blob: 0494766fc574ec52fb7a1a197e7d639d9a4289ef [file] [log] [blame]
Greg Kroah-Hartmanb2441312017-11-01 15:07:57 +01001// SPDX-License-Identifier: GPL-2.0
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -07003 * 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.
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 */
8
9#include <linux/kernel.h>
Paul Gortmaker8bc3bcc2011-11-16 21:29:17 -050010#include <linux/export.h>
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070011#include <linux/bitops.h>
Jason A. Donenfeld9a1536b2022-01-11 18:58:43 +010012#include <linux/string.h>
Eric Biggersa24d22b2020-11-12 21:20:21 -080013#include <crypto/sha1.h>
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070014#include <asm/unaligned.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070016/*
17 * If you have 32 registers or more, the compiler can (and should)
18 * try to change the array[] accesses into registers. However, on
19 * machines with less than ~25 registers, that won't really work,
20 * and at least gcc will make an unholy mess of it.
21 *
22 * So to avoid that mess which just slows things down, we force
23 * the stores to memory to actually happen (we might be better off
24 * with a 'W(t)=(val);asm("":"+m" (W(t))' there instead, as
25 * suggested by Artur Skawina - that will also make gcc unable to
26 * try to do the silly "optimize away loads" part because it won't
27 * see what the value will be).
28 *
29 * Ben Herrenschmidt reports that on PPC, the C version comes close
30 * to the optimized asm with this (ie on PPC you don't want that
31 * 'volatile', since there are lots of registers).
32 *
33 * On ARM we get the best code generation by forcing a full memory barrier
34 * between each SHA_ROUND, otherwise gcc happily get wild with spilling and
35 * the stack frame size simply explode and performance goes down the drain.
36 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070038#ifdef CONFIG_X86
39 #define setW(x, val) (*(volatile __u32 *)&W(x) = (val))
40#elif defined(CONFIG_ARM)
41 #define setW(x, val) do { W(x) = (val); __asm__("":::"memory"); } while (0)
42#else
43 #define setW(x, val) (W(x) = (val))
44#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070046/* This "rolls" over the 512-bit array */
47#define W(x) (array[(x)&15])
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070049/*
50 * Where do we get the source from? The first 16 iterations get it from
51 * the input data, the next mix it from the 512-bit array.
52 */
53#define SHA_SRC(t) get_unaligned_be32((__u32 *)data + t)
54#define SHA_MIX(t) rol32(W(t+13) ^ W(t+8) ^ W(t+2) ^ W(t), 1)
55
56#define SHA_ROUND(t, input, fn, constant, A, B, C, D, E) do { \
57 __u32 TEMP = input(t); setW(t, TEMP); \
58 E += TEMP + rol32(A,5) + (fn) + (constant); \
Jason A. Donenfeld9a1536b2022-01-11 18:58:43 +010059 B = ror32(B, 2); \
60 TEMP = E; E = D; D = C; C = B; B = A; A = TEMP; } while (0)
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070061
62#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#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#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#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#define T_60_79(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0xca62c1d6, A, B, C, D, E )
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
Robert P. J. Day72fd4a32007-02-10 01:45:59 -080068/**
Eric Biggers6b0b0fa2020-05-02 11:24:25 -070069 * sha1_transform - single block SHA1 transform (deprecated)
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 *
71 * @digest: 160 bit digest to update
72 * @data: 512 bits of data to hash
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070073 * @array: 16 words of workspace (see note)
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 *
Eric Biggers6b0b0fa2020-05-02 11:24:25 -070075 * This function executes SHA-1's internal compression function. It updates the
76 * 160-bit internal state (@digest) with a single 512-bit data block (@data).
77 *
78 * Don't use this function. SHA-1 is no longer considered secure. And even if
79 * you do have to use SHA-1, this isn't the correct way to hash something with
80 * SHA-1 as this doesn't handle padding and finalization.
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 *
82 * Note: If the hash is security sensitive, the caller should be sure
83 * to clear the workspace. This is left to the caller to avoid
84 * unnecessary clears between chained hashing operations.
85 */
Eric Biggers6b0b0fa2020-05-02 11:24:25 -070086void sha1_transform(__u32 *digest, const char *data, __u32 *array)
Linus Torvalds1da177e2005-04-16 15:20:36 -070087{
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070088 __u32 A, B, C, D, E;
Jason A. Donenfeld9a1536b2022-01-11 18:58:43 +010089 unsigned int i = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070091 A = digest[0];
92 B = digest[1];
93 C = digest[2];
94 D = digest[3];
95 E = digest[4];
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -070097 /* Round 1 - iterations 0-16 take their input from 'data' */
Jason A. Donenfeld9a1536b2022-01-11 18:58:43 +010098 for (; i < 16; ++i)
99 T_0_15(i, A, B, C, D, E);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -0700101 /* Round 1 - tail. Input from 512-bit mixing array */
Jason A. Donenfeld9a1536b2022-01-11 18:58:43 +0100102 for (; i < 20; ++i)
103 T_16_19(i, A, B, C, D, E);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -0700105 /* Round 2 */
Jason A. Donenfeld9a1536b2022-01-11 18:58:43 +0100106 for (; i < 40; ++i)
107 T_20_39(i, A, B, C, D, E);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -0700109 /* Round 3 */
Jason A. Donenfeld9a1536b2022-01-11 18:58:43 +0100110 for (; i < 60; ++i)
111 T_40_59(i, A, B, C, D, E);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -0700113 /* Round 4 */
Jason A. Donenfeld9a1536b2022-01-11 18:58:43 +0100114 for (; i < 80; ++i)
115 T_60_79(i, A, B, C, D, E);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
Mandeep Singh Baines1eb19a12011-08-05 18:46:27 -0700117 digest[0] += A;
118 digest[1] += B;
119 digest[2] += C;
120 digest[3] += D;
121 digest[4] += E;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122}
Eric Biggers6b0b0fa2020-05-02 11:24:25 -0700123EXPORT_SYMBOL(sha1_transform);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800125/**
Eric Biggers6b0b0fa2020-05-02 11:24:25 -0700126 * sha1_init - initialize the vectors for a SHA1 digest
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 * @buf: vector to initialize
128 */
Eric Biggers6b0b0fa2020-05-02 11:24:25 -0700129void sha1_init(__u32 *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130{
131 buf[0] = 0x67452301;
132 buf[1] = 0xefcdab89;
133 buf[2] = 0x98badcfe;
134 buf[3] = 0x10325476;
135 buf[4] = 0xc3d2e1f0;
136}
Eric Biggers6b0b0fa2020-05-02 11:24:25 -0700137EXPORT_SYMBOL(sha1_init);