2 * Copyright (c) 2006 Pawel Jakub Dawidek <pjd@FreeBSD.org>
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * $FreeBSD: src/sys/crypto/via/padlock_hash.c,v 1.4 2009/05/27 09:52:12 vanhu Exp $
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/kernel.h>
32 #include <sys/module.h>
33 #include <sys/malloc.h>
34 #include <sys/libkern.h>
35 #include <sys/endian.h>
36 #if defined(__x86_64__) || defined(__i386__)
37 #include <machine/cpufunc.h>
38 #include <machine/cputypes.h>
39 #include <machine/md_var.h>
40 #include <machine/specialreg.h>
43 #include <opencrypto/cryptodev.h>
44 #include <opencrypto/cryptosoft.h> /* for hmac_ipad_buffer and hmac_opad_buffer */
45 #include <opencrypto/xform.h>
47 #include <crypto/via/padlock.h>
50 * Implementation notes.
52 * Some VIA CPUs provides SHA1 and SHA256 acceleration.
53 * We implement all HMAC algorithms provided by crypto(9) framework, but we do
54 * the crypto work in software unless this is HMAC/SHA1 or HMAC/SHA256 and
55 * our CPU can accelerate it.
57 * Additional CPU instructions, which preform SHA1 and SHA256 are one-shot
58 * functions - we have only one chance to give the data, CPU itself will add
59 * the padding and calculate hash automatically.
60 * This means, it is not possible to implement common init(), update(), final()
62 * The way I've choosen is to keep adding data to the buffer on update()
63 * (reallocating the buffer if necessary) and call XSHA{1,256} instruction on
67 struct padlock_sha_ctx {
72 CTASSERT(sizeof(struct padlock_sha_ctx) <= sizeof(union authctx));
74 static void padlock_sha_init(struct padlock_sha_ctx *ctx);
75 static int padlock_sha_update(struct padlock_sha_ctx *ctx, uint8_t *buf,
77 static void padlock_sha1_final(uint8_t *hash, struct padlock_sha_ctx *ctx);
78 static void padlock_sha256_final(uint8_t *hash, struct padlock_sha_ctx *ctx);
80 static struct auth_hash padlock_hmac_sha1 = {
81 CRYPTO_SHA1_HMAC, "HMAC-SHA1",
82 20, SHA1_HASH_LEN, SHA1_HMAC_BLOCK_LEN, sizeof(struct padlock_sha_ctx),
83 (void (*)(void *))padlock_sha_init,
84 (int (*)(void *, uint8_t *, uint16_t))padlock_sha_update,
85 (void (*)(uint8_t *, void *))padlock_sha1_final
88 static struct auth_hash padlock_hmac_sha256 = {
89 CRYPTO_SHA2_256_HMAC, "HMAC-SHA2-256",
90 32, SHA2_256_HASH_LEN, SHA2_256_HMAC_BLOCK_LEN, sizeof(struct padlock_sha_ctx),
91 (void (*)(void *))padlock_sha_init,
92 (int (*)(void *, uint8_t *, uint16_t))padlock_sha_update,
93 (void (*)(uint8_t *, void *))padlock_sha256_final
96 MALLOC_DECLARE(M_PADLOCK);
99 padlock_output_block(uint32_t *src, uint32_t *dst, size_t count)
103 *dst++ = bswap32(*src++);
107 padlock_do_sha1(const u_char *in, u_char *out, int count)
109 u_char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */
110 u_char *result = PADLOCK_ALIGN(buf);
112 ((uint32_t *)result)[0] = 0x67452301;
113 ((uint32_t *)result)[1] = 0xEFCDAB89;
114 ((uint32_t *)result)[2] = 0x98BADCFE;
115 ((uint32_t *)result)[3] = 0x10325476;
116 ((uint32_t *)result)[4] = 0xC3D2E1F0;
119 ".byte 0xf3, 0x0f, 0xa6, 0xc8" /* rep xsha1 */
120 : "+S"(in), "+D"(result)
124 padlock_output_block((uint32_t *)result, (uint32_t *)out,
125 SHA1_HASH_LEN / sizeof(uint32_t));
129 padlock_do_sha256(const char *in, char *out, int count)
131 char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */
132 char *result = PADLOCK_ALIGN(buf);
134 ((uint32_t *)result)[0] = 0x6A09E667;
135 ((uint32_t *)result)[1] = 0xBB67AE85;
136 ((uint32_t *)result)[2] = 0x3C6EF372;
137 ((uint32_t *)result)[3] = 0xA54FF53A;
138 ((uint32_t *)result)[4] = 0x510E527F;
139 ((uint32_t *)result)[5] = 0x9B05688C;
140 ((uint32_t *)result)[6] = 0x1F83D9AB;
141 ((uint32_t *)result)[7] = 0x5BE0CD19;
144 ".byte 0xf3, 0x0f, 0xa6, 0xd0" /* rep xsha256 */
145 : "+S"(in), "+D"(result)
149 padlock_output_block((uint32_t *)result, (uint32_t *)out,
150 SHA2_256_HASH_LEN / sizeof(uint32_t));
154 padlock_sha_init(struct padlock_sha_ctx *ctx)
163 padlock_sha_update(struct padlock_sha_ctx *ctx, uint8_t *buf, uint16_t bufsize)
166 if (ctx->psc_size - ctx->psc_offset < bufsize) {
167 ctx->psc_size = MAX(ctx->psc_size * 2, ctx->psc_size + bufsize);
168 ctx->psc_buf = krealloc(ctx->psc_buf, ctx->psc_size, M_PADLOCK,
170 if(ctx->psc_buf == NULL)
173 bcopy(buf, ctx->psc_buf + ctx->psc_offset, bufsize);
174 ctx->psc_offset += bufsize;
179 padlock_sha_free(struct padlock_sha_ctx *ctx)
182 if (ctx->psc_buf != NULL) {
183 //bzero(ctx->psc_buf, ctx->psc_size);
184 kfree(ctx->psc_buf, M_PADLOCK);
192 padlock_sha1_final(uint8_t *hash, struct padlock_sha_ctx *ctx)
195 padlock_do_sha1(ctx->psc_buf, hash, ctx->psc_offset);
196 padlock_sha_free(ctx);
200 padlock_sha256_final(uint8_t *hash, struct padlock_sha_ctx *ctx)
203 padlock_do_sha256(ctx->psc_buf, hash, ctx->psc_offset);
204 padlock_sha_free(ctx);
208 padlock_copy_ctx(struct auth_hash *axf, void *sctx, void *dctx)
211 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 &&
212 (axf->type == CRYPTO_SHA1_HMAC ||
213 axf->type == CRYPTO_SHA2_256_HMAC)) {
214 struct padlock_sha_ctx *spctx = sctx, *dpctx = dctx;
216 dpctx->psc_offset = spctx->psc_offset;
217 dpctx->psc_size = spctx->psc_size;
218 dpctx->psc_buf = kmalloc(dpctx->psc_size, M_PADLOCK, M_WAITOK);
219 bcopy(spctx->psc_buf, dpctx->psc_buf, dpctx->psc_size);
221 bcopy(sctx, dctx, axf->ctxsize);
226 padlock_free_ctx(struct auth_hash *axf, void *ctx)
229 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 &&
230 (axf->type == CRYPTO_SHA1_HMAC ||
231 axf->type == CRYPTO_SHA2_256_HMAC)) {
232 padlock_sha_free(ctx);
237 padlock_hash_key_setup(struct padlock_session *ses, caddr_t key, int klen)
239 struct auth_hash *axf;
246 * Try to free contexts before using them, because
247 * padlock_hash_key_setup() can be called twice - once from
248 * padlock_newsession() and again from padlock_process().
250 padlock_free_ctx(axf, ses->ses_ictx);
251 padlock_free_ctx(axf, ses->ses_octx);
253 for (i = 0; i < klen; i++)
254 key[i] ^= HMAC_IPAD_VAL;
256 axf->Init(ses->ses_ictx);
257 axf->Update(ses->ses_ictx, key, klen);
258 axf->Update(ses->ses_ictx, hmac_ipad_buffer, axf->blocksize - klen);
260 for (i = 0; i < klen; i++)
261 key[i] ^= (HMAC_IPAD_VAL ^ HMAC_OPAD_VAL);
263 axf->Init(ses->ses_octx);
264 axf->Update(ses->ses_octx, key, klen);
265 axf->Update(ses->ses_octx, hmac_opad_buffer, axf->blocksize - klen);
267 for (i = 0; i < klen; i++)
268 key[i] ^= HMAC_OPAD_VAL;
272 * Compute keyed-hash authenticator.
275 padlock_authcompute(struct padlock_session *ses, struct cryptodesc *crd,
276 caddr_t buf, int flags)
278 u_char hash[HASH_MAX_LEN];
279 struct auth_hash *axf;
285 padlock_copy_ctx(axf, ses->ses_ictx, &ctx);
286 error = crypto_apply(flags, buf, crd->crd_skip, crd->crd_len,
287 (int (*)(void *, void *, unsigned int))axf->Update, (caddr_t)&ctx);
289 padlock_free_ctx(axf, &ctx);
292 axf->Final(hash, &ctx);
294 padlock_copy_ctx(axf, ses->ses_octx, &ctx);
295 axf->Update(&ctx, hash, axf->hashsize);
296 axf->Final(hash, &ctx);
298 /* Inject the authentication data */
299 crypto_copyback(flags, buf, crd->crd_inject,
300 ses->ses_mlen == 0 ? axf->hashsize : ses->ses_mlen, hash);
305 padlock_hash_setup(struct padlock_session *ses, struct cryptoini *macini)
308 ses->ses_mlen = macini->cri_mlen;
310 /* Find software structure which describes HMAC algorithm. */
311 switch (macini->cri_alg) {
312 case CRYPTO_NULL_HMAC:
313 ses->ses_axf = &auth_hash_null;
315 case CRYPTO_MD5_HMAC:
316 ses->ses_axf = &auth_hash_hmac_md5;
318 case CRYPTO_SHA1_HMAC:
319 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0)
320 ses->ses_axf = &padlock_hmac_sha1;
322 ses->ses_axf = &auth_hash_hmac_sha1;
324 case CRYPTO_RIPEMD160_HMAC:
325 ses->ses_axf = &auth_hash_hmac_ripemd_160;
327 case CRYPTO_SHA2_256_HMAC:
328 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0)
329 ses->ses_axf = &padlock_hmac_sha256;
331 ses->ses_axf = &auth_hash_hmac_sha2_256;
333 case CRYPTO_SHA2_384_HMAC:
334 ses->ses_axf = &auth_hash_hmac_sha2_384;
336 case CRYPTO_SHA2_512_HMAC:
337 ses->ses_axf = &auth_hash_hmac_sha2_512;
341 /* Allocate memory for HMAC inner and outer contexts. */
342 ses->ses_ictx = kmalloc(ses->ses_axf->ctxsize, M_PADLOCK,
344 ses->ses_octx = kmalloc(ses->ses_axf->ctxsize, M_PADLOCK,
346 if (ses->ses_ictx == NULL || ses->ses_octx == NULL)
349 /* Setup key if given. */
350 if (macini->cri_key != NULL) {
351 padlock_hash_key_setup(ses, macini->cri_key,
358 padlock_hash_process(struct padlock_session *ses, struct cryptodesc *maccrd,
363 if ((maccrd->crd_flags & CRD_F_KEY_EXPLICIT) != 0)
364 padlock_hash_key_setup(ses, maccrd->crd_key, maccrd->crd_klen);
366 error = padlock_authcompute(ses, maccrd, crp->crp_buf, crp->crp_flags);
371 padlock_hash_free(struct padlock_session *ses)
374 if (ses->ses_ictx != NULL) {
375 padlock_free_ctx(ses->ses_axf, ses->ses_ictx);
376 bzero(ses->ses_ictx, ses->ses_axf->ctxsize);
377 kfree(ses->ses_ictx, M_PADLOCK);
378 ses->ses_ictx = NULL;
380 if (ses->ses_octx != NULL) {
381 padlock_free_ctx(ses->ses_axf, ses->ses_octx);
382 bzero(ses->ses_octx, ses->ses_axf->ctxsize);
383 kfree(ses->ses_octx, M_PADLOCK);
384 ses->ses_octx = NULL;