1// SPDX-License-Identifier: GPL-2.0
2/*
3 * SHA1 routine optimized to do word accesses rather than byte accesses,
4 * and to avoid unnecessary copies into the context array.
5 *
6 * This was based on the git SHA1 implementation.
7 */
8
9#include <linux/kernel.h>
10#include <linux/export.h>
11#include <linux/bitops.h>
12#include <linux/cryptohash.h>
13#include <asm/unaligned.h>
14
15/*
16 * If you have 32 registers or more, the compiler can (and should)
17 * try to change the array[] accesses into registers. However, on
18 * machines with less than ~25 registers, that won't really work,
19 * and at least gcc will make an unholy mess of it.
20 *
21 * So to avoid that mess which just slows things down, we force
22 * the stores to memory to actually happen (we might be better off
23 * with a 'W(t)=(val);asm("":"+m" (W(t))' there instead, as
24 * suggested by Artur Skawina - that will also make gcc unable to
25 * try to do the silly "optimize away loads" part because it won't
26 * see what the value will be).
27 *
28 * Ben Herrenschmidt reports that on PPC, the C version comes close
29 * to the optimized asm with this (ie on PPC you don't want that
30 * 'volatile', since there are lots of registers).
31 *
32 * On ARM we get the best code generation by forcing a full memory barrier
33 * between each SHA_ROUND, otherwise gcc happily get wild with spilling and
34 * the stack frame size simply explode and performance goes down the drain.
35 */
36
37#ifdef CONFIG_X86
38 #define setW(x, val) (*(volatile __u32 *)&W(x) = (val))
39#elif defined(CONFIG_ARM)
40 #define setW(x, val) do { W(x) = (val); __asm__("":::"memory"); } while (0)
41#else
42 #define setW(x, val) (W(x) = (val))
43#endif
44
45/* This "rolls" over the 512-bit array */
46#define W(x) (array[(x)&15])
47
48/*
49 * Where do we get the source from? The first 16 iterations get it from
50 * the input data, the next mix it from the 512-bit array.
51 */
52#define SHA_SRC(t) get_unaligned_be32((__u32 *)data + t)
53#define SHA_MIX(t) rol32(W(t+13) ^ W(t+8) ^ W(t+2) ^ W(t), 1)
54
55#define SHA_ROUND(t, input, fn, constant, A, B, C, D, E) do { \
56 __u32 TEMP = input(t); setW(t, TEMP); \
57 E += TEMP + rol32(A,5) + (fn) + (constant); \
58 B = ror32(B, 2); } while (0)
59
60#define T_0_15(t, A, B, C, D, E) SHA_ROUND(t, SHA_SRC, (((C^D)&B)^D) , 0x5a827999, A, B, C, D, E )
61#define T_16_19(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (((C^D)&B)^D) , 0x5a827999, A, B, C, D, E )
62#define T_20_39(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0x6ed9eba1, A, B, C, D, E )
63#define T_40_59(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, ((B&C)+(D&(B^C))) , 0x8f1bbcdc, A, B, C, D, E )
64#define T_60_79(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0xca62c1d6, A, B, C, D, E )
65
66/**
67 * sha_transform - single block SHA1 transform
68 *
69 * @digest: 160 bit digest to update
70 * @data: 512 bits of data to hash
71 * @array: 16 words of workspace (see note)
72 *
73 * This function generates a SHA1 digest for a single 512-bit block.
74 * Be warned, it does not handle padding and message digest, do not
75 * confuse it with the full FIPS 180-1 digest algorithm for variable
76 * length messages.
77 *
78 * Note: If the hash is security sensitive, the caller should be sure
79 * to clear the workspace. This is left to the caller to avoid
80 * unnecessary clears between chained hashing operations.
81 */
82void sha_transform(__u32 *digest, const char *data, __u32 *array)
83{
84 __u32 A, B, C, D, E;
85
86 A = digest[0];
87 B = digest[1];
88 C = digest[2];
89 D = digest[3];
90 E = digest[4];
91
92 /* Round 1 - iterations 0-16 take their input from 'data' */
93 T_0_15( 0, A, B, C, D, E);
94 T_0_15( 1, E, A, B, C, D);
95 T_0_15( 2, D, E, A, B, C);
96 T_0_15( 3, C, D, E, A, B);
97 T_0_15( 4, B, C, D, E, A);
98 T_0_15( 5, A, B, C, D, E);
99 T_0_15( 6, E, A, B, C, D);
100 T_0_15( 7, D, E, A, B, C);
101 T_0_15( 8, C, D, E, A, B);
102 T_0_15( 9, B, C, D, E, A);
103 T_0_15(10, A, B, C, D, E);
104 T_0_15(11, E, A, B, C, D);
105 T_0_15(12, D, E, A, B, C);
106 T_0_15(13, C, D, E, A, B);
107 T_0_15(14, B, C, D, E, A);
108 T_0_15(15, A, B, C, D, E);
109
110 /* Round 1 - tail. Input from 512-bit mixing array */
111 T_16_19(16, E, A, B, C, D);
112 T_16_19(17, D, E, A, B, C);
113 T_16_19(18, C, D, E, A, B);
114 T_16_19(19, B, C, D, E, A);
115
116 /* Round 2 */
117 T_20_39(20, A, B, C, D, E);
118 T_20_39(21, E, A, B, C, D);
119 T_20_39(22, D, E, A, B, C);
120 T_20_39(23, C, D, E, A, B);
121 T_20_39(24, B, C, D, E, A);
122 T_20_39(25, A, B, C, D, E);
123 T_20_39(26, E, A, B, C, D);
124 T_20_39(27, D, E, A, B, C);
125 T_20_39(28, C, D, E, A, B);
126 T_20_39(29, B, C, D, E, A);
127 T_20_39(30, A, B, C, D, E);
128 T_20_39(31, E, A, B, C, D);
129 T_20_39(32, D, E, A, B, C);
130 T_20_39(33, C, D, E, A, B);
131 T_20_39(34, B, C, D, E, A);
132 T_20_39(35, A, B, C, D, E);
133 T_20_39(36, E, A, B, C, D);
134 T_20_39(37, D, E, A, B, C);
135 T_20_39(38, C, D, E, A, B);
136 T_20_39(39, B, C, D, E, A);
137
138 /* Round 3 */
139 T_40_59(40, A, B, C, D, E);
140 T_40_59(41, E, A, B, C, D);
141 T_40_59(42, D, E, A, B, C);
142 T_40_59(43, C, D, E, A, B);
143 T_40_59(44, B, C, D, E, A);
144 T_40_59(45, A, B, C, D, E);
145 T_40_59(46, E, A, B, C, D);
146 T_40_59(47, D, E, A, B, C);
147 T_40_59(48, C, D, E, A, B);
148 T_40_59(49, B, C, D, E, A);
149 T_40_59(50, A, B, C, D, E);
150 T_40_59(51, E, A, B, C, D);
151 T_40_59(52, D, E, A, B, C);
152 T_40_59(53, C, D, E, A, B);
153 T_40_59(54, B, C, D, E, A);
154 T_40_59(55, A, B, C, D, E);
155 T_40_59(56, E, A, B, C, D);
156 T_40_59(57, D, E, A, B, C);
157 T_40_59(58, C, D, E, A, B);
158 T_40_59(59, B, C, D, E, A);
159
160 /* Round 4 */
161 T_60_79(60, A, B, C, D, E);
162 T_60_79(61, E, A, B, C, D);
163 T_60_79(62, D, E, A, B, C);
164 T_60_79(63, C, D, E, A, B);
165 T_60_79(64, B, C, D, E, A);
166 T_60_79(65, A, B, C, D, E);
167 T_60_79(66, E, A, B, C, D);
168 T_60_79(67, D, E, A, B, C);
169 T_60_79(68, C, D, E, A, B);
170 T_60_79(69, B, C, D, E, A);
171 T_60_79(70, A, B, C, D, E);
172 T_60_79(71, E, A, B, C, D);
173 T_60_79(72, D, E, A, B, C);
174 T_60_79(73, C, D, E, A, B);
175 T_60_79(74, B, C, D, E, A);
176 T_60_79(75, A, B, C, D, E);
177 T_60_79(76, E, A, B, C, D);
178 T_60_79(77, D, E, A, B, C);
179 T_60_79(78, C, D, E, A, B);
180 T_60_79(79, B, C, D, E, A);
181
182 digest[0] += A;
183 digest[1] += B;
184 digest[2] += C;
185 digest[3] += D;
186 digest[4] += E;
187}
188EXPORT_SYMBOL(sha_transform);
189
190/**
191 * sha_init - initialize the vectors for a SHA1 digest
192 * @buf: vector to initialize
193 */
194void sha_init(__u32 *buf)
195{
196 buf[0] = 0x67452301;
197 buf[1] = 0xefcdab89;
198 buf[2] = 0x98badcfe;
199 buf[3] = 0x10325476;
200 buf[4] = 0xc3d2e1f0;
201}
202EXPORT_SYMBOL(sha_init);
203