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/cryptosoft.h> /* for hmac_ipad_buffer and hmac_opad_buffer */ 48 #include <opencrypto/xform.h> 49 50 #include <crypto/via/padlock.h> 51 52 /* 53 * Implementation notes. 54 * 55 * Some VIA CPUs provides SHA1 and SHA256 acceleration. 56 * We implement all HMAC algorithms provided by crypto(9) framework, but we do 57 * the crypto work in software unless this is HMAC/SHA1 or HMAC/SHA256 and 58 * our CPU can accelerate it. 59 * 60 * Additional CPU instructions, which preform SHA1 and SHA256 are one-shot 61 * functions - we have only one chance to give the data, CPU itself will add 62 * the padding and calculate hash automatically. 63 * This means, it is not possible to implement common init(), update(), final() 64 * methods. 65 * The way I've choosen is to keep adding data to the buffer on update() 66 * (reallocating the buffer if necessary) and call XSHA{1,256} instruction on 67 * final(). 68 */ 69 70 struct padlock_sha_ctx { 71 uint8_t *psc_buf; 72 int psc_offset; 73 int psc_size; 74 }; 75 CTASSERT(sizeof(struct padlock_sha_ctx) <= sizeof(union authctx)); 76 77 static void padlock_sha_init(struct padlock_sha_ctx *ctx); 78 static int padlock_sha_update(struct padlock_sha_ctx *ctx, const uint8_t *buf, 79 uint16_t bufsize); 80 static void padlock_sha1_final(uint8_t *hash, struct padlock_sha_ctx *ctx); 81 static void padlock_sha256_final(uint8_t *hash, struct padlock_sha_ctx *ctx); 82 83 static struct auth_hash padlock_hmac_sha1 = { 84 .type = CRYPTO_SHA1_HMAC, 85 .name = "HMAC-SHA1", 86 .keysize = SHA1_BLOCK_LEN, 87 .hashsize = SHA1_HASH_LEN, 88 .ctxsize = sizeof(struct padlock_sha_ctx), 89 .blocksize = SHA1_BLOCK_LEN, 90 .Init = (void (*)(void *))padlock_sha_init, 91 .Update = (int (*)(void *, const uint8_t *, uint16_t))padlock_sha_update, 92 .Final = (void (*)(uint8_t *, void *))padlock_sha1_final, 93 }; 94 95 static struct auth_hash padlock_hmac_sha256 = { 96 .type = CRYPTO_SHA2_256_HMAC, 97 .name = "HMAC-SHA2-256", 98 .keysize = SHA2_256_BLOCK_LEN, 99 .hashsize = SHA2_256_HASH_LEN, 100 .ctxsize = sizeof(struct padlock_sha_ctx), 101 .blocksize = SHA2_256_BLOCK_LEN, 102 .Init = (void (*)(void *))padlock_sha_init, 103 .Update = (int (*)(void *, const uint8_t *, uint16_t))padlock_sha_update, 104 .Final = (void (*)(uint8_t *, void *))padlock_sha256_final, 105 }; 106 107 MALLOC_DECLARE(M_PADLOCK); 108 109 static __inline void 110 padlock_output_block(uint32_t *src, uint32_t *dst, size_t count) 111 { 112 113 while (count-- > 0) 114 *dst++ = bswap32(*src++); 115 } 116 117 static void 118 padlock_do_sha1(const u_char *in, u_char *out, int count) 119 { 120 u_char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */ 121 u_char *result = PADLOCK_ALIGN(buf); 122 123 ((uint32_t *)result)[0] = 0x67452301; 124 ((uint32_t *)result)[1] = 0xEFCDAB89; 125 ((uint32_t *)result)[2] = 0x98BADCFE; 126 ((uint32_t *)result)[3] = 0x10325476; 127 ((uint32_t *)result)[4] = 0xC3D2E1F0; 128 129 #ifdef __GNUCLIKE_ASM 130 __asm __volatile( 131 ".byte 0xf3, 0x0f, 0xa6, 0xc8" /* rep xsha1 */ 132 : "+S"(in), "+D"(result) 133 : "c"(count), "a"(0) 134 ); 135 #endif 136 137 padlock_output_block((uint32_t *)result, (uint32_t *)out, 138 SHA1_HASH_LEN / sizeof(uint32_t)); 139 } 140 141 static void 142 padlock_do_sha256(const char *in, char *out, int count) 143 { 144 char buf[128+16]; /* PadLock needs at least 128 bytes buffer. */ 145 char *result = PADLOCK_ALIGN(buf); 146 147 ((uint32_t *)result)[0] = 0x6A09E667; 148 ((uint32_t *)result)[1] = 0xBB67AE85; 149 ((uint32_t *)result)[2] = 0x3C6EF372; 150 ((uint32_t *)result)[3] = 0xA54FF53A; 151 ((uint32_t *)result)[4] = 0x510E527F; 152 ((uint32_t *)result)[5] = 0x9B05688C; 153 ((uint32_t *)result)[6] = 0x1F83D9AB; 154 ((uint32_t *)result)[7] = 0x5BE0CD19; 155 156 #ifdef __GNUCLIKE_ASM 157 __asm __volatile( 158 ".byte 0xf3, 0x0f, 0xa6, 0xd0" /* rep xsha256 */ 159 : "+S"(in), "+D"(result) 160 : "c"(count), "a"(0) 161 ); 162 #endif 163 164 padlock_output_block((uint32_t *)result, (uint32_t *)out, 165 SHA2_256_HASH_LEN / sizeof(uint32_t)); 166 } 167 168 static void 169 padlock_sha_init(struct padlock_sha_ctx *ctx) 170 { 171 172 ctx->psc_buf = NULL; 173 ctx->psc_offset = 0; 174 ctx->psc_size = 0; 175 } 176 177 static int 178 padlock_sha_update(struct padlock_sha_ctx *ctx, const uint8_t *buf, uint16_t bufsize) 179 { 180 181 if (ctx->psc_size - ctx->psc_offset < bufsize) { 182 ctx->psc_size = MAX(ctx->psc_size * 2, ctx->psc_size + bufsize); 183 ctx->psc_buf = realloc(ctx->psc_buf, ctx->psc_size, M_PADLOCK, 184 M_NOWAIT); 185 if(ctx->psc_buf == NULL) 186 return (ENOMEM); 187 } 188 bcopy(buf, ctx->psc_buf + ctx->psc_offset, bufsize); 189 ctx->psc_offset += bufsize; 190 return (0); 191 } 192 193 static void 194 padlock_sha_free(struct padlock_sha_ctx *ctx) 195 { 196 197 if (ctx->psc_buf != NULL) { 198 //bzero(ctx->psc_buf, ctx->psc_size); 199 free(ctx->psc_buf, M_PADLOCK); 200 ctx->psc_buf = NULL; 201 ctx->psc_offset = 0; 202 ctx->psc_size = 0; 203 } 204 } 205 206 static void 207 padlock_sha1_final(uint8_t *hash, struct padlock_sha_ctx *ctx) 208 { 209 210 padlock_do_sha1(ctx->psc_buf, hash, ctx->psc_offset); 211 padlock_sha_free(ctx); 212 } 213 214 static void 215 padlock_sha256_final(uint8_t *hash, struct padlock_sha_ctx *ctx) 216 { 217 218 padlock_do_sha256(ctx->psc_buf, hash, ctx->psc_offset); 219 padlock_sha_free(ctx); 220 } 221 222 static void 223 padlock_copy_ctx(struct auth_hash *axf, void *sctx, void *dctx) 224 { 225 226 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 && 227 (axf->type == CRYPTO_SHA1_HMAC || 228 axf->type == CRYPTO_SHA2_256_HMAC)) { 229 struct padlock_sha_ctx *spctx = sctx, *dpctx = dctx; 230 231 dpctx->psc_offset = spctx->psc_offset; 232 dpctx->psc_size = spctx->psc_size; 233 dpctx->psc_buf = malloc(dpctx->psc_size, M_PADLOCK, M_WAITOK); 234 bcopy(spctx->psc_buf, dpctx->psc_buf, dpctx->psc_size); 235 } else { 236 bcopy(sctx, dctx, axf->ctxsize); 237 } 238 } 239 240 static void 241 padlock_free_ctx(struct auth_hash *axf, void *ctx) 242 { 243 244 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0 && 245 (axf->type == CRYPTO_SHA1_HMAC || 246 axf->type == CRYPTO_SHA2_256_HMAC)) { 247 padlock_sha_free(ctx); 248 } 249 } 250 251 static void 252 padlock_hash_key_setup(struct padlock_session *ses, caddr_t key, int klen) 253 { 254 struct auth_hash *axf; 255 int i; 256 257 klen /= 8; 258 axf = ses->ses_axf; 259 260 /* 261 * Try to free contexts before using them, because 262 * padlock_hash_key_setup() can be called twice - once from 263 * padlock_newsession() and again from padlock_process(). 264 */ 265 padlock_free_ctx(axf, ses->ses_ictx); 266 padlock_free_ctx(axf, ses->ses_octx); 267 268 for (i = 0; i < klen; i++) 269 key[i] ^= HMAC_IPAD_VAL; 270 271 axf->Init(ses->ses_ictx); 272 axf->Update(ses->ses_ictx, key, klen); 273 axf->Update(ses->ses_ictx, hmac_ipad_buffer, axf->blocksize - klen); 274 275 for (i = 0; i < klen; i++) 276 key[i] ^= (HMAC_IPAD_VAL ^ HMAC_OPAD_VAL); 277 278 axf->Init(ses->ses_octx); 279 axf->Update(ses->ses_octx, key, klen); 280 axf->Update(ses->ses_octx, hmac_opad_buffer, axf->blocksize - klen); 281 282 for (i = 0; i < klen; i++) 283 key[i] ^= HMAC_OPAD_VAL; 284 } 285 286 /* 287 * Compute keyed-hash authenticator. 288 */ 289 static int 290 padlock_authcompute(struct padlock_session *ses, struct cryptodesc *crd, 291 caddr_t buf, int flags) 292 { 293 u_char hash[HASH_MAX_LEN]; 294 struct auth_hash *axf; 295 union authctx ctx; 296 int error; 297 298 axf = ses->ses_axf; 299 300 padlock_copy_ctx(axf, ses->ses_ictx, &ctx); 301 error = crypto_apply(flags, buf, crd->crd_skip, crd->crd_len, 302 (int (*)(void *, void *, unsigned int))axf->Update, (caddr_t)&ctx); 303 if (error != 0) { 304 padlock_free_ctx(axf, &ctx); 305 return (error); 306 } 307 axf->Final(hash, &ctx); 308 309 padlock_copy_ctx(axf, ses->ses_octx, &ctx); 310 axf->Update(&ctx, hash, axf->hashsize); 311 axf->Final(hash, &ctx); 312 313 /* Inject the authentication data */ 314 crypto_copyback(flags, buf, crd->crd_inject, 315 ses->ses_mlen == 0 ? axf->hashsize : ses->ses_mlen, hash); 316 return (0); 317 } 318 319 int 320 padlock_hash_setup(struct padlock_session *ses, struct cryptoini *macini) 321 { 322 323 ses->ses_mlen = macini->cri_mlen; 324 325 /* Find software structure which describes HMAC algorithm. */ 326 switch (macini->cri_alg) { 327 case CRYPTO_NULL_HMAC: 328 ses->ses_axf = &auth_hash_null; 329 break; 330 case CRYPTO_MD5_HMAC: 331 ses->ses_axf = &auth_hash_hmac_md5; 332 break; 333 case CRYPTO_SHA1_HMAC: 334 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0) 335 ses->ses_axf = &padlock_hmac_sha1; 336 else 337 ses->ses_axf = &auth_hash_hmac_sha1; 338 break; 339 case CRYPTO_RIPEMD160_HMAC: 340 ses->ses_axf = &auth_hash_hmac_ripemd_160; 341 break; 342 case CRYPTO_SHA2_256_HMAC: 343 if ((via_feature_xcrypt & VIA_HAS_SHA) != 0) 344 ses->ses_axf = &padlock_hmac_sha256; 345 else 346 ses->ses_axf = &auth_hash_hmac_sha2_256; 347 break; 348 case CRYPTO_SHA2_384_HMAC: 349 ses->ses_axf = &auth_hash_hmac_sha2_384; 350 break; 351 case CRYPTO_SHA2_512_HMAC: 352 ses->ses_axf = &auth_hash_hmac_sha2_512; 353 break; 354 } 355 356 /* Allocate memory for HMAC inner and outer contexts. */ 357 ses->ses_ictx = malloc(ses->ses_axf->ctxsize, M_PADLOCK, 358 M_ZERO | M_NOWAIT); 359 ses->ses_octx = malloc(ses->ses_axf->ctxsize, M_PADLOCK, 360 M_ZERO | M_NOWAIT); 361 if (ses->ses_ictx == NULL || ses->ses_octx == NULL) 362 return (ENOMEM); 363 364 /* Setup key if given. */ 365 if (macini->cri_key != NULL) { 366 padlock_hash_key_setup(ses, macini->cri_key, 367 macini->cri_klen); 368 } 369 return (0); 370 } 371 372 int 373 padlock_hash_process(struct padlock_session *ses, struct cryptodesc *maccrd, 374 struct cryptop *crp) 375 { 376 struct thread *td; 377 int error; 378 379 td = curthread; 380 fpu_kern_enter(td, ses->ses_fpu_ctx, FPU_KERN_NORMAL | FPU_KERN_KTHR); 381 if ((maccrd->crd_flags & CRD_F_KEY_EXPLICIT) != 0) 382 padlock_hash_key_setup(ses, maccrd->crd_key, maccrd->crd_klen); 383 384 error = padlock_authcompute(ses, maccrd, crp->crp_buf, crp->crp_flags); 385 fpu_kern_leave(td, ses->ses_fpu_ctx); 386 return (error); 387 } 388 389 void 390 padlock_hash_free(struct padlock_session *ses) 391 { 392 393 if (ses->ses_ictx != NULL) { 394 padlock_free_ctx(ses->ses_axf, ses->ses_ictx); 395 bzero(ses->ses_ictx, ses->ses_axf->ctxsize); 396 free(ses->ses_ictx, M_PADLOCK); 397 ses->ses_ictx = NULL; 398 } 399 if (ses->ses_octx != NULL) { 400 padlock_free_ctx(ses->ses_axf, ses->ses_octx); 401 bzero(ses->ses_octx, ses->ses_axf->ctxsize); 402 free(ses->ses_octx, M_PADLOCK); 403 ses->ses_octx = NULL; 404 } 405 } 406