1 /*- 2 * Copyright (c) 2011 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_SII, "md_sii_data", "GEOM_RAID SiI metadata"); 47 48 struct sii_raid_conf { 49 uint16_t ata_params_00_53[54]; 50 uint64_t total_sectors; /* 54 - 57 */ 51 uint16_t ata_params_58_81[72]; 52 uint16_t product_id; /* 130 */ 53 uint16_t vendor_id; /* 131 */ 54 uint16_t version_minor; /* 132 */ 55 uint16_t version_major; /* 133 */ 56 uint8_t timestamp[6]; /* 134 - 136 */ 57 uint16_t strip_sectors; /* 137 */ 58 uint16_t dummy_2; 59 uint8_t disk_number; /* 139 */ 60 uint8_t type; 61 #define SII_T_RAID0 0x00 62 #define SII_T_RAID1 0x01 63 #define SII_T_RAID01 0x02 64 #define SII_T_SPARE 0x03 65 #define SII_T_CONCAT 0x04 66 #define SII_T_RAID5 0x10 67 #define SII_T_RESERVED 0xfd 68 #define SII_T_JBOD 0xff 69 70 uint8_t raid0_disks; /* 140 */ 71 uint8_t raid0_ident; 72 uint8_t raid1_disks; /* 141 */ 73 uint8_t raid1_ident; 74 uint64_t rebuild_lba; /* 142 - 145 */ 75 uint32_t generation; /* 146 - 147 */ 76 uint8_t disk_status; /* 148 */ 77 #define SII_S_CURRENT 0x01 78 #define SII_S_REBUILD 0x02 79 #define SII_S_DROPPED 0x03 80 #define SII_S_REMOVED 0x04 81 82 uint8_t raid_status; 83 #define SII_S_ONLINE 0x01 84 #define SII_S_AVAILABLE 0x02 85 86 uint8_t raid_location; /* 149 */ 87 uint8_t disk_location; 88 uint8_t auto_rebuild; /* 150 */ 89 #define SII_R_REBUILD 0x00 90 #define SII_R_NOREBUILD 0xff 91 92 uint8_t dummy_3; 93 uint8_t name[16]; /* 151 - 158 */ 94 uint16_t checksum; /* 159 */ 95 uint16_t ata_params_160_255[96]; 96 } __packed; 97 98 struct g_raid_md_sii_perdisk { 99 struct sii_raid_conf *pd_meta; 100 int pd_disk_pos; 101 off_t pd_disk_size; 102 }; 103 104 struct g_raid_md_sii_object { 105 struct g_raid_md_object mdio_base; 106 uint8_t mdio_timestamp[6]; 107 uint8_t mdio_location; 108 uint32_t mdio_generation; 109 struct sii_raid_conf *mdio_meta; 110 struct callout mdio_start_co; /* STARTING state timer. */ 111 int mdio_total_disks; 112 int mdio_disks_present; 113 int mdio_started; 114 int mdio_incomplete; 115 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */ 116 }; 117 118 static g_raid_md_create_t g_raid_md_create_sii; 119 static g_raid_md_taste_t g_raid_md_taste_sii; 120 static g_raid_md_event_t g_raid_md_event_sii; 121 static g_raid_md_ctl_t g_raid_md_ctl_sii; 122 static g_raid_md_write_t g_raid_md_write_sii; 123 static g_raid_md_fail_disk_t g_raid_md_fail_disk_sii; 124 static g_raid_md_free_disk_t g_raid_md_free_disk_sii; 125 static g_raid_md_free_t g_raid_md_free_sii; 126 127 static kobj_method_t g_raid_md_sii_methods[] = { 128 KOBJMETHOD(g_raid_md_create, g_raid_md_create_sii), 129 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_sii), 130 KOBJMETHOD(g_raid_md_event, g_raid_md_event_sii), 131 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_sii), 132 KOBJMETHOD(g_raid_md_write, g_raid_md_write_sii), 133 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_sii), 134 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_sii), 135 KOBJMETHOD(g_raid_md_free, g_raid_md_free_sii), 136 { 0, 0 } 137 }; 138 139 static struct g_raid_md_class g_raid_md_sii_class = { 140 "SiI", 141 g_raid_md_sii_methods, 142 sizeof(struct g_raid_md_sii_object), 143 .mdc_enable = 1, 144 .mdc_priority = 100 145 }; 146 147 static void 148 g_raid_md_sii_print(struct sii_raid_conf *meta) 149 { 150 151 if (g_raid_debug < 1) 152 return; 153 154 printf("********* ATA SiI RAID Metadata *********\n"); 155 printf("total_sectors %llu\n", 156 (long long unsigned)meta->total_sectors); 157 printf("product_id 0x%04x\n", meta->product_id); 158 printf("vendor_id 0x%04x\n", meta->vendor_id); 159 printf("version_minor 0x%04x\n", meta->version_minor); 160 printf("version_major 0x%04x\n", meta->version_major); 161 printf("timestamp 0x%02x%02x%02x%02x%02x%02x\n", 162 meta->timestamp[5], meta->timestamp[4], meta->timestamp[3], 163 meta->timestamp[2], meta->timestamp[1], meta->timestamp[0]); 164 printf("strip_sectors %d\n", meta->strip_sectors); 165 printf("disk_number %d\n", meta->disk_number); 166 printf("type 0x%02x\n", meta->type); 167 printf("raid0_disks %d\n", meta->raid0_disks); 168 printf("raid0_ident %d\n", meta->raid0_ident); 169 printf("raid1_disks %d\n", meta->raid1_disks); 170 printf("raid1_ident %d\n", meta->raid1_ident); 171 printf("rebuild_lba %llu\n", 172 (long long unsigned)meta->rebuild_lba); 173 printf("generation %d\n", meta->generation); 174 printf("disk_status %d\n", meta->disk_status); 175 printf("raid_status %d\n", meta->raid_status); 176 printf("raid_location %d\n", meta->raid_location); 177 printf("disk_location %d\n", meta->disk_location); 178 printf("auto_rebuild %d\n", meta->auto_rebuild); 179 printf("name <%.16s>\n", meta->name); 180 printf("checksum 0x%04x\n", meta->checksum); 181 printf("=================================================\n"); 182 } 183 184 static struct sii_raid_conf * 185 sii_meta_copy(struct sii_raid_conf *meta) 186 { 187 struct sii_raid_conf *nmeta; 188 189 nmeta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK); 190 memcpy(nmeta, meta, sizeof(*meta)); 191 return (nmeta); 192 } 193 194 static int 195 sii_meta_total_disks(struct sii_raid_conf *meta) 196 { 197 198 switch (meta->type) { 199 case SII_T_RAID0: 200 case SII_T_RAID5: 201 case SII_T_CONCAT: 202 return (meta->raid0_disks); 203 case SII_T_RAID1: 204 return (meta->raid1_disks); 205 case SII_T_RAID01: 206 return (meta->raid0_disks * meta->raid1_disks); 207 case SII_T_SPARE: 208 case SII_T_JBOD: 209 return (1); 210 } 211 return (0); 212 } 213 214 static int 215 sii_meta_disk_pos(struct sii_raid_conf *meta, struct sii_raid_conf *pdmeta) 216 { 217 218 if (pdmeta->type == SII_T_SPARE) 219 return (-3); 220 221 if (memcmp(&meta->timestamp, &pdmeta->timestamp, 6) != 0) 222 return (-1); 223 224 switch (pdmeta->type) { 225 case SII_T_RAID0: 226 case SII_T_RAID1: 227 case SII_T_RAID5: 228 case SII_T_CONCAT: 229 return (pdmeta->disk_number); 230 case SII_T_RAID01: 231 return (pdmeta->raid1_ident * pdmeta->raid1_disks + 232 pdmeta->raid0_ident); 233 case SII_T_JBOD: 234 return (0); 235 } 236 return (-1); 237 } 238 239 static void 240 sii_meta_get_name(struct sii_raid_conf *meta, char *buf) 241 { 242 int i; 243 244 strncpy(buf, meta->name, 16); 245 buf[16] = 0; 246 for (i = 15; i >= 0; i--) { 247 if (buf[i] > 0x20) 248 break; 249 buf[i] = 0; 250 } 251 } 252 253 static void 254 sii_meta_put_name(struct sii_raid_conf *meta, char *buf) 255 { 256 257 memset(meta->name, 0x20, 16); 258 memcpy(meta->name, buf, MIN(strlen(buf), 16)); 259 } 260 261 static struct sii_raid_conf * 262 sii_meta_read(struct g_consumer *cp) 263 { 264 struct g_provider *pp; 265 struct sii_raid_conf *meta; 266 char *buf; 267 int error, i; 268 uint16_t checksum, *ptr; 269 270 pp = cp->provider; 271 272 /* Read the anchor sector. */ 273 buf = g_read_data(cp, 274 pp->mediasize - pp->sectorsize, pp->sectorsize, &error); 275 if (buf == NULL) { 276 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).", 277 pp->name, error); 278 return (NULL); 279 } 280 meta = (struct sii_raid_conf *)buf; 281 282 /* Check vendor ID. */ 283 if (meta->vendor_id != 0x1095) { 284 G_RAID_DEBUG(1, "SiI vendor ID check failed on %s (0x%04x)", 285 pp->name, meta->vendor_id); 286 g_free(buf); 287 return (NULL); 288 } 289 290 /* Check metadata major version. */ 291 if (meta->version_major != 2) { 292 G_RAID_DEBUG(1, "SiI version check failed on %s (%d.%d)", 293 pp->name, meta->version_major, meta->version_minor); 294 g_free(buf); 295 return (NULL); 296 } 297 meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK); 298 memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize)); 299 g_free(buf); 300 301 /* Check metadata checksum. */ 302 for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i <= 159; i++) 303 checksum += *ptr++; 304 if (checksum != 0) { 305 G_RAID_DEBUG(1, "SiI checksum check failed on %s", pp->name); 306 free(meta, M_MD_SII); 307 return (NULL); 308 } 309 310 /* Check raid type. */ 311 if (meta->type != SII_T_RAID0 && meta->type != SII_T_RAID1 && 312 meta->type != SII_T_RAID01 && meta->type != SII_T_SPARE && 313 meta->type != SII_T_RAID5 && meta->type != SII_T_CONCAT && 314 meta->type != SII_T_JBOD) { 315 G_RAID_DEBUG(1, "SiI unknown RAID level on %s (0x%02x)", 316 pp->name, meta->type); 317 free(meta, M_MD_SII); 318 return (NULL); 319 } 320 321 return (meta); 322 } 323 324 static int 325 sii_meta_write(struct g_consumer *cp, struct sii_raid_conf *meta) 326 { 327 struct g_provider *pp; 328 char *buf; 329 int error, i; 330 uint16_t checksum, *ptr; 331 332 pp = cp->provider; 333 334 /* Recalculate checksum for case if metadata were changed. */ 335 meta->checksum = 0; 336 for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i < 159; i++) 337 checksum += *ptr++; 338 meta->checksum -= checksum; 339 340 /* Create and fill buffer. */ 341 buf = malloc(pp->sectorsize, M_MD_SII, M_WAITOK | M_ZERO); 342 memcpy(buf, meta, sizeof(*meta)); 343 344 /* Write 4 copies of metadata. */ 345 for (i = 0; i < 4; i++) { 346 error = g_write_data(cp, 347 pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)), 348 buf, pp->sectorsize); 349 if (error != 0) { 350 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).", 351 pp->name, error); 352 break; 353 } 354 } 355 356 free(buf, M_MD_SII); 357 return (error); 358 } 359 360 static int 361 sii_meta_erase(struct g_consumer *cp) 362 { 363 struct g_provider *pp; 364 char *buf; 365 int error, i; 366 367 pp = cp->provider; 368 buf = malloc(pp->sectorsize, M_MD_SII, M_WAITOK | M_ZERO); 369 /* Write 4 copies of metadata. */ 370 for (i = 0; i < 4; i++) { 371 error = g_write_data(cp, 372 pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)), 373 buf, pp->sectorsize); 374 if (error != 0) { 375 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).", 376 pp->name, error); 377 } 378 } 379 free(buf, M_MD_SII); 380 return (error); 381 } 382 383 static int 384 sii_meta_write_spare(struct g_consumer *cp) 385 { 386 struct sii_raid_conf *meta; 387 int error; 388 389 meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK | M_ZERO); 390 meta->total_sectors = cp->provider->mediasize / 391 cp->provider->sectorsize - 0x800; 392 meta->vendor_id = 0x1095; 393 meta->version_minor = 0; 394 meta->version_major = 2; 395 meta->timestamp[0] = arc4random(); 396 meta->timestamp[1] = arc4random(); 397 meta->timestamp[2] = arc4random(); 398 meta->timestamp[3] = arc4random(); 399 meta->timestamp[4] = arc4random(); 400 meta->timestamp[5] = arc4random(); 401 meta->type = SII_T_SPARE; 402 meta->generation = 1; 403 meta->raid1_ident = 0xff; 404 meta->raid_location = arc4random(); 405 error = sii_meta_write(cp, meta); 406 free(meta, M_MD_SII); 407 return (error); 408 } 409 410 static struct g_raid_disk * 411 g_raid_md_sii_get_disk(struct g_raid_softc *sc, int id) 412 { 413 struct g_raid_disk *disk; 414 struct g_raid_md_sii_perdisk *pd; 415 416 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 417 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 418 if (pd->pd_disk_pos == id) 419 break; 420 } 421 return (disk); 422 } 423 424 static int 425 g_raid_md_sii_supported(int level, int qual, int disks, int force) 426 { 427 428 if (disks > 8) 429 return (0); 430 switch (level) { 431 case G_RAID_VOLUME_RL_RAID0: 432 if (disks < 1) 433 return (0); 434 if (!force && (disks < 2 || disks > 6)) 435 return (0); 436 break; 437 case G_RAID_VOLUME_RL_RAID1: 438 if (disks < 1) 439 return (0); 440 if (!force && (disks != 2)) 441 return (0); 442 break; 443 case G_RAID_VOLUME_RL_RAID1E: 444 if (disks < 2) 445 return (0); 446 if (disks % 2 != 0) 447 return (0); 448 if (!force && (disks < 4)) 449 return (0); 450 break; 451 case G_RAID_VOLUME_RL_SINGLE: 452 if (disks != 1) 453 return (0); 454 break; 455 case G_RAID_VOLUME_RL_CONCAT: 456 if (disks < 2) 457 return (0); 458 break; 459 case G_RAID_VOLUME_RL_RAID5: 460 if (disks < 3) 461 return (0); 462 if (qual != G_RAID_VOLUME_RLQ_R5LS) 463 return (0); 464 break; 465 default: 466 return (0); 467 } 468 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE) 469 return (0); 470 return (1); 471 } 472 473 static int 474 g_raid_md_sii_start_disk(struct g_raid_disk *disk) 475 { 476 struct g_raid_softc *sc; 477 struct g_raid_subdisk *sd, *tmpsd; 478 struct g_raid_disk *olddisk, *tmpdisk; 479 struct g_raid_md_object *md; 480 struct g_raid_md_sii_object *mdi; 481 struct g_raid_md_sii_perdisk *pd, *oldpd; 482 struct sii_raid_conf *meta; 483 int disk_pos, resurrection = 0; 484 485 sc = disk->d_softc; 486 md = sc->sc_md; 487 mdi = (struct g_raid_md_sii_object *)md; 488 meta = mdi->mdio_meta; 489 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 490 olddisk = NULL; 491 492 /* Find disk position in metadata by it's serial. */ 493 if (pd->pd_meta != NULL) 494 disk_pos = sii_meta_disk_pos(meta, pd->pd_meta); 495 else 496 disk_pos = -3; 497 if (disk_pos < 0) { 498 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk"); 499 /* If we are in the start process, that's all for now. */ 500 if (!mdi->mdio_started) 501 goto nofit; 502 /* 503 * If we have already started - try to get use of the disk. 504 * Try to replace OFFLINE disks first, then FAILED. 505 */ 506 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) { 507 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE && 508 tmpdisk->d_state != G_RAID_DISK_S_FAILED) 509 continue; 510 /* Make sure this disk is big enough. */ 511 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) { 512 if (sd->sd_offset + sd->sd_size + 512 > 513 pd->pd_disk_size) { 514 G_RAID_DEBUG1(1, sc, 515 "Disk too small (%ju < %ju)", 516 pd->pd_disk_size, 517 sd->sd_offset + sd->sd_size + 512); 518 break; 519 } 520 } 521 if (sd != NULL) 522 continue; 523 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) { 524 olddisk = tmpdisk; 525 break; 526 } else if (olddisk == NULL) 527 olddisk = tmpdisk; 528 } 529 if (olddisk == NULL) { 530 nofit: 531 if (disk_pos == -3 || pd->pd_disk_pos == -3) { 532 g_raid_change_disk_state(disk, 533 G_RAID_DISK_S_SPARE); 534 return (1); 535 } else { 536 g_raid_change_disk_state(disk, 537 G_RAID_DISK_S_STALE); 538 return (0); 539 } 540 } 541 oldpd = (struct g_raid_md_sii_perdisk *)olddisk->d_md_data; 542 disk_pos = oldpd->pd_disk_pos; 543 resurrection = 1; 544 } 545 546 if (olddisk == NULL) { 547 /* Find placeholder by position. */ 548 olddisk = g_raid_md_sii_get_disk(sc, disk_pos); 549 if (olddisk == NULL) 550 panic("No disk at position %d!", disk_pos); 551 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) { 552 G_RAID_DEBUG1(1, sc, "More then one disk for pos %d", 553 disk_pos); 554 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE); 555 return (0); 556 } 557 oldpd = (struct g_raid_md_sii_perdisk *)olddisk->d_md_data; 558 } 559 560 /* Replace failed disk or placeholder with new disk. */ 561 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) { 562 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next); 563 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 564 sd->sd_disk = disk; 565 } 566 oldpd->pd_disk_pos = -2; 567 pd->pd_disk_pos = disk_pos; 568 569 /* If it was placeholder -- destroy it. */ 570 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) { 571 g_raid_destroy_disk(olddisk); 572 } else { 573 /* Otherwise, make it STALE_FAILED. */ 574 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED); 575 } 576 577 /* Welcome the new disk. */ 578 if (resurrection) 579 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); 580 else if (pd->pd_meta->disk_status == SII_S_CURRENT || 581 pd->pd_meta->disk_status == SII_S_REBUILD) 582 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); 583 else 584 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED); 585 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 586 587 /* 588 * Different disks may have different sizes, 589 * in concat mode. Update from real disk size. 590 */ 591 if (meta->type == SII_T_CONCAT || meta->type == SII_T_JBOD) 592 sd->sd_size = pd->pd_disk_size - 0x800 * 512; 593 594 if (resurrection) { 595 /* New or ex-spare disk. */ 596 g_raid_change_subdisk_state(sd, 597 G_RAID_SUBDISK_S_NEW); 598 } else if (pd->pd_meta->disk_status == SII_S_REBUILD) { 599 /* Rebuilding disk. */ 600 g_raid_change_subdisk_state(sd, 601 G_RAID_SUBDISK_S_REBUILD); 602 if (pd->pd_meta->generation == meta->generation) 603 sd->sd_rebuild_pos = pd->pd_meta->rebuild_lba * 512; 604 else 605 sd->sd_rebuild_pos = 0; 606 } else if (pd->pd_meta->disk_status == SII_S_CURRENT) { 607 if (pd->pd_meta->raid_status == SII_S_ONLINE || 608 pd->pd_meta->generation != meta->generation) { 609 /* Dirty or resyncing disk. */ 610 g_raid_change_subdisk_state(sd, 611 G_RAID_SUBDISK_S_STALE); 612 } else { 613 /* Up to date disk. */ 614 g_raid_change_subdisk_state(sd, 615 G_RAID_SUBDISK_S_ACTIVE); 616 } 617 } else { 618 g_raid_change_subdisk_state(sd, 619 G_RAID_SUBDISK_S_FAILED); 620 } 621 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 622 G_RAID_EVENT_SUBDISK); 623 } 624 625 /* Update status of our need for spare. */ 626 if (mdi->mdio_started) { 627 mdi->mdio_incomplete = 628 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < 629 mdi->mdio_total_disks); 630 } 631 632 return (resurrection); 633 } 634 635 static void 636 g_disk_md_sii_retaste(void *arg, int pending) 637 { 638 639 G_RAID_DEBUG(1, "Array is not complete, trying to retaste."); 640 g_retaste(&g_raid_class); 641 free(arg, M_MD_SII); 642 } 643 644 static void 645 g_raid_md_sii_refill(struct g_raid_softc *sc) 646 { 647 struct g_raid_md_object *md; 648 struct g_raid_md_sii_object *mdi; 649 struct g_raid_disk *disk; 650 struct task *task; 651 int update, na; 652 653 md = sc->sc_md; 654 mdi = (struct g_raid_md_sii_object *)md; 655 update = 0; 656 do { 657 /* Make sure we miss anything. */ 658 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE); 659 if (na == mdi->mdio_total_disks) 660 break; 661 662 G_RAID_DEBUG1(1, md->mdo_softc, 663 "Array is not complete (%d of %d), " 664 "trying to refill.", na, mdi->mdio_total_disks); 665 666 /* Try to get use some of STALE disks. */ 667 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 668 if (disk->d_state == G_RAID_DISK_S_STALE) { 669 update += g_raid_md_sii_start_disk(disk); 670 if (disk->d_state == G_RAID_DISK_S_ACTIVE) 671 break; 672 } 673 } 674 if (disk != NULL) 675 continue; 676 677 /* Try to get use some of SPARE disks. */ 678 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 679 if (disk->d_state == G_RAID_DISK_S_SPARE) { 680 update += g_raid_md_sii_start_disk(disk); 681 if (disk->d_state == G_RAID_DISK_S_ACTIVE) 682 break; 683 } 684 } 685 } while (disk != NULL); 686 687 /* Write new metadata if we changed something. */ 688 if (update) 689 g_raid_md_write_sii(md, NULL, NULL, NULL); 690 691 /* Update status of our need for spare. */ 692 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < 693 mdi->mdio_total_disks); 694 695 /* Request retaste hoping to find spare. */ 696 if (mdi->mdio_incomplete) { 697 task = malloc(sizeof(struct task), 698 M_MD_SII, M_WAITOK | M_ZERO); 699 TASK_INIT(task, 0, g_disk_md_sii_retaste, task); 700 taskqueue_enqueue(taskqueue_swi, task); 701 } 702 } 703 704 static void 705 g_raid_md_sii_start(struct g_raid_softc *sc) 706 { 707 struct g_raid_md_object *md; 708 struct g_raid_md_sii_object *mdi; 709 struct g_raid_md_sii_perdisk *pd; 710 struct sii_raid_conf *meta; 711 struct g_raid_volume *vol; 712 struct g_raid_subdisk *sd; 713 struct g_raid_disk *disk, *best; 714 off_t size; 715 int j, disk_pos; 716 uint32_t gendiff, bestgendiff; 717 char buf[17]; 718 719 md = sc->sc_md; 720 mdi = (struct g_raid_md_sii_object *)md; 721 meta = mdi->mdio_meta; 722 723 /* Create volumes and subdisks. */ 724 sii_meta_get_name(meta, buf); 725 vol = g_raid_create_volume(sc, buf, -1); 726 vol->v_mediasize = (off_t)meta->total_sectors * 512; 727 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE; 728 if (meta->type == SII_T_RAID0) { 729 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0; 730 size = vol->v_mediasize / mdi->mdio_total_disks; 731 } else if (meta->type == SII_T_RAID1) { 732 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1; 733 size = vol->v_mediasize; 734 } else if (meta->type == SII_T_RAID01) { 735 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E; 736 size = vol->v_mediasize / (mdi->mdio_total_disks / 2); 737 } else if (meta->type == SII_T_CONCAT) { 738 if (mdi->mdio_total_disks == 1) 739 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE; 740 else 741 vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT; 742 size = 0; 743 } else if (meta->type == SII_T_RAID5) { 744 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5; 745 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LS; 746 size = vol->v_mediasize / (mdi->mdio_total_disks - 1); 747 } else if (meta->type == SII_T_JBOD) { 748 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE; 749 size = 0; 750 } else { 751 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN; 752 size = 0; 753 } 754 vol->v_strip_size = meta->strip_sectors * 512; //ZZZ 755 vol->v_disks_count = mdi->mdio_total_disks; 756 vol->v_sectorsize = 512; //ZZZ 757 for (j = 0; j < vol->v_disks_count; j++) { 758 sd = &vol->v_subdisks[j]; 759 sd->sd_offset = 0; 760 sd->sd_size = size; 761 } 762 g_raid_start_volume(vol); 763 764 /* Create disk placeholders to store data for later writing. */ 765 for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) { 766 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO); 767 pd->pd_disk_pos = disk_pos; 768 disk = g_raid_create_disk(sc); 769 disk->d_md_data = (void *)pd; 770 disk->d_state = G_RAID_DISK_S_OFFLINE; 771 sd = &vol->v_subdisks[disk_pos]; 772 sd->sd_disk = disk; 773 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 774 } 775 776 /* 777 * Make all disks found till the moment take their places 778 * in order of their generation numbers. 779 */ 780 do { 781 best = NULL; 782 bestgendiff = 0xffffffff; 783 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 784 if (disk->d_state != G_RAID_DISK_S_NONE) 785 continue; 786 pd = disk->d_md_data; 787 if (pd->pd_meta == NULL) 788 gendiff = 0xfffffffe; 789 else 790 gendiff = meta->generation - 791 pd->pd_meta->generation; 792 if (gendiff < bestgendiff) { 793 best = disk; 794 bestgendiff = gendiff; 795 } 796 } 797 if (best != NULL) 798 g_raid_md_sii_start_disk(best); 799 } while (best != NULL); 800 801 mdi->mdio_started = 1; 802 G_RAID_DEBUG1(0, sc, "Array started."); 803 g_raid_md_write_sii(md, NULL, NULL, NULL); 804 805 /* Pickup any STALE/SPARE disks to refill array if needed. */ 806 g_raid_md_sii_refill(sc); 807 808 g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME); 809 810 callout_stop(&mdi->mdio_start_co); 811 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount); 812 root_mount_rel(mdi->mdio_rootmount); 813 mdi->mdio_rootmount = NULL; 814 } 815 816 static void 817 g_raid_md_sii_new_disk(struct g_raid_disk *disk) 818 { 819 struct g_raid_softc *sc; 820 struct g_raid_md_object *md; 821 struct g_raid_md_sii_object *mdi; 822 struct sii_raid_conf *pdmeta; 823 struct g_raid_md_sii_perdisk *pd; 824 825 sc = disk->d_softc; 826 md = sc->sc_md; 827 mdi = (struct g_raid_md_sii_object *)md; 828 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 829 pdmeta = pd->pd_meta; 830 831 if (mdi->mdio_started) { 832 if (g_raid_md_sii_start_disk(disk)) 833 g_raid_md_write_sii(md, NULL, NULL, NULL); 834 } else { 835 if (mdi->mdio_meta == NULL || 836 ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) { 837 G_RAID_DEBUG1(1, sc, "Newer disk"); 838 if (mdi->mdio_meta != NULL) 839 free(mdi->mdio_meta, M_MD_SII); 840 mdi->mdio_meta = sii_meta_copy(pdmeta); 841 mdi->mdio_generation = mdi->mdio_meta->generation; 842 mdi->mdio_total_disks = sii_meta_total_disks(pdmeta); 843 mdi->mdio_disks_present = 1; 844 } else if (pdmeta->generation == mdi->mdio_generation) { 845 mdi->mdio_disks_present++; 846 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)", 847 mdi->mdio_disks_present, 848 mdi->mdio_total_disks); 849 } else { 850 G_RAID_DEBUG1(1, sc, "Older disk"); 851 } 852 853 /* If we collected all needed disks - start array. */ 854 if (mdi->mdio_disks_present == mdi->mdio_total_disks) 855 g_raid_md_sii_start(sc); 856 } 857 } 858 859 static void 860 g_raid_sii_go(void *arg) 861 { 862 struct g_raid_softc *sc; 863 struct g_raid_md_object *md; 864 struct g_raid_md_sii_object *mdi; 865 866 sc = arg; 867 md = sc->sc_md; 868 mdi = (struct g_raid_md_sii_object *)md; 869 if (!mdi->mdio_started) { 870 G_RAID_DEBUG1(0, sc, "Force array start due to timeout."); 871 g_raid_event_send(sc, G_RAID_NODE_E_START, 0); 872 } 873 } 874 875 static int 876 g_raid_md_create_sii(struct g_raid_md_object *md, struct g_class *mp, 877 struct g_geom **gp) 878 { 879 struct g_raid_softc *sc; 880 struct g_raid_md_sii_object *mdi; 881 char name[32]; 882 883 mdi = (struct g_raid_md_sii_object *)md; 884 mdi->mdio_timestamp[5] = arc4random(); 885 mdi->mdio_timestamp[4] = arc4random(); 886 mdi->mdio_timestamp[3] = arc4random(); 887 mdi->mdio_timestamp[2] = arc4random(); 888 mdi->mdio_timestamp[1] = arc4random(); 889 mdi->mdio_timestamp[0] = arc4random(); 890 mdi->mdio_location = arc4random(); 891 mdi->mdio_generation = 0; 892 snprintf(name, sizeof(name), "SiI-%02x%02x%02x%02x%02x%02x", 893 mdi->mdio_timestamp[5], mdi->mdio_timestamp[4], 894 mdi->mdio_timestamp[3], mdi->mdio_timestamp[2], 895 mdi->mdio_timestamp[1], mdi->mdio_timestamp[0]); 896 sc = g_raid_create_node(mp, name, md); 897 if (sc == NULL) 898 return (G_RAID_MD_TASTE_FAIL); 899 md->mdo_softc = sc; 900 *gp = sc->sc_geom; 901 return (G_RAID_MD_TASTE_NEW); 902 } 903 904 static int 905 g_raid_md_taste_sii(struct g_raid_md_object *md, struct g_class *mp, 906 struct g_consumer *cp, struct g_geom **gp) 907 { 908 struct g_consumer *rcp; 909 struct g_provider *pp; 910 struct g_raid_md_sii_object *mdi, *mdi1; 911 struct g_raid_softc *sc; 912 struct g_raid_disk *disk; 913 struct sii_raid_conf *meta; 914 struct g_raid_md_sii_perdisk *pd; 915 struct g_geom *geom; 916 int disk_pos, result, spare, len; 917 char name[32]; 918 uint16_t vendor; 919 920 G_RAID_DEBUG(1, "Tasting SiI on %s", cp->provider->name); 921 mdi = (struct g_raid_md_sii_object *)md; 922 pp = cp->provider; 923 924 /* Read metadata from device. */ 925 meta = NULL; 926 g_topology_unlock(); 927 vendor = 0xffff; 928 len = sizeof(vendor); 929 if (pp->geom->rank == 1) 930 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor); 931 meta = sii_meta_read(cp); 932 g_topology_lock(); 933 if (meta == NULL) { 934 if (g_raid_aggressive_spare) { 935 if (vendor == 0x1095) { 936 G_RAID_DEBUG(1, 937 "No SiI metadata, forcing spare."); 938 spare = 2; 939 goto search; 940 } else { 941 G_RAID_DEBUG(1, 942 "SiI vendor mismatch 0x%04x != 0x1095", 943 vendor); 944 } 945 } 946 return (G_RAID_MD_TASTE_FAIL); 947 } 948 949 /* Check this disk position in obtained metadata. */ 950 disk_pos = sii_meta_disk_pos(meta, meta); 951 if (disk_pos == -1) { 952 G_RAID_DEBUG(1, "SiI disk position not found"); 953 goto fail1; 954 } 955 956 /* Metadata valid. Print it. */ 957 g_raid_md_sii_print(meta); 958 G_RAID_DEBUG(1, "SiI disk position %d", disk_pos); 959 spare = (meta->type == SII_T_SPARE) ? 1 : 0; 960 961 search: 962 /* Search for matching node. */ 963 sc = NULL; 964 mdi1 = NULL; 965 LIST_FOREACH(geom, &mp->geom, geom) { 966 sc = geom->softc; 967 if (sc == NULL) 968 continue; 969 if (sc->sc_stopping != 0) 970 continue; 971 if (sc->sc_md->mdo_class != md->mdo_class) 972 continue; 973 mdi1 = (struct g_raid_md_sii_object *)sc->sc_md; 974 if (spare) { 975 if (mdi1->mdio_incomplete) 976 break; 977 } else { 978 if (mdi1->mdio_location == meta->raid_location && 979 memcmp(&mdi1->mdio_timestamp, 980 &meta->timestamp, 6) == 0) 981 break; 982 } 983 } 984 985 /* Found matching node. */ 986 if (geom != NULL) { 987 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name); 988 result = G_RAID_MD_TASTE_EXISTING; 989 990 } else if (spare) { /* Not found needy node -- left for later. */ 991 G_RAID_DEBUG(1, "Spare is not needed at this time"); 992 goto fail1; 993 994 } else { /* Not found matching node -- create one. */ 995 result = G_RAID_MD_TASTE_NEW; 996 memcpy(&mdi->mdio_timestamp, &meta->timestamp, 6); 997 mdi->mdio_location = meta->raid_location; 998 snprintf(name, sizeof(name), "SiI-%02x%02x%02x%02x%02x%02x", 999 mdi->mdio_timestamp[5], mdi->mdio_timestamp[4], 1000 mdi->mdio_timestamp[3], mdi->mdio_timestamp[2], 1001 mdi->mdio_timestamp[1], mdi->mdio_timestamp[0]); 1002 sc = g_raid_create_node(mp, name, md); 1003 md->mdo_softc = sc; 1004 geom = sc->sc_geom; 1005 callout_init(&mdi->mdio_start_co, 1); 1006 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz, 1007 g_raid_sii_go, sc); 1008 mdi->mdio_rootmount = root_mount_hold("GRAID-SiI"); 1009 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount); 1010 } 1011 1012 /* There is no return after this point, so we close passed consumer. */ 1013 g_access(cp, -1, 0, 0); 1014 1015 rcp = g_new_consumer(geom); 1016 rcp->flags |= G_CF_DIRECT_RECEIVE; 1017 g_attach(rcp, pp); 1018 if (g_access(rcp, 1, 1, 1) != 0) 1019 ; //goto fail1; 1020 1021 g_topology_unlock(); 1022 sx_xlock(&sc->sc_lock); 1023 1024 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO); 1025 pd->pd_meta = meta; 1026 if (spare == 2) { 1027 pd->pd_disk_pos = -3; 1028 } else { 1029 pd->pd_disk_pos = -1; 1030 } 1031 pd->pd_disk_size = pp->mediasize; 1032 disk = g_raid_create_disk(sc); 1033 disk->d_md_data = (void *)pd; 1034 disk->d_consumer = rcp; 1035 rcp->private = disk; 1036 1037 g_raid_get_disk_info(disk); 1038 1039 g_raid_md_sii_new_disk(disk); 1040 1041 sx_xunlock(&sc->sc_lock); 1042 g_topology_lock(); 1043 *gp = geom; 1044 return (result); 1045 fail1: 1046 free(meta, M_MD_SII); 1047 return (G_RAID_MD_TASTE_FAIL); 1048 } 1049 1050 static int 1051 g_raid_md_event_sii(struct g_raid_md_object *md, 1052 struct g_raid_disk *disk, u_int event) 1053 { 1054 struct g_raid_softc *sc; 1055 struct g_raid_subdisk *sd; 1056 struct g_raid_md_sii_object *mdi; 1057 struct g_raid_md_sii_perdisk *pd; 1058 1059 sc = md->mdo_softc; 1060 mdi = (struct g_raid_md_sii_object *)md; 1061 if (disk == NULL) { 1062 switch (event) { 1063 case G_RAID_NODE_E_START: 1064 if (!mdi->mdio_started) 1065 g_raid_md_sii_start(sc); 1066 return (0); 1067 } 1068 return (-1); 1069 } 1070 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1071 switch (event) { 1072 case G_RAID_DISK_E_DISCONNECTED: 1073 /* If disk was assigned, just update statuses. */ 1074 if (pd->pd_disk_pos >= 0) { 1075 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1076 if (disk->d_consumer) { 1077 g_raid_kill_consumer(sc, disk->d_consumer); 1078 disk->d_consumer = NULL; 1079 } 1080 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 1081 g_raid_change_subdisk_state(sd, 1082 G_RAID_SUBDISK_S_NONE); 1083 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED, 1084 G_RAID_EVENT_SUBDISK); 1085 } 1086 } else { 1087 /* Otherwise -- delete. */ 1088 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE); 1089 g_raid_destroy_disk(disk); 1090 } 1091 1092 /* Write updated metadata to all disks. */ 1093 g_raid_md_write_sii(md, NULL, NULL, NULL); 1094 1095 /* Check if anything left except placeholders. */ 1096 if (g_raid_ndisks(sc, -1) == 1097 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 1098 g_raid_destroy_node(sc, 0); 1099 else 1100 g_raid_md_sii_refill(sc); 1101 return (0); 1102 } 1103 return (-2); 1104 } 1105 1106 static int 1107 g_raid_md_ctl_sii(struct g_raid_md_object *md, 1108 struct gctl_req *req) 1109 { 1110 struct g_raid_softc *sc; 1111 struct g_raid_volume *vol; 1112 struct g_raid_subdisk *sd; 1113 struct g_raid_disk *disk; 1114 struct g_raid_md_sii_object *mdi; 1115 struct g_raid_md_sii_perdisk *pd; 1116 struct g_consumer *cp; 1117 struct g_provider *pp; 1118 char arg[16]; 1119 const char *verb, *volname, *levelname, *diskname; 1120 int *nargs, *force; 1121 off_t size, sectorsize, strip; 1122 intmax_t *sizearg, *striparg; 1123 int numdisks, i, len, level, qual, update; 1124 int error; 1125 1126 sc = md->mdo_softc; 1127 mdi = (struct g_raid_md_sii_object *)md; 1128 verb = gctl_get_param(req, "verb", NULL); 1129 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs)); 1130 error = 0; 1131 if (strcmp(verb, "label") == 0) { 1132 1133 if (*nargs < 4) { 1134 gctl_error(req, "Invalid number of arguments."); 1135 return (-1); 1136 } 1137 volname = gctl_get_asciiparam(req, "arg1"); 1138 if (volname == NULL) { 1139 gctl_error(req, "No volume name."); 1140 return (-2); 1141 } 1142 levelname = gctl_get_asciiparam(req, "arg2"); 1143 if (levelname == NULL) { 1144 gctl_error(req, "No RAID level."); 1145 return (-3); 1146 } 1147 if (strcasecmp(levelname, "RAID5") == 0) 1148 levelname = "RAID5-LS"; 1149 if (g_raid_volume_str2level(levelname, &level, &qual)) { 1150 gctl_error(req, "Unknown RAID level '%s'.", levelname); 1151 return (-4); 1152 } 1153 numdisks = *nargs - 3; 1154 force = gctl_get_paraml(req, "force", sizeof(*force)); 1155 if (!g_raid_md_sii_supported(level, qual, numdisks, 1156 force ? *force : 0)) { 1157 gctl_error(req, "Unsupported RAID level " 1158 "(0x%02x/0x%02x), or number of disks (%d).", 1159 level, qual, numdisks); 1160 return (-5); 1161 } 1162 1163 /* Search for disks, connect them and probe. */ 1164 size = 0x7fffffffffffffffllu; 1165 sectorsize = 0; 1166 for (i = 0; i < numdisks; i++) { 1167 snprintf(arg, sizeof(arg), "arg%d", i + 3); 1168 diskname = gctl_get_asciiparam(req, arg); 1169 if (diskname == NULL) { 1170 gctl_error(req, "No disk name (%s).", arg); 1171 error = -6; 1172 break; 1173 } 1174 if (strcmp(diskname, "NONE") == 0) { 1175 cp = NULL; 1176 pp = NULL; 1177 } else { 1178 g_topology_lock(); 1179 cp = g_raid_open_consumer(sc, diskname); 1180 if (cp == NULL) { 1181 gctl_error(req, "Can't open '%s'.", 1182 diskname); 1183 g_topology_unlock(); 1184 error = -7; 1185 break; 1186 } 1187 pp = cp->provider; 1188 } 1189 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO); 1190 pd->pd_disk_pos = i; 1191 disk = g_raid_create_disk(sc); 1192 disk->d_md_data = (void *)pd; 1193 disk->d_consumer = cp; 1194 if (cp == NULL) 1195 continue; 1196 cp->private = disk; 1197 g_topology_unlock(); 1198 1199 g_raid_get_disk_info(disk); 1200 1201 pd->pd_disk_size = pp->mediasize; 1202 if (size > pp->mediasize) 1203 size = pp->mediasize; 1204 if (sectorsize < pp->sectorsize) 1205 sectorsize = pp->sectorsize; 1206 } 1207 if (error != 0) 1208 return (error); 1209 1210 if (sectorsize <= 0) { 1211 gctl_error(req, "Can't get sector size."); 1212 return (-8); 1213 } 1214 1215 /* Reserve space for metadata. */ 1216 size -= 0x800 * sectorsize; 1217 1218 /* Handle size argument. */ 1219 len = sizeof(*sizearg); 1220 sizearg = gctl_get_param(req, "size", &len); 1221 if (sizearg != NULL && len == sizeof(*sizearg) && 1222 *sizearg > 0) { 1223 if (*sizearg > size) { 1224 gctl_error(req, "Size too big %lld > %lld.", 1225 (long long)*sizearg, (long long)size); 1226 return (-9); 1227 } 1228 size = *sizearg; 1229 } 1230 1231 /* Handle strip argument. */ 1232 strip = 131072; 1233 len = sizeof(*striparg); 1234 striparg = gctl_get_param(req, "strip", &len); 1235 if (striparg != NULL && len == sizeof(*striparg) && 1236 *striparg > 0) { 1237 if (*striparg < sectorsize) { 1238 gctl_error(req, "Strip size too small."); 1239 return (-10); 1240 } 1241 if (*striparg % sectorsize != 0) { 1242 gctl_error(req, "Incorrect strip size."); 1243 return (-11); 1244 } 1245 if (strip > 65535 * sectorsize) { 1246 gctl_error(req, "Strip size too big."); 1247 return (-12); 1248 } 1249 strip = *striparg; 1250 } 1251 1252 /* Round size down to strip or sector. */ 1253 if (level == G_RAID_VOLUME_RL_RAID1) 1254 size -= (size % sectorsize); 1255 else if (level == G_RAID_VOLUME_RL_RAID1E && 1256 (numdisks & 1) != 0) 1257 size -= (size % (2 * strip)); 1258 else 1259 size -= (size % strip); 1260 if (size <= 0) { 1261 gctl_error(req, "Size too small."); 1262 return (-13); 1263 } 1264 if (size > 0xffffffffffffllu * sectorsize) { 1265 gctl_error(req, "Size too big."); 1266 return (-14); 1267 } 1268 1269 /* We have all we need, create things: volume, ... */ 1270 mdi->mdio_total_disks = numdisks; 1271 mdi->mdio_started = 1; 1272 vol = g_raid_create_volume(sc, volname, -1); 1273 vol->v_md_data = (void *)(intptr_t)0; 1274 vol->v_raid_level = level; 1275 vol->v_raid_level_qualifier = qual; 1276 vol->v_strip_size = strip; 1277 vol->v_disks_count = numdisks; 1278 if (level == G_RAID_VOLUME_RL_RAID0 || 1279 level == G_RAID_VOLUME_RL_CONCAT || 1280 level == G_RAID_VOLUME_RL_SINGLE) 1281 vol->v_mediasize = size * numdisks; 1282 else if (level == G_RAID_VOLUME_RL_RAID1) 1283 vol->v_mediasize = size; 1284 else if (level == G_RAID_VOLUME_RL_RAID5) 1285 vol->v_mediasize = size * (numdisks - 1); 1286 else { /* RAID1E */ 1287 vol->v_mediasize = ((size * numdisks) / strip / 2) * 1288 strip; 1289 } 1290 vol->v_sectorsize = sectorsize; 1291 g_raid_start_volume(vol); 1292 1293 /* , and subdisks. */ 1294 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1295 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1296 sd = &vol->v_subdisks[pd->pd_disk_pos]; 1297 sd->sd_disk = disk; 1298 sd->sd_offset = 0; 1299 sd->sd_size = size; 1300 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 1301 if (sd->sd_disk->d_consumer != NULL) { 1302 g_raid_change_disk_state(disk, 1303 G_RAID_DISK_S_ACTIVE); 1304 g_raid_change_subdisk_state(sd, 1305 G_RAID_SUBDISK_S_ACTIVE); 1306 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 1307 G_RAID_EVENT_SUBDISK); 1308 } else { 1309 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1310 } 1311 } 1312 1313 /* Write metadata based on created entities. */ 1314 G_RAID_DEBUG1(0, sc, "Array started."); 1315 g_raid_md_write_sii(md, NULL, NULL, NULL); 1316 1317 /* Pickup any STALE/SPARE disks to refill array if needed. */ 1318 g_raid_md_sii_refill(sc); 1319 1320 g_raid_event_send(vol, G_RAID_VOLUME_E_START, 1321 G_RAID_EVENT_VOLUME); 1322 return (0); 1323 } 1324 if (strcmp(verb, "delete") == 0) { 1325 1326 /* Check if some volume is still open. */ 1327 force = gctl_get_paraml(req, "force", sizeof(*force)); 1328 if (force != NULL && *force == 0 && 1329 g_raid_nopens(sc) != 0) { 1330 gctl_error(req, "Some volume is still open."); 1331 return (-4); 1332 } 1333 1334 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1335 if (disk->d_consumer) 1336 sii_meta_erase(disk->d_consumer); 1337 } 1338 g_raid_destroy_node(sc, 0); 1339 return (0); 1340 } 1341 if (strcmp(verb, "remove") == 0 || 1342 strcmp(verb, "fail") == 0) { 1343 if (*nargs < 2) { 1344 gctl_error(req, "Invalid number of arguments."); 1345 return (-1); 1346 } 1347 for (i = 1; i < *nargs; i++) { 1348 snprintf(arg, sizeof(arg), "arg%d", i); 1349 diskname = gctl_get_asciiparam(req, arg); 1350 if (diskname == NULL) { 1351 gctl_error(req, "No disk name (%s).", arg); 1352 error = -2; 1353 break; 1354 } 1355 if (strncmp(diskname, "/dev/", 5) == 0) 1356 diskname += 5; 1357 1358 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1359 if (disk->d_consumer != NULL && 1360 disk->d_consumer->provider != NULL && 1361 strcmp(disk->d_consumer->provider->name, 1362 diskname) == 0) 1363 break; 1364 } 1365 if (disk == NULL) { 1366 gctl_error(req, "Disk '%s' not found.", 1367 diskname); 1368 error = -3; 1369 break; 1370 } 1371 1372 if (strcmp(verb, "fail") == 0) { 1373 g_raid_md_fail_disk_sii(md, NULL, disk); 1374 continue; 1375 } 1376 1377 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1378 1379 /* Erase metadata on deleting disk. */ 1380 sii_meta_erase(disk->d_consumer); 1381 1382 /* If disk was assigned, just update statuses. */ 1383 if (pd->pd_disk_pos >= 0) { 1384 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1385 g_raid_kill_consumer(sc, disk->d_consumer); 1386 disk->d_consumer = NULL; 1387 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 1388 g_raid_change_subdisk_state(sd, 1389 G_RAID_SUBDISK_S_NONE); 1390 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED, 1391 G_RAID_EVENT_SUBDISK); 1392 } 1393 } else { 1394 /* Otherwise -- delete. */ 1395 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE); 1396 g_raid_destroy_disk(disk); 1397 } 1398 } 1399 1400 /* Write updated metadata to remaining disks. */ 1401 g_raid_md_write_sii(md, NULL, NULL, NULL); 1402 1403 /* Check if anything left except placeholders. */ 1404 if (g_raid_ndisks(sc, -1) == 1405 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 1406 g_raid_destroy_node(sc, 0); 1407 else 1408 g_raid_md_sii_refill(sc); 1409 return (error); 1410 } 1411 if (strcmp(verb, "insert") == 0) { 1412 if (*nargs < 2) { 1413 gctl_error(req, "Invalid number of arguments."); 1414 return (-1); 1415 } 1416 update = 0; 1417 for (i = 1; i < *nargs; i++) { 1418 /* Get disk name. */ 1419 snprintf(arg, sizeof(arg), "arg%d", i); 1420 diskname = gctl_get_asciiparam(req, arg); 1421 if (diskname == NULL) { 1422 gctl_error(req, "No disk name (%s).", arg); 1423 error = -3; 1424 break; 1425 } 1426 1427 /* Try to find provider with specified name. */ 1428 g_topology_lock(); 1429 cp = g_raid_open_consumer(sc, diskname); 1430 if (cp == NULL) { 1431 gctl_error(req, "Can't open disk '%s'.", 1432 diskname); 1433 g_topology_unlock(); 1434 error = -4; 1435 break; 1436 } 1437 pp = cp->provider; 1438 1439 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO); 1440 pd->pd_disk_pos = -3; 1441 pd->pd_disk_size = pp->mediasize; 1442 1443 disk = g_raid_create_disk(sc); 1444 disk->d_consumer = cp; 1445 disk->d_md_data = (void *)pd; 1446 cp->private = disk; 1447 g_topology_unlock(); 1448 1449 g_raid_get_disk_info(disk); 1450 1451 /* Welcome the "new" disk. */ 1452 update += g_raid_md_sii_start_disk(disk); 1453 if (disk->d_state == G_RAID_DISK_S_SPARE) { 1454 sii_meta_write_spare(cp); 1455 g_raid_destroy_disk(disk); 1456 } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) { 1457 gctl_error(req, "Disk '%s' doesn't fit.", 1458 diskname); 1459 g_raid_destroy_disk(disk); 1460 error = -8; 1461 break; 1462 } 1463 } 1464 1465 /* Write new metadata if we changed something. */ 1466 if (update) 1467 g_raid_md_write_sii(md, NULL, NULL, NULL); 1468 return (error); 1469 } 1470 gctl_error(req, "Command '%s' is not supported.", verb); 1471 return (-100); 1472 } 1473 1474 static int 1475 g_raid_md_write_sii(struct g_raid_md_object *md, struct g_raid_volume *tvol, 1476 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk) 1477 { 1478 struct g_raid_softc *sc; 1479 struct g_raid_volume *vol; 1480 struct g_raid_subdisk *sd; 1481 struct g_raid_disk *disk; 1482 struct g_raid_md_sii_object *mdi; 1483 struct g_raid_md_sii_perdisk *pd; 1484 struct sii_raid_conf *meta; 1485 int i; 1486 1487 sc = md->mdo_softc; 1488 mdi = (struct g_raid_md_sii_object *)md; 1489 1490 if (sc->sc_stopping == G_RAID_DESTROY_HARD) 1491 return (0); 1492 1493 /* Bump generation. Newly written metadata may differ from previous. */ 1494 mdi->mdio_generation++; 1495 1496 /* There is only one volume. */ 1497 vol = TAILQ_FIRST(&sc->sc_volumes); 1498 1499 /* Fill global fields. */ 1500 meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK | M_ZERO); 1501 if (mdi->mdio_meta) 1502 memcpy(meta, mdi->mdio_meta, sizeof(*meta)); 1503 meta->total_sectors = vol->v_mediasize / vol->v_sectorsize; 1504 meta->vendor_id = 0x1095; 1505 meta->version_minor = 0; 1506 meta->version_major = 2; 1507 memcpy(&meta->timestamp, &mdi->mdio_timestamp, 6); 1508 meta->strip_sectors = vol->v_strip_size / vol->v_sectorsize; 1509 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0) { 1510 meta->type = SII_T_RAID0; 1511 meta->raid0_disks = vol->v_disks_count; 1512 meta->raid1_disks = 0xff; 1513 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) { 1514 meta->type = SII_T_RAID1; 1515 meta->raid0_disks = 0xff; 1516 meta->raid1_disks = vol->v_disks_count; 1517 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) { 1518 meta->type = SII_T_RAID01; 1519 meta->raid0_disks = vol->v_disks_count / 2; 1520 meta->raid1_disks = 2; 1521 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT || 1522 vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE) { 1523 meta->type = SII_T_JBOD; 1524 meta->raid0_disks = vol->v_disks_count; 1525 meta->raid1_disks = 0xff; 1526 } else { 1527 meta->type = SII_T_RAID5; 1528 meta->raid0_disks = vol->v_disks_count; 1529 meta->raid1_disks = 0xff; 1530 } 1531 meta->generation = mdi->mdio_generation; 1532 meta->raid_status = vol->v_dirty ? SII_S_ONLINE : SII_S_AVAILABLE; 1533 for (i = 0; i < vol->v_disks_count; i++) { 1534 sd = &vol->v_subdisks[i]; 1535 if (sd->sd_state == G_RAID_SUBDISK_S_STALE || 1536 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) 1537 meta->raid_status = SII_S_ONLINE; 1538 } 1539 meta->raid_location = mdi->mdio_location; 1540 sii_meta_put_name(meta, vol->v_name); 1541 1542 /* We are done. Print meta data and store them to disks. */ 1543 if (mdi->mdio_meta != NULL) 1544 free(mdi->mdio_meta, M_MD_SII); 1545 mdi->mdio_meta = meta; 1546 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1547 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1548 if (disk->d_state != G_RAID_DISK_S_ACTIVE) 1549 continue; 1550 if (pd->pd_meta != NULL) { 1551 free(pd->pd_meta, M_MD_SII); 1552 pd->pd_meta = NULL; 1553 } 1554 pd->pd_meta = sii_meta_copy(meta); 1555 if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) { 1556 if (sd->sd_state < G_RAID_SUBDISK_S_NEW) 1557 pd->pd_meta->disk_status = SII_S_DROPPED; 1558 else if (sd->sd_state < G_RAID_SUBDISK_S_STALE) { 1559 pd->pd_meta->disk_status = SII_S_REBUILD; 1560 pd->pd_meta->rebuild_lba = 1561 sd->sd_rebuild_pos / vol->v_sectorsize; 1562 } else 1563 pd->pd_meta->disk_status = SII_S_CURRENT; 1564 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) { 1565 pd->pd_meta->disk_number = sd->sd_pos; 1566 pd->pd_meta->raid0_ident = 0xff; 1567 pd->pd_meta->raid1_ident = 0; 1568 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) { 1569 pd->pd_meta->disk_number = sd->sd_pos / meta->raid1_disks; 1570 pd->pd_meta->raid0_ident = sd->sd_pos % meta->raid1_disks; 1571 pd->pd_meta->raid1_ident = sd->sd_pos / meta->raid1_disks; 1572 } else { 1573 pd->pd_meta->disk_number = sd->sd_pos; 1574 pd->pd_meta->raid0_ident = 0; 1575 pd->pd_meta->raid1_ident = 0xff; 1576 } 1577 } 1578 G_RAID_DEBUG(1, "Writing SiI metadata to %s", 1579 g_raid_get_diskname(disk)); 1580 g_raid_md_sii_print(pd->pd_meta); 1581 sii_meta_write(disk->d_consumer, pd->pd_meta); 1582 } 1583 return (0); 1584 } 1585 1586 static int 1587 g_raid_md_fail_disk_sii(struct g_raid_md_object *md, 1588 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk) 1589 { 1590 struct g_raid_softc *sc; 1591 struct g_raid_md_sii_perdisk *pd; 1592 struct g_raid_subdisk *sd; 1593 1594 sc = md->mdo_softc; 1595 pd = (struct g_raid_md_sii_perdisk *)tdisk->d_md_data; 1596 1597 /* We can't fail disk that is not a part of array now. */ 1598 if (pd->pd_disk_pos < 0) 1599 return (-1); 1600 1601 /* 1602 * Mark disk as failed in metadata and try to write that metadata 1603 * to the disk itself to prevent it's later resurrection as STALE. 1604 */ 1605 if (tdisk->d_consumer) { 1606 if (pd->pd_meta) { 1607 pd->pd_meta->disk_status = SII_S_REMOVED; 1608 sii_meta_write(tdisk->d_consumer, pd->pd_meta); 1609 } else 1610 sii_meta_erase(tdisk->d_consumer); 1611 } 1612 1613 /* Change states. */ 1614 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED); 1615 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) { 1616 g_raid_change_subdisk_state(sd, 1617 G_RAID_SUBDISK_S_FAILED); 1618 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED, 1619 G_RAID_EVENT_SUBDISK); 1620 } 1621 1622 /* Write updated metadata to remaining disks. */ 1623 g_raid_md_write_sii(md, NULL, NULL, tdisk); 1624 1625 /* Check if anything left except placeholders. */ 1626 if (g_raid_ndisks(sc, -1) == 1627 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 1628 g_raid_destroy_node(sc, 0); 1629 else 1630 g_raid_md_sii_refill(sc); 1631 return (0); 1632 } 1633 1634 static int 1635 g_raid_md_free_disk_sii(struct g_raid_md_object *md, 1636 struct g_raid_disk *disk) 1637 { 1638 struct g_raid_md_sii_perdisk *pd; 1639 1640 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1641 if (pd->pd_meta != NULL) { 1642 free(pd->pd_meta, M_MD_SII); 1643 pd->pd_meta = NULL; 1644 } 1645 free(pd, M_MD_SII); 1646 disk->d_md_data = NULL; 1647 return (0); 1648 } 1649 1650 static int 1651 g_raid_md_free_sii(struct g_raid_md_object *md) 1652 { 1653 struct g_raid_md_sii_object *mdi; 1654 1655 mdi = (struct g_raid_md_sii_object *)md; 1656 if (!mdi->mdio_started) { 1657 mdi->mdio_started = 0; 1658 callout_stop(&mdi->mdio_start_co); 1659 G_RAID_DEBUG1(1, md->mdo_softc, 1660 "root_mount_rel %p", mdi->mdio_rootmount); 1661 root_mount_rel(mdi->mdio_rootmount); 1662 mdi->mdio_rootmount = NULL; 1663 } 1664 if (mdi->mdio_meta != NULL) { 1665 free(mdi->mdio_meta, M_MD_SII); 1666 mdi->mdio_meta = NULL; 1667 } 1668 return (0); 1669 } 1670 1671 G_RAID_MD_DECLARE(sii, "SiI"); 1672