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(void *vctx); 77 static int padlock_sha_update(void *vctx, const void *buf, u_int bufsize); 78 static void padlock_sha1_final(uint8_t *hash, void *vctx); 79 static void padlock_sha256_final(uint8_t *hash, void *vctx); 80 81 static const struct auth_hash padlock_hmac_sha1 = { 82 .type = CRYPTO_SHA1_HMAC, 83 .name = "HMAC-SHA1", 84 .keysize = SHA1_BLOCK_LEN, 85 .hashsize = SHA1_HASH_LEN, 86 .ctxsize = sizeof(struct padlock_sha_ctx), 87 .blocksize = SHA1_BLOCK_LEN, 88 .Init = padlock_sha_init, 89 .Update = padlock_sha_update, 90 .Final = padlock_sha1_final, 91 }; 92 93 static const struct auth_hash padlock_hmac_sha256 = { 94 .type = CRYPTO_SHA2_256_HMAC, 95 .name = "HMAC-SHA2-256", 96 .keysize = SHA2_256_BLOCK_LEN, 97 .hashsize = SHA2_256_HASH_LEN, 98 .ctxsize = sizeof(struct padlock_sha_ctx), 99 .blocksize = SHA2_256_BLOCK_LEN, 100 .Init = padlock_sha_init, 101 .Update = padlock_sha_update, 102 .Final = padlock_sha256_final, 103 }; 104 105 MALLOC_DECLARE(M_PADLOCK); 106 107 static __inline void 108 padlock_output_block(uint32_t *src, uint32_t *dst, size_t count) 109 { 110 111 while (count-- > 0) 112 *dst++ = bswap32(*src++); 113 } 114 115 static void 116 padlock_do_sha1(const u_char *in, u_char *out, int count) 117 { 118 u_char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */ 119 u_char *result = PADLOCK_ALIGN(buf); 120 121 ((uint32_t *)result)[0] = 0x67452301; 122 ((uint32_t *)result)[1] = 0xEFCDAB89; 123 ((uint32_t *)result)[2] = 0x98BADCFE; 124 ((uint32_t *)result)[3] = 0x10325476; 125 ((uint32_t *)result)[4] = 0xC3D2E1F0; 126 127 __asm __volatile( 128 ".byte 0xf3, 0x0f, 0xa6, 0xc8" /* rep xsha1 */ 129 : "+S"(in), "+D"(result) 130 : "c"(count), "a"(0) 131 ); 132 133 padlock_output_block((uint32_t *)result, (uint32_t *)out, 134 SHA1_HASH_LEN / sizeof(uint32_t)); 135 } 136 137 static void 138 padlock_do_sha256(const char *in, char *out, int count) 139 { 140 char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */ 141 char *result = PADLOCK_ALIGN(buf); 142 143 ((uint32_t *)result)[0] = 0x6A09E667; 144 ((uint32_t *)result)[1] = 0xBB67AE85; 145 ((uint32_t *)result)[2] = 0x3C6EF372; 146 ((uint32_t *)result)[3] = 0xA54FF53A; 147 ((uint32_t *)result)[4] = 0x510E527F; 148 ((uint32_t *)result)[5] = 0x9B05688C; 149 ((uint32_t *)result)[6] = 0x1F83D9AB; 150 ((uint32_t *)result)[7] = 0x5BE0CD19; 151 152 __asm __volatile( 153 ".byte 0xf3, 0x0f, 0xa6, 0xd0" /* rep xsha256 */ 154 : "+S"(in), "+D"(result) 155 : "c"(count), "a"(0) 156 ); 157 158 padlock_output_block((uint32_t *)result, (uint32_t *)out, 159 SHA2_256_HASH_LEN / sizeof(uint32_t)); 160 } 161 162 static void 163 padlock_sha_init(void *vctx) 164 { 165 struct padlock_sha_ctx *ctx; 166 167 ctx = vctx; 168 ctx->psc_buf = NULL; 169 ctx->psc_offset = 0; 170 ctx->psc_size = 0; 171 } 172 173 static int 174 padlock_sha_update(void *vctx, const void *buf, u_int bufsize) 175 { 176 struct padlock_sha_ctx *ctx; 177 178 ctx = vctx; 179 if (ctx->psc_size - ctx->psc_offset < bufsize) { 180 ctx->psc_size = MAX(ctx->psc_size * 2, ctx->psc_size + bufsize); 181 ctx->psc_buf = realloc(ctx->psc_buf, ctx->psc_size, M_PADLOCK, 182 M_NOWAIT); 183 if(ctx->psc_buf == NULL) 184 return (ENOMEM); 185 } 186 bcopy(buf, ctx->psc_buf + ctx->psc_offset, bufsize); 187 ctx->psc_offset += bufsize; 188 return (0); 189 } 190 191 static void 192 padlock_sha_free(void *vctx) 193 { 194 struct padlock_sha_ctx *ctx; 195 196 ctx = vctx; 197 if (ctx->psc_buf != NULL) { 198 zfree(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, void *vctx) 207 { 208 struct padlock_sha_ctx *ctx; 209 210 ctx = vctx; 211 padlock_do_sha1(ctx->psc_buf, hash, ctx->psc_offset); 212 padlock_sha_free(ctx); 213 } 214 215 static void 216 padlock_sha256_final(uint8_t *hash, void *vctx) 217 { 218 struct padlock_sha_ctx *ctx; 219 220 ctx = vctx; 221 padlock_do_sha256(ctx->psc_buf, hash, ctx->psc_offset); 222 padlock_sha_free(ctx); 223 } 224 225 static void 226 padlock_copy_ctx(const struct auth_hash *axf, void *sctx, void *dctx) 227 { 228 229 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 && 230 (axf->type == CRYPTO_SHA1_HMAC || 231 axf->type == CRYPTO_SHA2_256_HMAC)) { 232 struct padlock_sha_ctx *spctx = sctx, *dpctx = dctx; 233 234 dpctx->psc_offset = spctx->psc_offset; 235 dpctx->psc_size = spctx->psc_size; 236 dpctx->psc_buf = malloc(dpctx->psc_size, M_PADLOCK, M_WAITOK); 237 bcopy(spctx->psc_buf, dpctx->psc_buf, dpctx->psc_size); 238 } else { 239 bcopy(sctx, dctx, axf->ctxsize); 240 } 241 } 242 243 static void 244 padlock_free_ctx(const struct auth_hash *axf, void *ctx) 245 { 246 247 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 && 248 (axf->type == CRYPTO_SHA1_HMAC || 249 axf->type == CRYPTO_SHA2_256_HMAC)) { 250 padlock_sha_free(ctx); 251 } 252 } 253 254 static void 255 padlock_hash_key_setup(struct padlock_session *ses, const uint8_t *key, 256 int klen) 257 { 258 const struct auth_hash *axf; 259 260 axf = ses->ses_axf; 261 262 /* 263 * Try to free contexts before using them, because 264 * padlock_hash_key_setup() can be called twice - once from 265 * padlock_newsession() and again from padlock_process(). 266 */ 267 padlock_free_ctx(axf, ses->ses_ictx); 268 padlock_free_ctx(axf, ses->ses_octx); 269 270 hmac_init_ipad(axf, key, klen, ses->ses_ictx); 271 hmac_init_opad(axf, key, klen, ses->ses_octx); 272 } 273 274 /* 275 * Compute keyed-hash authenticator. 276 */ 277 static int 278 padlock_authcompute(struct padlock_session *ses, struct cryptop *crp) 279 { 280 u_char hash[HASH_MAX_LEN], hash2[HASH_MAX_LEN]; 281 const 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(crp, crp->crp_aad_start, crp->crp_aad_length, 289 axf->Update, &ctx); 290 if (error != 0) { 291 padlock_free_ctx(axf, &ctx); 292 return (error); 293 } 294 error = crypto_apply(crp, crp->crp_payload_start, 295 crp->crp_payload_length, axf->Update, &ctx); 296 if (error != 0) { 297 padlock_free_ctx(axf, &ctx); 298 return (error); 299 } 300 axf->Final(hash, &ctx); 301 302 padlock_copy_ctx(axf, ses->ses_octx, &ctx); 303 axf->Update(&ctx, hash, axf->hashsize); 304 axf->Final(hash, &ctx); 305 306 if (crp->crp_op & CRYPTO_OP_VERIFY_DIGEST) { 307 crypto_copydata(crp, crp->crp_digest_start, ses->ses_mlen, 308 hash2); 309 if (timingsafe_bcmp(hash, hash2, ses->ses_mlen) != 0) 310 return (EBADMSG); 311 } else 312 crypto_copyback(crp, crp->crp_digest_start, ses->ses_mlen, 313 hash); 314 return (0); 315 } 316 317 /* Find software structure which describes HMAC algorithm. */ 318 static const struct auth_hash * 319 padlock_hash_lookup(int alg) 320 { 321 const struct auth_hash *axf; 322 323 switch (alg) { 324 case CRYPTO_NULL_HMAC: 325 axf = &auth_hash_null; 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 zfree(ses->ses_ictx, M_PADLOCK); 414 ses->ses_ictx = NULL; 415 } 416 if (ses->ses_octx != NULL) { 417 padlock_free_ctx(ses->ses_axf, ses->ses_octx); 418 zfree(ses->ses_octx, M_PADLOCK); 419 ses->ses_octx = NULL; 420 } 421 } 422