blob: 185fad49d86f0c114411f4f4bc879bdbd9102874 [file] [log] [blame]
Huang Ying0e1227d2009-10-19 11:53:06 +09001/*
2 * Accelerated GHASH implementation with Intel PCLMULQDQ-NI
3 * instructions. This file contains accelerated part of ghash
4 * implementation. More information about PCLMULQDQ can be found at:
5 *
6 * http://software.intel.com/en-us/articles/carry-less-multiplication-and-its-usage-for-computing-the-gcm-mode/
7 *
8 * Copyright (c) 2009 Intel Corp.
9 * Author: Huang Ying <ying.huang@intel.com>
10 * Vinodh Gopal
11 * Erdinc Ozturk
12 * Deniz Karakoyunlu
13 *
14 * This program is free software; you can redistribute it and/or modify it
15 * under the terms of the GNU General Public License version 2 as published
16 * by the Free Software Foundation.
17 */
18
19#include <linux/linkage.h>
Huang Ying564ec0e2009-11-23 19:55:22 +080020#include <asm/inst.h>
Huang Ying0e1227d2009-10-19 11:53:06 +090021
Jiri Kosina68ee8712009-11-23 20:19:47 +080022.data
23
Huang Ying0e1227d2009-10-19 11:53:06 +090024.align 16
25.Lbswap_mask:
26 .octa 0x000102030405060708090a0b0c0d0e0f
Huang Ying0e1227d2009-10-19 11:53:06 +090027
28#define DATA %xmm0
29#define SHASH %xmm1
30#define T1 %xmm2
31#define T2 %xmm3
32#define T3 %xmm4
33#define BSWAP %xmm5
34#define IN1 %xmm6
35
36.text
37
38/*
39 * __clmul_gf128mul_ble: internal ABI
40 * input:
41 * DATA: operand1
42 * SHASH: operand2, hash_key << 1 mod poly
43 * output:
44 * DATA: operand1 * operand2 mod poly
45 * changed:
46 * T1
47 * T2
48 * T3
49 */
50__clmul_gf128mul_ble:
51 movaps DATA, T1
52 pshufd $0b01001110, DATA, T2
53 pshufd $0b01001110, SHASH, T3
54 pxor DATA, T2
55 pxor SHASH, T3
56
Huang Ying564ec0e2009-11-23 19:55:22 +080057 PCLMULQDQ 0x00 SHASH DATA # DATA = a0 * b0
58 PCLMULQDQ 0x11 SHASH T1 # T1 = a1 * b1
59 PCLMULQDQ 0x00 T3 T2 # T2 = (a1 + a0) * (b1 + b0)
Huang Ying0e1227d2009-10-19 11:53:06 +090060 pxor DATA, T2
61 pxor T1, T2 # T2 = a0 * b1 + a1 * b0
62
63 movaps T2, T3
64 pslldq $8, T3
65 psrldq $8, T2
66 pxor T3, DATA
67 pxor T2, T1 # <T1:DATA> is result of
68 # carry-less multiplication
69
70 # first phase of the reduction
71 movaps DATA, T3
72 psllq $1, T3
73 pxor DATA, T3
74 psllq $5, T3
75 pxor DATA, T3
76 psllq $57, T3
77 movaps T3, T2
78 pslldq $8, T2
79 psrldq $8, T3
80 pxor T2, DATA
81 pxor T3, T1
82
83 # second phase of the reduction
84 movaps DATA, T2
85 psrlq $5, T2
86 pxor DATA, T2
87 psrlq $1, T2
88 pxor DATA, T2
89 psrlq $1, T2
90 pxor T2, T1
91 pxor T1, DATA
92 ret
Jussi Kivilinnab05d3f32013-01-19 13:39:26 +020093ENDPROC(__clmul_gf128mul_ble)
Huang Ying0e1227d2009-10-19 11:53:06 +090094
95/* void clmul_ghash_mul(char *dst, const be128 *shash) */
96ENTRY(clmul_ghash_mul)
97 movups (%rdi), DATA
98 movups (%rsi), SHASH
99 movaps .Lbswap_mask, BSWAP
Huang Ying564ec0e2009-11-23 19:55:22 +0800100 PSHUFB_XMM BSWAP DATA
Huang Ying0e1227d2009-10-19 11:53:06 +0900101 call __clmul_gf128mul_ble
Huang Ying564ec0e2009-11-23 19:55:22 +0800102 PSHUFB_XMM BSWAP DATA
Huang Ying0e1227d2009-10-19 11:53:06 +0900103 movups DATA, (%rdi)
104 ret
Jussi Kivilinnab05d3f32013-01-19 13:39:26 +0200105ENDPROC(clmul_ghash_mul)
Huang Ying0e1227d2009-10-19 11:53:06 +0900106
107/*
108 * void clmul_ghash_update(char *dst, const char *src, unsigned int srclen,
109 * const be128 *shash);
110 */
111ENTRY(clmul_ghash_update)
112 cmp $16, %rdx
113 jb .Lupdate_just_ret # check length
114 movaps .Lbswap_mask, BSWAP
115 movups (%rdi), DATA
116 movups (%rcx), SHASH
Huang Ying564ec0e2009-11-23 19:55:22 +0800117 PSHUFB_XMM BSWAP DATA
Huang Ying0e1227d2009-10-19 11:53:06 +0900118.align 4
119.Lupdate_loop:
120 movups (%rsi), IN1
Huang Ying564ec0e2009-11-23 19:55:22 +0800121 PSHUFB_XMM BSWAP IN1
Huang Ying0e1227d2009-10-19 11:53:06 +0900122 pxor IN1, DATA
123 call __clmul_gf128mul_ble
124 sub $16, %rdx
125 add $16, %rsi
126 cmp $16, %rdx
127 jge .Lupdate_loop
Huang Ying564ec0e2009-11-23 19:55:22 +0800128 PSHUFB_XMM BSWAP DATA
Huang Ying0e1227d2009-10-19 11:53:06 +0900129 movups DATA, (%rdi)
130.Lupdate_just_ret:
131 ret
Jussi Kivilinnab05d3f32013-01-19 13:39:26 +0200132ENDPROC(clmul_ghash_update)