crypto: blowfish - add AVX2/x86_64 implementation of blowfish cipher

Patch adds AVX2/x86-64 implementation of Blowfish cipher, requiring 32 parallel
blocks for input (256 bytes). Table look-ups are performed using vpgatherdd
instruction directly from vector registers and thus should be faster than
earlier implementations.

Signed-off-by: Jussi Kivilinna <jussi.kivilinna@iki.fi>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

1 files changed, 449 insertions, 0 deletions

diff --git a/arch/x86/crypto/blowfish-avx2-asm_64.S b/arch/x86/crypto/blowfish-avx2-asm_64.S
new file mode 100644
index 0000000..784452e
--- /dev/null
+++ b/arch/x86/crypto/blowfish-avx2-asm_64.S
@@ -0,0 +1,449 @@
+/*
+ * 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)