1 | // SPDX-License-Identifier: GPL-2.0-or-later |
2 | /* |
3 | * Glue code for SHA-1 implementation for SPE instructions (PPC) |
4 | * |
5 | * Based on generic implementation. |
6 | * |
7 | * Copyright (c) 2015 Markus Stockhausen <stockhausen@collogia.de> |
8 | */ |
9 | |
10 | #include <crypto/internal/hash.h> |
11 | #include <linux/init.h> |
12 | #include <linux/module.h> |
13 | #include <linux/mm.h> |
14 | #include <linux/types.h> |
15 | #include <crypto/sha1.h> |
16 | #include <crypto/sha1_base.h> |
17 | #include <asm/byteorder.h> |
18 | #include <asm/switch_to.h> |
19 | #include <linux/hardirq.h> |
20 | |
21 | /* |
22 | * MAX_BYTES defines the number of bytes that are allowed to be processed |
23 | * between preempt_disable() and preempt_enable(). SHA1 takes ~1000 |
24 | * operations per 64 bytes. e500 cores can issue two arithmetic instructions |
25 | * per clock cycle using one 32/64 bit unit (SU1) and one 32 bit unit (SU2). |
26 | * Thus 2KB of input data will need an estimated maximum of 18,000 cycles. |
27 | * Headroom for cache misses included. Even with the low end model clocked |
28 | * at 667 MHz this equals to a critical time window of less than 27us. |
29 | * |
30 | */ |
31 | #define MAX_BYTES 2048 |
32 | |
33 | extern void ppc_spe_sha1_transform(u32 *state, const u8 *src, u32 blocks); |
34 | |
35 | static void spe_begin(void) |
36 | { |
37 | /* We just start SPE operations and will save SPE registers later. */ |
38 | preempt_disable(); |
39 | enable_kernel_spe(); |
40 | } |
41 | |
42 | static void spe_end(void) |
43 | { |
44 | disable_kernel_spe(); |
45 | /* reenable preemption */ |
46 | preempt_enable(); |
47 | } |
48 | |
49 | static inline void ppc_sha1_clear_context(struct sha1_state *sctx) |
50 | { |
51 | int count = sizeof(struct sha1_state) >> 2; |
52 | u32 *ptr = (u32 *)sctx; |
53 | |
54 | /* make sure we can clear the fast way */ |
55 | BUILD_BUG_ON(sizeof(struct sha1_state) % 4); |
56 | do { *ptr++ = 0; } while (--count); |
57 | } |
58 | |
59 | static int ppc_spe_sha1_update(struct shash_desc *desc, const u8 *data, |
60 | unsigned int len) |
61 | { |
62 | struct sha1_state *sctx = shash_desc_ctx(desc); |
63 | const unsigned int offset = sctx->count & 0x3f; |
64 | const unsigned int avail = 64 - offset; |
65 | unsigned int bytes; |
66 | const u8 *src = data; |
67 | |
68 | if (avail > len) { |
69 | sctx->count += len; |
70 | memcpy((char *)sctx->buffer + offset, src, len); |
71 | return 0; |
72 | } |
73 | |
74 | sctx->count += len; |
75 | |
76 | if (offset) { |
77 | memcpy((char *)sctx->buffer + offset, src, avail); |
78 | |
79 | spe_begin(); |
80 | ppc_spe_sha1_transform(state: sctx->state, src: (const u8 *)sctx->buffer, blocks: 1); |
81 | spe_end(); |
82 | |
83 | len -= avail; |
84 | src += avail; |
85 | } |
86 | |
87 | while (len > 63) { |
88 | bytes = (len > MAX_BYTES) ? MAX_BYTES : len; |
89 | bytes = bytes & ~0x3f; |
90 | |
91 | spe_begin(); |
92 | ppc_spe_sha1_transform(state: sctx->state, src, blocks: bytes >> 6); |
93 | spe_end(); |
94 | |
95 | src += bytes; |
96 | len -= bytes; |
97 | } |
98 | |
99 | memcpy((char *)sctx->buffer, src, len); |
100 | return 0; |
101 | } |
102 | |
103 | static int ppc_spe_sha1_final(struct shash_desc *desc, u8 *out) |
104 | { |
105 | struct sha1_state *sctx = shash_desc_ctx(desc); |
106 | const unsigned int offset = sctx->count & 0x3f; |
107 | char *p = (char *)sctx->buffer + offset; |
108 | int padlen; |
109 | __be64 *pbits = (__be64 *)(((char *)&sctx->buffer) + 56); |
110 | __be32 *dst = (__be32 *)out; |
111 | |
112 | padlen = 55 - offset; |
113 | *p++ = 0x80; |
114 | |
115 | spe_begin(); |
116 | |
117 | if (padlen < 0) { |
118 | memset(p, 0x00, padlen + sizeof (u64)); |
119 | ppc_spe_sha1_transform(state: sctx->state, src: sctx->buffer, blocks: 1); |
120 | p = (char *)sctx->buffer; |
121 | padlen = 56; |
122 | } |
123 | |
124 | memset(p, 0, padlen); |
125 | *pbits = cpu_to_be64(sctx->count << 3); |
126 | ppc_spe_sha1_transform(state: sctx->state, src: sctx->buffer, blocks: 1); |
127 | |
128 | spe_end(); |
129 | |
130 | dst[0] = cpu_to_be32(sctx->state[0]); |
131 | dst[1] = cpu_to_be32(sctx->state[1]); |
132 | dst[2] = cpu_to_be32(sctx->state[2]); |
133 | dst[3] = cpu_to_be32(sctx->state[3]); |
134 | dst[4] = cpu_to_be32(sctx->state[4]); |
135 | |
136 | ppc_sha1_clear_context(sctx); |
137 | return 0; |
138 | } |
139 | |
140 | static int ppc_spe_sha1_export(struct shash_desc *desc, void *out) |
141 | { |
142 | struct sha1_state *sctx = shash_desc_ctx(desc); |
143 | |
144 | memcpy(out, sctx, sizeof(*sctx)); |
145 | return 0; |
146 | } |
147 | |
148 | static int ppc_spe_sha1_import(struct shash_desc *desc, const void *in) |
149 | { |
150 | struct sha1_state *sctx = shash_desc_ctx(desc); |
151 | |
152 | memcpy(sctx, in, sizeof(*sctx)); |
153 | return 0; |
154 | } |
155 | |
156 | static struct shash_alg alg = { |
157 | .digestsize = SHA1_DIGEST_SIZE, |
158 | .init = sha1_base_init, |
159 | .update = ppc_spe_sha1_update, |
160 | .final = ppc_spe_sha1_final, |
161 | .export = ppc_spe_sha1_export, |
162 | .import = ppc_spe_sha1_import, |
163 | .descsize = sizeof(struct sha1_state), |
164 | .statesize = sizeof(struct sha1_state), |
165 | .base = { |
166 | .cra_name = "sha1" , |
167 | .cra_driver_name= "sha1-ppc-spe" , |
168 | .cra_priority = 300, |
169 | .cra_blocksize = SHA1_BLOCK_SIZE, |
170 | .cra_module = THIS_MODULE, |
171 | } |
172 | }; |
173 | |
174 | static int __init ppc_spe_sha1_mod_init(void) |
175 | { |
176 | return crypto_register_shash(alg: &alg); |
177 | } |
178 | |
179 | static void __exit ppc_spe_sha1_mod_fini(void) |
180 | { |
181 | crypto_unregister_shash(alg: &alg); |
182 | } |
183 | |
184 | module_init(ppc_spe_sha1_mod_init); |
185 | module_exit(ppc_spe_sha1_mod_fini); |
186 | |
187 | MODULE_LICENSE("GPL" ); |
188 | MODULE_DESCRIPTION("SHA1 Secure Hash Algorithm, SPE optimized" ); |
189 | |
190 | MODULE_ALIAS_CRYPTO("sha1" ); |
191 | MODULE_ALIAS_CRYPTO("sha1-ppc-spe" ); |
192 | |