++ b/arch/x86/crypto/crct10dif-pcl-asm_64.S @@ -0,0 +1,643 @@ +######################################################################## +# Implement fast CRC-T10DIF computation with SSE and PCLMULQDQ instructions +# +# Copyright (c) 2013, Intel Corporation +# +# Authors: +# Erdinc Ozturk <erdinc.ozturk@intel.com> +# Vinodh Gopal <vinodh.gopal@intel.com> +# James Guilford <james.guilford@intel.com> +# Tim Chen <tim.c.chen@linux.intel.com> +# +# This software is available to you under a choice of one of two +# licenses. You may choose to be licensed under the terms of the GNU +# General Public License (GPL) Version 2, available from the file +# COPYING in the main directory of this source tree, or the +# OpenIB.org BSD license below: +# +# Redistribution and use in source and binary forms, with or without +# modification, are permitted provided that the following conditions are +# met: +# +# * Redistributions of source code must retain the above copyright +# notice, this list of conditions and the following disclaimer. +# +# * Redistributions in binary form must reproduce the above copyright +# notice, this list of conditions and the following disclaimer in the +# documentation and/or other materials provided with the +# distribution. +# +# * Neither the name of the Intel Corporation nor the names of its +# contributors may be used to endorse or promote products derived from +# this software without specific prior written permission. +# +# +# THIS SOFTWARE IS PROVIDED BY INTEL CORPORATION ""AS IS"" AND ANY +# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE +# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR +# PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL CORPORATION OR +# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, +# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING +# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +######################################################################## +# Function API: +# UINT16 crc_t10dif_pcl( +# UINT16 init_crc, //initial CRC value, 16 bits +# const unsigned char *buf, //buffer pointer to calculate CRC on +# UINT64 len //buffer length in bytes (64-bit data) +# ); +# +# Reference paper titled "Fast CRC Computation for Generic +# Polynomials Using PCLMULQDQ Instruction" +# URL: http://www.intel.com/content/dam/www/public/us/en/documents +# /white-papers/fast-crc-computation-generic-polynomials-pclmulqdq-paper.pdf +# +# + +#include <linux/linkage.h> + +.text + +#define arg1 %rdi +#define arg2 %rsi +#define arg3 %rdx + +#define arg1_low32 %edi + +ENTRY(crc_t10dif_pcl) +.align 16 + + # adjust the 16-bit initial_crc value, scale it to 32 bits + shl $16, arg1_low32 + + # Allocate Stack Space + mov %rsp, %rcx + sub $16*2, %rsp + # align stack to 16 byte boundary + and $~(0x10 - 1), %rsp + + # check if smaller than 256 + cmp $256, arg3 + + # for sizes less than 128, we can't fold 64B at a time... + jl _less_than_128 + + + # load the initial crc value + movd arg1_low32, %xmm10 # initial crc + + # crc value does not need to be byte-reflected, but it needs + # to be moved to the high part of the register. + # because data will be byte-reflected and will align with + # initial crc at correct place. + pslldq $12, %xmm10 + + movdqa SHUF_MASK(%rip), %xmm11 + # receive the initial 64B data, xor the initial crc value + movdqu 16*0(arg2), %xmm0 + movdqu 16*1(arg2), %xmm1 + movdqu 16*2(arg2), %xmm2 + movdqu 16*3(arg2), %xmm3 + movdqu 16*4(arg2), %xmm4 + movdqu 16*5(arg2), %xmm5 + movdqu 16*6(arg2), %xmm6 + movdqu 16*7(arg2), %xmm7 + + pshufb %xmm11, %xmm0 + # XOR the initial_crc value + pxor %xmm10, %xmm0 + pshufb %xmm11, %xmm1 + pshufb %xmm11, %xmm2 + pshufb %xmm11, %xmm3 + pshufb %xmm11, %xmm4 + pshufb %xmm11, %xmm5 + pshufb %xmm11, %xmm6 + pshufb %xmm11, %xmm7 + + movdqa rk3(%rip), %xmm10 #xmm10 has rk3 and rk4 + #imm value of pclmulqdq instruction + #will determine which constant to use + + ################################################################# + # we subtract 256 instead of 128 to save one instruction from the loop + sub $256, arg3 + + # at this section of the code, there is 64*x+y (0<=y<64) bytes of + # buffer. The _fold_64_B_loop will fold 64B at a time + # until we have 64+y Bytes of buffer + + + # fold 64B at a time. This section of the code folds 4 xmm + # registers in parallel +_fold_64_B_loop: + + # update the buffer pointer + add $128, arg2 # buf += 64# + + movdqu 16*0(arg2), %xmm9 + movdqu 16*1(arg2), %xmm12 + pshufb %xmm11, %xmm9 + pshufb %xmm11, %xmm12 + movdqa %xmm0, %xmm8 + movdqa %xmm1, %xmm13 + pclmulqdq $0x0 , %xmm10, %xmm0 + pclmulqdq $0x11, %xmm10, %xmm8 + pclmulqdq $0x0 , %xmm10, %xmm1 + pclmulqdq $0x11, %xmm10, %xmm13 + pxor %xmm9 , %xmm0 + xorps %xmm8 , %xmm0 + pxor %xmm12, %xmm1 + xorps %xmm13, %xmm1 + + movdqu 16*2(arg2), %xmm9 + movdqu 16*3(arg2), %xmm12 + pshufb %xmm11, %xmm9 + pshufb %xmm11, %xmm12 + movdqa %xmm2, %xmm8 + movdqa %xmm3, %xmm13 + pclmulqdq $0x0, %xmm10, %xmm2 + pclmulqdq $0x11, %xmm10, %xmm8 + pclmulqdq $0x0, %xmm10, %xmm3 + pclmulqdq $0x11, %xmm10, %xmm13 + pxor %xmm9 , %xmm2 + xorps %xmm8 , %xmm2 + pxor %xmm12, %xmm3 + xorps %xmm13, %xmm3 + + movdqu 16*4(arg2), %xmm9 + movdqu 16*5(arg2), %xmm12 + pshufb %xmm11, %xmm9 + pshufb %xmm11, %xmm12 + movdqa %xmm4, %xmm8 + movdqa %xmm5, %xmm13 + pclmulqdq $0x0, %xmm10, %xmm4 + pclmulqdq $0x11, %xmm10, %xmm8 + pclmulqdq $0x0, %xmm10, %xmm5 + pclmulqdq $0x11, %xmm10, %xmm13 + pxor %xmm9 , %xmm4 + xorps %xmm8 , %xmm4 + pxor %xmm12, %xmm5 + xorps %xmm13, %xmm5 + + movdqu 16*6(arg2), %xmm9 + movdqu 16*7(arg2), %xmm12 + pshufb %xmm11, %xmm9 + pshufb %xmm11, %xmm12 + movdqa %xmm6 , %xmm8 + movdqa %xmm7 , %xmm13 + pclmulqdq $0x0 , %xmm10, %xmm6 + pclmulqdq $0x11, %xmm10, %xmm8 + pclmulqdq $0x0 , %xmm10, %xmm7 + pclmulqdq $0x11, %xmm10, %xmm13 + pxor %xmm9 , %xmm6 + xorps %xmm8 , %xmm6 + pxor %xmm12, %xmm7 + xorps %xmm13, %xmm7 + + sub $128, arg3 + + # check if there is another 64B in the buffer to be able to fold + jge _fold_64_B_loop + ################################################################## + + + add $128, arg2 + # at this point, the buffer pointer is pointing at the last y Bytes + # of the buffer the 64B of folded data is in 4 of the xmm + # registers: xmm0, xmm1, xmm2, xmm3 + + + # fold the 8 xmm registers to 1 xmm register with different constants + + movdqa rk9(%rip), %xmm10 + movdqa %xmm0, %xmm8 + pclmulqdq $0x11, %xmm10, %xmm0 + pclmulqdq $0x0 , %xmm10, %xmm8 + pxor %xmm8, %xmm7 + xorps %xmm0, %xmm7 + + movdqa rk11(%rip), %xmm10 + movdqa %xmm1, %xmm8 + pclmulqdq $0x11, %xmm10, %xmm1 + pclmulqdq $0x0 , %xmm10, %xmm8 + pxor %xmm8, %xmm7 + xorps %xmm1, %xmm7 + + movdqa rk13(%rip), %xmm10 + movdqa %xmm2, %xmm8 + pclmulqdq $0x11, %xmm10, %xmm2 + pclmulqdq $0x0 , %xmm10, %xmm8 + pxor %xmm8, %xmm7 + pxor %xmm2, %xmm7 + + movdqa rk15(%rip), %xmm10 + movdqa %xmm3, %xmm8 + pclmulqdq $0x11, %xmm10, %xmm3 + pclmulqdq $0x0 , %xmm10, %xmm8 + pxor %xmm8, %xmm7 + xorps %xmm3, %xmm7 + + movdqa rk17(%rip), %xmm10 + movdqa %xmm4, %xmm8 + pclmulqdq $0x11, %xmm10, %xmm4 + pclmulqdq $0x0 , %xmm10, %xmm8 + pxor %xmm8, %xmm7 + pxor %xmm4, %xmm7 + + movdqa rk19(%rip), %xmm10 + movdqa %xmm5, %xmm8 + pclmulqdq $0x11, %xmm10, %xmm5 + pclmulqdq $0x0 , %xmm10, %xmm8 + pxor %xmm8, %xmm7 + xorps %xmm5, %xmm7 + + movdqa rk1(%rip), %xmm10 #xmm10 has rk1 and rk2 + #imm value of pclmulqdq instruction + #will determine which constant to use + movdqa %xmm6, %xmm8 + pclmulqdq $0x11, %xmm10, %xmm6 + pclmulqdq $0x0 , %xmm10, %xmm8 + pxor %xmm8, %xmm7 + pxor %xmm6, %xmm7 + + + # instead of 64, we add 48 to the loop counter to save 1 instruction + # from the loop instead of a cmp instruction, we use the negative + # flag with the jl instruction + add $128-16, arg3 + jl _final_reduction_for_128 + + # now we have 16+y bytes left to reduce. 16 Bytes is in register xmm7 + # and the rest is in memory. We can fold 16 bytes at a time if y>=16 + # continue folding 16B at a time + +_16B_reduction_loop: + movdqa %xmm7, %xmm8 + pclmulqdq $0x11, %xmm10, %xmm7 + pclmulqdq $0x0 , %xmm10, %xmm8 + pxor %xmm8, %xmm7 + movdqu (arg2), %xmm0 + pshufb %xmm11, %xmm0 + pxor %xmm0 , %xmm7 + add $16, arg2 + sub $16, arg3 + # instead of a cmp instruction, we utilize the flags with the + # jge instruction equivalent of: cmp arg3, 16-16 + # check if there is any more 16B in the buffer to be able to fold + jge _16B_reduction_loop + + #now we have 16+z bytes left to reduce, where 0<= z < 16. + #first, we reduce the data in the xmm7 register + + +_final_reduction_for_128: + # check if any more data to fold. If not, compute the CRC of + # the final 128 bits + add $16, arg3 + je _128_done + + # here we are getting data that is less than 16 bytes. + # since we know that there was data before the pointer, we can + # offset the input pointer before the actual point, to receive + # exactly 16 bytes. after that the registers need to be adjusted. +_get_last_two_xmms: + movdqa %xmm7, %xmm2 + + movdqu -16(arg2, arg3), %xmm1 + pshufb %xmm11, %xmm1 + + # get rid of the extra data that was loaded before + # load the shift constant + lea pshufb_shf_table+16(%rip), %rax + sub arg3, %rax + movdqu (%rax), %xmm0 + + # shift xmm2 to the left by arg3 bytes + pshufb %xmm0, %xmm2 + + # shift xmm7 to the right by 16-arg3 bytes + pxor mask1(%rip), %xmm0 + pshufb %xmm0, %xmm7 + pblendvb %xmm2, %xmm1 #xmm0 is implicit + + # fold 16 Bytes + movdqa %xmm1, %xmm2 + movdqa %xmm7, %xmm8 + pclmulqdq $0x11, %xmm10, %xmm7 + pclmulqdq $0x0 , %xmm10, %xmm8 + pxor %xmm8, %xmm7 + pxor %xmm2, %xmm7 + +_128_done: + # compute crc of a 128-bit value + movdqa rk5(%rip), %xmm10 # rk5 and rk6 in xmm10 + movdqa %xmm7, %xmm0 + + #64b fold + pclmulqdq $0x1, %xmm10, %xmm7 + pslldq $8 , %xmm0 + pxor %xmm0, %xmm7 + + #32b fold + movdqa %xmm7, %xmm0 + + pand mask2(%rip), %xmm0 + + psrldq $12, %xmm7 + pclmulqdq $0x10, %xmm10, %xmm7 + pxor %xmm0, %xmm7 + + #barrett reduction +_barrett: + movdqa rk7(%rip), %xmm10 # rk7 and rk8 in xmm10 + movdqa %xmm7, %xmm0 + pclmulqdq $0x01, %xmm10, %xmm7 + pslldq $4, %xmm7 + pclmulqdq $0x11, %xmm10, %xmm7 + + pslldq $4, %xmm7 + pxor %xmm0, %xmm7 + pextrd $1, %xmm7, %eax + +_cleanup: + # scale the result back to 16 bits + shr $16, %eax + mov %rcx, %rsp + ret + +######################################################################## + +.align 16 +_less_than_128: + + # check if there is enough buffer to be able to fold 16B at a time + cmp $32, arg3 + jl _less_than_32 + movdqa SHUF_MASK(%rip), %xmm11 + + # now if there is, load the constants + movdqa rk1(%rip), %xmm10 # rk1 and rk2 in xmm10 + + movd arg1_low32, %xmm0 # get the initial crc value + pslldq $12, %xmm0 # align it to its correct place + movdqu (arg2), %xmm7 # load the plaintext + pshufb %xmm11, %xmm7 # byte-reflect the plaintext + pxor %xmm0, %xmm7 + + + # update the buffer pointer + add $16, arg2 + + # update the counter. subtract 32 instead of 16 to save one + # instruction from the loop + sub $32, arg3 + + jmp _16B_reduction_loop + + +.align 16 +_less_than_32: + # mov initial crc to the return value. this is necessary for + # zero-length buffers. + mov arg1_low32, %eax + test arg3, arg3 + je _cleanup + + movdqa SHUF_MASK(%rip), %xmm11 + + movd arg1_low32, %xmm0 # get the initial crc value + pslldq $12, %xmm0 # align it to its correct place + + cmp $16, arg3 + je _exact_16_left + jl _less_than_16_left + + movdqu (arg2), %xmm7 # load the plaintext + pshufb %xmm11, %xmm7 # byte-reflect the plaintext + pxor %xmm0 , %xmm7 # xor the initial crc value + add $16, arg2 + sub $16, arg3 + movdqa rk1(%rip), %xmm10 # rk1 and rk2 in xmm10 + jmp _get_last_two_xmms + + +.align 16 +_less_than_16_left: + # use stack space to load data less than 16 bytes, zero-out + # the 16B in memory first. + + pxor %xmm1, %xmm1 + mov %rsp, %r11 + movdqa %xmm1, (%r11) + + cmp $4, arg3 + jl _only_less_than_4 + + # backup the counter value + mov arg3, %r9 + cmp $8, arg3 + jl _less_than_8_left + + # load 8 Bytes + mov (arg2), %rax + mov %rax, (%r11) + add $8, %r11 + sub $8, arg3 + add $8, arg2 +_less_than_8_left: + + cmp $4, arg3 + jl _less_than_4_left + + # load 4 Bytes + mov (arg2), %eax + mov %eax, (%r11) + add $4, %r11 + sub $4, arg3 + add $4, arg2 +_less_than_4_left: + + cmp $2, arg3 + jl _less_than_2_left + + # load 2 Bytes + mov (arg2), %ax + mov %ax, (%r11) + add $2, %r11 + sub $2, arg3 + add $2, arg2 +_less_than_2_left: + cmp $1, arg3 + jl _zero_left + + # load 1 Byte + mov (arg2), %al + mov %al, (%r11) +_zero_left: + movdqa (%rsp), %xmm7 + pshufb %xmm11, %xmm7 + pxor %xmm0 , %xmm7 # xor the initial crc value + + # shl r9, 4 + lea pshufb_shf_table+16(%rip), %rax + sub %r9, %rax + movdqu (%rax), %xmm0 + pxor mask1(%rip), %xmm0 + + pshufb %xmm0, %xmm7 + jmp _128_done + +.align 16 +_exact_16_left: + movdqu (arg2), %xmm7 + pshufb %xmm11, %xmm7 + pxor %xmm0 , %xmm7 # xor the initial crc value + + jmp _128_done + +_only_less_than_4: + cmp $3, arg3 + jl _only_less_than_3 + + # load 3 Bytes + mov (arg2), %al + mov %al, (%r11) + + mov 1(arg2), %al + mov %al, 1(%r11) + + mov 2(arg2), %al + mov %al, 2(%r11) + + movdqa (%rsp), %xmm7 + pshufb %xmm11, %xmm7 + pxor %xmm0 , %xmm7 # xor the initial crc value + + psrldq $5, %xmm7 + + jmp _barrett +_only_less_than_3: + cmp $2, arg3 + jl _only_less_than_2 + + # load 2 Bytes + mov (arg2), %al + mov %al, (%r11) + + mov 1(arg2), %al + mov %al, 1(%r11) + + movdqa (%rsp), %xmm7 + pshufb %xmm11, %xmm7 + pxor %xmm0 , %xmm7 # xor the initial crc value + + psrldq $6, %xmm7 + + jmp _barrett +_only_less_than_2: + + # load 1 Byte + mov (arg2), %al + mov %al, (%r11) + + movdqa (%rsp), %xmm7 + pshufb %xmm11, %xmm7 + pxor %xmm0 , %xmm7 # xor the initial crc value + + psrldq $7, %xmm7 + + jmp _barrett + +ENDPROC(crc_t10dif_pcl) + +.data + +# precomputed constants +# these constants are precomputed from the poly: +# 0x8bb70000 (0x8bb7 scaled to 32 bits) +.align 16 +# Q = 0x18BB70000 +# rk1 = 2^(32*3) mod Q << 32 +# rk2 = 2^(32*5) mod Q << 32 +# rk3 = 2^(32*15) mod Q << 32 +# rk4 = 2^(32*17) mod Q << 32 +# rk5 = 2^(32*3) mod Q << 32 +# rk6 = 2^(32*2) mod Q << 32 +# rk7 = floor(2^64/Q) +# rk8 = Q +rk1: +.quad 0x2d56000000000000 +rk2: +.quad 0x06df000000000000 +rk3: +.quad 0x9d9d000000000000 +rk4: +.quad 0x7cf5000000000000 +rk5: +.quad 0x2d56000000000000 +rk6: +.quad 0x1368000000000000 +rk7: +.quad 0x00000001f65a57f8 +rk8: +.quad 0x000000018bb70000 + +rk9: +.quad 0xceae000000000000 +rk10: +.quad 0xbfd6000000000000 +rk11: +.quad 0x1e16000000000000 +rk12: +.quad 0x713c000000000000 +rk13: +.quad 0xf7f9000000000000 +rk14: +.quad 0x80a6000000000000 +rk15: +.quad 0x044c000000000000 +rk16: +.quad 0xe658000000000000 +rk17: +.quad 0xad18000000000000 +rk18: +.quad 0xa497000000000000 +rk19: +.quad 0x6ee3000000000000 +rk20: +.quad 0xe7b5000000000000 + + + +mask1: +.octa 0x80808080808080808080808080808080 +mask2: +.octa 0x00000000FFFFFFFFFFFFFFFFFFFFFFFF + +SHUF_MASK: +.octa 0x000102030405060708090A0B0C0D0E0F + +pshufb_shf_table: +# use these values for shift constants for the pshufb instruction +# different alignments result in values as shown: +# DDQ 0x008f8e8d8c8b8a898887868584838281 # shl 15 (16-1) / shr1 +# DDQ 0x01008f8e8d8c8b8a8988878685848382 # shl 14 (16-3) / shr2 +# DDQ 0x0201008f8e8d8c8b8a89888786858483 # shl 13 (16-4) / shr3 +# DDQ 0x030201008f8e8d8c8b8a898887868584 # shl 12 (16-4) / shr4 +# DDQ 0x04030201008f8e8d8c8b8a8988878685 # shl 11 (16-5) / shr5 +# DDQ 0x0504030201008f8e8d8c8b8a89888786 # shl 10 (16-6) / shr6 +# DDQ 0x060504030201008f8e8d8c8b8a898887 # shl 9 (16-7) / shr7 +# DDQ 0x07060504030201008f8e8d8c8b8a8988 # shl 8 (16-8) / shr8 +# DDQ 0x0807060504030201008f8e8d8c8b8a89 # shl 7 (16-9) / shr9 +# DDQ 0x090807060504030201008f8e8d8c8b8a # shl 6 (16-10) / shr10 +# DDQ 0x0a090807060504030201008f8e8d8c8b # shl 5 (16-11) / shr11 +# DDQ 0x0b0a090807060504030201008f8e8d8c # shl 4 (16-12) / shr12 +# DDQ 0x0c0b0a090807060504030201008f8e8d # shl 3 (16-13) / shr13 +# DDQ 0x0d0c0b0a090807060504030201008f8e # shl 2 (16-14) / shr14 +# DDQ 0x0e0d0c0b0a090807060504030201008f # shl 1 (16-15) / shr15 +.octa 0x8f8e8d8c8b8a89888786858483828100 +.octa 0x000e0d0c0b0a09080706050403020100 diff --git a/arch/x86/crypto/crct10dif-pclmul_glue.c b/arch/x86/crypto/crct10dif-pclmul_glue.c new file mode 100644 index 000000000000..7845d7fd54c0 --- /dev/null +++ b/arch/x86/crypto/crct10dif-pclmul_glue.c @@ -0,0 +1,151 @@ +/* + * Cryptographic API. + * + * T10 Data Integrity Field CRC16 Crypto Transform using PCLMULQDQ Instructions + * + * Copyright (C) 2013 Intel Corporation + * Author: Tim Chen <tim.c.chen@linux.intel.com> + * + * 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. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + */ + +#include <linux/types.h> +#include <linux/module.h> +#include <linux/crc-t10dif.h> +#include <crypto/internal/hash.h> +#include <linux/init.h> +#include <linux/string.h> +#include <linux/kernel.h> +#include <asm/i387.h> +#include <asm/cpufeature.h> +#include <asm/cpu_device_id.h> + +asmlinkage __u16 crc_t10dif_pcl(__u16 crc, const unsigned char *buf, + size_t len); + +struct chksum_desc_ctx { + __u16 crc; +}; + +/* + * Steps through buffer one byte at at time, calculates reflected + * crc using table. + */ + +static int chksum_init(struct shash_desc *desc) +{ + struct chksum_desc_ctx *ctx = shash_desc_ctx(desc); + + ctx->crc = 0; + + return 0; +} + +static int chksum_update(struct shash_desc *desc, const u8 *data, + unsigned int length) +{ + struct chksum_desc_ctx *ctx = shash_desc_ctx(desc); + + if (irq_fpu_usable()) { + kernel_fpu_begin(); + ctx->crc = crc_t10dif_pcl(ctx->crc, data, length); + kernel_fpu_end(); + } else + ctx->crc = crc_t10dif_generic(ctx->crc, data, length); + return 0; +} + +static int chksum_final(struct shash_desc *desc, u8 *out) +{ + struct chksum_desc_ctx *ctx = shash_desc_ctx(desc); + + *(__u16 *)out = ctx->crc; + return 0; +} + +static int __chksum_finup(__u16 *crcp, const u8 *data, unsigned int len, + u8 *out) +{ + if (irq_fpu_usable()) { + kernel_fpu_begin(); + *(__u16 *)out = crc_t10dif_pcl(*crcp, data, len); + kernel_fpu_end(); + } else + *(__u16 *)out = crc_t10dif_generic(*crcp, data, len); + return 0; +} + +static int chksum_finup(struct shash_desc *desc, const u8 *data, + unsigned int len, u8 *out) +{ + struct chksum_desc_ctx *ctx = shash_desc_ctx(desc); + + return __chksum_finup(&ctx->crc, data, len, out); +} + +static int chksum_digest(struct shash_desc *desc, const u8 *data, + unsigned int length, u8 *out) +{ + struct chksum_desc_ctx *ctx = shash_desc_ctx(desc); + + return __chksum_finup(&ctx->crc, data, length, out); +} + +static struct shash_alg alg = { + .digestsize = CRC_T10DIF_DIGEST_SIZE, + .init = chksum_init, + .update = chksum_update, + .final = chksum_final, + .finup = chksum_finup, + .digest = chksum_digest, + .descsize = sizeof(struct chksum_desc_ctx), + .base = { + .cra_name = "crct10dif", + .cra_driver_name = "crct10dif-pclmul", + .cra_priority = 200, + .cra_blocksize = CRC_T10DIF_BLOCK_SIZE, + .cra_module = THIS_MODULE, + } +}; + +static const struct x86_cpu_id crct10dif_cpu_id[] = { + X86_FEATURE_MATCH(X86_FEATURE_PCLMULQDQ), + {} +}; +MODULE_DEVICE_TABLE(x86cpu, crct10dif_cpu_id); + +static int __init crct10dif_intel_mod_init(void) +{ + if (!x86_match_cpu(crct10dif_cpu_id)) + return -ENODEV; + + return crypto_register_shash(&alg); +} + +static void __exit crct10dif_intel_mod_fini(void) +{ + crypto_unregister_shash(&alg); +} + +module_init(crct10dif_intel_mod_init); +module_exit(crct10dif_intel_mod_fini); + +MODULE_AUTHOR("Tim Chen <tim.c.chen@linux.intel.com>"); +MODULE_DESCRIPTION("T10 DIF CRC calculation accelerated with PCLMULQDQ."); +MODULE_LICENSE("GPL"); + +MODULE_ALIAS("crct10dif"); +MODULE_ALIAS("crct10dif-pclmul"); diff --git a/arch/x86/crypto/sha256_ssse3_glue.c b/arch/x86/crypto/sha256_ssse3_glue.c index 597d4da69656..50226c4b86ed 100644 --- a/arch/x86/crypto/sha256_ssse3_glue.c +++ b/arch/x86/crypto/sha256_ssse3_glue.c @@ -187,7 +187,36 @@ static int sha256_ssse3_import(struct shash_desc *desc, const void *in) return 0; } -static struct shash_alg alg = { +static int sha224_ssse3_init(struct shash_desc *desc) +{ + struct sha256_state *sctx = shash_desc_ctx(desc); + + sctx->state[0] = SHA224_H0; + sctx->state[1] = SHA224_H1; + sctx->state[2] = SHA224_H2; + sctx->state[3] = SHA224_H3; + sctx->state[4] = SHA224_H4; + sctx->state[5] = SHA224_H5; + sctx->state[6] = SHA224_H6; + sctx->state[7] = SHA224_H7; + sctx->count = 0; + + return 0; +} + +static int sha224_ssse3_final(struct shash_desc *desc, u8 *hash) +{ + u8 D[SHA256_DIGEST_SIZE]; + + sha256_ssse3_final(desc, D); + + memcpy(hash, D, SHA224_DIGEST_SIZE); + memset(D, 0, SHA256_DIGEST_SIZE); + + return 0; +} + +static struct shash_alg algs[] = { { .digestsize = SHA256_DIGEST_SIZE, .init = sha256_ssse3_init, .update = sha256_ssse3_update, @@ -204,7 +233,24 @@ static struct shash_alg alg = { .cra_blocksize = SHA256_BLOCK_SIZE, .cra_module = THIS_MODULE, } -}; +}, { + .digestsize = SHA224_DIGEST_SIZE, + .init = sha224_ssse3_init, + .update = sha256_ssse3_update, + .final = sha224_ssse3_final, + .export = sha256_ssse3_export, + .import = sha256_ssse3_import, + .descsize = sizeof(struct sha256_state), + .statesize = sizeof(struct sha256_state), + .base = { + .cra_name = "sha224", + .cra_driver_name = "sha224-ssse3", + .cra_priority = 150, + .cra_flags = CRYPTO_ALG_TYPE_SHASH, + .cra_blocksize = SHA224_BLOCK_SIZE, + .cra_module = THIS_MODULE, + } +} }; #ifdef CONFIG_AS_AVX static bool __init avx_usable(void) @@ -227,7 +273,7 @@ static bool __init avx_usable(void) static int __init sha256_ssse3_mod_init(void) { - /* test for SSE3 first */ + /* test for SSSE3 first */ if (cpu_has_ssse3) sha256_transform_asm = sha256_transform_ssse3; @@ -254,7 +300,7 @@ static int __init sha256_ssse3_mod_init(void) else #endif pr_info("Using SSSE3 optimized SHA-256 implementation\n"); - return crypto_register_shash(&alg); + return crypto_register_shashes(algs, ARRAY_SIZE(algs)); } pr_info("Neither AVX nor SSSE3 is available/usable.\n"); @@ -263,7 +309,7 @@ static int __init sha256_ssse3_mod_init(void) static void __exit sha256_ssse3_mod_fini(void) { - crypto_unregister_shash(&alg); + crypto_unregister_shashes(algs, ARRAY_SIZE(algs)); } module_init(sha256_ssse3_mod_init); @@ -273,3 +319,4 @@ MODULE_LICENSE("GPL"); MODULE_DESCRIPTION("SHA256 Secure Hash Algorithm, Supplemental SSE3 accelerated"); MODULE_ALIAS("sha256"); +MODULE_ALIAS("sha384"); diff --git a/arch/x86/crypto/sha512_ssse3_glue.c b/arch/x86/crypto/sha512_ssse3_glue.c index 6cbd8df348d2..f30cd10293f0 100644 --- a/arch/x86/crypto/sha512_ssse3_glue.c +++ b/arch/x86/crypto/sha512_ssse3_glue.c @@ -194,7 +194,37 @@ static int sha512_ssse3_import(struct shash_desc *desc, const void *in) return 0; } -static struct shash_alg alg = { +static int sha384_ssse3_init(struct shash_desc *desc) +{ + struct sha512_state *sctx = shash_desc_ctx(desc); + + sctx->state[0] = SHA384_H0; + sctx->state[1] = SHA384_H1; + sctx->state[2] = SHA384_H2; + sctx->state[3] = SHA384_H3; + sctx->state[4] = SHA384_H4; + sctx->state[5] = SHA384_H5; + sctx->state[6] = SHA384_H6; + sctx->state[7] = SHA384_H7; + + sctx->count[0] = sctx->count[1] = 0; + + return 0; +} + +static int sha384_ssse3_final(struct shash_desc *desc, u8 *hash) +{ + u8 D[SHA512_DIGEST_SIZE]; + + sha512_ssse3_final(desc, D); + + memcpy(hash, D, SHA384_DIGEST_SIZE); + memset(D, 0, SHA512_DIGEST_SIZE); + + return 0; +} + +static struct shash_alg algs[] = { { .digestsize = SHA512_DIGEST_SIZE, .init = sha512_ssse3_init, .update = sha512_ssse3_update, @@ -211,7 +241,24 @@ static struct shash_alg alg = { .cra_blocksize = SHA512_BLOCK_SIZE, .cra_module = THIS_MODULE, } -}; +}, { + .digestsize = SHA384_DIGEST_SIZE, + .init = sha384_ssse3_init, + .update = sha512_ssse3_update, + .final = sha384_ssse3_final, + .export = sha512_ssse3_export, + .import = sha512_ssse3_import, + .descsize = sizeof(struct sha512_state), + .statesize = sizeof(struct sha512_state), + .base = { + .cra_name = "sha384", + .cra_driver_name = "sha384-ssse3", + .cra_priority = 150, + .cra_flags = CRYPTO_ALG_TYPE_SHASH, + .cra_blocksize = SHA384_BLOCK_SIZE, + .cra_module = THIS_MODULE, + } +} }; #ifdef CONFIG_AS_AVX static bool __init avx_usable(void) @@ -234,7 +281,7 @@ static bool __init avx_usable(void) static int __init sha512_ssse3_mod_init(void) { - /* test for SSE3 first */ + /* test for SSSE3 first */ if (cpu_has_ssse3) sha512_transform_asm = sha512_transform_ssse3; @@ -261,7 +308,7 @@ static int __init sha512_ssse3_mod_init(void) else #endif pr_info("Using SSSE3 optimized SHA-512 implementation\n"); - return crypto_register_shash(&alg); + return crypto_register_shashes(algs, ARRAY_SIZE(algs)); } pr_info("Neither AVX nor SSSE3 is available/usable.\n"); @@ -270,7 +317,7 @@ static int __init sha512_ssse3_mod_init(void) static void __exit sha512_ssse3_mod_fini(void) { - crypto_unregister_shash(&alg); + crypto_unregister_shashes(algs, ARRAY_SIZE(algs)); } module_init(sha512_ssse3_mod_init); @@ -280,3 +327,4 @@ MODULE_LICENSE("GPL"); MODULE_DESCRIPTION("SHA512 Secure Hash Algorithm, Supplemental SSE3 accelerated"); MODULE_ALIAS("sha512"); +MODULE_ALIAS("sha384"); diff --git a/arch/x86/crypto/twofish-avx2-asm_64.S b/arch/x86/crypto/twofish-avx2-asm_64.S deleted file mode 100644 index e1a83b9cd389..000000000000 --- a/arch/x86/crypto/twofish-avx2-asm_64.S +++ /dev/null @@ -1,600 +0,0 @@ -/* - * x86_64/AVX2 assembler optimized version of Twofish - * - * Copyright © 2012-2013 Jussi Kivilinna <jussi.kivilinna@iki.fi> - * - * 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> -#include "glue_helper-asm-avx2.S" - -.file "twofish-avx2-asm_64.S" - -.data -.align 16 - -.Lvpshufb_mask0: -.long 0x80808000 -.long 0x80808004 -.long 0x80808008 -.long 0x8080800c - -.Lbswap128_mask: - .byte 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0 -.Lxts_gf128mul_and_shl1_mask_0: - .byte 0x87, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 -.Lxts_gf128mul_and_shl1_mask_1: - .byte 0x0e, 1, 0, 0, 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0, 0 - -.text - -/* structure of crypto context */ -#define s0 0 -#define s1 1024 -#define s2 2048 -#define s3 3072 -#define w 4096 -#define k 4128 - -/* register macros */ -#define CTX %rdi - -#define RS0 CTX -#define RS1 %r8 -#define RS2 %r9 -#define RS3 %r10 -#define RK %r11 -#define RW %rax -#define RROUND %r12 -#define RROUNDd %r12d - -#define RA0 %ymm8 -#define RB0 %ymm9 -#define RC0 %ymm10 -#define RD0 %ymm11 -#define RA1 %ymm12 -#define RB1 %ymm13 -#define RC1 %ymm14 -#define RD1 %ymm15 - -/* temp regs */ -#define RX0 %ymm0 -#define RY0 %ymm1 -#define RX1 %ymm2 -#define RY1 %ymm3 -#define RT0 %ymm4 -#define RIDX %ymm5 - -#define RX0x %xmm0 -#define RY0x %xmm1 -#define RX1x %xmm2 -#define RY1x %xmm3 -#define RT0x %xmm4 - -/* vpgatherdd mask and '-1' */ -#define RNOT %ymm6 - -/* byte mask, (-1 >> 24) */ -#define RBYTE %ymm7 - -/********************************************************************** - 16-way AVX2 twofish - **********************************************************************/ -#define init_round_constants() \ - vpcmpeqd RNOT, RNOT, RNOT; \ - vpsrld $24, RNOT, RBYTE; \ - leaq k(CTX), RK; \ - leaq w(CTX), RW; \ - leaq s1(CTX), RS1; \ - leaq s2(CTX), RS2; \ - leaq s3(CTX), RS3; \ - -#define g16(ab, rs0, rs1, rs2, rs3, xy) \ - vpand RBYTE, ab ## 0, RIDX; \ - vpgatherdd RNOT, (rs0, RIDX, 4), xy ## 0; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - \ - vpand RBYTE, ab ## 1, RIDX; \ - vpgatherdd RNOT, (rs0, RIDX, 4), xy ## 1; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - \ - vpsrld $8, ab ## 0, RIDX; \ - vpand RBYTE, RIDX, RIDX; \ - vpgatherdd RNOT, (rs1, RIDX, 4), RT0; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - vpxor RT0, xy ## 0, xy ## 0; \ - \ - vpsrld $8, ab ## 1, RIDX; \ - vpand RBYTE, RIDX, RIDX; \ - vpgatherdd RNOT, (rs1, RIDX, 4), RT0; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - vpxor RT0, xy ## 1, xy ## 1; \ - \ - vpsrld $16, ab ## 0, RIDX; \ - vpand RBYTE, RIDX, RIDX; \ - vpgatherdd RNOT, (rs2, RIDX, 4), RT0; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - vpxor RT0, xy ## 0, xy ## 0; \ - \ - vpsrld $16, ab ## 1, RIDX; \ - vpand RBYTE, RIDX, RIDX; \ - vpgatherdd RNOT, (rs2, RIDX, 4), RT0; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - vpxor RT0, xy ## 1, xy ## 1; \ - \ - vpsrld $24, ab ## 0, RIDX; \ - vpgatherdd RNOT, (rs3, RIDX, 4), RT0; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - vpxor RT0, xy ## 0, xy ## 0; \ - \ - vpsrld $24, ab ## 1, RIDX; \ - vpgatherdd RNOT, (rs3, RIDX, 4), RT0; \ - vpcmpeqd RNOT, RNOT, RNOT; \ - vpxor RT0, xy ## 1, xy ## 1; - -#define g1_16(a, x) \ - g16(a, RS0, RS1, RS2, RS3, x); - -#define g2_16(b, y) \ - g16(b, RS1, RS2, RS3, RS0, y); - -#define encrypt_round_end16(a, b, c, d, nk) \ - vpaddd RY0, RX0, RX0; \ - vpaddd RX0, RY0, RY0; \ - vpbroadcastd nk(RK,RROUND,8), RT0; \ - vpaddd RT0, RX0, RX0; \ - vpbroadcastd 4+nk(RK,RROUND,8), RT0; \ - vpaddd RT0, RY0, RY0; \ - \ - vpxor RY0, d ## 0, d ## 0; \ - \ - vpxor RX0, c ## 0, c ## 0; \ - vpsrld $1, c ## 0, RT0; \ - vpslld $31, c ## 0, c ## 0; \ - vpor RT0, c ## 0, c ## 0; \ - \ - vpaddd RY1, RX1, RX1; \ - vpaddd RX1, RY1, RY1; \ - vpbroadcastd nk(RK,RROUND,8), RT0; \ - vpaddd RT0, RX1, RX1; \ - vpbroadcastd 4+nk(RK,RROUND,8), RT0; \ - vpaddd RT0, RY1, RY1; \ - \ - vpxor RY1, d ## 1, d ## 1; \ - \ - vpxor RX1, c ## 1, c ## 1; \ - vpsrld $1, c ## 1, RT0; \ - vpslld $31, c ## 1, c ## 1; \ - vpor RT0, c ## 1, c ## 1; \ - -#define encrypt_round16(a, b, c, d, nk) \ - g2_16(b, RY); \ - \ - vpslld $1, b ## 0, RT0; \ - vpsrld $31, b ## 0, b ## 0; \ - vpor RT0, b ## 0, b ## 0; \ - \ - vpslld $1, b ## 1, RT0; \ - vpsrld $31, b ## 1, b ## 1; \ - vpor RT0, b ## 1, b ## 1; \ - \ - g1_16(a, RX); \ - \ - encrypt_round_end16(a, b, c, d, nk); - -#define encrypt_round_first16(a, b, c, d, nk) \ - vpslld $1, d ## 0, RT0; \ - vpsrld $31, d ## 0, d ## 0; \ - vpor RT0, d ## 0, d ## 0; \ - \ - vpslld $1, d ## 1, RT0; \ - vpsrld $31, d ## 1, d ## 1; \ - vpor RT0, d ## 1, d ## 1; \ - \ - encrypt_round16(a, b, c, d, nk); - -#define encrypt_round_last16(a, b, c, d, nk) \ - g2_16(b, RY); \ - \ - g1_16(a, RX); \ - \ - encrypt_round_end16(a, b, c, d, nk); - -#define decrypt_round_end16(a, b, c, d, nk) \ - vpaddd RY0, RX0, RX0; \ - vpaddd RX0, RY0, RY0; \ - vpbroadcastd nk(RK,RROUND,8), RT0; \ - vpaddd RT0, RX0, RX0; \ - vpbroadcastd 4+nk(RK,RROUND,8), RT0; \ - vpaddd RT0, RY0, RY0; \ - \ - vpxor RX0, c ## 0, c ## 0; \ - \ - vpxor RY0, d ## 0, d ## 0; \ - vpsrld $1, d ## 0, RT0; \ - vpslld $31, d ## 0, d ## 0; \ - vpor RT0, d ## 0, d ## 0; \ - \ - vpaddd RY1, RX1, RX1; \ - vpaddd RX1, RY1, RY1; \ - vpbroadcastd nk(RK,RROUND,8), RT0; \ - vpaddd RT0, RX1, RX1; \ - vpbroadcastd 4+nk(RK,RROUND,8), RT0; \ - vpaddd RT0, RY1, RY1; \ - \ - vpxor RX1, c ## 1, c ## 1; \ - \ - vpxor RY1, d ## 1, d ## 1; \ - vpsrld $1, d ## 1, RT0; \ - vpslld $31, d ## 1, d ## 1; \ - vpor RT0, d ## 1, d ## 1; - -#define decrypt_round16(a, b, c, d, nk) \ - g1_16(a, RX); \ - \ - vpslld $1, a ## 0, RT0; \ - vpsrld $31, a ## 0, a ## 0; \ - vpor RT0, a ## 0, a ## 0; \ - \ - vpslld $1, a ## 1, RT0; \ - vpsrld $31, a ## 1, a ## 1; \ - vpor RT0, a ## 1, a ## 1; \ - \ - g2_16(b, RY); \ - \ - decrypt_round_end16(a, b, c, d, nk); - -#define decrypt_round_first16(a, b, c, d, nk) \ - vpslld $1, c ## 0, RT0; \ - vpsrld $31, c ## 0, c ## 0; \ - vpor RT0, c ## 0, c ## 0; \ - \ - vpslld $1, c ## 1, RT0; \ - vpsrld $31, c ## 1, c ## 1; \ - vpor RT0, c ## 1, c ## 1; \ - \ - decrypt_round16(a, b, c, d, nk) - -#define decrypt_round_last16(a, b, c, d, nk) \ - g1_16(a, RX); \ - \ - g2_16(b, RY); \ - \ - decrypt_round_end16(a, b, c, d, nk); - -#define encrypt_cycle16() \ - encrypt_round16(RA, RB, RC, RD, 0); \ - encrypt_round16(RC, RD, RA, RB, 8); - -#define encrypt_cycle_first16() \ - encrypt_round_first16(RA, RB, RC, RD, 0); \ - encrypt_round16(RC, RD, RA, RB, 8); - -#define encrypt_cycle_last16() \ - encrypt_round16(RA, RB, RC, RD, 0); \ - encrypt_round_last16(RC, RD, RA, RB, 8); - -#define decrypt_cycle16(n) \ - decrypt_round16(RC, RD, RA, RB, 8); \ - decrypt_round16(RA, RB, RC, RD, 0); - -#define decrypt_cycle_first16(n) \ - decrypt_round_first16(RC, RD, RA, RB, 8); \ - decrypt_round16(RA, RB, RC, RD, 0); - -#define decrypt_cycle_last16(n) \ - decrypt_round16(RC, RD, RA, RB, 8); \ - decrypt_round_last16(RA, RB, RC, RD, 0); - -#define transpose_4x4(x0,x1,x2,x3,t1,t2) \ - vpunpckhdq x1, x0, t2; \ - vpunpckldq x1, x0, x0; \ - \ - vpunpckldq x3, x2, t1; \ - vpunpckhdq x3, x2, x2; \ - \ - vpunpckhqdq t1, x0, x1; \ - vpunpcklqdq t1, x0, x0; \ - \ - vpunpckhqdq x2, t2, x3; \ - vpunpcklqdq x2, t2, x2; - -#define read_blocks8(offs,a,b,c,d) \ - transpose_4x4(a, b, c, d, RX0, RY0); - -#define write_blocks8(offs,a,b,c,d) \ - transpose_4x4(a, b, c, d, RX0, RY0); - -#define inpack_enc8(a,b,c,d) \ - vpbroadcastd 4*0(RW), RT0; \ - vpxor RT0, a, a; \ - \ - vpbroadcastd 4*1(RW), RT0; \ - vpxor RT0, b, b; \ - \ - vpbroadcastd 4*2(RW), RT0; \ - vpxor RT0, c, c; \ - \ - vpbroadcastd 4*3(RW), RT0; \ - vpxor RT0, d, d; - -#define outunpack_enc8(a,b,c,d) \ - vpbroadcastd 4*4(RW), RX0; \ - vpbroadcastd 4*5(RW), RY0; \ - vpxor RX0, c, RX0; \ - vpxor RY0, d, RY0; \ - \ - vpbroadcastd 4*6(RW), RT0; \ - vpxor RT0, a, c; \ - vpbroadcastd 4*7(RW), RT0; \ - vpxor RT0, b, d; \ - \ - vmovdqa RX0, a; \ - vmovdqa RY0, b; - -#define inpack_dec8(a,b,c,d) \ - vpbroadcastd 4*4(RW), RX0; \ - vpbroadcastd 4*5(RW), RY0; \ - vpxor RX0, a, RX0; \ - vpxor RY0, b, RY0; \ - \ - vpbroadcastd 4*6(RW), RT0; \ - vpxor RT0, c, a; \ - vpbroadcastd 4*7(RW), RT0; \ - vpxor RT0, d, b; \ - \ - vmovdqa RX0, c; \ - vmovdqa RY0, d; - -#define outunpack_dec8(a,b,c,d) \ - vpbroadcastd 4*0(RW), RT0; \ - vpxor RT0, a, a; \ - \ - vpbroadcastd 4*1(RW), RT0; \ - vpxor RT0, b, b; \ - \ - vpbroadcastd 4*2(RW), RT0; \ - vpxor RT0, c, c; \ - \ - vpbroadcastd 4*3(RW), RT0; \ - vpxor RT0, d, d; - -#define read_blocks16(a,b,c,d) \ - read_blocks8(0, a ## 0, b ## 0, c ## 0, d ## 0); \ - read_blocks8(8, a ## 1, b ## 1, c ## 1, d ## 1); - -#define write_blocks16(a,b,c,d) \ - write_blocks8(0, a ## 0, b ## 0, c ## 0, d ## 0); \ - write_blocks8(8, a ## 1, b ## 1, c ## 1, d ## 1); - -#define xor_blocks16(a,b,c,d) \ - xor_blocks8(0, a ## 0, b ## 0, c ## 0, d ## 0); \ - xor_blocks8(8, a ## 1, b ## 1, c ## 1, d ## 1); - -#define inpack_enc16(a,b,c,d) \ - inpack_enc8(a ## 0, b ## 0, c ## 0, d ## 0); \ - inpack_enc8(a ## 1, b ## 1, c ## 1, d ## 1); - -#define outunpack_enc16(a,b,c,d) \ - outunpack_enc8(a ## 0, b ## 0, c ## 0, d ## 0); \ - outunpack_enc8(a ## 1, b ## 1, c ## 1, d ## 1); - -#define inpack_dec16(a,b,c,d) \ - inpack_dec8(a ## 0, b ## 0, c ## 0, d ## 0); \ - inpack_dec8(a ## 1, b ## 1, c ## 1, d ## 1); - -#define outunpack_dec16(a,b,c,d) \ - outunpack_dec8(a ## 0, b ## 0, c ## 0, d ## 0); \ - outunpack_dec8(a ## 1, b ## 1, c ## 1, d ## 1); - -.align 8 -__twofish_enc_blk16: - /* input: - * %rdi: ctx, CTX - * RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1: plaintext - * output: - * RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1: ciphertext - */ - init_round_constants(); - - read_blocks16(RA, RB, RC, RD); - inpack_enc16(RA, RB, RC, RD); - - xorl RROUNDd, RROUNDd; - encrypt_cycle_first16(); - movl $2, RROUNDd; - -.align 4 -.L__enc_loop: - encrypt_cycle16(); - - addl $2, RROUNDd; - cmpl $14, RROUNDd; - jne .L__enc_loop; - - encrypt_cycle_last16(); - - outunpack_enc16(RA, RB, RC, RD); - write_blocks16(RA, RB, RC, RD); - - ret; -ENDPROC(__twofish_enc_blk16) - -.align 8 -__twofish_dec_blk16: - /* input: - * %rdi: ctx, CTX - * RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1: ciphertext - * output: - * RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1: plaintext - */ - init_round_constants(); - - read_blocks16(RA, RB, RC, RD); - inpack_dec16(RA, RB, RC, RD); - - movl $14, RROUNDd; - decrypt_cycle_first16(); - movl $12, RROUNDd; - -.align 4 -.L__dec_loop: - decrypt_cycle16(); - - addl $-2, RROUNDd; - jnz .L__dec_loop; - - decrypt_cycle_last16(); - - outunpack_dec16(RA, RB, RC, RD); - write_blocks16(RA, RB, RC, RD); - - ret; -ENDPROC(__twofish_dec_blk16) - -ENTRY(twofish_ecb_enc_16way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst - * %rdx: src - */ - - vzeroupper; - pushq %r12; - - load_16way(%rdx, RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1); - - call __twofish_enc_blk16; - - store_16way(%rsi, RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1); - - popq %r12; - vzeroupper; - - ret; -ENDPROC(twofish_ecb_enc_16way) - -ENTRY(twofish_ecb_dec_16way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst - * %rdx: src - */ - - vzeroupper; - pushq %r12; - - load_16way(%rdx, RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1); - - call __twofish_dec_blk16; - - store_16way(%rsi, RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1); - - popq %r12; - vzeroupper; - - ret; -ENDPROC(twofish_ecb_dec_16way) - -ENTRY(twofish_cbc_dec_16way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst - * %rdx: src - */ - - vzeroupper; - pushq %r12; - - load_16way(%rdx, RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1); - - call __twofish_dec_blk16; - - store_cbc_16way(%rdx, %rsi, RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1, - RX0); - - popq %r12; - vzeroupper; - - ret; -ENDPROC(twofish_cbc_dec_16way) - -ENTRY(twofish_ctr_16way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst (16 blocks) - * %rdx: src (16 blocks) - * %rcx: iv (little endian, 128bit) - */ - - vzeroupper; - pushq %r12; - - load_ctr_16way(%rcx, .Lbswap128_mask, RA0, RB0, RC0, RD0, RA1, RB1, RC1, - RD1, RX0, RX0x, RX1, RX1x, RY0, RY0x, RY1, RY1x, RNOT, - RBYTE); - - call __twofish_enc_blk16; - - store_ctr_16way(%rdx, %rsi, RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1); - - popq %r12; - vzeroupper; - - ret; -ENDPROC(twofish_ctr_16way) - -.align 8 -twofish_xts_crypt_16way: - /* input: - * %rdi: ctx, CTX - * %rsi: dst (16 blocks) - * %rdx: src (16 blocks) - * %rcx: iv (t ⊕ αⁿ ∈ GF(2¹²⁸)) - * %r8: pointer to __twofish_enc_blk16 or __twofish_dec_blk16 - */ - - vzeroupper; - pushq %r12; - - load_xts_16way(%rcx, %rdx, %rsi, RA0, RB0, RC0, RD0, RA1, RB1, RC1, - RD1, RX0, RX0x, RX1, RX1x, RY0, RY0x, RY1, RY1x, RNOT, - .Lxts_gf128mul_and_shl1_mask_0, - .Lxts_gf128mul_and_shl1_mask_1); - - call *%r8; - - store_xts_16way(%rsi, RA0, RB0, RC0, RD0, RA1, RB1, RC1, RD1); - - popq %r12; - vzeroupper; - - ret; -ENDPROC(twofish_xts_crypt_16way) - -ENTRY(twofish_xts_enc_16way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst (16 blocks) - * %rdx: src (16 blocks) - * %rcx: iv (t ⊕ αⁿ ∈ GF(2¹²⁸)) - */ - leaq __twofish_enc_blk16, %r8; - jmp twofish_xts_crypt_16way; -ENDPROC(twofish_xts_enc_16way) - -ENTRY(twofish_xts_dec_16way) - /* input: - * %rdi: ctx, CTX - * %rsi: dst (16 blocks) - * %rdx: src (16 blocks) - * %rcx: iv (t ⊕ αⁿ ∈ GF(2¹²⁸)) - */ - leaq __twofish_dec_blk16, %r8; - jmp twofish_xts_crypt_16way; -ENDPROC(twofish_xts_dec_16way) diff --git a/arch/x86/crypto/twofish_avx2_glue.c b/arch/x86/crypto/twofish_avx2_glue.c deleted file mode 100644 index ce33b5be64ee..000000000000 --- a/arch/x86/crypto/twofish_avx2_glue.c +++ /dev/null @@ -1,584 +0,0 @@ -/* - * Glue Code for x86_64/AVX2 assembler optimized version of Twofish - * - * Copyright © 2012-2013 Jussi Kivilinna <jussi.kivilinna@mbnet.fi> - * - * 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/module.h> -#include <linux/types.h> -#include <linux/crypto.h> -#include <linux/err.h> -#include <crypto/algapi.h> -#include <crypto/ctr.h> -#include <crypto/twofish.h> -#include <crypto/lrw.h> -#include <crypto/xts.h> -#include <asm/xcr.h> -#include <asm/xsave.h> -#include <asm/crypto/twofish.h> -#include <asm/crypto/ablk_helper.h> -#include <asm/crypto/glue_helper.h> -#include <crypto/scatterwalk.h> - -#define TF_AVX2_PARALLEL_BLOCKS 16 - -/* 16-way AVX2 parallel cipher functions */ -asmlinkage void twofish_ecb_enc_16way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src); -asmlinkage void twofish_ecb_dec_16way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src); -asmlinkage void twofish_cbc_dec_16way(void *ctx, u128 *dst, const u128 *src); - -asmlinkage void twofish_ctr_16way(void *ctx, u128 *dst, const u128 *src, - le128 *iv); - -asmlinkage void twofish_xts_enc_16way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src, le128 *iv); -asmlinkage void twofish_xts_dec_16way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src, le128 *iv); - -static inline void twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src) -{ - __twofish_enc_blk_3way(ctx, dst, src, false); -} - -static const struct common_glue_ctx twofish_enc = { - .num_funcs = 4, - .fpu_blocks_limit = 8, - - .funcs = { { - .num_blocks = 16, - .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_ecb_enc_16way) } - }, { - .num_blocks = 8, - .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_ecb_enc_8way) } - }, { - .num_blocks = 3, - .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_enc_blk_3way) } - }, { - .num_blocks = 1, - .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_enc_blk) } - } } -}; - -static const struct common_glue_ctx twofish_ctr = { - .num_funcs = 4, - .fpu_blocks_limit = 8, - - .funcs = { { - .num_blocks = 16, - .fn_u = { .ctr = GLUE_CTR_FUNC_CAST(twofish_ctr_16way) } - }, { - .num_blocks = 8, - .fn_u = { .ctr = GLUE_CTR_FUNC_CAST(twofish_ctr_8way) } - }, { - .num_blocks = 3, - .fn_u = { .ctr = GLUE_CTR_FUNC_CAST(twofish_enc_blk_ctr_3way) } - }, { - .num_blocks = 1, - .fn_u = { .ctr = GLUE_CTR_FUNC_CAST(twofish_enc_blk_ctr) } - } } -}; - -static const struct common_glue_ctx twofish_enc_xts = { - .num_funcs = 3, - .fpu_blocks_limit = 8, - - .funcs = { { - .num_blocks = 16, - .fn_u = { .xts = GLUE_XTS_FUNC_CAST(twofish_xts_enc_16way) } - }, { - .num_blocks = 8, - .fn_u = { .xts = GLUE_XTS_FUNC_CAST(twofish_xts_enc_8way) } - }, { - .num_blocks = 1, - .fn_u = { .xts = GLUE_XTS_FUNC_CAST(twofish_xts_enc) } - } } -}; - -static const struct common_glue_ctx twofish_dec = { - .num_funcs = 4, - .fpu_blocks_limit = 8, - - .funcs = { { - .num_blocks = 16, - .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_ecb_dec_16way) } - }, { - .num_blocks = 8, - .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_ecb_dec_8way) } - }, { - .num_blocks = 3, - .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_dec_blk_3way) } - }, { - .num_blocks = 1, - .fn_u = { .ecb = GLUE_FUNC_CAST(twofish_dec_blk) } - } } -}; - -static const struct common_glue_ctx twofish_dec_cbc = { - .num_funcs = 4, - .fpu_blocks_limit = 8, - - .funcs = { { - .num_blocks = 16, - .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(twofish_cbc_dec_16way) } - }, { - .num_blocks = 8, - .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(twofish_cbc_dec_8way) } - }, { - .num_blocks = 3, - .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(twofish_dec_blk_cbc_3way) } - }, { - .num_blocks = 1, - .fn_u = { .cbc = GLUE_CBC_FUNC_CAST(twofish_dec_blk) } - } } -}; - -static const struct common_glue_ctx twofish_dec_xts = { - .num_funcs = 3, - .fpu_blocks_limit = 8, - - .funcs = { { - .num_blocks = 16, - .fn_u = { .xts = GLUE_XTS_FUNC_CAST(twofish_xts_dec_16way) } - }, { - .num_blocks = 8, - .fn_u = { .xts = GLUE_XTS_FUNC_CAST(twofish_xts_dec_8way) } - }, { - .num_blocks = 1, - .fn_u = { .xts = GLUE_XTS_FUNC_CAST(twofish_xts_dec) } - } } -}; - -static int ecb_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - return glue_ecb_crypt_128bit(&twofish_enc, desc, dst, src, nbytes); -} - -static int ecb_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - return glue_ecb_crypt_128bit(&twofish_dec, desc, dst, src, nbytes); -} - -static int cbc_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - return glue_cbc_encrypt_128bit(GLUE_FUNC_CAST(twofish_enc_blk), desc, - dst, src, nbytes); -} - -static int cbc_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - return glue_cbc_decrypt_128bit(&twofish_dec_cbc, desc, dst, src, - nbytes); -} - -static int ctr_crypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - return glue_ctr_crypt_128bit(&twofish_ctr, desc, dst, src, nbytes); -} - -static inline bool twofish_fpu_begin(bool fpu_enabled, unsigned int nbytes) -{ - /* since reusing AVX functions, starts using FPU at 8 parallel blocks */ - return glue_fpu_begin(TF_BLOCK_SIZE, 8, NULL, fpu_enabled, nbytes); -} - -static inline void twofish_fpu_end(bool fpu_enabled) -{ - glue_fpu_end(fpu_enabled); -} - -struct crypt_priv { - struct twofish_ctx *ctx; - bool fpu_enabled; -}; - -static void encrypt_callback(void *priv, u8 *srcdst, unsigned int nbytes) -{ - const unsigned int bsize = TF_BLOCK_SIZE; - struct crypt_priv *ctx = priv; - int i; - - ctx->fpu_enabled = twofish_fpu_begin(ctx->fpu_enabled, nbytes); - - while (nbytes >= TF_AVX2_PARALLEL_BLOCKS * bsize) { - twofish_ecb_enc_16way(ctx->ctx, srcdst, srcdst); - srcdst += bsize * TF_AVX2_PARALLEL_BLOCKS; - nbytes -= bsize * TF_AVX2_PARALLEL_BLOCKS; - } - - while (nbytes >= 8 * bsize) { - twofish_ecb_enc_8way(ctx->ctx, srcdst, srcdst); - srcdst += bsize * 8; - nbytes -= bsize * 8; - } - - while (nbytes >= 3 * bsize) { - twofish_enc_blk_3way(ctx->ctx, srcdst, srcdst); - srcdst += bsize * 3; - nbytes -= bsize * 3; - } - - for (i = 0; i < nbytes / bsize; i++, srcdst += bsize) - twofish_enc_blk(ctx->ctx, srcdst, srcdst); -} - -static void decrypt_callback(void *priv, u8 *srcdst, unsigned int nbytes) -{ - const unsigned int bsize = TF_BLOCK_SIZE; - struct crypt_priv *ctx = priv; - int i; - - ctx->fpu_enabled = twofish_fpu_begin(ctx->fpu_enabled, nbytes); - - while (nbytes >= TF_AVX2_PARALLEL_BLOCKS * bsize) { - twofish_ecb_dec_16way(ctx->ctx, srcdst, srcdst); - srcdst += bsize * TF_AVX2_PARALLEL_BLOCKS; - nbytes -= bsize * TF_AVX2_PARALLEL_BLOCKS; - } - - while (nbytes >= 8 * bsize) { - twofish_ecb_dec_8way(ctx->ctx, srcdst, srcdst); - srcdst += bsize * 8; - nbytes -= bsize * 8; - } - - while (nbytes >= 3 * bsize) { - twofish_dec_blk_3way(ctx->ctx, srcdst, srcdst); - srcdst += bsize * 3; - nbytes -= bsize * 3; - } - - for (i = 0; i < nbytes / bsize; i++, srcdst += bsize) - twofish_dec_blk(ctx->ctx, srcdst, srcdst); -} - -static int lrw_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - struct twofish_lrw_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); - be128 buf[TF_AVX2_PARALLEL_BLOCKS]; - struct crypt_priv crypt_ctx = { - .ctx = &ctx->twofish_ctx, - .fpu_enabled = false, - }; - struct lrw_crypt_req req = { - .tbuf = buf, - .tbuflen = sizeof(buf), - - .table_ctx = &ctx->lrw_table, - .crypt_ctx = &crypt_ctx, - .crypt_fn = encrypt_callback, - }; - int ret; - - desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; - ret = lrw_crypt(desc, dst, src, nbytes, &req); - twofish_fpu_end(crypt_ctx.fpu_enabled); - - return ret; -} - -static int lrw_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - struct twofish_lrw_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); - be128 buf[TF_AVX2_PARALLEL_BLOCKS]; - struct crypt_priv crypt_ctx = { - .ctx = &ctx->twofish_ctx, - .fpu_enabled = false, - }; - struct lrw_crypt_req req = { - .tbuf = buf, - .tbuflen = sizeof(buf), - - .table_ctx = &ctx->lrw_table, - .crypt_ctx = &crypt_ctx, - .crypt_fn = decrypt_callback, - }; - int ret; - - desc->flags &= ~CRYPTO_TFM_REQ_MAY_SLEEP; - ret = lrw_crypt(desc, dst, src, nbytes, &req); - twofish_fpu_end(crypt_ctx.fpu_enabled); - - return ret; -} - -static int xts_encrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - struct twofish_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); - - return glue_xts_crypt_128bit(&twofish_enc_xts, desc, dst, src, nbytes, - XTS_TWEAK_CAST(twofish_enc_blk), - &ctx->tweak_ctx, &ctx->crypt_ctx); -} - -static int xts_decrypt(struct blkcipher_desc *desc, struct scatterlist *dst, - struct scatterlist *src, unsigned int nbytes) -{ - struct twofish_xts_ctx *ctx = crypto_blkcipher_ctx(desc->tfm); - - return glue_xts_crypt_128bit(&twofish_dec_xts, desc, dst, src, nbytes, - XTS_TWEAK_CAST(twofish_enc_blk), - &ctx->tweak_ctx, &ctx->crypt_ctx); -} - -static struct crypto_alg tf_algs[10] = { { - .cra_name = "__ecb-twofish-avx2", - .cra_driver_name = "__driver-ecb-twofish-avx2", - .cra_priority = 0, - .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, - .cra_blocksize = TF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct twofish_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_blkcipher_type, - .cra_module = THIS_MODULE, - .cra_u = { - .blkcipher = { - .min_keysize = TF_MIN_KEY_SIZE, - .max_keysize = TF_MAX_KEY_SIZE, - .setkey = twofish_setkey, - .encrypt = ecb_encrypt, - .decrypt = ecb_decrypt, - }, - }, -}, { - .cra_name = "__cbc-twofish-avx2", - .cra_driver_name = "__driver-cbc-twofish-avx2", - .cra_priority = 0, - .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, - .cra_blocksize = TF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct twofish_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_blkcipher_type, - .cra_module = THIS_MODULE, - .cra_u = { - .blkcipher = { - .min_keysize = TF_MIN_KEY_SIZE, - .max_keysize = TF_MAX_KEY_SIZE, - .setkey = twofish_setkey, - .encrypt = cbc_encrypt, - .decrypt = cbc_decrypt, - }, - }, -}, { - .cra_name = "__ctr-twofish-avx2", - .cra_driver_name = "__driver-ctr-twofish-avx2", - .cra_priority = 0, - .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, - .cra_blocksize = 1, - .cra_ctxsize = sizeof(struct twofish_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_blkcipher_type, - .cra_module = THIS_MODULE, - .cra_u = { - .blkcipher = { - .min_keysize = TF_MIN_KEY_SIZE, - .max_keysize = TF_MAX_KEY_SIZE, - .ivsize = TF_BLOCK_SIZE, - .setkey = twofish_setkey, - .encrypt = ctr_crypt, - .decrypt = ctr_crypt, - }, - }, -}, { - .cra_name = "__lrw-twofish-avx2", - .cra_driver_name = "__driver-lrw-twofish-avx2", - .cra_priority = 0, - .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, - .cra_blocksize = TF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct twofish_lrw_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_blkcipher_type, - .cra_module = THIS_MODULE, - .cra_exit = lrw_twofish_exit_tfm, - .cra_u = { - .blkcipher = { - .min_keysize = TF_MIN_KEY_SIZE + - TF_BLOCK_SIZE, - .max_keysize = TF_MAX_KEY_SIZE + - TF_BLOCK_SIZE, - .ivsize = TF_BLOCK_SIZE, - .setkey = lrw_twofish_setkey, - .encrypt = lrw_encrypt, - .decrypt = lrw_decrypt, - }, - }, -}, { - .cra_name = "__xts-twofish-avx2", - .cra_driver_name = "__driver-xts-twofish-avx2", - .cra_priority = 0, - .cra_flags = CRYPTO_ALG_TYPE_BLKCIPHER, - .cra_blocksize = TF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct twofish_xts_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_blkcipher_type, - .cra_module = THIS_MODULE, - .cra_u = { - .blkcipher = { - .min_keysize = TF_MIN_KEY_SIZE * 2, - .max_keysize = TF_MAX_KEY_SIZE * 2, - .ivsize = TF_BLOCK_SIZE, - .setkey = xts_twofish_setkey, - .encrypt = xts_encrypt, - .decrypt = xts_decrypt, - }, - }, -}, { - .cra_name = "ecb(twofish)", - .cra_driver_name = "ecb-twofish-avx2", - .cra_priority = 500, - .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, - .cra_blocksize = TF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct async_helper_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_ablkcipher_type, - .cra_module = THIS_MODULE, - .cra_init = ablk_init, - .cra_exit = ablk_exit, - .cra_u = { - .ablkcipher = { - .min_keysize = TF_MIN_KEY_SIZE, - .max_keysize = TF_MAX_KEY_SIZE, - .setkey = ablk_set_key, - .encrypt = ablk_encrypt, - .decrypt = ablk_decrypt, - }, - }, -}, { - .cra_name = "cbc(twofish)", - .cra_driver_name = "cbc-twofish-avx2", - .cra_priority = 500, - .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, - .cra_blocksize = TF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct async_helper_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_ablkcipher_type, - .cra_module = THIS_MODULE, - .cra_init = ablk_init, - .cra_exit = ablk_exit, - .cra_u = { - .ablkcipher = { - .min_keysize = TF_MIN_KEY_SIZE, - .max_keysize = TF_MAX_KEY_SIZE, - .ivsize = TF_BLOCK_SIZE, - .setkey = ablk_set_key, - .encrypt = __ablk_encrypt, - .decrypt = ablk_decrypt, - }, - }, -}, { - .cra_name = "ctr(twofish)", - .cra_driver_name = "ctr-twofish-avx2", - .cra_priority = 500, - .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, - .cra_blocksize = 1, - .cra_ctxsize = sizeof(struct async_helper_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_ablkcipher_type, - .cra_module = THIS_MODULE, - .cra_init = ablk_init, - .cra_exit = ablk_exit, - .cra_u = { - .ablkcipher = { - .min_keysize = TF_MIN_KEY_SIZE, - .max_keysize = TF_MAX_KEY_SIZE, - .ivsize = TF_BLOCK_SIZE, - .setkey = ablk_set_key, - .encrypt = ablk_encrypt, - .decrypt = ablk_encrypt, - .geniv = "chainiv", - }, - }, -}, { - .cra_name = "lrw(twofish)", - .cra_driver_name = "lrw-twofish-avx2", - .cra_priority = 500, - .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, - .cra_blocksize = TF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct async_helper_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_ablkcipher_type, - .cra_module = THIS_MODULE, - .cra_init = ablk_init, - .cra_exit = ablk_exit, - .cra_u = { - .ablkcipher = { - .min_keysize = TF_MIN_KEY_SIZE + - TF_BLOCK_SIZE, - .max_keysize = TF_MAX_KEY_SIZE + - TF_BLOCK_SIZE, - .ivsize = TF_BLOCK_SIZE, - .setkey = ablk_set_key, - .encrypt = ablk_encrypt, - .decrypt = ablk_decrypt, - }, - }, -}, { - .cra_name = "xts(twofish)", - .cra_driver_name = "xts-twofish-avx2", - .cra_priority = 500, - .cra_flags = CRYPTO_ALG_TYPE_ABLKCIPHER | CRYPTO_ALG_ASYNC, - .cra_blocksize = TF_BLOCK_SIZE, - .cra_ctxsize = sizeof(struct async_helper_ctx), - .cra_alignmask = 0, - .cra_type = &crypto_ablkcipher_type, - .cra_module = THIS_MODULE, - .cra_init = ablk_init, - .cra_exit = ablk_exit, - .cra_u = { - .ablkcipher = { - .min_keysize = TF_MIN_KEY_SIZE * 2, - .max_keysize = TF_MAX_KEY_SIZE * 2, - .ivsize = TF_BLOCK_SIZE, - .setkey = ablk_set_key, - .encrypt = ablk_encrypt, - .decrypt = ablk_decrypt, - }, - }, -} }; - -static int __init init(void) -{ - u64 xcr0; - - if (!cpu_has_avx2 || !cpu_has_osxsave) { - pr_info("AVX2 instructions are not detected.\n"); - return -ENODEV; - } - - xcr0 = xgetbv(XCR_XFEATURE_ENABLED_MASK); - if ((xcr0 & (XSTATE_SSE | XSTATE_YMM)) != (XSTATE_SSE | XSTATE_YMM)) { - pr_info("AVX2 detected but unusable.\n"); - return -ENODEV; - } - - return crypto_register_algs(tf_algs, ARRAY_SIZE(tf_algs)); -} - -static void __exit fini(void) -{ - crypto_unregister_algs(tf_algs, ARRAY_SIZE(tf_algs)); -} - -module_init(init); -module_exit(fini); - -MODULE_LICENSE("GPL"); -MODULE_DESCRIPTION("Twofish Cipher Algorithm, AVX2 optimized"); -MODULE_ALIAS("twofish"); -MODULE_ALIAS("twofish-asm"); diff --git a/arch/x86/crypto/twofish_avx_glue.c b/arch/x86/crypto/twofish_avx_glue.c index 2047a562f6b3..a62ba541884e 100644 --- a/arch/x86/crypto/twofish_avx_glue.c +++ b/arch/x86/crypto/twofish_avx_glue.c @@ -50,26 +50,18 @@ /* 8-way parallel cipher functions */ asmlinkage void twofish_ecb_enc_8way(struct twofish_ctx *ctx, u8 *dst, const u8 *src); -EXPORT_SYMBOL_GPL(twofish_ecb_enc_8way); - asmlinkage void twofish_ecb_dec_8way(struct twofish_ctx *ctx, u8 *dst, const u8 *src); -EXPORT_SYMBOL_GPL(twofish_ecb_dec_8way); asmlinkage void twofish_cbc_dec_8way(struct twofish_ctx *ctx, u8 *dst, const u8 *src); -EXPORT_SYMBOL_GPL(twofish_cbc_dec_8way); - asmlinkage void twofish_ctr_8way(struct twofish_ctx *ctx, u8 *dst, const u8 *src, le128 *iv); -EXPORT_SYMBOL_GPL(twofish_ctr_8way); asmlinkage void twofish_xts_enc_8way(struct twofish_ctx *ctx, u8 *dst, const u8 *src, le128 *iv); -EXPORT_SYMBOL_GPL(twofish_xts_enc_8way); asmlinkage void twofish_xts_dec_8way(struct twofish_ctx *ctx, u8 *dst, const u8 *src, le128 *iv); -EXPORT_SYMBOL_GPL(twofish_xts_dec_8way); static inline void twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst, const u8 *src) @@ -77,19 +69,17 @@ static inline void twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst, __twofish_enc_blk_3way(ctx, dst, src, false); } -void twofish_xts_enc(void *ctx, u128 *dst, const u128 *src, le128 *iv) +static void twofish_xts_enc(void *ctx, u128 *dst, const u128 *src, le128 *iv) { glue_xts_crypt_128bit_one(ctx, dst, src, iv, GLUE_FUNC_CAST(twofish_enc_blk)); } -EXPORT_SYMBOL_GPL(twofish_xts_enc); -void twofish_xts_dec(void *ctx, u128 *dst, const u128 *src, le128 *iv) +static void twofish_xts_dec(void *ctx, u128 *dst, const u128 *src, le128 *iv) { glue_xts_crypt_128bit_one(ctx, dst, src, iv, GLUE_FUNC_CAST(twofish_dec_blk)); } -EXPORT_SYMBOL_GPL(twofish_xts_dec); static const struct common_glue_ctx twofish_enc = { diff --git a/arch/x86/include/asm/crypto/blowfish.h b/arch/x86/include/asm/crypto/blowfish.h deleted file mode 100644 index f097b2face10..000000000000 --- a/arch/x86/include/asm/crypto/blowfish.h +++ /dev/null @@ -1,43 +0,0 @@ -#ifndef ASM_X86_BLOWFISH_H -#define ASM_X86_BLOWFISH_H - -#include <linux/crypto.h> -#include <crypto/blowfish.h> - -#define BF_PARALLEL_BLOCKS 4 - -/* regular block cipher functions */ -asmlinkage void __blowfish_enc_blk(struct bf_ctx *ctx, u8 *dst, const u8 *src, - bool xor); -asmlinkage void blowfish_dec_blk(struct bf_ctx *ctx, u8 *dst, const u8 *src); - -/* 4-way parallel cipher functions */ -asmlinkage void __blowfish_enc_blk_4way(struct bf_ctx *ctx, u8 *dst, - const u8 *src, bool xor); -asmlinkage void blowfish_dec_blk_4way(struct bf_ctx *ctx, u8 *dst, - const u8 *src); - -static inline void blowfish_enc_blk(struct bf_ctx *ctx, u8 *dst, const u8 *src) -{ - __blowfish_enc_blk(ctx, dst, src, false); -} - -static inline void blowfish_enc_blk_xor(struct bf_ctx *ctx, u8 *dst, - const u8 *src) -{ - __blowfish_enc_blk(ctx, dst, src, true); -} - -static inline void blowfish_enc_blk_4way(struct bf_ctx *ctx, u8 *dst, - const u8 *src) -{ - __blowfish_enc_blk_4way(ctx, dst, src, false); -} - -static inline void blowfish_enc_blk_xor_4way(struct bf_ctx *ctx, u8 *dst, - const u8 *src) -{ - __blowfish_enc_blk_4way(ctx, dst, src, true); -} - -#endif diff --git a/arch/x86/include/asm/crypto/twofish.h b/arch/x86/include/asm/crypto/twofish.h index e655c6029b45..878c51ceebb5 100644 --- a/arch/x86/include/asm/crypto/twofish.h +++ b/arch/x86/include/asm/crypto/twofish.h @@ -28,20 +28,6 @@ asmlinkage void __twofish_enc_blk_3way(struct twofish_ctx *ctx, u8 *dst, asmlinkage void twofish_dec_blk_3way(struct twofish_ctx *ctx, u8 *dst, const u8 *src); -/* 8-way parallel cipher functions */ -asmlinkage void twofish_ecb_enc_8way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src); -asmlinkage void twofish_ecb_dec_8way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src); -asmlinkage void twofish_cbc_dec_8way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src); -asmlinkage void twofish_ctr_8way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src, le128 *iv); -asmlinkage void twofish_xts_enc_8way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src, le128 *iv); -asmlinkage void twofish_xts_dec_8way(struct twofish_ctx *ctx, u8 *dst, - const u8 *src, le128 *iv); - /* helpers from twofish_x86_64-3way module */ extern void twofish_dec_blk_cbc_3way(void *ctx, u128 *dst, const u128 *src); extern void twofish_enc_blk_ctr(void *ctx, u128 *dst, const u128 *src, @@ -57,8 +43,4 @@ extern void lrw_twofish_exit_tfm(struct crypto_tfm *tfm); extern int xts_twofish_setkey(struct crypto_tfm *tfm, const u8 *key, unsigned int keylen); -/* helpers from twofish-avx module */ -extern void twofish_xts_enc(void *ctx, u128 *dst, const u128 *src, le128 *iv); -extern void twofish_xts_dec(void *ctx, u128 *dst, const u128 *src, le128 *iv); - #endif /* ASM_X86_TWOFISH_H */ |