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
39extern void ppc_spe_sha256_transform(u32 *state, const u8 *src, u32 blocks);
40
41static 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
48static void spe_end(void)
49{
50	/* reenable preemption */
51	preempt_enable();
52}
53
54static inline void ppc_sha256_clear_context(struct sha256_state *sctx)
55{
56	int count = sizeof(struct sha256_state) >> 2;
57	u32 *ptr = (u32 *)sctx;
58
59	/* make sure we can clear the fast way */
60	BUILD_BUG_ON(sizeof(struct sha256_state) % 4);
61	do { *ptr++ = 0; } while (--count);
62}
63
64static int ppc_spe_sha256_init(struct shash_desc *desc)
65{
66	struct sha256_state *sctx = shash_desc_ctx(desc);
67
68	sctx->state[0] = SHA256_H0;
69	sctx->state[1] = SHA256_H1;
70	sctx->state[2] = SHA256_H2;
71	sctx->state[3] = SHA256_H3;
72	sctx->state[4] = SHA256_H4;
73	sctx->state[5] = SHA256_H5;
74	sctx->state[6] = SHA256_H6;
75	sctx->state[7] = SHA256_H7;
76	sctx->count = 0;
77
78	return 0;
79}
80
81static int ppc_spe_sha224_init(struct shash_desc *desc)
82{
83	struct sha256_state *sctx = shash_desc_ctx(desc);
84
85	sctx->state[0] = SHA224_H0;
86	sctx->state[1] = SHA224_H1;
87	sctx->state[2] = SHA224_H2;
88	sctx->state[3] = SHA224_H3;
89	sctx->state[4] = SHA224_H4;
90	sctx->state[5] = SHA224_H5;
91	sctx->state[6] = SHA224_H6;
92	sctx->state[7] = SHA224_H7;
93	sctx->count = 0;
94
95	return 0;
96}
97
98static int ppc_spe_sha256_update(struct shash_desc *desc, const u8 *data,
99			unsigned int len)
100{
101	struct sha256_state *sctx = shash_desc_ctx(desc);
102	const unsigned int offset = sctx->count & 0x3f;
103	const unsigned int avail = 64 - offset;
104	unsigned int bytes;
105	const u8 *src = data;
106
107	if (avail > len) {
108		sctx->count += len;
109		memcpy((char *)sctx->buf + offset, src, len);
110		return 0;
111	}
112
113	sctx->count += len;
114
115	if (offset) {
116		memcpy((char *)sctx->buf + offset, src, avail);
117
118		spe_begin();
119		ppc_spe_sha256_transform(sctx->state, (const u8 *)sctx->buf, 1);
120		spe_end();
121
122		len -= avail;
123		src += avail;
124	}
125
126	while (len > 63) {
127		/* cut input data into smaller blocks */
128		bytes = (len > MAX_BYTES) ? MAX_BYTES : len;
129		bytes = bytes & ~0x3f;
130
131		spe_begin();
132		ppc_spe_sha256_transform(sctx->state, src, bytes >> 6);
133		spe_end();
134
135		src += bytes;
136		len -= bytes;
137	};
138
139	memcpy((char *)sctx->buf, src, len);
140	return 0;
141}
142
143static int ppc_spe_sha256_final(struct shash_desc *desc, u8 *out)
144{
145	struct sha256_state *sctx = shash_desc_ctx(desc);
146	const unsigned int offset = sctx->count & 0x3f;
147	char *p = (char *)sctx->buf + offset;
148	int padlen;
149	__be64 *pbits = (__be64 *)(((char *)&sctx->buf) + 56);
150	__be32 *dst = (__be32 *)out;
151
152	padlen = 55 - offset;
153	*p++ = 0x80;
154
155	spe_begin();
156
157	if (padlen < 0) {
158		memset(p, 0x00, padlen + sizeof (u64));
159		ppc_spe_sha256_transform(sctx->state, sctx->buf, 1);
160		p = (char *)sctx->buf;
161		padlen = 56;
162	}
163
164	memset(p, 0, padlen);
165	*pbits = cpu_to_be64(sctx->count << 3);
166	ppc_spe_sha256_transform(sctx->state, sctx->buf, 1);
167
168	spe_end();
169
170	dst[0] = cpu_to_be32(sctx->state[0]);
171	dst[1] = cpu_to_be32(sctx->state[1]);
172	dst[2] = cpu_to_be32(sctx->state[2]);
173	dst[3] = cpu_to_be32(sctx->state[3]);
174	dst[4] = cpu_to_be32(sctx->state[4]);
175	dst[5] = cpu_to_be32(sctx->state[5]);
176	dst[6] = cpu_to_be32(sctx->state[6]);
177	dst[7] = cpu_to_be32(sctx->state[7]);
178
179	ppc_sha256_clear_context(sctx);
180	return 0;
181}
182
183static int ppc_spe_sha224_final(struct shash_desc *desc, u8 *out)
184{
185	u32 D[SHA256_DIGEST_SIZE >> 2];
186	__be32 *dst = (__be32 *)out;
187
188	ppc_spe_sha256_final(desc, (u8 *)D);
189
190	/* avoid bytewise memcpy */
191	dst[0] = D[0];
192	dst[1] = D[1];
193	dst[2] = D[2];
194	dst[3] = D[3];
195	dst[4] = D[4];
196	dst[5] = D[5];
197	dst[6] = D[6];
198
199	/* clear sensitive data */
200	memzero_explicit(D, SHA256_DIGEST_SIZE);
201	return 0;
202}
203
204static int ppc_spe_sha256_export(struct shash_desc *desc, void *out)
205{
206	struct sha256_state *sctx = shash_desc_ctx(desc);
207
208	memcpy(out, sctx, sizeof(*sctx));
209	return 0;
210}
211
212static int ppc_spe_sha256_import(struct shash_desc *desc, const void *in)
213{
214	struct sha256_state *sctx = shash_desc_ctx(desc);
215
216	memcpy(sctx, in, sizeof(*sctx));
217	return 0;
218}
219
220static struct shash_alg algs[2] = { {
221	.digestsize	=	SHA256_DIGEST_SIZE,
222	.init		=	ppc_spe_sha256_init,
223	.update		=	ppc_spe_sha256_update,
224	.final		=	ppc_spe_sha256_final,
225	.export		=	ppc_spe_sha256_export,
226	.import		=	ppc_spe_sha256_import,
227	.descsize	=	sizeof(struct sha256_state),
228	.statesize	=	sizeof(struct sha256_state),
229	.base		=	{
230		.cra_name	=	"sha256",
231		.cra_driver_name=	"sha256-ppc-spe",
232		.cra_priority	=	300,
233		.cra_flags	=	CRYPTO_ALG_TYPE_SHASH,
234		.cra_blocksize	=	SHA256_BLOCK_SIZE,
235		.cra_module	=	THIS_MODULE,
236	}
237}, {
238	.digestsize	=	SHA224_DIGEST_SIZE,
239	.init		=	ppc_spe_sha224_init,
240	.update		=	ppc_spe_sha256_update,
241	.final		=	ppc_spe_sha224_final,
242	.export		=	ppc_spe_sha256_export,
243	.import		=	ppc_spe_sha256_import,
244	.descsize	=	sizeof(struct sha256_state),
245	.statesize	=	sizeof(struct sha256_state),
246	.base		=	{
247		.cra_name	=	"sha224",
248		.cra_driver_name=	"sha224-ppc-spe",
249		.cra_priority	=	300,
250		.cra_flags	=	CRYPTO_ALG_TYPE_SHASH,
251		.cra_blocksize	=	SHA224_BLOCK_SIZE,
252		.cra_module	=	THIS_MODULE,
253	}
254} };
255
256static int __init ppc_spe_sha256_mod_init(void)
257{
258	return crypto_register_shashes(algs, ARRAY_SIZE(algs));
259}
260
261static void __exit ppc_spe_sha256_mod_fini(void)
262{
263	crypto_unregister_shashes(algs, ARRAY_SIZE(algs));
264}
265
266module_init(ppc_spe_sha256_mod_init);
267module_exit(ppc_spe_sha256_mod_fini);
268
269MODULE_LICENSE("GPL");
270MODULE_DESCRIPTION("SHA-224 and SHA-256 Secure Hash Algorithm, SPE optimized");
271
272MODULE_ALIAS_CRYPTO("sha224");
273MODULE_ALIAS_CRYPTO("sha224-ppc-spe");
274MODULE_ALIAS_CRYPTO("sha256");
275MODULE_ALIAS_CRYPTO("sha256-ppc-spe");
276