Commit | Line | Data |
---|---|---|
c147028c MS |
1 | /* |
2 | * Glue code for SHA-256 implementation for SPE instructions (PPC) | |
3 | * | |
4 | * Based on generic implementation. The assembler module takes care | |
5 | * about the SPE registers so it can run from interrupt context. | |
6 | * | |
7 | * Copyright (c) 2015 Markus Stockhausen <stockhausen@collogia.de> | |
8 | * | |
9 | * This program is free software; you can redistribute it and/or modify it | |
10 | * under the terms of the GNU General Public License as published by the Free | |
11 | * Software Foundation; either version 2 of the License, or (at your option) | |
12 | * any later version. | |
13 | * | |
14 | */ | |
15 | ||
16 | #include <crypto/internal/hash.h> | |
17 | #include <linux/init.h> | |
18 | #include <linux/module.h> | |
19 | #include <linux/mm.h> | |
20 | #include <linux/cryptohash.h> | |
21 | #include <linux/types.h> | |
22 | #include <crypto/sha.h> | |
23 | #include <asm/byteorder.h> | |
24 | #include <asm/switch_to.h> | |
25 | #include <linux/hardirq.h> | |
26 | ||
27 | /* | |
28 | * MAX_BYTES defines the number of bytes that are allowed to be processed | |
29 | * between preempt_disable() and preempt_enable(). SHA256 takes ~2,000 | |
30 | * operations per 64 bytes. e500 cores can issue two arithmetic instructions | |
31 | * per clock cycle using one 32/64 bit unit (SU1) and one 32 bit unit (SU2). | |
32 | * Thus 1KB of input data will need an estimated maximum of 18,000 cycles. | |
33 | * Headroom for cache misses included. Even with the low end model clocked | |
34 | * at 667 MHz this equals to a critical time window of less than 27us. | |
35 | * | |
36 | */ | |
37 | #define MAX_BYTES 1024 | |
38 | ||
39 | extern void ppc_spe_sha256_transform(u32 *state, const u8 *src, u32 blocks); | |
40 | ||
41 | static void spe_begin(void) | |
42 | { | |
43 | /* We just start SPE operations and will save SPE registers later. */ | |
44 | preempt_disable(); | |
45 | enable_kernel_spe(); | |
46 | } | |
47 | ||
48 | static void spe_end(void) | |
49 | { | |
dc4fbba1 | 50 | disable_kernel_spe(); |
c147028c MS |
51 | /* reenable preemption */ |
52 | preempt_enable(); | |
53 | } | |
54 | ||
55 | static inline void ppc_sha256_clear_context(struct sha256_state *sctx) | |
56 | { | |
57 | int count = sizeof(struct sha256_state) >> 2; | |
58 | u32 *ptr = (u32 *)sctx; | |
59 | ||
60 | /* make sure we can clear the fast way */ | |
61 | BUILD_BUG_ON(sizeof(struct sha256_state) % 4); | |
62 | do { *ptr++ = 0; } while (--count); | |
63 | } | |
64 | ||
65 | static int ppc_spe_sha256_init(struct shash_desc *desc) | |
66 | { | |
67 | struct sha256_state *sctx = shash_desc_ctx(desc); | |
68 | ||
69 | sctx->state[0] = SHA256_H0; | |
70 | sctx->state[1] = SHA256_H1; | |
71 | sctx->state[2] = SHA256_H2; | |
72 | sctx->state[3] = SHA256_H3; | |
73 | sctx->state[4] = SHA256_H4; | |
74 | sctx->state[5] = SHA256_H5; | |
75 | sctx->state[6] = SHA256_H6; | |
76 | sctx->state[7] = SHA256_H7; | |
77 | sctx->count = 0; | |
78 | ||
79 | return 0; | |
80 | } | |
81 | ||
82 | static int ppc_spe_sha224_init(struct shash_desc *desc) | |
83 | { | |
84 | struct sha256_state *sctx = shash_desc_ctx(desc); | |
85 | ||
86 | sctx->state[0] = SHA224_H0; | |
87 | sctx->state[1] = SHA224_H1; | |
88 | sctx->state[2] = SHA224_H2; | |
89 | sctx->state[3] = SHA224_H3; | |
90 | sctx->state[4] = SHA224_H4; | |
91 | sctx->state[5] = SHA224_H5; | |
92 | sctx->state[6] = SHA224_H6; | |
93 | sctx->state[7] = SHA224_H7; | |
94 | sctx->count = 0; | |
95 | ||
96 | return 0; | |
97 | } | |
98 | ||
99 | static int ppc_spe_sha256_update(struct shash_desc *desc, const u8 *data, | |
100 | unsigned int len) | |
101 | { | |
102 | struct sha256_state *sctx = shash_desc_ctx(desc); | |
103 | const unsigned int offset = sctx->count & 0x3f; | |
104 | const unsigned int avail = 64 - offset; | |
105 | unsigned int bytes; | |
106 | const u8 *src = data; | |
107 | ||
108 | if (avail > len) { | |
109 | sctx->count += len; | |
110 | memcpy((char *)sctx->buf + offset, src, len); | |
111 | return 0; | |
112 | } | |
113 | ||
114 | sctx->count += len; | |
115 | ||
116 | if (offset) { | |
117 | memcpy((char *)sctx->buf + offset, src, avail); | |
118 | ||
119 | spe_begin(); | |
120 | ppc_spe_sha256_transform(sctx->state, (const u8 *)sctx->buf, 1); | |
121 | spe_end(); | |
122 | ||
123 | len -= avail; | |
124 | src += avail; | |
125 | } | |
126 | ||
127 | while (len > 63) { | |
128 | /* cut input data into smaller blocks */ | |
129 | bytes = (len > MAX_BYTES) ? MAX_BYTES : len; | |
130 | bytes = bytes & ~0x3f; | |
131 | ||
132 | spe_begin(); | |
133 | ppc_spe_sha256_transform(sctx->state, src, bytes >> 6); | |
134 | spe_end(); | |
135 | ||
136 | src += bytes; | |
137 | len -= bytes; | |
138 | }; | |
139 | ||
140 | memcpy((char *)sctx->buf, src, len); | |
141 | return 0; | |
142 | } | |
143 | ||
144 | static int ppc_spe_sha256_final(struct shash_desc *desc, u8 *out) | |
145 | { | |
146 | struct sha256_state *sctx = shash_desc_ctx(desc); | |
147 | const unsigned int offset = sctx->count & 0x3f; | |
148 | char *p = (char *)sctx->buf + offset; | |
149 | int padlen; | |
150 | __be64 *pbits = (__be64 *)(((char *)&sctx->buf) + 56); | |
151 | __be32 *dst = (__be32 *)out; | |
152 | ||
153 | padlen = 55 - offset; | |
154 | *p++ = 0x80; | |
155 | ||
156 | spe_begin(); | |
157 | ||
158 | if (padlen < 0) { | |
159 | memset(p, 0x00, padlen + sizeof (u64)); | |
160 | ppc_spe_sha256_transform(sctx->state, sctx->buf, 1); | |
161 | p = (char *)sctx->buf; | |
162 | padlen = 56; | |
163 | } | |
164 | ||
165 | memset(p, 0, padlen); | |
166 | *pbits = cpu_to_be64(sctx->count << 3); | |
167 | ppc_spe_sha256_transform(sctx->state, sctx->buf, 1); | |
168 | ||
169 | spe_end(); | |
170 | ||
171 | dst[0] = cpu_to_be32(sctx->state[0]); | |
172 | dst[1] = cpu_to_be32(sctx->state[1]); | |
173 | dst[2] = cpu_to_be32(sctx->state[2]); | |
174 | dst[3] = cpu_to_be32(sctx->state[3]); | |
175 | dst[4] = cpu_to_be32(sctx->state[4]); | |
176 | dst[5] = cpu_to_be32(sctx->state[5]); | |
177 | dst[6] = cpu_to_be32(sctx->state[6]); | |
178 | dst[7] = cpu_to_be32(sctx->state[7]); | |
179 | ||
180 | ppc_sha256_clear_context(sctx); | |
181 | return 0; | |
182 | } | |
183 | ||
184 | static int ppc_spe_sha224_final(struct shash_desc *desc, u8 *out) | |
185 | { | |
186 | u32 D[SHA256_DIGEST_SIZE >> 2]; | |
187 | __be32 *dst = (__be32 *)out; | |
188 | ||
189 | ppc_spe_sha256_final(desc, (u8 *)D); | |
190 | ||
191 | /* avoid bytewise memcpy */ | |
192 | dst[0] = D[0]; | |
193 | dst[1] = D[1]; | |
194 | dst[2] = D[2]; | |
195 | dst[3] = D[3]; | |
196 | dst[4] = D[4]; | |
197 | dst[5] = D[5]; | |
198 | dst[6] = D[6]; | |
199 | ||
200 | /* clear sensitive data */ | |
201 | memzero_explicit(D, SHA256_DIGEST_SIZE); | |
202 | return 0; | |
203 | } | |
204 | ||
205 | static int ppc_spe_sha256_export(struct shash_desc *desc, void *out) | |
206 | { | |
207 | struct sha256_state *sctx = shash_desc_ctx(desc); | |
208 | ||
209 | memcpy(out, sctx, sizeof(*sctx)); | |
210 | return 0; | |
211 | } | |
212 | ||
213 | static int ppc_spe_sha256_import(struct shash_desc *desc, const void *in) | |
214 | { | |
215 | struct sha256_state *sctx = shash_desc_ctx(desc); | |
216 | ||
217 | memcpy(sctx, in, sizeof(*sctx)); | |
218 | return 0; | |
219 | } | |
220 | ||
221 | static struct shash_alg algs[2] = { { | |
222 | .digestsize = SHA256_DIGEST_SIZE, | |
223 | .init = ppc_spe_sha256_init, | |
224 | .update = ppc_spe_sha256_update, | |
225 | .final = ppc_spe_sha256_final, | |
226 | .export = ppc_spe_sha256_export, | |
227 | .import = ppc_spe_sha256_import, | |
228 | .descsize = sizeof(struct sha256_state), | |
229 | .statesize = sizeof(struct sha256_state), | |
230 | .base = { | |
231 | .cra_name = "sha256", | |
232 | .cra_driver_name= "sha256-ppc-spe", | |
233 | .cra_priority = 300, | |
234 | .cra_flags = CRYPTO_ALG_TYPE_SHASH, | |
235 | .cra_blocksize = SHA256_BLOCK_SIZE, | |
236 | .cra_module = THIS_MODULE, | |
237 | } | |
238 | }, { | |
239 | .digestsize = SHA224_DIGEST_SIZE, | |
240 | .init = ppc_spe_sha224_init, | |
241 | .update = ppc_spe_sha256_update, | |
242 | .final = ppc_spe_sha224_final, | |
243 | .export = ppc_spe_sha256_export, | |
244 | .import = ppc_spe_sha256_import, | |
245 | .descsize = sizeof(struct sha256_state), | |
246 | .statesize = sizeof(struct sha256_state), | |
247 | .base = { | |
248 | .cra_name = "sha224", | |
249 | .cra_driver_name= "sha224-ppc-spe", | |
250 | .cra_priority = 300, | |
251 | .cra_flags = CRYPTO_ALG_TYPE_SHASH, | |
252 | .cra_blocksize = SHA224_BLOCK_SIZE, | |
253 | .cra_module = THIS_MODULE, | |
254 | } | |
255 | } }; | |
256 | ||
257 | static int __init ppc_spe_sha256_mod_init(void) | |
258 | { | |
259 | return crypto_register_shashes(algs, ARRAY_SIZE(algs)); | |
260 | } | |
261 | ||
262 | static void __exit ppc_spe_sha256_mod_fini(void) | |
263 | { | |
264 | crypto_unregister_shashes(algs, ARRAY_SIZE(algs)); | |
265 | } | |
266 | ||
267 | module_init(ppc_spe_sha256_mod_init); | |
268 | module_exit(ppc_spe_sha256_mod_fini); | |
269 | ||
270 | MODULE_LICENSE("GPL"); | |
271 | MODULE_DESCRIPTION("SHA-224 and SHA-256 Secure Hash Algorithm, SPE optimized"); | |
272 | ||
273 | MODULE_ALIAS_CRYPTO("sha224"); | |
274 | MODULE_ALIAS_CRYPTO("sha224-ppc-spe"); | |
275 | MODULE_ALIAS_CRYPTO("sha256"); | |
276 | MODULE_ALIAS_CRYPTO("sha256-ppc-spe"); |