1 /*- 2 * Copyright (c) 2006 Pawel Jakub Dawidek <pjd@FreeBSD.org> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 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. 13 * 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 24 * SUCH DAMAGE. 25 * 26 * $FreeBSD: src/sys/crypto/via/padlock_hash.c,v 1.4 2009/05/27 09:52:12 vanhu Exp $ 27 */ 28 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> 41 #endif 42 43 #include <opencrypto/cryptodev.h> 44 #include <opencrypto/cryptosoft.h> /* for hmac_ipad_buffer and hmac_opad_buffer */ 45 #include <opencrypto/xform.h> 46 47 #include <dev/crypto/padlock/padlock.h> 48 49 /* 50 * Implementation notes. 51 * 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. 56 * 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() 61 * methods. 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 64 * final(). 65 */ 66 67 struct padlock_sha_ctx { 68 uint8_t *psc_buf; 69 int psc_offset; 70 int psc_size; 71 }; 72 CTASSERT(sizeof(struct padlock_sha_ctx) <= sizeof(union authctx)); 73 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, 76 uint16_t bufsize); 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); 79 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 NULL, NULL, 85 (int (*)(void *, uint8_t *, uint16_t))padlock_sha_update, 86 (void (*)(uint8_t *, void *))padlock_sha1_final 87 }; 88 89 static struct auth_hash padlock_hmac_sha256 = { 90 CRYPTO_SHA2_256_HMAC, "HMAC-SHA2-256", 91 32, SHA2_256_HASH_LEN, SHA2_256_HMAC_BLOCK_LEN, sizeof(struct padlock_sha_ctx), 92 (void (*)(void *))padlock_sha_init, 93 NULL, NULL, 94 (int (*)(void *, uint8_t *, uint16_t))padlock_sha_update, 95 (void (*)(uint8_t *, void *))padlock_sha256_final 96 }; 97 98 MALLOC_DECLARE(M_PADLOCK); 99 100 static __inline void 101 padlock_output_block(uint32_t *src, uint32_t *dst, size_t count) 102 { 103 104 while (count-- > 0) 105 *dst++ = bswap32(*src++); 106 } 107 108 static void 109 padlock_do_sha1(const u_char *in, u_char *out, int count) 110 { 111 u_char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */ 112 u_char *result = PADLOCK_ALIGN(buf); 113 114 ((uint32_t *)result)[0] = 0x67452301; 115 ((uint32_t *)result)[1] = 0xEFCDAB89; 116 ((uint32_t *)result)[2] = 0x98BADCFE; 117 ((uint32_t *)result)[3] = 0x10325476; 118 ((uint32_t *)result)[4] = 0xC3D2E1F0; 119 120 __asm __volatile( 121 ".byte 0xf3, 0x0f, 0xa6, 0xc8" /* rep xsha1 */ 122 : "+S"(in), "+D"(result) 123 : "c"(count), "a"(0) 124 ); 125 126 padlock_output_block((uint32_t *)result, (uint32_t *)out, 127 SHA1_HASH_LEN / sizeof(uint32_t)); 128 } 129 130 static void 131 padlock_do_sha256(const char *in, char *out, int count) 132 { 133 char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */ 134 char *result = PADLOCK_ALIGN(buf); 135 136 ((uint32_t *)result)[0] = 0x6A09E667; 137 ((uint32_t *)result)[1] = 0xBB67AE85; 138 ((uint32_t *)result)[2] = 0x3C6EF372; 139 ((uint32_t *)result)[3] = 0xA54FF53A; 140 ((uint32_t *)result)[4] = 0x510E527F; 141 ((uint32_t *)result)[5] = 0x9B05688C; 142 ((uint32_t *)result)[6] = 0x1F83D9AB; 143 ((uint32_t *)result)[7] = 0x5BE0CD19; 144 145 __asm __volatile( 146 ".byte 0xf3, 0x0f, 0xa6, 0xd0" /* rep xsha256 */ 147 : "+S"(in), "+D"(result) 148 : "c"(count), "a"(0) 149 ); 150 151 padlock_output_block((uint32_t *)result, (uint32_t *)out, 152 SHA2_256_HASH_LEN / sizeof(uint32_t)); 153 } 154 155 static void 156 padlock_sha_init(struct padlock_sha_ctx *ctx) 157 { 158 159 ctx->psc_buf = NULL; 160 ctx->psc_offset = 0; 161 ctx->psc_size = 0; 162 } 163 164 static int 165 padlock_sha_update(struct padlock_sha_ctx *ctx, uint8_t *buf, uint16_t bufsize) 166 { 167 168 if (ctx->psc_size - ctx->psc_offset < bufsize) { 169 ctx->psc_size = MAX(ctx->psc_size * 2, ctx->psc_size + bufsize); 170 ctx->psc_buf = krealloc(ctx->psc_buf, ctx->psc_size, M_PADLOCK, 171 M_NOWAIT); 172 if(ctx->psc_buf == NULL) 173 return (ENOMEM); 174 } 175 bcopy(buf, ctx->psc_buf + ctx->psc_offset, bufsize); 176 ctx->psc_offset += bufsize; 177 return (0); 178 } 179 180 static void 181 padlock_sha_free(struct padlock_sha_ctx *ctx) 182 { 183 184 if (ctx->psc_buf != NULL) { 185 //bzero(ctx->psc_buf, ctx->psc_size); 186 kfree(ctx->psc_buf, M_PADLOCK); 187 ctx->psc_buf = NULL; 188 ctx->psc_offset = 0; 189 ctx->psc_size = 0; 190 } 191 } 192 193 static void 194 padlock_sha1_final(uint8_t *hash, struct padlock_sha_ctx *ctx) 195 { 196 197 padlock_do_sha1(ctx->psc_buf, hash, ctx->psc_offset); 198 padlock_sha_free(ctx); 199 } 200 201 static void 202 padlock_sha256_final(uint8_t *hash, struct padlock_sha_ctx *ctx) 203 { 204 205 padlock_do_sha256(ctx->psc_buf, hash, ctx->psc_offset); 206 padlock_sha_free(ctx); 207 } 208 209 static void 210 padlock_copy_ctx(struct auth_hash *axf, void *sctx, void *dctx) 211 { 212 213 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 && 214 (axf->type == CRYPTO_SHA1_HMAC || 215 axf->type == CRYPTO_SHA2_256_HMAC)) { 216 struct padlock_sha_ctx *spctx = sctx, *dpctx = dctx; 217 218 dpctx->psc_offset = spctx->psc_offset; 219 dpctx->psc_size = spctx->psc_size; 220 dpctx->psc_buf = kmalloc(dpctx->psc_size, M_PADLOCK, M_WAITOK); 221 bcopy(spctx->psc_buf, dpctx->psc_buf, dpctx->psc_size); 222 } else { 223 bcopy(sctx, dctx, axf->ctxsize); 224 } 225 } 226 227 static void 228 padlock_free_ctx(struct auth_hash *axf, void *ctx) 229 { 230 231 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 && 232 (axf->type == CRYPTO_SHA1_HMAC || 233 axf->type == CRYPTO_SHA2_256_HMAC)) { 234 padlock_sha_free(ctx); 235 } 236 } 237 238 static void 239 padlock_hash_key_setup(struct padlock_session *ses, caddr_t key, int klen) 240 { 241 struct auth_hash *axf; 242 int i; 243 244 klen /= 8; 245 axf = ses->ses_axf; 246 247 /* 248 * Try to free contexts before using them, because 249 * padlock_hash_key_setup() can be called twice - once from 250 * padlock_newsession() and again from padlock_process(). 251 */ 252 padlock_free_ctx(axf, ses->ses_ictx); 253 padlock_free_ctx(axf, ses->ses_octx); 254 255 for (i = 0; i < klen; i++) 256 key[i] ^= HMAC_IPAD_VAL; 257 258 axf->Init(ses->ses_ictx); 259 axf->Update(ses->ses_ictx, key, klen); 260 axf->Update(ses->ses_ictx, hmac_ipad_buffer, axf->blocksize - klen); 261 262 for (i = 0; i < klen; i++) 263 key[i] ^= (HMAC_IPAD_VAL ^ HMAC_OPAD_VAL); 264 265 axf->Init(ses->ses_octx); 266 axf->Update(ses->ses_octx, key, klen); 267 axf->Update(ses->ses_octx, hmac_opad_buffer, axf->blocksize - klen); 268 269 for (i = 0; i < klen; i++) 270 key[i] ^= HMAC_OPAD_VAL; 271 } 272 273 /* 274 * Compute keyed-hash authenticator. 275 */ 276 static int 277 padlock_authcompute(struct padlock_session *ses, struct cryptodesc *crd, 278 caddr_t buf, int flags) 279 { 280 u_char hash[HASH_MAX_LEN]; 281 struct auth_hash *axf; 282 union authctx ctx; 283 int error; 284 285 axf = ses->ses_axf; 286 287 padlock_copy_ctx(axf, ses->ses_ictx, &ctx); 288 error = crypto_apply(flags, buf, crd->crd_skip, crd->crd_len, 289 (int (*)(void *, void *, unsigned int))axf->Update, (caddr_t)&ctx); 290 if (error != 0) { 291 padlock_free_ctx(axf, &ctx); 292 return (error); 293 } 294 axf->Final(hash, &ctx); 295 296 padlock_copy_ctx(axf, ses->ses_octx, &ctx); 297 axf->Update(&ctx, hash, axf->hashsize); 298 axf->Final(hash, &ctx); 299 300 /* Inject the authentication data */ 301 crypto_copyback(flags, buf, crd->crd_inject, 302 ses->ses_mlen == 0 ? axf->hashsize : ses->ses_mlen, hash); 303 return (0); 304 } 305 306 int 307 padlock_hash_setup(struct padlock_session *ses, struct cryptoini *macini) 308 { 309 310 ses->ses_mlen = macini->cri_mlen; 311 312 /* Find software structure which describes HMAC algorithm. */ 313 switch (macini->cri_alg) { 314 case CRYPTO_NULL_HMAC: 315 ses->ses_axf = &auth_hash_null; 316 break; 317 case CRYPTO_MD5_HMAC: 318 ses->ses_axf = &auth_hash_hmac_md5; 319 break; 320 case CRYPTO_SHA1_HMAC: 321 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0) 322 ses->ses_axf = &padlock_hmac_sha1; 323 else 324 ses->ses_axf = &auth_hash_hmac_sha1; 325 break; 326 case CRYPTO_RIPEMD160_HMAC: 327 ses->ses_axf = &auth_hash_hmac_ripemd_160; 328 break; 329 case CRYPTO_SHA2_256_HMAC: 330 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0) 331 ses->ses_axf = &padlock_hmac_sha256; 332 else 333 ses->ses_axf = &auth_hash_hmac_sha2_256; 334 break; 335 case CRYPTO_SHA2_384_HMAC: 336 ses->ses_axf = &auth_hash_hmac_sha2_384; 337 break; 338 case CRYPTO_SHA2_512_HMAC: 339 ses->ses_axf = &auth_hash_hmac_sha2_512; 340 break; 341 } 342 343 /* Allocate memory for HMAC inner and outer contexts. */ 344 ses->ses_ictx = kmalloc(ses->ses_axf->ctxsize, M_PADLOCK, 345 M_ZERO | M_NOWAIT); 346 ses->ses_octx = kmalloc(ses->ses_axf->ctxsize, M_PADLOCK, 347 M_ZERO | M_NOWAIT); 348 if (ses->ses_ictx == NULL || ses->ses_octx == NULL) 349 return (ENOMEM); 350 351 /* Setup key if given. */ 352 if (macini->cri_key != NULL) { 353 padlock_hash_key_setup(ses, macini->cri_key, 354 macini->cri_klen); 355 } 356 return (0); 357 } 358 359 int 360 padlock_hash_process(struct padlock_session *ses, struct cryptodesc *maccrd, 361 struct cryptop *crp) 362 { 363 int error; 364 365 if ((maccrd->crd_flags & CRD_F_KEY_EXPLICIT) != 0) 366 padlock_hash_key_setup(ses, maccrd->crd_key, maccrd->crd_klen); 367 368 error = padlock_authcompute(ses, maccrd, crp->crp_buf, crp->crp_flags); 369 return (error); 370 } 371 372 void 373 padlock_hash_free(struct padlock_session *ses) 374 { 375 376 if (ses->ses_ictx != NULL) { 377 padlock_free_ctx(ses->ses_axf, ses->ses_ictx); 378 bzero(ses->ses_ictx, ses->ses_axf->ctxsize); 379 kfree(ses->ses_ictx, M_PADLOCK); 380 ses->ses_ictx = NULL; 381 } 382 if (ses->ses_octx != NULL) { 383 padlock_free_ctx(ses->ses_axf, ses->ses_octx); 384 bzero(ses->ses_octx, ses->ses_axf->ctxsize); 385 kfree(ses->ses_octx, M_PADLOCK); 386 ses->ses_octx = NULL; 387 } 388 } 389