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 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include <sys/param.h> 31 #include <sys/systm.h> 32 #include <sys/kernel.h> 33 #include <sys/module.h> 34 #include <sys/malloc.h> 35 #include <sys/libkern.h> 36 #include <sys/endian.h> 37 #include <sys/pcpu.h> 38 #if defined(__amd64__) || defined(__i386__) 39 #include <machine/cpufunc.h> 40 #include <machine/cputypes.h> 41 #include <machine/md_var.h> 42 #include <machine/specialreg.h> 43 #endif 44 #include <machine/pcb.h> 45 46 #include <opencrypto/cryptodev.h> 47 #include <opencrypto/xform.h> 48 49 #include <crypto/via/padlock.h> 50 51 /* 52 * Implementation notes. 53 * 54 * Some VIA CPUs provides SHA1 and SHA256 acceleration. 55 * We implement all HMAC algorithms provided by crypto(9) framework, but we do 56 * the crypto work in software unless this is HMAC/SHA1 or HMAC/SHA256 and 57 * our CPU can accelerate it. 58 * 59 * Additional CPU instructions, which preform SHA1 and SHA256 are one-shot 60 * functions - we have only one chance to give the data, CPU itself will add 61 * the padding and calculate hash automatically. 62 * This means, it is not possible to implement common init(), update(), final() 63 * methods. 64 * The way I've choosen is to keep adding data to the buffer on update() 65 * (reallocating the buffer if necessary) and call XSHA{1,256} instruction on 66 * final(). 67 */ 68 69 struct padlock_sha_ctx { 70 uint8_t *psc_buf; 71 int psc_offset; 72 int psc_size; 73 }; 74 CTASSERT(sizeof(struct padlock_sha_ctx) <= sizeof(union authctx)); 75 76 static void padlock_sha_init(struct padlock_sha_ctx *ctx); 77 static int padlock_sha_update(struct padlock_sha_ctx *ctx, const uint8_t *buf, 78 uint16_t bufsize); 79 static void padlock_sha1_final(uint8_t *hash, struct padlock_sha_ctx *ctx); 80 static void padlock_sha256_final(uint8_t *hash, struct padlock_sha_ctx *ctx); 81 82 static struct auth_hash padlock_hmac_sha1 = { 83 .type = CRYPTO_SHA1_HMAC, 84 .name = "HMAC-SHA1", 85 .keysize = SHA1_BLOCK_LEN, 86 .hashsize = SHA1_HASH_LEN, 87 .ctxsize = sizeof(struct padlock_sha_ctx), 88 .blocksize = SHA1_BLOCK_LEN, 89 .Init = (void (*)(void *))padlock_sha_init, 90 .Update = (int (*)(void *, const uint8_t *, uint16_t))padlock_sha_update, 91 .Final = (void (*)(uint8_t *, void *))padlock_sha1_final, 92 }; 93 94 static struct auth_hash padlock_hmac_sha256 = { 95 .type = CRYPTO_SHA2_256_HMAC, 96 .name = "HMAC-SHA2-256", 97 .keysize = SHA2_256_BLOCK_LEN, 98 .hashsize = SHA2_256_HASH_LEN, 99 .ctxsize = sizeof(struct padlock_sha_ctx), 100 .blocksize = SHA2_256_BLOCK_LEN, 101 .Init = (void (*)(void *))padlock_sha_init, 102 .Update = (int (*)(void *, const uint8_t *, uint16_t))padlock_sha_update, 103 .Final = (void (*)(uint8_t *, void *))padlock_sha256_final, 104 }; 105 106 MALLOC_DECLARE(M_PADLOCK); 107 108 static __inline void 109 padlock_output_block(uint32_t *src, uint32_t *dst, size_t count) 110 { 111 112 while (count-- > 0) 113 *dst++ = bswap32(*src++); 114 } 115 116 static void 117 padlock_do_sha1(const u_char *in, u_char *out, int count) 118 { 119 u_char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */ 120 u_char *result = PADLOCK_ALIGN(buf); 121 122 ((uint32_t *)result)[0] = 0x67452301; 123 ((uint32_t *)result)[1] = 0xEFCDAB89; 124 ((uint32_t *)result)[2] = 0x98BADCFE; 125 ((uint32_t *)result)[3] = 0x10325476; 126 ((uint32_t *)result)[4] = 0xC3D2E1F0; 127 128 #ifdef __GNUCLIKE_ASM 129 __asm __volatile( 130 ".byte 0xf3, 0x0f, 0xa6, 0xc8" /* rep xsha1 */ 131 : "+S"(in), "+D"(result) 132 : "c"(count), "a"(0) 133 ); 134 #endif 135 136 padlock_output_block((uint32_t *)result, (uint32_t *)out, 137 SHA1_HASH_LEN / sizeof(uint32_t)); 138 } 139 140 static void 141 padlock_do_sha256(const char *in, char *out, int count) 142 { 143 char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */ 144 char *result = PADLOCK_ALIGN(buf); 145 146 ((uint32_t *)result)[0] = 0x6A09E667; 147 ((uint32_t *)result)[1] = 0xBB67AE85; 148 ((uint32_t *)result)[2] = 0x3C6EF372; 149 ((uint32_t *)result)[3] = 0xA54FF53A; 150 ((uint32_t *)result)[4] = 0x510E527F; 151 ((uint32_t *)result)[5] = 0x9B05688C; 152 ((uint32_t *)result)[6] = 0x1F83D9AB; 153 ((uint32_t *)result)[7] = 0x5BE0CD19; 154 155 #ifdef __GNUCLIKE_ASM 156 __asm __volatile( 157 ".byte 0xf3, 0x0f, 0xa6, 0xd0" /* rep xsha256 */ 158 : "+S"(in), "+D"(result) 159 : "c"(count), "a"(0) 160 ); 161 #endif 162 163 padlock_output_block((uint32_t *)result, (uint32_t *)out, 164 SHA2_256_HASH_LEN / sizeof(uint32_t)); 165 } 166 167 static void 168 padlock_sha_init(struct padlock_sha_ctx *ctx) 169 { 170 171 ctx->psc_buf = NULL; 172 ctx->psc_offset = 0; 173 ctx->psc_size = 0; 174 } 175 176 static int 177 padlock_sha_update(struct padlock_sha_ctx *ctx, const uint8_t *buf, uint16_t bufsize) 178 { 179 180 if (ctx->psc_size - ctx->psc_offset < bufsize) { 181 ctx->psc_size = MAX(ctx->psc_size * 2, ctx->psc_size + bufsize); 182 ctx->psc_buf = realloc(ctx->psc_buf, ctx->psc_size, M_PADLOCK, 183 M_NOWAIT); 184 if(ctx->psc_buf == NULL) 185 return (ENOMEM); 186 } 187 bcopy(buf, ctx->psc_buf + ctx->psc_offset, bufsize); 188 ctx->psc_offset += bufsize; 189 return (0); 190 } 191 192 static void 193 padlock_sha_free(struct padlock_sha_ctx *ctx) 194 { 195 196 if (ctx->psc_buf != NULL) { 197 //bzero(ctx->psc_buf, ctx->psc_size); 198 free(ctx->psc_buf, M_PADLOCK); 199 ctx->psc_buf = NULL; 200 ctx->psc_offset = 0; 201 ctx->psc_size = 0; 202 } 203 } 204 205 static void 206 padlock_sha1_final(uint8_t *hash, struct padlock_sha_ctx *ctx) 207 { 208 209 padlock_do_sha1(ctx->psc_buf, hash, ctx->psc_offset); 210 padlock_sha_free(ctx); 211 } 212 213 static void 214 padlock_sha256_final(uint8_t *hash, struct padlock_sha_ctx *ctx) 215 { 216 217 padlock_do_sha256(ctx->psc_buf, hash, ctx->psc_offset); 218 padlock_sha_free(ctx); 219 } 220 221 static void 222 padlock_copy_ctx(struct auth_hash *axf, void *sctx, void *dctx) 223 { 224 225 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 && 226 (axf->type == CRYPTO_SHA1_HMAC || 227 axf->type == CRYPTO_SHA2_256_HMAC)) { 228 struct padlock_sha_ctx *spctx = sctx, *dpctx = dctx; 229 230 dpctx->psc_offset = spctx->psc_offset; 231 dpctx->psc_size = spctx->psc_size; 232 dpctx->psc_buf = malloc(dpctx->psc_size, M_PADLOCK, M_WAITOK); 233 bcopy(spctx->psc_buf, dpctx->psc_buf, dpctx->psc_size); 234 } else { 235 bcopy(sctx, dctx, axf->ctxsize); 236 } 237 } 238 239 static void 240 padlock_free_ctx(struct auth_hash *axf, void *ctx) 241 { 242 243 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 && 244 (axf->type == CRYPTO_SHA1_HMAC || 245 axf->type == CRYPTO_SHA2_256_HMAC)) { 246 padlock_sha_free(ctx); 247 } 248 } 249 250 static void 251 padlock_hash_key_setup(struct padlock_session *ses, const uint8_t *key, 252 int klen) 253 { 254 struct auth_hash *axf; 255 256 axf = ses->ses_axf; 257 258 /* 259 * Try to free contexts before using them, because 260 * padlock_hash_key_setup() can be called twice - once from 261 * padlock_newsession() and again from padlock_process(). 262 */ 263 padlock_free_ctx(axf, ses->ses_ictx); 264 padlock_free_ctx(axf, ses->ses_octx); 265 266 hmac_init_ipad(axf, key, klen, ses->ses_ictx); 267 hmac_init_opad(axf, key, klen, ses->ses_octx); 268 } 269 270 /* 271 * Compute keyed-hash authenticator. 272 */ 273 static int 274 padlock_authcompute(struct padlock_session *ses, struct cryptop *crp) 275 { 276 u_char hash[HASH_MAX_LEN], hash2[HASH_MAX_LEN]; 277 struct auth_hash *axf; 278 union authctx ctx; 279 int error; 280 281 axf = ses->ses_axf; 282 283 padlock_copy_ctx(axf, ses->ses_ictx, &ctx); 284 error = crypto_apply(crp, crp->crp_aad_start, crp->crp_aad_length, 285 (int (*)(void *, void *, unsigned int))axf->Update, (caddr_t)&ctx); 286 if (error != 0) { 287 padlock_free_ctx(axf, &ctx); 288 return (error); 289 } 290 error = crypto_apply(crp, crp->crp_payload_start, 291 crp->crp_payload_length, 292 (int (*)(void *, void *, unsigned int))axf->Update, (caddr_t)&ctx); 293 if (error != 0) { 294 padlock_free_ctx(axf, &ctx); 295 return (error); 296 } 297 axf->Final(hash, &ctx); 298 299 padlock_copy_ctx(axf, ses->ses_octx, &ctx); 300 axf->Update(&ctx, hash, axf->hashsize); 301 axf->Final(hash, &ctx); 302 303 if (crp->crp_op & CRYPTO_OP_VERIFY_DIGEST) { 304 crypto_copydata(crp, crp->crp_digest_start, ses->ses_mlen, 305 hash2); 306 if (timingsafe_bcmp(hash, hash2, ses->ses_mlen) != 0) 307 return (EBADMSG); 308 } else 309 crypto_copyback(crp, crp->crp_digest_start, ses->ses_mlen, 310 hash); 311 return (0); 312 } 313 314 /* Find software structure which describes HMAC algorithm. */ 315 static struct auth_hash * 316 padlock_hash_lookup(int alg) 317 { 318 struct auth_hash *axf; 319 320 switch (alg) { 321 case CRYPTO_NULL_HMAC: 322 axf = &auth_hash_null; 323 break; 324 case CRYPTO_MD5_HMAC: 325 axf = &auth_hash_hmac_md5; 326 break; 327 case CRYPTO_SHA1_HMAC: 328 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0) 329 axf = &padlock_hmac_sha1; 330 else 331 axf = &auth_hash_hmac_sha1; 332 break; 333 case CRYPTO_RIPEMD160_HMAC: 334 axf = &auth_hash_hmac_ripemd_160; 335 break; 336 case CRYPTO_SHA2_256_HMAC: 337 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0) 338 axf = &padlock_hmac_sha256; 339 else 340 axf = &auth_hash_hmac_sha2_256; 341 break; 342 case CRYPTO_SHA2_384_HMAC: 343 axf = &auth_hash_hmac_sha2_384; 344 break; 345 case CRYPTO_SHA2_512_HMAC: 346 axf = &auth_hash_hmac_sha2_512; 347 break; 348 default: 349 axf = NULL; 350 break; 351 } 352 return (axf); 353 } 354 355 bool 356 padlock_hash_check(const struct crypto_session_params *csp) 357 { 358 359 return (padlock_hash_lookup(csp->csp_auth_alg) != NULL); 360 } 361 362 int 363 padlock_hash_setup(struct padlock_session *ses, 364 const struct crypto_session_params *csp) 365 { 366 367 ses->ses_axf = padlock_hash_lookup(csp->csp_auth_alg); 368 if (csp->csp_auth_mlen == 0) 369 ses->ses_mlen = ses->ses_axf->hashsize; 370 else 371 ses->ses_mlen = csp->csp_auth_mlen; 372 373 /* Allocate memory for HMAC inner and outer contexts. */ 374 ses->ses_ictx = malloc(ses->ses_axf->ctxsize, M_PADLOCK, 375 M_ZERO | M_NOWAIT); 376 ses->ses_octx = malloc(ses->ses_axf->ctxsize, M_PADLOCK, 377 M_ZERO | M_NOWAIT); 378 if (ses->ses_ictx == NULL || ses->ses_octx == NULL) 379 return (ENOMEM); 380 381 /* Setup key if given. */ 382 if (csp->csp_auth_key != NULL) { 383 padlock_hash_key_setup(ses, csp->csp_auth_key, 384 csp->csp_auth_klen); 385 } 386 return (0); 387 } 388 389 int 390 padlock_hash_process(struct padlock_session *ses, struct cryptop *crp, 391 const struct crypto_session_params *csp) 392 { 393 struct thread *td; 394 int error; 395 396 td = curthread; 397 fpu_kern_enter(td, ses->ses_fpu_ctx, FPU_KERN_NORMAL | FPU_KERN_KTHR); 398 if (crp->crp_auth_key != NULL) 399 padlock_hash_key_setup(ses, crp->crp_auth_key, 400 csp->csp_auth_klen); 401 402 error = padlock_authcompute(ses, crp); 403 fpu_kern_leave(td, ses->ses_fpu_ctx); 404 return (error); 405 } 406 407 void 408 padlock_hash_free(struct padlock_session *ses) 409 { 410 411 if (ses->ses_ictx != NULL) { 412 padlock_free_ctx(ses->ses_axf, ses->ses_ictx); 413 bzero(ses->ses_ictx, ses->ses_axf->ctxsize); 414 free(ses->ses_ictx, M_PADLOCK); 415 ses->ses_ictx = NULL; 416 } 417 if (ses->ses_octx != NULL) { 418 padlock_free_ctx(ses->ses_axf, ses->ses_octx); 419 bzero(ses->ses_octx, ses->ses_axf->ctxsize); 420 free(ses->ses_octx, M_PADLOCK); 421 ses->ses_octx = NULL; 422 } 423 } 424