1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright (c) 2013, 2015 by Delphix. All rights reserved. 24 * Copyright (c) 2013, Joyent, Inc. All rights reserved. 25 * Copyright 2013 Saso Kiselkov. All rights reserved. 26 */ 27 28 #include <sys/zfs_context.h> 29 #include <sys/spa.h> 30 #include <sys/spa_impl.h> 31 #include <sys/zio.h> 32 #include <sys/zio_checksum.h> 33 #include <sys/zil.h> 34 #include <zfs_fletcher.h> 35 36 /* 37 * Checksum vectors. 38 * 39 * In the SPA, everything is checksummed. We support checksum vectors 40 * for three distinct reasons: 41 * 42 * 1. Different kinds of data need different levels of protection. 43 * For SPA metadata, we always want a very strong checksum. 44 * For user data, we let users make the trade-off between speed 45 * and checksum strength. 46 * 47 * 2. Cryptographic hash and MAC algorithms are an area of active research. 48 * It is likely that in future hash functions will be at least as strong 49 * as current best-of-breed, and may be substantially faster as well. 50 * We want the ability to take advantage of these new hashes as soon as 51 * they become available. 52 * 53 * 3. If someone develops hardware that can compute a strong hash quickly, 54 * we want the ability to take advantage of that hardware. 55 * 56 * Of course, we don't want a checksum upgrade to invalidate existing 57 * data, so we store the checksum *function* in eight bits of the bp. 58 * This gives us room for up to 256 different checksum functions. 59 * 60 * When writing a block, we always checksum it with the latest-and-greatest 61 * checksum function of the appropriate strength. When reading a block, 62 * we compare the expected checksum against the actual checksum, which we 63 * compute via the checksum function specified by BP_GET_CHECKSUM(bp). 64 * 65 * SALTED CHECKSUMS 66 * 67 * To enable the use of less secure hash algorithms with dedup, we 68 * introduce the notion of salted checksums (MACs, really). A salted 69 * checksum is fed both a random 256-bit value (the salt) and the data 70 * to be checksummed. This salt is kept secret (stored on the pool, but 71 * never shown to the user). Thus even if an attacker knew of collision 72 * weaknesses in the hash algorithm, they won't be able to mount a known 73 * plaintext attack on the DDT, since the actual hash value cannot be 74 * known ahead of time. How the salt is used is algorithm-specific 75 * (some might simply prefix it to the data block, others might need to 76 * utilize a full-blown HMAC). On disk the salt is stored in a ZAP 77 * object in the MOS (DMU_POOL_CHECKSUM_SALT). 78 * 79 * CONTEXT TEMPLATES 80 * 81 * Some hashing algorithms need to perform a substantial amount of 82 * initialization work (e.g. salted checksums above may need to pre-hash 83 * the salt) before being able to process data. Performing this 84 * redundant work for each block would be wasteful, so we instead allow 85 * a checksum algorithm to do the work once (the first time it's used) 86 * and then keep this pre-initialized context as a template inside the 87 * spa_t (spa_cksum_tmpls). If the zio_checksum_info_t contains 88 * non-NULL ci_tmpl_init and ci_tmpl_free callbacks, they are used to 89 * construct and destruct the pre-initialized checksum context. The 90 * pre-initialized context is then reused during each checksum 91 * invocation and passed to the checksum function. 92 */ 93 94 /*ARGSUSED*/ 95 static void 96 zio_checksum_off(const void *buf, uint64_t size, 97 const void *ctx_template, zio_cksum_t *zcp) 98 { 99 ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0); 100 } 101 102 zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS] = { 103 {{NULL, NULL}, NULL, NULL, 0, "inherit"}, 104 {{NULL, NULL}, NULL, NULL, 0, "on"}, 105 {{zio_checksum_off, zio_checksum_off}, 106 NULL, NULL, 0, "off"}, 107 {{zio_checksum_SHA256, zio_checksum_SHA256}, 108 NULL, NULL, ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_EMBEDDED, 109 "label"}, 110 {{zio_checksum_SHA256, zio_checksum_SHA256}, 111 NULL, NULL, ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_EMBEDDED, 112 "gang_header"}, 113 {{fletcher_2_native, fletcher_2_byteswap}, 114 NULL, NULL, ZCHECKSUM_FLAG_EMBEDDED, "zilog"}, 115 {{fletcher_2_native, fletcher_2_byteswap}, 116 NULL, NULL, 0, "fletcher2"}, 117 {{fletcher_4_native, fletcher_4_byteswap}, 118 NULL, NULL, ZCHECKSUM_FLAG_METADATA, "fletcher4"}, 119 {{zio_checksum_SHA256, zio_checksum_SHA256}, 120 NULL, NULL, ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_DEDUP | 121 ZCHECKSUM_FLAG_NOPWRITE, "sha256"}, 122 {{fletcher_4_native, fletcher_4_byteswap}, 123 NULL, NULL, ZCHECKSUM_FLAG_EMBEDDED, "zilog2"}, 124 {{zio_checksum_off, zio_checksum_off}, 125 NULL, NULL, 0, "noparity"}, 126 #ifndef __NetBSD__ 127 {{zio_checksum_SHA512_native, zio_checksum_SHA512_byteswap}, 128 NULL, NULL, ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_DEDUP | 129 ZCHECKSUM_FLAG_NOPWRITE, "sha512"}, 130 {{zio_checksum_skein_native, zio_checksum_skein_byteswap}, 131 zio_checksum_skein_tmpl_init, zio_checksum_skein_tmpl_free, 132 ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_DEDUP | 133 ZCHECKSUM_FLAG_SALTED | ZCHECKSUM_FLAG_NOPWRITE, "skein"}, 134 #endif 135 #ifdef illumos 136 {{zio_checksum_edonr_native, zio_checksum_edonr_byteswap}, 137 zio_checksum_edonr_tmpl_init, zio_checksum_edonr_tmpl_free, 138 ZCHECKSUM_FLAG_METADATA | ZCHECKSUM_FLAG_SALTED | 139 ZCHECKSUM_FLAG_NOPWRITE, "edonr"}, 140 #endif 141 }; 142 143 /* 144 * The flag corresponding to the "verify" in dedup=[checksum,]verify 145 * must be cleared first, so callers should use ZIO_CHECKSUM_MASK. 146 */ 147 spa_feature_t 148 zio_checksum_to_feature(enum zio_checksum cksum) 149 { 150 VERIFY((cksum & ~ZIO_CHECKSUM_MASK) == 0); 151 152 switch (cksum) { 153 #ifndef __NetBSD__ 154 case ZIO_CHECKSUM_SHA512: 155 return (SPA_FEATURE_SHA512); 156 case ZIO_CHECKSUM_SKEIN: 157 return (SPA_FEATURE_SKEIN); 158 #endif 159 #ifdef illumos 160 case ZIO_CHECKSUM_EDONR: 161 return (SPA_FEATURE_EDONR); 162 #endif 163 } 164 return (SPA_FEATURE_NONE); 165 } 166 167 enum zio_checksum 168 zio_checksum_select(enum zio_checksum child, enum zio_checksum parent) 169 { 170 ASSERT(child < ZIO_CHECKSUM_FUNCTIONS); 171 ASSERT(parent < ZIO_CHECKSUM_FUNCTIONS); 172 ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON); 173 174 if (child == ZIO_CHECKSUM_INHERIT) 175 return (parent); 176 177 if (child == ZIO_CHECKSUM_ON) 178 return (ZIO_CHECKSUM_ON_VALUE); 179 180 return (child); 181 } 182 183 enum zio_checksum 184 zio_checksum_dedup_select(spa_t *spa, enum zio_checksum child, 185 enum zio_checksum parent) 186 { 187 ASSERT((child & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS); 188 ASSERT((parent & ZIO_CHECKSUM_MASK) < ZIO_CHECKSUM_FUNCTIONS); 189 ASSERT(parent != ZIO_CHECKSUM_INHERIT && parent != ZIO_CHECKSUM_ON); 190 191 if (child == ZIO_CHECKSUM_INHERIT) 192 return (parent); 193 194 if (child == ZIO_CHECKSUM_ON) 195 return (spa_dedup_checksum(spa)); 196 197 if (child == (ZIO_CHECKSUM_ON | ZIO_CHECKSUM_VERIFY)) 198 return (spa_dedup_checksum(spa) | ZIO_CHECKSUM_VERIFY); 199 200 ASSERT((zio_checksum_table[child & ZIO_CHECKSUM_MASK].ci_flags & 201 ZCHECKSUM_FLAG_DEDUP) || 202 (child & ZIO_CHECKSUM_VERIFY) || child == ZIO_CHECKSUM_OFF); 203 204 return (child); 205 } 206 207 /* 208 * Set the external verifier for a gang block based on <vdev, offset, txg>, 209 * a tuple which is guaranteed to be unique for the life of the pool. 210 */ 211 static void 212 zio_checksum_gang_verifier(zio_cksum_t *zcp, blkptr_t *bp) 213 { 214 dva_t *dva = BP_IDENTITY(bp); 215 uint64_t txg = BP_PHYSICAL_BIRTH(bp); 216 217 ASSERT(BP_IS_GANG(bp)); 218 219 ZIO_SET_CHECKSUM(zcp, DVA_GET_VDEV(dva), DVA_GET_OFFSET(dva), txg, 0); 220 } 221 222 /* 223 * Set the external verifier for a label block based on its offset. 224 * The vdev is implicit, and the txg is unknowable at pool open time -- 225 * hence the logic in vdev_uberblock_load() to find the most recent copy. 226 */ 227 static void 228 zio_checksum_label_verifier(zio_cksum_t *zcp, uint64_t offset) 229 { 230 ZIO_SET_CHECKSUM(zcp, offset, 0, 0, 0); 231 } 232 233 /* 234 * Calls the template init function of a checksum which supports context 235 * templates and installs the template into the spa_t. 236 */ 237 static void 238 zio_checksum_template_init(enum zio_checksum checksum, spa_t *spa) 239 { 240 zio_checksum_info_t *ci = &zio_checksum_table[checksum]; 241 242 if (ci->ci_tmpl_init == NULL) 243 return; 244 if (spa->spa_cksum_tmpls[checksum] != NULL) 245 return; 246 247 VERIFY(ci->ci_tmpl_free != NULL); 248 mutex_enter(&spa->spa_cksum_tmpls_lock); 249 if (spa->spa_cksum_tmpls[checksum] == NULL) { 250 spa->spa_cksum_tmpls[checksum] = 251 ci->ci_tmpl_init(&spa->spa_cksum_salt); 252 VERIFY(spa->spa_cksum_tmpls[checksum] != NULL); 253 } 254 mutex_exit(&spa->spa_cksum_tmpls_lock); 255 } 256 257 /* 258 * Generate the checksum. 259 */ 260 void 261 zio_checksum_compute(zio_t *zio, enum zio_checksum checksum, 262 void *data, uint64_t size) 263 { 264 blkptr_t *bp = zio->io_bp; 265 uint64_t offset = zio->io_offset; 266 zio_checksum_info_t *ci = &zio_checksum_table[checksum]; 267 zio_cksum_t cksum; 268 spa_t *spa = zio->io_spa; 269 270 ASSERT((uint_t)checksum < ZIO_CHECKSUM_FUNCTIONS); 271 ASSERT(ci->ci_func[0] != NULL); 272 273 zio_checksum_template_init(checksum, spa); 274 275 if (ci->ci_flags & ZCHECKSUM_FLAG_EMBEDDED) { 276 zio_eck_t *eck; 277 278 if (checksum == ZIO_CHECKSUM_ZILOG2) { 279 zil_chain_t *zilc = data; 280 281 size = P2ROUNDUP_TYPED(zilc->zc_nused, ZIL_MIN_BLKSZ, 282 uint64_t); 283 eck = &zilc->zc_eck; 284 } else { 285 eck = (zio_eck_t *)((char *)data + size) - 1; 286 } 287 if (checksum == ZIO_CHECKSUM_GANG_HEADER) 288 zio_checksum_gang_verifier(&eck->zec_cksum, bp); 289 else if (checksum == ZIO_CHECKSUM_LABEL) 290 zio_checksum_label_verifier(&eck->zec_cksum, offset); 291 else 292 bp->blk_cksum = eck->zec_cksum; 293 eck->zec_magic = ZEC_MAGIC; 294 ci->ci_func[0](data, size, spa->spa_cksum_tmpls[checksum], 295 &cksum); 296 eck->zec_cksum = cksum; 297 } else { 298 ci->ci_func[0](data, size, spa->spa_cksum_tmpls[checksum], 299 &bp->blk_cksum); 300 } 301 } 302 303 int 304 zio_checksum_error_impl(spa_t *spa, blkptr_t *bp, enum zio_checksum checksum, 305 void *data, uint64_t size, uint64_t offset, zio_bad_cksum_t *info) 306 { 307 zio_checksum_info_t *ci = &zio_checksum_table[checksum]; 308 zio_cksum_t actual_cksum, expected_cksum; 309 int byteswap; 310 311 if (checksum >= ZIO_CHECKSUM_FUNCTIONS || ci->ci_func[0] == NULL) 312 return (SET_ERROR(EINVAL)); 313 314 zio_checksum_template_init(checksum, spa); 315 316 if (ci->ci_flags & ZCHECKSUM_FLAG_EMBEDDED) { 317 zio_eck_t *eck; 318 zio_cksum_t verifier; 319 320 if (checksum == ZIO_CHECKSUM_ZILOG2) { 321 zil_chain_t *zilc = data; 322 uint64_t nused; 323 324 eck = &zilc->zc_eck; 325 if (eck->zec_magic == ZEC_MAGIC) 326 nused = zilc->zc_nused; 327 else if (eck->zec_magic == BSWAP_64(ZEC_MAGIC)) 328 nused = BSWAP_64(zilc->zc_nused); 329 else 330 return (SET_ERROR(ECKSUM)); 331 332 if (nused > size) 333 return (SET_ERROR(ECKSUM)); 334 335 size = P2ROUNDUP_TYPED(nused, ZIL_MIN_BLKSZ, uint64_t); 336 } else { 337 eck = (zio_eck_t *)((char *)data + size) - 1; 338 } 339 340 if (checksum == ZIO_CHECKSUM_GANG_HEADER) 341 zio_checksum_gang_verifier(&verifier, bp); 342 else if (checksum == ZIO_CHECKSUM_LABEL) 343 zio_checksum_label_verifier(&verifier, offset); 344 else 345 verifier = bp->blk_cksum; 346 347 byteswap = (eck->zec_magic == BSWAP_64(ZEC_MAGIC)); 348 349 if (byteswap) 350 byteswap_uint64_array(&verifier, sizeof (zio_cksum_t)); 351 352 expected_cksum = eck->zec_cksum; 353 eck->zec_cksum = verifier; 354 ci->ci_func[byteswap](data, size, 355 spa->spa_cksum_tmpls[checksum], &actual_cksum); 356 eck->zec_cksum = expected_cksum; 357 358 if (byteswap) { 359 byteswap_uint64_array(&expected_cksum, 360 sizeof (zio_cksum_t)); 361 } 362 } else { 363 byteswap = BP_SHOULD_BYTESWAP(bp); 364 expected_cksum = bp->blk_cksum; 365 ci->ci_func[byteswap](data, size, 366 spa->spa_cksum_tmpls[checksum], &actual_cksum); 367 } 368 369 if (info != NULL) { 370 info->zbc_expected = expected_cksum; 371 info->zbc_actual = actual_cksum; 372 info->zbc_checksum_name = ci->ci_name; 373 info->zbc_byteswapped = byteswap; 374 info->zbc_injected = 0; 375 info->zbc_has_cksum = 1; 376 } 377 378 if (!ZIO_CHECKSUM_EQUAL(actual_cksum, expected_cksum)) 379 return (SET_ERROR(ECKSUM)); 380 381 return (0); 382 } 383 384 int 385 zio_checksum_error(zio_t *zio, zio_bad_cksum_t *info) 386 { 387 blkptr_t *bp = zio->io_bp; 388 uint_t checksum = (bp == NULL ? zio->io_prop.zp_checksum : 389 (BP_IS_GANG(bp) ? ZIO_CHECKSUM_GANG_HEADER : BP_GET_CHECKSUM(bp))); 390 int error; 391 uint64_t size = (bp == NULL ? zio->io_size : 392 (BP_IS_GANG(bp) ? SPA_GANGBLOCKSIZE : BP_GET_PSIZE(bp))); 393 uint64_t offset = zio->io_offset; 394 void *data = zio->io_data; 395 spa_t *spa = zio->io_spa; 396 397 error = zio_checksum_error_impl(spa, bp, checksum, data, size, 398 offset, info); 399 if (error != 0 && zio_injection_enabled && !zio->io_error && 400 (error = zio_handle_fault_injection(zio, ECKSUM)) != 0) { 401 402 info->zbc_injected = 1; 403 return (error); 404 } 405 return (error); 406 } 407 408 /* 409 * Called by a spa_t that's about to be deallocated. This steps through 410 * all of the checksum context templates and deallocates any that were 411 * initialized using the algorithm-specific template init function. 412 */ 413 void 414 zio_checksum_templates_free(spa_t *spa) 415 { 416 for (enum zio_checksum checksum = 0; 417 checksum < ZIO_CHECKSUM_FUNCTIONS; checksum++) { 418 if (spa->spa_cksum_tmpls[checksum] != NULL) { 419 zio_checksum_info_t *ci = &zio_checksum_table[checksum]; 420 421 VERIFY(ci->ci_tmpl_free != NULL); 422 ci->ci_tmpl_free(spa->spa_cksum_tmpls[checksum]); 423 spa->spa_cksum_tmpls[checksum] = NULL; 424 } 425 } 426 } 427