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 /* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2013 by Delphix. All rights reserved. 25 * Copyright (c) 2013, Joyent, Inc. All rights reserved. 26 */ 27 28 #include <sys/zfs_context.h> 29 #include <sys/spa.h> 30 #include <sys/vdev_impl.h> 31 #include <sys/vdev_disk.h> 32 #include <sys/vdev_file.h> 33 #include <sys/vdev_raidz.h> 34 #include <sys/zio.h> 35 #include <sys/zio_checksum.h> 36 #include <sys/fs/zfs.h> 37 #include <sys/fm/fs/zfs.h> 38 39 /* 40 * Virtual device vector for RAID-Z. 41 * 42 * This vdev supports single, double, and triple parity. For single parity, 43 * we use a simple XOR of all the data columns. For double or triple parity, 44 * we use a special case of Reed-Solomon coding. This extends the 45 * technique described in "The mathematics of RAID-6" by H. Peter Anvin by 46 * drawing on the system described in "A Tutorial on Reed-Solomon Coding for 47 * Fault-Tolerance in RAID-like Systems" by James S. Plank on which the 48 * former is also based. The latter is designed to provide higher performance 49 * for writes. 50 * 51 * Note that the Plank paper claimed to support arbitrary N+M, but was then 52 * amended six years later identifying a critical flaw that invalidates its 53 * claims. Nevertheless, the technique can be adapted to work for up to 54 * triple parity. For additional parity, the amendment "Note: Correction to 55 * the 1997 Tutorial on Reed-Solomon Coding" by James S. Plank and Ying Ding 56 * is viable, but the additional complexity means that write performance will 57 * suffer. 58 * 59 * All of the methods above operate on a Galois field, defined over the 60 * integers mod 2^N. In our case we choose N=8 for GF(8) so that all elements 61 * can be expressed with a single byte. Briefly, the operations on the 62 * field are defined as follows: 63 * 64 * o addition (+) is represented by a bitwise XOR 65 * o subtraction (-) is therefore identical to addition: A + B = A - B 66 * o multiplication of A by 2 is defined by the following bitwise expression: 67 * 68 * (A * 2)_7 = A_6 69 * (A * 2)_6 = A_5 70 * (A * 2)_5 = A_4 71 * (A * 2)_4 = A_3 + A_7 72 * (A * 2)_3 = A_2 + A_7 73 * (A * 2)_2 = A_1 + A_7 74 * (A * 2)_1 = A_0 75 * (A * 2)_0 = A_7 76 * 77 * In C, multiplying by 2 is therefore ((a << 1) ^ ((a & 0x80) ? 0x1d : 0)). 78 * As an aside, this multiplication is derived from the error correcting 79 * primitive polynomial x^8 + x^4 + x^3 + x^2 + 1. 80 * 81 * Observe that any number in the field (except for 0) can be expressed as a 82 * power of 2 -- a generator for the field. We store a table of the powers of 83 * 2 and logs base 2 for quick look ups, and exploit the fact that A * B can 84 * be rewritten as 2^(log_2(A) + log_2(B)) (where '+' is normal addition rather 85 * than field addition). The inverse of a field element A (A^-1) is therefore 86 * A ^ (255 - 1) = A^254. 87 * 88 * The up-to-three parity columns, P, Q, R over several data columns, 89 * D_0, ... D_n-1, can be expressed by field operations: 90 * 91 * P = D_0 + D_1 + ... + D_n-2 + D_n-1 92 * Q = 2^n-1 * D_0 + 2^n-2 * D_1 + ... + 2^1 * D_n-2 + 2^0 * D_n-1 93 * = ((...((D_0) * 2 + D_1) * 2 + ...) * 2 + D_n-2) * 2 + D_n-1 94 * R = 4^n-1 * D_0 + 4^n-2 * D_1 + ... + 4^1 * D_n-2 + 4^0 * D_n-1 95 * = ((...((D_0) * 4 + D_1) * 4 + ...) * 4 + D_n-2) * 4 + D_n-1 96 * 97 * We chose 1, 2, and 4 as our generators because 1 corresponds to the trival 98 * XOR operation, and 2 and 4 can be computed quickly and generate linearly- 99 * independent coefficients. (There are no additional coefficients that have 100 * this property which is why the uncorrected Plank method breaks down.) 101 * 102 * See the reconstruction code below for how P, Q and R can used individually 103 * or in concert to recover missing data columns. 104 */ 105 106 typedef struct raidz_col { 107 uint64_t rc_devidx; /* child device index for I/O */ 108 uint64_t rc_offset; /* device offset */ 109 uint64_t rc_size; /* I/O size */ 110 void *rc_data; /* I/O data */ 111 void *rc_gdata; /* used to store the "good" version */ 112 int rc_error; /* I/O error for this device */ 113 uint8_t rc_tried; /* Did we attempt this I/O column? */ 114 uint8_t rc_skipped; /* Did we skip this I/O column? */ 115 } raidz_col_t; 116 117 typedef struct raidz_map { 118 uint64_t rm_cols; /* Regular column count */ 119 uint64_t rm_scols; /* Count including skipped columns */ 120 uint64_t rm_bigcols; /* Number of oversized columns */ 121 uint64_t rm_asize; /* Actual total I/O size */ 122 uint64_t rm_missingdata; /* Count of missing data devices */ 123 uint64_t rm_missingparity; /* Count of missing parity devices */ 124 uint64_t rm_firstdatacol; /* First data column/parity count */ 125 uint64_t rm_nskip; /* Skipped sectors for padding */ 126 uint64_t rm_skipstart; /* Column index of padding start */ 127 void *rm_datacopy; /* rm_asize-buffer of copied data */ 128 uintptr_t rm_reports; /* # of referencing checksum reports */ 129 uint8_t rm_freed; /* map no longer has referencing ZIO */ 130 uint8_t rm_ecksuminjected; /* checksum error was injected */ 131 raidz_col_t rm_col[1]; /* Flexible array of I/O columns */ 132 } raidz_map_t; 133 134 #define VDEV_RAIDZ_P 0 135 #define VDEV_RAIDZ_Q 1 136 #define VDEV_RAIDZ_R 2 137 138 #define VDEV_RAIDZ_MUL_2(x) (((x) << 1) ^ (((x) & 0x80) ? 0x1d : 0)) 139 #define VDEV_RAIDZ_MUL_4(x) (VDEV_RAIDZ_MUL_2(VDEV_RAIDZ_MUL_2(x))) 140 141 /* 142 * We provide a mechanism to perform the field multiplication operation on a 143 * 64-bit value all at once rather than a byte at a time. This works by 144 * creating a mask from the top bit in each byte and using that to 145 * conditionally apply the XOR of 0x1d. 146 */ 147 #define VDEV_RAIDZ_64MUL_2(x, mask) \ 148 { \ 149 (mask) = (x) & 0x8080808080808080ULL; \ 150 (mask) = ((mask) << 1) - ((mask) >> 7); \ 151 (x) = (((x) << 1) & 0xfefefefefefefefeULL) ^ \ 152 ((mask) & 0x1d1d1d1d1d1d1d1d); \ 153 } 154 155 #define VDEV_RAIDZ_64MUL_4(x, mask) \ 156 { \ 157 VDEV_RAIDZ_64MUL_2((x), mask); \ 158 VDEV_RAIDZ_64MUL_2((x), mask); \ 159 } 160 161 #define VDEV_LABEL_OFFSET(x) (x + VDEV_LABEL_START_SIZE) 162 163 /* 164 * Force reconstruction to use the general purpose method. 165 */ 166 int vdev_raidz_default_to_general; 167 168 /* Powers of 2 in the Galois field defined above. */ 169 static const uint8_t vdev_raidz_pow2[256] = { 170 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 171 0x1d, 0x3a, 0x74, 0xe8, 0xcd, 0x87, 0x13, 0x26, 172 0x4c, 0x98, 0x2d, 0x5a, 0xb4, 0x75, 0xea, 0xc9, 173 0x8f, 0x03, 0x06, 0x0c, 0x18, 0x30, 0x60, 0xc0, 174 0x9d, 0x27, 0x4e, 0x9c, 0x25, 0x4a, 0x94, 0x35, 175 0x6a, 0xd4, 0xb5, 0x77, 0xee, 0xc1, 0x9f, 0x23, 176 0x46, 0x8c, 0x05, 0x0a, 0x14, 0x28, 0x50, 0xa0, 177 0x5d, 0xba, 0x69, 0xd2, 0xb9, 0x6f, 0xde, 0xa1, 178 0x5f, 0xbe, 0x61, 0xc2, 0x99, 0x2f, 0x5e, 0xbc, 179 0x65, 0xca, 0x89, 0x0f, 0x1e, 0x3c, 0x78, 0xf0, 180 0xfd, 0xe7, 0xd3, 0xbb, 0x6b, 0xd6, 0xb1, 0x7f, 181 0xfe, 0xe1, 0xdf, 0xa3, 0x5b, 0xb6, 0x71, 0xe2, 182 0xd9, 0xaf, 0x43, 0x86, 0x11, 0x22, 0x44, 0x88, 183 0x0d, 0x1a, 0x34, 0x68, 0xd0, 0xbd, 0x67, 0xce, 184 0x81, 0x1f, 0x3e, 0x7c, 0xf8, 0xed, 0xc7, 0x93, 185 0x3b, 0x76, 0xec, 0xc5, 0x97, 0x33, 0x66, 0xcc, 186 0x85, 0x17, 0x2e, 0x5c, 0xb8, 0x6d, 0xda, 0xa9, 187 0x4f, 0x9e, 0x21, 0x42, 0x84, 0x15, 0x2a, 0x54, 188 0xa8, 0x4d, 0x9a, 0x29, 0x52, 0xa4, 0x55, 0xaa, 189 0x49, 0x92, 0x39, 0x72, 0xe4, 0xd5, 0xb7, 0x73, 190 0xe6, 0xd1, 0xbf, 0x63, 0xc6, 0x91, 0x3f, 0x7e, 191 0xfc, 0xe5, 0xd7, 0xb3, 0x7b, 0xf6, 0xf1, 0xff, 192 0xe3, 0xdb, 0xab, 0x4b, 0x96, 0x31, 0x62, 0xc4, 193 0x95, 0x37, 0x6e, 0xdc, 0xa5, 0x57, 0xae, 0x41, 194 0x82, 0x19, 0x32, 0x64, 0xc8, 0x8d, 0x07, 0x0e, 195 0x1c, 0x38, 0x70, 0xe0, 0xdd, 0xa7, 0x53, 0xa6, 196 0x51, 0xa2, 0x59, 0xb2, 0x79, 0xf2, 0xf9, 0xef, 197 0xc3, 0x9b, 0x2b, 0x56, 0xac, 0x45, 0x8a, 0x09, 198 0x12, 0x24, 0x48, 0x90, 0x3d, 0x7a, 0xf4, 0xf5, 199 0xf7, 0xf3, 0xfb, 0xeb, 0xcb, 0x8b, 0x0b, 0x16, 200 0x2c, 0x58, 0xb0, 0x7d, 0xfa, 0xe9, 0xcf, 0x83, 201 0x1b, 0x36, 0x6c, 0xd8, 0xad, 0x47, 0x8e, 0x01 202 }; 203 /* Logs of 2 in the Galois field defined above. */ 204 static const uint8_t vdev_raidz_log2[256] = { 205 0x00, 0x00, 0x01, 0x19, 0x02, 0x32, 0x1a, 0xc6, 206 0x03, 0xdf, 0x33, 0xee, 0x1b, 0x68, 0xc7, 0x4b, 207 0x04, 0x64, 0xe0, 0x0e, 0x34, 0x8d, 0xef, 0x81, 208 0x1c, 0xc1, 0x69, 0xf8, 0xc8, 0x08, 0x4c, 0x71, 209 0x05, 0x8a, 0x65, 0x2f, 0xe1, 0x24, 0x0f, 0x21, 210 0x35, 0x93, 0x8e, 0xda, 0xf0, 0x12, 0x82, 0x45, 211 0x1d, 0xb5, 0xc2, 0x7d, 0x6a, 0x27, 0xf9, 0xb9, 212 0xc9, 0x9a, 0x09, 0x78, 0x4d, 0xe4, 0x72, 0xa6, 213 0x06, 0xbf, 0x8b, 0x62, 0x66, 0xdd, 0x30, 0xfd, 214 0xe2, 0x98, 0x25, 0xb3, 0x10, 0x91, 0x22, 0x88, 215 0x36, 0xd0, 0x94, 0xce, 0x8f, 0x96, 0xdb, 0xbd, 216 0xf1, 0xd2, 0x13, 0x5c, 0x83, 0x38, 0x46, 0x40, 217 0x1e, 0x42, 0xb6, 0xa3, 0xc3, 0x48, 0x7e, 0x6e, 218 0x6b, 0x3a, 0x28, 0x54, 0xfa, 0x85, 0xba, 0x3d, 219 0xca, 0x5e, 0x9b, 0x9f, 0x0a, 0x15, 0x79, 0x2b, 220 0x4e, 0xd4, 0xe5, 0xac, 0x73, 0xf3, 0xa7, 0x57, 221 0x07, 0x70, 0xc0, 0xf7, 0x8c, 0x80, 0x63, 0x0d, 222 0x67, 0x4a, 0xde, 0xed, 0x31, 0xc5, 0xfe, 0x18, 223 0xe3, 0xa5, 0x99, 0x77, 0x26, 0xb8, 0xb4, 0x7c, 224 0x11, 0x44, 0x92, 0xd9, 0x23, 0x20, 0x89, 0x2e, 225 0x37, 0x3f, 0xd1, 0x5b, 0x95, 0xbc, 0xcf, 0xcd, 226 0x90, 0x87, 0x97, 0xb2, 0xdc, 0xfc, 0xbe, 0x61, 227 0xf2, 0x56, 0xd3, 0xab, 0x14, 0x2a, 0x5d, 0x9e, 228 0x84, 0x3c, 0x39, 0x53, 0x47, 0x6d, 0x41, 0xa2, 229 0x1f, 0x2d, 0x43, 0xd8, 0xb7, 0x7b, 0xa4, 0x76, 230 0xc4, 0x17, 0x49, 0xec, 0x7f, 0x0c, 0x6f, 0xf6, 231 0x6c, 0xa1, 0x3b, 0x52, 0x29, 0x9d, 0x55, 0xaa, 232 0xfb, 0x60, 0x86, 0xb1, 0xbb, 0xcc, 0x3e, 0x5a, 233 0xcb, 0x59, 0x5f, 0xb0, 0x9c, 0xa9, 0xa0, 0x51, 234 0x0b, 0xf5, 0x16, 0xeb, 0x7a, 0x75, 0x2c, 0xd7, 235 0x4f, 0xae, 0xd5, 0xe9, 0xe6, 0xe7, 0xad, 0xe8, 236 0x74, 0xd6, 0xf4, 0xea, 0xa8, 0x50, 0x58, 0xaf, 237 }; 238 239 static void vdev_raidz_generate_parity(raidz_map_t *rm); 240 241 /* 242 * Multiply a given number by 2 raised to the given power. 243 */ 244 static uint8_t 245 vdev_raidz_exp2(uint_t a, int exp) 246 { 247 if (a == 0) 248 return (0); 249 250 ASSERT(exp >= 0); 251 ASSERT(vdev_raidz_log2[a] > 0 || a == 1); 252 253 exp += vdev_raidz_log2[a]; 254 if (exp > 255) 255 exp -= 255; 256 257 return (vdev_raidz_pow2[exp]); 258 } 259 260 static void 261 vdev_raidz_map_free(raidz_map_t *rm) 262 { 263 int c; 264 size_t size; 265 266 for (c = 0; c < rm->rm_firstdatacol; c++) { 267 zio_buf_free(rm->rm_col[c].rc_data, rm->rm_col[c].rc_size); 268 269 if (rm->rm_col[c].rc_gdata != NULL) 270 zio_buf_free(rm->rm_col[c].rc_gdata, 271 rm->rm_col[c].rc_size); 272 } 273 274 size = 0; 275 for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) 276 size += rm->rm_col[c].rc_size; 277 278 if (rm->rm_datacopy != NULL) 279 zio_buf_free(rm->rm_datacopy, size); 280 281 kmem_free(rm, offsetof(raidz_map_t, rm_col[rm->rm_scols])); 282 } 283 284 static void 285 vdev_raidz_map_free_vsd(zio_t *zio) 286 { 287 raidz_map_t *rm = zio->io_vsd; 288 289 ASSERT0(rm->rm_freed); 290 rm->rm_freed = 1; 291 292 if (rm->rm_reports == 0) 293 vdev_raidz_map_free(rm); 294 } 295 296 /*ARGSUSED*/ 297 static void 298 vdev_raidz_cksum_free(void *arg, size_t ignored) 299 { 300 raidz_map_t *rm = arg; 301 302 ASSERT3U(rm->rm_reports, >, 0); 303 304 if (--rm->rm_reports == 0 && rm->rm_freed != 0) 305 vdev_raidz_map_free(rm); 306 } 307 308 static void 309 vdev_raidz_cksum_finish(zio_cksum_report_t *zcr, const void *good_data) 310 { 311 raidz_map_t *rm = zcr->zcr_cbdata; 312 size_t c = zcr->zcr_cbinfo; 313 size_t x; 314 315 const char *good = NULL; 316 const char *bad = rm->rm_col[c].rc_data; 317 318 if (good_data == NULL) { 319 zfs_ereport_finish_checksum(zcr, NULL, NULL, B_FALSE); 320 return; 321 } 322 323 if (c < rm->rm_firstdatacol) { 324 /* 325 * The first time through, calculate the parity blocks for 326 * the good data (this relies on the fact that the good 327 * data never changes for a given logical ZIO) 328 */ 329 if (rm->rm_col[0].rc_gdata == NULL) { 330 char *bad_parity[VDEV_RAIDZ_MAXPARITY]; 331 char *buf; 332 333 /* 334 * Set up the rm_col[]s to generate the parity for 335 * good_data, first saving the parity bufs and 336 * replacing them with buffers to hold the result. 337 */ 338 for (x = 0; x < rm->rm_firstdatacol; x++) { 339 bad_parity[x] = rm->rm_col[x].rc_data; 340 rm->rm_col[x].rc_data = rm->rm_col[x].rc_gdata = 341 zio_buf_alloc(rm->rm_col[x].rc_size); 342 } 343 344 /* fill in the data columns from good_data */ 345 buf = (char *)good_data; 346 for (; x < rm->rm_cols; x++) { 347 rm->rm_col[x].rc_data = buf; 348 buf += rm->rm_col[x].rc_size; 349 } 350 351 /* 352 * Construct the parity from the good data. 353 */ 354 vdev_raidz_generate_parity(rm); 355 356 /* restore everything back to its original state */ 357 for (x = 0; x < rm->rm_firstdatacol; x++) 358 rm->rm_col[x].rc_data = bad_parity[x]; 359 360 buf = rm->rm_datacopy; 361 for (x = rm->rm_firstdatacol; x < rm->rm_cols; x++) { 362 rm->rm_col[x].rc_data = buf; 363 buf += rm->rm_col[x].rc_size; 364 } 365 } 366 367 ASSERT3P(rm->rm_col[c].rc_gdata, !=, NULL); 368 good = rm->rm_col[c].rc_gdata; 369 } else { 370 /* adjust good_data to point at the start of our column */ 371 good = good_data; 372 373 for (x = rm->rm_firstdatacol; x < c; x++) 374 good += rm->rm_col[x].rc_size; 375 } 376 377 /* we drop the ereport if it ends up that the data was good */ 378 zfs_ereport_finish_checksum(zcr, good, bad, B_TRUE); 379 } 380 381 /* 382 * Invoked indirectly by zfs_ereport_start_checksum(), called 383 * below when our read operation fails completely. The main point 384 * is to keep a copy of everything we read from disk, so that at 385 * vdev_raidz_cksum_finish() time we can compare it with the good data. 386 */ 387 static void 388 vdev_raidz_cksum_report(zio_t *zio, zio_cksum_report_t *zcr, void *arg) 389 { 390 size_t c = (size_t)(uintptr_t)arg; 391 caddr_t buf; 392 393 raidz_map_t *rm = zio->io_vsd; 394 size_t size; 395 396 /* set up the report and bump the refcount */ 397 zcr->zcr_cbdata = rm; 398 zcr->zcr_cbinfo = c; 399 zcr->zcr_finish = vdev_raidz_cksum_finish; 400 zcr->zcr_free = vdev_raidz_cksum_free; 401 402 rm->rm_reports++; 403 ASSERT3U(rm->rm_reports, >, 0); 404 405 if (rm->rm_datacopy != NULL) 406 return; 407 408 /* 409 * It's the first time we're called for this raidz_map_t, so we need 410 * to copy the data aside; there's no guarantee that our zio's buffer 411 * won't be re-used for something else. 412 * 413 * Our parity data is already in separate buffers, so there's no need 414 * to copy them. 415 */ 416 417 size = 0; 418 for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) 419 size += rm->rm_col[c].rc_size; 420 421 buf = rm->rm_datacopy = zio_buf_alloc(size); 422 423 for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) { 424 raidz_col_t *col = &rm->rm_col[c]; 425 426 bcopy(col->rc_data, buf, col->rc_size); 427 col->rc_data = buf; 428 429 buf += col->rc_size; 430 } 431 ASSERT3P(buf - (caddr_t)rm->rm_datacopy, ==, size); 432 } 433 434 static const zio_vsd_ops_t vdev_raidz_vsd_ops = { 435 vdev_raidz_map_free_vsd, 436 vdev_raidz_cksum_report 437 }; 438 439 /* 440 * Divides the IO evenly across all child vdevs; usually, dcols is 441 * the number of children in the target vdev. 442 */ 443 static raidz_map_t * 444 vdev_raidz_map_alloc(caddr_t data, uint64_t size, uint64_t offset, 445 uint64_t unit_shift, uint64_t dcols, uint64_t nparity) 446 { 447 raidz_map_t *rm; 448 /* The starting RAIDZ (parent) vdev sector of the block. */ 449 uint64_t b = offset >> unit_shift; 450 /* The zio's size in units of the vdev's minimum sector size. */ 451 uint64_t s = size >> unit_shift; 452 /* The first column for this stripe. */ 453 uint64_t f = b % dcols; 454 /* The starting byte offset on each child vdev. */ 455 uint64_t o = (b / dcols) << unit_shift; 456 uint64_t q, r, c, bc, col, acols, scols, coff, devidx, asize, tot; 457 458 /* 459 * "Quotient": The number of data sectors for this stripe on all but 460 * the "big column" child vdevs that also contain "remainder" data. 461 */ 462 q = s / (dcols - nparity); 463 464 /* 465 * "Remainder": The number of partial stripe data sectors in this I/O. 466 * This will add a sector to some, but not all, child vdevs. 467 */ 468 r = s - q * (dcols - nparity); 469 470 /* The number of "big columns" - those which contain remainder data. */ 471 bc = (r == 0 ? 0 : r + nparity); 472 473 /* 474 * The total number of data and parity sectors associated with 475 * this I/O. 476 */ 477 tot = s + nparity * (q + (r == 0 ? 0 : 1)); 478 479 /* acols: The columns that will be accessed. */ 480 /* scols: The columns that will be accessed or skipped. */ 481 if (q == 0) { 482 /* Our I/O request doesn't span all child vdevs. */ 483 acols = bc; 484 scols = MIN(dcols, roundup(bc, nparity + 1)); 485 } else { 486 acols = dcols; 487 scols = dcols; 488 } 489 490 ASSERT3U(acols, <=, scols); 491 492 rm = kmem_alloc(offsetof(raidz_map_t, rm_col[scols]), KM_SLEEP); 493 494 rm->rm_cols = acols; 495 rm->rm_scols = scols; 496 rm->rm_bigcols = bc; 497 rm->rm_skipstart = bc; 498 rm->rm_missingdata = 0; 499 rm->rm_missingparity = 0; 500 rm->rm_firstdatacol = nparity; 501 rm->rm_datacopy = NULL; 502 rm->rm_reports = 0; 503 rm->rm_freed = 0; 504 rm->rm_ecksuminjected = 0; 505 506 asize = 0; 507 508 for (c = 0; c < scols; c++) { 509 col = f + c; 510 coff = o; 511 if (col >= dcols) { 512 col -= dcols; 513 coff += 1ULL << unit_shift; 514 } 515 rm->rm_col[c].rc_devidx = col; 516 rm->rm_col[c].rc_offset = coff; 517 rm->rm_col[c].rc_data = NULL; 518 rm->rm_col[c].rc_gdata = NULL; 519 rm->rm_col[c].rc_error = 0; 520 rm->rm_col[c].rc_tried = 0; 521 rm->rm_col[c].rc_skipped = 0; 522 523 if (c >= acols) 524 rm->rm_col[c].rc_size = 0; 525 else if (c < bc) 526 rm->rm_col[c].rc_size = (q + 1) << unit_shift; 527 else 528 rm->rm_col[c].rc_size = q << unit_shift; 529 530 asize += rm->rm_col[c].rc_size; 531 } 532 533 ASSERT3U(asize, ==, tot << unit_shift); 534 rm->rm_asize = roundup(asize, (nparity + 1) << unit_shift); 535 rm->rm_nskip = roundup(tot, nparity + 1) - tot; 536 ASSERT3U(rm->rm_asize - asize, ==, rm->rm_nskip << unit_shift); 537 ASSERT3U(rm->rm_nskip, <=, nparity); 538 539 for (c = 0; c < rm->rm_firstdatacol; c++) 540 rm->rm_col[c].rc_data = zio_buf_alloc(rm->rm_col[c].rc_size); 541 542 rm->rm_col[c].rc_data = data; 543 544 for (c = c + 1; c < acols; c++) 545 rm->rm_col[c].rc_data = (char *)rm->rm_col[c - 1].rc_data + 546 rm->rm_col[c - 1].rc_size; 547 548 /* 549 * If all data stored spans all columns, there's a danger that parity 550 * will always be on the same device and, since parity isn't read 551 * during normal operation, that that device's I/O bandwidth won't be 552 * used effectively. We therefore switch the parity every 1MB. 553 * 554 * ... at least that was, ostensibly, the theory. As a practical 555 * matter unless we juggle the parity between all devices evenly, we 556 * won't see any benefit. Further, occasional writes that aren't a 557 * multiple of the LCM of the number of children and the minimum 558 * stripe width are sufficient to avoid pessimal behavior. 559 * Unfortunately, this decision created an implicit on-disk format 560 * requirement that we need to support for all eternity, but only 561 * for single-parity RAID-Z. 562 * 563 * If we intend to skip a sector in the zeroth column for padding 564 * we must make sure to note this swap. We will never intend to 565 * skip the first column since at least one data and one parity 566 * column must appear in each row. 567 */ 568 ASSERT(rm->rm_cols >= 2); 569 ASSERT(rm->rm_col[0].rc_size == rm->rm_col[1].rc_size); 570 571 if (rm->rm_firstdatacol == 1 && (offset & (1ULL << 20))) { 572 devidx = rm->rm_col[0].rc_devidx; 573 o = rm->rm_col[0].rc_offset; 574 rm->rm_col[0].rc_devidx = rm->rm_col[1].rc_devidx; 575 rm->rm_col[0].rc_offset = rm->rm_col[1].rc_offset; 576 rm->rm_col[1].rc_devidx = devidx; 577 rm->rm_col[1].rc_offset = o; 578 579 if (rm->rm_skipstart == 0) 580 rm->rm_skipstart = 1; 581 } 582 583 return (rm); 584 } 585 586 static void 587 vdev_raidz_generate_parity_p(raidz_map_t *rm) 588 { 589 uint64_t *p, *src, pcount, ccount, i; 590 int c; 591 592 pcount = rm->rm_col[VDEV_RAIDZ_P].rc_size / sizeof (src[0]); 593 594 for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) { 595 src = rm->rm_col[c].rc_data; 596 p = rm->rm_col[VDEV_RAIDZ_P].rc_data; 597 ccount = rm->rm_col[c].rc_size / sizeof (src[0]); 598 599 if (c == rm->rm_firstdatacol) { 600 ASSERT(ccount == pcount); 601 for (i = 0; i < ccount; i++, src++, p++) { 602 *p = *src; 603 } 604 } else { 605 ASSERT(ccount <= pcount); 606 for (i = 0; i < ccount; i++, src++, p++) { 607 *p ^= *src; 608 } 609 } 610 } 611 } 612 613 static void 614 vdev_raidz_generate_parity_pq(raidz_map_t *rm) 615 { 616 uint64_t *p, *q, *src, pcnt, ccnt, mask, i; 617 int c; 618 619 pcnt = rm->rm_col[VDEV_RAIDZ_P].rc_size / sizeof (src[0]); 620 ASSERT(rm->rm_col[VDEV_RAIDZ_P].rc_size == 621 rm->rm_col[VDEV_RAIDZ_Q].rc_size); 622 623 for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) { 624 src = rm->rm_col[c].rc_data; 625 p = rm->rm_col[VDEV_RAIDZ_P].rc_data; 626 q = rm->rm_col[VDEV_RAIDZ_Q].rc_data; 627 628 ccnt = rm->rm_col[c].rc_size / sizeof (src[0]); 629 630 if (c == rm->rm_firstdatacol) { 631 ASSERT(ccnt == pcnt || ccnt == 0); 632 for (i = 0; i < ccnt; i++, src++, p++, q++) { 633 *p = *src; 634 *q = *src; 635 } 636 for (; i < pcnt; i++, src++, p++, q++) { 637 *p = 0; 638 *q = 0; 639 } 640 } else { 641 ASSERT(ccnt <= pcnt); 642 643 /* 644 * Apply the algorithm described above by multiplying 645 * the previous result and adding in the new value. 646 */ 647 for (i = 0; i < ccnt; i++, src++, p++, q++) { 648 *p ^= *src; 649 650 VDEV_RAIDZ_64MUL_2(*q, mask); 651 *q ^= *src; 652 } 653 654 /* 655 * Treat short columns as though they are full of 0s. 656 * Note that there's therefore nothing needed for P. 657 */ 658 for (; i < pcnt; i++, q++) { 659 VDEV_RAIDZ_64MUL_2(*q, mask); 660 } 661 } 662 } 663 } 664 665 static void 666 vdev_raidz_generate_parity_pqr(raidz_map_t *rm) 667 { 668 uint64_t *p, *q, *r, *src, pcnt, ccnt, mask, i; 669 int c; 670 671 pcnt = rm->rm_col[VDEV_RAIDZ_P].rc_size / sizeof (src[0]); 672 ASSERT(rm->rm_col[VDEV_RAIDZ_P].rc_size == 673 rm->rm_col[VDEV_RAIDZ_Q].rc_size); 674 ASSERT(rm->rm_col[VDEV_RAIDZ_P].rc_size == 675 rm->rm_col[VDEV_RAIDZ_R].rc_size); 676 677 for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) { 678 src = rm->rm_col[c].rc_data; 679 p = rm->rm_col[VDEV_RAIDZ_P].rc_data; 680 q = rm->rm_col[VDEV_RAIDZ_Q].rc_data; 681 r = rm->rm_col[VDEV_RAIDZ_R].rc_data; 682 683 ccnt = rm->rm_col[c].rc_size / sizeof (src[0]); 684 685 if (c == rm->rm_firstdatacol) { 686 ASSERT(ccnt == pcnt || ccnt == 0); 687 for (i = 0; i < ccnt; i++, src++, p++, q++, r++) { 688 *p = *src; 689 *q = *src; 690 *r = *src; 691 } 692 for (; i < pcnt; i++, src++, p++, q++, r++) { 693 *p = 0; 694 *q = 0; 695 *r = 0; 696 } 697 } else { 698 ASSERT(ccnt <= pcnt); 699 700 /* 701 * Apply the algorithm described above by multiplying 702 * the previous result and adding in the new value. 703 */ 704 for (i = 0; i < ccnt; i++, src++, p++, q++, r++) { 705 *p ^= *src; 706 707 VDEV_RAIDZ_64MUL_2(*q, mask); 708 *q ^= *src; 709 710 VDEV_RAIDZ_64MUL_4(*r, mask); 711 *r ^= *src; 712 } 713 714 /* 715 * Treat short columns as though they are full of 0s. 716 * Note that there's therefore nothing needed for P. 717 */ 718 for (; i < pcnt; i++, q++, r++) { 719 VDEV_RAIDZ_64MUL_2(*q, mask); 720 VDEV_RAIDZ_64MUL_4(*r, mask); 721 } 722 } 723 } 724 } 725 726 /* 727 * Generate RAID parity in the first virtual columns according to the number of 728 * parity columns available. 729 */ 730 static void 731 vdev_raidz_generate_parity(raidz_map_t *rm) 732 { 733 switch (rm->rm_firstdatacol) { 734 case 1: 735 vdev_raidz_generate_parity_p(rm); 736 break; 737 case 2: 738 vdev_raidz_generate_parity_pq(rm); 739 break; 740 case 3: 741 vdev_raidz_generate_parity_pqr(rm); 742 break; 743 default: 744 cmn_err(CE_PANIC, "invalid RAID-Z configuration"); 745 } 746 } 747 748 static int 749 vdev_raidz_reconstruct_p(raidz_map_t *rm, int *tgts, int ntgts) 750 { 751 uint64_t *dst, *src, xcount, ccount, count, i; 752 int x = tgts[0]; 753 int c; 754 755 ASSERT(ntgts == 1); 756 ASSERT(x >= rm->rm_firstdatacol); 757 ASSERT(x < rm->rm_cols); 758 759 xcount = rm->rm_col[x].rc_size / sizeof (src[0]); 760 ASSERT(xcount <= rm->rm_col[VDEV_RAIDZ_P].rc_size / sizeof (src[0])); 761 ASSERT(xcount > 0); 762 763 src = rm->rm_col[VDEV_RAIDZ_P].rc_data; 764 dst = rm->rm_col[x].rc_data; 765 for (i = 0; i < xcount; i++, dst++, src++) { 766 *dst = *src; 767 } 768 769 for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) { 770 src = rm->rm_col[c].rc_data; 771 dst = rm->rm_col[x].rc_data; 772 773 if (c == x) 774 continue; 775 776 ccount = rm->rm_col[c].rc_size / sizeof (src[0]); 777 count = MIN(ccount, xcount); 778 779 for (i = 0; i < count; i++, dst++, src++) { 780 *dst ^= *src; 781 } 782 } 783 784 return (1 << VDEV_RAIDZ_P); 785 } 786 787 static int 788 vdev_raidz_reconstruct_q(raidz_map_t *rm, int *tgts, int ntgts) 789 { 790 uint64_t *dst, *src, xcount, ccount, count, mask, i; 791 uint8_t *b; 792 int x = tgts[0]; 793 int c, j, exp; 794 795 ASSERT(ntgts == 1); 796 797 xcount = rm->rm_col[x].rc_size / sizeof (src[0]); 798 ASSERT(xcount <= rm->rm_col[VDEV_RAIDZ_Q].rc_size / sizeof (src[0])); 799 800 for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) { 801 src = rm->rm_col[c].rc_data; 802 dst = rm->rm_col[x].rc_data; 803 804 if (c == x) 805 ccount = 0; 806 else 807 ccount = rm->rm_col[c].rc_size / sizeof (src[0]); 808 809 count = MIN(ccount, xcount); 810 811 if (c == rm->rm_firstdatacol) { 812 for (i = 0; i < count; i++, dst++, src++) { 813 *dst = *src; 814 } 815 for (; i < xcount; i++, dst++) { 816 *dst = 0; 817 } 818 819 } else { 820 for (i = 0; i < count; i++, dst++, src++) { 821 VDEV_RAIDZ_64MUL_2(*dst, mask); 822 *dst ^= *src; 823 } 824 825 for (; i < xcount; i++, dst++) { 826 VDEV_RAIDZ_64MUL_2(*dst, mask); 827 } 828 } 829 } 830 831 src = rm->rm_col[VDEV_RAIDZ_Q].rc_data; 832 dst = rm->rm_col[x].rc_data; 833 exp = 255 - (rm->rm_cols - 1 - x); 834 835 for (i = 0; i < xcount; i++, dst++, src++) { 836 *dst ^= *src; 837 for (j = 0, b = (uint8_t *)dst; j < 8; j++, b++) { 838 *b = vdev_raidz_exp2(*b, exp); 839 } 840 } 841 842 return (1 << VDEV_RAIDZ_Q); 843 } 844 845 static int 846 vdev_raidz_reconstruct_pq(raidz_map_t *rm, int *tgts, int ntgts) 847 { 848 uint8_t *p, *q, *pxy, *qxy, *xd, *yd, tmp, a, b, aexp, bexp; 849 void *pdata, *qdata; 850 uint64_t xsize, ysize, i; 851 int x = tgts[0]; 852 int y = tgts[1]; 853 854 ASSERT(ntgts == 2); 855 ASSERT(x < y); 856 ASSERT(x >= rm->rm_firstdatacol); 857 ASSERT(y < rm->rm_cols); 858 859 ASSERT(rm->rm_col[x].rc_size >= rm->rm_col[y].rc_size); 860 861 /* 862 * Move the parity data aside -- we're going to compute parity as 863 * though columns x and y were full of zeros -- Pxy and Qxy. We want to 864 * reuse the parity generation mechanism without trashing the actual 865 * parity so we make those columns appear to be full of zeros by 866 * setting their lengths to zero. 867 */ 868 pdata = rm->rm_col[VDEV_RAIDZ_P].rc_data; 869 qdata = rm->rm_col[VDEV_RAIDZ_Q].rc_data; 870 xsize = rm->rm_col[x].rc_size; 871 ysize = rm->rm_col[y].rc_size; 872 873 rm->rm_col[VDEV_RAIDZ_P].rc_data = 874 zio_buf_alloc(rm->rm_col[VDEV_RAIDZ_P].rc_size); 875 rm->rm_col[VDEV_RAIDZ_Q].rc_data = 876 zio_buf_alloc(rm->rm_col[VDEV_RAIDZ_Q].rc_size); 877 rm->rm_col[x].rc_size = 0; 878 rm->rm_col[y].rc_size = 0; 879 880 vdev_raidz_generate_parity_pq(rm); 881 882 rm->rm_col[x].rc_size = xsize; 883 rm->rm_col[y].rc_size = ysize; 884 885 p = pdata; 886 q = qdata; 887 pxy = rm->rm_col[VDEV_RAIDZ_P].rc_data; 888 qxy = rm->rm_col[VDEV_RAIDZ_Q].rc_data; 889 xd = rm->rm_col[x].rc_data; 890 yd = rm->rm_col[y].rc_data; 891 892 /* 893 * We now have: 894 * Pxy = P + D_x + D_y 895 * Qxy = Q + 2^(ndevs - 1 - x) * D_x + 2^(ndevs - 1 - y) * D_y 896 * 897 * We can then solve for D_x: 898 * D_x = A * (P + Pxy) + B * (Q + Qxy) 899 * where 900 * A = 2^(x - y) * (2^(x - y) + 1)^-1 901 * B = 2^(ndevs - 1 - x) * (2^(x - y) + 1)^-1 902 * 903 * With D_x in hand, we can easily solve for D_y: 904 * D_y = P + Pxy + D_x 905 */ 906 907 a = vdev_raidz_pow2[255 + x - y]; 908 b = vdev_raidz_pow2[255 - (rm->rm_cols - 1 - x)]; 909 tmp = 255 - vdev_raidz_log2[a ^ 1]; 910 911 aexp = vdev_raidz_log2[vdev_raidz_exp2(a, tmp)]; 912 bexp = vdev_raidz_log2[vdev_raidz_exp2(b, tmp)]; 913 914 for (i = 0; i < xsize; i++, p++, q++, pxy++, qxy++, xd++, yd++) { 915 *xd = vdev_raidz_exp2(*p ^ *pxy, aexp) ^ 916 vdev_raidz_exp2(*q ^ *qxy, bexp); 917 918 if (i < ysize) 919 *yd = *p ^ *pxy ^ *xd; 920 } 921 922 zio_buf_free(rm->rm_col[VDEV_RAIDZ_P].rc_data, 923 rm->rm_col[VDEV_RAIDZ_P].rc_size); 924 zio_buf_free(rm->rm_col[VDEV_RAIDZ_Q].rc_data, 925 rm->rm_col[VDEV_RAIDZ_Q].rc_size); 926 927 /* 928 * Restore the saved parity data. 929 */ 930 rm->rm_col[VDEV_RAIDZ_P].rc_data = pdata; 931 rm->rm_col[VDEV_RAIDZ_Q].rc_data = qdata; 932 933 return ((1 << VDEV_RAIDZ_P) | (1 << VDEV_RAIDZ_Q)); 934 } 935 936 /* BEGIN CSTYLED */ 937 /* 938 * In the general case of reconstruction, we must solve the system of linear 939 * equations defined by the coeffecients used to generate parity as well as 940 * the contents of the data and parity disks. This can be expressed with 941 * vectors for the original data (D) and the actual data (d) and parity (p) 942 * and a matrix composed of the identity matrix (I) and a dispersal matrix (V): 943 * 944 * __ __ __ __ 945 * | | __ __ | p_0 | 946 * | V | | D_0 | | p_m-1 | 947 * | | x | : | = | d_0 | 948 * | I | | D_n-1 | | : | 949 * | | ~~ ~~ | d_n-1 | 950 * ~~ ~~ ~~ ~~ 951 * 952 * I is simply a square identity matrix of size n, and V is a vandermonde 953 * matrix defined by the coeffecients we chose for the various parity columns 954 * (1, 2, 4). Note that these values were chosen both for simplicity, speedy 955 * computation as well as linear separability. 956 * 957 * __ __ __ __ 958 * | 1 .. 1 1 1 | | p_0 | 959 * | 2^n-1 .. 4 2 1 | __ __ | : | 960 * | 4^n-1 .. 16 4 1 | | D_0 | | p_m-1 | 961 * | 1 .. 0 0 0 | | D_1 | | d_0 | 962 * | 0 .. 0 0 0 | x | D_2 | = | d_1 | 963 * | : : : : | | : | | d_2 | 964 * | 0 .. 1 0 0 | | D_n-1 | | : | 965 * | 0 .. 0 1 0 | ~~ ~~ | : | 966 * | 0 .. 0 0 1 | | d_n-1 | 967 * ~~ ~~ ~~ ~~ 968 * 969 * Note that I, V, d, and p are known. To compute D, we must invert the 970 * matrix and use the known data and parity values to reconstruct the unknown 971 * data values. We begin by removing the rows in V|I and d|p that correspond 972 * to failed or missing columns; we then make V|I square (n x n) and d|p 973 * sized n by removing rows corresponding to unused parity from the bottom up 974 * to generate (V|I)' and (d|p)'. We can then generate the inverse of (V|I)' 975 * using Gauss-Jordan elimination. In the example below we use m=3 parity 976 * columns, n=8 data columns, with errors in d_1, d_2, and p_1: 977 * __ __ 978 * | 1 1 1 1 1 1 1 1 | 979 * | 128 64 32 16 8 4 2 1 | <-----+-+-- missing disks 980 * | 19 205 116 29 64 16 4 1 | / / 981 * | 1 0 0 0 0 0 0 0 | / / 982 * | 0 1 0 0 0 0 0 0 | <--' / 983 * (V|I) = | 0 0 1 0 0 0 0 0 | <---' 984 * | 0 0 0 1 0 0 0 0 | 985 * | 0 0 0 0 1 0 0 0 | 986 * | 0 0 0 0 0 1 0 0 | 987 * | 0 0 0 0 0 0 1 0 | 988 * | 0 0 0 0 0 0 0 1 | 989 * ~~ ~~ 990 * __ __ 991 * | 1 1 1 1 1 1 1 1 | 992 * | 19 205 116 29 64 16 4 1 | 993 * | 1 0 0 0 0 0 0 0 | 994 * (V|I)' = | 0 0 0 1 0 0 0 0 | 995 * | 0 0 0 0 1 0 0 0 | 996 * | 0 0 0 0 0 1 0 0 | 997 * | 0 0 0 0 0 0 1 0 | 998 * | 0 0 0 0 0 0 0 1 | 999 * ~~ ~~ 1000 * 1001 * Here we employ Gauss-Jordan elimination to find the inverse of (V|I)'. We 1002 * have carefully chosen the seed values 1, 2, and 4 to ensure that this 1003 * matrix is not singular. 1004 * __ __ 1005 * | 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 | 1006 * | 19 205 116 29 64 16 4 1 0 1 0 0 0 0 0 0 | 1007 * | 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 | 1008 * | 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 | 1009 * | 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 | 1010 * | 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 | 1011 * | 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 | 1012 * | 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 | 1013 * ~~ ~~ 1014 * __ __ 1015 * | 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 | 1016 * | 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 | 1017 * | 19 205 116 29 64 16 4 1 0 1 0 0 0 0 0 0 | 1018 * | 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 | 1019 * | 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 | 1020 * | 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 | 1021 * | 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 | 1022 * | 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 | 1023 * ~~ ~~ 1024 * __ __ 1025 * | 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 | 1026 * | 0 1 1 0 0 0 0 0 1 0 1 1 1 1 1 1 | 1027 * | 0 205 116 0 0 0 0 0 0 1 19 29 64 16 4 1 | 1028 * | 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 | 1029 * | 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 | 1030 * | 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 | 1031 * | 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 | 1032 * | 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 | 1033 * ~~ ~~ 1034 * __ __ 1035 * | 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 | 1036 * | 0 1 1 0 0 0 0 0 1 0 1 1 1 1 1 1 | 1037 * | 0 0 185 0 0 0 0 0 205 1 222 208 141 221 201 204 | 1038 * | 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 | 1039 * | 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 | 1040 * | 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 | 1041 * | 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 | 1042 * | 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 | 1043 * ~~ ~~ 1044 * __ __ 1045 * | 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 | 1046 * | 0 1 1 0 0 0 0 0 1 0 1 1 1 1 1 1 | 1047 * | 0 0 1 0 0 0 0 0 166 100 4 40 158 168 216 209 | 1048 * | 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 | 1049 * | 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 | 1050 * | 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 | 1051 * | 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 | 1052 * | 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 | 1053 * ~~ ~~ 1054 * __ __ 1055 * | 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 | 1056 * | 0 1 0 0 0 0 0 0 167 100 5 41 159 169 217 208 | 1057 * | 0 0 1 0 0 0 0 0 166 100 4 40 158 168 216 209 | 1058 * | 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 | 1059 * | 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 | 1060 * | 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 | 1061 * | 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 | 1062 * | 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 | 1063 * ~~ ~~ 1064 * __ __ 1065 * | 0 0 1 0 0 0 0 0 | 1066 * | 167 100 5 41 159 169 217 208 | 1067 * | 166 100 4 40 158 168 216 209 | 1068 * (V|I)'^-1 = | 0 0 0 1 0 0 0 0 | 1069 * | 0 0 0 0 1 0 0 0 | 1070 * | 0 0 0 0 0 1 0 0 | 1071 * | 0 0 0 0 0 0 1 0 | 1072 * | 0 0 0 0 0 0 0 1 | 1073 * ~~ ~~ 1074 * 1075 * We can then simply compute D = (V|I)'^-1 x (d|p)' to discover the values 1076 * of the missing data. 1077 * 1078 * As is apparent from the example above, the only non-trivial rows in the 1079 * inverse matrix correspond to the data disks that we're trying to 1080 * reconstruct. Indeed, those are the only rows we need as the others would 1081 * only be useful for reconstructing data known or assumed to be valid. For 1082 * that reason, we only build the coefficients in the rows that correspond to 1083 * targeted columns. 1084 */ 1085 /* END CSTYLED */ 1086 1087 static void 1088 vdev_raidz_matrix_init(raidz_map_t *rm, int n, int nmap, int *map, 1089 uint8_t **rows) 1090 { 1091 int i, j; 1092 int pow; 1093 1094 ASSERT(n == rm->rm_cols - rm->rm_firstdatacol); 1095 1096 /* 1097 * Fill in the missing rows of interest. 1098 */ 1099 for (i = 0; i < nmap; i++) { 1100 ASSERT3S(0, <=, map[i]); 1101 ASSERT3S(map[i], <=, 2); 1102 1103 pow = map[i] * n; 1104 if (pow > 255) 1105 pow -= 255; 1106 ASSERT(pow <= 255); 1107 1108 for (j = 0; j < n; j++) { 1109 pow -= map[i]; 1110 if (pow < 0) 1111 pow += 255; 1112 rows[i][j] = vdev_raidz_pow2[pow]; 1113 } 1114 } 1115 } 1116 1117 static void 1118 vdev_raidz_matrix_invert(raidz_map_t *rm, int n, int nmissing, int *missing, 1119 uint8_t **rows, uint8_t **invrows, const uint8_t *used) 1120 { 1121 int i, j, ii, jj; 1122 uint8_t log; 1123 1124 /* 1125 * Assert that the first nmissing entries from the array of used 1126 * columns correspond to parity columns and that subsequent entries 1127 * correspond to data columns. 1128 */ 1129 for (i = 0; i < nmissing; i++) { 1130 ASSERT3S(used[i], <, rm->rm_firstdatacol); 1131 } 1132 for (; i < n; i++) { 1133 ASSERT3S(used[i], >=, rm->rm_firstdatacol); 1134 } 1135 1136 /* 1137 * First initialize the storage where we'll compute the inverse rows. 1138 */ 1139 for (i = 0; i < nmissing; i++) { 1140 for (j = 0; j < n; j++) { 1141 invrows[i][j] = (i == j) ? 1 : 0; 1142 } 1143 } 1144 1145 /* 1146 * Subtract all trivial rows from the rows of consequence. 1147 */ 1148 for (i = 0; i < nmissing; i++) { 1149 for (j = nmissing; j < n; j++) { 1150 ASSERT3U(used[j], >=, rm->rm_firstdatacol); 1151 jj = used[j] - rm->rm_firstdatacol; 1152 ASSERT3S(jj, <, n); 1153 invrows[i][j] = rows[i][jj]; 1154 rows[i][jj] = 0; 1155 } 1156 } 1157 1158 /* 1159 * For each of the rows of interest, we must normalize it and subtract 1160 * a multiple of it from the other rows. 1161 */ 1162 for (i = 0; i < nmissing; i++) { 1163 for (j = 0; j < missing[i]; j++) { 1164 ASSERT0(rows[i][j]); 1165 } 1166 ASSERT3U(rows[i][missing[i]], !=, 0); 1167 1168 /* 1169 * Compute the inverse of the first element and multiply each 1170 * element in the row by that value. 1171 */ 1172 log = 255 - vdev_raidz_log2[rows[i][missing[i]]]; 1173 1174 for (j = 0; j < n; j++) { 1175 rows[i][j] = vdev_raidz_exp2(rows[i][j], log); 1176 invrows[i][j] = vdev_raidz_exp2(invrows[i][j], log); 1177 } 1178 1179 for (ii = 0; ii < nmissing; ii++) { 1180 if (i == ii) 1181 continue; 1182 1183 ASSERT3U(rows[ii][missing[i]], !=, 0); 1184 1185 log = vdev_raidz_log2[rows[ii][missing[i]]]; 1186 1187 for (j = 0; j < n; j++) { 1188 rows[ii][j] ^= 1189 vdev_raidz_exp2(rows[i][j], log); 1190 invrows[ii][j] ^= 1191 vdev_raidz_exp2(invrows[i][j], log); 1192 } 1193 } 1194 } 1195 1196 /* 1197 * Verify that the data that is left in the rows are properly part of 1198 * an identity matrix. 1199 */ 1200 for (i = 0; i < nmissing; i++) { 1201 for (j = 0; j < n; j++) { 1202 if (j == missing[i]) { 1203 ASSERT3U(rows[i][j], ==, 1); 1204 } else { 1205 ASSERT0(rows[i][j]); 1206 } 1207 } 1208 } 1209 } 1210 1211 static void 1212 vdev_raidz_matrix_reconstruct(raidz_map_t *rm, int n, int nmissing, 1213 int *missing, uint8_t **invrows, const uint8_t *used) 1214 { 1215 int i, j, x, cc, c; 1216 uint8_t *src; 1217 uint64_t ccount; 1218 uint8_t *dst[VDEV_RAIDZ_MAXPARITY]; 1219 uint64_t dcount[VDEV_RAIDZ_MAXPARITY]; 1220 uint8_t log = 0; 1221 uint8_t val; 1222 int ll; 1223 uint8_t *invlog[VDEV_RAIDZ_MAXPARITY]; 1224 uint8_t *p, *pp; 1225 size_t psize; 1226 1227 psize = sizeof (invlog[0][0]) * n * nmissing; 1228 p = kmem_alloc(psize, KM_SLEEP); 1229 1230 for (pp = p, i = 0; i < nmissing; i++) { 1231 invlog[i] = pp; 1232 pp += n; 1233 } 1234 1235 for (i = 0; i < nmissing; i++) { 1236 for (j = 0; j < n; j++) { 1237 ASSERT3U(invrows[i][j], !=, 0); 1238 invlog[i][j] = vdev_raidz_log2[invrows[i][j]]; 1239 } 1240 } 1241 1242 for (i = 0; i < n; i++) { 1243 c = used[i]; 1244 ASSERT3U(c, <, rm->rm_cols); 1245 1246 src = rm->rm_col[c].rc_data; 1247 ccount = rm->rm_col[c].rc_size; 1248 for (j = 0; j < nmissing; j++) { 1249 cc = missing[j] + rm->rm_firstdatacol; 1250 ASSERT3U(cc, >=, rm->rm_firstdatacol); 1251 ASSERT3U(cc, <, rm->rm_cols); 1252 ASSERT3U(cc, !=, c); 1253 1254 dst[j] = rm->rm_col[cc].rc_data; 1255 dcount[j] = rm->rm_col[cc].rc_size; 1256 } 1257 1258 ASSERT(ccount >= rm->rm_col[missing[0]].rc_size || i > 0); 1259 1260 for (x = 0; x < ccount; x++, src++) { 1261 if (*src != 0) 1262 log = vdev_raidz_log2[*src]; 1263 1264 for (cc = 0; cc < nmissing; cc++) { 1265 if (x >= dcount[cc]) 1266 continue; 1267 1268 if (*src == 0) { 1269 val = 0; 1270 } else { 1271 if ((ll = log + invlog[cc][i]) >= 255) 1272 ll -= 255; 1273 val = vdev_raidz_pow2[ll]; 1274 } 1275 1276 if (i == 0) 1277 dst[cc][x] = val; 1278 else 1279 dst[cc][x] ^= val; 1280 } 1281 } 1282 } 1283 1284 kmem_free(p, psize); 1285 } 1286 1287 static int 1288 vdev_raidz_reconstruct_general(raidz_map_t *rm, int *tgts, int ntgts) 1289 { 1290 int n, i, c, t, tt; 1291 int nmissing_rows; 1292 int missing_rows[VDEV_RAIDZ_MAXPARITY]; 1293 int parity_map[VDEV_RAIDZ_MAXPARITY]; 1294 1295 uint8_t *p, *pp; 1296 size_t psize; 1297 1298 uint8_t *rows[VDEV_RAIDZ_MAXPARITY]; 1299 uint8_t *invrows[VDEV_RAIDZ_MAXPARITY]; 1300 uint8_t *used; 1301 1302 int code = 0; 1303 1304 1305 n = rm->rm_cols - rm->rm_firstdatacol; 1306 1307 /* 1308 * Figure out which data columns are missing. 1309 */ 1310 nmissing_rows = 0; 1311 for (t = 0; t < ntgts; t++) { 1312 if (tgts[t] >= rm->rm_firstdatacol) { 1313 missing_rows[nmissing_rows++] = 1314 tgts[t] - rm->rm_firstdatacol; 1315 } 1316 } 1317 1318 /* 1319 * Figure out which parity columns to use to help generate the missing 1320 * data columns. 1321 */ 1322 for (tt = 0, c = 0, i = 0; i < nmissing_rows; c++) { 1323 ASSERT(tt < ntgts); 1324 ASSERT(c < rm->rm_firstdatacol); 1325 1326 /* 1327 * Skip any targeted parity columns. 1328 */ 1329 if (c == tgts[tt]) { 1330 tt++; 1331 continue; 1332 } 1333 1334 code |= 1 << c; 1335 1336 parity_map[i] = c; 1337 i++; 1338 } 1339 1340 ASSERT(code != 0); 1341 ASSERT3U(code, <, 1 << VDEV_RAIDZ_MAXPARITY); 1342 1343 psize = (sizeof (rows[0][0]) + sizeof (invrows[0][0])) * 1344 nmissing_rows * n + sizeof (used[0]) * n; 1345 p = kmem_alloc(psize, KM_SLEEP); 1346 1347 for (pp = p, i = 0; i < nmissing_rows; i++) { 1348 rows[i] = pp; 1349 pp += n; 1350 invrows[i] = pp; 1351 pp += n; 1352 } 1353 used = pp; 1354 1355 for (i = 0; i < nmissing_rows; i++) { 1356 used[i] = parity_map[i]; 1357 } 1358 1359 for (tt = 0, c = rm->rm_firstdatacol; c < rm->rm_cols; c++) { 1360 if (tt < nmissing_rows && 1361 c == missing_rows[tt] + rm->rm_firstdatacol) { 1362 tt++; 1363 continue; 1364 } 1365 1366 ASSERT3S(i, <, n); 1367 used[i] = c; 1368 i++; 1369 } 1370 1371 /* 1372 * Initialize the interesting rows of the matrix. 1373 */ 1374 vdev_raidz_matrix_init(rm, n, nmissing_rows, parity_map, rows); 1375 1376 /* 1377 * Invert the matrix. 1378 */ 1379 vdev_raidz_matrix_invert(rm, n, nmissing_rows, missing_rows, rows, 1380 invrows, used); 1381 1382 /* 1383 * Reconstruct the missing data using the generated matrix. 1384 */ 1385 vdev_raidz_matrix_reconstruct(rm, n, nmissing_rows, missing_rows, 1386 invrows, used); 1387 1388 kmem_free(p, psize); 1389 1390 return (code); 1391 } 1392 1393 static int 1394 vdev_raidz_reconstruct(raidz_map_t *rm, int *t, int nt) 1395 { 1396 int tgts[VDEV_RAIDZ_MAXPARITY], *dt; 1397 int ntgts; 1398 int i, c; 1399 int code; 1400 int nbadparity, nbaddata; 1401 int parity_valid[VDEV_RAIDZ_MAXPARITY]; 1402 1403 /* 1404 * The tgts list must already be sorted. 1405 */ 1406 for (i = 1; i < nt; i++) { 1407 ASSERT(t[i] > t[i - 1]); 1408 } 1409 1410 nbadparity = rm->rm_firstdatacol; 1411 nbaddata = rm->rm_cols - nbadparity; 1412 ntgts = 0; 1413 for (i = 0, c = 0; c < rm->rm_cols; c++) { 1414 if (c < rm->rm_firstdatacol) 1415 parity_valid[c] = B_FALSE; 1416 1417 if (i < nt && c == t[i]) { 1418 tgts[ntgts++] = c; 1419 i++; 1420 } else if (rm->rm_col[c].rc_error != 0) { 1421 tgts[ntgts++] = c; 1422 } else if (c >= rm->rm_firstdatacol) { 1423 nbaddata--; 1424 } else { 1425 parity_valid[c] = B_TRUE; 1426 nbadparity--; 1427 } 1428 } 1429 1430 ASSERT(ntgts >= nt); 1431 ASSERT(nbaddata >= 0); 1432 ASSERT(nbaddata + nbadparity == ntgts); 1433 1434 dt = &tgts[nbadparity]; 1435 1436 /* 1437 * See if we can use any of our optimized reconstruction routines. 1438 */ 1439 if (!vdev_raidz_default_to_general) { 1440 switch (nbaddata) { 1441 case 1: 1442 if (parity_valid[VDEV_RAIDZ_P]) 1443 return (vdev_raidz_reconstruct_p(rm, dt, 1)); 1444 1445 ASSERT(rm->rm_firstdatacol > 1); 1446 1447 if (parity_valid[VDEV_RAIDZ_Q]) 1448 return (vdev_raidz_reconstruct_q(rm, dt, 1)); 1449 1450 ASSERT(rm->rm_firstdatacol > 2); 1451 break; 1452 1453 case 2: 1454 ASSERT(rm->rm_firstdatacol > 1); 1455 1456 if (parity_valid[VDEV_RAIDZ_P] && 1457 parity_valid[VDEV_RAIDZ_Q]) 1458 return (vdev_raidz_reconstruct_pq(rm, dt, 2)); 1459 1460 ASSERT(rm->rm_firstdatacol > 2); 1461 1462 break; 1463 } 1464 } 1465 1466 code = vdev_raidz_reconstruct_general(rm, tgts, ntgts); 1467 ASSERT(code < (1 << VDEV_RAIDZ_MAXPARITY)); 1468 ASSERT(code > 0); 1469 return (code); 1470 } 1471 1472 static int 1473 vdev_raidz_open(vdev_t *vd, uint64_t *asize, uint64_t *max_asize, 1474 uint64_t *ashift) 1475 { 1476 vdev_t *cvd; 1477 uint64_t nparity = vd->vdev_nparity; 1478 int c; 1479 int lasterror = 0; 1480 int numerrors = 0; 1481 1482 ASSERT(nparity > 0); 1483 1484 if (nparity > VDEV_RAIDZ_MAXPARITY || 1485 vd->vdev_children < nparity + 1) { 1486 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; 1487 return (SET_ERROR(EINVAL)); 1488 } 1489 1490 vdev_open_children(vd); 1491 1492 for (c = 0; c < vd->vdev_children; c++) { 1493 cvd = vd->vdev_child[c]; 1494 1495 if (cvd->vdev_open_error != 0) { 1496 lasterror = cvd->vdev_open_error; 1497 numerrors++; 1498 continue; 1499 } 1500 1501 *asize = MIN(*asize - 1, cvd->vdev_asize - 1) + 1; 1502 *max_asize = MIN(*max_asize - 1, cvd->vdev_max_asize - 1) + 1; 1503 *ashift = MAX(*ashift, cvd->vdev_ashift); 1504 } 1505 1506 *asize *= vd->vdev_children; 1507 *max_asize *= vd->vdev_children; 1508 1509 if (numerrors > nparity) { 1510 vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS; 1511 return (lasterror); 1512 } 1513 1514 return (0); 1515 } 1516 1517 static void 1518 vdev_raidz_close(vdev_t *vd) 1519 { 1520 int c; 1521 1522 for (c = 0; c < vd->vdev_children; c++) 1523 vdev_close(vd->vdev_child[c]); 1524 } 1525 1526 /* 1527 * Handle a read or write I/O to a RAID-Z dump device. 1528 * 1529 * The dump device is in a unique situation compared to other ZFS datasets: 1530 * writing to this device should be as simple and fast as possible. In 1531 * addition, durability matters much less since the dump will be extracted 1532 * once the machine reboots. For that reason, this function eschews parity for 1533 * performance and simplicity. The dump device uses the checksum setting 1534 * ZIO_CHECKSUM_NOPARITY to indicate that parity is not maintained for this 1535 * dataset. 1536 * 1537 * Blocks of size 128 KB have been preallocated for this volume. I/Os less than 1538 * 128 KB will not fill an entire block; in addition, they may not be properly 1539 * aligned. In that case, this function uses the preallocated 128 KB block and 1540 * omits reading or writing any "empty" portions of that block, as opposed to 1541 * allocating a fresh appropriately-sized block. 1542 * 1543 * Looking at an example of a 32 KB I/O to a RAID-Z vdev with 5 child vdevs: 1544 * 1545 * vdev_raidz_io_start(data, size: 32 KB, offset: 64 KB) 1546 * 1547 * If this were a standard RAID-Z dataset, a block of at least 40 KB would be 1548 * allocated which spans all five child vdevs. 8 KB of data would be written to 1549 * each of four vdevs, with the fifth containing the parity bits. 1550 * 1551 * parity data data data data 1552 * | PP | XX | XX | XX | XX | 1553 * ^ ^ ^ ^ ^ 1554 * | | | | | 1555 * 8 KB parity ------8 KB data blocks------ 1556 * 1557 * However, when writing to the dump device, the behavior is different: 1558 * 1559 * vdev_raidz_physio(data, size: 32 KB, offset: 64 KB) 1560 * 1561 * Unlike the normal RAID-Z case in which the block is allocated based on the 1562 * I/O size, reads and writes here always use a 128 KB logical I/O size. If the 1563 * I/O size is less than 128 KB, only the actual portions of data are written. 1564 * In this example the data is written to the third data vdev since that vdev 1565 * contains the offset [64 KB, 96 KB). 1566 * 1567 * parity data data data data 1568 * | | | | XX | | 1569 * ^ 1570 * | 1571 * 32 KB data block 1572 * 1573 * As a result, an individual I/O may not span all child vdevs; moreover, a 1574 * small I/O may only operate on a single child vdev. 1575 * 1576 * Note that since there are no parity bits calculated or written, this format 1577 * remains the same no matter how many parity bits are used in a normal RAID-Z 1578 * stripe. On a RAID-Z3 configuration with seven child vdevs, the example above 1579 * would look like: 1580 * 1581 * parity parity parity data data data data 1582 * | | | | | | XX | | 1583 * ^ 1584 * | 1585 * 32 KB data block 1586 */ 1587 int 1588 vdev_raidz_physio(vdev_t *vd, caddr_t data, size_t size, 1589 uint64_t offset, uint64_t origoffset, boolean_t doread, boolean_t isdump) 1590 { 1591 vdev_t *tvd = vd->vdev_top; 1592 vdev_t *cvd; 1593 raidz_map_t *rm; 1594 raidz_col_t *rc; 1595 int c, err = 0; 1596 1597 uint64_t start, end, colstart, colend; 1598 uint64_t coloffset, colsize, colskip; 1599 1600 int flags = doread ? B_READ : B_WRITE; 1601 1602 #ifdef _KERNEL 1603 1604 /* 1605 * Don't write past the end of the block 1606 */ 1607 VERIFY3U(offset + size, <=, origoffset + SPA_MAXBLOCKSIZE); 1608 1609 start = offset; 1610 end = start + size; 1611 1612 /* 1613 * Allocate a RAID-Z map for this block. Note that this block starts 1614 * from the "original" offset, this is, the offset of the extent which 1615 * contains the requisite offset of the data being read or written. 1616 * 1617 * Even if this I/O operation doesn't span the full block size, let's 1618 * treat the on-disk format as if the only blocks are the complete 128 1619 * KB size. 1620 */ 1621 rm = vdev_raidz_map_alloc(data - (offset - origoffset), 1622 SPA_MAXBLOCKSIZE, origoffset, tvd->vdev_ashift, vd->vdev_children, 1623 vd->vdev_nparity); 1624 1625 coloffset = origoffset; 1626 1627 for (c = rm->rm_firstdatacol; c < rm->rm_cols; 1628 c++, coloffset += rc->rc_size) { 1629 rc = &rm->rm_col[c]; 1630 cvd = vd->vdev_child[rc->rc_devidx]; 1631 1632 /* 1633 * Find the start and end of this column in the RAID-Z map, 1634 * keeping in mind that the stated size and offset of the 1635 * operation may not fill the entire column for this vdev. 1636 * 1637 * If any portion of the data spans this column, issue the 1638 * appropriate operation to the vdev. 1639 */ 1640 if (coloffset + rc->rc_size <= start) 1641 continue; 1642 if (coloffset >= end) 1643 continue; 1644 1645 colstart = MAX(coloffset, start); 1646 colend = MIN(end, coloffset + rc->rc_size); 1647 colsize = colend - colstart; 1648 colskip = colstart - coloffset; 1649 1650 VERIFY3U(colsize, <=, rc->rc_size); 1651 VERIFY3U(colskip, <=, rc->rc_size); 1652 1653 /* 1654 * Note that the child vdev will have a vdev label at the start 1655 * of its range of offsets, hence the need for 1656 * VDEV_LABEL_OFFSET(). See zio_vdev_child_io() for another 1657 * example of why this calculation is needed. 1658 */ 1659 if ((err = vdev_disk_physio(cvd, 1660 ((char *)rc->rc_data) + colskip, colsize, 1661 VDEV_LABEL_OFFSET(rc->rc_offset) + colskip, 1662 flags, isdump)) != 0) 1663 break; 1664 } 1665 1666 vdev_raidz_map_free(rm); 1667 #endif /* KERNEL */ 1668 1669 return (err); 1670 } 1671 1672 static uint64_t 1673 vdev_raidz_asize(vdev_t *vd, uint64_t psize) 1674 { 1675 uint64_t asize; 1676 uint64_t ashift = vd->vdev_top->vdev_ashift; 1677 uint64_t cols = vd->vdev_children; 1678 uint64_t nparity = vd->vdev_nparity; 1679 1680 asize = ((psize - 1) >> ashift) + 1; 1681 asize += nparity * ((asize + cols - nparity - 1) / (cols - nparity)); 1682 asize = roundup(asize, nparity + 1) << ashift; 1683 1684 return (asize); 1685 } 1686 1687 static void 1688 vdev_raidz_child_done(zio_t *zio) 1689 { 1690 raidz_col_t *rc = zio->io_private; 1691 1692 rc->rc_error = zio->io_error; 1693 rc->rc_tried = 1; 1694 rc->rc_skipped = 0; 1695 } 1696 1697 /* 1698 * Start an IO operation on a RAIDZ VDev 1699 * 1700 * Outline: 1701 * - For write operations: 1702 * 1. Generate the parity data 1703 * 2. Create child zio write operations to each column's vdev, for both 1704 * data and parity. 1705 * 3. If the column skips any sectors for padding, create optional dummy 1706 * write zio children for those areas to improve aggregation continuity. 1707 * - For read operations: 1708 * 1. Create child zio read operations to each data column's vdev to read 1709 * the range of data required for zio. 1710 * 2. If this is a scrub or resilver operation, or if any of the data 1711 * vdevs have had errors, then create zio read operations to the parity 1712 * columns' VDevs as well. 1713 */ 1714 static int 1715 vdev_raidz_io_start(zio_t *zio) 1716 { 1717 vdev_t *vd = zio->io_vd; 1718 vdev_t *tvd = vd->vdev_top; 1719 vdev_t *cvd; 1720 raidz_map_t *rm; 1721 raidz_col_t *rc; 1722 int c, i; 1723 1724 rm = vdev_raidz_map_alloc(zio->io_data, zio->io_size, zio->io_offset, 1725 tvd->vdev_ashift, vd->vdev_children, 1726 vd->vdev_nparity); 1727 1728 zio->io_vsd = rm; 1729 zio->io_vsd_ops = &vdev_raidz_vsd_ops; 1730 1731 ASSERT3U(rm->rm_asize, ==, vdev_psize_to_asize(vd, zio->io_size)); 1732 1733 if (zio->io_type == ZIO_TYPE_WRITE) { 1734 vdev_raidz_generate_parity(rm); 1735 1736 for (c = 0; c < rm->rm_cols; c++) { 1737 rc = &rm->rm_col[c]; 1738 cvd = vd->vdev_child[rc->rc_devidx]; 1739 zio_nowait(zio_vdev_child_io(zio, NULL, cvd, 1740 rc->rc_offset, rc->rc_data, rc->rc_size, 1741 zio->io_type, zio->io_priority, 0, 1742 vdev_raidz_child_done, rc)); 1743 } 1744 1745 /* 1746 * Generate optional I/Os for any skipped sectors to improve 1747 * aggregation contiguity. 1748 */ 1749 for (c = rm->rm_skipstart, i = 0; i < rm->rm_nskip; c++, i++) { 1750 ASSERT(c <= rm->rm_scols); 1751 if (c == rm->rm_scols) 1752 c = 0; 1753 rc = &rm->rm_col[c]; 1754 cvd = vd->vdev_child[rc->rc_devidx]; 1755 zio_nowait(zio_vdev_child_io(zio, NULL, cvd, 1756 rc->rc_offset + rc->rc_size, NULL, 1757 1 << tvd->vdev_ashift, 1758 zio->io_type, zio->io_priority, 1759 ZIO_FLAG_NODATA | ZIO_FLAG_OPTIONAL, NULL, NULL)); 1760 } 1761 1762 return (ZIO_PIPELINE_CONTINUE); 1763 } 1764 1765 ASSERT(zio->io_type == ZIO_TYPE_READ); 1766 1767 /* 1768 * Iterate over the columns in reverse order so that we hit the parity 1769 * last -- any errors along the way will force us to read the parity. 1770 */ 1771 for (c = rm->rm_cols - 1; c >= 0; c--) { 1772 rc = &rm->rm_col[c]; 1773 cvd = vd->vdev_child[rc->rc_devidx]; 1774 if (!vdev_readable(cvd)) { 1775 if (c >= rm->rm_firstdatacol) 1776 rm->rm_missingdata++; 1777 else 1778 rm->rm_missingparity++; 1779 rc->rc_error = SET_ERROR(ENXIO); 1780 rc->rc_tried = 1; /* don't even try */ 1781 rc->rc_skipped = 1; 1782 continue; 1783 } 1784 if (vdev_dtl_contains(cvd, DTL_MISSING, zio->io_txg, 1)) { 1785 if (c >= rm->rm_firstdatacol) 1786 rm->rm_missingdata++; 1787 else 1788 rm->rm_missingparity++; 1789 rc->rc_error = SET_ERROR(ESTALE); 1790 rc->rc_skipped = 1; 1791 continue; 1792 } 1793 if (c >= rm->rm_firstdatacol || rm->rm_missingdata > 0 || 1794 (zio->io_flags & (ZIO_FLAG_SCRUB | ZIO_FLAG_RESILVER))) { 1795 zio_nowait(zio_vdev_child_io(zio, NULL, cvd, 1796 rc->rc_offset, rc->rc_data, rc->rc_size, 1797 zio->io_type, zio->io_priority, 0, 1798 vdev_raidz_child_done, rc)); 1799 } 1800 } 1801 1802 return (ZIO_PIPELINE_CONTINUE); 1803 } 1804 1805 1806 /* 1807 * Report a checksum error for a child of a RAID-Z device. 1808 */ 1809 static void 1810 raidz_checksum_error(zio_t *zio, raidz_col_t *rc, void *bad_data) 1811 { 1812 vdev_t *vd = zio->io_vd->vdev_child[rc->rc_devidx]; 1813 1814 if (!(zio->io_flags & ZIO_FLAG_SPECULATIVE)) { 1815 zio_bad_cksum_t zbc; 1816 raidz_map_t *rm = zio->io_vsd; 1817 1818 mutex_enter(&vd->vdev_stat_lock); 1819 vd->vdev_stat.vs_checksum_errors++; 1820 mutex_exit(&vd->vdev_stat_lock); 1821 1822 zbc.zbc_has_cksum = 0; 1823 zbc.zbc_injected = rm->rm_ecksuminjected; 1824 1825 zfs_ereport_post_checksum(zio->io_spa, vd, zio, 1826 rc->rc_offset, rc->rc_size, rc->rc_data, bad_data, 1827 &zbc); 1828 } 1829 } 1830 1831 /* 1832 * We keep track of whether or not there were any injected errors, so that 1833 * any ereports we generate can note it. 1834 */ 1835 static int 1836 raidz_checksum_verify(zio_t *zio) 1837 { 1838 zio_bad_cksum_t zbc; 1839 raidz_map_t *rm = zio->io_vsd; 1840 1841 int ret = zio_checksum_error(zio, &zbc); 1842 if (ret != 0 && zbc.zbc_injected != 0) 1843 rm->rm_ecksuminjected = 1; 1844 1845 return (ret); 1846 } 1847 1848 /* 1849 * Generate the parity from the data columns. If we tried and were able to 1850 * read the parity without error, verify that the generated parity matches the 1851 * data we read. If it doesn't, we fire off a checksum error. Return the 1852 * number such failures. 1853 */ 1854 static int 1855 raidz_parity_verify(zio_t *zio, raidz_map_t *rm) 1856 { 1857 void *orig[VDEV_RAIDZ_MAXPARITY]; 1858 int c, ret = 0; 1859 raidz_col_t *rc; 1860 1861 blkptr_t *bp = zio->io_bp; 1862 enum zio_checksum checksum = (bp == NULL ? zio->io_prop.zp_checksum : 1863 (BP_IS_GANG(bp) ? ZIO_CHECKSUM_GANG_HEADER : BP_GET_CHECKSUM(bp))); 1864 1865 if (checksum == ZIO_CHECKSUM_NOPARITY) 1866 return (ret); 1867 1868 for (c = 0; c < rm->rm_firstdatacol; c++) { 1869 rc = &rm->rm_col[c]; 1870 if (!rc->rc_tried || rc->rc_error != 0) 1871 continue; 1872 orig[c] = zio_buf_alloc(rc->rc_size); 1873 bcopy(rc->rc_data, orig[c], rc->rc_size); 1874 } 1875 1876 vdev_raidz_generate_parity(rm); 1877 1878 for (c = 0; c < rm->rm_firstdatacol; c++) { 1879 rc = &rm->rm_col[c]; 1880 if (!rc->rc_tried || rc->rc_error != 0) 1881 continue; 1882 if (bcmp(orig[c], rc->rc_data, rc->rc_size) != 0) { 1883 raidz_checksum_error(zio, rc, orig[c]); 1884 rc->rc_error = SET_ERROR(ECKSUM); 1885 ret++; 1886 } 1887 zio_buf_free(orig[c], rc->rc_size); 1888 } 1889 1890 return (ret); 1891 } 1892 1893 /* 1894 * Keep statistics on all the ways that we used parity to correct data. 1895 */ 1896 static uint64_t raidz_corrected[1 << VDEV_RAIDZ_MAXPARITY]; 1897 1898 static int 1899 vdev_raidz_worst_error(raidz_map_t *rm) 1900 { 1901 int error = 0; 1902 1903 for (int c = 0; c < rm->rm_cols; c++) 1904 error = zio_worst_error(error, rm->rm_col[c].rc_error); 1905 1906 return (error); 1907 } 1908 1909 /* 1910 * Iterate over all combinations of bad data and attempt a reconstruction. 1911 * Note that the algorithm below is non-optimal because it doesn't take into 1912 * account how reconstruction is actually performed. For example, with 1913 * triple-parity RAID-Z the reconstruction procedure is the same if column 4 1914 * is targeted as invalid as if columns 1 and 4 are targeted since in both 1915 * cases we'd only use parity information in column 0. 1916 */ 1917 static int 1918 vdev_raidz_combrec(zio_t *zio, int total_errors, int data_errors) 1919 { 1920 raidz_map_t *rm = zio->io_vsd; 1921 raidz_col_t *rc; 1922 void *orig[VDEV_RAIDZ_MAXPARITY]; 1923 int tstore[VDEV_RAIDZ_MAXPARITY + 2]; 1924 int *tgts = &tstore[1]; 1925 int current, next, i, c, n; 1926 int code, ret = 0; 1927 1928 ASSERT(total_errors < rm->rm_firstdatacol); 1929 1930 /* 1931 * This simplifies one edge condition. 1932 */ 1933 tgts[-1] = -1; 1934 1935 for (n = 1; n <= rm->rm_firstdatacol - total_errors; n++) { 1936 /* 1937 * Initialize the targets array by finding the first n columns 1938 * that contain no error. 1939 * 1940 * If there were no data errors, we need to ensure that we're 1941 * always explicitly attempting to reconstruct at least one 1942 * data column. To do this, we simply push the highest target 1943 * up into the data columns. 1944 */ 1945 for (c = 0, i = 0; i < n; i++) { 1946 if (i == n - 1 && data_errors == 0 && 1947 c < rm->rm_firstdatacol) { 1948 c = rm->rm_firstdatacol; 1949 } 1950 1951 while (rm->rm_col[c].rc_error != 0) { 1952 c++; 1953 ASSERT3S(c, <, rm->rm_cols); 1954 } 1955 1956 tgts[i] = c++; 1957 } 1958 1959 /* 1960 * Setting tgts[n] simplifies the other edge condition. 1961 */ 1962 tgts[n] = rm->rm_cols; 1963 1964 /* 1965 * These buffers were allocated in previous iterations. 1966 */ 1967 for (i = 0; i < n - 1; i++) { 1968 ASSERT(orig[i] != NULL); 1969 } 1970 1971 orig[n - 1] = zio_buf_alloc(rm->rm_col[0].rc_size); 1972 1973 current = 0; 1974 next = tgts[current]; 1975 1976 while (current != n) { 1977 tgts[current] = next; 1978 current = 0; 1979 1980 /* 1981 * Save off the original data that we're going to 1982 * attempt to reconstruct. 1983 */ 1984 for (i = 0; i < n; i++) { 1985 ASSERT(orig[i] != NULL); 1986 c = tgts[i]; 1987 ASSERT3S(c, >=, 0); 1988 ASSERT3S(c, <, rm->rm_cols); 1989 rc = &rm->rm_col[c]; 1990 bcopy(rc->rc_data, orig[i], rc->rc_size); 1991 } 1992 1993 /* 1994 * Attempt a reconstruction and exit the outer loop on 1995 * success. 1996 */ 1997 code = vdev_raidz_reconstruct(rm, tgts, n); 1998 if (raidz_checksum_verify(zio) == 0) { 1999 atomic_inc_64(&raidz_corrected[code]); 2000 2001 for (i = 0; i < n; i++) { 2002 c = tgts[i]; 2003 rc = &rm->rm_col[c]; 2004 ASSERT(rc->rc_error == 0); 2005 if (rc->rc_tried) 2006 raidz_checksum_error(zio, rc, 2007 orig[i]); 2008 rc->rc_error = SET_ERROR(ECKSUM); 2009 } 2010 2011 ret = code; 2012 goto done; 2013 } 2014 2015 /* 2016 * Restore the original data. 2017 */ 2018 for (i = 0; i < n; i++) { 2019 c = tgts[i]; 2020 rc = &rm->rm_col[c]; 2021 bcopy(orig[i], rc->rc_data, rc->rc_size); 2022 } 2023 2024 do { 2025 /* 2026 * Find the next valid column after the current 2027 * position.. 2028 */ 2029 for (next = tgts[current] + 1; 2030 next < rm->rm_cols && 2031 rm->rm_col[next].rc_error != 0; next++) 2032 continue; 2033 2034 ASSERT(next <= tgts[current + 1]); 2035 2036 /* 2037 * If that spot is available, we're done here. 2038 */ 2039 if (next != tgts[current + 1]) 2040 break; 2041 2042 /* 2043 * Otherwise, find the next valid column after 2044 * the previous position. 2045 */ 2046 for (c = tgts[current - 1] + 1; 2047 rm->rm_col[c].rc_error != 0; c++) 2048 continue; 2049 2050 tgts[current] = c; 2051 current++; 2052 2053 } while (current != n); 2054 } 2055 } 2056 n--; 2057 done: 2058 for (i = 0; i < n; i++) { 2059 zio_buf_free(orig[i], rm->rm_col[0].rc_size); 2060 } 2061 2062 return (ret); 2063 } 2064 2065 /* 2066 * Complete an IO operation on a RAIDZ VDev 2067 * 2068 * Outline: 2069 * - For write operations: 2070 * 1. Check for errors on the child IOs. 2071 * 2. Return, setting an error code if too few child VDevs were written 2072 * to reconstruct the data later. Note that partial writes are 2073 * considered successful if they can be reconstructed at all. 2074 * - For read operations: 2075 * 1. Check for errors on the child IOs. 2076 * 2. If data errors occurred: 2077 * a. Try to reassemble the data from the parity available. 2078 * b. If we haven't yet read the parity drives, read them now. 2079 * c. If all parity drives have been read but the data still doesn't 2080 * reassemble with a correct checksum, then try combinatorial 2081 * reconstruction. 2082 * d. If that doesn't work, return an error. 2083 * 3. If there were unexpected errors or this is a resilver operation, 2084 * rewrite the vdevs that had errors. 2085 */ 2086 static void 2087 vdev_raidz_io_done(zio_t *zio) 2088 { 2089 vdev_t *vd = zio->io_vd; 2090 vdev_t *cvd; 2091 raidz_map_t *rm = zio->io_vsd; 2092 raidz_col_t *rc; 2093 int unexpected_errors = 0; 2094 int parity_errors = 0; 2095 int parity_untried = 0; 2096 int data_errors = 0; 2097 int total_errors = 0; 2098 int n, c; 2099 int tgts[VDEV_RAIDZ_MAXPARITY]; 2100 int code; 2101 2102 ASSERT(zio->io_bp != NULL); /* XXX need to add code to enforce this */ 2103 2104 ASSERT(rm->rm_missingparity <= rm->rm_firstdatacol); 2105 ASSERT(rm->rm_missingdata <= rm->rm_cols - rm->rm_firstdatacol); 2106 2107 for (c = 0; c < rm->rm_cols; c++) { 2108 rc = &rm->rm_col[c]; 2109 2110 if (rc->rc_error) { 2111 ASSERT(rc->rc_error != ECKSUM); /* child has no bp */ 2112 2113 if (c < rm->rm_firstdatacol) 2114 parity_errors++; 2115 else 2116 data_errors++; 2117 2118 if (!rc->rc_skipped) 2119 unexpected_errors++; 2120 2121 total_errors++; 2122 } else if (c < rm->rm_firstdatacol && !rc->rc_tried) { 2123 parity_untried++; 2124 } 2125 } 2126 2127 if (zio->io_type == ZIO_TYPE_WRITE) { 2128 /* 2129 * XXX -- for now, treat partial writes as a success. 2130 * (If we couldn't write enough columns to reconstruct 2131 * the data, the I/O failed. Otherwise, good enough.) 2132 * 2133 * Now that we support write reallocation, it would be better 2134 * to treat partial failure as real failure unless there are 2135 * no non-degraded top-level vdevs left, and not update DTLs 2136 * if we intend to reallocate. 2137 */ 2138 /* XXPOLICY */ 2139 if (total_errors > rm->rm_firstdatacol) 2140 zio->io_error = vdev_raidz_worst_error(rm); 2141 2142 return; 2143 } 2144 2145 ASSERT(zio->io_type == ZIO_TYPE_READ); 2146 /* 2147 * There are three potential phases for a read: 2148 * 1. produce valid data from the columns read 2149 * 2. read all disks and try again 2150 * 3. perform combinatorial reconstruction 2151 * 2152 * Each phase is progressively both more expensive and less likely to 2153 * occur. If we encounter more errors than we can repair or all phases 2154 * fail, we have no choice but to return an error. 2155 */ 2156 2157 /* 2158 * If the number of errors we saw was correctable -- less than or equal 2159 * to the number of parity disks read -- attempt to produce data that 2160 * has a valid checksum. Naturally, this case applies in the absence of 2161 * any errors. 2162 */ 2163 if (total_errors <= rm->rm_firstdatacol - parity_untried) { 2164 if (data_errors == 0) { 2165 if (raidz_checksum_verify(zio) == 0) { 2166 /* 2167 * If we read parity information (unnecessarily 2168 * as it happens since no reconstruction was 2169 * needed) regenerate and verify the parity. 2170 * We also regenerate parity when resilvering 2171 * so we can write it out to the failed device 2172 * later. 2173 */ 2174 if (parity_errors + parity_untried < 2175 rm->rm_firstdatacol || 2176 (zio->io_flags & ZIO_FLAG_RESILVER)) { 2177 n = raidz_parity_verify(zio, rm); 2178 unexpected_errors += n; 2179 ASSERT(parity_errors + n <= 2180 rm->rm_firstdatacol); 2181 } 2182 goto done; 2183 } 2184 } else { 2185 /* 2186 * We either attempt to read all the parity columns or 2187 * none of them. If we didn't try to read parity, we 2188 * wouldn't be here in the correctable case. There must 2189 * also have been fewer parity errors than parity 2190 * columns or, again, we wouldn't be in this code path. 2191 */ 2192 ASSERT(parity_untried == 0); 2193 ASSERT(parity_errors < rm->rm_firstdatacol); 2194 2195 /* 2196 * Identify the data columns that reported an error. 2197 */ 2198 n = 0; 2199 for (c = rm->rm_firstdatacol; c < rm->rm_cols; c++) { 2200 rc = &rm->rm_col[c]; 2201 if (rc->rc_error != 0) { 2202 ASSERT(n < VDEV_RAIDZ_MAXPARITY); 2203 tgts[n++] = c; 2204 } 2205 } 2206 2207 ASSERT(rm->rm_firstdatacol >= n); 2208 2209 code = vdev_raidz_reconstruct(rm, tgts, n); 2210 2211 if (raidz_checksum_verify(zio) == 0) { 2212 atomic_inc_64(&raidz_corrected[code]); 2213 2214 /* 2215 * If we read more parity disks than were used 2216 * for reconstruction, confirm that the other 2217 * parity disks produced correct data. This 2218 * routine is suboptimal in that it regenerates 2219 * the parity that we already used in addition 2220 * to the parity that we're attempting to 2221 * verify, but this should be a relatively 2222 * uncommon case, and can be optimized if it 2223 * becomes a problem. Note that we regenerate 2224 * parity when resilvering so we can write it 2225 * out to failed devices later. 2226 */ 2227 if (parity_errors < rm->rm_firstdatacol - n || 2228 (zio->io_flags & ZIO_FLAG_RESILVER)) { 2229 n = raidz_parity_verify(zio, rm); 2230 unexpected_errors += n; 2231 ASSERT(parity_errors + n <= 2232 rm->rm_firstdatacol); 2233 } 2234 2235 goto done; 2236 } 2237 } 2238 } 2239 2240 /* 2241 * This isn't a typical situation -- either we got a read error or 2242 * a child silently returned bad data. Read every block so we can 2243 * try again with as much data and parity as we can track down. If 2244 * we've already been through once before, all children will be marked 2245 * as tried so we'll proceed to combinatorial reconstruction. 2246 */ 2247 unexpected_errors = 1; 2248 rm->rm_missingdata = 0; 2249 rm->rm_missingparity = 0; 2250 2251 for (c = 0; c < rm->rm_cols; c++) { 2252 if (rm->rm_col[c].rc_tried) 2253 continue; 2254 2255 zio_vdev_io_redone(zio); 2256 do { 2257 rc = &rm->rm_col[c]; 2258 if (rc->rc_tried) 2259 continue; 2260 zio_nowait(zio_vdev_child_io(zio, NULL, 2261 vd->vdev_child[rc->rc_devidx], 2262 rc->rc_offset, rc->rc_data, rc->rc_size, 2263 zio->io_type, zio->io_priority, 0, 2264 vdev_raidz_child_done, rc)); 2265 } while (++c < rm->rm_cols); 2266 2267 return; 2268 } 2269 2270 /* 2271 * At this point we've attempted to reconstruct the data given the 2272 * errors we detected, and we've attempted to read all columns. There 2273 * must, therefore, be one or more additional problems -- silent errors 2274 * resulting in invalid data rather than explicit I/O errors resulting 2275 * in absent data. We check if there is enough additional data to 2276 * possibly reconstruct the data and then perform combinatorial 2277 * reconstruction over all possible combinations. If that fails, 2278 * we're cooked. 2279 */ 2280 if (total_errors > rm->rm_firstdatacol) { 2281 zio->io_error = vdev_raidz_worst_error(rm); 2282 2283 } else if (total_errors < rm->rm_firstdatacol && 2284 (code = vdev_raidz_combrec(zio, total_errors, data_errors)) != 0) { 2285 /* 2286 * If we didn't use all the available parity for the 2287 * combinatorial reconstruction, verify that the remaining 2288 * parity is correct. 2289 */ 2290 if (code != (1 << rm->rm_firstdatacol) - 1) 2291 (void) raidz_parity_verify(zio, rm); 2292 } else { 2293 /* 2294 * We're here because either: 2295 * 2296 * total_errors == rm_first_datacol, or 2297 * vdev_raidz_combrec() failed 2298 * 2299 * In either case, there is enough bad data to prevent 2300 * reconstruction. 2301 * 2302 * Start checksum ereports for all children which haven't 2303 * failed, and the IO wasn't speculative. 2304 */ 2305 zio->io_error = SET_ERROR(ECKSUM); 2306 2307 if (!(zio->io_flags & ZIO_FLAG_SPECULATIVE)) { 2308 for (c = 0; c < rm->rm_cols; c++) { 2309 rc = &rm->rm_col[c]; 2310 if (rc->rc_error == 0) { 2311 zio_bad_cksum_t zbc; 2312 zbc.zbc_has_cksum = 0; 2313 zbc.zbc_injected = 2314 rm->rm_ecksuminjected; 2315 2316 zfs_ereport_start_checksum( 2317 zio->io_spa, 2318 vd->vdev_child[rc->rc_devidx], 2319 zio, rc->rc_offset, rc->rc_size, 2320 (void *)(uintptr_t)c, &zbc); 2321 } 2322 } 2323 } 2324 } 2325 2326 done: 2327 zio_checksum_verified(zio); 2328 2329 if (zio->io_error == 0 && spa_writeable(zio->io_spa) && 2330 (unexpected_errors || (zio->io_flags & ZIO_FLAG_RESILVER))) { 2331 /* 2332 * Use the good data we have in hand to repair damaged children. 2333 */ 2334 for (c = 0; c < rm->rm_cols; c++) { 2335 rc = &rm->rm_col[c]; 2336 cvd = vd->vdev_child[rc->rc_devidx]; 2337 2338 if (rc->rc_error == 0) 2339 continue; 2340 2341 zio_nowait(zio_vdev_child_io(zio, NULL, cvd, 2342 rc->rc_offset, rc->rc_data, rc->rc_size, 2343 ZIO_TYPE_WRITE, ZIO_PRIORITY_ASYNC_WRITE, 2344 ZIO_FLAG_IO_REPAIR | (unexpected_errors ? 2345 ZIO_FLAG_SELF_HEAL : 0), NULL, NULL)); 2346 } 2347 } 2348 } 2349 2350 static void 2351 vdev_raidz_state_change(vdev_t *vd, int faulted, int degraded) 2352 { 2353 if (faulted > vd->vdev_nparity) 2354 vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN, 2355 VDEV_AUX_NO_REPLICAS); 2356 else if (degraded + faulted != 0) 2357 vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE); 2358 else 2359 vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE); 2360 } 2361 2362 vdev_ops_t vdev_raidz_ops = { 2363 vdev_raidz_open, 2364 vdev_raidz_close, 2365 vdev_raidz_asize, 2366 vdev_raidz_io_start, 2367 vdev_raidz_io_done, 2368 vdev_raidz_state_change, 2369 NULL, 2370 NULL, 2371 VDEV_TYPE_RAIDZ, /* name of this vdev type */ 2372 B_FALSE /* not a leaf vdev */ 2373 }; 2374