2 * SHA-512 routines supporting the Power 7+ Nest Accelerators driver
4 * Copyright (C) 2011-2012 International Business Machines Inc.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; version 2 only.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
19 * Author: Kent Yoder <yoder1@us.ibm.com>
22 #include <crypto/internal/hash.h>
23 #include <crypto/sha.h>
24 #include <linux/module.h>
27 #include "nx_csbcpb.h"
31 static int nx_crypto_ctx_sha512_init(struct crypto_tfm
*tfm
)
33 struct nx_crypto_ctx
*nx_ctx
= crypto_tfm_ctx(tfm
);
36 err
= nx_crypto_ctx_sha_init(tfm
);
40 nx_ctx_init(nx_ctx
, HCOP_FC_SHA
);
42 nx_ctx
->ap
= &nx_ctx
->props
[NX_PROPS_SHA512
];
44 NX_CPB_SET_DIGEST_SIZE(nx_ctx
->csbcpb
, NX_DS_SHA512
);
49 static int nx_sha512_init(struct shash_desc
*desc
)
51 struct sha512_state
*sctx
= shash_desc_ctx(desc
);
53 memset(sctx
, 0, sizeof *sctx
);
55 sctx
->state
[0] = __cpu_to_be64(SHA512_H0
);
56 sctx
->state
[1] = __cpu_to_be64(SHA512_H1
);
57 sctx
->state
[2] = __cpu_to_be64(SHA512_H2
);
58 sctx
->state
[3] = __cpu_to_be64(SHA512_H3
);
59 sctx
->state
[4] = __cpu_to_be64(SHA512_H4
);
60 sctx
->state
[5] = __cpu_to_be64(SHA512_H5
);
61 sctx
->state
[6] = __cpu_to_be64(SHA512_H6
);
62 sctx
->state
[7] = __cpu_to_be64(SHA512_H7
);
68 static int nx_sha512_update(struct shash_desc
*desc
, const u8
*data
,
71 struct sha512_state
*sctx
= shash_desc_ctx(desc
);
72 struct nx_crypto_ctx
*nx_ctx
= crypto_tfm_ctx(&desc
->tfm
->base
);
73 struct nx_csbcpb
*csbcpb
= (struct nx_csbcpb
*)nx_ctx
->csbcpb
;
75 u64 to_process
, leftover
= 0, total
;
76 unsigned long irq_flags
;
80 u64 buf_len
= (sctx
->count
[0] % SHA512_BLOCK_SIZE
);
82 spin_lock_irqsave(&nx_ctx
->lock
, irq_flags
);
84 /* 2 cases for total data len:
85 * 1: < SHA512_BLOCK_SIZE: copy into state, return 0
86 * 2: >= SHA512_BLOCK_SIZE: process X blocks, copy in leftover
88 total
= (sctx
->count
[0] % SHA512_BLOCK_SIZE
) + len
;
89 if (total
< SHA512_BLOCK_SIZE
) {
90 memcpy(sctx
->buf
+ buf_len
, data
, len
);
91 sctx
->count
[0] += len
;
95 memcpy(csbcpb
->cpb
.sha512
.message_digest
, sctx
->state
, SHA512_DIGEST_SIZE
);
96 NX_CPB_FDM(csbcpb
) |= NX_FDM_INTERMEDIATE
;
97 NX_CPB_FDM(csbcpb
) |= NX_FDM_CONTINUATION
;
99 max_sg_len
= min_t(u64
, nx_ctx
->ap
->sglen
,
100 nx_driver
.of
.max_sg_len
/sizeof(struct nx_sg
));
101 max_sg_len
= min_t(u64
, max_sg_len
,
102 nx_ctx
->ap
->databytelen
/NX_PAGE_SIZE
);
104 data_len
= SHA512_DIGEST_SIZE
;
105 out_sg
= nx_build_sg_list(nx_ctx
->out_sg
, (u8
*)sctx
->state
,
106 &data_len
, max_sg_len
);
107 nx_ctx
->op
.outlen
= (nx_ctx
->out_sg
- out_sg
) * sizeof(struct nx_sg
);
109 if (data_len
!= SHA512_DIGEST_SIZE
) {
116 struct nx_sg
*in_sg
= nx_ctx
->in_sg
;
120 in_sg
= nx_build_sg_list(in_sg
,
122 &data_len
, max_sg_len
);
124 if (data_len
!= buf_len
) {
128 used_sgs
= in_sg
- nx_ctx
->in_sg
;
131 /* to_process: SHA512_BLOCK_SIZE aligned chunk to be
132 * processed in this iteration. This value is restricted
133 * by sg list limits and number of sgs we already used
134 * for leftover data. (see above)
135 * In ideal case, we could allow NX_PAGE_SIZE * max_sg_len,
136 * but because data may not be aligned, we need to account
138 to_process
= min_t(u64
, total
,
139 (max_sg_len
- 1 - used_sgs
) * NX_PAGE_SIZE
);
140 to_process
= to_process
& ~(SHA512_BLOCK_SIZE
- 1);
142 data_len
= to_process
- buf_len
;
143 in_sg
= nx_build_sg_list(in_sg
, (u8
*) data
,
144 &data_len
, max_sg_len
);
146 nx_ctx
->op
.inlen
= (nx_ctx
->in_sg
- in_sg
) * sizeof(struct nx_sg
);
148 if (data_len
!= (to_process
- buf_len
)) {
153 to_process
= data_len
+ buf_len
;
154 leftover
= total
- to_process
;
157 * we've hit the nx chip previously and we're updating
158 * again, so copy over the partial digest.
160 memcpy(csbcpb
->cpb
.sha512
.input_partial_digest
,
161 csbcpb
->cpb
.sha512
.message_digest
,
164 if (!nx_ctx
->op
.inlen
|| !nx_ctx
->op
.outlen
) {
169 rc
= nx_hcall_sync(nx_ctx
, &nx_ctx
->op
,
170 desc
->flags
& CRYPTO_TFM_REQ_MAY_SLEEP
);
174 atomic_inc(&(nx_ctx
->stats
->sha512_ops
));
177 data
+= to_process
- buf_len
;
180 } while (leftover
>= SHA512_BLOCK_SIZE
);
182 /* copy the leftover back into the state struct */
184 memcpy(sctx
->buf
, data
, leftover
);
185 sctx
->count
[0] += len
;
186 memcpy(sctx
->state
, csbcpb
->cpb
.sha512
.message_digest
, SHA512_DIGEST_SIZE
);
188 spin_unlock_irqrestore(&nx_ctx
->lock
, irq_flags
);
192 static int nx_sha512_final(struct shash_desc
*desc
, u8
*out
)
194 struct sha512_state
*sctx
= shash_desc_ctx(desc
);
195 struct nx_crypto_ctx
*nx_ctx
= crypto_tfm_ctx(&desc
->tfm
->base
);
196 struct nx_csbcpb
*csbcpb
= (struct nx_csbcpb
*)nx_ctx
->csbcpb
;
197 struct nx_sg
*in_sg
, *out_sg
;
200 unsigned long irq_flags
;
204 spin_lock_irqsave(&nx_ctx
->lock
, irq_flags
);
206 max_sg_len
= min_t(u64
, nx_ctx
->ap
->sglen
,
207 nx_driver
.of
.max_sg_len
/sizeof(struct nx_sg
));
208 max_sg_len
= min_t(u64
, max_sg_len
,
209 nx_ctx
->ap
->databytelen
/NX_PAGE_SIZE
);
211 /* final is represented by continuing the operation and indicating that
212 * this is not an intermediate operation */
213 if (sctx
->count
[0] >= SHA512_BLOCK_SIZE
) {
214 /* we've hit the nx chip previously, now we're finalizing,
215 * so copy over the partial digest */
216 memcpy(csbcpb
->cpb
.sha512
.input_partial_digest
, sctx
->state
,
218 NX_CPB_FDM(csbcpb
) &= ~NX_FDM_INTERMEDIATE
;
219 NX_CPB_FDM(csbcpb
) |= NX_FDM_CONTINUATION
;
221 NX_CPB_FDM(csbcpb
) &= ~NX_FDM_INTERMEDIATE
;
222 NX_CPB_FDM(csbcpb
) &= ~NX_FDM_CONTINUATION
;
225 NX_CPB_FDM(csbcpb
) &= ~NX_FDM_INTERMEDIATE
;
227 count0
= sctx
->count
[0] * 8;
229 csbcpb
->cpb
.sha512
.message_bit_length_lo
= count0
;
231 len
= sctx
->count
[0] & (SHA512_BLOCK_SIZE
- 1);
232 in_sg
= nx_build_sg_list(nx_ctx
->in_sg
, sctx
->buf
, &len
,
235 if (len
!= (sctx
->count
[0] & (SHA512_BLOCK_SIZE
- 1))) {
240 len
= SHA512_DIGEST_SIZE
;
241 out_sg
= nx_build_sg_list(nx_ctx
->out_sg
, out
, &len
,
244 nx_ctx
->op
.inlen
= (nx_ctx
->in_sg
- in_sg
) * sizeof(struct nx_sg
);
245 nx_ctx
->op
.outlen
= (nx_ctx
->out_sg
- out_sg
) * sizeof(struct nx_sg
);
247 if (!nx_ctx
->op
.outlen
) {
252 rc
= nx_hcall_sync(nx_ctx
, &nx_ctx
->op
,
253 desc
->flags
& CRYPTO_TFM_REQ_MAY_SLEEP
);
257 atomic_inc(&(nx_ctx
->stats
->sha512_ops
));
258 atomic64_add(sctx
->count
[0], &(nx_ctx
->stats
->sha512_bytes
));
260 memcpy(out
, csbcpb
->cpb
.sha512
.message_digest
, SHA512_DIGEST_SIZE
);
262 spin_unlock_irqrestore(&nx_ctx
->lock
, irq_flags
);
266 static int nx_sha512_export(struct shash_desc
*desc
, void *out
)
268 struct sha512_state
*sctx
= shash_desc_ctx(desc
);
270 memcpy(out
, sctx
, sizeof(*sctx
));
275 static int nx_sha512_import(struct shash_desc
*desc
, const void *in
)
277 struct sha512_state
*sctx
= shash_desc_ctx(desc
);
279 memcpy(sctx
, in
, sizeof(*sctx
));
284 struct shash_alg nx_shash_sha512_alg
= {
285 .digestsize
= SHA512_DIGEST_SIZE
,
286 .init
= nx_sha512_init
,
287 .update
= nx_sha512_update
,
288 .final
= nx_sha512_final
,
289 .export
= nx_sha512_export
,
290 .import
= nx_sha512_import
,
291 .descsize
= sizeof(struct sha512_state
),
292 .statesize
= sizeof(struct sha512_state
),
294 .cra_name
= "sha512",
295 .cra_driver_name
= "sha512-nx",
297 .cra_blocksize
= SHA512_BLOCK_SIZE
,
298 .cra_module
= THIS_MODULE
,
299 .cra_ctxsize
= sizeof(struct nx_crypto_ctx
),
300 .cra_init
= nx_crypto_ctx_sha512_init
,
301 .cra_exit
= nx_crypto_ctx_exit
,