1*4882a593Smuzhiyun // SPDX-License-Identifier: GPL-2.0
2*4882a593Smuzhiyun /*
3*4882a593Smuzhiyun * SHA1 routine optimized to do word accesses rather than byte accesses,
4*4882a593Smuzhiyun * and to avoid unnecessary copies into the context array.
5*4882a593Smuzhiyun *
6*4882a593Smuzhiyun * This was based on the git SHA1 implementation.
7*4882a593Smuzhiyun */
8*4882a593Smuzhiyun
9*4882a593Smuzhiyun #include <linux/kernel.h>
10*4882a593Smuzhiyun #include <linux/export.h>
11*4882a593Smuzhiyun #include <linux/bitops.h>
12*4882a593Smuzhiyun #include <linux/string.h>
13*4882a593Smuzhiyun #include <crypto/sha.h>
14*4882a593Smuzhiyun #include <asm/unaligned.h>
15*4882a593Smuzhiyun
16*4882a593Smuzhiyun /*
17*4882a593Smuzhiyun * If you have 32 registers or more, the compiler can (and should)
18*4882a593Smuzhiyun * try to change the array[] accesses into registers. However, on
19*4882a593Smuzhiyun * machines with less than ~25 registers, that won't really work,
20*4882a593Smuzhiyun * and at least gcc will make an unholy mess of it.
21*4882a593Smuzhiyun *
22*4882a593Smuzhiyun * So to avoid that mess which just slows things down, we force
23*4882a593Smuzhiyun * the stores to memory to actually happen (we might be better off
24*4882a593Smuzhiyun * with a 'W(t)=(val);asm("":"+m" (W(t))' there instead, as
25*4882a593Smuzhiyun * suggested by Artur Skawina - that will also make gcc unable to
26*4882a593Smuzhiyun * try to do the silly "optimize away loads" part because it won't
27*4882a593Smuzhiyun * see what the value will be).
28*4882a593Smuzhiyun *
29*4882a593Smuzhiyun * Ben Herrenschmidt reports that on PPC, the C version comes close
30*4882a593Smuzhiyun * to the optimized asm with this (ie on PPC you don't want that
31*4882a593Smuzhiyun * 'volatile', since there are lots of registers).
32*4882a593Smuzhiyun *
33*4882a593Smuzhiyun * On ARM we get the best code generation by forcing a full memory barrier
34*4882a593Smuzhiyun * between each SHA_ROUND, otherwise gcc happily get wild with spilling and
35*4882a593Smuzhiyun * the stack frame size simply explode and performance goes down the drain.
36*4882a593Smuzhiyun */
37*4882a593Smuzhiyun
38*4882a593Smuzhiyun #ifdef CONFIG_X86
39*4882a593Smuzhiyun #define setW(x, val) (*(volatile __u32 *)&W(x) = (val))
40*4882a593Smuzhiyun #elif defined(CONFIG_ARM)
41*4882a593Smuzhiyun #define setW(x, val) do { W(x) = (val); __asm__("":::"memory"); } while (0)
42*4882a593Smuzhiyun #else
43*4882a593Smuzhiyun #define setW(x, val) (W(x) = (val))
44*4882a593Smuzhiyun #endif
45*4882a593Smuzhiyun
46*4882a593Smuzhiyun /* This "rolls" over the 512-bit array */
47*4882a593Smuzhiyun #define W(x) (array[(x)&15])
48*4882a593Smuzhiyun
49*4882a593Smuzhiyun /*
50*4882a593Smuzhiyun * Where do we get the source from? The first 16 iterations get it from
51*4882a593Smuzhiyun * the input data, the next mix it from the 512-bit array.
52*4882a593Smuzhiyun */
53*4882a593Smuzhiyun #define SHA_SRC(t) get_unaligned_be32((__u32 *)data + t)
54*4882a593Smuzhiyun #define SHA_MIX(t) rol32(W(t+13) ^ W(t+8) ^ W(t+2) ^ W(t), 1)
55*4882a593Smuzhiyun
56*4882a593Smuzhiyun #define SHA_ROUND(t, input, fn, constant, A, B, C, D, E) do { \
57*4882a593Smuzhiyun __u32 TEMP = input(t); setW(t, TEMP); \
58*4882a593Smuzhiyun E += TEMP + rol32(A,5) + (fn) + (constant); \
59*4882a593Smuzhiyun B = ror32(B, 2); \
60*4882a593Smuzhiyun TEMP = E; E = D; D = C; C = B; B = A; A = TEMP; } while (0)
61*4882a593Smuzhiyun
62*4882a593Smuzhiyun #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 )
63*4882a593Smuzhiyun #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 )
64*4882a593Smuzhiyun #define T_20_39(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0x6ed9eba1, A, B, C, D, E )
65*4882a593Smuzhiyun #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 )
66*4882a593Smuzhiyun #define T_60_79(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0xca62c1d6, A, B, C, D, E )
67*4882a593Smuzhiyun
68*4882a593Smuzhiyun /**
69*4882a593Smuzhiyun * sha1_transform - single block SHA1 transform (deprecated)
70*4882a593Smuzhiyun *
71*4882a593Smuzhiyun * @digest: 160 bit digest to update
72*4882a593Smuzhiyun * @data: 512 bits of data to hash
73*4882a593Smuzhiyun * @array: 16 words of workspace (see note)
74*4882a593Smuzhiyun *
75*4882a593Smuzhiyun * This function executes SHA-1's internal compression function. It updates the
76*4882a593Smuzhiyun * 160-bit internal state (@digest) with a single 512-bit data block (@data).
77*4882a593Smuzhiyun *
78*4882a593Smuzhiyun * Don't use this function. SHA-1 is no longer considered secure. And even if
79*4882a593Smuzhiyun * you do have to use SHA-1, this isn't the correct way to hash something with
80*4882a593Smuzhiyun * SHA-1 as this doesn't handle padding and finalization.
81*4882a593Smuzhiyun *
82*4882a593Smuzhiyun * Note: If the hash is security sensitive, the caller should be sure
83*4882a593Smuzhiyun * to clear the workspace. This is left to the caller to avoid
84*4882a593Smuzhiyun * unnecessary clears between chained hashing operations.
85*4882a593Smuzhiyun */
sha1_transform(__u32 * digest,const char * data,__u32 * array)86*4882a593Smuzhiyun void sha1_transform(__u32 *digest, const char *data, __u32 *array)
87*4882a593Smuzhiyun {
88*4882a593Smuzhiyun __u32 A, B, C, D, E;
89*4882a593Smuzhiyun unsigned int i = 0;
90*4882a593Smuzhiyun
91*4882a593Smuzhiyun A = digest[0];
92*4882a593Smuzhiyun B = digest[1];
93*4882a593Smuzhiyun C = digest[2];
94*4882a593Smuzhiyun D = digest[3];
95*4882a593Smuzhiyun E = digest[4];
96*4882a593Smuzhiyun
97*4882a593Smuzhiyun /* Round 1 - iterations 0-16 take their input from 'data' */
98*4882a593Smuzhiyun for (; i < 16; ++i)
99*4882a593Smuzhiyun T_0_15(i, A, B, C, D, E);
100*4882a593Smuzhiyun
101*4882a593Smuzhiyun /* Round 1 - tail. Input from 512-bit mixing array */
102*4882a593Smuzhiyun for (; i < 20; ++i)
103*4882a593Smuzhiyun T_16_19(i, A, B, C, D, E);
104*4882a593Smuzhiyun
105*4882a593Smuzhiyun /* Round 2 */
106*4882a593Smuzhiyun for (; i < 40; ++i)
107*4882a593Smuzhiyun T_20_39(i, A, B, C, D, E);
108*4882a593Smuzhiyun
109*4882a593Smuzhiyun /* Round 3 */
110*4882a593Smuzhiyun for (; i < 60; ++i)
111*4882a593Smuzhiyun T_40_59(i, A, B, C, D, E);
112*4882a593Smuzhiyun
113*4882a593Smuzhiyun /* Round 4 */
114*4882a593Smuzhiyun for (; i < 80; ++i)
115*4882a593Smuzhiyun T_60_79(i, A, B, C, D, E);
116*4882a593Smuzhiyun
117*4882a593Smuzhiyun digest[0] += A;
118*4882a593Smuzhiyun digest[1] += B;
119*4882a593Smuzhiyun digest[2] += C;
120*4882a593Smuzhiyun digest[3] += D;
121*4882a593Smuzhiyun digest[4] += E;
122*4882a593Smuzhiyun }
123*4882a593Smuzhiyun EXPORT_SYMBOL(sha1_transform);
124*4882a593Smuzhiyun
125*4882a593Smuzhiyun /**
126*4882a593Smuzhiyun * sha1_init - initialize the vectors for a SHA1 digest
127*4882a593Smuzhiyun * @buf: vector to initialize
128*4882a593Smuzhiyun */
sha1_init(__u32 * buf)129*4882a593Smuzhiyun void sha1_init(__u32 *buf)
130*4882a593Smuzhiyun {
131*4882a593Smuzhiyun buf[0] = 0x67452301;
132*4882a593Smuzhiyun buf[1] = 0xefcdab89;
133*4882a593Smuzhiyun buf[2] = 0x98badcfe;
134*4882a593Smuzhiyun buf[3] = 0x10325476;
135*4882a593Smuzhiyun buf[4] = 0xc3d2e1f0;
136*4882a593Smuzhiyun }
137*4882a593Smuzhiyun EXPORT_SYMBOL(sha1_init);
138