1 /*- 2 * Copyright (c) 2010 Alexander Motin <mav@FreeBSD.org> 3 * Copyright (c) 2000 - 2008 Søren Schmidt <sos@FreeBSD.org> 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND 16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 */ 27 28 #include <sys/cdefs.h> 29 __FBSDID("$FreeBSD$"); 30 31 #include <sys/param.h> 32 #include <sys/bio.h> 33 #include <sys/endian.h> 34 #include <sys/kernel.h> 35 #include <sys/kobj.h> 36 #include <sys/limits.h> 37 #include <sys/lock.h> 38 #include <sys/malloc.h> 39 #include <sys/mutex.h> 40 #include <sys/systm.h> 41 #include <sys/taskqueue.h> 42 #include <geom/geom.h> 43 #include "geom/raid/g_raid.h" 44 #include "g_raid_md_if.h" 45 46 static MALLOC_DEFINE(M_MD_INTEL, "md_intel_data", "GEOM_RAID Intel metadata"); 47 48 struct intel_raid_map { 49 uint32_t offset; 50 uint32_t disk_sectors; 51 uint32_t stripe_count; 52 uint16_t strip_sectors; 53 uint8_t status; 54 #define INTEL_S_READY 0x00 55 #define INTEL_S_UNINITIALIZED 0x01 56 #define INTEL_S_DEGRADED 0x02 57 #define INTEL_S_FAILURE 0x03 58 59 uint8_t type; 60 #define INTEL_T_RAID0 0x00 61 #define INTEL_T_RAID1 0x01 62 #define INTEL_T_RAID5 0x05 63 64 uint8_t total_disks; 65 uint8_t total_domains; 66 uint8_t failed_disk_num; 67 uint8_t ddf; 68 uint32_t offset_hi; 69 uint32_t disk_sectors_hi; 70 uint32_t stripe_count_hi; 71 uint32_t filler_2[4]; 72 uint32_t disk_idx[1]; /* total_disks entries. */ 73 #define INTEL_DI_IDX 0x00ffffff 74 #define INTEL_DI_RBLD 0x01000000 75 } __packed; 76 77 struct intel_raid_vol { 78 uint8_t name[16]; 79 u_int64_t total_sectors __packed; 80 uint32_t state; 81 #define INTEL_ST_BOOTABLE 0x00000001 82 #define INTEL_ST_BOOT_DEVICE 0x00000002 83 #define INTEL_ST_READ_COALESCING 0x00000004 84 #define INTEL_ST_WRITE_COALESCING 0x00000008 85 #define INTEL_ST_LAST_SHUTDOWN_DIRTY 0x00000010 86 #define INTEL_ST_HIDDEN_AT_BOOT 0x00000020 87 #define INTEL_ST_CURRENTLY_HIDDEN 0x00000040 88 #define INTEL_ST_VERIFY_AND_FIX 0x00000080 89 #define INTEL_ST_MAP_STATE_UNINIT 0x00000100 90 #define INTEL_ST_NO_AUTO_RECOVERY 0x00000200 91 #define INTEL_ST_CLONE_N_GO 0x00000400 92 #define INTEL_ST_CLONE_MAN_SYNC 0x00000800 93 #define INTEL_ST_CNG_MASTER_DISK_NUM 0x00001000 94 uint32_t reserved; 95 uint8_t migr_priority; 96 uint8_t num_sub_vols; 97 uint8_t tid; 98 uint8_t cng_master_disk; 99 uint16_t cache_policy; 100 uint8_t cng_state; 101 uint8_t cng_sub_state; 102 uint32_t filler_0[10]; 103 104 uint32_t curr_migr_unit; 105 uint32_t checkpoint_id; 106 uint8_t migr_state; 107 uint8_t migr_type; 108 #define INTEL_MT_INIT 0 109 #define INTEL_MT_REBUILD 1 110 #define INTEL_MT_VERIFY 2 111 #define INTEL_MT_GEN_MIGR 3 112 #define INTEL_MT_STATE_CHANGE 4 113 #define INTEL_MT_REPAIR 5 114 uint8_t dirty; 115 uint8_t fs_state; 116 uint16_t verify_errors; 117 uint16_t bad_blocks; 118 uint32_t curr_migr_unit_hi; 119 uint32_t filler_1[3]; 120 struct intel_raid_map map[1]; /* 2 entries if migr_state != 0. */ 121 } __packed; 122 123 struct intel_raid_disk { 124 #define INTEL_SERIAL_LEN 16 125 uint8_t serial[INTEL_SERIAL_LEN]; 126 uint32_t sectors; 127 uint32_t id; 128 uint32_t flags; 129 #define INTEL_F_SPARE 0x01 130 #define INTEL_F_ASSIGNED 0x02 131 #define INTEL_F_FAILED 0x04 132 #define INTEL_F_ONLINE 0x08 133 uint32_t owner_cfg_num; 134 uint32_t sectors_hi; 135 uint32_t filler[3]; 136 } __packed; 137 138 struct intel_raid_conf { 139 uint8_t intel_id[24]; 140 #define INTEL_MAGIC "Intel Raid ISM Cfg Sig. " 141 142 uint8_t version[6]; 143 #define INTEL_VERSION_1000 "1.0.00" /* RAID0 */ 144 #define INTEL_VERSION_1100 "1.1.00" /* RAID1 */ 145 #define INTEL_VERSION_1200 "1.2.00" /* Many volumes */ 146 #define INTEL_VERSION_1201 "1.2.01" /* 3 or 4 disks */ 147 #define INTEL_VERSION_1202 "1.2.02" /* RAID5 */ 148 #define INTEL_VERSION_1204 "1.2.04" /* 5 or 6 disks */ 149 #define INTEL_VERSION_1206 "1.2.06" /* CNG */ 150 #define INTEL_VERSION_1300 "1.3.00" /* Attributes */ 151 152 uint8_t dummy_0[2]; 153 uint32_t checksum; 154 uint32_t config_size; 155 uint32_t config_id; 156 uint32_t generation; 157 uint32_t error_log_size; 158 uint32_t attributes; 159 #define INTEL_ATTR_RAID0 0x00000001 160 #define INTEL_ATTR_RAID1 0x00000002 161 #define INTEL_ATTR_RAID10 0x00000004 162 #define INTEL_ATTR_RAID1E 0x00000008 163 #define INTEL_ATTR_RAID5 0x00000010 164 #define INTEL_ATTR_RAIDCNG 0x00000020 165 #define INTEL_ATTR_2TB 0x20000000 166 #define INTEL_ATTR_PM 0x40000000 167 #define INTEL_ATTR_CHECKSUM 0x80000000 168 169 uint8_t total_disks; 170 uint8_t total_volumes; 171 uint8_t dummy_2[2]; 172 uint32_t filler_0[39]; 173 struct intel_raid_disk disk[1]; /* total_disks entries. */ 174 /* Here goes total_volumes of struct intel_raid_vol. */ 175 } __packed; 176 177 #define INTEL_MAX_MD_SIZE(ndisks) \ 178 (sizeof(struct intel_raid_conf) + \ 179 sizeof(struct intel_raid_disk) * (ndisks - 1) + \ 180 sizeof(struct intel_raid_vol) * 2 + \ 181 sizeof(struct intel_raid_map) * 2 + \ 182 sizeof(uint32_t) * (ndisks - 1) * 4) 183 184 struct g_raid_md_intel_perdisk { 185 struct intel_raid_conf *pd_meta; 186 int pd_disk_pos; 187 struct intel_raid_disk pd_disk_meta; 188 }; 189 190 struct g_raid_md_intel_object { 191 struct g_raid_md_object mdio_base; 192 uint32_t mdio_config_id; 193 uint32_t mdio_generation; 194 struct intel_raid_conf *mdio_meta; 195 struct callout mdio_start_co; /* STARTING state timer. */ 196 int mdio_disks_present; 197 int mdio_started; 198 int mdio_incomplete; 199 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */ 200 }; 201 202 static g_raid_md_create_t g_raid_md_create_intel; 203 static g_raid_md_taste_t g_raid_md_taste_intel; 204 static g_raid_md_event_t g_raid_md_event_intel; 205 static g_raid_md_ctl_t g_raid_md_ctl_intel; 206 static g_raid_md_write_t g_raid_md_write_intel; 207 static g_raid_md_fail_disk_t g_raid_md_fail_disk_intel; 208 static g_raid_md_free_disk_t g_raid_md_free_disk_intel; 209 static g_raid_md_free_t g_raid_md_free_intel; 210 211 static kobj_method_t g_raid_md_intel_methods[] = { 212 KOBJMETHOD(g_raid_md_create, g_raid_md_create_intel), 213 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_intel), 214 KOBJMETHOD(g_raid_md_event, g_raid_md_event_intel), 215 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_intel), 216 KOBJMETHOD(g_raid_md_write, g_raid_md_write_intel), 217 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_intel), 218 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_intel), 219 KOBJMETHOD(g_raid_md_free, g_raid_md_free_intel), 220 { 0, 0 } 221 }; 222 223 static struct g_raid_md_class g_raid_md_intel_class = { 224 "Intel", 225 g_raid_md_intel_methods, 226 sizeof(struct g_raid_md_intel_object), 227 .mdc_priority = 100 228 }; 229 230 231 static struct intel_raid_map * 232 intel_get_map(struct intel_raid_vol *mvol, int i) 233 { 234 struct intel_raid_map *mmap; 235 236 if (i > (mvol->migr_state ? 1 : 0)) 237 return (NULL); 238 mmap = &mvol->map[0]; 239 for (; i > 0; i--) { 240 mmap = (struct intel_raid_map *) 241 &mmap->disk_idx[mmap->total_disks]; 242 } 243 return ((struct intel_raid_map *)mmap); 244 } 245 246 static struct intel_raid_vol * 247 intel_get_volume(struct intel_raid_conf *meta, int i) 248 { 249 struct intel_raid_vol *mvol; 250 struct intel_raid_map *mmap; 251 252 if (i > 1) 253 return (NULL); 254 mvol = (struct intel_raid_vol *)&meta->disk[meta->total_disks]; 255 for (; i > 0; i--) { 256 mmap = intel_get_map(mvol, mvol->migr_state ? 1 : 0); 257 mvol = (struct intel_raid_vol *) 258 &mmap->disk_idx[mmap->total_disks]; 259 } 260 return (mvol); 261 } 262 263 static off_t 264 intel_get_map_offset(struct intel_raid_map *mmap) 265 { 266 off_t offset = (off_t)mmap->offset_hi << 32; 267 268 offset += mmap->offset; 269 return (offset); 270 } 271 272 static void 273 intel_set_map_offset(struct intel_raid_map *mmap, off_t offset) 274 { 275 276 mmap->offset = offset & 0xffffffff; 277 mmap->offset_hi = offset >> 32; 278 } 279 280 static off_t 281 intel_get_map_disk_sectors(struct intel_raid_map *mmap) 282 { 283 off_t disk_sectors = (off_t)mmap->disk_sectors_hi << 32; 284 285 disk_sectors += mmap->disk_sectors; 286 return (disk_sectors); 287 } 288 289 static void 290 intel_set_map_disk_sectors(struct intel_raid_map *mmap, off_t disk_sectors) 291 { 292 293 mmap->disk_sectors = disk_sectors & 0xffffffff; 294 mmap->disk_sectors_hi = disk_sectors >> 32; 295 } 296 297 static void 298 intel_set_map_stripe_count(struct intel_raid_map *mmap, off_t stripe_count) 299 { 300 301 mmap->stripe_count = stripe_count & 0xffffffff; 302 mmap->stripe_count_hi = stripe_count >> 32; 303 } 304 305 static off_t 306 intel_get_disk_sectors(struct intel_raid_disk *disk) 307 { 308 off_t sectors = (off_t)disk->sectors_hi << 32; 309 310 sectors += disk->sectors; 311 return (sectors); 312 } 313 314 static void 315 intel_set_disk_sectors(struct intel_raid_disk *disk, off_t sectors) 316 { 317 318 disk->sectors = sectors & 0xffffffff; 319 disk->sectors_hi = sectors >> 32; 320 } 321 322 static off_t 323 intel_get_vol_curr_migr_unit(struct intel_raid_vol *vol) 324 { 325 off_t curr_migr_unit = (off_t)vol->curr_migr_unit_hi << 32; 326 327 curr_migr_unit += vol->curr_migr_unit; 328 return (curr_migr_unit); 329 } 330 331 static void 332 intel_set_vol_curr_migr_unit(struct intel_raid_vol *vol, off_t curr_migr_unit) 333 { 334 335 vol->curr_migr_unit = curr_migr_unit & 0xffffffff; 336 vol->curr_migr_unit_hi = curr_migr_unit >> 32; 337 } 338 339 static void 340 g_raid_md_intel_print(struct intel_raid_conf *meta) 341 { 342 struct intel_raid_vol *mvol; 343 struct intel_raid_map *mmap; 344 int i, j, k; 345 346 if (g_raid_debug < 1) 347 return; 348 349 printf("********* ATA Intel MatrixRAID Metadata *********\n"); 350 printf("intel_id <%.24s>\n", meta->intel_id); 351 printf("version <%.6s>\n", meta->version); 352 printf("checksum 0x%08x\n", meta->checksum); 353 printf("config_size 0x%08x\n", meta->config_size); 354 printf("config_id 0x%08x\n", meta->config_id); 355 printf("generation 0x%08x\n", meta->generation); 356 printf("attributes 0x%08x\n", meta->attributes); 357 printf("total_disks %u\n", meta->total_disks); 358 printf("total_volumes %u\n", meta->total_volumes); 359 printf("DISK# serial disk_sectors disk_sectors_hi disk_id flags\n"); 360 for (i = 0; i < meta->total_disks; i++ ) { 361 printf(" %d <%.16s> %u %u 0x%08x 0x%08x\n", i, 362 meta->disk[i].serial, meta->disk[i].sectors, 363 meta->disk[i].sectors_hi, 364 meta->disk[i].id, meta->disk[i].flags); 365 } 366 for (i = 0; i < meta->total_volumes; i++) { 367 mvol = intel_get_volume(meta, i); 368 printf(" ****** Volume %d ******\n", i); 369 printf(" name %.16s\n", mvol->name); 370 printf(" total_sectors %ju\n", mvol->total_sectors); 371 printf(" state %u\n", mvol->state); 372 printf(" reserved %u\n", mvol->reserved); 373 printf(" curr_migr_unit %u\n", mvol->curr_migr_unit); 374 printf(" curr_migr_unit_hi %u\n", mvol->curr_migr_unit_hi); 375 printf(" checkpoint_id %u\n", mvol->checkpoint_id); 376 printf(" migr_state %u\n", mvol->migr_state); 377 printf(" migr_type %u\n", mvol->migr_type); 378 printf(" dirty %u\n", mvol->dirty); 379 380 for (j = 0; j < (mvol->migr_state ? 2 : 1); j++) { 381 printf(" *** Map %d ***\n", j); 382 mmap = intel_get_map(mvol, j); 383 printf(" offset %u\n", mmap->offset); 384 printf(" offset_hi %u\n", mmap->offset_hi); 385 printf(" disk_sectors %u\n", mmap->disk_sectors); 386 printf(" disk_sectors_hi %u\n", mmap->disk_sectors_hi); 387 printf(" stripe_count %u\n", mmap->stripe_count); 388 printf(" stripe_count_hi %u\n", mmap->stripe_count_hi); 389 printf(" strip_sectors %u\n", mmap->strip_sectors); 390 printf(" status %u\n", mmap->status); 391 printf(" type %u\n", mmap->type); 392 printf(" total_disks %u\n", mmap->total_disks); 393 printf(" total_domains %u\n", mmap->total_domains); 394 printf(" failed_disk_num %u\n", mmap->failed_disk_num); 395 printf(" ddf %u\n", mmap->ddf); 396 printf(" disk_idx "); 397 for (k = 0; k < mmap->total_disks; k++) 398 printf(" 0x%08x", mmap->disk_idx[k]); 399 printf("\n"); 400 } 401 } 402 printf("=================================================\n"); 403 } 404 405 static struct intel_raid_conf * 406 intel_meta_copy(struct intel_raid_conf *meta) 407 { 408 struct intel_raid_conf *nmeta; 409 410 nmeta = malloc(meta->config_size, M_MD_INTEL, M_WAITOK); 411 memcpy(nmeta, meta, meta->config_size); 412 return (nmeta); 413 } 414 415 static int 416 intel_meta_find_disk(struct intel_raid_conf *meta, char *serial) 417 { 418 int pos; 419 420 for (pos = 0; pos < meta->total_disks; pos++) { 421 if (strncmp(meta->disk[pos].serial, 422 serial, INTEL_SERIAL_LEN) == 0) 423 return (pos); 424 } 425 return (-1); 426 } 427 428 static struct intel_raid_conf * 429 intel_meta_read(struct g_consumer *cp) 430 { 431 struct g_provider *pp; 432 struct intel_raid_conf *meta; 433 struct intel_raid_vol *mvol; 434 struct intel_raid_map *mmap; 435 char *buf; 436 int error, i, j, k, left, size; 437 uint32_t checksum, *ptr; 438 439 pp = cp->provider; 440 441 /* Read the anchor sector. */ 442 buf = g_read_data(cp, 443 pp->mediasize - pp->sectorsize * 2, pp->sectorsize, &error); 444 if (buf == NULL) { 445 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).", 446 pp->name, error); 447 return (NULL); 448 } 449 meta = (struct intel_raid_conf *)buf; 450 451 /* Check if this is an Intel RAID struct */ 452 if (strncmp(meta->intel_id, INTEL_MAGIC, strlen(INTEL_MAGIC))) { 453 G_RAID_DEBUG(1, "Intel signature check failed on %s", pp->name); 454 g_free(buf); 455 return (NULL); 456 } 457 if (meta->config_size > 65536 || 458 meta->config_size < sizeof(struct intel_raid_conf)) { 459 G_RAID_DEBUG(1, "Intel metadata size looks wrong: %d", 460 meta->config_size); 461 g_free(buf); 462 return (NULL); 463 } 464 size = meta->config_size; 465 meta = malloc(size, M_MD_INTEL, M_WAITOK); 466 memcpy(meta, buf, min(size, pp->sectorsize)); 467 g_free(buf); 468 469 /* Read all the rest, if needed. */ 470 if (meta->config_size > pp->sectorsize) { 471 left = (meta->config_size - 1) / pp->sectorsize; 472 buf = g_read_data(cp, 473 pp->mediasize - pp->sectorsize * (2 + left), 474 pp->sectorsize * left, &error); 475 if (buf == NULL) { 476 G_RAID_DEBUG(1, "Cannot read remaining metadata" 477 " part from %s (error=%d).", 478 pp->name, error); 479 free(meta, M_MD_INTEL); 480 return (NULL); 481 } 482 memcpy(((char *)meta) + pp->sectorsize, buf, 483 pp->sectorsize * left); 484 g_free(buf); 485 } 486 487 /* Check metadata checksum. */ 488 for (checksum = 0, ptr = (uint32_t *)meta, i = 0; 489 i < (meta->config_size / sizeof(uint32_t)); i++) { 490 checksum += *ptr++; 491 } 492 checksum -= meta->checksum; 493 if (checksum != meta->checksum) { 494 G_RAID_DEBUG(1, "Intel checksum check failed on %s", pp->name); 495 free(meta, M_MD_INTEL); 496 return (NULL); 497 } 498 499 /* Validate metadata size. */ 500 size = sizeof(struct intel_raid_conf) + 501 sizeof(struct intel_raid_disk) * (meta->total_disks - 1) + 502 sizeof(struct intel_raid_vol) * meta->total_volumes; 503 if (size > meta->config_size) { 504 badsize: 505 G_RAID_DEBUG(1, "Intel metadata size incorrect %d < %d", 506 meta->config_size, size); 507 free(meta, M_MD_INTEL); 508 return (NULL); 509 } 510 for (i = 0; i < meta->total_volumes; i++) { 511 mvol = intel_get_volume(meta, i); 512 mmap = intel_get_map(mvol, 0); 513 size += 4 * (mmap->total_disks - 1); 514 if (size > meta->config_size) 515 goto badsize; 516 if (mvol->migr_state) { 517 size += sizeof(struct intel_raid_map); 518 if (size > meta->config_size) 519 goto badsize; 520 mmap = intel_get_map(mvol, 1); 521 size += 4 * (mmap->total_disks - 1); 522 if (size > meta->config_size) 523 goto badsize; 524 } 525 } 526 527 /* Validate disk indexes. */ 528 for (i = 0; i < meta->total_volumes; i++) { 529 mvol = intel_get_volume(meta, i); 530 for (j = 0; j < (mvol->migr_state ? 2 : 1); j++) { 531 mmap = intel_get_map(mvol, j); 532 for (k = 0; k < mmap->total_disks; k++) { 533 if ((mmap->disk_idx[k] & INTEL_DI_IDX) > 534 meta->total_disks) { 535 G_RAID_DEBUG(1, "Intel metadata disk" 536 " index %d too big (>%d)", 537 mmap->disk_idx[k] & INTEL_DI_IDX, 538 meta->total_disks); 539 free(meta, M_MD_INTEL); 540 return (NULL); 541 } 542 } 543 } 544 } 545 546 /* Validate migration types. */ 547 for (i = 0; i < meta->total_volumes; i++) { 548 mvol = intel_get_volume(meta, i); 549 if (mvol->migr_state && 550 mvol->migr_type != INTEL_MT_INIT && 551 mvol->migr_type != INTEL_MT_REBUILD && 552 mvol->migr_type != INTEL_MT_VERIFY && 553 mvol->migr_type != INTEL_MT_REPAIR) { 554 G_RAID_DEBUG(1, "Intel metadata has unsupported" 555 " migration type %d", mvol->migr_type); 556 free(meta, M_MD_INTEL); 557 return (NULL); 558 } 559 } 560 561 return (meta); 562 } 563 564 static int 565 intel_meta_write(struct g_consumer *cp, struct intel_raid_conf *meta) 566 { 567 struct g_provider *pp; 568 char *buf; 569 int error, i, sectors; 570 uint32_t checksum, *ptr; 571 572 pp = cp->provider; 573 574 /* Recalculate checksum for case if metadata were changed. */ 575 meta->checksum = 0; 576 for (checksum = 0, ptr = (uint32_t *)meta, i = 0; 577 i < (meta->config_size / sizeof(uint32_t)); i++) { 578 checksum += *ptr++; 579 } 580 meta->checksum = checksum; 581 582 /* Create and fill buffer. */ 583 sectors = (meta->config_size + pp->sectorsize - 1) / pp->sectorsize; 584 buf = malloc(sectors * pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO); 585 if (sectors > 1) { 586 memcpy(buf, ((char *)meta) + pp->sectorsize, 587 (sectors - 1) * pp->sectorsize); 588 } 589 memcpy(buf + (sectors - 1) * pp->sectorsize, meta, pp->sectorsize); 590 591 error = g_write_data(cp, 592 pp->mediasize - pp->sectorsize * (1 + sectors), 593 buf, pp->sectorsize * sectors); 594 if (error != 0) { 595 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).", 596 pp->name, error); 597 } 598 599 free(buf, M_MD_INTEL); 600 return (error); 601 } 602 603 static int 604 intel_meta_erase(struct g_consumer *cp) 605 { 606 struct g_provider *pp; 607 char *buf; 608 int error; 609 610 pp = cp->provider; 611 buf = malloc(pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO); 612 error = g_write_data(cp, 613 pp->mediasize - 2 * pp->sectorsize, 614 buf, pp->sectorsize); 615 if (error != 0) { 616 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).", 617 pp->name, error); 618 } 619 free(buf, M_MD_INTEL); 620 return (error); 621 } 622 623 static int 624 intel_meta_write_spare(struct g_consumer *cp, struct intel_raid_disk *d) 625 { 626 struct intel_raid_conf *meta; 627 int error; 628 629 /* Fill anchor and single disk. */ 630 meta = malloc(INTEL_MAX_MD_SIZE(1), M_MD_INTEL, M_WAITOK | M_ZERO); 631 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1); 632 memcpy(&meta->version[0], INTEL_VERSION_1000, 633 sizeof(INTEL_VERSION_1000) - 1); 634 meta->config_size = INTEL_MAX_MD_SIZE(1); 635 meta->config_id = arc4random(); 636 meta->generation = 1; 637 meta->total_disks = 1; 638 meta->disk[0] = *d; 639 error = intel_meta_write(cp, meta); 640 free(meta, M_MD_INTEL); 641 return (error); 642 } 643 644 static struct g_raid_disk * 645 g_raid_md_intel_get_disk(struct g_raid_softc *sc, int id) 646 { 647 struct g_raid_disk *disk; 648 struct g_raid_md_intel_perdisk *pd; 649 650 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 651 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 652 if (pd->pd_disk_pos == id) 653 break; 654 } 655 return (disk); 656 } 657 658 static int 659 g_raid_md_intel_supported(int level, int qual, int disks, int force) 660 { 661 662 switch (level) { 663 case G_RAID_VOLUME_RL_RAID0: 664 if (disks < 1) 665 return (0); 666 if (!force && (disks < 2 || disks > 6)) 667 return (0); 668 break; 669 case G_RAID_VOLUME_RL_RAID1: 670 if (disks < 1) 671 return (0); 672 if (!force && (disks != 2)) 673 return (0); 674 break; 675 case G_RAID_VOLUME_RL_RAID1E: 676 if (disks < 2) 677 return (0); 678 if (!force && (disks != 4)) 679 return (0); 680 break; 681 case G_RAID_VOLUME_RL_RAID5: 682 if (disks < 3) 683 return (0); 684 if (!force && disks > 6) 685 return (0); 686 if (qual != G_RAID_VOLUME_RLQ_R5LA) 687 return (0); 688 break; 689 default: 690 return (0); 691 } 692 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE) 693 return (0); 694 return (1); 695 } 696 697 static struct g_raid_volume * 698 g_raid_md_intel_get_volume(struct g_raid_softc *sc, int id) 699 { 700 struct g_raid_volume *mvol; 701 702 TAILQ_FOREACH(mvol, &sc->sc_volumes, v_next) { 703 if ((intptr_t)(mvol->v_md_data) == id) 704 break; 705 } 706 return (mvol); 707 } 708 709 static int 710 g_raid_md_intel_start_disk(struct g_raid_disk *disk) 711 { 712 struct g_raid_softc *sc; 713 struct g_raid_subdisk *sd, *tmpsd; 714 struct g_raid_disk *olddisk, *tmpdisk; 715 struct g_raid_md_object *md; 716 struct g_raid_md_intel_object *mdi; 717 struct g_raid_md_intel_perdisk *pd, *oldpd; 718 struct intel_raid_conf *meta; 719 struct intel_raid_vol *mvol; 720 struct intel_raid_map *mmap0, *mmap1; 721 int disk_pos, resurrection = 0; 722 723 sc = disk->d_softc; 724 md = sc->sc_md; 725 mdi = (struct g_raid_md_intel_object *)md; 726 meta = mdi->mdio_meta; 727 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 728 olddisk = NULL; 729 730 /* Find disk position in metadata by it's serial. */ 731 disk_pos = intel_meta_find_disk(meta, pd->pd_disk_meta.serial); 732 if (disk_pos < 0) { 733 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk"); 734 /* Failed stale disk is useless for us. */ 735 if (pd->pd_disk_meta.flags & INTEL_F_FAILED) { 736 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE_FAILED); 737 return (0); 738 } 739 /* If we are in the start process, that's all for now. */ 740 if (!mdi->mdio_started) 741 goto nofit; 742 /* 743 * If we have already started - try to get use of the disk. 744 * Try to replace OFFLINE disks first, then FAILED. 745 */ 746 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) { 747 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE && 748 tmpdisk->d_state != G_RAID_DISK_S_FAILED) 749 continue; 750 /* Make sure this disk is big enough. */ 751 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) { 752 off_t disk_sectors = 753 intel_get_disk_sectors(&pd->pd_disk_meta); 754 755 if (sd->sd_offset + sd->sd_size + 4096 > 756 disk_sectors * 512) { 757 G_RAID_DEBUG1(1, sc, 758 "Disk too small (%llu < %llu)", 759 (unsigned long long) 760 disk_sectors * 512, 761 (unsigned long long) 762 sd->sd_offset + sd->sd_size + 4096); 763 break; 764 } 765 } 766 if (sd != NULL) 767 continue; 768 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) { 769 olddisk = tmpdisk; 770 break; 771 } else if (olddisk == NULL) 772 olddisk = tmpdisk; 773 } 774 if (olddisk == NULL) { 775 nofit: 776 if (pd->pd_disk_meta.flags & INTEL_F_SPARE) { 777 g_raid_change_disk_state(disk, 778 G_RAID_DISK_S_SPARE); 779 return (1); 780 } else { 781 g_raid_change_disk_state(disk, 782 G_RAID_DISK_S_STALE); 783 return (0); 784 } 785 } 786 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data; 787 disk_pos = oldpd->pd_disk_pos; 788 resurrection = 1; 789 } 790 791 if (olddisk == NULL) { 792 /* Find placeholder by position. */ 793 olddisk = g_raid_md_intel_get_disk(sc, disk_pos); 794 if (olddisk == NULL) 795 panic("No disk at position %d!", disk_pos); 796 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) { 797 G_RAID_DEBUG1(1, sc, "More then one disk for pos %d", 798 disk_pos); 799 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE); 800 return (0); 801 } 802 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data; 803 } 804 805 /* Replace failed disk or placeholder with new disk. */ 806 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) { 807 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next); 808 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 809 sd->sd_disk = disk; 810 } 811 oldpd->pd_disk_pos = -2; 812 pd->pd_disk_pos = disk_pos; 813 814 /* If it was placeholder -- destroy it. */ 815 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) { 816 g_raid_destroy_disk(olddisk); 817 } else { 818 /* Otherwise, make it STALE_FAILED. */ 819 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED); 820 /* Update global metadata just in case. */ 821 memcpy(&meta->disk[disk_pos], &pd->pd_disk_meta, 822 sizeof(struct intel_raid_disk)); 823 } 824 825 /* Welcome the new disk. */ 826 if (resurrection) 827 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); 828 else if (meta->disk[disk_pos].flags & INTEL_F_FAILED) 829 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED); 830 else if (meta->disk[disk_pos].flags & INTEL_F_SPARE) 831 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE); 832 else 833 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); 834 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 835 mvol = intel_get_volume(meta, 836 (uintptr_t)(sd->sd_volume->v_md_data)); 837 mmap0 = intel_get_map(mvol, 0); 838 if (mvol->migr_state) 839 mmap1 = intel_get_map(mvol, 1); 840 else 841 mmap1 = mmap0; 842 843 if (resurrection) { 844 /* Stale disk, almost same as new. */ 845 g_raid_change_subdisk_state(sd, 846 G_RAID_SUBDISK_S_NEW); 847 } else if (meta->disk[disk_pos].flags & INTEL_F_FAILED) { 848 /* Failed disk, almost useless. */ 849 g_raid_change_subdisk_state(sd, 850 G_RAID_SUBDISK_S_FAILED); 851 } else if (mvol->migr_state == 0) { 852 if (mmap0->status == INTEL_S_UNINITIALIZED) { 853 /* Freshly created uninitialized volume. */ 854 g_raid_change_subdisk_state(sd, 855 G_RAID_SUBDISK_S_UNINITIALIZED); 856 } else if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 857 /* Freshly inserted disk. */ 858 g_raid_change_subdisk_state(sd, 859 G_RAID_SUBDISK_S_NEW); 860 } else if (mvol->dirty) { 861 /* Dirty volume (unclean shutdown). */ 862 g_raid_change_subdisk_state(sd, 863 G_RAID_SUBDISK_S_STALE); 864 } else { 865 /* Up to date disk. */ 866 g_raid_change_subdisk_state(sd, 867 G_RAID_SUBDISK_S_ACTIVE); 868 } 869 } else if (mvol->migr_type == INTEL_MT_INIT || 870 mvol->migr_type == INTEL_MT_REBUILD) { 871 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 872 /* Freshly inserted disk. */ 873 g_raid_change_subdisk_state(sd, 874 G_RAID_SUBDISK_S_NEW); 875 } else if (mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 876 /* Rebuilding disk. */ 877 g_raid_change_subdisk_state(sd, 878 G_RAID_SUBDISK_S_REBUILD); 879 if (mvol->dirty) { 880 sd->sd_rebuild_pos = 0; 881 } else { 882 sd->sd_rebuild_pos = 883 intel_get_vol_curr_migr_unit(mvol) * 884 sd->sd_volume->v_strip_size * 885 mmap0->total_domains; 886 } 887 } else if (mvol->dirty) { 888 /* Dirty volume (unclean shutdown). */ 889 g_raid_change_subdisk_state(sd, 890 G_RAID_SUBDISK_S_STALE); 891 } else { 892 /* Up to date disk. */ 893 g_raid_change_subdisk_state(sd, 894 G_RAID_SUBDISK_S_ACTIVE); 895 } 896 } else if (mvol->migr_type == INTEL_MT_VERIFY || 897 mvol->migr_type == INTEL_MT_REPAIR) { 898 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 899 /* Freshly inserted disk. */ 900 g_raid_change_subdisk_state(sd, 901 G_RAID_SUBDISK_S_NEW); 902 } else if (mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 903 /* Resyncing disk. */ 904 g_raid_change_subdisk_state(sd, 905 G_RAID_SUBDISK_S_RESYNC); 906 if (mvol->dirty) { 907 sd->sd_rebuild_pos = 0; 908 } else { 909 sd->sd_rebuild_pos = 910 intel_get_vol_curr_migr_unit(mvol) * 911 sd->sd_volume->v_strip_size * 912 mmap0->total_domains; 913 } 914 } else if (mvol->dirty) { 915 /* Dirty volume (unclean shutdown). */ 916 g_raid_change_subdisk_state(sd, 917 G_RAID_SUBDISK_S_STALE); 918 } else { 919 /* Up to date disk. */ 920 g_raid_change_subdisk_state(sd, 921 G_RAID_SUBDISK_S_ACTIVE); 922 } 923 } 924 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 925 G_RAID_EVENT_SUBDISK); 926 } 927 928 /* Update status of our need for spare. */ 929 if (mdi->mdio_started) { 930 mdi->mdio_incomplete = 931 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < 932 meta->total_disks); 933 } 934 935 return (resurrection); 936 } 937 938 static void 939 g_disk_md_intel_retaste(void *arg, int pending) 940 { 941 942 G_RAID_DEBUG(1, "Array is not complete, trying to retaste."); 943 g_retaste(&g_raid_class); 944 free(arg, M_MD_INTEL); 945 } 946 947 static void 948 g_raid_md_intel_refill(struct g_raid_softc *sc) 949 { 950 struct g_raid_md_object *md; 951 struct g_raid_md_intel_object *mdi; 952 struct intel_raid_conf *meta; 953 struct g_raid_disk *disk; 954 struct task *task; 955 int update, na; 956 957 md = sc->sc_md; 958 mdi = (struct g_raid_md_intel_object *)md; 959 meta = mdi->mdio_meta; 960 update = 0; 961 do { 962 /* Make sure we miss anything. */ 963 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE); 964 if (na == meta->total_disks) 965 break; 966 967 G_RAID_DEBUG1(1, md->mdo_softc, 968 "Array is not complete (%d of %d), " 969 "trying to refill.", na, meta->total_disks); 970 971 /* Try to get use some of STALE disks. */ 972 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 973 if (disk->d_state == G_RAID_DISK_S_STALE) { 974 update += g_raid_md_intel_start_disk(disk); 975 if (disk->d_state == G_RAID_DISK_S_ACTIVE) 976 break; 977 } 978 } 979 if (disk != NULL) 980 continue; 981 982 /* Try to get use some of SPARE disks. */ 983 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 984 if (disk->d_state == G_RAID_DISK_S_SPARE) { 985 update += g_raid_md_intel_start_disk(disk); 986 if (disk->d_state == G_RAID_DISK_S_ACTIVE) 987 break; 988 } 989 } 990 } while (disk != NULL); 991 992 /* Write new metadata if we changed something. */ 993 if (update) { 994 g_raid_md_write_intel(md, NULL, NULL, NULL); 995 meta = mdi->mdio_meta; 996 } 997 998 /* Update status of our need for spare. */ 999 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < 1000 meta->total_disks); 1001 1002 /* Request retaste hoping to find spare. */ 1003 if (mdi->mdio_incomplete) { 1004 task = malloc(sizeof(struct task), 1005 M_MD_INTEL, M_WAITOK | M_ZERO); 1006 TASK_INIT(task, 0, g_disk_md_intel_retaste, task); 1007 taskqueue_enqueue(taskqueue_swi, task); 1008 } 1009 } 1010 1011 static void 1012 g_raid_md_intel_start(struct g_raid_softc *sc) 1013 { 1014 struct g_raid_md_object *md; 1015 struct g_raid_md_intel_object *mdi; 1016 struct g_raid_md_intel_perdisk *pd; 1017 struct intel_raid_conf *meta; 1018 struct intel_raid_vol *mvol; 1019 struct intel_raid_map *mmap; 1020 struct g_raid_volume *vol; 1021 struct g_raid_subdisk *sd; 1022 struct g_raid_disk *disk; 1023 int i, j, disk_pos; 1024 1025 md = sc->sc_md; 1026 mdi = (struct g_raid_md_intel_object *)md; 1027 meta = mdi->mdio_meta; 1028 1029 /* Create volumes and subdisks. */ 1030 for (i = 0; i < meta->total_volumes; i++) { 1031 mvol = intel_get_volume(meta, i); 1032 mmap = intel_get_map(mvol, 0); 1033 vol = g_raid_create_volume(sc, mvol->name, -1); 1034 vol->v_md_data = (void *)(intptr_t)i; 1035 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE; 1036 if (mmap->type == INTEL_T_RAID0) 1037 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0; 1038 else if (mmap->type == INTEL_T_RAID1 && 1039 mmap->total_domains >= 2 && 1040 mmap->total_domains <= mmap->total_disks) { 1041 /* Assume total_domains is correct. */ 1042 if (mmap->total_domains == mmap->total_disks) 1043 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1; 1044 else 1045 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E; 1046 } else if (mmap->type == INTEL_T_RAID1) { 1047 /* total_domains looks wrong. */ 1048 if (mmap->total_disks <= 2) 1049 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1; 1050 else 1051 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E; 1052 } else if (mmap->type == INTEL_T_RAID5) { 1053 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5; 1054 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LA; 1055 } else 1056 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN; 1057 vol->v_strip_size = (u_int)mmap->strip_sectors * 512; //ZZZ 1058 vol->v_disks_count = mmap->total_disks; 1059 vol->v_mediasize = (off_t)mvol->total_sectors * 512; //ZZZ 1060 vol->v_sectorsize = 512; //ZZZ 1061 for (j = 0; j < vol->v_disks_count; j++) { 1062 sd = &vol->v_subdisks[j]; 1063 sd->sd_offset = intel_get_map_offset(mmap) * 512; //ZZZ 1064 sd->sd_size = intel_get_map_disk_sectors(mmap) * 512; //ZZZ 1065 } 1066 g_raid_start_volume(vol); 1067 } 1068 1069 /* Create disk placeholders to store data for later writing. */ 1070 for (disk_pos = 0; disk_pos < meta->total_disks; disk_pos++) { 1071 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO); 1072 pd->pd_disk_pos = disk_pos; 1073 pd->pd_disk_meta = meta->disk[disk_pos]; 1074 disk = g_raid_create_disk(sc); 1075 disk->d_md_data = (void *)pd; 1076 disk->d_state = G_RAID_DISK_S_OFFLINE; 1077 for (i = 0; i < meta->total_volumes; i++) { 1078 mvol = intel_get_volume(meta, i); 1079 mmap = intel_get_map(mvol, 0); 1080 for (j = 0; j < mmap->total_disks; j++) { 1081 if ((mmap->disk_idx[j] & INTEL_DI_IDX) == disk_pos) 1082 break; 1083 } 1084 if (j == mmap->total_disks) 1085 continue; 1086 vol = g_raid_md_intel_get_volume(sc, i); 1087 sd = &vol->v_subdisks[j]; 1088 sd->sd_disk = disk; 1089 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 1090 } 1091 } 1092 1093 /* Make all disks found till the moment take their places. */ 1094 do { 1095 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1096 if (disk->d_state == G_RAID_DISK_S_NONE) { 1097 g_raid_md_intel_start_disk(disk); 1098 break; 1099 } 1100 } 1101 } while (disk != NULL); 1102 1103 mdi->mdio_started = 1; 1104 G_RAID_DEBUG1(0, sc, "Array started."); 1105 g_raid_md_write_intel(md, NULL, NULL, NULL); 1106 1107 /* Pickup any STALE/SPARE disks to refill array if needed. */ 1108 g_raid_md_intel_refill(sc); 1109 1110 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 1111 g_raid_event_send(vol, G_RAID_VOLUME_E_START, 1112 G_RAID_EVENT_VOLUME); 1113 } 1114 1115 callout_stop(&mdi->mdio_start_co); 1116 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount); 1117 root_mount_rel(mdi->mdio_rootmount); 1118 mdi->mdio_rootmount = NULL; 1119 } 1120 1121 static void 1122 g_raid_md_intel_new_disk(struct g_raid_disk *disk) 1123 { 1124 struct g_raid_softc *sc; 1125 struct g_raid_md_object *md; 1126 struct g_raid_md_intel_object *mdi; 1127 struct intel_raid_conf *pdmeta; 1128 struct g_raid_md_intel_perdisk *pd; 1129 1130 sc = disk->d_softc; 1131 md = sc->sc_md; 1132 mdi = (struct g_raid_md_intel_object *)md; 1133 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 1134 pdmeta = pd->pd_meta; 1135 1136 if (mdi->mdio_started) { 1137 if (g_raid_md_intel_start_disk(disk)) 1138 g_raid_md_write_intel(md, NULL, NULL, NULL); 1139 } else { 1140 /* If we haven't started yet - check metadata freshness. */ 1141 if (mdi->mdio_meta == NULL || 1142 ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) { 1143 G_RAID_DEBUG1(1, sc, "Newer disk"); 1144 if (mdi->mdio_meta != NULL) 1145 free(mdi->mdio_meta, M_MD_INTEL); 1146 mdi->mdio_meta = intel_meta_copy(pdmeta); 1147 mdi->mdio_generation = mdi->mdio_meta->generation; 1148 mdi->mdio_disks_present = 1; 1149 } else if (pdmeta->generation == mdi->mdio_generation) { 1150 mdi->mdio_disks_present++; 1151 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)", 1152 mdi->mdio_disks_present, 1153 mdi->mdio_meta->total_disks); 1154 } else { 1155 G_RAID_DEBUG1(1, sc, "Older disk"); 1156 } 1157 /* If we collected all needed disks - start array. */ 1158 if (mdi->mdio_disks_present == mdi->mdio_meta->total_disks) 1159 g_raid_md_intel_start(sc); 1160 } 1161 } 1162 1163 static void 1164 g_raid_intel_go(void *arg) 1165 { 1166 struct g_raid_softc *sc; 1167 struct g_raid_md_object *md; 1168 struct g_raid_md_intel_object *mdi; 1169 1170 sc = arg; 1171 md = sc->sc_md; 1172 mdi = (struct g_raid_md_intel_object *)md; 1173 if (!mdi->mdio_started) { 1174 G_RAID_DEBUG1(0, sc, "Force array start due to timeout."); 1175 g_raid_event_send(sc, G_RAID_NODE_E_START, 0); 1176 } 1177 } 1178 1179 static int 1180 g_raid_md_create_intel(struct g_raid_md_object *md, struct g_class *mp, 1181 struct g_geom **gp) 1182 { 1183 struct g_raid_softc *sc; 1184 struct g_raid_md_intel_object *mdi; 1185 char name[16]; 1186 1187 mdi = (struct g_raid_md_intel_object *)md; 1188 mdi->mdio_config_id = arc4random(); 1189 mdi->mdio_generation = 0; 1190 snprintf(name, sizeof(name), "Intel-%08x", mdi->mdio_config_id); 1191 sc = g_raid_create_node(mp, name, md); 1192 if (sc == NULL) 1193 return (G_RAID_MD_TASTE_FAIL); 1194 md->mdo_softc = sc; 1195 *gp = sc->sc_geom; 1196 return (G_RAID_MD_TASTE_NEW); 1197 } 1198 1199 /* 1200 * Return the last N characters of the serial label. The Linux and 1201 * ataraid(7) code always uses the last 16 characters of the label to 1202 * store into the Intel meta format. Generalize this to N characters 1203 * since that's easy. Labels can be up to 20 characters for SATA drives 1204 * and up 251 characters for SAS drives. Since intel controllers don't 1205 * support SAS drives, just stick with the SATA limits for stack friendliness. 1206 */ 1207 static int 1208 g_raid_md_get_label(struct g_consumer *cp, char *serial, int serlen) 1209 { 1210 char serial_buffer[24]; 1211 int len, error; 1212 1213 len = sizeof(serial_buffer); 1214 error = g_io_getattr("GEOM::ident", cp, &len, serial_buffer); 1215 if (error != 0) 1216 return (error); 1217 len = strlen(serial_buffer); 1218 if (len > serlen) 1219 len -= serlen; 1220 else 1221 len = 0; 1222 strncpy(serial, serial_buffer + len, serlen); 1223 return (0); 1224 } 1225 1226 static int 1227 g_raid_md_taste_intel(struct g_raid_md_object *md, struct g_class *mp, 1228 struct g_consumer *cp, struct g_geom **gp) 1229 { 1230 struct g_consumer *rcp; 1231 struct g_provider *pp; 1232 struct g_raid_md_intel_object *mdi, *mdi1; 1233 struct g_raid_softc *sc; 1234 struct g_raid_disk *disk; 1235 struct intel_raid_conf *meta; 1236 struct g_raid_md_intel_perdisk *pd; 1237 struct g_geom *geom; 1238 int error, disk_pos, result, spare, len; 1239 char serial[INTEL_SERIAL_LEN]; 1240 char name[16]; 1241 uint16_t vendor; 1242 1243 G_RAID_DEBUG(1, "Tasting Intel on %s", cp->provider->name); 1244 mdi = (struct g_raid_md_intel_object *)md; 1245 pp = cp->provider; 1246 1247 /* Read metadata from device. */ 1248 meta = NULL; 1249 vendor = 0xffff; 1250 disk_pos = 0; 1251 if (g_access(cp, 1, 0, 0) != 0) 1252 return (G_RAID_MD_TASTE_FAIL); 1253 g_topology_unlock(); 1254 error = g_raid_md_get_label(cp, serial, sizeof(serial)); 1255 if (error != 0) { 1256 G_RAID_DEBUG(1, "Cannot get serial number from %s (error=%d).", 1257 pp->name, error); 1258 goto fail2; 1259 } 1260 len = 2; 1261 if (pp->geom->rank == 1) 1262 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor); 1263 meta = intel_meta_read(cp); 1264 g_topology_lock(); 1265 g_access(cp, -1, 0, 0); 1266 if (meta == NULL) { 1267 if (g_raid_aggressive_spare) { 1268 if (vendor != 0x8086) { 1269 G_RAID_DEBUG(1, 1270 "Intel vendor mismatch 0x%04x != 0x8086", 1271 vendor); 1272 } else { 1273 G_RAID_DEBUG(1, 1274 "No Intel metadata, forcing spare."); 1275 spare = 2; 1276 goto search; 1277 } 1278 } 1279 return (G_RAID_MD_TASTE_FAIL); 1280 } 1281 1282 /* Check this disk position in obtained metadata. */ 1283 disk_pos = intel_meta_find_disk(meta, serial); 1284 if (disk_pos < 0) { 1285 G_RAID_DEBUG(1, "Intel serial '%s' not found", serial); 1286 goto fail1; 1287 } 1288 if (intel_get_disk_sectors(&meta->disk[disk_pos]) != 1289 (pp->mediasize / pp->sectorsize)) { 1290 G_RAID_DEBUG(1, "Intel size mismatch %ju != %ju", 1291 intel_get_disk_sectors(&meta->disk[disk_pos]), 1292 (off_t)(pp->mediasize / pp->sectorsize)); 1293 goto fail1; 1294 } 1295 1296 /* Metadata valid. Print it. */ 1297 g_raid_md_intel_print(meta); 1298 G_RAID_DEBUG(1, "Intel disk position %d", disk_pos); 1299 spare = meta->disk[disk_pos].flags & INTEL_F_SPARE; 1300 1301 search: 1302 /* Search for matching node. */ 1303 sc = NULL; 1304 mdi1 = NULL; 1305 LIST_FOREACH(geom, &mp->geom, geom) { 1306 sc = geom->softc; 1307 if (sc == NULL) 1308 continue; 1309 if (sc->sc_stopping != 0) 1310 continue; 1311 if (sc->sc_md->mdo_class != md->mdo_class) 1312 continue; 1313 mdi1 = (struct g_raid_md_intel_object *)sc->sc_md; 1314 if (spare) { 1315 if (mdi1->mdio_incomplete) 1316 break; 1317 } else { 1318 if (mdi1->mdio_config_id == meta->config_id) 1319 break; 1320 } 1321 } 1322 1323 /* Found matching node. */ 1324 if (geom != NULL) { 1325 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name); 1326 result = G_RAID_MD_TASTE_EXISTING; 1327 1328 } else if (spare) { /* Not found needy node -- left for later. */ 1329 G_RAID_DEBUG(1, "Spare is not needed at this time"); 1330 goto fail1; 1331 1332 } else { /* Not found matching node -- create one. */ 1333 result = G_RAID_MD_TASTE_NEW; 1334 mdi->mdio_config_id = meta->config_id; 1335 snprintf(name, sizeof(name), "Intel-%08x", meta->config_id); 1336 sc = g_raid_create_node(mp, name, md); 1337 md->mdo_softc = sc; 1338 geom = sc->sc_geom; 1339 callout_init(&mdi->mdio_start_co, 1); 1340 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz, 1341 g_raid_intel_go, sc); 1342 mdi->mdio_rootmount = root_mount_hold("GRAID-Intel"); 1343 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount); 1344 } 1345 1346 rcp = g_new_consumer(geom); 1347 g_attach(rcp, pp); 1348 if (g_access(rcp, 1, 1, 1) != 0) 1349 ; //goto fail1; 1350 1351 g_topology_unlock(); 1352 sx_xlock(&sc->sc_lock); 1353 1354 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO); 1355 pd->pd_meta = meta; 1356 pd->pd_disk_pos = -1; 1357 if (spare == 2) { 1358 memcpy(&pd->pd_disk_meta.serial[0], serial, INTEL_SERIAL_LEN); 1359 intel_set_disk_sectors(&pd->pd_disk_meta, 1360 pp->mediasize / pp->sectorsize); 1361 pd->pd_disk_meta.id = 0; 1362 pd->pd_disk_meta.flags = INTEL_F_SPARE; 1363 } else { 1364 pd->pd_disk_meta = meta->disk[disk_pos]; 1365 } 1366 disk = g_raid_create_disk(sc); 1367 disk->d_md_data = (void *)pd; 1368 disk->d_consumer = rcp; 1369 rcp->private = disk; 1370 1371 /* Read kernel dumping information. */ 1372 disk->d_kd.offset = 0; 1373 disk->d_kd.length = OFF_MAX; 1374 len = sizeof(disk->d_kd); 1375 error = g_io_getattr("GEOM::kerneldump", rcp, &len, &disk->d_kd); 1376 if (disk->d_kd.di.dumper == NULL) 1377 G_RAID_DEBUG1(2, sc, "Dumping not supported by %s: %d.", 1378 rcp->provider->name, error); 1379 1380 g_raid_md_intel_new_disk(disk); 1381 1382 sx_xunlock(&sc->sc_lock); 1383 g_topology_lock(); 1384 *gp = geom; 1385 return (result); 1386 fail2: 1387 g_topology_lock(); 1388 g_access(cp, -1, 0, 0); 1389 fail1: 1390 free(meta, M_MD_INTEL); 1391 return (G_RAID_MD_TASTE_FAIL); 1392 } 1393 1394 static int 1395 g_raid_md_event_intel(struct g_raid_md_object *md, 1396 struct g_raid_disk *disk, u_int event) 1397 { 1398 struct g_raid_softc *sc; 1399 struct g_raid_subdisk *sd; 1400 struct g_raid_md_intel_object *mdi; 1401 struct g_raid_md_intel_perdisk *pd; 1402 1403 sc = md->mdo_softc; 1404 mdi = (struct g_raid_md_intel_object *)md; 1405 if (disk == NULL) { 1406 switch (event) { 1407 case G_RAID_NODE_E_START: 1408 if (!mdi->mdio_started) 1409 g_raid_md_intel_start(sc); 1410 return (0); 1411 } 1412 return (-1); 1413 } 1414 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 1415 switch (event) { 1416 case G_RAID_DISK_E_DISCONNECTED: 1417 /* If disk was assigned, just update statuses. */ 1418 if (pd->pd_disk_pos >= 0) { 1419 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1420 if (disk->d_consumer) { 1421 g_raid_kill_consumer(sc, disk->d_consumer); 1422 disk->d_consumer = NULL; 1423 } 1424 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 1425 g_raid_change_subdisk_state(sd, 1426 G_RAID_SUBDISK_S_NONE); 1427 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED, 1428 G_RAID_EVENT_SUBDISK); 1429 } 1430 } else { 1431 /* Otherwise -- delete. */ 1432 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE); 1433 g_raid_destroy_disk(disk); 1434 } 1435 1436 /* Write updated metadata to all disks. */ 1437 g_raid_md_write_intel(md, NULL, NULL, NULL); 1438 1439 /* Check if anything left except placeholders. */ 1440 if (g_raid_ndisks(sc, -1) == 1441 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 1442 g_raid_destroy_node(sc, 0); 1443 else 1444 g_raid_md_intel_refill(sc); 1445 return (0); 1446 } 1447 return (-2); 1448 } 1449 1450 static int 1451 g_raid_md_ctl_intel(struct g_raid_md_object *md, 1452 struct gctl_req *req) 1453 { 1454 struct g_raid_softc *sc; 1455 struct g_raid_volume *vol, *vol1; 1456 struct g_raid_subdisk *sd; 1457 struct g_raid_disk *disk; 1458 struct g_raid_md_intel_object *mdi; 1459 struct g_raid_md_intel_perdisk *pd; 1460 struct g_consumer *cp; 1461 struct g_provider *pp; 1462 char arg[16], serial[INTEL_SERIAL_LEN]; 1463 const char *verb, *volname, *levelname, *diskname; 1464 char *tmp; 1465 int *nargs, *force; 1466 off_t off, size, sectorsize, strip, disk_sectors; 1467 intmax_t *sizearg, *striparg; 1468 int numdisks, i, len, level, qual, update; 1469 int error; 1470 1471 sc = md->mdo_softc; 1472 mdi = (struct g_raid_md_intel_object *)md; 1473 verb = gctl_get_param(req, "verb", NULL); 1474 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs)); 1475 error = 0; 1476 if (strcmp(verb, "label") == 0) { 1477 1478 if (*nargs < 4) { 1479 gctl_error(req, "Invalid number of arguments."); 1480 return (-1); 1481 } 1482 volname = gctl_get_asciiparam(req, "arg1"); 1483 if (volname == NULL) { 1484 gctl_error(req, "No volume name."); 1485 return (-2); 1486 } 1487 levelname = gctl_get_asciiparam(req, "arg2"); 1488 if (levelname == NULL) { 1489 gctl_error(req, "No RAID level."); 1490 return (-3); 1491 } 1492 if (strcasecmp(levelname, "RAID5") == 0) 1493 levelname = "RAID5-LA"; 1494 if (g_raid_volume_str2level(levelname, &level, &qual)) { 1495 gctl_error(req, "Unknown RAID level '%s'.", levelname); 1496 return (-4); 1497 } 1498 numdisks = *nargs - 3; 1499 force = gctl_get_paraml(req, "force", sizeof(*force)); 1500 if (!g_raid_md_intel_supported(level, qual, numdisks, 1501 force ? *force : 0)) { 1502 gctl_error(req, "Unsupported RAID level " 1503 "(0x%02x/0x%02x), or number of disks (%d).", 1504 level, qual, numdisks); 1505 return (-5); 1506 } 1507 1508 /* Search for disks, connect them and probe. */ 1509 size = 0x7fffffffffffffffllu; 1510 sectorsize = 0; 1511 for (i = 0; i < numdisks; i++) { 1512 snprintf(arg, sizeof(arg), "arg%d", i + 3); 1513 diskname = gctl_get_asciiparam(req, arg); 1514 if (diskname == NULL) { 1515 gctl_error(req, "No disk name (%s).", arg); 1516 error = -6; 1517 break; 1518 } 1519 if (strcmp(diskname, "NONE") == 0) { 1520 cp = NULL; 1521 pp = NULL; 1522 } else { 1523 g_topology_lock(); 1524 cp = g_raid_open_consumer(sc, diskname); 1525 if (cp == NULL) { 1526 gctl_error(req, "Can't open disk '%s'.", 1527 diskname); 1528 g_topology_unlock(); 1529 error = -7; 1530 break; 1531 } 1532 pp = cp->provider; 1533 } 1534 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO); 1535 pd->pd_disk_pos = i; 1536 disk = g_raid_create_disk(sc); 1537 disk->d_md_data = (void *)pd; 1538 disk->d_consumer = cp; 1539 if (cp == NULL) { 1540 strcpy(&pd->pd_disk_meta.serial[0], "NONE"); 1541 pd->pd_disk_meta.id = 0xffffffff; 1542 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED; 1543 continue; 1544 } 1545 cp->private = disk; 1546 g_topology_unlock(); 1547 1548 error = g_raid_md_get_label(cp, 1549 &pd->pd_disk_meta.serial[0], INTEL_SERIAL_LEN); 1550 if (error != 0) { 1551 gctl_error(req, 1552 "Can't get serial for provider '%s'.", 1553 diskname); 1554 error = -8; 1555 break; 1556 } 1557 1558 /* Read kernel dumping information. */ 1559 disk->d_kd.offset = 0; 1560 disk->d_kd.length = OFF_MAX; 1561 len = sizeof(disk->d_kd); 1562 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd); 1563 if (disk->d_kd.di.dumper == NULL) 1564 G_RAID_DEBUG1(2, sc, 1565 "Dumping not supported by %s.", 1566 cp->provider->name); 1567 1568 intel_set_disk_sectors(&pd->pd_disk_meta, 1569 pp->mediasize / pp->sectorsize); 1570 if (size > pp->mediasize) 1571 size = pp->mediasize; 1572 if (sectorsize < pp->sectorsize) 1573 sectorsize = pp->sectorsize; 1574 pd->pd_disk_meta.id = 0; 1575 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED | INTEL_F_ONLINE; 1576 } 1577 if (error != 0) 1578 return (error); 1579 1580 if (sectorsize <= 0) { 1581 gctl_error(req, "Can't get sector size."); 1582 return (-8); 1583 } 1584 1585 /* Reserve some space for metadata. */ 1586 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize; 1587 1588 /* Handle size argument. */ 1589 len = sizeof(*sizearg); 1590 sizearg = gctl_get_param(req, "size", &len); 1591 if (sizearg != NULL && len == sizeof(*sizearg) && 1592 *sizearg > 0) { 1593 if (*sizearg > size) { 1594 gctl_error(req, "Size too big %lld > %lld.", 1595 (long long)*sizearg, (long long)size); 1596 return (-9); 1597 } 1598 size = *sizearg; 1599 } 1600 1601 /* Handle strip argument. */ 1602 strip = 131072; 1603 len = sizeof(*striparg); 1604 striparg = gctl_get_param(req, "strip", &len); 1605 if (striparg != NULL && len == sizeof(*striparg) && 1606 *striparg > 0) { 1607 if (*striparg < sectorsize) { 1608 gctl_error(req, "Strip size too small."); 1609 return (-10); 1610 } 1611 if (*striparg % sectorsize != 0) { 1612 gctl_error(req, "Incorrect strip size."); 1613 return (-11); 1614 } 1615 if (strip > 65535 * sectorsize) { 1616 gctl_error(req, "Strip size too big."); 1617 return (-12); 1618 } 1619 strip = *striparg; 1620 } 1621 1622 /* Round size down to strip or sector. */ 1623 if (level == G_RAID_VOLUME_RL_RAID1) 1624 size -= (size % sectorsize); 1625 else if (level == G_RAID_VOLUME_RL_RAID1E && 1626 (numdisks & 1) != 0) 1627 size -= (size % (2 * strip)); 1628 else 1629 size -= (size % strip); 1630 if (size <= 0) { 1631 gctl_error(req, "Size too small."); 1632 return (-13); 1633 } 1634 1635 /* We have all we need, create things: volume, ... */ 1636 mdi->mdio_started = 1; 1637 vol = g_raid_create_volume(sc, volname, -1); 1638 vol->v_md_data = (void *)(intptr_t)0; 1639 vol->v_raid_level = level; 1640 vol->v_raid_level_qualifier = qual; 1641 vol->v_strip_size = strip; 1642 vol->v_disks_count = numdisks; 1643 if (level == G_RAID_VOLUME_RL_RAID0) 1644 vol->v_mediasize = size * numdisks; 1645 else if (level == G_RAID_VOLUME_RL_RAID1) 1646 vol->v_mediasize = size; 1647 else if (level == G_RAID_VOLUME_RL_RAID5) 1648 vol->v_mediasize = size * (numdisks - 1); 1649 else { /* RAID1E */ 1650 vol->v_mediasize = ((size * numdisks) / strip / 2) * 1651 strip; 1652 } 1653 vol->v_sectorsize = sectorsize; 1654 g_raid_start_volume(vol); 1655 1656 /* , and subdisks. */ 1657 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1658 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 1659 sd = &vol->v_subdisks[pd->pd_disk_pos]; 1660 sd->sd_disk = disk; 1661 sd->sd_offset = 0; 1662 sd->sd_size = size; 1663 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 1664 if (sd->sd_disk->d_consumer != NULL) { 1665 g_raid_change_disk_state(disk, 1666 G_RAID_DISK_S_ACTIVE); 1667 if (level == G_RAID_VOLUME_RL_RAID5) 1668 g_raid_change_subdisk_state(sd, 1669 G_RAID_SUBDISK_S_UNINITIALIZED); 1670 else 1671 g_raid_change_subdisk_state(sd, 1672 G_RAID_SUBDISK_S_ACTIVE); 1673 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 1674 G_RAID_EVENT_SUBDISK); 1675 } else { 1676 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1677 } 1678 } 1679 1680 /* Write metadata based on created entities. */ 1681 G_RAID_DEBUG1(0, sc, "Array started."); 1682 g_raid_md_write_intel(md, NULL, NULL, NULL); 1683 1684 /* Pickup any STALE/SPARE disks to refill array if needed. */ 1685 g_raid_md_intel_refill(sc); 1686 1687 g_raid_event_send(vol, G_RAID_VOLUME_E_START, 1688 G_RAID_EVENT_VOLUME); 1689 return (0); 1690 } 1691 if (strcmp(verb, "add") == 0) { 1692 1693 if (*nargs != 3) { 1694 gctl_error(req, "Invalid number of arguments."); 1695 return (-1); 1696 } 1697 volname = gctl_get_asciiparam(req, "arg1"); 1698 if (volname == NULL) { 1699 gctl_error(req, "No volume name."); 1700 return (-2); 1701 } 1702 levelname = gctl_get_asciiparam(req, "arg2"); 1703 if (levelname == NULL) { 1704 gctl_error(req, "No RAID level."); 1705 return (-3); 1706 } 1707 if (strcasecmp(levelname, "RAID5") == 0) 1708 levelname = "RAID5-LA"; 1709 if (g_raid_volume_str2level(levelname, &level, &qual)) { 1710 gctl_error(req, "Unknown RAID level '%s'.", levelname); 1711 return (-4); 1712 } 1713 1714 /* Look for existing volumes. */ 1715 i = 0; 1716 vol1 = NULL; 1717 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 1718 vol1 = vol; 1719 i++; 1720 } 1721 if (i > 1) { 1722 gctl_error(req, "Maximum two volumes supported."); 1723 return (-6); 1724 } 1725 if (vol1 == NULL) { 1726 gctl_error(req, "At least one volume must exist."); 1727 return (-7); 1728 } 1729 1730 numdisks = vol1->v_disks_count; 1731 force = gctl_get_paraml(req, "force", sizeof(*force)); 1732 if (!g_raid_md_intel_supported(level, qual, numdisks, 1733 force ? *force : 0)) { 1734 gctl_error(req, "Unsupported RAID level " 1735 "(0x%02x/0x%02x), or number of disks (%d).", 1736 level, qual, numdisks); 1737 return (-5); 1738 } 1739 1740 /* Collect info about present disks. */ 1741 size = 0x7fffffffffffffffllu; 1742 sectorsize = 512; 1743 for (i = 0; i < numdisks; i++) { 1744 disk = vol1->v_subdisks[i].sd_disk; 1745 pd = (struct g_raid_md_intel_perdisk *) 1746 disk->d_md_data; 1747 disk_sectors = 1748 intel_get_disk_sectors(&pd->pd_disk_meta); 1749 1750 if (disk_sectors * 512 < size) 1751 size = disk_sectors * 512; 1752 if (disk->d_consumer != NULL && 1753 disk->d_consumer->provider != NULL && 1754 disk->d_consumer->provider->sectorsize > 1755 sectorsize) { 1756 sectorsize = 1757 disk->d_consumer->provider->sectorsize; 1758 } 1759 } 1760 1761 /* Reserve some space for metadata. */ 1762 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize; 1763 1764 /* Decide insert before or after. */ 1765 sd = &vol1->v_subdisks[0]; 1766 if (sd->sd_offset > 1767 size - (sd->sd_offset + sd->sd_size)) { 1768 off = 0; 1769 size = sd->sd_offset; 1770 } else { 1771 off = sd->sd_offset + sd->sd_size; 1772 size = size - (sd->sd_offset + sd->sd_size); 1773 } 1774 1775 /* Handle strip argument. */ 1776 strip = 131072; 1777 len = sizeof(*striparg); 1778 striparg = gctl_get_param(req, "strip", &len); 1779 if (striparg != NULL && len == sizeof(*striparg) && 1780 *striparg > 0) { 1781 if (*striparg < sectorsize) { 1782 gctl_error(req, "Strip size too small."); 1783 return (-10); 1784 } 1785 if (*striparg % sectorsize != 0) { 1786 gctl_error(req, "Incorrect strip size."); 1787 return (-11); 1788 } 1789 if (strip > 65535 * sectorsize) { 1790 gctl_error(req, "Strip size too big."); 1791 return (-12); 1792 } 1793 strip = *striparg; 1794 } 1795 1796 /* Round offset up to strip. */ 1797 if (off % strip != 0) { 1798 size -= strip - off % strip; 1799 off += strip - off % strip; 1800 } 1801 1802 /* Handle size argument. */ 1803 len = sizeof(*sizearg); 1804 sizearg = gctl_get_param(req, "size", &len); 1805 if (sizearg != NULL && len == sizeof(*sizearg) && 1806 *sizearg > 0) { 1807 if (*sizearg > size) { 1808 gctl_error(req, "Size too big %lld > %lld.", 1809 (long long)*sizearg, (long long)size); 1810 return (-9); 1811 } 1812 size = *sizearg; 1813 } 1814 1815 /* Round size down to strip or sector. */ 1816 if (level == G_RAID_VOLUME_RL_RAID1) 1817 size -= (size % sectorsize); 1818 else 1819 size -= (size % strip); 1820 if (size <= 0) { 1821 gctl_error(req, "Size too small."); 1822 return (-13); 1823 } 1824 if (size > 0xffffffffllu * sectorsize) { 1825 gctl_error(req, "Size too big."); 1826 return (-14); 1827 } 1828 1829 /* We have all we need, create things: volume, ... */ 1830 vol = g_raid_create_volume(sc, volname, -1); 1831 vol->v_md_data = (void *)(intptr_t)i; 1832 vol->v_raid_level = level; 1833 vol->v_raid_level_qualifier = qual; 1834 vol->v_strip_size = strip; 1835 vol->v_disks_count = numdisks; 1836 if (level == G_RAID_VOLUME_RL_RAID0) 1837 vol->v_mediasize = size * numdisks; 1838 else if (level == G_RAID_VOLUME_RL_RAID1) 1839 vol->v_mediasize = size; 1840 else if (level == G_RAID_VOLUME_RL_RAID5) 1841 vol->v_mediasize = size * (numdisks - 1); 1842 else { /* RAID1E */ 1843 vol->v_mediasize = ((size * numdisks) / strip / 2) * 1844 strip; 1845 } 1846 vol->v_sectorsize = sectorsize; 1847 g_raid_start_volume(vol); 1848 1849 /* , and subdisks. */ 1850 for (i = 0; i < numdisks; i++) { 1851 disk = vol1->v_subdisks[i].sd_disk; 1852 sd = &vol->v_subdisks[i]; 1853 sd->sd_disk = disk; 1854 sd->sd_offset = off; 1855 sd->sd_size = size; 1856 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 1857 if (disk->d_state == G_RAID_DISK_S_ACTIVE) { 1858 if (level == G_RAID_VOLUME_RL_RAID5) 1859 g_raid_change_subdisk_state(sd, 1860 G_RAID_SUBDISK_S_UNINITIALIZED); 1861 else 1862 g_raid_change_subdisk_state(sd, 1863 G_RAID_SUBDISK_S_ACTIVE); 1864 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 1865 G_RAID_EVENT_SUBDISK); 1866 } 1867 } 1868 1869 /* Write metadata based on created entities. */ 1870 g_raid_md_write_intel(md, NULL, NULL, NULL); 1871 1872 g_raid_event_send(vol, G_RAID_VOLUME_E_START, 1873 G_RAID_EVENT_VOLUME); 1874 return (0); 1875 } 1876 if (strcmp(verb, "delete") == 0) { 1877 1878 /* Full node destruction. */ 1879 if (*nargs == 1) { 1880 /* Check if some volume is still open. */ 1881 force = gctl_get_paraml(req, "force", sizeof(*force)); 1882 if (force != NULL && *force == 0 && 1883 g_raid_nopens(sc) != 0) { 1884 gctl_error(req, "Some volume is still open."); 1885 return (-4); 1886 } 1887 1888 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1889 if (disk->d_consumer) 1890 intel_meta_erase(disk->d_consumer); 1891 } 1892 g_raid_destroy_node(sc, 0); 1893 return (0); 1894 } 1895 1896 /* Destroy specified volume. If it was last - all node. */ 1897 if (*nargs != 2) { 1898 gctl_error(req, "Invalid number of arguments."); 1899 return (-1); 1900 } 1901 volname = gctl_get_asciiparam(req, "arg1"); 1902 if (volname == NULL) { 1903 gctl_error(req, "No volume name."); 1904 return (-2); 1905 } 1906 1907 /* Search for volume. */ 1908 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 1909 if (strcmp(vol->v_name, volname) == 0) 1910 break; 1911 } 1912 if (vol == NULL) { 1913 i = strtol(volname, &tmp, 10); 1914 if (verb != volname && tmp[0] == 0) { 1915 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 1916 if (vol->v_global_id == i) 1917 break; 1918 } 1919 } 1920 } 1921 if (vol == NULL) { 1922 gctl_error(req, "Volume '%s' not found.", volname); 1923 return (-3); 1924 } 1925 1926 /* Check if volume is still open. */ 1927 force = gctl_get_paraml(req, "force", sizeof(*force)); 1928 if (force != NULL && *force == 0 && 1929 vol->v_provider_open != 0) { 1930 gctl_error(req, "Volume is still open."); 1931 return (-4); 1932 } 1933 1934 /* Destroy volume and potentially node. */ 1935 i = 0; 1936 TAILQ_FOREACH(vol1, &sc->sc_volumes, v_next) 1937 i++; 1938 if (i >= 2) { 1939 g_raid_destroy_volume(vol); 1940 g_raid_md_write_intel(md, NULL, NULL, NULL); 1941 } else { 1942 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1943 if (disk->d_consumer) 1944 intel_meta_erase(disk->d_consumer); 1945 } 1946 g_raid_destroy_node(sc, 0); 1947 } 1948 return (0); 1949 } 1950 if (strcmp(verb, "remove") == 0 || 1951 strcmp(verb, "fail") == 0) { 1952 if (*nargs < 2) { 1953 gctl_error(req, "Invalid number of arguments."); 1954 return (-1); 1955 } 1956 for (i = 1; i < *nargs; i++) { 1957 snprintf(arg, sizeof(arg), "arg%d", i); 1958 diskname = gctl_get_asciiparam(req, arg); 1959 if (diskname == NULL) { 1960 gctl_error(req, "No disk name (%s).", arg); 1961 error = -2; 1962 break; 1963 } 1964 if (strncmp(diskname, "/dev/", 5) == 0) 1965 diskname += 5; 1966 1967 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1968 if (disk->d_consumer != NULL && 1969 disk->d_consumer->provider != NULL && 1970 strcmp(disk->d_consumer->provider->name, 1971 diskname) == 0) 1972 break; 1973 } 1974 if (disk == NULL) { 1975 gctl_error(req, "Disk '%s' not found.", 1976 diskname); 1977 error = -3; 1978 break; 1979 } 1980 1981 if (strcmp(verb, "fail") == 0) { 1982 g_raid_md_fail_disk_intel(md, NULL, disk); 1983 continue; 1984 } 1985 1986 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 1987 1988 /* Erase metadata on deleting disk. */ 1989 intel_meta_erase(disk->d_consumer); 1990 1991 /* If disk was assigned, just update statuses. */ 1992 if (pd->pd_disk_pos >= 0) { 1993 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1994 g_raid_kill_consumer(sc, disk->d_consumer); 1995 disk->d_consumer = NULL; 1996 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 1997 g_raid_change_subdisk_state(sd, 1998 G_RAID_SUBDISK_S_NONE); 1999 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED, 2000 G_RAID_EVENT_SUBDISK); 2001 } 2002 } else { 2003 /* Otherwise -- delete. */ 2004 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE); 2005 g_raid_destroy_disk(disk); 2006 } 2007 } 2008 2009 /* Write updated metadata to remaining disks. */ 2010 g_raid_md_write_intel(md, NULL, NULL, NULL); 2011 2012 /* Check if anything left except placeholders. */ 2013 if (g_raid_ndisks(sc, -1) == 2014 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 2015 g_raid_destroy_node(sc, 0); 2016 else 2017 g_raid_md_intel_refill(sc); 2018 return (error); 2019 } 2020 if (strcmp(verb, "insert") == 0) { 2021 if (*nargs < 2) { 2022 gctl_error(req, "Invalid number of arguments."); 2023 return (-1); 2024 } 2025 update = 0; 2026 for (i = 1; i < *nargs; i++) { 2027 /* Get disk name. */ 2028 snprintf(arg, sizeof(arg), "arg%d", i); 2029 diskname = gctl_get_asciiparam(req, arg); 2030 if (diskname == NULL) { 2031 gctl_error(req, "No disk name (%s).", arg); 2032 error = -3; 2033 break; 2034 } 2035 2036 /* Try to find provider with specified name. */ 2037 g_topology_lock(); 2038 cp = g_raid_open_consumer(sc, diskname); 2039 if (cp == NULL) { 2040 gctl_error(req, "Can't open disk '%s'.", 2041 diskname); 2042 g_topology_unlock(); 2043 error = -4; 2044 break; 2045 } 2046 pp = cp->provider; 2047 g_topology_unlock(); 2048 2049 /* Read disk serial. */ 2050 error = g_raid_md_get_label(cp, 2051 &serial[0], INTEL_SERIAL_LEN); 2052 if (error != 0) { 2053 gctl_error(req, 2054 "Can't get serial for provider '%s'.", 2055 diskname); 2056 g_raid_kill_consumer(sc, cp); 2057 error = -7; 2058 break; 2059 } 2060 2061 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO); 2062 pd->pd_disk_pos = -1; 2063 2064 disk = g_raid_create_disk(sc); 2065 disk->d_consumer = cp; 2066 disk->d_md_data = (void *)pd; 2067 cp->private = disk; 2068 2069 /* Read kernel dumping information. */ 2070 disk->d_kd.offset = 0; 2071 disk->d_kd.length = OFF_MAX; 2072 len = sizeof(disk->d_kd); 2073 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd); 2074 if (disk->d_kd.di.dumper == NULL) 2075 G_RAID_DEBUG1(2, sc, 2076 "Dumping not supported by %s.", 2077 cp->provider->name); 2078 2079 memcpy(&pd->pd_disk_meta.serial[0], &serial[0], 2080 INTEL_SERIAL_LEN); 2081 intel_set_disk_sectors(&pd->pd_disk_meta, 2082 pp->mediasize / pp->sectorsize); 2083 pd->pd_disk_meta.id = 0; 2084 pd->pd_disk_meta.flags = INTEL_F_SPARE; 2085 2086 /* Welcome the "new" disk. */ 2087 update += g_raid_md_intel_start_disk(disk); 2088 if (disk->d_state == G_RAID_DISK_S_SPARE) { 2089 intel_meta_write_spare(cp, &pd->pd_disk_meta); 2090 g_raid_destroy_disk(disk); 2091 } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) { 2092 gctl_error(req, "Disk '%s' doesn't fit.", 2093 diskname); 2094 g_raid_destroy_disk(disk); 2095 error = -8; 2096 break; 2097 } 2098 } 2099 2100 /* Write new metadata if we changed something. */ 2101 if (update) 2102 g_raid_md_write_intel(md, NULL, NULL, NULL); 2103 return (error); 2104 } 2105 return (-100); 2106 } 2107 2108 static int 2109 g_raid_md_write_intel(struct g_raid_md_object *md, struct g_raid_volume *tvol, 2110 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk) 2111 { 2112 struct g_raid_softc *sc; 2113 struct g_raid_volume *vol; 2114 struct g_raid_subdisk *sd; 2115 struct g_raid_disk *disk; 2116 struct g_raid_md_intel_object *mdi; 2117 struct g_raid_md_intel_perdisk *pd; 2118 struct intel_raid_conf *meta; 2119 struct intel_raid_vol *mvol; 2120 struct intel_raid_map *mmap0, *mmap1; 2121 off_t sectorsize = 512, pos; 2122 const char *version, *cv; 2123 int vi, sdi, numdisks, len, state, stale; 2124 2125 sc = md->mdo_softc; 2126 mdi = (struct g_raid_md_intel_object *)md; 2127 2128 if (sc->sc_stopping == G_RAID_DESTROY_HARD) 2129 return (0); 2130 2131 /* Bump generation. Newly written metadata may differ from previous. */ 2132 mdi->mdio_generation++; 2133 2134 /* Count number of disks. */ 2135 numdisks = 0; 2136 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 2137 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 2138 if (pd->pd_disk_pos < 0) 2139 continue; 2140 numdisks++; 2141 if (disk->d_state == G_RAID_DISK_S_ACTIVE) { 2142 pd->pd_disk_meta.flags = 2143 INTEL_F_ONLINE | INTEL_F_ASSIGNED; 2144 } else if (disk->d_state == G_RAID_DISK_S_FAILED) { 2145 pd->pd_disk_meta.flags = INTEL_F_FAILED | INTEL_F_ASSIGNED; 2146 } else { 2147 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED; 2148 if (pd->pd_disk_meta.id != 0xffffffff) { 2149 pd->pd_disk_meta.id = 0xffffffff; 2150 len = strlen(pd->pd_disk_meta.serial); 2151 len = min(len, INTEL_SERIAL_LEN - 3); 2152 strcpy(pd->pd_disk_meta.serial + len, ":0"); 2153 } 2154 } 2155 } 2156 2157 /* Fill anchor and disks. */ 2158 meta = malloc(INTEL_MAX_MD_SIZE(numdisks), 2159 M_MD_INTEL, M_WAITOK | M_ZERO); 2160 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1); 2161 meta->config_size = INTEL_MAX_MD_SIZE(numdisks); 2162 meta->config_id = mdi->mdio_config_id; 2163 meta->generation = mdi->mdio_generation; 2164 meta->attributes = INTEL_ATTR_CHECKSUM; 2165 meta->total_disks = numdisks; 2166 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 2167 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 2168 if (pd->pd_disk_pos < 0) 2169 continue; 2170 meta->disk[pd->pd_disk_pos] = pd->pd_disk_meta; 2171 } 2172 2173 /* Fill volumes and maps. */ 2174 vi = 0; 2175 version = INTEL_VERSION_1000; 2176 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 2177 if (vol->v_stopping) 2178 continue; 2179 mvol = intel_get_volume(meta, vi); 2180 2181 /* New metadata may have different volumes order. */ 2182 vol->v_md_data = (void *)(intptr_t)vi; 2183 2184 for (sdi = 0; sdi < vol->v_disks_count; sdi++) { 2185 sd = &vol->v_subdisks[sdi]; 2186 if (sd->sd_disk != NULL) 2187 break; 2188 } 2189 if (sdi >= vol->v_disks_count) 2190 panic("No any filled subdisk in volume"); 2191 if (vol->v_mediasize >= 0x20000000000llu) 2192 meta->attributes |= INTEL_ATTR_2TB; 2193 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0) 2194 meta->attributes |= INTEL_ATTR_RAID0; 2195 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) 2196 meta->attributes |= INTEL_ATTR_RAID1; 2197 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5) 2198 meta->attributes |= INTEL_ATTR_RAID5; 2199 else 2200 meta->attributes |= INTEL_ATTR_RAID10; 2201 2202 if (meta->attributes & INTEL_ATTR_2TB) 2203 cv = INTEL_VERSION_1300; 2204 // else if (dev->status == DEV_CLONE_N_GO) 2205 // cv = INTEL_VERSION_1206; 2206 else if (vol->v_disks_count > 4) 2207 cv = INTEL_VERSION_1204; 2208 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5) 2209 cv = INTEL_VERSION_1202; 2210 else if (vol->v_disks_count > 2) 2211 cv = INTEL_VERSION_1201; 2212 else if (vi > 0) 2213 cv = INTEL_VERSION_1200; 2214 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) 2215 cv = INTEL_VERSION_1100; 2216 else 2217 cv = INTEL_VERSION_1000; 2218 if (strcmp(cv, version) > 0) 2219 version = cv; 2220 2221 strlcpy(&mvol->name[0], vol->v_name, sizeof(mvol->name)); 2222 mvol->total_sectors = vol->v_mediasize / sectorsize; 2223 2224 /* Check for any recovery in progress. */ 2225 state = G_RAID_SUBDISK_S_ACTIVE; 2226 pos = 0x7fffffffffffffffllu; 2227 stale = 0; 2228 for (sdi = 0; sdi < vol->v_disks_count; sdi++) { 2229 sd = &vol->v_subdisks[sdi]; 2230 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD) 2231 state = G_RAID_SUBDISK_S_REBUILD; 2232 else if (sd->sd_state == G_RAID_SUBDISK_S_RESYNC && 2233 state != G_RAID_SUBDISK_S_REBUILD) 2234 state = G_RAID_SUBDISK_S_RESYNC; 2235 else if (sd->sd_state == G_RAID_SUBDISK_S_STALE) 2236 stale = 1; 2237 if ((sd->sd_state == G_RAID_SUBDISK_S_REBUILD || 2238 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) && 2239 sd->sd_rebuild_pos < pos) 2240 pos = sd->sd_rebuild_pos; 2241 } 2242 if (state == G_RAID_SUBDISK_S_REBUILD) { 2243 mvol->migr_state = 1; 2244 mvol->migr_type = INTEL_MT_REBUILD; 2245 } else if (state == G_RAID_SUBDISK_S_RESYNC) { 2246 mvol->migr_state = 1; 2247 /* mvol->migr_type = INTEL_MT_REPAIR; */ 2248 mvol->migr_type = INTEL_MT_VERIFY; 2249 mvol->state |= INTEL_ST_VERIFY_AND_FIX; 2250 } else 2251 mvol->migr_state = 0; 2252 mvol->dirty = (vol->v_dirty || stale); 2253 2254 mmap0 = intel_get_map(mvol, 0); 2255 2256 /* Write map / common part of two maps. */ 2257 intel_set_map_offset(mmap0, sd->sd_offset / sectorsize); 2258 intel_set_map_disk_sectors(mmap0, sd->sd_size / sectorsize); 2259 mmap0->strip_sectors = vol->v_strip_size / sectorsize; 2260 if (vol->v_state == G_RAID_VOLUME_S_BROKEN) 2261 mmap0->status = INTEL_S_FAILURE; 2262 else if (vol->v_state == G_RAID_VOLUME_S_DEGRADED) 2263 mmap0->status = INTEL_S_DEGRADED; 2264 else if (g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_UNINITIALIZED) 2265 == g_raid_nsubdisks(vol, -1)) 2266 mmap0->status = INTEL_S_UNINITIALIZED; 2267 else 2268 mmap0->status = INTEL_S_READY; 2269 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0) 2270 mmap0->type = INTEL_T_RAID0; 2271 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1 || 2272 vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) 2273 mmap0->type = INTEL_T_RAID1; 2274 else 2275 mmap0->type = INTEL_T_RAID5; 2276 mmap0->total_disks = vol->v_disks_count; 2277 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) 2278 mmap0->total_domains = vol->v_disks_count; 2279 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) 2280 mmap0->total_domains = 2; 2281 else 2282 mmap0->total_domains = 1; 2283 intel_set_map_stripe_count(mmap0, 2284 sd->sd_size / vol->v_strip_size / mmap0->total_domains); 2285 mmap0->failed_disk_num = 0xff; 2286 mmap0->ddf = 1; 2287 2288 /* If there are two maps - copy common and update. */ 2289 if (mvol->migr_state) { 2290 intel_set_vol_curr_migr_unit(mvol, 2291 pos / vol->v_strip_size / mmap0->total_domains); 2292 mmap1 = intel_get_map(mvol, 1); 2293 memcpy(mmap1, mmap0, sizeof(struct intel_raid_map)); 2294 mmap0->status = INTEL_S_READY; 2295 } else 2296 mmap1 = NULL; 2297 2298 /* Write disk indexes and put rebuild flags. */ 2299 for (sdi = 0; sdi < vol->v_disks_count; sdi++) { 2300 sd = &vol->v_subdisks[sdi]; 2301 pd = (struct g_raid_md_intel_perdisk *) 2302 sd->sd_disk->d_md_data; 2303 mmap0->disk_idx[sdi] = pd->pd_disk_pos; 2304 if (mvol->migr_state) 2305 mmap1->disk_idx[sdi] = pd->pd_disk_pos; 2306 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD || 2307 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) { 2308 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD; 2309 } else if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE && 2310 sd->sd_state != G_RAID_SUBDISK_S_STALE && 2311 sd->sd_state != G_RAID_SUBDISK_S_UNINITIALIZED) { 2312 mmap0->disk_idx[sdi] |= INTEL_DI_RBLD; 2313 if (mvol->migr_state) 2314 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD; 2315 } 2316 if ((sd->sd_state == G_RAID_SUBDISK_S_NONE || 2317 sd->sd_state == G_RAID_SUBDISK_S_FAILED) && 2318 mmap0->failed_disk_num == 0xff) { 2319 mmap0->failed_disk_num = sdi; 2320 if (mvol->migr_state) 2321 mmap1->failed_disk_num = sdi; 2322 } 2323 } 2324 vi++; 2325 } 2326 meta->total_volumes = vi; 2327 if (strcmp(version, INTEL_VERSION_1300) != 0) 2328 meta->attributes &= INTEL_ATTR_CHECKSUM; 2329 memcpy(&meta->version[0], version, sizeof(INTEL_VERSION_1000) - 1); 2330 2331 /* We are done. Print meta data and store them to disks. */ 2332 g_raid_md_intel_print(meta); 2333 if (mdi->mdio_meta != NULL) 2334 free(mdi->mdio_meta, M_MD_INTEL); 2335 mdi->mdio_meta = meta; 2336 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 2337 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 2338 if (disk->d_state != G_RAID_DISK_S_ACTIVE) 2339 continue; 2340 if (pd->pd_meta != NULL) { 2341 free(pd->pd_meta, M_MD_INTEL); 2342 pd->pd_meta = NULL; 2343 } 2344 pd->pd_meta = intel_meta_copy(meta); 2345 intel_meta_write(disk->d_consumer, meta); 2346 } 2347 return (0); 2348 } 2349 2350 static int 2351 g_raid_md_fail_disk_intel(struct g_raid_md_object *md, 2352 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk) 2353 { 2354 struct g_raid_softc *sc; 2355 struct g_raid_md_intel_object *mdi; 2356 struct g_raid_md_intel_perdisk *pd; 2357 struct g_raid_subdisk *sd; 2358 2359 sc = md->mdo_softc; 2360 mdi = (struct g_raid_md_intel_object *)md; 2361 pd = (struct g_raid_md_intel_perdisk *)tdisk->d_md_data; 2362 2363 /* We can't fail disk that is not a part of array now. */ 2364 if (pd->pd_disk_pos < 0) 2365 return (-1); 2366 2367 /* 2368 * Mark disk as failed in metadata and try to write that metadata 2369 * to the disk itself to prevent it's later resurrection as STALE. 2370 */ 2371 mdi->mdio_meta->disk[pd->pd_disk_pos].flags = INTEL_F_FAILED; 2372 pd->pd_disk_meta.flags = INTEL_F_FAILED; 2373 g_raid_md_intel_print(mdi->mdio_meta); 2374 if (tdisk->d_consumer) 2375 intel_meta_write(tdisk->d_consumer, mdi->mdio_meta); 2376 2377 /* Change states. */ 2378 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED); 2379 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) { 2380 g_raid_change_subdisk_state(sd, 2381 G_RAID_SUBDISK_S_FAILED); 2382 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED, 2383 G_RAID_EVENT_SUBDISK); 2384 } 2385 2386 /* Write updated metadata to remaining disks. */ 2387 g_raid_md_write_intel(md, NULL, NULL, tdisk); 2388 2389 /* Check if anything left except placeholders. */ 2390 if (g_raid_ndisks(sc, -1) == 2391 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 2392 g_raid_destroy_node(sc, 0); 2393 else 2394 g_raid_md_intel_refill(sc); 2395 return (0); 2396 } 2397 2398 static int 2399 g_raid_md_free_disk_intel(struct g_raid_md_object *md, 2400 struct g_raid_disk *disk) 2401 { 2402 struct g_raid_md_intel_perdisk *pd; 2403 2404 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 2405 if (pd->pd_meta != NULL) { 2406 free(pd->pd_meta, M_MD_INTEL); 2407 pd->pd_meta = NULL; 2408 } 2409 free(pd, M_MD_INTEL); 2410 disk->d_md_data = NULL; 2411 return (0); 2412 } 2413 2414 static int 2415 g_raid_md_free_intel(struct g_raid_md_object *md) 2416 { 2417 struct g_raid_md_intel_object *mdi; 2418 2419 mdi = (struct g_raid_md_intel_object *)md; 2420 if (!mdi->mdio_started) { 2421 mdi->mdio_started = 0; 2422 callout_stop(&mdi->mdio_start_co); 2423 G_RAID_DEBUG1(1, md->mdo_softc, 2424 "root_mount_rel %p", mdi->mdio_rootmount); 2425 root_mount_rel(mdi->mdio_rootmount); 2426 mdi->mdio_rootmount = NULL; 2427 } 2428 if (mdi->mdio_meta != NULL) { 2429 free(mdi->mdio_meta, M_MD_INTEL); 2430 mdi->mdio_meta = NULL; 2431 } 2432 return (0); 2433 } 2434 2435 G_RAID_MD_DECLARE(g_raid_md_intel); 2436