md_intel.c revision 235092
1/*- 2 * Copyright (c) 2010 Alexander Motin <mav@FreeBSD.org> 3 * Copyright (c) 2000 - 2008 S��ren 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: stable/9/sys/geom/raid/md_intel.c 235092 2012-05-06 15:55:01Z mav $"); 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 46static MALLOC_DEFINE(M_MD_INTEL, "md_intel_data", "GEOM_RAID Intel metadata"); 47 48struct 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 77struct 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 123struct 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 138struct 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 184struct 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 190struct 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 202static g_raid_md_create_t g_raid_md_create_intel; 203static g_raid_md_taste_t g_raid_md_taste_intel; 204static g_raid_md_event_t g_raid_md_event_intel; 205static g_raid_md_ctl_t g_raid_md_ctl_intel; 206static g_raid_md_write_t g_raid_md_write_intel; 207static g_raid_md_fail_disk_t g_raid_md_fail_disk_intel; 208static g_raid_md_free_disk_t g_raid_md_free_disk_intel; 209static g_raid_md_free_t g_raid_md_free_intel; 210 211static 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 223static 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 231static struct intel_raid_map * 232intel_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 246static struct intel_raid_vol * 247intel_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 263static off_t 264intel_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 272static void 273intel_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 280static off_t 281intel_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 289static void 290intel_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 297static void 298intel_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 305static off_t 306intel_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 314static void 315intel_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 322static off_t 323intel_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 331static void 332intel_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 339static void 340g_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 405static struct intel_raid_conf * 406intel_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 415static int 416intel_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 428static struct intel_raid_conf * 429intel_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) { 504badsize: 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 564static int 565intel_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 603static int 604intel_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 623static int 624intel_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 644static struct g_raid_disk * 645g_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 658static int 659g_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 break; 687 default: 688 return (0); 689 } 690 if (qual != G_RAID_VOLUME_RLQ_NONE) 691 return (0); 692 return (1); 693} 694 695static struct g_raid_volume * 696g_raid_md_intel_get_volume(struct g_raid_softc *sc, int id) 697{ 698 struct g_raid_volume *mvol; 699 700 TAILQ_FOREACH(mvol, &sc->sc_volumes, v_next) { 701 if ((intptr_t)(mvol->v_md_data) == id) 702 break; 703 } 704 return (mvol); 705} 706 707static int 708g_raid_md_intel_start_disk(struct g_raid_disk *disk) 709{ 710 struct g_raid_softc *sc; 711 struct g_raid_subdisk *sd, *tmpsd; 712 struct g_raid_disk *olddisk, *tmpdisk; 713 struct g_raid_md_object *md; 714 struct g_raid_md_intel_object *mdi; 715 struct g_raid_md_intel_perdisk *pd, *oldpd; 716 struct intel_raid_conf *meta; 717 struct intel_raid_vol *mvol; 718 struct intel_raid_map *mmap0, *mmap1; 719 int disk_pos, resurrection = 0; 720 721 sc = disk->d_softc; 722 md = sc->sc_md; 723 mdi = (struct g_raid_md_intel_object *)md; 724 meta = mdi->mdio_meta; 725 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 726 olddisk = NULL; 727 728 /* Find disk position in metadata by it's serial. */ 729 disk_pos = intel_meta_find_disk(meta, pd->pd_disk_meta.serial); 730 if (disk_pos < 0) { 731 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk"); 732 /* Failed stale disk is useless for us. */ 733 if (pd->pd_disk_meta.flags & INTEL_F_FAILED) { 734 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE_FAILED); 735 return (0); 736 } 737 /* If we are in the start process, that's all for now. */ 738 if (!mdi->mdio_started) 739 goto nofit; 740 /* 741 * If we have already started - try to get use of the disk. 742 * Try to replace OFFLINE disks first, then FAILED. 743 */ 744 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) { 745 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE && 746 tmpdisk->d_state != G_RAID_DISK_S_FAILED) 747 continue; 748 /* Make sure this disk is big enough. */ 749 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) { 750 off_t disk_sectors = 751 intel_get_disk_sectors(&pd->pd_disk_meta); 752 753 if (sd->sd_offset + sd->sd_size + 4096 > 754 disk_sectors * 512) { 755 G_RAID_DEBUG1(1, sc, 756 "Disk too small (%llu < %llu)", 757 (unsigned long long) 758 disk_sectors * 512, 759 (unsigned long long) 760 sd->sd_offset + sd->sd_size + 4096); 761 break; 762 } 763 } 764 if (sd != NULL) 765 continue; 766 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) { 767 olddisk = tmpdisk; 768 break; 769 } else if (olddisk == NULL) 770 olddisk = tmpdisk; 771 } 772 if (olddisk == NULL) { 773nofit: 774 if (pd->pd_disk_meta.flags & INTEL_F_SPARE) { 775 g_raid_change_disk_state(disk, 776 G_RAID_DISK_S_SPARE); 777 return (1); 778 } else { 779 g_raid_change_disk_state(disk, 780 G_RAID_DISK_S_STALE); 781 return (0); 782 } 783 } 784 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data; 785 disk_pos = oldpd->pd_disk_pos; 786 resurrection = 1; 787 } 788 789 if (olddisk == NULL) { 790 /* Find placeholder by position. */ 791 olddisk = g_raid_md_intel_get_disk(sc, disk_pos); 792 if (olddisk == NULL) 793 panic("No disk at position %d!", disk_pos); 794 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) { 795 G_RAID_DEBUG1(1, sc, "More then one disk for pos %d", 796 disk_pos); 797 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE); 798 return (0); 799 } 800 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data; 801 } 802 803 /* Replace failed disk or placeholder with new disk. */ 804 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) { 805 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next); 806 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 807 sd->sd_disk = disk; 808 } 809 oldpd->pd_disk_pos = -2; 810 pd->pd_disk_pos = disk_pos; 811 812 /* If it was placeholder -- destroy it. */ 813 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) { 814 g_raid_destroy_disk(olddisk); 815 } else { 816 /* Otherwise, make it STALE_FAILED. */ 817 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED); 818 /* Update global metadata just in case. */ 819 memcpy(&meta->disk[disk_pos], &pd->pd_disk_meta, 820 sizeof(struct intel_raid_disk)); 821 } 822 823 /* Welcome the new disk. */ 824 if (resurrection) 825 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); 826 else if (meta->disk[disk_pos].flags & INTEL_F_FAILED) 827 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED); 828 else if (meta->disk[disk_pos].flags & INTEL_F_SPARE) 829 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE); 830 else 831 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); 832 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 833 mvol = intel_get_volume(meta, 834 (uintptr_t)(sd->sd_volume->v_md_data)); 835 mmap0 = intel_get_map(mvol, 0); 836 if (mvol->migr_state) 837 mmap1 = intel_get_map(mvol, 1); 838 else 839 mmap1 = mmap0; 840 841 if (resurrection) { 842 /* Stale disk, almost same as new. */ 843 g_raid_change_subdisk_state(sd, 844 G_RAID_SUBDISK_S_NEW); 845 } else if (meta->disk[disk_pos].flags & INTEL_F_FAILED) { 846 /* Failed disk, almost useless. */ 847 g_raid_change_subdisk_state(sd, 848 G_RAID_SUBDISK_S_FAILED); 849 } else if (mvol->migr_state == 0) { 850 if (mmap0->status == INTEL_S_UNINITIALIZED) { 851 /* Freshly created uninitialized volume. */ 852 g_raid_change_subdisk_state(sd, 853 G_RAID_SUBDISK_S_UNINITIALIZED); 854 } else if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 855 /* Freshly inserted disk. */ 856 g_raid_change_subdisk_state(sd, 857 G_RAID_SUBDISK_S_NEW); 858 } else if (mvol->dirty) { 859 /* Dirty volume (unclean shutdown). */ 860 g_raid_change_subdisk_state(sd, 861 G_RAID_SUBDISK_S_STALE); 862 } else { 863 /* Up to date disk. */ 864 g_raid_change_subdisk_state(sd, 865 G_RAID_SUBDISK_S_ACTIVE); 866 } 867 } else if (mvol->migr_type == INTEL_MT_INIT || 868 mvol->migr_type == INTEL_MT_REBUILD) { 869 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 870 /* Freshly inserted disk. */ 871 g_raid_change_subdisk_state(sd, 872 G_RAID_SUBDISK_S_NEW); 873 } else if (mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 874 /* Rebuilding disk. */ 875 g_raid_change_subdisk_state(sd, 876 G_RAID_SUBDISK_S_REBUILD); 877 if (mvol->dirty) { 878 sd->sd_rebuild_pos = 0; 879 } else { 880 sd->sd_rebuild_pos = 881 intel_get_vol_curr_migr_unit(mvol) * 882 sd->sd_volume->v_strip_size * 883 mmap0->total_domains; 884 } 885 } else if (mvol->dirty) { 886 /* Dirty volume (unclean shutdown). */ 887 g_raid_change_subdisk_state(sd, 888 G_RAID_SUBDISK_S_STALE); 889 } else { 890 /* Up to date disk. */ 891 g_raid_change_subdisk_state(sd, 892 G_RAID_SUBDISK_S_ACTIVE); 893 } 894 } else if (mvol->migr_type == INTEL_MT_VERIFY || 895 mvol->migr_type == INTEL_MT_REPAIR) { 896 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 897 /* Freshly inserted disk. */ 898 g_raid_change_subdisk_state(sd, 899 G_RAID_SUBDISK_S_NEW); 900 } else if (mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) { 901 /* Resyncing disk. */ 902 g_raid_change_subdisk_state(sd, 903 G_RAID_SUBDISK_S_RESYNC); 904 if (mvol->dirty) { 905 sd->sd_rebuild_pos = 0; 906 } else { 907 sd->sd_rebuild_pos = 908 intel_get_vol_curr_migr_unit(mvol) * 909 sd->sd_volume->v_strip_size * 910 mmap0->total_domains; 911 } 912 } else if (mvol->dirty) { 913 /* Dirty volume (unclean shutdown). */ 914 g_raid_change_subdisk_state(sd, 915 G_RAID_SUBDISK_S_STALE); 916 } else { 917 /* Up to date disk. */ 918 g_raid_change_subdisk_state(sd, 919 G_RAID_SUBDISK_S_ACTIVE); 920 } 921 } 922 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 923 G_RAID_EVENT_SUBDISK); 924 } 925 926 /* Update status of our need for spare. */ 927 if (mdi->mdio_started) { 928 mdi->mdio_incomplete = 929 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < 930 meta->total_disks); 931 } 932 933 return (resurrection); 934} 935 936static void 937g_disk_md_intel_retaste(void *arg, int pending) 938{ 939 940 G_RAID_DEBUG(1, "Array is not complete, trying to retaste."); 941 g_retaste(&g_raid_class); 942 free(arg, M_MD_INTEL); 943} 944 945static void 946g_raid_md_intel_refill(struct g_raid_softc *sc) 947{ 948 struct g_raid_md_object *md; 949 struct g_raid_md_intel_object *mdi; 950 struct intel_raid_conf *meta; 951 struct g_raid_disk *disk; 952 struct task *task; 953 int update, na; 954 955 md = sc->sc_md; 956 mdi = (struct g_raid_md_intel_object *)md; 957 meta = mdi->mdio_meta; 958 update = 0; 959 do { 960 /* Make sure we miss anything. */ 961 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE); 962 if (na == meta->total_disks) 963 break; 964 965 G_RAID_DEBUG1(1, md->mdo_softc, 966 "Array is not complete (%d of %d), " 967 "trying to refill.", na, meta->total_disks); 968 969 /* Try to get use some of STALE disks. */ 970 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 971 if (disk->d_state == G_RAID_DISK_S_STALE) { 972 update += g_raid_md_intel_start_disk(disk); 973 if (disk->d_state == G_RAID_DISK_S_ACTIVE) 974 break; 975 } 976 } 977 if (disk != NULL) 978 continue; 979 980 /* Try to get use some of SPARE disks. */ 981 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 982 if (disk->d_state == G_RAID_DISK_S_SPARE) { 983 update += g_raid_md_intel_start_disk(disk); 984 if (disk->d_state == G_RAID_DISK_S_ACTIVE) 985 break; 986 } 987 } 988 } while (disk != NULL); 989 990 /* Write new metadata if we changed something. */ 991 if (update) { 992 g_raid_md_write_intel(md, NULL, NULL, NULL); 993 meta = mdi->mdio_meta; 994 } 995 996 /* Update status of our need for spare. */ 997 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < 998 meta->total_disks); 999 1000 /* Request retaste hoping to find spare. */ 1001 if (mdi->mdio_incomplete) { 1002 task = malloc(sizeof(struct task), 1003 M_MD_INTEL, M_WAITOK | M_ZERO); 1004 TASK_INIT(task, 0, g_disk_md_intel_retaste, task); 1005 taskqueue_enqueue(taskqueue_swi, task); 1006 } 1007} 1008 1009static void 1010g_raid_md_intel_start(struct g_raid_softc *sc) 1011{ 1012 struct g_raid_md_object *md; 1013 struct g_raid_md_intel_object *mdi; 1014 struct g_raid_md_intel_perdisk *pd; 1015 struct intel_raid_conf *meta; 1016 struct intel_raid_vol *mvol; 1017 struct intel_raid_map *mmap; 1018 struct g_raid_volume *vol; 1019 struct g_raid_subdisk *sd; 1020 struct g_raid_disk *disk; 1021 int i, j, disk_pos; 1022 1023 md = sc->sc_md; 1024 mdi = (struct g_raid_md_intel_object *)md; 1025 meta = mdi->mdio_meta; 1026 1027 /* Create volumes and subdisks. */ 1028 for (i = 0; i < meta->total_volumes; i++) { 1029 mvol = intel_get_volume(meta, i); 1030 mmap = intel_get_map(mvol, 0); 1031 vol = g_raid_create_volume(sc, mvol->name, -1); 1032 vol->v_md_data = (void *)(intptr_t)i; 1033 if (mmap->type == INTEL_T_RAID0) 1034 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0; 1035 else if (mmap->type == INTEL_T_RAID1 && 1036 mmap->total_domains >= 2 && 1037 mmap->total_domains <= mmap->total_disks) { 1038 /* Assume total_domains is correct. */ 1039 if (mmap->total_domains == mmap->total_disks) 1040 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1; 1041 else 1042 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E; 1043 } else if (mmap->type == INTEL_T_RAID1) { 1044 /* total_domains looks wrong. */ 1045 if (mmap->total_disks <= 2) 1046 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1; 1047 else 1048 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E; 1049 } else if (mmap->type == INTEL_T_RAID5) 1050 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5; 1051 else 1052 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN; 1053 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE; 1054 vol->v_strip_size = (u_int)mmap->strip_sectors * 512; //ZZZ 1055 vol->v_disks_count = mmap->total_disks; 1056 vol->v_mediasize = (off_t)mvol->total_sectors * 512; //ZZZ 1057 vol->v_sectorsize = 512; //ZZZ 1058 for (j = 0; j < vol->v_disks_count; j++) { 1059 sd = &vol->v_subdisks[j]; 1060 sd->sd_offset = intel_get_map_offset(mmap) * 512; //ZZZ 1061 sd->sd_size = intel_get_map_disk_sectors(mmap) * 512; //ZZZ 1062 } 1063 g_raid_start_volume(vol); 1064 } 1065 1066 /* Create disk placeholders to store data for later writing. */ 1067 for (disk_pos = 0; disk_pos < meta->total_disks; disk_pos++) { 1068 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO); 1069 pd->pd_disk_pos = disk_pos; 1070 pd->pd_disk_meta = meta->disk[disk_pos]; 1071 disk = g_raid_create_disk(sc); 1072 disk->d_md_data = (void *)pd; 1073 disk->d_state = G_RAID_DISK_S_OFFLINE; 1074 for (i = 0; i < meta->total_volumes; i++) { 1075 mvol = intel_get_volume(meta, i); 1076 mmap = intel_get_map(mvol, 0); 1077 for (j = 0; j < mmap->total_disks; j++) { 1078 if ((mmap->disk_idx[j] & INTEL_DI_IDX) == disk_pos) 1079 break; 1080 } 1081 if (j == mmap->total_disks) 1082 continue; 1083 vol = g_raid_md_intel_get_volume(sc, i); 1084 sd = &vol->v_subdisks[j]; 1085 sd->sd_disk = disk; 1086 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 1087 } 1088 } 1089 1090 /* Make all disks found till the moment take their places. */ 1091 do { 1092 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1093 if (disk->d_state == G_RAID_DISK_S_NONE) { 1094 g_raid_md_intel_start_disk(disk); 1095 break; 1096 } 1097 } 1098 } while (disk != NULL); 1099 1100 mdi->mdio_started = 1; 1101 G_RAID_DEBUG1(0, sc, "Array started."); 1102 g_raid_md_write_intel(md, NULL, NULL, NULL); 1103 1104 /* Pickup any STALE/SPARE disks to refill array if needed. */ 1105 g_raid_md_intel_refill(sc); 1106 1107 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 1108 g_raid_event_send(vol, G_RAID_VOLUME_E_START, 1109 G_RAID_EVENT_VOLUME); 1110 } 1111 1112 callout_stop(&mdi->mdio_start_co); 1113 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount); 1114 root_mount_rel(mdi->mdio_rootmount); 1115 mdi->mdio_rootmount = NULL; 1116} 1117 1118static void 1119g_raid_md_intel_new_disk(struct g_raid_disk *disk) 1120{ 1121 struct g_raid_softc *sc; 1122 struct g_raid_md_object *md; 1123 struct g_raid_md_intel_object *mdi; 1124 struct intel_raid_conf *pdmeta; 1125 struct g_raid_md_intel_perdisk *pd; 1126 1127 sc = disk->d_softc; 1128 md = sc->sc_md; 1129 mdi = (struct g_raid_md_intel_object *)md; 1130 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 1131 pdmeta = pd->pd_meta; 1132 1133 if (mdi->mdio_started) { 1134 if (g_raid_md_intel_start_disk(disk)) 1135 g_raid_md_write_intel(md, NULL, NULL, NULL); 1136 } else { 1137 /* If we haven't started yet - check metadata freshness. */ 1138 if (mdi->mdio_meta == NULL || 1139 ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) { 1140 G_RAID_DEBUG1(1, sc, "Newer disk"); 1141 if (mdi->mdio_meta != NULL) 1142 free(mdi->mdio_meta, M_MD_INTEL); 1143 mdi->mdio_meta = intel_meta_copy(pdmeta); 1144 mdi->mdio_generation = mdi->mdio_meta->generation; 1145 mdi->mdio_disks_present = 1; 1146 } else if (pdmeta->generation == mdi->mdio_generation) { 1147 mdi->mdio_disks_present++; 1148 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)", 1149 mdi->mdio_disks_present, 1150 mdi->mdio_meta->total_disks); 1151 } else { 1152 G_RAID_DEBUG1(1, sc, "Older disk"); 1153 } 1154 /* If we collected all needed disks - start array. */ 1155 if (mdi->mdio_disks_present == mdi->mdio_meta->total_disks) 1156 g_raid_md_intel_start(sc); 1157 } 1158} 1159 1160static void 1161g_raid_intel_go(void *arg) 1162{ 1163 struct g_raid_softc *sc; 1164 struct g_raid_md_object *md; 1165 struct g_raid_md_intel_object *mdi; 1166 1167 sc = arg; 1168 md = sc->sc_md; 1169 mdi = (struct g_raid_md_intel_object *)md; 1170 if (!mdi->mdio_started) { 1171 G_RAID_DEBUG1(0, sc, "Force array start due to timeout."); 1172 g_raid_event_send(sc, G_RAID_NODE_E_START, 0); 1173 } 1174} 1175 1176static int 1177g_raid_md_create_intel(struct g_raid_md_object *md, struct g_class *mp, 1178 struct g_geom **gp) 1179{ 1180 struct g_raid_softc *sc; 1181 struct g_raid_md_intel_object *mdi; 1182 char name[16]; 1183 1184 mdi = (struct g_raid_md_intel_object *)md; 1185 mdi->mdio_config_id = arc4random(); 1186 mdi->mdio_generation = 0; 1187 snprintf(name, sizeof(name), "Intel-%08x", mdi->mdio_config_id); 1188 sc = g_raid_create_node(mp, name, md); 1189 if (sc == NULL) 1190 return (G_RAID_MD_TASTE_FAIL); 1191 md->mdo_softc = sc; 1192 *gp = sc->sc_geom; 1193 return (G_RAID_MD_TASTE_NEW); 1194} 1195 1196/* 1197 * Return the last N characters of the serial label. The Linux and 1198 * ataraid(7) code always uses the last 16 characters of the label to 1199 * store into the Intel meta format. Generalize this to N characters 1200 * since that's easy. Labels can be up to 20 characters for SATA drives 1201 * and up 251 characters for SAS drives. Since intel controllers don't 1202 * support SAS drives, just stick with the SATA limits for stack friendliness. 1203 */ 1204static int 1205g_raid_md_get_label(struct g_consumer *cp, char *serial, int serlen) 1206{ 1207 char serial_buffer[24]; 1208 int len, error; 1209 1210 len = sizeof(serial_buffer); 1211 error = g_io_getattr("GEOM::ident", cp, &len, serial_buffer); 1212 if (error != 0) 1213 return (error); 1214 len = strlen(serial_buffer); 1215 if (len > serlen) 1216 len -= serlen; 1217 else 1218 len = 0; 1219 strncpy(serial, serial_buffer + len, serlen); 1220 return (0); 1221} 1222 1223static int 1224g_raid_md_taste_intel(struct g_raid_md_object *md, struct g_class *mp, 1225 struct g_consumer *cp, struct g_geom **gp) 1226{ 1227 struct g_consumer *rcp; 1228 struct g_provider *pp; 1229 struct g_raid_md_intel_object *mdi, *mdi1; 1230 struct g_raid_softc *sc; 1231 struct g_raid_disk *disk; 1232 struct intel_raid_conf *meta; 1233 struct g_raid_md_intel_perdisk *pd; 1234 struct g_geom *geom; 1235 int error, disk_pos, result, spare, len; 1236 char serial[INTEL_SERIAL_LEN]; 1237 char name[16]; 1238 uint16_t vendor; 1239 1240 G_RAID_DEBUG(1, "Tasting Intel on %s", cp->provider->name); 1241 mdi = (struct g_raid_md_intel_object *)md; 1242 pp = cp->provider; 1243 1244 /* Read metadata from device. */ 1245 meta = NULL; 1246 vendor = 0xffff; 1247 disk_pos = 0; 1248 if (g_access(cp, 1, 0, 0) != 0) 1249 return (G_RAID_MD_TASTE_FAIL); 1250 g_topology_unlock(); 1251 error = g_raid_md_get_label(cp, serial, sizeof(serial)); 1252 if (error != 0) { 1253 G_RAID_DEBUG(1, "Cannot get serial number from %s (error=%d).", 1254 pp->name, error); 1255 goto fail2; 1256 } 1257 len = 2; 1258 if (pp->geom->rank == 1) 1259 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor); 1260 meta = intel_meta_read(cp); 1261 g_topology_lock(); 1262 g_access(cp, -1, 0, 0); 1263 if (meta == NULL) { 1264 if (g_raid_aggressive_spare) { 1265 if (vendor != 0x8086) { 1266 G_RAID_DEBUG(1, 1267 "Intel vendor mismatch 0x%04x != 0x8086", 1268 vendor); 1269 } else { 1270 G_RAID_DEBUG(1, 1271 "No Intel metadata, forcing spare."); 1272 spare = 2; 1273 goto search; 1274 } 1275 } 1276 return (G_RAID_MD_TASTE_FAIL); 1277 } 1278 1279 /* Check this disk position in obtained metadata. */ 1280 disk_pos = intel_meta_find_disk(meta, serial); 1281 if (disk_pos < 0) { 1282 G_RAID_DEBUG(1, "Intel serial '%s' not found", serial); 1283 goto fail1; 1284 } 1285 if (intel_get_disk_sectors(&meta->disk[disk_pos]) != 1286 (pp->mediasize / pp->sectorsize)) { 1287 G_RAID_DEBUG(1, "Intel size mismatch %ju != %ju", 1288 intel_get_disk_sectors(&meta->disk[disk_pos]), 1289 (off_t)(pp->mediasize / pp->sectorsize)); 1290 goto fail1; 1291 } 1292 1293 /* Metadata valid. Print it. */ 1294 g_raid_md_intel_print(meta); 1295 G_RAID_DEBUG(1, "Intel disk position %d", disk_pos); 1296 spare = meta->disk[disk_pos].flags & INTEL_F_SPARE; 1297 1298search: 1299 /* Search for matching node. */ 1300 sc = NULL; 1301 mdi1 = NULL; 1302 LIST_FOREACH(geom, &mp->geom, geom) { 1303 sc = geom->softc; 1304 if (sc == NULL) 1305 continue; 1306 if (sc->sc_stopping != 0) 1307 continue; 1308 if (sc->sc_md->mdo_class != md->mdo_class) 1309 continue; 1310 mdi1 = (struct g_raid_md_intel_object *)sc->sc_md; 1311 if (spare) { 1312 if (mdi1->mdio_incomplete) 1313 break; 1314 } else { 1315 if (mdi1->mdio_config_id == meta->config_id) 1316 break; 1317 } 1318 } 1319 1320 /* Found matching node. */ 1321 if (geom != NULL) { 1322 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name); 1323 result = G_RAID_MD_TASTE_EXISTING; 1324 1325 } else if (spare) { /* Not found needy node -- left for later. */ 1326 G_RAID_DEBUG(1, "Spare is not needed at this time"); 1327 goto fail1; 1328 1329 } else { /* Not found matching node -- create one. */ 1330 result = G_RAID_MD_TASTE_NEW; 1331 mdi->mdio_config_id = meta->config_id; 1332 snprintf(name, sizeof(name), "Intel-%08x", meta->config_id); 1333 sc = g_raid_create_node(mp, name, md); 1334 md->mdo_softc = sc; 1335 geom = sc->sc_geom; 1336 callout_init(&mdi->mdio_start_co, 1); 1337 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz, 1338 g_raid_intel_go, sc); 1339 mdi->mdio_rootmount = root_mount_hold("GRAID-Intel"); 1340 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount); 1341 } 1342 1343 rcp = g_new_consumer(geom); 1344 g_attach(rcp, pp); 1345 if (g_access(rcp, 1, 1, 1) != 0) 1346 ; //goto fail1; 1347 1348 g_topology_unlock(); 1349 sx_xlock(&sc->sc_lock); 1350 1351 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO); 1352 pd->pd_meta = meta; 1353 pd->pd_disk_pos = -1; 1354 if (spare == 2) { 1355 memcpy(&pd->pd_disk_meta.serial[0], serial, INTEL_SERIAL_LEN); 1356 intel_set_disk_sectors(&pd->pd_disk_meta, 1357 pp->mediasize / pp->sectorsize); 1358 pd->pd_disk_meta.id = 0; 1359 pd->pd_disk_meta.flags = INTEL_F_SPARE; 1360 } else { 1361 pd->pd_disk_meta = meta->disk[disk_pos]; 1362 } 1363 disk = g_raid_create_disk(sc); 1364 disk->d_md_data = (void *)pd; 1365 disk->d_consumer = rcp; 1366 rcp->private = disk; 1367 1368 /* Read kernel dumping information. */ 1369 disk->d_kd.offset = 0; 1370 disk->d_kd.length = OFF_MAX; 1371 len = sizeof(disk->d_kd); 1372 error = g_io_getattr("GEOM::kerneldump", rcp, &len, &disk->d_kd); 1373 if (disk->d_kd.di.dumper == NULL) 1374 G_RAID_DEBUG1(2, sc, "Dumping not supported by %s: %d.", 1375 rcp->provider->name, error); 1376 1377 g_raid_md_intel_new_disk(disk); 1378 1379 sx_xunlock(&sc->sc_lock); 1380 g_topology_lock(); 1381 *gp = geom; 1382 return (result); 1383fail2: 1384 g_topology_lock(); 1385 g_access(cp, -1, 0, 0); 1386fail1: 1387 free(meta, M_MD_INTEL); 1388 return (G_RAID_MD_TASTE_FAIL); 1389} 1390 1391static int 1392g_raid_md_event_intel(struct g_raid_md_object *md, 1393 struct g_raid_disk *disk, u_int event) 1394{ 1395 struct g_raid_softc *sc; 1396 struct g_raid_subdisk *sd; 1397 struct g_raid_md_intel_object *mdi; 1398 struct g_raid_md_intel_perdisk *pd; 1399 1400 sc = md->mdo_softc; 1401 mdi = (struct g_raid_md_intel_object *)md; 1402 if (disk == NULL) { 1403 switch (event) { 1404 case G_RAID_NODE_E_START: 1405 if (!mdi->mdio_started) 1406 g_raid_md_intel_start(sc); 1407 return (0); 1408 } 1409 return (-1); 1410 } 1411 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 1412 switch (event) { 1413 case G_RAID_DISK_E_DISCONNECTED: 1414 /* If disk was assigned, just update statuses. */ 1415 if (pd->pd_disk_pos >= 0) { 1416 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1417 if (disk->d_consumer) { 1418 g_raid_kill_consumer(sc, disk->d_consumer); 1419 disk->d_consumer = NULL; 1420 } 1421 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 1422 g_raid_change_subdisk_state(sd, 1423 G_RAID_SUBDISK_S_NONE); 1424 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED, 1425 G_RAID_EVENT_SUBDISK); 1426 } 1427 } else { 1428 /* Otherwise -- delete. */ 1429 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE); 1430 g_raid_destroy_disk(disk); 1431 } 1432 1433 /* Write updated metadata to all disks. */ 1434 g_raid_md_write_intel(md, NULL, NULL, NULL); 1435 1436 /* Check if anything left except placeholders. */ 1437 if (g_raid_ndisks(sc, -1) == 1438 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 1439 g_raid_destroy_node(sc, 0); 1440 else 1441 g_raid_md_intel_refill(sc); 1442 return (0); 1443 } 1444 return (-2); 1445} 1446 1447static int 1448g_raid_md_ctl_intel(struct g_raid_md_object *md, 1449 struct gctl_req *req) 1450{ 1451 struct g_raid_softc *sc; 1452 struct g_raid_volume *vol, *vol1; 1453 struct g_raid_subdisk *sd; 1454 struct g_raid_disk *disk; 1455 struct g_raid_md_intel_object *mdi; 1456 struct g_raid_md_intel_perdisk *pd; 1457 struct g_consumer *cp; 1458 struct g_provider *pp; 1459 char arg[16], serial[INTEL_SERIAL_LEN]; 1460 const char *verb, *volname, *levelname, *diskname; 1461 char *tmp; 1462 int *nargs, *force; 1463 off_t off, size, sectorsize, strip, disk_sectors; 1464 intmax_t *sizearg, *striparg; 1465 int numdisks, i, len, level, qual, update; 1466 int error; 1467 1468 sc = md->mdo_softc; 1469 mdi = (struct g_raid_md_intel_object *)md; 1470 verb = gctl_get_param(req, "verb", NULL); 1471 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs)); 1472 error = 0; 1473 if (strcmp(verb, "label") == 0) { 1474 1475 if (*nargs < 4) { 1476 gctl_error(req, "Invalid number of arguments."); 1477 return (-1); 1478 } 1479 volname = gctl_get_asciiparam(req, "arg1"); 1480 if (volname == NULL) { 1481 gctl_error(req, "No volume name."); 1482 return (-2); 1483 } 1484 levelname = gctl_get_asciiparam(req, "arg2"); 1485 if (levelname == NULL) { 1486 gctl_error(req, "No RAID level."); 1487 return (-3); 1488 } 1489 if (g_raid_volume_str2level(levelname, &level, &qual)) { 1490 gctl_error(req, "Unknown RAID level '%s'.", levelname); 1491 return (-4); 1492 } 1493 numdisks = *nargs - 3; 1494 force = gctl_get_paraml(req, "force", sizeof(*force)); 1495 if (!g_raid_md_intel_supported(level, qual, numdisks, 1496 force ? *force : 0)) { 1497 gctl_error(req, "Unsupported RAID level " 1498 "(0x%02x/0x%02x), or number of disks (%d).", 1499 level, qual, numdisks); 1500 return (-5); 1501 } 1502 1503 /* Search for disks, connect them and probe. */ 1504 size = 0x7fffffffffffffffllu; 1505 sectorsize = 0; 1506 for (i = 0; i < numdisks; i++) { 1507 snprintf(arg, sizeof(arg), "arg%d", i + 3); 1508 diskname = gctl_get_asciiparam(req, arg); 1509 if (diskname == NULL) { 1510 gctl_error(req, "No disk name (%s).", arg); 1511 error = -6; 1512 break; 1513 } 1514 if (strcmp(diskname, "NONE") == 0) { 1515 cp = NULL; 1516 pp = NULL; 1517 } else { 1518 g_topology_lock(); 1519 cp = g_raid_open_consumer(sc, diskname); 1520 if (cp == NULL) { 1521 gctl_error(req, "Can't open disk '%s'.", 1522 diskname); 1523 g_topology_unlock(); 1524 error = -7; 1525 break; 1526 } 1527 pp = cp->provider; 1528 } 1529 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO); 1530 pd->pd_disk_pos = i; 1531 disk = g_raid_create_disk(sc); 1532 disk->d_md_data = (void *)pd; 1533 disk->d_consumer = cp; 1534 if (cp == NULL) { 1535 strcpy(&pd->pd_disk_meta.serial[0], "NONE"); 1536 pd->pd_disk_meta.id = 0xffffffff; 1537 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED; 1538 continue; 1539 } 1540 cp->private = disk; 1541 g_topology_unlock(); 1542 1543 error = g_raid_md_get_label(cp, 1544 &pd->pd_disk_meta.serial[0], INTEL_SERIAL_LEN); 1545 if (error != 0) { 1546 gctl_error(req, 1547 "Can't get serial for provider '%s'.", 1548 diskname); 1549 error = -8; 1550 break; 1551 } 1552 1553 /* Read kernel dumping information. */ 1554 disk->d_kd.offset = 0; 1555 disk->d_kd.length = OFF_MAX; 1556 len = sizeof(disk->d_kd); 1557 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd); 1558 if (disk->d_kd.di.dumper == NULL) 1559 G_RAID_DEBUG1(2, sc, 1560 "Dumping not supported by %s.", 1561 cp->provider->name); 1562 1563 intel_set_disk_sectors(&pd->pd_disk_meta, 1564 pp->mediasize / pp->sectorsize); 1565 if (size > pp->mediasize) 1566 size = pp->mediasize; 1567 if (sectorsize < pp->sectorsize) 1568 sectorsize = pp->sectorsize; 1569 pd->pd_disk_meta.id = 0; 1570 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED | INTEL_F_ONLINE; 1571 } 1572 if (error != 0) 1573 return (error); 1574 1575 if (sectorsize <= 0) { 1576 gctl_error(req, "Can't get sector size."); 1577 return (-8); 1578 } 1579 1580 /* Reserve some space for metadata. */ 1581 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize; 1582 1583 /* Handle size argument. */ 1584 len = sizeof(*sizearg); 1585 sizearg = gctl_get_param(req, "size", &len); 1586 if (sizearg != NULL && len == sizeof(*sizearg) && 1587 *sizearg > 0) { 1588 if (*sizearg > size) { 1589 gctl_error(req, "Size too big %lld > %lld.", 1590 (long long)*sizearg, (long long)size); 1591 return (-9); 1592 } 1593 size = *sizearg; 1594 } 1595 1596 /* Handle strip argument. */ 1597 strip = 131072; 1598 len = sizeof(*striparg); 1599 striparg = gctl_get_param(req, "strip", &len); 1600 if (striparg != NULL && len == sizeof(*striparg) && 1601 *striparg > 0) { 1602 if (*striparg < sectorsize) { 1603 gctl_error(req, "Strip size too small."); 1604 return (-10); 1605 } 1606 if (*striparg % sectorsize != 0) { 1607 gctl_error(req, "Incorrect strip size."); 1608 return (-11); 1609 } 1610 if (strip > 65535 * sectorsize) { 1611 gctl_error(req, "Strip size too big."); 1612 return (-12); 1613 } 1614 strip = *striparg; 1615 } 1616 1617 /* Round size down to strip or sector. */ 1618 if (level == G_RAID_VOLUME_RL_RAID1) 1619 size -= (size % sectorsize); 1620 else if (level == G_RAID_VOLUME_RL_RAID1E && 1621 (numdisks & 1) != 0) 1622 size -= (size % (2 * strip)); 1623 else 1624 size -= (size % strip); 1625 if (size <= 0) { 1626 gctl_error(req, "Size too small."); 1627 return (-13); 1628 } 1629 1630 /* We have all we need, create things: volume, ... */ 1631 mdi->mdio_started = 1; 1632 vol = g_raid_create_volume(sc, volname, -1); 1633 vol->v_md_data = (void *)(intptr_t)0; 1634 vol->v_raid_level = level; 1635 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE; 1636 vol->v_strip_size = strip; 1637 vol->v_disks_count = numdisks; 1638 if (level == G_RAID_VOLUME_RL_RAID0) 1639 vol->v_mediasize = size * numdisks; 1640 else if (level == G_RAID_VOLUME_RL_RAID1) 1641 vol->v_mediasize = size; 1642 else if (level == G_RAID_VOLUME_RL_RAID5) 1643 vol->v_mediasize = size * (numdisks - 1); 1644 else { /* RAID1E */ 1645 vol->v_mediasize = ((size * numdisks) / strip / 2) * 1646 strip; 1647 } 1648 vol->v_sectorsize = sectorsize; 1649 g_raid_start_volume(vol); 1650 1651 /* , and subdisks. */ 1652 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1653 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 1654 sd = &vol->v_subdisks[pd->pd_disk_pos]; 1655 sd->sd_disk = disk; 1656 sd->sd_offset = 0; 1657 sd->sd_size = size; 1658 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 1659 if (sd->sd_disk->d_consumer != NULL) { 1660 g_raid_change_disk_state(disk, 1661 G_RAID_DISK_S_ACTIVE); 1662 g_raid_change_subdisk_state(sd, 1663 G_RAID_SUBDISK_S_ACTIVE); 1664 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 1665 G_RAID_EVENT_SUBDISK); 1666 } else { 1667 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1668 } 1669 } 1670 1671 /* Write metadata based on created entities. */ 1672 G_RAID_DEBUG1(0, sc, "Array started."); 1673 g_raid_md_write_intel(md, NULL, NULL, NULL); 1674 1675 /* Pickup any STALE/SPARE disks to refill array if needed. */ 1676 g_raid_md_intel_refill(sc); 1677 1678 g_raid_event_send(vol, G_RAID_VOLUME_E_START, 1679 G_RAID_EVENT_VOLUME); 1680 return (0); 1681 } 1682 if (strcmp(verb, "add") == 0) { 1683 1684 if (*nargs != 3) { 1685 gctl_error(req, "Invalid number of arguments."); 1686 return (-1); 1687 } 1688 volname = gctl_get_asciiparam(req, "arg1"); 1689 if (volname == NULL) { 1690 gctl_error(req, "No volume name."); 1691 return (-2); 1692 } 1693 levelname = gctl_get_asciiparam(req, "arg2"); 1694 if (levelname == NULL) { 1695 gctl_error(req, "No RAID level."); 1696 return (-3); 1697 } 1698 if (g_raid_volume_str2level(levelname, &level, &qual)) { 1699 gctl_error(req, "Unknown RAID level '%s'.", levelname); 1700 return (-4); 1701 } 1702 1703 /* Look for existing volumes. */ 1704 i = 0; 1705 vol1 = NULL; 1706 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 1707 vol1 = vol; 1708 i++; 1709 } 1710 if (i > 1) { 1711 gctl_error(req, "Maximum two volumes supported."); 1712 return (-6); 1713 } 1714 if (vol1 == NULL) { 1715 gctl_error(req, "At least one volume must exist."); 1716 return (-7); 1717 } 1718 1719 numdisks = vol1->v_disks_count; 1720 force = gctl_get_paraml(req, "force", sizeof(*force)); 1721 if (!g_raid_md_intel_supported(level, qual, numdisks, 1722 force ? *force : 0)) { 1723 gctl_error(req, "Unsupported RAID level " 1724 "(0x%02x/0x%02x), or number of disks (%d).", 1725 level, qual, numdisks); 1726 return (-5); 1727 } 1728 1729 /* Collect info about present disks. */ 1730 size = 0x7fffffffffffffffllu; 1731 sectorsize = 512; 1732 for (i = 0; i < numdisks; i++) { 1733 disk = vol1->v_subdisks[i].sd_disk; 1734 pd = (struct g_raid_md_intel_perdisk *) 1735 disk->d_md_data; 1736 disk_sectors = 1737 intel_get_disk_sectors(&pd->pd_disk_meta); 1738 1739 if (disk_sectors * 512 < size) 1740 size = disk_sectors * 512; 1741 if (disk->d_consumer != NULL && 1742 disk->d_consumer->provider != NULL && 1743 disk->d_consumer->provider->sectorsize > 1744 sectorsize) { 1745 sectorsize = 1746 disk->d_consumer->provider->sectorsize; 1747 } 1748 } 1749 1750 /* Reserve some space for metadata. */ 1751 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize; 1752 1753 /* Decide insert before or after. */ 1754 sd = &vol1->v_subdisks[0]; 1755 if (sd->sd_offset > 1756 size - (sd->sd_offset + sd->sd_size)) { 1757 off = 0; 1758 size = sd->sd_offset; 1759 } else { 1760 off = sd->sd_offset + sd->sd_size; 1761 size = size - (sd->sd_offset + sd->sd_size); 1762 } 1763 1764 /* Handle strip argument. */ 1765 strip = 131072; 1766 len = sizeof(*striparg); 1767 striparg = gctl_get_param(req, "strip", &len); 1768 if (striparg != NULL && len == sizeof(*striparg) && 1769 *striparg > 0) { 1770 if (*striparg < sectorsize) { 1771 gctl_error(req, "Strip size too small."); 1772 return (-10); 1773 } 1774 if (*striparg % sectorsize != 0) { 1775 gctl_error(req, "Incorrect strip size."); 1776 return (-11); 1777 } 1778 if (strip > 65535 * sectorsize) { 1779 gctl_error(req, "Strip size too big."); 1780 return (-12); 1781 } 1782 strip = *striparg; 1783 } 1784 1785 /* Round offset up to strip. */ 1786 if (off % strip != 0) { 1787 size -= strip - off % strip; 1788 off += strip - off % strip; 1789 } 1790 1791 /* Handle size argument. */ 1792 len = sizeof(*sizearg); 1793 sizearg = gctl_get_param(req, "size", &len); 1794 if (sizearg != NULL && len == sizeof(*sizearg) && 1795 *sizearg > 0) { 1796 if (*sizearg > size) { 1797 gctl_error(req, "Size too big %lld > %lld.", 1798 (long long)*sizearg, (long long)size); 1799 return (-9); 1800 } 1801 size = *sizearg; 1802 } 1803 1804 /* Round size down to strip or sector. */ 1805 if (level == G_RAID_VOLUME_RL_RAID1) 1806 size -= (size % sectorsize); 1807 else 1808 size -= (size % strip); 1809 if (size <= 0) { 1810 gctl_error(req, "Size too small."); 1811 return (-13); 1812 } 1813 if (size > 0xffffffffllu * sectorsize) { 1814 gctl_error(req, "Size too big."); 1815 return (-14); 1816 } 1817 1818 /* We have all we need, create things: volume, ... */ 1819 vol = g_raid_create_volume(sc, volname, -1); 1820 vol->v_md_data = (void *)(intptr_t)i; 1821 vol->v_raid_level = level; 1822 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE; 1823 vol->v_strip_size = strip; 1824 vol->v_disks_count = numdisks; 1825 if (level == G_RAID_VOLUME_RL_RAID0) 1826 vol->v_mediasize = size * numdisks; 1827 else if (level == G_RAID_VOLUME_RL_RAID1) 1828 vol->v_mediasize = size; 1829 else if (level == G_RAID_VOLUME_RL_RAID5) 1830 vol->v_mediasize = size * (numdisks - 1); 1831 else { /* RAID1E */ 1832 vol->v_mediasize = ((size * numdisks) / strip / 2) * 1833 strip; 1834 } 1835 vol->v_sectorsize = sectorsize; 1836 g_raid_start_volume(vol); 1837 1838 /* , and subdisks. */ 1839 for (i = 0; i < numdisks; i++) { 1840 disk = vol1->v_subdisks[i].sd_disk; 1841 sd = &vol->v_subdisks[i]; 1842 sd->sd_disk = disk; 1843 sd->sd_offset = off; 1844 sd->sd_size = size; 1845 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 1846 if (disk->d_state == G_RAID_DISK_S_ACTIVE) { 1847 g_raid_change_subdisk_state(sd, 1848 G_RAID_SUBDISK_S_ACTIVE); 1849 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 1850 G_RAID_EVENT_SUBDISK); 1851 } 1852 } 1853 1854 /* Write metadata based on created entities. */ 1855 g_raid_md_write_intel(md, NULL, NULL, NULL); 1856 1857 g_raid_event_send(vol, G_RAID_VOLUME_E_START, 1858 G_RAID_EVENT_VOLUME); 1859 return (0); 1860 } 1861 if (strcmp(verb, "delete") == 0) { 1862 1863 /* Full node destruction. */ 1864 if (*nargs == 1) { 1865 /* Check if some volume is still open. */ 1866 force = gctl_get_paraml(req, "force", sizeof(*force)); 1867 if (force != NULL && *force == 0 && 1868 g_raid_nopens(sc) != 0) { 1869 gctl_error(req, "Some volume is still open."); 1870 return (-4); 1871 } 1872 1873 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1874 if (disk->d_consumer) 1875 intel_meta_erase(disk->d_consumer); 1876 } 1877 g_raid_destroy_node(sc, 0); 1878 return (0); 1879 } 1880 1881 /* Destroy specified volume. If it was last - all node. */ 1882 if (*nargs != 2) { 1883 gctl_error(req, "Invalid number of arguments."); 1884 return (-1); 1885 } 1886 volname = gctl_get_asciiparam(req, "arg1"); 1887 if (volname == NULL) { 1888 gctl_error(req, "No volume name."); 1889 return (-2); 1890 } 1891 1892 /* Search for volume. */ 1893 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 1894 if (strcmp(vol->v_name, volname) == 0) 1895 break; 1896 } 1897 if (vol == NULL) { 1898 i = strtol(volname, &tmp, 10); 1899 if (verb != volname && tmp[0] == 0) { 1900 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 1901 if (vol->v_global_id == i) 1902 break; 1903 } 1904 } 1905 } 1906 if (vol == NULL) { 1907 gctl_error(req, "Volume '%s' not found.", volname); 1908 return (-3); 1909 } 1910 1911 /* Check if volume is still open. */ 1912 force = gctl_get_paraml(req, "force", sizeof(*force)); 1913 if (force != NULL && *force == 0 && 1914 vol->v_provider_open != 0) { 1915 gctl_error(req, "Volume is still open."); 1916 return (-4); 1917 } 1918 1919 /* Destroy volume and potentially node. */ 1920 i = 0; 1921 TAILQ_FOREACH(vol1, &sc->sc_volumes, v_next) 1922 i++; 1923 if (i >= 2) { 1924 g_raid_destroy_volume(vol); 1925 g_raid_md_write_intel(md, NULL, NULL, NULL); 1926 } else { 1927 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1928 if (disk->d_consumer) 1929 intel_meta_erase(disk->d_consumer); 1930 } 1931 g_raid_destroy_node(sc, 0); 1932 } 1933 return (0); 1934 } 1935 if (strcmp(verb, "remove") == 0 || 1936 strcmp(verb, "fail") == 0) { 1937 if (*nargs < 2) { 1938 gctl_error(req, "Invalid number of arguments."); 1939 return (-1); 1940 } 1941 for (i = 1; i < *nargs; i++) { 1942 snprintf(arg, sizeof(arg), "arg%d", i); 1943 diskname = gctl_get_asciiparam(req, arg); 1944 if (diskname == NULL) { 1945 gctl_error(req, "No disk name (%s).", arg); 1946 error = -2; 1947 break; 1948 } 1949 if (strncmp(diskname, "/dev/", 5) == 0) 1950 diskname += 5; 1951 1952 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1953 if (disk->d_consumer != NULL && 1954 disk->d_consumer->provider != NULL && 1955 strcmp(disk->d_consumer->provider->name, 1956 diskname) == 0) 1957 break; 1958 } 1959 if (disk == NULL) { 1960 gctl_error(req, "Disk '%s' not found.", 1961 diskname); 1962 error = -3; 1963 break; 1964 } 1965 1966 if (strcmp(verb, "fail") == 0) { 1967 g_raid_md_fail_disk_intel(md, NULL, disk); 1968 continue; 1969 } 1970 1971 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 1972 1973 /* Erase metadata on deleting disk. */ 1974 intel_meta_erase(disk->d_consumer); 1975 1976 /* If disk was assigned, just update statuses. */ 1977 if (pd->pd_disk_pos >= 0) { 1978 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1979 g_raid_kill_consumer(sc, disk->d_consumer); 1980 disk->d_consumer = NULL; 1981 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 1982 g_raid_change_subdisk_state(sd, 1983 G_RAID_SUBDISK_S_NONE); 1984 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED, 1985 G_RAID_EVENT_SUBDISK); 1986 } 1987 } else { 1988 /* Otherwise -- delete. */ 1989 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE); 1990 g_raid_destroy_disk(disk); 1991 } 1992 } 1993 1994 /* Write updated metadata to remaining disks. */ 1995 g_raid_md_write_intel(md, NULL, NULL, NULL); 1996 1997 /* Check if anything left except placeholders. */ 1998 if (g_raid_ndisks(sc, -1) == 1999 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 2000 g_raid_destroy_node(sc, 0); 2001 else 2002 g_raid_md_intel_refill(sc); 2003 return (error); 2004 } 2005 if (strcmp(verb, "insert") == 0) { 2006 if (*nargs < 2) { 2007 gctl_error(req, "Invalid number of arguments."); 2008 return (-1); 2009 } 2010 update = 0; 2011 for (i = 1; i < *nargs; i++) { 2012 /* Get disk name. */ 2013 snprintf(arg, sizeof(arg), "arg%d", i); 2014 diskname = gctl_get_asciiparam(req, arg); 2015 if (diskname == NULL) { 2016 gctl_error(req, "No disk name (%s).", arg); 2017 error = -3; 2018 break; 2019 } 2020 2021 /* Try to find provider with specified name. */ 2022 g_topology_lock(); 2023 cp = g_raid_open_consumer(sc, diskname); 2024 if (cp == NULL) { 2025 gctl_error(req, "Can't open disk '%s'.", 2026 diskname); 2027 g_topology_unlock(); 2028 error = -4; 2029 break; 2030 } 2031 pp = cp->provider; 2032 g_topology_unlock(); 2033 2034 /* Read disk serial. */ 2035 error = g_raid_md_get_label(cp, 2036 &serial[0], INTEL_SERIAL_LEN); 2037 if (error != 0) { 2038 gctl_error(req, 2039 "Can't get serial for provider '%s'.", 2040 diskname); 2041 g_raid_kill_consumer(sc, cp); 2042 error = -7; 2043 break; 2044 } 2045 2046 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO); 2047 pd->pd_disk_pos = -1; 2048 2049 disk = g_raid_create_disk(sc); 2050 disk->d_consumer = cp; 2051 disk->d_md_data = (void *)pd; 2052 cp->private = disk; 2053 2054 /* Read kernel dumping information. */ 2055 disk->d_kd.offset = 0; 2056 disk->d_kd.length = OFF_MAX; 2057 len = sizeof(disk->d_kd); 2058 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd); 2059 if (disk->d_kd.di.dumper == NULL) 2060 G_RAID_DEBUG1(2, sc, 2061 "Dumping not supported by %s.", 2062 cp->provider->name); 2063 2064 memcpy(&pd->pd_disk_meta.serial[0], &serial[0], 2065 INTEL_SERIAL_LEN); 2066 intel_set_disk_sectors(&pd->pd_disk_meta, 2067 pp->mediasize / pp->sectorsize); 2068 pd->pd_disk_meta.id = 0; 2069 pd->pd_disk_meta.flags = INTEL_F_SPARE; 2070 2071 /* Welcome the "new" disk. */ 2072 update += g_raid_md_intel_start_disk(disk); 2073 if (disk->d_state == G_RAID_DISK_S_SPARE) { 2074 intel_meta_write_spare(cp, &pd->pd_disk_meta); 2075 g_raid_destroy_disk(disk); 2076 } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) { 2077 gctl_error(req, "Disk '%s' doesn't fit.", 2078 diskname); 2079 g_raid_destroy_disk(disk); 2080 error = -8; 2081 break; 2082 } 2083 } 2084 2085 /* Write new metadata if we changed something. */ 2086 if (update) 2087 g_raid_md_write_intel(md, NULL, NULL, NULL); 2088 return (error); 2089 } 2090 return (-100); 2091} 2092 2093static int 2094g_raid_md_write_intel(struct g_raid_md_object *md, struct g_raid_volume *tvol, 2095 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk) 2096{ 2097 struct g_raid_softc *sc; 2098 struct g_raid_volume *vol; 2099 struct g_raid_subdisk *sd; 2100 struct g_raid_disk *disk; 2101 struct g_raid_md_intel_object *mdi; 2102 struct g_raid_md_intel_perdisk *pd; 2103 struct intel_raid_conf *meta; 2104 struct intel_raid_vol *mvol; 2105 struct intel_raid_map *mmap0, *mmap1; 2106 off_t sectorsize = 512, pos; 2107 const char *version, *cv; 2108 int vi, sdi, numdisks, len, state, stale; 2109 2110 sc = md->mdo_softc; 2111 mdi = (struct g_raid_md_intel_object *)md; 2112 2113 if (sc->sc_stopping == G_RAID_DESTROY_HARD) 2114 return (0); 2115 2116 /* Bump generation. Newly written metadata may differ from previous. */ 2117 mdi->mdio_generation++; 2118 2119 /* Count number of disks. */ 2120 numdisks = 0; 2121 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 2122 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 2123 if (pd->pd_disk_pos < 0) 2124 continue; 2125 numdisks++; 2126 if (disk->d_state == G_RAID_DISK_S_ACTIVE) { 2127 pd->pd_disk_meta.flags = 2128 INTEL_F_ONLINE | INTEL_F_ASSIGNED; 2129 } else if (disk->d_state == G_RAID_DISK_S_FAILED) { 2130 pd->pd_disk_meta.flags = INTEL_F_FAILED | INTEL_F_ASSIGNED; 2131 } else { 2132 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED; 2133 if (pd->pd_disk_meta.id != 0xffffffff) { 2134 pd->pd_disk_meta.id = 0xffffffff; 2135 len = strlen(pd->pd_disk_meta.serial); 2136 len = min(len, INTEL_SERIAL_LEN - 3); 2137 strcpy(pd->pd_disk_meta.serial + len, ":0"); 2138 } 2139 } 2140 } 2141 2142 /* Fill anchor and disks. */ 2143 meta = malloc(INTEL_MAX_MD_SIZE(numdisks), 2144 M_MD_INTEL, M_WAITOK | M_ZERO); 2145 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1); 2146 meta->config_size = INTEL_MAX_MD_SIZE(numdisks); 2147 meta->config_id = mdi->mdio_config_id; 2148 meta->generation = mdi->mdio_generation; 2149 meta->attributes = INTEL_ATTR_CHECKSUM; 2150 meta->total_disks = numdisks; 2151 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 2152 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 2153 if (pd->pd_disk_pos < 0) 2154 continue; 2155 meta->disk[pd->pd_disk_pos] = pd->pd_disk_meta; 2156 } 2157 2158 /* Fill volumes and maps. */ 2159 vi = 0; 2160 version = INTEL_VERSION_1000; 2161 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) { 2162 if (vol->v_stopping) 2163 continue; 2164 mvol = intel_get_volume(meta, vi); 2165 2166 /* New metadata may have different volumes order. */ 2167 vol->v_md_data = (void *)(intptr_t)vi; 2168 2169 for (sdi = 0; sdi < vol->v_disks_count; sdi++) { 2170 sd = &vol->v_subdisks[sdi]; 2171 if (sd->sd_disk != NULL) 2172 break; 2173 } 2174 if (sdi >= vol->v_disks_count) 2175 panic("No any filled subdisk in volume"); 2176 if (vol->v_mediasize >= 0x20000000000llu) 2177 meta->attributes |= INTEL_ATTR_2TB; 2178 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0) 2179 meta->attributes |= INTEL_ATTR_RAID0; 2180 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) 2181 meta->attributes |= INTEL_ATTR_RAID1; 2182 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5) 2183 meta->attributes |= INTEL_ATTR_RAID5; 2184 else 2185 meta->attributes |= INTEL_ATTR_RAID10; 2186 2187 if (meta->attributes & INTEL_ATTR_2TB) 2188 cv = INTEL_VERSION_1300; 2189// else if (dev->status == DEV_CLONE_N_GO) 2190// cv = INTEL_VERSION_1206; 2191 else if (vol->v_disks_count > 4) 2192 cv = INTEL_VERSION_1204; 2193 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5) 2194 cv = INTEL_VERSION_1202; 2195 else if (vol->v_disks_count > 2) 2196 cv = INTEL_VERSION_1201; 2197 else if (vi > 0) 2198 cv = INTEL_VERSION_1200; 2199 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) 2200 cv = INTEL_VERSION_1100; 2201 else 2202 cv = INTEL_VERSION_1000; 2203 if (strcmp(cv, version) > 0) 2204 version = cv; 2205 2206 strlcpy(&mvol->name[0], vol->v_name, sizeof(mvol->name)); 2207 mvol->total_sectors = vol->v_mediasize / sectorsize; 2208 2209 /* Check for any recovery in progress. */ 2210 state = G_RAID_SUBDISK_S_ACTIVE; 2211 pos = 0x7fffffffffffffffllu; 2212 stale = 0; 2213 for (sdi = 0; sdi < vol->v_disks_count; sdi++) { 2214 sd = &vol->v_subdisks[sdi]; 2215 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD) 2216 state = G_RAID_SUBDISK_S_REBUILD; 2217 else if (sd->sd_state == G_RAID_SUBDISK_S_RESYNC && 2218 state != G_RAID_SUBDISK_S_REBUILD) 2219 state = G_RAID_SUBDISK_S_RESYNC; 2220 else if (sd->sd_state == G_RAID_SUBDISK_S_STALE) 2221 stale = 1; 2222 if ((sd->sd_state == G_RAID_SUBDISK_S_REBUILD || 2223 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) && 2224 sd->sd_rebuild_pos < pos) 2225 pos = sd->sd_rebuild_pos; 2226 } 2227 if (state == G_RAID_SUBDISK_S_REBUILD) { 2228 mvol->migr_state = 1; 2229 mvol->migr_type = INTEL_MT_REBUILD; 2230 } else if (state == G_RAID_SUBDISK_S_RESYNC) { 2231 mvol->migr_state = 1; 2232 /* mvol->migr_type = INTEL_MT_REPAIR; */ 2233 mvol->migr_type = INTEL_MT_VERIFY; 2234 mvol->state |= INTEL_ST_VERIFY_AND_FIX; 2235 } else 2236 mvol->migr_state = 0; 2237 mvol->dirty = (vol->v_dirty || stale); 2238 2239 mmap0 = intel_get_map(mvol, 0); 2240 2241 /* Write map / common part of two maps. */ 2242 intel_set_map_offset(mmap0, sd->sd_offset / sectorsize); 2243 intel_set_map_disk_sectors(mmap0, sd->sd_size / sectorsize); 2244 mmap0->strip_sectors = vol->v_strip_size / sectorsize; 2245 if (vol->v_state == G_RAID_VOLUME_S_BROKEN) 2246 mmap0->status = INTEL_S_FAILURE; 2247 else if (vol->v_state == G_RAID_VOLUME_S_DEGRADED) 2248 mmap0->status = INTEL_S_DEGRADED; 2249 else 2250 mmap0->status = INTEL_S_READY; 2251 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0) 2252 mmap0->type = INTEL_T_RAID0; 2253 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1 || 2254 vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) 2255 mmap0->type = INTEL_T_RAID1; 2256 else 2257 mmap0->type = INTEL_T_RAID5; 2258 mmap0->total_disks = vol->v_disks_count; 2259 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) 2260 mmap0->total_domains = vol->v_disks_count; 2261 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) 2262 mmap0->total_domains = 2; 2263 else 2264 mmap0->total_domains = 1; 2265 intel_set_map_stripe_count(mmap0, 2266 sd->sd_size / vol->v_strip_size / mmap0->total_domains); 2267 mmap0->failed_disk_num = 0xff; 2268 mmap0->ddf = 1; 2269 2270 /* If there are two maps - copy common and update. */ 2271 if (mvol->migr_state) { 2272 intel_set_vol_curr_migr_unit(mvol, 2273 pos / vol->v_strip_size / mmap0->total_domains); 2274 mmap1 = intel_get_map(mvol, 1); 2275 memcpy(mmap1, mmap0, sizeof(struct intel_raid_map)); 2276 mmap0->status = INTEL_S_READY; 2277 } else 2278 mmap1 = NULL; 2279 2280 /* Write disk indexes and put rebuild flags. */ 2281 for (sdi = 0; sdi < vol->v_disks_count; sdi++) { 2282 sd = &vol->v_subdisks[sdi]; 2283 pd = (struct g_raid_md_intel_perdisk *) 2284 sd->sd_disk->d_md_data; 2285 mmap0->disk_idx[sdi] = pd->pd_disk_pos; 2286 if (mvol->migr_state) 2287 mmap1->disk_idx[sdi] = pd->pd_disk_pos; 2288 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD || 2289 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) { 2290 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD; 2291 } else if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE && 2292 sd->sd_state != G_RAID_SUBDISK_S_STALE) { 2293 mmap0->disk_idx[sdi] |= INTEL_DI_RBLD; 2294 if (mvol->migr_state) 2295 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD; 2296 } 2297 if ((sd->sd_state == G_RAID_SUBDISK_S_NONE || 2298 sd->sd_state == G_RAID_SUBDISK_S_FAILED) && 2299 mmap0->failed_disk_num == 0xff) { 2300 mmap0->failed_disk_num = sdi; 2301 if (mvol->migr_state) 2302 mmap1->failed_disk_num = sdi; 2303 } 2304 } 2305 vi++; 2306 } 2307 meta->total_volumes = vi; 2308 if (strcmp(version, INTEL_VERSION_1300) != 0) 2309 meta->attributes &= INTEL_ATTR_CHECKSUM; 2310 memcpy(&meta->version[0], version, sizeof(INTEL_VERSION_1000) - 1); 2311 2312 /* We are done. Print meta data and store them to disks. */ 2313 g_raid_md_intel_print(meta); 2314 if (mdi->mdio_meta != NULL) 2315 free(mdi->mdio_meta, M_MD_INTEL); 2316 mdi->mdio_meta = meta; 2317 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 2318 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 2319 if (disk->d_state != G_RAID_DISK_S_ACTIVE) 2320 continue; 2321 if (pd->pd_meta != NULL) { 2322 free(pd->pd_meta, M_MD_INTEL); 2323 pd->pd_meta = NULL; 2324 } 2325 pd->pd_meta = intel_meta_copy(meta); 2326 intel_meta_write(disk->d_consumer, meta); 2327 } 2328 return (0); 2329} 2330 2331static int 2332g_raid_md_fail_disk_intel(struct g_raid_md_object *md, 2333 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk) 2334{ 2335 struct g_raid_softc *sc; 2336 struct g_raid_md_intel_object *mdi; 2337 struct g_raid_md_intel_perdisk *pd; 2338 struct g_raid_subdisk *sd; 2339 2340 sc = md->mdo_softc; 2341 mdi = (struct g_raid_md_intel_object *)md; 2342 pd = (struct g_raid_md_intel_perdisk *)tdisk->d_md_data; 2343 2344 /* We can't fail disk that is not a part of array now. */ 2345 if (pd->pd_disk_pos < 0) 2346 return (-1); 2347 2348 /* 2349 * Mark disk as failed in metadata and try to write that metadata 2350 * to the disk itself to prevent it's later resurrection as STALE. 2351 */ 2352 mdi->mdio_meta->disk[pd->pd_disk_pos].flags = INTEL_F_FAILED; 2353 pd->pd_disk_meta.flags = INTEL_F_FAILED; 2354 g_raid_md_intel_print(mdi->mdio_meta); 2355 if (tdisk->d_consumer) 2356 intel_meta_write(tdisk->d_consumer, mdi->mdio_meta); 2357 2358 /* Change states. */ 2359 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED); 2360 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) { 2361 g_raid_change_subdisk_state(sd, 2362 G_RAID_SUBDISK_S_FAILED); 2363 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED, 2364 G_RAID_EVENT_SUBDISK); 2365 } 2366 2367 /* Write updated metadata to remaining disks. */ 2368 g_raid_md_write_intel(md, NULL, NULL, tdisk); 2369 2370 /* Check if anything left except placeholders. */ 2371 if (g_raid_ndisks(sc, -1) == 2372 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 2373 g_raid_destroy_node(sc, 0); 2374 else 2375 g_raid_md_intel_refill(sc); 2376 return (0); 2377} 2378 2379static int 2380g_raid_md_free_disk_intel(struct g_raid_md_object *md, 2381 struct g_raid_disk *disk) 2382{ 2383 struct g_raid_md_intel_perdisk *pd; 2384 2385 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data; 2386 if (pd->pd_meta != NULL) { 2387 free(pd->pd_meta, M_MD_INTEL); 2388 pd->pd_meta = NULL; 2389 } 2390 free(pd, M_MD_INTEL); 2391 disk->d_md_data = NULL; 2392 return (0); 2393} 2394 2395static int 2396g_raid_md_free_intel(struct g_raid_md_object *md) 2397{ 2398 struct g_raid_md_intel_object *mdi; 2399 2400 mdi = (struct g_raid_md_intel_object *)md; 2401 if (!mdi->mdio_started) { 2402 mdi->mdio_started = 0; 2403 callout_stop(&mdi->mdio_start_co); 2404 G_RAID_DEBUG1(1, md->mdo_softc, 2405 "root_mount_rel %p", mdi->mdio_rootmount); 2406 root_mount_rel(mdi->mdio_rootmount); 2407 mdi->mdio_rootmount = NULL; 2408 } 2409 if (mdi->mdio_meta != NULL) { 2410 free(mdi->mdio_meta, M_MD_INTEL); 2411 mdi->mdio_meta = NULL; 2412 } 2413 return (0); 2414} 2415 2416G_RAID_MD_DECLARE(g_raid_md_intel); 2417