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