1 /*- 2 * Copyright (c) 2004-2006 Pawel Jakub Dawidek <pjd@FreeBSD.org> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include <sys/param.h> 31 #include <sys/systm.h> 32 #include <sys/kernel.h> 33 #include <sys/module.h> 34 #include <sys/limits.h> 35 #include <sys/lock.h> 36 #include <sys/mutex.h> 37 #include <sys/bio.h> 38 #include <sys/sbuf.h> 39 #include <sys/sysctl.h> 40 #include <sys/malloc.h> 41 #include <sys/eventhandler.h> 42 #include <vm/uma.h> 43 #include <geom/geom.h> 44 #include <sys/proc.h> 45 #include <sys/kthread.h> 46 #include <sys/sched.h> 47 #include <geom/raid3/g_raid3.h> 48 49 FEATURE(geom_raid3, "GEOM RAID-3 functionality"); 50 51 static MALLOC_DEFINE(M_RAID3, "raid3_data", "GEOM_RAID3 Data"); 52 53 SYSCTL_DECL(_kern_geom); 54 static SYSCTL_NODE(_kern_geom, OID_AUTO, raid3, CTLFLAG_RW, 0, 55 "GEOM_RAID3 stuff"); 56 u_int g_raid3_debug = 0; 57 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, debug, CTLFLAG_RWTUN, &g_raid3_debug, 0, 58 "Debug level"); 59 static u_int g_raid3_timeout = 4; 60 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, timeout, CTLFLAG_RWTUN, &g_raid3_timeout, 61 0, "Time to wait on all raid3 components"); 62 static u_int g_raid3_idletime = 5; 63 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, idletime, CTLFLAG_RWTUN, 64 &g_raid3_idletime, 0, "Mark components as clean when idling"); 65 static u_int g_raid3_disconnect_on_failure = 1; 66 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, disconnect_on_failure, CTLFLAG_RWTUN, 67 &g_raid3_disconnect_on_failure, 0, "Disconnect component on I/O failure."); 68 static u_int g_raid3_syncreqs = 2; 69 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, sync_requests, CTLFLAG_RDTUN, 70 &g_raid3_syncreqs, 0, "Parallel synchronization I/O requests."); 71 static u_int g_raid3_use_malloc = 0; 72 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, use_malloc, CTLFLAG_RDTUN, 73 &g_raid3_use_malloc, 0, "Use malloc(9) instead of uma(9)."); 74 75 static u_int g_raid3_n64k = 50; 76 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, n64k, CTLFLAG_RDTUN, &g_raid3_n64k, 0, 77 "Maximum number of 64kB allocations"); 78 static u_int g_raid3_n16k = 200; 79 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, n16k, CTLFLAG_RDTUN, &g_raid3_n16k, 0, 80 "Maximum number of 16kB allocations"); 81 static u_int g_raid3_n4k = 1200; 82 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, n4k, CTLFLAG_RDTUN, &g_raid3_n4k, 0, 83 "Maximum number of 4kB allocations"); 84 85 static SYSCTL_NODE(_kern_geom_raid3, OID_AUTO, stat, CTLFLAG_RW, 0, 86 "GEOM_RAID3 statistics"); 87 static u_int g_raid3_parity_mismatch = 0; 88 SYSCTL_UINT(_kern_geom_raid3_stat, OID_AUTO, parity_mismatch, CTLFLAG_RD, 89 &g_raid3_parity_mismatch, 0, "Number of failures in VERIFY mode"); 90 91 #define MSLEEP(ident, mtx, priority, wmesg, timeout) do { \ 92 G_RAID3_DEBUG(4, "%s: Sleeping %p.", __func__, (ident)); \ 93 msleep((ident), (mtx), (priority), (wmesg), (timeout)); \ 94 G_RAID3_DEBUG(4, "%s: Woken up %p.", __func__, (ident)); \ 95 } while (0) 96 97 static eventhandler_tag g_raid3_post_sync = NULL; 98 static int g_raid3_shutdown = 0; 99 100 static int g_raid3_destroy_geom(struct gctl_req *req, struct g_class *mp, 101 struct g_geom *gp); 102 static g_taste_t g_raid3_taste; 103 static void g_raid3_init(struct g_class *mp); 104 static void g_raid3_fini(struct g_class *mp); 105 106 struct g_class g_raid3_class = { 107 .name = G_RAID3_CLASS_NAME, 108 .version = G_VERSION, 109 .ctlreq = g_raid3_config, 110 .taste = g_raid3_taste, 111 .destroy_geom = g_raid3_destroy_geom, 112 .init = g_raid3_init, 113 .fini = g_raid3_fini 114 }; 115 116 117 static void g_raid3_destroy_provider(struct g_raid3_softc *sc); 118 static int g_raid3_update_disk(struct g_raid3_disk *disk, u_int state); 119 static void g_raid3_update_device(struct g_raid3_softc *sc, boolean_t force); 120 static void g_raid3_dumpconf(struct sbuf *sb, const char *indent, 121 struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp); 122 static void g_raid3_sync_stop(struct g_raid3_softc *sc, int type); 123 static int g_raid3_register_request(struct bio *pbp); 124 static void g_raid3_sync_release(struct g_raid3_softc *sc); 125 126 127 static const char * 128 g_raid3_disk_state2str(int state) 129 { 130 131 switch (state) { 132 case G_RAID3_DISK_STATE_NODISK: 133 return ("NODISK"); 134 case G_RAID3_DISK_STATE_NONE: 135 return ("NONE"); 136 case G_RAID3_DISK_STATE_NEW: 137 return ("NEW"); 138 case G_RAID3_DISK_STATE_ACTIVE: 139 return ("ACTIVE"); 140 case G_RAID3_DISK_STATE_STALE: 141 return ("STALE"); 142 case G_RAID3_DISK_STATE_SYNCHRONIZING: 143 return ("SYNCHRONIZING"); 144 case G_RAID3_DISK_STATE_DISCONNECTED: 145 return ("DISCONNECTED"); 146 default: 147 return ("INVALID"); 148 } 149 } 150 151 static const char * 152 g_raid3_device_state2str(int state) 153 { 154 155 switch (state) { 156 case G_RAID3_DEVICE_STATE_STARTING: 157 return ("STARTING"); 158 case G_RAID3_DEVICE_STATE_DEGRADED: 159 return ("DEGRADED"); 160 case G_RAID3_DEVICE_STATE_COMPLETE: 161 return ("COMPLETE"); 162 default: 163 return ("INVALID"); 164 } 165 } 166 167 const char * 168 g_raid3_get_diskname(struct g_raid3_disk *disk) 169 { 170 171 if (disk->d_consumer == NULL || disk->d_consumer->provider == NULL) 172 return ("[unknown]"); 173 return (disk->d_name); 174 } 175 176 static void * 177 g_raid3_alloc(struct g_raid3_softc *sc, size_t size, int flags) 178 { 179 void *ptr; 180 enum g_raid3_zones zone; 181 182 if (g_raid3_use_malloc || 183 (zone = g_raid3_zone(size)) == G_RAID3_NUM_ZONES) 184 ptr = malloc(size, M_RAID3, flags); 185 else { 186 ptr = uma_zalloc_arg(sc->sc_zones[zone].sz_zone, 187 &sc->sc_zones[zone], flags); 188 sc->sc_zones[zone].sz_requested++; 189 if (ptr == NULL) 190 sc->sc_zones[zone].sz_failed++; 191 } 192 return (ptr); 193 } 194 195 static void 196 g_raid3_free(struct g_raid3_softc *sc, void *ptr, size_t size) 197 { 198 enum g_raid3_zones zone; 199 200 if (g_raid3_use_malloc || 201 (zone = g_raid3_zone(size)) == G_RAID3_NUM_ZONES) 202 free(ptr, M_RAID3); 203 else { 204 uma_zfree_arg(sc->sc_zones[zone].sz_zone, 205 ptr, &sc->sc_zones[zone]); 206 } 207 } 208 209 static int 210 g_raid3_uma_ctor(void *mem, int size, void *arg, int flags) 211 { 212 struct g_raid3_zone *sz = arg; 213 214 if (sz->sz_max > 0 && sz->sz_inuse == sz->sz_max) 215 return (ENOMEM); 216 sz->sz_inuse++; 217 return (0); 218 } 219 220 static void 221 g_raid3_uma_dtor(void *mem, int size, void *arg) 222 { 223 struct g_raid3_zone *sz = arg; 224 225 sz->sz_inuse--; 226 } 227 228 #define g_raid3_xor(src, dst, size) \ 229 _g_raid3_xor((uint64_t *)(src), \ 230 (uint64_t *)(dst), (size_t)size) 231 static void 232 _g_raid3_xor(uint64_t *src, uint64_t *dst, size_t size) 233 { 234 235 KASSERT((size % 128) == 0, ("Invalid size: %zu.", size)); 236 for (; size > 0; size -= 128) { 237 *dst++ ^= (*src++); 238 *dst++ ^= (*src++); 239 *dst++ ^= (*src++); 240 *dst++ ^= (*src++); 241 *dst++ ^= (*src++); 242 *dst++ ^= (*src++); 243 *dst++ ^= (*src++); 244 *dst++ ^= (*src++); 245 *dst++ ^= (*src++); 246 *dst++ ^= (*src++); 247 *dst++ ^= (*src++); 248 *dst++ ^= (*src++); 249 *dst++ ^= (*src++); 250 *dst++ ^= (*src++); 251 *dst++ ^= (*src++); 252 *dst++ ^= (*src++); 253 } 254 } 255 256 static int 257 g_raid3_is_zero(struct bio *bp) 258 { 259 static const uint64_t zeros[] = { 260 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 261 }; 262 u_char *addr; 263 ssize_t size; 264 265 size = bp->bio_length; 266 addr = (u_char *)bp->bio_data; 267 for (; size > 0; size -= sizeof(zeros), addr += sizeof(zeros)) { 268 if (bcmp(addr, zeros, sizeof(zeros)) != 0) 269 return (0); 270 } 271 return (1); 272 } 273 274 /* 275 * --- Events handling functions --- 276 * Events in geom_raid3 are used to maintain disks and device status 277 * from one thread to simplify locking. 278 */ 279 static void 280 g_raid3_event_free(struct g_raid3_event *ep) 281 { 282 283 free(ep, M_RAID3); 284 } 285 286 int 287 g_raid3_event_send(void *arg, int state, int flags) 288 { 289 struct g_raid3_softc *sc; 290 struct g_raid3_disk *disk; 291 struct g_raid3_event *ep; 292 int error; 293 294 ep = malloc(sizeof(*ep), M_RAID3, M_WAITOK); 295 G_RAID3_DEBUG(4, "%s: Sending event %p.", __func__, ep); 296 if ((flags & G_RAID3_EVENT_DEVICE) != 0) { 297 disk = NULL; 298 sc = arg; 299 } else { 300 disk = arg; 301 sc = disk->d_softc; 302 } 303 ep->e_disk = disk; 304 ep->e_state = state; 305 ep->e_flags = flags; 306 ep->e_error = 0; 307 mtx_lock(&sc->sc_events_mtx); 308 TAILQ_INSERT_TAIL(&sc->sc_events, ep, e_next); 309 mtx_unlock(&sc->sc_events_mtx); 310 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, sc); 311 mtx_lock(&sc->sc_queue_mtx); 312 wakeup(sc); 313 wakeup(&sc->sc_queue); 314 mtx_unlock(&sc->sc_queue_mtx); 315 if ((flags & G_RAID3_EVENT_DONTWAIT) != 0) 316 return (0); 317 sx_assert(&sc->sc_lock, SX_XLOCKED); 318 G_RAID3_DEBUG(4, "%s: Sleeping %p.", __func__, ep); 319 sx_xunlock(&sc->sc_lock); 320 while ((ep->e_flags & G_RAID3_EVENT_DONE) == 0) { 321 mtx_lock(&sc->sc_events_mtx); 322 MSLEEP(ep, &sc->sc_events_mtx, PRIBIO | PDROP, "r3:event", 323 hz * 5); 324 } 325 error = ep->e_error; 326 g_raid3_event_free(ep); 327 sx_xlock(&sc->sc_lock); 328 return (error); 329 } 330 331 static struct g_raid3_event * 332 g_raid3_event_get(struct g_raid3_softc *sc) 333 { 334 struct g_raid3_event *ep; 335 336 mtx_lock(&sc->sc_events_mtx); 337 ep = TAILQ_FIRST(&sc->sc_events); 338 mtx_unlock(&sc->sc_events_mtx); 339 return (ep); 340 } 341 342 static void 343 g_raid3_event_remove(struct g_raid3_softc *sc, struct g_raid3_event *ep) 344 { 345 346 mtx_lock(&sc->sc_events_mtx); 347 TAILQ_REMOVE(&sc->sc_events, ep, e_next); 348 mtx_unlock(&sc->sc_events_mtx); 349 } 350 351 static void 352 g_raid3_event_cancel(struct g_raid3_disk *disk) 353 { 354 struct g_raid3_softc *sc; 355 struct g_raid3_event *ep, *tmpep; 356 357 sc = disk->d_softc; 358 sx_assert(&sc->sc_lock, SX_XLOCKED); 359 360 mtx_lock(&sc->sc_events_mtx); 361 TAILQ_FOREACH_SAFE(ep, &sc->sc_events, e_next, tmpep) { 362 if ((ep->e_flags & G_RAID3_EVENT_DEVICE) != 0) 363 continue; 364 if (ep->e_disk != disk) 365 continue; 366 TAILQ_REMOVE(&sc->sc_events, ep, e_next); 367 if ((ep->e_flags & G_RAID3_EVENT_DONTWAIT) != 0) 368 g_raid3_event_free(ep); 369 else { 370 ep->e_error = ECANCELED; 371 wakeup(ep); 372 } 373 } 374 mtx_unlock(&sc->sc_events_mtx); 375 } 376 377 /* 378 * Return the number of disks in the given state. 379 * If state is equal to -1, count all connected disks. 380 */ 381 u_int 382 g_raid3_ndisks(struct g_raid3_softc *sc, int state) 383 { 384 struct g_raid3_disk *disk; 385 u_int n, ndisks; 386 387 sx_assert(&sc->sc_lock, SX_LOCKED); 388 389 for (n = ndisks = 0; n < sc->sc_ndisks; n++) { 390 disk = &sc->sc_disks[n]; 391 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 392 continue; 393 if (state == -1 || disk->d_state == state) 394 ndisks++; 395 } 396 return (ndisks); 397 } 398 399 static u_int 400 g_raid3_nrequests(struct g_raid3_softc *sc, struct g_consumer *cp) 401 { 402 struct bio *bp; 403 u_int nreqs = 0; 404 405 mtx_lock(&sc->sc_queue_mtx); 406 TAILQ_FOREACH(bp, &sc->sc_queue.queue, bio_queue) { 407 if (bp->bio_from == cp) 408 nreqs++; 409 } 410 mtx_unlock(&sc->sc_queue_mtx); 411 return (nreqs); 412 } 413 414 static int 415 g_raid3_is_busy(struct g_raid3_softc *sc, struct g_consumer *cp) 416 { 417 418 if (cp->index > 0) { 419 G_RAID3_DEBUG(2, 420 "I/O requests for %s exist, can't destroy it now.", 421 cp->provider->name); 422 return (1); 423 } 424 if (g_raid3_nrequests(sc, cp) > 0) { 425 G_RAID3_DEBUG(2, 426 "I/O requests for %s in queue, can't destroy it now.", 427 cp->provider->name); 428 return (1); 429 } 430 return (0); 431 } 432 433 static void 434 g_raid3_destroy_consumer(void *arg, int flags __unused) 435 { 436 struct g_consumer *cp; 437 438 g_topology_assert(); 439 440 cp = arg; 441 G_RAID3_DEBUG(1, "Consumer %s destroyed.", cp->provider->name); 442 g_detach(cp); 443 g_destroy_consumer(cp); 444 } 445 446 static void 447 g_raid3_kill_consumer(struct g_raid3_softc *sc, struct g_consumer *cp) 448 { 449 struct g_provider *pp; 450 int retaste_wait; 451 452 g_topology_assert(); 453 454 cp->private = NULL; 455 if (g_raid3_is_busy(sc, cp)) 456 return; 457 G_RAID3_DEBUG(2, "Consumer %s destroyed.", cp->provider->name); 458 pp = cp->provider; 459 retaste_wait = 0; 460 if (cp->acw == 1) { 461 if ((pp->geom->flags & G_GEOM_WITHER) == 0) 462 retaste_wait = 1; 463 } 464 G_RAID3_DEBUG(2, "Access %s r%dw%de%d = %d", pp->name, -cp->acr, 465 -cp->acw, -cp->ace, 0); 466 if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0) 467 g_access(cp, -cp->acr, -cp->acw, -cp->ace); 468 if (retaste_wait) { 469 /* 470 * After retaste event was send (inside g_access()), we can send 471 * event to detach and destroy consumer. 472 * A class, which has consumer to the given provider connected 473 * will not receive retaste event for the provider. 474 * This is the way how I ignore retaste events when I close 475 * consumers opened for write: I detach and destroy consumer 476 * after retaste event is sent. 477 */ 478 g_post_event(g_raid3_destroy_consumer, cp, M_WAITOK, NULL); 479 return; 480 } 481 G_RAID3_DEBUG(1, "Consumer %s destroyed.", pp->name); 482 g_detach(cp); 483 g_destroy_consumer(cp); 484 } 485 486 static int 487 g_raid3_connect_disk(struct g_raid3_disk *disk, struct g_provider *pp) 488 { 489 struct g_consumer *cp; 490 int error; 491 492 g_topology_assert_not(); 493 KASSERT(disk->d_consumer == NULL, 494 ("Disk already connected (device %s).", disk->d_softc->sc_name)); 495 496 g_topology_lock(); 497 cp = g_new_consumer(disk->d_softc->sc_geom); 498 error = g_attach(cp, pp); 499 if (error != 0) { 500 g_destroy_consumer(cp); 501 g_topology_unlock(); 502 return (error); 503 } 504 error = g_access(cp, 1, 1, 1); 505 g_topology_unlock(); 506 if (error != 0) { 507 g_detach(cp); 508 g_destroy_consumer(cp); 509 G_RAID3_DEBUG(0, "Cannot open consumer %s (error=%d).", 510 pp->name, error); 511 return (error); 512 } 513 disk->d_consumer = cp; 514 disk->d_consumer->private = disk; 515 disk->d_consumer->index = 0; 516 G_RAID3_DEBUG(2, "Disk %s connected.", g_raid3_get_diskname(disk)); 517 return (0); 518 } 519 520 static void 521 g_raid3_disconnect_consumer(struct g_raid3_softc *sc, struct g_consumer *cp) 522 { 523 524 g_topology_assert(); 525 526 if (cp == NULL) 527 return; 528 if (cp->provider != NULL) 529 g_raid3_kill_consumer(sc, cp); 530 else 531 g_destroy_consumer(cp); 532 } 533 534 /* 535 * Initialize disk. This means allocate memory, create consumer, attach it 536 * to the provider and open access (r1w1e1) to it. 537 */ 538 static struct g_raid3_disk * 539 g_raid3_init_disk(struct g_raid3_softc *sc, struct g_provider *pp, 540 struct g_raid3_metadata *md, int *errorp) 541 { 542 struct g_raid3_disk *disk; 543 int error; 544 545 disk = &sc->sc_disks[md->md_no]; 546 error = g_raid3_connect_disk(disk, pp); 547 if (error != 0) { 548 if (errorp != NULL) 549 *errorp = error; 550 return (NULL); 551 } 552 disk->d_state = G_RAID3_DISK_STATE_NONE; 553 disk->d_flags = md->md_dflags; 554 if (md->md_provider[0] != '\0') 555 disk->d_flags |= G_RAID3_DISK_FLAG_HARDCODED; 556 disk->d_sync.ds_consumer = NULL; 557 disk->d_sync.ds_offset = md->md_sync_offset; 558 disk->d_sync.ds_offset_done = md->md_sync_offset; 559 disk->d_genid = md->md_genid; 560 disk->d_sync.ds_syncid = md->md_syncid; 561 if (errorp != NULL) 562 *errorp = 0; 563 return (disk); 564 } 565 566 static void 567 g_raid3_destroy_disk(struct g_raid3_disk *disk) 568 { 569 struct g_raid3_softc *sc; 570 571 g_topology_assert_not(); 572 sc = disk->d_softc; 573 sx_assert(&sc->sc_lock, SX_XLOCKED); 574 575 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 576 return; 577 g_raid3_event_cancel(disk); 578 switch (disk->d_state) { 579 case G_RAID3_DISK_STATE_SYNCHRONIZING: 580 if (sc->sc_syncdisk != NULL) 581 g_raid3_sync_stop(sc, 1); 582 /* FALLTHROUGH */ 583 case G_RAID3_DISK_STATE_NEW: 584 case G_RAID3_DISK_STATE_STALE: 585 case G_RAID3_DISK_STATE_ACTIVE: 586 g_topology_lock(); 587 g_raid3_disconnect_consumer(sc, disk->d_consumer); 588 g_topology_unlock(); 589 disk->d_consumer = NULL; 590 break; 591 default: 592 KASSERT(0 == 1, ("Wrong disk state (%s, %s).", 593 g_raid3_get_diskname(disk), 594 g_raid3_disk_state2str(disk->d_state))); 595 } 596 disk->d_state = G_RAID3_DISK_STATE_NODISK; 597 } 598 599 static void 600 g_raid3_destroy_device(struct g_raid3_softc *sc) 601 { 602 struct g_raid3_event *ep; 603 struct g_raid3_disk *disk; 604 struct g_geom *gp; 605 struct g_consumer *cp; 606 u_int n; 607 608 g_topology_assert_not(); 609 sx_assert(&sc->sc_lock, SX_XLOCKED); 610 611 gp = sc->sc_geom; 612 if (sc->sc_provider != NULL) 613 g_raid3_destroy_provider(sc); 614 for (n = 0; n < sc->sc_ndisks; n++) { 615 disk = &sc->sc_disks[n]; 616 if (disk->d_state != G_RAID3_DISK_STATE_NODISK) { 617 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 618 g_raid3_update_metadata(disk); 619 g_raid3_destroy_disk(disk); 620 } 621 } 622 while ((ep = g_raid3_event_get(sc)) != NULL) { 623 g_raid3_event_remove(sc, ep); 624 if ((ep->e_flags & G_RAID3_EVENT_DONTWAIT) != 0) 625 g_raid3_event_free(ep); 626 else { 627 ep->e_error = ECANCELED; 628 ep->e_flags |= G_RAID3_EVENT_DONE; 629 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, ep); 630 mtx_lock(&sc->sc_events_mtx); 631 wakeup(ep); 632 mtx_unlock(&sc->sc_events_mtx); 633 } 634 } 635 callout_drain(&sc->sc_callout); 636 cp = LIST_FIRST(&sc->sc_sync.ds_geom->consumer); 637 g_topology_lock(); 638 if (cp != NULL) 639 g_raid3_disconnect_consumer(sc, cp); 640 g_wither_geom(sc->sc_sync.ds_geom, ENXIO); 641 G_RAID3_DEBUG(0, "Device %s destroyed.", gp->name); 642 g_wither_geom(gp, ENXIO); 643 g_topology_unlock(); 644 if (!g_raid3_use_malloc) { 645 uma_zdestroy(sc->sc_zones[G_RAID3_ZONE_64K].sz_zone); 646 uma_zdestroy(sc->sc_zones[G_RAID3_ZONE_16K].sz_zone); 647 uma_zdestroy(sc->sc_zones[G_RAID3_ZONE_4K].sz_zone); 648 } 649 mtx_destroy(&sc->sc_queue_mtx); 650 mtx_destroy(&sc->sc_events_mtx); 651 sx_xunlock(&sc->sc_lock); 652 sx_destroy(&sc->sc_lock); 653 } 654 655 static void 656 g_raid3_orphan(struct g_consumer *cp) 657 { 658 struct g_raid3_disk *disk; 659 660 g_topology_assert(); 661 662 disk = cp->private; 663 if (disk == NULL) 664 return; 665 disk->d_softc->sc_bump_id = G_RAID3_BUMP_SYNCID; 666 g_raid3_event_send(disk, G_RAID3_DISK_STATE_DISCONNECTED, 667 G_RAID3_EVENT_DONTWAIT); 668 } 669 670 static int 671 g_raid3_write_metadata(struct g_raid3_disk *disk, struct g_raid3_metadata *md) 672 { 673 struct g_raid3_softc *sc; 674 struct g_consumer *cp; 675 off_t offset, length; 676 u_char *sector; 677 int error = 0; 678 679 g_topology_assert_not(); 680 sc = disk->d_softc; 681 sx_assert(&sc->sc_lock, SX_LOCKED); 682 683 cp = disk->d_consumer; 684 KASSERT(cp != NULL, ("NULL consumer (%s).", sc->sc_name)); 685 KASSERT(cp->provider != NULL, ("NULL provider (%s).", sc->sc_name)); 686 KASSERT(cp->acr >= 1 && cp->acw >= 1 && cp->ace >= 1, 687 ("Consumer %s closed? (r%dw%de%d).", cp->provider->name, cp->acr, 688 cp->acw, cp->ace)); 689 length = cp->provider->sectorsize; 690 offset = cp->provider->mediasize - length; 691 sector = malloc((size_t)length, M_RAID3, M_WAITOK | M_ZERO); 692 if (md != NULL) 693 raid3_metadata_encode(md, sector); 694 error = g_write_data(cp, offset, sector, length); 695 free(sector, M_RAID3); 696 if (error != 0) { 697 if ((disk->d_flags & G_RAID3_DISK_FLAG_BROKEN) == 0) { 698 G_RAID3_DEBUG(0, "Cannot write metadata on %s " 699 "(device=%s, error=%d).", 700 g_raid3_get_diskname(disk), sc->sc_name, error); 701 disk->d_flags |= G_RAID3_DISK_FLAG_BROKEN; 702 } else { 703 G_RAID3_DEBUG(1, "Cannot write metadata on %s " 704 "(device=%s, error=%d).", 705 g_raid3_get_diskname(disk), sc->sc_name, error); 706 } 707 if (g_raid3_disconnect_on_failure && 708 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE) { 709 sc->sc_bump_id |= G_RAID3_BUMP_GENID; 710 g_raid3_event_send(disk, 711 G_RAID3_DISK_STATE_DISCONNECTED, 712 G_RAID3_EVENT_DONTWAIT); 713 } 714 } 715 return (error); 716 } 717 718 int 719 g_raid3_clear_metadata(struct g_raid3_disk *disk) 720 { 721 int error; 722 723 g_topology_assert_not(); 724 sx_assert(&disk->d_softc->sc_lock, SX_LOCKED); 725 726 error = g_raid3_write_metadata(disk, NULL); 727 if (error == 0) { 728 G_RAID3_DEBUG(2, "Metadata on %s cleared.", 729 g_raid3_get_diskname(disk)); 730 } else { 731 G_RAID3_DEBUG(0, 732 "Cannot clear metadata on disk %s (error=%d).", 733 g_raid3_get_diskname(disk), error); 734 } 735 return (error); 736 } 737 738 void 739 g_raid3_fill_metadata(struct g_raid3_disk *disk, struct g_raid3_metadata *md) 740 { 741 struct g_raid3_softc *sc; 742 struct g_provider *pp; 743 744 sc = disk->d_softc; 745 strlcpy(md->md_magic, G_RAID3_MAGIC, sizeof(md->md_magic)); 746 md->md_version = G_RAID3_VERSION; 747 strlcpy(md->md_name, sc->sc_name, sizeof(md->md_name)); 748 md->md_id = sc->sc_id; 749 md->md_all = sc->sc_ndisks; 750 md->md_genid = sc->sc_genid; 751 md->md_mediasize = sc->sc_mediasize; 752 md->md_sectorsize = sc->sc_sectorsize; 753 md->md_mflags = (sc->sc_flags & G_RAID3_DEVICE_FLAG_MASK); 754 md->md_no = disk->d_no; 755 md->md_syncid = disk->d_sync.ds_syncid; 756 md->md_dflags = (disk->d_flags & G_RAID3_DISK_FLAG_MASK); 757 if (disk->d_state != G_RAID3_DISK_STATE_SYNCHRONIZING) 758 md->md_sync_offset = 0; 759 else { 760 md->md_sync_offset = 761 disk->d_sync.ds_offset_done / (sc->sc_ndisks - 1); 762 } 763 if (disk->d_consumer != NULL && disk->d_consumer->provider != NULL) 764 pp = disk->d_consumer->provider; 765 else 766 pp = NULL; 767 if ((disk->d_flags & G_RAID3_DISK_FLAG_HARDCODED) != 0 && pp != NULL) 768 strlcpy(md->md_provider, pp->name, sizeof(md->md_provider)); 769 else 770 bzero(md->md_provider, sizeof(md->md_provider)); 771 if (pp != NULL) 772 md->md_provsize = pp->mediasize; 773 else 774 md->md_provsize = 0; 775 } 776 777 void 778 g_raid3_update_metadata(struct g_raid3_disk *disk) 779 { 780 struct g_raid3_softc *sc; 781 struct g_raid3_metadata md; 782 int error; 783 784 g_topology_assert_not(); 785 sc = disk->d_softc; 786 sx_assert(&sc->sc_lock, SX_LOCKED); 787 788 g_raid3_fill_metadata(disk, &md); 789 error = g_raid3_write_metadata(disk, &md); 790 if (error == 0) { 791 G_RAID3_DEBUG(2, "Metadata on %s updated.", 792 g_raid3_get_diskname(disk)); 793 } else { 794 G_RAID3_DEBUG(0, 795 "Cannot update metadata on disk %s (error=%d).", 796 g_raid3_get_diskname(disk), error); 797 } 798 } 799 800 static void 801 g_raid3_bump_syncid(struct g_raid3_softc *sc) 802 { 803 struct g_raid3_disk *disk; 804 u_int n; 805 806 g_topology_assert_not(); 807 sx_assert(&sc->sc_lock, SX_XLOCKED); 808 KASSERT(g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) > 0, 809 ("%s called with no active disks (device=%s).", __func__, 810 sc->sc_name)); 811 812 sc->sc_syncid++; 813 G_RAID3_DEBUG(1, "Device %s: syncid bumped to %u.", sc->sc_name, 814 sc->sc_syncid); 815 for (n = 0; n < sc->sc_ndisks; n++) { 816 disk = &sc->sc_disks[n]; 817 if (disk->d_state == G_RAID3_DISK_STATE_ACTIVE || 818 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 819 disk->d_sync.ds_syncid = sc->sc_syncid; 820 g_raid3_update_metadata(disk); 821 } 822 } 823 } 824 825 static void 826 g_raid3_bump_genid(struct g_raid3_softc *sc) 827 { 828 struct g_raid3_disk *disk; 829 u_int n; 830 831 g_topology_assert_not(); 832 sx_assert(&sc->sc_lock, SX_XLOCKED); 833 KASSERT(g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) > 0, 834 ("%s called with no active disks (device=%s).", __func__, 835 sc->sc_name)); 836 837 sc->sc_genid++; 838 G_RAID3_DEBUG(1, "Device %s: genid bumped to %u.", sc->sc_name, 839 sc->sc_genid); 840 for (n = 0; n < sc->sc_ndisks; n++) { 841 disk = &sc->sc_disks[n]; 842 if (disk->d_state == G_RAID3_DISK_STATE_ACTIVE || 843 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 844 disk->d_genid = sc->sc_genid; 845 g_raid3_update_metadata(disk); 846 } 847 } 848 } 849 850 static int 851 g_raid3_idle(struct g_raid3_softc *sc, int acw) 852 { 853 struct g_raid3_disk *disk; 854 u_int i; 855 int timeout; 856 857 g_topology_assert_not(); 858 sx_assert(&sc->sc_lock, SX_XLOCKED); 859 860 if (sc->sc_provider == NULL) 861 return (0); 862 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOFAILSYNC) != 0) 863 return (0); 864 if (sc->sc_idle) 865 return (0); 866 if (sc->sc_writes > 0) 867 return (0); 868 if (acw > 0 || (acw == -1 && sc->sc_provider->acw > 0)) { 869 timeout = g_raid3_idletime - (time_uptime - sc->sc_last_write); 870 if (!g_raid3_shutdown && timeout > 0) 871 return (timeout); 872 } 873 sc->sc_idle = 1; 874 for (i = 0; i < sc->sc_ndisks; i++) { 875 disk = &sc->sc_disks[i]; 876 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) 877 continue; 878 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as clean.", 879 g_raid3_get_diskname(disk), sc->sc_name); 880 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 881 g_raid3_update_metadata(disk); 882 } 883 return (0); 884 } 885 886 static void 887 g_raid3_unidle(struct g_raid3_softc *sc) 888 { 889 struct g_raid3_disk *disk; 890 u_int i; 891 892 g_topology_assert_not(); 893 sx_assert(&sc->sc_lock, SX_XLOCKED); 894 895 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOFAILSYNC) != 0) 896 return; 897 sc->sc_idle = 0; 898 sc->sc_last_write = time_uptime; 899 for (i = 0; i < sc->sc_ndisks; i++) { 900 disk = &sc->sc_disks[i]; 901 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) 902 continue; 903 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as dirty.", 904 g_raid3_get_diskname(disk), sc->sc_name); 905 disk->d_flags |= G_RAID3_DISK_FLAG_DIRTY; 906 g_raid3_update_metadata(disk); 907 } 908 } 909 910 /* 911 * Treat bio_driver1 field in parent bio as list head and field bio_caller1 912 * in child bio as pointer to the next element on the list. 913 */ 914 #define G_RAID3_HEAD_BIO(pbp) (pbp)->bio_driver1 915 916 #define G_RAID3_NEXT_BIO(cbp) (cbp)->bio_caller1 917 918 #define G_RAID3_FOREACH_BIO(pbp, bp) \ 919 for ((bp) = G_RAID3_HEAD_BIO(pbp); (bp) != NULL; \ 920 (bp) = G_RAID3_NEXT_BIO(bp)) 921 922 #define G_RAID3_FOREACH_SAFE_BIO(pbp, bp, tmpbp) \ 923 for ((bp) = G_RAID3_HEAD_BIO(pbp); \ 924 (bp) != NULL && ((tmpbp) = G_RAID3_NEXT_BIO(bp), 1); \ 925 (bp) = (tmpbp)) 926 927 static void 928 g_raid3_init_bio(struct bio *pbp) 929 { 930 931 G_RAID3_HEAD_BIO(pbp) = NULL; 932 } 933 934 static void 935 g_raid3_remove_bio(struct bio *cbp) 936 { 937 struct bio *pbp, *bp; 938 939 pbp = cbp->bio_parent; 940 if (G_RAID3_HEAD_BIO(pbp) == cbp) 941 G_RAID3_HEAD_BIO(pbp) = G_RAID3_NEXT_BIO(cbp); 942 else { 943 G_RAID3_FOREACH_BIO(pbp, bp) { 944 if (G_RAID3_NEXT_BIO(bp) == cbp) { 945 G_RAID3_NEXT_BIO(bp) = G_RAID3_NEXT_BIO(cbp); 946 break; 947 } 948 } 949 } 950 G_RAID3_NEXT_BIO(cbp) = NULL; 951 } 952 953 static void 954 g_raid3_replace_bio(struct bio *sbp, struct bio *dbp) 955 { 956 struct bio *pbp, *bp; 957 958 g_raid3_remove_bio(sbp); 959 pbp = dbp->bio_parent; 960 G_RAID3_NEXT_BIO(sbp) = G_RAID3_NEXT_BIO(dbp); 961 if (G_RAID3_HEAD_BIO(pbp) == dbp) 962 G_RAID3_HEAD_BIO(pbp) = sbp; 963 else { 964 G_RAID3_FOREACH_BIO(pbp, bp) { 965 if (G_RAID3_NEXT_BIO(bp) == dbp) { 966 G_RAID3_NEXT_BIO(bp) = sbp; 967 break; 968 } 969 } 970 } 971 G_RAID3_NEXT_BIO(dbp) = NULL; 972 } 973 974 static void 975 g_raid3_destroy_bio(struct g_raid3_softc *sc, struct bio *cbp) 976 { 977 struct bio *bp, *pbp; 978 size_t size; 979 980 pbp = cbp->bio_parent; 981 pbp->bio_children--; 982 KASSERT(cbp->bio_data != NULL, ("NULL bio_data")); 983 size = pbp->bio_length / (sc->sc_ndisks - 1); 984 g_raid3_free(sc, cbp->bio_data, size); 985 if (G_RAID3_HEAD_BIO(pbp) == cbp) { 986 G_RAID3_HEAD_BIO(pbp) = G_RAID3_NEXT_BIO(cbp); 987 G_RAID3_NEXT_BIO(cbp) = NULL; 988 g_destroy_bio(cbp); 989 } else { 990 G_RAID3_FOREACH_BIO(pbp, bp) { 991 if (G_RAID3_NEXT_BIO(bp) == cbp) 992 break; 993 } 994 if (bp != NULL) { 995 KASSERT(G_RAID3_NEXT_BIO(bp) != NULL, 996 ("NULL bp->bio_driver1")); 997 G_RAID3_NEXT_BIO(bp) = G_RAID3_NEXT_BIO(cbp); 998 G_RAID3_NEXT_BIO(cbp) = NULL; 999 } 1000 g_destroy_bio(cbp); 1001 } 1002 } 1003 1004 static struct bio * 1005 g_raid3_clone_bio(struct g_raid3_softc *sc, struct bio *pbp) 1006 { 1007 struct bio *bp, *cbp; 1008 size_t size; 1009 int memflag; 1010 1011 cbp = g_clone_bio(pbp); 1012 if (cbp == NULL) 1013 return (NULL); 1014 size = pbp->bio_length / (sc->sc_ndisks - 1); 1015 if ((pbp->bio_cflags & G_RAID3_BIO_CFLAG_REGULAR) != 0) 1016 memflag = M_WAITOK; 1017 else 1018 memflag = M_NOWAIT; 1019 cbp->bio_data = g_raid3_alloc(sc, size, memflag); 1020 if (cbp->bio_data == NULL) { 1021 pbp->bio_children--; 1022 g_destroy_bio(cbp); 1023 return (NULL); 1024 } 1025 G_RAID3_NEXT_BIO(cbp) = NULL; 1026 if (G_RAID3_HEAD_BIO(pbp) == NULL) 1027 G_RAID3_HEAD_BIO(pbp) = cbp; 1028 else { 1029 G_RAID3_FOREACH_BIO(pbp, bp) { 1030 if (G_RAID3_NEXT_BIO(bp) == NULL) { 1031 G_RAID3_NEXT_BIO(bp) = cbp; 1032 break; 1033 } 1034 } 1035 } 1036 return (cbp); 1037 } 1038 1039 static void 1040 g_raid3_scatter(struct bio *pbp) 1041 { 1042 struct g_raid3_softc *sc; 1043 struct g_raid3_disk *disk; 1044 struct bio *bp, *cbp, *tmpbp; 1045 off_t atom, cadd, padd, left; 1046 int first; 1047 1048 sc = pbp->bio_to->geom->softc; 1049 bp = NULL; 1050 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_NOPARITY) == 0) { 1051 /* 1052 * Find bio for which we should calculate data. 1053 */ 1054 G_RAID3_FOREACH_BIO(pbp, cbp) { 1055 if ((cbp->bio_cflags & G_RAID3_BIO_CFLAG_PARITY) != 0) { 1056 bp = cbp; 1057 break; 1058 } 1059 } 1060 KASSERT(bp != NULL, ("NULL parity bio.")); 1061 } 1062 atom = sc->sc_sectorsize / (sc->sc_ndisks - 1); 1063 cadd = padd = 0; 1064 for (left = pbp->bio_length; left > 0; left -= sc->sc_sectorsize) { 1065 G_RAID3_FOREACH_BIO(pbp, cbp) { 1066 if (cbp == bp) 1067 continue; 1068 bcopy(pbp->bio_data + padd, cbp->bio_data + cadd, atom); 1069 padd += atom; 1070 } 1071 cadd += atom; 1072 } 1073 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_NOPARITY) == 0) { 1074 /* 1075 * Calculate parity. 1076 */ 1077 first = 1; 1078 G_RAID3_FOREACH_SAFE_BIO(pbp, cbp, tmpbp) { 1079 if (cbp == bp) 1080 continue; 1081 if (first) { 1082 bcopy(cbp->bio_data, bp->bio_data, 1083 bp->bio_length); 1084 first = 0; 1085 } else { 1086 g_raid3_xor(cbp->bio_data, bp->bio_data, 1087 bp->bio_length); 1088 } 1089 if ((cbp->bio_cflags & G_RAID3_BIO_CFLAG_NODISK) != 0) 1090 g_raid3_destroy_bio(sc, cbp); 1091 } 1092 } 1093 G_RAID3_FOREACH_SAFE_BIO(pbp, cbp, tmpbp) { 1094 struct g_consumer *cp; 1095 1096 disk = cbp->bio_caller2; 1097 cp = disk->d_consumer; 1098 cbp->bio_to = cp->provider; 1099 G_RAID3_LOGREQ(3, cbp, "Sending request."); 1100 KASSERT(cp->acr >= 1 && cp->acw >= 1 && cp->ace >= 1, 1101 ("Consumer %s not opened (r%dw%de%d).", cp->provider->name, 1102 cp->acr, cp->acw, cp->ace)); 1103 cp->index++; 1104 sc->sc_writes++; 1105 g_io_request(cbp, cp); 1106 } 1107 } 1108 1109 static void 1110 g_raid3_gather(struct bio *pbp) 1111 { 1112 struct g_raid3_softc *sc; 1113 struct g_raid3_disk *disk; 1114 struct bio *xbp, *fbp, *cbp; 1115 off_t atom, cadd, padd, left; 1116 1117 sc = pbp->bio_to->geom->softc; 1118 /* 1119 * Find bio for which we have to calculate data. 1120 * While going through this path, check if all requests 1121 * succeeded, if not, deny whole request. 1122 * If we're in COMPLETE mode, we allow one request to fail, 1123 * so if we find one, we're sending it to the parity consumer. 1124 * If there are more failed requests, we deny whole request. 1125 */ 1126 xbp = fbp = NULL; 1127 G_RAID3_FOREACH_BIO(pbp, cbp) { 1128 if ((cbp->bio_cflags & G_RAID3_BIO_CFLAG_PARITY) != 0) { 1129 KASSERT(xbp == NULL, ("More than one parity bio.")); 1130 xbp = cbp; 1131 } 1132 if (cbp->bio_error == 0) 1133 continue; 1134 /* 1135 * Found failed request. 1136 */ 1137 if (fbp == NULL) { 1138 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_DEGRADED) != 0) { 1139 /* 1140 * We are already in degraded mode, so we can't 1141 * accept any failures. 1142 */ 1143 if (pbp->bio_error == 0) 1144 pbp->bio_error = cbp->bio_error; 1145 } else { 1146 fbp = cbp; 1147 } 1148 } else { 1149 /* 1150 * Next failed request, that's too many. 1151 */ 1152 if (pbp->bio_error == 0) 1153 pbp->bio_error = fbp->bio_error; 1154 } 1155 disk = cbp->bio_caller2; 1156 if (disk == NULL) 1157 continue; 1158 if ((disk->d_flags & G_RAID3_DISK_FLAG_BROKEN) == 0) { 1159 disk->d_flags |= G_RAID3_DISK_FLAG_BROKEN; 1160 G_RAID3_LOGREQ(0, cbp, "Request failed (error=%d).", 1161 cbp->bio_error); 1162 } else { 1163 G_RAID3_LOGREQ(1, cbp, "Request failed (error=%d).", 1164 cbp->bio_error); 1165 } 1166 if (g_raid3_disconnect_on_failure && 1167 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE) { 1168 sc->sc_bump_id |= G_RAID3_BUMP_GENID; 1169 g_raid3_event_send(disk, 1170 G_RAID3_DISK_STATE_DISCONNECTED, 1171 G_RAID3_EVENT_DONTWAIT); 1172 } 1173 } 1174 if (pbp->bio_error != 0) 1175 goto finish; 1176 if (fbp != NULL && (pbp->bio_pflags & G_RAID3_BIO_PFLAG_VERIFY) != 0) { 1177 pbp->bio_pflags &= ~G_RAID3_BIO_PFLAG_VERIFY; 1178 if (xbp != fbp) 1179 g_raid3_replace_bio(xbp, fbp); 1180 g_raid3_destroy_bio(sc, fbp); 1181 } else if (fbp != NULL) { 1182 struct g_consumer *cp; 1183 1184 /* 1185 * One request failed, so send the same request to 1186 * the parity consumer. 1187 */ 1188 disk = pbp->bio_driver2; 1189 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) { 1190 pbp->bio_error = fbp->bio_error; 1191 goto finish; 1192 } 1193 pbp->bio_pflags |= G_RAID3_BIO_PFLAG_DEGRADED; 1194 pbp->bio_inbed--; 1195 fbp->bio_flags &= ~(BIO_DONE | BIO_ERROR); 1196 if (disk->d_no == sc->sc_ndisks - 1) 1197 fbp->bio_cflags |= G_RAID3_BIO_CFLAG_PARITY; 1198 fbp->bio_error = 0; 1199 fbp->bio_completed = 0; 1200 fbp->bio_children = 0; 1201 fbp->bio_inbed = 0; 1202 cp = disk->d_consumer; 1203 fbp->bio_caller2 = disk; 1204 fbp->bio_to = cp->provider; 1205 G_RAID3_LOGREQ(3, fbp, "Sending request (recover)."); 1206 KASSERT(cp->acr >= 1 && cp->acw >= 1 && cp->ace >= 1, 1207 ("Consumer %s not opened (r%dw%de%d).", cp->provider->name, 1208 cp->acr, cp->acw, cp->ace)); 1209 cp->index++; 1210 g_io_request(fbp, cp); 1211 return; 1212 } 1213 if (xbp != NULL) { 1214 /* 1215 * Calculate parity. 1216 */ 1217 G_RAID3_FOREACH_BIO(pbp, cbp) { 1218 if ((cbp->bio_cflags & G_RAID3_BIO_CFLAG_PARITY) != 0) 1219 continue; 1220 g_raid3_xor(cbp->bio_data, xbp->bio_data, 1221 xbp->bio_length); 1222 } 1223 xbp->bio_cflags &= ~G_RAID3_BIO_CFLAG_PARITY; 1224 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_VERIFY) != 0) { 1225 if (!g_raid3_is_zero(xbp)) { 1226 g_raid3_parity_mismatch++; 1227 pbp->bio_error = EIO; 1228 goto finish; 1229 } 1230 g_raid3_destroy_bio(sc, xbp); 1231 } 1232 } 1233 atom = sc->sc_sectorsize / (sc->sc_ndisks - 1); 1234 cadd = padd = 0; 1235 for (left = pbp->bio_length; left > 0; left -= sc->sc_sectorsize) { 1236 G_RAID3_FOREACH_BIO(pbp, cbp) { 1237 bcopy(cbp->bio_data + cadd, pbp->bio_data + padd, atom); 1238 pbp->bio_completed += atom; 1239 padd += atom; 1240 } 1241 cadd += atom; 1242 } 1243 finish: 1244 if (pbp->bio_error == 0) 1245 G_RAID3_LOGREQ(3, pbp, "Request finished."); 1246 else { 1247 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_VERIFY) != 0) 1248 G_RAID3_LOGREQ(1, pbp, "Verification error."); 1249 else 1250 G_RAID3_LOGREQ(0, pbp, "Request failed."); 1251 } 1252 pbp->bio_pflags &= ~G_RAID3_BIO_PFLAG_MASK; 1253 while ((cbp = G_RAID3_HEAD_BIO(pbp)) != NULL) 1254 g_raid3_destroy_bio(sc, cbp); 1255 g_io_deliver(pbp, pbp->bio_error); 1256 } 1257 1258 static void 1259 g_raid3_done(struct bio *bp) 1260 { 1261 struct g_raid3_softc *sc; 1262 1263 sc = bp->bio_from->geom->softc; 1264 bp->bio_cflags |= G_RAID3_BIO_CFLAG_REGULAR; 1265 G_RAID3_LOGREQ(3, bp, "Regular request done (error=%d).", bp->bio_error); 1266 mtx_lock(&sc->sc_queue_mtx); 1267 bioq_insert_head(&sc->sc_queue, bp); 1268 mtx_unlock(&sc->sc_queue_mtx); 1269 wakeup(sc); 1270 wakeup(&sc->sc_queue); 1271 } 1272 1273 static void 1274 g_raid3_regular_request(struct bio *cbp) 1275 { 1276 struct g_raid3_softc *sc; 1277 struct g_raid3_disk *disk; 1278 struct bio *pbp; 1279 1280 g_topology_assert_not(); 1281 1282 pbp = cbp->bio_parent; 1283 sc = pbp->bio_to->geom->softc; 1284 cbp->bio_from->index--; 1285 if (cbp->bio_cmd == BIO_WRITE) 1286 sc->sc_writes--; 1287 disk = cbp->bio_from->private; 1288 if (disk == NULL) { 1289 g_topology_lock(); 1290 g_raid3_kill_consumer(sc, cbp->bio_from); 1291 g_topology_unlock(); 1292 } 1293 1294 G_RAID3_LOGREQ(3, cbp, "Request finished."); 1295 pbp->bio_inbed++; 1296 KASSERT(pbp->bio_inbed <= pbp->bio_children, 1297 ("bio_inbed (%u) is bigger than bio_children (%u).", pbp->bio_inbed, 1298 pbp->bio_children)); 1299 if (pbp->bio_inbed != pbp->bio_children) 1300 return; 1301 switch (pbp->bio_cmd) { 1302 case BIO_READ: 1303 g_raid3_gather(pbp); 1304 break; 1305 case BIO_WRITE: 1306 case BIO_DELETE: 1307 { 1308 int error = 0; 1309 1310 pbp->bio_completed = pbp->bio_length; 1311 while ((cbp = G_RAID3_HEAD_BIO(pbp)) != NULL) { 1312 if (cbp->bio_error == 0) { 1313 g_raid3_destroy_bio(sc, cbp); 1314 continue; 1315 } 1316 1317 if (error == 0) 1318 error = cbp->bio_error; 1319 else if (pbp->bio_error == 0) { 1320 /* 1321 * Next failed request, that's too many. 1322 */ 1323 pbp->bio_error = error; 1324 } 1325 1326 disk = cbp->bio_caller2; 1327 if (disk == NULL) { 1328 g_raid3_destroy_bio(sc, cbp); 1329 continue; 1330 } 1331 1332 if ((disk->d_flags & G_RAID3_DISK_FLAG_BROKEN) == 0) { 1333 disk->d_flags |= G_RAID3_DISK_FLAG_BROKEN; 1334 G_RAID3_LOGREQ(0, cbp, 1335 "Request failed (error=%d).", 1336 cbp->bio_error); 1337 } else { 1338 G_RAID3_LOGREQ(1, cbp, 1339 "Request failed (error=%d).", 1340 cbp->bio_error); 1341 } 1342 if (g_raid3_disconnect_on_failure && 1343 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE) { 1344 sc->sc_bump_id |= G_RAID3_BUMP_GENID; 1345 g_raid3_event_send(disk, 1346 G_RAID3_DISK_STATE_DISCONNECTED, 1347 G_RAID3_EVENT_DONTWAIT); 1348 } 1349 g_raid3_destroy_bio(sc, cbp); 1350 } 1351 if (pbp->bio_error == 0) 1352 G_RAID3_LOGREQ(3, pbp, "Request finished."); 1353 else 1354 G_RAID3_LOGREQ(0, pbp, "Request failed."); 1355 pbp->bio_pflags &= ~G_RAID3_BIO_PFLAG_DEGRADED; 1356 pbp->bio_pflags &= ~G_RAID3_BIO_PFLAG_NOPARITY; 1357 bioq_remove(&sc->sc_inflight, pbp); 1358 /* Release delayed sync requests if possible. */ 1359 g_raid3_sync_release(sc); 1360 g_io_deliver(pbp, pbp->bio_error); 1361 break; 1362 } 1363 } 1364 } 1365 1366 static void 1367 g_raid3_sync_done(struct bio *bp) 1368 { 1369 struct g_raid3_softc *sc; 1370 1371 G_RAID3_LOGREQ(3, bp, "Synchronization request delivered."); 1372 sc = bp->bio_from->geom->softc; 1373 bp->bio_cflags |= G_RAID3_BIO_CFLAG_SYNC; 1374 mtx_lock(&sc->sc_queue_mtx); 1375 bioq_insert_head(&sc->sc_queue, bp); 1376 mtx_unlock(&sc->sc_queue_mtx); 1377 wakeup(sc); 1378 wakeup(&sc->sc_queue); 1379 } 1380 1381 static void 1382 g_raid3_flush(struct g_raid3_softc *sc, struct bio *bp) 1383 { 1384 struct bio_queue_head queue; 1385 struct g_raid3_disk *disk; 1386 struct g_consumer *cp; 1387 struct bio *cbp; 1388 u_int i; 1389 1390 bioq_init(&queue); 1391 for (i = 0; i < sc->sc_ndisks; i++) { 1392 disk = &sc->sc_disks[i]; 1393 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) 1394 continue; 1395 cbp = g_clone_bio(bp); 1396 if (cbp == NULL) { 1397 for (cbp = bioq_first(&queue); cbp != NULL; 1398 cbp = bioq_first(&queue)) { 1399 bioq_remove(&queue, cbp); 1400 g_destroy_bio(cbp); 1401 } 1402 if (bp->bio_error == 0) 1403 bp->bio_error = ENOMEM; 1404 g_io_deliver(bp, bp->bio_error); 1405 return; 1406 } 1407 bioq_insert_tail(&queue, cbp); 1408 cbp->bio_done = g_std_done; 1409 cbp->bio_caller1 = disk; 1410 cbp->bio_to = disk->d_consumer->provider; 1411 } 1412 for (cbp = bioq_first(&queue); cbp != NULL; cbp = bioq_first(&queue)) { 1413 bioq_remove(&queue, cbp); 1414 G_RAID3_LOGREQ(3, cbp, "Sending request."); 1415 disk = cbp->bio_caller1; 1416 cbp->bio_caller1 = NULL; 1417 cp = disk->d_consumer; 1418 KASSERT(cp->acr >= 1 && cp->acw >= 1 && cp->ace >= 1, 1419 ("Consumer %s not opened (r%dw%de%d).", cp->provider->name, 1420 cp->acr, cp->acw, cp->ace)); 1421 g_io_request(cbp, disk->d_consumer); 1422 } 1423 } 1424 1425 static void 1426 g_raid3_start(struct bio *bp) 1427 { 1428 struct g_raid3_softc *sc; 1429 1430 sc = bp->bio_to->geom->softc; 1431 /* 1432 * If sc == NULL or there are no valid disks, provider's error 1433 * should be set and g_raid3_start() should not be called at all. 1434 */ 1435 KASSERT(sc != NULL && (sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 1436 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE), 1437 ("Provider's error should be set (error=%d)(device=%s).", 1438 bp->bio_to->error, bp->bio_to->name)); 1439 G_RAID3_LOGREQ(3, bp, "Request received."); 1440 1441 switch (bp->bio_cmd) { 1442 case BIO_READ: 1443 case BIO_WRITE: 1444 case BIO_DELETE: 1445 break; 1446 case BIO_FLUSH: 1447 g_raid3_flush(sc, bp); 1448 return; 1449 case BIO_GETATTR: 1450 default: 1451 g_io_deliver(bp, EOPNOTSUPP); 1452 return; 1453 } 1454 mtx_lock(&sc->sc_queue_mtx); 1455 bioq_insert_tail(&sc->sc_queue, bp); 1456 mtx_unlock(&sc->sc_queue_mtx); 1457 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, sc); 1458 wakeup(sc); 1459 } 1460 1461 /* 1462 * Return TRUE if the given request is colliding with a in-progress 1463 * synchronization request. 1464 */ 1465 static int 1466 g_raid3_sync_collision(struct g_raid3_softc *sc, struct bio *bp) 1467 { 1468 struct g_raid3_disk *disk; 1469 struct bio *sbp; 1470 off_t rstart, rend, sstart, send; 1471 int i; 1472 1473 disk = sc->sc_syncdisk; 1474 if (disk == NULL) 1475 return (0); 1476 rstart = bp->bio_offset; 1477 rend = bp->bio_offset + bp->bio_length; 1478 for (i = 0; i < g_raid3_syncreqs; i++) { 1479 sbp = disk->d_sync.ds_bios[i]; 1480 if (sbp == NULL) 1481 continue; 1482 sstart = sbp->bio_offset; 1483 send = sbp->bio_length; 1484 if (sbp->bio_cmd == BIO_WRITE) { 1485 sstart *= sc->sc_ndisks - 1; 1486 send *= sc->sc_ndisks - 1; 1487 } 1488 send += sstart; 1489 if (rend > sstart && rstart < send) 1490 return (1); 1491 } 1492 return (0); 1493 } 1494 1495 /* 1496 * Return TRUE if the given sync request is colliding with a in-progress regular 1497 * request. 1498 */ 1499 static int 1500 g_raid3_regular_collision(struct g_raid3_softc *sc, struct bio *sbp) 1501 { 1502 off_t rstart, rend, sstart, send; 1503 struct bio *bp; 1504 1505 if (sc->sc_syncdisk == NULL) 1506 return (0); 1507 sstart = sbp->bio_offset; 1508 send = sstart + sbp->bio_length; 1509 TAILQ_FOREACH(bp, &sc->sc_inflight.queue, bio_queue) { 1510 rstart = bp->bio_offset; 1511 rend = bp->bio_offset + bp->bio_length; 1512 if (rend > sstart && rstart < send) 1513 return (1); 1514 } 1515 return (0); 1516 } 1517 1518 /* 1519 * Puts request onto delayed queue. 1520 */ 1521 static void 1522 g_raid3_regular_delay(struct g_raid3_softc *sc, struct bio *bp) 1523 { 1524 1525 G_RAID3_LOGREQ(2, bp, "Delaying request."); 1526 bioq_insert_head(&sc->sc_regular_delayed, bp); 1527 } 1528 1529 /* 1530 * Puts synchronization request onto delayed queue. 1531 */ 1532 static void 1533 g_raid3_sync_delay(struct g_raid3_softc *sc, struct bio *bp) 1534 { 1535 1536 G_RAID3_LOGREQ(2, bp, "Delaying synchronization request."); 1537 bioq_insert_tail(&sc->sc_sync_delayed, bp); 1538 } 1539 1540 /* 1541 * Releases delayed regular requests which don't collide anymore with sync 1542 * requests. 1543 */ 1544 static void 1545 g_raid3_regular_release(struct g_raid3_softc *sc) 1546 { 1547 struct bio *bp, *bp2; 1548 1549 TAILQ_FOREACH_SAFE(bp, &sc->sc_regular_delayed.queue, bio_queue, bp2) { 1550 if (g_raid3_sync_collision(sc, bp)) 1551 continue; 1552 bioq_remove(&sc->sc_regular_delayed, bp); 1553 G_RAID3_LOGREQ(2, bp, "Releasing delayed request (%p).", bp); 1554 mtx_lock(&sc->sc_queue_mtx); 1555 bioq_insert_head(&sc->sc_queue, bp); 1556 #if 0 1557 /* 1558 * wakeup() is not needed, because this function is called from 1559 * the worker thread. 1560 */ 1561 wakeup(&sc->sc_queue); 1562 #endif 1563 mtx_unlock(&sc->sc_queue_mtx); 1564 } 1565 } 1566 1567 /* 1568 * Releases delayed sync requests which don't collide anymore with regular 1569 * requests. 1570 */ 1571 static void 1572 g_raid3_sync_release(struct g_raid3_softc *sc) 1573 { 1574 struct bio *bp, *bp2; 1575 1576 TAILQ_FOREACH_SAFE(bp, &sc->sc_sync_delayed.queue, bio_queue, bp2) { 1577 if (g_raid3_regular_collision(sc, bp)) 1578 continue; 1579 bioq_remove(&sc->sc_sync_delayed, bp); 1580 G_RAID3_LOGREQ(2, bp, 1581 "Releasing delayed synchronization request."); 1582 g_io_request(bp, bp->bio_from); 1583 } 1584 } 1585 1586 /* 1587 * Handle synchronization requests. 1588 * Every synchronization request is two-steps process: first, READ request is 1589 * send to active provider and then WRITE request (with read data) to the provider 1590 * being synchronized. When WRITE is finished, new synchronization request is 1591 * send. 1592 */ 1593 static void 1594 g_raid3_sync_request(struct bio *bp) 1595 { 1596 struct g_raid3_softc *sc; 1597 struct g_raid3_disk *disk; 1598 1599 bp->bio_from->index--; 1600 sc = bp->bio_from->geom->softc; 1601 disk = bp->bio_from->private; 1602 if (disk == NULL) { 1603 sx_xunlock(&sc->sc_lock); /* Avoid recursion on sc_lock. */ 1604 g_topology_lock(); 1605 g_raid3_kill_consumer(sc, bp->bio_from); 1606 g_topology_unlock(); 1607 free(bp->bio_data, M_RAID3); 1608 g_destroy_bio(bp); 1609 sx_xlock(&sc->sc_lock); 1610 return; 1611 } 1612 1613 /* 1614 * Synchronization request. 1615 */ 1616 switch (bp->bio_cmd) { 1617 case BIO_READ: 1618 { 1619 struct g_consumer *cp; 1620 u_char *dst, *src; 1621 off_t left; 1622 u_int atom; 1623 1624 if (bp->bio_error != 0) { 1625 G_RAID3_LOGREQ(0, bp, 1626 "Synchronization request failed (error=%d).", 1627 bp->bio_error); 1628 g_destroy_bio(bp); 1629 return; 1630 } 1631 G_RAID3_LOGREQ(3, bp, "Synchronization request finished."); 1632 atom = sc->sc_sectorsize / (sc->sc_ndisks - 1); 1633 dst = src = bp->bio_data; 1634 if (disk->d_no == sc->sc_ndisks - 1) { 1635 u_int n; 1636 1637 /* Parity component. */ 1638 for (left = bp->bio_length; left > 0; 1639 left -= sc->sc_sectorsize) { 1640 bcopy(src, dst, atom); 1641 src += atom; 1642 for (n = 1; n < sc->sc_ndisks - 1; n++) { 1643 g_raid3_xor(src, dst, atom); 1644 src += atom; 1645 } 1646 dst += atom; 1647 } 1648 } else { 1649 /* Regular component. */ 1650 src += atom * disk->d_no; 1651 for (left = bp->bio_length; left > 0; 1652 left -= sc->sc_sectorsize) { 1653 bcopy(src, dst, atom); 1654 src += sc->sc_sectorsize; 1655 dst += atom; 1656 } 1657 } 1658 bp->bio_driver1 = bp->bio_driver2 = NULL; 1659 bp->bio_pflags = 0; 1660 bp->bio_offset /= sc->sc_ndisks - 1; 1661 bp->bio_length /= sc->sc_ndisks - 1; 1662 bp->bio_cmd = BIO_WRITE; 1663 bp->bio_cflags = 0; 1664 bp->bio_children = bp->bio_inbed = 0; 1665 cp = disk->d_consumer; 1666 KASSERT(cp->acr >= 1 && cp->acw >= 1 && cp->ace >= 1, 1667 ("Consumer %s not opened (r%dw%de%d).", cp->provider->name, 1668 cp->acr, cp->acw, cp->ace)); 1669 cp->index++; 1670 g_io_request(bp, cp); 1671 return; 1672 } 1673 case BIO_WRITE: 1674 { 1675 struct g_raid3_disk_sync *sync; 1676 off_t boffset, moffset; 1677 void *data; 1678 int i; 1679 1680 if (bp->bio_error != 0) { 1681 G_RAID3_LOGREQ(0, bp, 1682 "Synchronization request failed (error=%d).", 1683 bp->bio_error); 1684 g_destroy_bio(bp); 1685 sc->sc_bump_id |= G_RAID3_BUMP_GENID; 1686 g_raid3_event_send(disk, 1687 G_RAID3_DISK_STATE_DISCONNECTED, 1688 G_RAID3_EVENT_DONTWAIT); 1689 return; 1690 } 1691 G_RAID3_LOGREQ(3, bp, "Synchronization request finished."); 1692 sync = &disk->d_sync; 1693 if (sync->ds_offset == sc->sc_mediasize / (sc->sc_ndisks - 1) || 1694 sync->ds_consumer == NULL || 1695 (sc->sc_flags & G_RAID3_DEVICE_FLAG_DESTROY) != 0) { 1696 /* Don't send more synchronization requests. */ 1697 sync->ds_inflight--; 1698 if (sync->ds_bios != NULL) { 1699 i = (int)(uintptr_t)bp->bio_caller1; 1700 sync->ds_bios[i] = NULL; 1701 } 1702 free(bp->bio_data, M_RAID3); 1703 g_destroy_bio(bp); 1704 if (sync->ds_inflight > 0) 1705 return; 1706 if (sync->ds_consumer == NULL || 1707 (sc->sc_flags & G_RAID3_DEVICE_FLAG_DESTROY) != 0) { 1708 return; 1709 } 1710 /* 1711 * Disk up-to-date, activate it. 1712 */ 1713 g_raid3_event_send(disk, G_RAID3_DISK_STATE_ACTIVE, 1714 G_RAID3_EVENT_DONTWAIT); 1715 return; 1716 } 1717 1718 /* Send next synchronization request. */ 1719 data = bp->bio_data; 1720 g_reset_bio(bp); 1721 bp->bio_cmd = BIO_READ; 1722 bp->bio_offset = sync->ds_offset * (sc->sc_ndisks - 1); 1723 bp->bio_length = MIN(MAXPHYS, sc->sc_mediasize - bp->bio_offset); 1724 sync->ds_offset += bp->bio_length / (sc->sc_ndisks - 1); 1725 bp->bio_done = g_raid3_sync_done; 1726 bp->bio_data = data; 1727 bp->bio_from = sync->ds_consumer; 1728 bp->bio_to = sc->sc_provider; 1729 G_RAID3_LOGREQ(3, bp, "Sending synchronization request."); 1730 sync->ds_consumer->index++; 1731 /* 1732 * Delay the request if it is colliding with a regular request. 1733 */ 1734 if (g_raid3_regular_collision(sc, bp)) 1735 g_raid3_sync_delay(sc, bp); 1736 else 1737 g_io_request(bp, sync->ds_consumer); 1738 1739 /* Release delayed requests if possible. */ 1740 g_raid3_regular_release(sc); 1741 1742 /* Find the smallest offset. */ 1743 moffset = sc->sc_mediasize; 1744 for (i = 0; i < g_raid3_syncreqs; i++) { 1745 bp = sync->ds_bios[i]; 1746 boffset = bp->bio_offset; 1747 if (bp->bio_cmd == BIO_WRITE) 1748 boffset *= sc->sc_ndisks - 1; 1749 if (boffset < moffset) 1750 moffset = boffset; 1751 } 1752 if (sync->ds_offset_done + (MAXPHYS * 100) < moffset) { 1753 /* Update offset_done on every 100 blocks. */ 1754 sync->ds_offset_done = moffset; 1755 g_raid3_update_metadata(disk); 1756 } 1757 return; 1758 } 1759 default: 1760 KASSERT(1 == 0, ("Invalid command here: %u (device=%s)", 1761 bp->bio_cmd, sc->sc_name)); 1762 break; 1763 } 1764 } 1765 1766 static int 1767 g_raid3_register_request(struct bio *pbp) 1768 { 1769 struct g_raid3_softc *sc; 1770 struct g_raid3_disk *disk; 1771 struct g_consumer *cp; 1772 struct bio *cbp, *tmpbp; 1773 off_t offset, length; 1774 u_int n, ndisks; 1775 int round_robin, verify; 1776 1777 ndisks = 0; 1778 sc = pbp->bio_to->geom->softc; 1779 if ((pbp->bio_cflags & G_RAID3_BIO_CFLAG_REGSYNC) != 0 && 1780 sc->sc_syncdisk == NULL) { 1781 g_io_deliver(pbp, EIO); 1782 return (0); 1783 } 1784 g_raid3_init_bio(pbp); 1785 length = pbp->bio_length / (sc->sc_ndisks - 1); 1786 offset = pbp->bio_offset / (sc->sc_ndisks - 1); 1787 round_robin = verify = 0; 1788 switch (pbp->bio_cmd) { 1789 case BIO_READ: 1790 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_VERIFY) != 0 && 1791 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE) { 1792 pbp->bio_pflags |= G_RAID3_BIO_PFLAG_VERIFY; 1793 verify = 1; 1794 ndisks = sc->sc_ndisks; 1795 } else { 1796 verify = 0; 1797 ndisks = sc->sc_ndisks - 1; 1798 } 1799 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_ROUND_ROBIN) != 0 && 1800 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE) { 1801 round_robin = 1; 1802 } else { 1803 round_robin = 0; 1804 } 1805 KASSERT(!round_robin || !verify, 1806 ("ROUND-ROBIN and VERIFY are mutually exclusive.")); 1807 pbp->bio_driver2 = &sc->sc_disks[sc->sc_ndisks - 1]; 1808 break; 1809 case BIO_WRITE: 1810 case BIO_DELETE: 1811 /* 1812 * Delay the request if it is colliding with a synchronization 1813 * request. 1814 */ 1815 if (g_raid3_sync_collision(sc, pbp)) { 1816 g_raid3_regular_delay(sc, pbp); 1817 return (0); 1818 } 1819 1820 if (sc->sc_idle) 1821 g_raid3_unidle(sc); 1822 else 1823 sc->sc_last_write = time_uptime; 1824 1825 ndisks = sc->sc_ndisks; 1826 break; 1827 } 1828 for (n = 0; n < ndisks; n++) { 1829 disk = &sc->sc_disks[n]; 1830 cbp = g_raid3_clone_bio(sc, pbp); 1831 if (cbp == NULL) { 1832 while ((cbp = G_RAID3_HEAD_BIO(pbp)) != NULL) 1833 g_raid3_destroy_bio(sc, cbp); 1834 /* 1835 * To prevent deadlock, we must run back up 1836 * with the ENOMEM for failed requests of any 1837 * of our consumers. Our own sync requests 1838 * can stick around, as they are finite. 1839 */ 1840 if ((pbp->bio_cflags & 1841 G_RAID3_BIO_CFLAG_REGULAR) != 0) { 1842 g_io_deliver(pbp, ENOMEM); 1843 return (0); 1844 } 1845 return (ENOMEM); 1846 } 1847 cbp->bio_offset = offset; 1848 cbp->bio_length = length; 1849 cbp->bio_done = g_raid3_done; 1850 switch (pbp->bio_cmd) { 1851 case BIO_READ: 1852 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) { 1853 /* 1854 * Replace invalid component with the parity 1855 * component. 1856 */ 1857 disk = &sc->sc_disks[sc->sc_ndisks - 1]; 1858 cbp->bio_cflags |= G_RAID3_BIO_CFLAG_PARITY; 1859 pbp->bio_pflags |= G_RAID3_BIO_PFLAG_DEGRADED; 1860 } else if (round_robin && 1861 disk->d_no == sc->sc_round_robin) { 1862 /* 1863 * In round-robin mode skip one data component 1864 * and use parity component when reading. 1865 */ 1866 pbp->bio_driver2 = disk; 1867 disk = &sc->sc_disks[sc->sc_ndisks - 1]; 1868 cbp->bio_cflags |= G_RAID3_BIO_CFLAG_PARITY; 1869 sc->sc_round_robin++; 1870 round_robin = 0; 1871 } else if (verify && disk->d_no == sc->sc_ndisks - 1) { 1872 cbp->bio_cflags |= G_RAID3_BIO_CFLAG_PARITY; 1873 } 1874 break; 1875 case BIO_WRITE: 1876 case BIO_DELETE: 1877 if (disk->d_state == G_RAID3_DISK_STATE_ACTIVE || 1878 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 1879 if (n == ndisks - 1) { 1880 /* 1881 * Active parity component, mark it as such. 1882 */ 1883 cbp->bio_cflags |= 1884 G_RAID3_BIO_CFLAG_PARITY; 1885 } 1886 } else { 1887 pbp->bio_pflags |= G_RAID3_BIO_PFLAG_DEGRADED; 1888 if (n == ndisks - 1) { 1889 /* 1890 * Parity component is not connected, 1891 * so destroy its request. 1892 */ 1893 pbp->bio_pflags |= 1894 G_RAID3_BIO_PFLAG_NOPARITY; 1895 g_raid3_destroy_bio(sc, cbp); 1896 cbp = NULL; 1897 } else { 1898 cbp->bio_cflags |= 1899 G_RAID3_BIO_CFLAG_NODISK; 1900 disk = NULL; 1901 } 1902 } 1903 break; 1904 } 1905 if (cbp != NULL) 1906 cbp->bio_caller2 = disk; 1907 } 1908 switch (pbp->bio_cmd) { 1909 case BIO_READ: 1910 if (round_robin) { 1911 /* 1912 * If we are in round-robin mode and 'round_robin' is 1913 * still 1, it means, that we skipped parity component 1914 * for this read and must reset sc_round_robin field. 1915 */ 1916 sc->sc_round_robin = 0; 1917 } 1918 G_RAID3_FOREACH_SAFE_BIO(pbp, cbp, tmpbp) { 1919 disk = cbp->bio_caller2; 1920 cp = disk->d_consumer; 1921 cbp->bio_to = cp->provider; 1922 G_RAID3_LOGREQ(3, cbp, "Sending request."); 1923 KASSERT(cp->acr >= 1 && cp->acw >= 1 && cp->ace >= 1, 1924 ("Consumer %s not opened (r%dw%de%d).", 1925 cp->provider->name, cp->acr, cp->acw, cp->ace)); 1926 cp->index++; 1927 g_io_request(cbp, cp); 1928 } 1929 break; 1930 case BIO_WRITE: 1931 case BIO_DELETE: 1932 /* 1933 * Put request onto inflight queue, so we can check if new 1934 * synchronization requests don't collide with it. 1935 */ 1936 bioq_insert_tail(&sc->sc_inflight, pbp); 1937 1938 /* 1939 * Bump syncid on first write. 1940 */ 1941 if ((sc->sc_bump_id & G_RAID3_BUMP_SYNCID) != 0) { 1942 sc->sc_bump_id &= ~G_RAID3_BUMP_SYNCID; 1943 g_raid3_bump_syncid(sc); 1944 } 1945 g_raid3_scatter(pbp); 1946 break; 1947 } 1948 return (0); 1949 } 1950 1951 static int 1952 g_raid3_can_destroy(struct g_raid3_softc *sc) 1953 { 1954 struct g_geom *gp; 1955 struct g_consumer *cp; 1956 1957 g_topology_assert(); 1958 gp = sc->sc_geom; 1959 if (gp->softc == NULL) 1960 return (1); 1961 LIST_FOREACH(cp, &gp->consumer, consumer) { 1962 if (g_raid3_is_busy(sc, cp)) 1963 return (0); 1964 } 1965 gp = sc->sc_sync.ds_geom; 1966 LIST_FOREACH(cp, &gp->consumer, consumer) { 1967 if (g_raid3_is_busy(sc, cp)) 1968 return (0); 1969 } 1970 G_RAID3_DEBUG(2, "No I/O requests for %s, it can be destroyed.", 1971 sc->sc_name); 1972 return (1); 1973 } 1974 1975 static int 1976 g_raid3_try_destroy(struct g_raid3_softc *sc) 1977 { 1978 1979 g_topology_assert_not(); 1980 sx_assert(&sc->sc_lock, SX_XLOCKED); 1981 1982 if (sc->sc_rootmount != NULL) { 1983 G_RAID3_DEBUG(1, "root_mount_rel[%u] %p", __LINE__, 1984 sc->sc_rootmount); 1985 root_mount_rel(sc->sc_rootmount); 1986 sc->sc_rootmount = NULL; 1987 } 1988 1989 g_topology_lock(); 1990 if (!g_raid3_can_destroy(sc)) { 1991 g_topology_unlock(); 1992 return (0); 1993 } 1994 sc->sc_geom->softc = NULL; 1995 sc->sc_sync.ds_geom->softc = NULL; 1996 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_WAIT) != 0) { 1997 g_topology_unlock(); 1998 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, 1999 &sc->sc_worker); 2000 /* Unlock sc_lock here, as it can be destroyed after wakeup. */ 2001 sx_xunlock(&sc->sc_lock); 2002 wakeup(&sc->sc_worker); 2003 sc->sc_worker = NULL; 2004 } else { 2005 g_topology_unlock(); 2006 g_raid3_destroy_device(sc); 2007 free(sc->sc_disks, M_RAID3); 2008 free(sc, M_RAID3); 2009 } 2010 return (1); 2011 } 2012 2013 /* 2014 * Worker thread. 2015 */ 2016 static void 2017 g_raid3_worker(void *arg) 2018 { 2019 struct g_raid3_softc *sc; 2020 struct g_raid3_event *ep; 2021 struct bio *bp; 2022 int timeout; 2023 2024 sc = arg; 2025 thread_lock(curthread); 2026 sched_prio(curthread, PRIBIO); 2027 thread_unlock(curthread); 2028 2029 sx_xlock(&sc->sc_lock); 2030 for (;;) { 2031 G_RAID3_DEBUG(5, "%s: Let's see...", __func__); 2032 /* 2033 * First take a look at events. 2034 * This is important to handle events before any I/O requests. 2035 */ 2036 ep = g_raid3_event_get(sc); 2037 if (ep != NULL) { 2038 g_raid3_event_remove(sc, ep); 2039 if ((ep->e_flags & G_RAID3_EVENT_DEVICE) != 0) { 2040 /* Update only device status. */ 2041 G_RAID3_DEBUG(3, 2042 "Running event for device %s.", 2043 sc->sc_name); 2044 ep->e_error = 0; 2045 g_raid3_update_device(sc, 1); 2046 } else { 2047 /* Update disk status. */ 2048 G_RAID3_DEBUG(3, "Running event for disk %s.", 2049 g_raid3_get_diskname(ep->e_disk)); 2050 ep->e_error = g_raid3_update_disk(ep->e_disk, 2051 ep->e_state); 2052 if (ep->e_error == 0) 2053 g_raid3_update_device(sc, 0); 2054 } 2055 if ((ep->e_flags & G_RAID3_EVENT_DONTWAIT) != 0) { 2056 KASSERT(ep->e_error == 0, 2057 ("Error cannot be handled.")); 2058 g_raid3_event_free(ep); 2059 } else { 2060 ep->e_flags |= G_RAID3_EVENT_DONE; 2061 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, 2062 ep); 2063 mtx_lock(&sc->sc_events_mtx); 2064 wakeup(ep); 2065 mtx_unlock(&sc->sc_events_mtx); 2066 } 2067 if ((sc->sc_flags & 2068 G_RAID3_DEVICE_FLAG_DESTROY) != 0) { 2069 if (g_raid3_try_destroy(sc)) { 2070 curthread->td_pflags &= ~TDP_GEOM; 2071 G_RAID3_DEBUG(1, "Thread exiting."); 2072 kproc_exit(0); 2073 } 2074 } 2075 G_RAID3_DEBUG(5, "%s: I'm here 1.", __func__); 2076 continue; 2077 } 2078 /* 2079 * Check if we can mark array as CLEAN and if we can't take 2080 * how much seconds should we wait. 2081 */ 2082 timeout = g_raid3_idle(sc, -1); 2083 /* 2084 * Now I/O requests. 2085 */ 2086 /* Get first request from the queue. */ 2087 mtx_lock(&sc->sc_queue_mtx); 2088 bp = bioq_first(&sc->sc_queue); 2089 if (bp == NULL) { 2090 if ((sc->sc_flags & 2091 G_RAID3_DEVICE_FLAG_DESTROY) != 0) { 2092 mtx_unlock(&sc->sc_queue_mtx); 2093 if (g_raid3_try_destroy(sc)) { 2094 curthread->td_pflags &= ~TDP_GEOM; 2095 G_RAID3_DEBUG(1, "Thread exiting."); 2096 kproc_exit(0); 2097 } 2098 mtx_lock(&sc->sc_queue_mtx); 2099 } 2100 sx_xunlock(&sc->sc_lock); 2101 /* 2102 * XXX: We can miss an event here, because an event 2103 * can be added without sx-device-lock and without 2104 * mtx-queue-lock. Maybe I should just stop using 2105 * dedicated mutex for events synchronization and 2106 * stick with the queue lock? 2107 * The event will hang here until next I/O request 2108 * or next event is received. 2109 */ 2110 MSLEEP(sc, &sc->sc_queue_mtx, PRIBIO | PDROP, "r3:w1", 2111 timeout * hz); 2112 sx_xlock(&sc->sc_lock); 2113 G_RAID3_DEBUG(5, "%s: I'm here 4.", __func__); 2114 continue; 2115 } 2116 process: 2117 bioq_remove(&sc->sc_queue, bp); 2118 mtx_unlock(&sc->sc_queue_mtx); 2119 2120 if (bp->bio_from->geom == sc->sc_sync.ds_geom && 2121 (bp->bio_cflags & G_RAID3_BIO_CFLAG_SYNC) != 0) { 2122 g_raid3_sync_request(bp); /* READ */ 2123 } else if (bp->bio_to != sc->sc_provider) { 2124 if ((bp->bio_cflags & G_RAID3_BIO_CFLAG_REGULAR) != 0) 2125 g_raid3_regular_request(bp); 2126 else if ((bp->bio_cflags & G_RAID3_BIO_CFLAG_SYNC) != 0) 2127 g_raid3_sync_request(bp); /* WRITE */ 2128 else { 2129 KASSERT(0, 2130 ("Invalid request cflags=0x%hx to=%s.", 2131 bp->bio_cflags, bp->bio_to->name)); 2132 } 2133 } else if (g_raid3_register_request(bp) != 0) { 2134 mtx_lock(&sc->sc_queue_mtx); 2135 bioq_insert_head(&sc->sc_queue, bp); 2136 /* 2137 * We are short in memory, let see if there are finished 2138 * request we can free. 2139 */ 2140 TAILQ_FOREACH(bp, &sc->sc_queue.queue, bio_queue) { 2141 if (bp->bio_cflags & G_RAID3_BIO_CFLAG_REGULAR) 2142 goto process; 2143 } 2144 /* 2145 * No finished regular request, so at least keep 2146 * synchronization running. 2147 */ 2148 TAILQ_FOREACH(bp, &sc->sc_queue.queue, bio_queue) { 2149 if (bp->bio_cflags & G_RAID3_BIO_CFLAG_SYNC) 2150 goto process; 2151 } 2152 sx_xunlock(&sc->sc_lock); 2153 MSLEEP(&sc->sc_queue, &sc->sc_queue_mtx, PRIBIO | PDROP, 2154 "r3:lowmem", hz / 10); 2155 sx_xlock(&sc->sc_lock); 2156 } 2157 G_RAID3_DEBUG(5, "%s: I'm here 9.", __func__); 2158 } 2159 } 2160 2161 static void 2162 g_raid3_update_idle(struct g_raid3_softc *sc, struct g_raid3_disk *disk) 2163 { 2164 2165 sx_assert(&sc->sc_lock, SX_LOCKED); 2166 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOFAILSYNC) != 0) 2167 return; 2168 if (!sc->sc_idle && (disk->d_flags & G_RAID3_DISK_FLAG_DIRTY) == 0) { 2169 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as dirty.", 2170 g_raid3_get_diskname(disk), sc->sc_name); 2171 disk->d_flags |= G_RAID3_DISK_FLAG_DIRTY; 2172 } else if (sc->sc_idle && 2173 (disk->d_flags & G_RAID3_DISK_FLAG_DIRTY) != 0) { 2174 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as clean.", 2175 g_raid3_get_diskname(disk), sc->sc_name); 2176 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 2177 } 2178 } 2179 2180 static void 2181 g_raid3_sync_start(struct g_raid3_softc *sc) 2182 { 2183 struct g_raid3_disk *disk; 2184 struct g_consumer *cp; 2185 struct bio *bp; 2186 int error; 2187 u_int n; 2188 2189 g_topology_assert_not(); 2190 sx_assert(&sc->sc_lock, SX_XLOCKED); 2191 2192 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED, 2193 ("Device not in DEGRADED state (%s, %u).", sc->sc_name, 2194 sc->sc_state)); 2195 KASSERT(sc->sc_syncdisk == NULL, ("Syncdisk is not NULL (%s, %u).", 2196 sc->sc_name, sc->sc_state)); 2197 disk = NULL; 2198 for (n = 0; n < sc->sc_ndisks; n++) { 2199 if (sc->sc_disks[n].d_state != G_RAID3_DISK_STATE_SYNCHRONIZING) 2200 continue; 2201 disk = &sc->sc_disks[n]; 2202 break; 2203 } 2204 if (disk == NULL) 2205 return; 2206 2207 sx_xunlock(&sc->sc_lock); 2208 g_topology_lock(); 2209 cp = g_new_consumer(sc->sc_sync.ds_geom); 2210 error = g_attach(cp, sc->sc_provider); 2211 KASSERT(error == 0, 2212 ("Cannot attach to %s (error=%d).", sc->sc_name, error)); 2213 error = g_access(cp, 1, 0, 0); 2214 KASSERT(error == 0, ("Cannot open %s (error=%d).", sc->sc_name, error)); 2215 g_topology_unlock(); 2216 sx_xlock(&sc->sc_lock); 2217 2218 G_RAID3_DEBUG(0, "Device %s: rebuilding provider %s.", sc->sc_name, 2219 g_raid3_get_diskname(disk)); 2220 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOFAILSYNC) == 0) 2221 disk->d_flags |= G_RAID3_DISK_FLAG_DIRTY; 2222 KASSERT(disk->d_sync.ds_consumer == NULL, 2223 ("Sync consumer already exists (device=%s, disk=%s).", 2224 sc->sc_name, g_raid3_get_diskname(disk))); 2225 2226 disk->d_sync.ds_consumer = cp; 2227 disk->d_sync.ds_consumer->private = disk; 2228 disk->d_sync.ds_consumer->index = 0; 2229 sc->sc_syncdisk = disk; 2230 2231 /* 2232 * Allocate memory for synchronization bios and initialize them. 2233 */ 2234 disk->d_sync.ds_bios = malloc(sizeof(struct bio *) * g_raid3_syncreqs, 2235 M_RAID3, M_WAITOK); 2236 for (n = 0; n < g_raid3_syncreqs; n++) { 2237 bp = g_alloc_bio(); 2238 disk->d_sync.ds_bios[n] = bp; 2239 bp->bio_parent = NULL; 2240 bp->bio_cmd = BIO_READ; 2241 bp->bio_data = malloc(MAXPHYS, M_RAID3, M_WAITOK); 2242 bp->bio_cflags = 0; 2243 bp->bio_offset = disk->d_sync.ds_offset * (sc->sc_ndisks - 1); 2244 bp->bio_length = MIN(MAXPHYS, sc->sc_mediasize - bp->bio_offset); 2245 disk->d_sync.ds_offset += bp->bio_length / (sc->sc_ndisks - 1); 2246 bp->bio_done = g_raid3_sync_done; 2247 bp->bio_from = disk->d_sync.ds_consumer; 2248 bp->bio_to = sc->sc_provider; 2249 bp->bio_caller1 = (void *)(uintptr_t)n; 2250 } 2251 2252 /* Set the number of in-flight synchronization requests. */ 2253 disk->d_sync.ds_inflight = g_raid3_syncreqs; 2254 2255 /* 2256 * Fire off first synchronization requests. 2257 */ 2258 for (n = 0; n < g_raid3_syncreqs; n++) { 2259 bp = disk->d_sync.ds_bios[n]; 2260 G_RAID3_LOGREQ(3, bp, "Sending synchronization request."); 2261 disk->d_sync.ds_consumer->index++; 2262 /* 2263 * Delay the request if it is colliding with a regular request. 2264 */ 2265 if (g_raid3_regular_collision(sc, bp)) 2266 g_raid3_sync_delay(sc, bp); 2267 else 2268 g_io_request(bp, disk->d_sync.ds_consumer); 2269 } 2270 } 2271 2272 /* 2273 * Stop synchronization process. 2274 * type: 0 - synchronization finished 2275 * 1 - synchronization stopped 2276 */ 2277 static void 2278 g_raid3_sync_stop(struct g_raid3_softc *sc, int type) 2279 { 2280 struct g_raid3_disk *disk; 2281 struct g_consumer *cp; 2282 2283 g_topology_assert_not(); 2284 sx_assert(&sc->sc_lock, SX_LOCKED); 2285 2286 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED, 2287 ("Device not in DEGRADED state (%s, %u).", sc->sc_name, 2288 sc->sc_state)); 2289 disk = sc->sc_syncdisk; 2290 sc->sc_syncdisk = NULL; 2291 KASSERT(disk != NULL, ("No disk was synchronized (%s).", sc->sc_name)); 2292 KASSERT(disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING, 2293 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 2294 g_raid3_disk_state2str(disk->d_state))); 2295 if (disk->d_sync.ds_consumer == NULL) 2296 return; 2297 2298 if (type == 0) { 2299 G_RAID3_DEBUG(0, "Device %s: rebuilding provider %s finished.", 2300 sc->sc_name, g_raid3_get_diskname(disk)); 2301 } else /* if (type == 1) */ { 2302 G_RAID3_DEBUG(0, "Device %s: rebuilding provider %s stopped.", 2303 sc->sc_name, g_raid3_get_diskname(disk)); 2304 } 2305 free(disk->d_sync.ds_bios, M_RAID3); 2306 disk->d_sync.ds_bios = NULL; 2307 cp = disk->d_sync.ds_consumer; 2308 disk->d_sync.ds_consumer = NULL; 2309 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 2310 sx_xunlock(&sc->sc_lock); /* Avoid recursion on sc_lock. */ 2311 g_topology_lock(); 2312 g_raid3_kill_consumer(sc, cp); 2313 g_topology_unlock(); 2314 sx_xlock(&sc->sc_lock); 2315 } 2316 2317 static void 2318 g_raid3_launch_provider(struct g_raid3_softc *sc) 2319 { 2320 struct g_provider *pp; 2321 struct g_raid3_disk *disk; 2322 int n; 2323 2324 sx_assert(&sc->sc_lock, SX_LOCKED); 2325 2326 g_topology_lock(); 2327 pp = g_new_providerf(sc->sc_geom, "raid3/%s", sc->sc_name); 2328 pp->mediasize = sc->sc_mediasize; 2329 pp->sectorsize = sc->sc_sectorsize; 2330 pp->stripesize = 0; 2331 pp->stripeoffset = 0; 2332 for (n = 0; n < sc->sc_ndisks; n++) { 2333 disk = &sc->sc_disks[n]; 2334 if (disk->d_consumer && disk->d_consumer->provider && 2335 disk->d_consumer->provider->stripesize > pp->stripesize) { 2336 pp->stripesize = disk->d_consumer->provider->stripesize; 2337 pp->stripeoffset = disk->d_consumer->provider->stripeoffset; 2338 } 2339 } 2340 pp->stripesize *= sc->sc_ndisks - 1; 2341 pp->stripeoffset *= sc->sc_ndisks - 1; 2342 sc->sc_provider = pp; 2343 g_error_provider(pp, 0); 2344 g_topology_unlock(); 2345 G_RAID3_DEBUG(0, "Device %s launched (%u/%u).", pp->name, 2346 g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE), sc->sc_ndisks); 2347 2348 if (sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED) 2349 g_raid3_sync_start(sc); 2350 } 2351 2352 static void 2353 g_raid3_destroy_provider(struct g_raid3_softc *sc) 2354 { 2355 struct bio *bp; 2356 2357 g_topology_assert_not(); 2358 KASSERT(sc->sc_provider != NULL, ("NULL provider (device=%s).", 2359 sc->sc_name)); 2360 2361 g_topology_lock(); 2362 g_error_provider(sc->sc_provider, ENXIO); 2363 mtx_lock(&sc->sc_queue_mtx); 2364 while ((bp = bioq_first(&sc->sc_queue)) != NULL) { 2365 bioq_remove(&sc->sc_queue, bp); 2366 g_io_deliver(bp, ENXIO); 2367 } 2368 mtx_unlock(&sc->sc_queue_mtx); 2369 G_RAID3_DEBUG(0, "Device %s: provider %s destroyed.", sc->sc_name, 2370 sc->sc_provider->name); 2371 g_wither_provider(sc->sc_provider, ENXIO); 2372 g_topology_unlock(); 2373 sc->sc_provider = NULL; 2374 if (sc->sc_syncdisk != NULL) 2375 g_raid3_sync_stop(sc, 1); 2376 } 2377 2378 static void 2379 g_raid3_go(void *arg) 2380 { 2381 struct g_raid3_softc *sc; 2382 2383 sc = arg; 2384 G_RAID3_DEBUG(0, "Force device %s start due to timeout.", sc->sc_name); 2385 g_raid3_event_send(sc, 0, 2386 G_RAID3_EVENT_DONTWAIT | G_RAID3_EVENT_DEVICE); 2387 } 2388 2389 static u_int 2390 g_raid3_determine_state(struct g_raid3_disk *disk) 2391 { 2392 struct g_raid3_softc *sc; 2393 u_int state; 2394 2395 sc = disk->d_softc; 2396 if (sc->sc_syncid == disk->d_sync.ds_syncid) { 2397 if ((disk->d_flags & 2398 G_RAID3_DISK_FLAG_SYNCHRONIZING) == 0) { 2399 /* Disk does not need synchronization. */ 2400 state = G_RAID3_DISK_STATE_ACTIVE; 2401 } else { 2402 if ((sc->sc_flags & 2403 G_RAID3_DEVICE_FLAG_NOAUTOSYNC) == 0 || 2404 (disk->d_flags & 2405 G_RAID3_DISK_FLAG_FORCE_SYNC) != 0) { 2406 /* 2407 * We can start synchronization from 2408 * the stored offset. 2409 */ 2410 state = G_RAID3_DISK_STATE_SYNCHRONIZING; 2411 } else { 2412 state = G_RAID3_DISK_STATE_STALE; 2413 } 2414 } 2415 } else if (disk->d_sync.ds_syncid < sc->sc_syncid) { 2416 /* 2417 * Reset all synchronization data for this disk, 2418 * because if it even was synchronized, it was 2419 * synchronized to disks with different syncid. 2420 */ 2421 disk->d_flags |= G_RAID3_DISK_FLAG_SYNCHRONIZING; 2422 disk->d_sync.ds_offset = 0; 2423 disk->d_sync.ds_offset_done = 0; 2424 disk->d_sync.ds_syncid = sc->sc_syncid; 2425 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) == 0 || 2426 (disk->d_flags & G_RAID3_DISK_FLAG_FORCE_SYNC) != 0) { 2427 state = G_RAID3_DISK_STATE_SYNCHRONIZING; 2428 } else { 2429 state = G_RAID3_DISK_STATE_STALE; 2430 } 2431 } else /* if (sc->sc_syncid < disk->d_sync.ds_syncid) */ { 2432 /* 2433 * Not good, NOT GOOD! 2434 * It means that device was started on stale disks 2435 * and more fresh disk just arrive. 2436 * If there were writes, device is broken, sorry. 2437 * I think the best choice here is don't touch 2438 * this disk and inform the user loudly. 2439 */ 2440 G_RAID3_DEBUG(0, "Device %s was started before the freshest " 2441 "disk (%s) arrives!! It will not be connected to the " 2442 "running device.", sc->sc_name, 2443 g_raid3_get_diskname(disk)); 2444 g_raid3_destroy_disk(disk); 2445 state = G_RAID3_DISK_STATE_NONE; 2446 /* Return immediately, because disk was destroyed. */ 2447 return (state); 2448 } 2449 G_RAID3_DEBUG(3, "State for %s disk: %s.", 2450 g_raid3_get_diskname(disk), g_raid3_disk_state2str(state)); 2451 return (state); 2452 } 2453 2454 /* 2455 * Update device state. 2456 */ 2457 static void 2458 g_raid3_update_device(struct g_raid3_softc *sc, boolean_t force) 2459 { 2460 struct g_raid3_disk *disk; 2461 u_int state; 2462 2463 sx_assert(&sc->sc_lock, SX_XLOCKED); 2464 2465 switch (sc->sc_state) { 2466 case G_RAID3_DEVICE_STATE_STARTING: 2467 { 2468 u_int n, ndirty, ndisks, genid, syncid; 2469 2470 KASSERT(sc->sc_provider == NULL, 2471 ("Non-NULL provider in STARTING state (%s).", sc->sc_name)); 2472 /* 2473 * Are we ready? We are, if all disks are connected or 2474 * one disk is missing and 'force' is true. 2475 */ 2476 if (g_raid3_ndisks(sc, -1) + force == sc->sc_ndisks) { 2477 if (!force) 2478 callout_drain(&sc->sc_callout); 2479 } else { 2480 if (force) { 2481 /* 2482 * Timeout expired, so destroy device. 2483 */ 2484 sc->sc_flags |= G_RAID3_DEVICE_FLAG_DESTROY; 2485 G_RAID3_DEBUG(1, "root_mount_rel[%u] %p", 2486 __LINE__, sc->sc_rootmount); 2487 root_mount_rel(sc->sc_rootmount); 2488 sc->sc_rootmount = NULL; 2489 } 2490 return; 2491 } 2492 2493 /* 2494 * Find the biggest genid. 2495 */ 2496 genid = 0; 2497 for (n = 0; n < sc->sc_ndisks; n++) { 2498 disk = &sc->sc_disks[n]; 2499 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 2500 continue; 2501 if (disk->d_genid > genid) 2502 genid = disk->d_genid; 2503 } 2504 sc->sc_genid = genid; 2505 /* 2506 * Remove all disks without the biggest genid. 2507 */ 2508 for (n = 0; n < sc->sc_ndisks; n++) { 2509 disk = &sc->sc_disks[n]; 2510 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 2511 continue; 2512 if (disk->d_genid < genid) { 2513 G_RAID3_DEBUG(0, 2514 "Component %s (device %s) broken, skipping.", 2515 g_raid3_get_diskname(disk), sc->sc_name); 2516 g_raid3_destroy_disk(disk); 2517 } 2518 } 2519 2520 /* 2521 * There must be at least 'sc->sc_ndisks - 1' components 2522 * with the same syncid and without SYNCHRONIZING flag. 2523 */ 2524 2525 /* 2526 * Find the biggest syncid, number of valid components and 2527 * number of dirty components. 2528 */ 2529 ndirty = ndisks = syncid = 0; 2530 for (n = 0; n < sc->sc_ndisks; n++) { 2531 disk = &sc->sc_disks[n]; 2532 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 2533 continue; 2534 if ((disk->d_flags & G_RAID3_DISK_FLAG_DIRTY) != 0) 2535 ndirty++; 2536 if (disk->d_sync.ds_syncid > syncid) { 2537 syncid = disk->d_sync.ds_syncid; 2538 ndisks = 0; 2539 } else if (disk->d_sync.ds_syncid < syncid) { 2540 continue; 2541 } 2542 if ((disk->d_flags & 2543 G_RAID3_DISK_FLAG_SYNCHRONIZING) != 0) { 2544 continue; 2545 } 2546 ndisks++; 2547 } 2548 /* 2549 * Do we have enough valid components? 2550 */ 2551 if (ndisks + 1 < sc->sc_ndisks) { 2552 G_RAID3_DEBUG(0, 2553 "Device %s is broken, too few valid components.", 2554 sc->sc_name); 2555 sc->sc_flags |= G_RAID3_DEVICE_FLAG_DESTROY; 2556 return; 2557 } 2558 /* 2559 * If there is one DIRTY component and all disks are present, 2560 * mark it for synchronization. If there is more than one DIRTY 2561 * component, mark parity component for synchronization. 2562 */ 2563 if (ndisks == sc->sc_ndisks && ndirty == 1) { 2564 for (n = 0; n < sc->sc_ndisks; n++) { 2565 disk = &sc->sc_disks[n]; 2566 if ((disk->d_flags & 2567 G_RAID3_DISK_FLAG_DIRTY) == 0) { 2568 continue; 2569 } 2570 disk->d_flags |= 2571 G_RAID3_DISK_FLAG_SYNCHRONIZING; 2572 } 2573 } else if (ndisks == sc->sc_ndisks && ndirty > 1) { 2574 disk = &sc->sc_disks[sc->sc_ndisks - 1]; 2575 disk->d_flags |= G_RAID3_DISK_FLAG_SYNCHRONIZING; 2576 } 2577 2578 sc->sc_syncid = syncid; 2579 if (force) { 2580 /* Remember to bump syncid on first write. */ 2581 sc->sc_bump_id |= G_RAID3_BUMP_SYNCID; 2582 } 2583 if (ndisks == sc->sc_ndisks) 2584 state = G_RAID3_DEVICE_STATE_COMPLETE; 2585 else /* if (ndisks == sc->sc_ndisks - 1) */ 2586 state = G_RAID3_DEVICE_STATE_DEGRADED; 2587 G_RAID3_DEBUG(1, "Device %s state changed from %s to %s.", 2588 sc->sc_name, g_raid3_device_state2str(sc->sc_state), 2589 g_raid3_device_state2str(state)); 2590 sc->sc_state = state; 2591 for (n = 0; n < sc->sc_ndisks; n++) { 2592 disk = &sc->sc_disks[n]; 2593 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 2594 continue; 2595 state = g_raid3_determine_state(disk); 2596 g_raid3_event_send(disk, state, G_RAID3_EVENT_DONTWAIT); 2597 if (state == G_RAID3_DISK_STATE_STALE) 2598 sc->sc_bump_id |= G_RAID3_BUMP_SYNCID; 2599 } 2600 break; 2601 } 2602 case G_RAID3_DEVICE_STATE_DEGRADED: 2603 /* 2604 * Genid need to be bumped immediately, so do it here. 2605 */ 2606 if ((sc->sc_bump_id & G_RAID3_BUMP_GENID) != 0) { 2607 sc->sc_bump_id &= ~G_RAID3_BUMP_GENID; 2608 g_raid3_bump_genid(sc); 2609 } 2610 2611 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_NEW) > 0) 2612 return; 2613 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) < 2614 sc->sc_ndisks - 1) { 2615 if (sc->sc_provider != NULL) 2616 g_raid3_destroy_provider(sc); 2617 sc->sc_flags |= G_RAID3_DEVICE_FLAG_DESTROY; 2618 return; 2619 } 2620 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) == 2621 sc->sc_ndisks) { 2622 state = G_RAID3_DEVICE_STATE_COMPLETE; 2623 G_RAID3_DEBUG(1, 2624 "Device %s state changed from %s to %s.", 2625 sc->sc_name, g_raid3_device_state2str(sc->sc_state), 2626 g_raid3_device_state2str(state)); 2627 sc->sc_state = state; 2628 } 2629 if (sc->sc_provider == NULL) 2630 g_raid3_launch_provider(sc); 2631 if (sc->sc_rootmount != NULL) { 2632 G_RAID3_DEBUG(1, "root_mount_rel[%u] %p", __LINE__, 2633 sc->sc_rootmount); 2634 root_mount_rel(sc->sc_rootmount); 2635 sc->sc_rootmount = NULL; 2636 } 2637 break; 2638 case G_RAID3_DEVICE_STATE_COMPLETE: 2639 /* 2640 * Genid need to be bumped immediately, so do it here. 2641 */ 2642 if ((sc->sc_bump_id & G_RAID3_BUMP_GENID) != 0) { 2643 sc->sc_bump_id &= ~G_RAID3_BUMP_GENID; 2644 g_raid3_bump_genid(sc); 2645 } 2646 2647 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_NEW) > 0) 2648 return; 2649 KASSERT(g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) >= 2650 sc->sc_ndisks - 1, 2651 ("Too few ACTIVE components in COMPLETE state (device %s).", 2652 sc->sc_name)); 2653 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) == 2654 sc->sc_ndisks - 1) { 2655 state = G_RAID3_DEVICE_STATE_DEGRADED; 2656 G_RAID3_DEBUG(1, 2657 "Device %s state changed from %s to %s.", 2658 sc->sc_name, g_raid3_device_state2str(sc->sc_state), 2659 g_raid3_device_state2str(state)); 2660 sc->sc_state = state; 2661 } 2662 if (sc->sc_provider == NULL) 2663 g_raid3_launch_provider(sc); 2664 if (sc->sc_rootmount != NULL) { 2665 G_RAID3_DEBUG(1, "root_mount_rel[%u] %p", __LINE__, 2666 sc->sc_rootmount); 2667 root_mount_rel(sc->sc_rootmount); 2668 sc->sc_rootmount = NULL; 2669 } 2670 break; 2671 default: 2672 KASSERT(1 == 0, ("Wrong device state (%s, %s).", sc->sc_name, 2673 g_raid3_device_state2str(sc->sc_state))); 2674 break; 2675 } 2676 } 2677 2678 /* 2679 * Update disk state and device state if needed. 2680 */ 2681 #define DISK_STATE_CHANGED() G_RAID3_DEBUG(1, \ 2682 "Disk %s state changed from %s to %s (device %s).", \ 2683 g_raid3_get_diskname(disk), \ 2684 g_raid3_disk_state2str(disk->d_state), \ 2685 g_raid3_disk_state2str(state), sc->sc_name) 2686 static int 2687 g_raid3_update_disk(struct g_raid3_disk *disk, u_int state) 2688 { 2689 struct g_raid3_softc *sc; 2690 2691 sc = disk->d_softc; 2692 sx_assert(&sc->sc_lock, SX_XLOCKED); 2693 2694 again: 2695 G_RAID3_DEBUG(3, "Changing disk %s state from %s to %s.", 2696 g_raid3_get_diskname(disk), g_raid3_disk_state2str(disk->d_state), 2697 g_raid3_disk_state2str(state)); 2698 switch (state) { 2699 case G_RAID3_DISK_STATE_NEW: 2700 /* 2701 * Possible scenarios: 2702 * 1. New disk arrive. 2703 */ 2704 /* Previous state should be NONE. */ 2705 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NONE, 2706 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 2707 g_raid3_disk_state2str(disk->d_state))); 2708 DISK_STATE_CHANGED(); 2709 2710 disk->d_state = state; 2711 G_RAID3_DEBUG(1, "Device %s: provider %s detected.", 2712 sc->sc_name, g_raid3_get_diskname(disk)); 2713 if (sc->sc_state == G_RAID3_DEVICE_STATE_STARTING) 2714 break; 2715 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2716 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE, 2717 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2718 g_raid3_device_state2str(sc->sc_state), 2719 g_raid3_get_diskname(disk), 2720 g_raid3_disk_state2str(disk->d_state))); 2721 state = g_raid3_determine_state(disk); 2722 if (state != G_RAID3_DISK_STATE_NONE) 2723 goto again; 2724 break; 2725 case G_RAID3_DISK_STATE_ACTIVE: 2726 /* 2727 * Possible scenarios: 2728 * 1. New disk does not need synchronization. 2729 * 2. Synchronization process finished successfully. 2730 */ 2731 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2732 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE, 2733 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2734 g_raid3_device_state2str(sc->sc_state), 2735 g_raid3_get_diskname(disk), 2736 g_raid3_disk_state2str(disk->d_state))); 2737 /* Previous state should be NEW or SYNCHRONIZING. */ 2738 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NEW || 2739 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING, 2740 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 2741 g_raid3_disk_state2str(disk->d_state))); 2742 DISK_STATE_CHANGED(); 2743 2744 if (disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 2745 disk->d_flags &= ~G_RAID3_DISK_FLAG_SYNCHRONIZING; 2746 disk->d_flags &= ~G_RAID3_DISK_FLAG_FORCE_SYNC; 2747 g_raid3_sync_stop(sc, 0); 2748 } 2749 disk->d_state = state; 2750 disk->d_sync.ds_offset = 0; 2751 disk->d_sync.ds_offset_done = 0; 2752 g_raid3_update_idle(sc, disk); 2753 g_raid3_update_metadata(disk); 2754 G_RAID3_DEBUG(1, "Device %s: provider %s activated.", 2755 sc->sc_name, g_raid3_get_diskname(disk)); 2756 break; 2757 case G_RAID3_DISK_STATE_STALE: 2758 /* 2759 * Possible scenarios: 2760 * 1. Stale disk was connected. 2761 */ 2762 /* Previous state should be NEW. */ 2763 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NEW, 2764 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 2765 g_raid3_disk_state2str(disk->d_state))); 2766 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2767 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE, 2768 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2769 g_raid3_device_state2str(sc->sc_state), 2770 g_raid3_get_diskname(disk), 2771 g_raid3_disk_state2str(disk->d_state))); 2772 /* 2773 * STALE state is only possible if device is marked 2774 * NOAUTOSYNC. 2775 */ 2776 KASSERT((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0, 2777 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2778 g_raid3_device_state2str(sc->sc_state), 2779 g_raid3_get_diskname(disk), 2780 g_raid3_disk_state2str(disk->d_state))); 2781 DISK_STATE_CHANGED(); 2782 2783 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 2784 disk->d_state = state; 2785 g_raid3_update_metadata(disk); 2786 G_RAID3_DEBUG(0, "Device %s: provider %s is stale.", 2787 sc->sc_name, g_raid3_get_diskname(disk)); 2788 break; 2789 case G_RAID3_DISK_STATE_SYNCHRONIZING: 2790 /* 2791 * Possible scenarios: 2792 * 1. Disk which needs synchronization was connected. 2793 */ 2794 /* Previous state should be NEW. */ 2795 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NEW, 2796 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 2797 g_raid3_disk_state2str(disk->d_state))); 2798 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2799 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE, 2800 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2801 g_raid3_device_state2str(sc->sc_state), 2802 g_raid3_get_diskname(disk), 2803 g_raid3_disk_state2str(disk->d_state))); 2804 DISK_STATE_CHANGED(); 2805 2806 if (disk->d_state == G_RAID3_DISK_STATE_NEW) 2807 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 2808 disk->d_state = state; 2809 if (sc->sc_provider != NULL) { 2810 g_raid3_sync_start(sc); 2811 g_raid3_update_metadata(disk); 2812 } 2813 break; 2814 case G_RAID3_DISK_STATE_DISCONNECTED: 2815 /* 2816 * Possible scenarios: 2817 * 1. Device wasn't running yet, but disk disappear. 2818 * 2. Disk was active and disapppear. 2819 * 3. Disk disappear during synchronization process. 2820 */ 2821 if (sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2822 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE) { 2823 /* 2824 * Previous state should be ACTIVE, STALE or 2825 * SYNCHRONIZING. 2826 */ 2827 KASSERT(disk->d_state == G_RAID3_DISK_STATE_ACTIVE || 2828 disk->d_state == G_RAID3_DISK_STATE_STALE || 2829 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING, 2830 ("Wrong disk state (%s, %s).", 2831 g_raid3_get_diskname(disk), 2832 g_raid3_disk_state2str(disk->d_state))); 2833 } else if (sc->sc_state == G_RAID3_DEVICE_STATE_STARTING) { 2834 /* Previous state should be NEW. */ 2835 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NEW, 2836 ("Wrong disk state (%s, %s).", 2837 g_raid3_get_diskname(disk), 2838 g_raid3_disk_state2str(disk->d_state))); 2839 /* 2840 * Reset bumping syncid if disk disappeared in STARTING 2841 * state. 2842 */ 2843 if ((sc->sc_bump_id & G_RAID3_BUMP_SYNCID) != 0) 2844 sc->sc_bump_id &= ~G_RAID3_BUMP_SYNCID; 2845 #ifdef INVARIANTS 2846 } else { 2847 KASSERT(1 == 0, ("Wrong device state (%s, %s, %s, %s).", 2848 sc->sc_name, 2849 g_raid3_device_state2str(sc->sc_state), 2850 g_raid3_get_diskname(disk), 2851 g_raid3_disk_state2str(disk->d_state))); 2852 #endif 2853 } 2854 DISK_STATE_CHANGED(); 2855 G_RAID3_DEBUG(0, "Device %s: provider %s disconnected.", 2856 sc->sc_name, g_raid3_get_diskname(disk)); 2857 2858 g_raid3_destroy_disk(disk); 2859 break; 2860 default: 2861 KASSERT(1 == 0, ("Unknown state (%u).", state)); 2862 break; 2863 } 2864 return (0); 2865 } 2866 #undef DISK_STATE_CHANGED 2867 2868 int 2869 g_raid3_read_metadata(struct g_consumer *cp, struct g_raid3_metadata *md) 2870 { 2871 struct g_provider *pp; 2872 u_char *buf; 2873 int error; 2874 2875 g_topology_assert(); 2876 2877 error = g_access(cp, 1, 0, 0); 2878 if (error != 0) 2879 return (error); 2880 pp = cp->provider; 2881 g_topology_unlock(); 2882 /* Metadata are stored on last sector. */ 2883 buf = g_read_data(cp, pp->mediasize - pp->sectorsize, pp->sectorsize, 2884 &error); 2885 g_topology_lock(); 2886 g_access(cp, -1, 0, 0); 2887 if (buf == NULL) { 2888 G_RAID3_DEBUG(1, "Cannot read metadata from %s (error=%d).", 2889 cp->provider->name, error); 2890 return (error); 2891 } 2892 2893 /* Decode metadata. */ 2894 error = raid3_metadata_decode(buf, md); 2895 g_free(buf); 2896 if (strcmp(md->md_magic, G_RAID3_MAGIC) != 0) 2897 return (EINVAL); 2898 if (md->md_version > G_RAID3_VERSION) { 2899 G_RAID3_DEBUG(0, 2900 "Kernel module is too old to handle metadata from %s.", 2901 cp->provider->name); 2902 return (EINVAL); 2903 } 2904 if (error != 0) { 2905 G_RAID3_DEBUG(1, "MD5 metadata hash mismatch for provider %s.", 2906 cp->provider->name); 2907 return (error); 2908 } 2909 if (md->md_sectorsize > MAXPHYS) { 2910 G_RAID3_DEBUG(0, "The blocksize is too big."); 2911 return (EINVAL); 2912 } 2913 2914 return (0); 2915 } 2916 2917 static int 2918 g_raid3_check_metadata(struct g_raid3_softc *sc, struct g_provider *pp, 2919 struct g_raid3_metadata *md) 2920 { 2921 2922 if (md->md_no >= sc->sc_ndisks) { 2923 G_RAID3_DEBUG(1, "Invalid disk %s number (no=%u), skipping.", 2924 pp->name, md->md_no); 2925 return (EINVAL); 2926 } 2927 if (sc->sc_disks[md->md_no].d_state != G_RAID3_DISK_STATE_NODISK) { 2928 G_RAID3_DEBUG(1, "Disk %s (no=%u) already exists, skipping.", 2929 pp->name, md->md_no); 2930 return (EEXIST); 2931 } 2932 if (md->md_all != sc->sc_ndisks) { 2933 G_RAID3_DEBUG(1, 2934 "Invalid '%s' field on disk %s (device %s), skipping.", 2935 "md_all", pp->name, sc->sc_name); 2936 return (EINVAL); 2937 } 2938 if ((md->md_mediasize % md->md_sectorsize) != 0) { 2939 G_RAID3_DEBUG(1, "Invalid metadata (mediasize %% sectorsize != " 2940 "0) on disk %s (device %s), skipping.", pp->name, 2941 sc->sc_name); 2942 return (EINVAL); 2943 } 2944 if (md->md_mediasize != sc->sc_mediasize) { 2945 G_RAID3_DEBUG(1, 2946 "Invalid '%s' field on disk %s (device %s), skipping.", 2947 "md_mediasize", pp->name, sc->sc_name); 2948 return (EINVAL); 2949 } 2950 if ((md->md_mediasize % (sc->sc_ndisks - 1)) != 0) { 2951 G_RAID3_DEBUG(1, 2952 "Invalid '%s' field on disk %s (device %s), skipping.", 2953 "md_mediasize", pp->name, sc->sc_name); 2954 return (EINVAL); 2955 } 2956 if ((sc->sc_mediasize / (sc->sc_ndisks - 1)) > pp->mediasize) { 2957 G_RAID3_DEBUG(1, 2958 "Invalid size of disk %s (device %s), skipping.", pp->name, 2959 sc->sc_name); 2960 return (EINVAL); 2961 } 2962 if ((md->md_sectorsize / pp->sectorsize) < sc->sc_ndisks - 1) { 2963 G_RAID3_DEBUG(1, 2964 "Invalid '%s' field on disk %s (device %s), skipping.", 2965 "md_sectorsize", pp->name, sc->sc_name); 2966 return (EINVAL); 2967 } 2968 if (md->md_sectorsize != sc->sc_sectorsize) { 2969 G_RAID3_DEBUG(1, 2970 "Invalid '%s' field on disk %s (device %s), skipping.", 2971 "md_sectorsize", pp->name, sc->sc_name); 2972 return (EINVAL); 2973 } 2974 if ((sc->sc_sectorsize % pp->sectorsize) != 0) { 2975 G_RAID3_DEBUG(1, 2976 "Invalid sector size of disk %s (device %s), skipping.", 2977 pp->name, sc->sc_name); 2978 return (EINVAL); 2979 } 2980 if ((md->md_mflags & ~G_RAID3_DEVICE_FLAG_MASK) != 0) { 2981 G_RAID3_DEBUG(1, 2982 "Invalid device flags on disk %s (device %s), skipping.", 2983 pp->name, sc->sc_name); 2984 return (EINVAL); 2985 } 2986 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_VERIFY) != 0 && 2987 (md->md_mflags & G_RAID3_DEVICE_FLAG_ROUND_ROBIN) != 0) { 2988 /* 2989 * VERIFY and ROUND-ROBIN options are mutally exclusive. 2990 */ 2991 G_RAID3_DEBUG(1, "Both VERIFY and ROUND-ROBIN flags exist on " 2992 "disk %s (device %s), skipping.", pp->name, sc->sc_name); 2993 return (EINVAL); 2994 } 2995 if ((md->md_dflags & ~G_RAID3_DISK_FLAG_MASK) != 0) { 2996 G_RAID3_DEBUG(1, 2997 "Invalid disk flags on disk %s (device %s), skipping.", 2998 pp->name, sc->sc_name); 2999 return (EINVAL); 3000 } 3001 return (0); 3002 } 3003 3004 int 3005 g_raid3_add_disk(struct g_raid3_softc *sc, struct g_provider *pp, 3006 struct g_raid3_metadata *md) 3007 { 3008 struct g_raid3_disk *disk; 3009 int error; 3010 3011 g_topology_assert_not(); 3012 G_RAID3_DEBUG(2, "Adding disk %s.", pp->name); 3013 3014 error = g_raid3_check_metadata(sc, pp, md); 3015 if (error != 0) 3016 return (error); 3017 if (sc->sc_state != G_RAID3_DEVICE_STATE_STARTING && 3018 md->md_genid < sc->sc_genid) { 3019 G_RAID3_DEBUG(0, "Component %s (device %s) broken, skipping.", 3020 pp->name, sc->sc_name); 3021 return (EINVAL); 3022 } 3023 disk = g_raid3_init_disk(sc, pp, md, &error); 3024 if (disk == NULL) 3025 return (error); 3026 error = g_raid3_event_send(disk, G_RAID3_DISK_STATE_NEW, 3027 G_RAID3_EVENT_WAIT); 3028 if (error != 0) 3029 return (error); 3030 if (md->md_version < G_RAID3_VERSION) { 3031 G_RAID3_DEBUG(0, "Upgrading metadata on %s (v%d->v%d).", 3032 pp->name, md->md_version, G_RAID3_VERSION); 3033 g_raid3_update_metadata(disk); 3034 } 3035 return (0); 3036 } 3037 3038 static void 3039 g_raid3_destroy_delayed(void *arg, int flag) 3040 { 3041 struct g_raid3_softc *sc; 3042 int error; 3043 3044 if (flag == EV_CANCEL) { 3045 G_RAID3_DEBUG(1, "Destroying canceled."); 3046 return; 3047 } 3048 sc = arg; 3049 g_topology_unlock(); 3050 sx_xlock(&sc->sc_lock); 3051 KASSERT((sc->sc_flags & G_RAID3_DEVICE_FLAG_DESTROY) == 0, 3052 ("DESTROY flag set on %s.", sc->sc_name)); 3053 KASSERT((sc->sc_flags & G_RAID3_DEVICE_FLAG_DESTROYING) != 0, 3054 ("DESTROYING flag not set on %s.", sc->sc_name)); 3055 G_RAID3_DEBUG(0, "Destroying %s (delayed).", sc->sc_name); 3056 error = g_raid3_destroy(sc, G_RAID3_DESTROY_SOFT); 3057 if (error != 0) { 3058 G_RAID3_DEBUG(0, "Cannot destroy %s.", sc->sc_name); 3059 sx_xunlock(&sc->sc_lock); 3060 } 3061 g_topology_lock(); 3062 } 3063 3064 static int 3065 g_raid3_access(struct g_provider *pp, int acr, int acw, int ace) 3066 { 3067 struct g_raid3_softc *sc; 3068 int dcr, dcw, dce, error = 0; 3069 3070 g_topology_assert(); 3071 G_RAID3_DEBUG(2, "Access request for %s: r%dw%de%d.", pp->name, acr, 3072 acw, ace); 3073 3074 sc = pp->geom->softc; 3075 if (sc == NULL && acr <= 0 && acw <= 0 && ace <= 0) 3076 return (0); 3077 KASSERT(sc != NULL, ("NULL softc (provider=%s).", pp->name)); 3078 3079 dcr = pp->acr + acr; 3080 dcw = pp->acw + acw; 3081 dce = pp->ace + ace; 3082 3083 g_topology_unlock(); 3084 sx_xlock(&sc->sc_lock); 3085 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_DESTROY) != 0 || 3086 g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) < sc->sc_ndisks - 1) { 3087 if (acr > 0 || acw > 0 || ace > 0) 3088 error = ENXIO; 3089 goto end; 3090 } 3091 if (dcw == 0) 3092 g_raid3_idle(sc, dcw); 3093 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_DESTROYING) != 0) { 3094 if (acr > 0 || acw > 0 || ace > 0) { 3095 error = ENXIO; 3096 goto end; 3097 } 3098 if (dcr == 0 && dcw == 0 && dce == 0) { 3099 g_post_event(g_raid3_destroy_delayed, sc, M_WAITOK, 3100 sc, NULL); 3101 } 3102 } 3103 end: 3104 sx_xunlock(&sc->sc_lock); 3105 g_topology_lock(); 3106 return (error); 3107 } 3108 3109 static struct g_geom * 3110 g_raid3_create(struct g_class *mp, const struct g_raid3_metadata *md) 3111 { 3112 struct g_raid3_softc *sc; 3113 struct g_geom *gp; 3114 int error, timeout; 3115 u_int n; 3116 3117 g_topology_assert(); 3118 G_RAID3_DEBUG(1, "Creating device %s (id=%u).", md->md_name, md->md_id); 3119 3120 /* One disk is minimum. */ 3121 if (md->md_all < 1) 3122 return (NULL); 3123 /* 3124 * Action geom. 3125 */ 3126 gp = g_new_geomf(mp, "%s", md->md_name); 3127 sc = malloc(sizeof(*sc), M_RAID3, M_WAITOK | M_ZERO); 3128 sc->sc_disks = malloc(sizeof(struct g_raid3_disk) * md->md_all, M_RAID3, 3129 M_WAITOK | M_ZERO); 3130 gp->start = g_raid3_start; 3131 gp->orphan = g_raid3_orphan; 3132 gp->access = g_raid3_access; 3133 gp->dumpconf = g_raid3_dumpconf; 3134 3135 sc->sc_id = md->md_id; 3136 sc->sc_mediasize = md->md_mediasize; 3137 sc->sc_sectorsize = md->md_sectorsize; 3138 sc->sc_ndisks = md->md_all; 3139 sc->sc_round_robin = 0; 3140 sc->sc_flags = md->md_mflags; 3141 sc->sc_bump_id = 0; 3142 sc->sc_idle = 1; 3143 sc->sc_last_write = time_uptime; 3144 sc->sc_writes = 0; 3145 for (n = 0; n < sc->sc_ndisks; n++) { 3146 sc->sc_disks[n].d_softc = sc; 3147 sc->sc_disks[n].d_no = n; 3148 sc->sc_disks[n].d_state = G_RAID3_DISK_STATE_NODISK; 3149 } 3150 sx_init(&sc->sc_lock, "graid3:lock"); 3151 bioq_init(&sc->sc_queue); 3152 mtx_init(&sc->sc_queue_mtx, "graid3:queue", NULL, MTX_DEF); 3153 bioq_init(&sc->sc_regular_delayed); 3154 bioq_init(&sc->sc_inflight); 3155 bioq_init(&sc->sc_sync_delayed); 3156 TAILQ_INIT(&sc->sc_events); 3157 mtx_init(&sc->sc_events_mtx, "graid3:events", NULL, MTX_DEF); 3158 callout_init(&sc->sc_callout, 1); 3159 sc->sc_state = G_RAID3_DEVICE_STATE_STARTING; 3160 gp->softc = sc; 3161 sc->sc_geom = gp; 3162 sc->sc_provider = NULL; 3163 /* 3164 * Synchronization geom. 3165 */ 3166 gp = g_new_geomf(mp, "%s.sync", md->md_name); 3167 gp->softc = sc; 3168 gp->orphan = g_raid3_orphan; 3169 sc->sc_sync.ds_geom = gp; 3170 3171 if (!g_raid3_use_malloc) { 3172 sc->sc_zones[G_RAID3_ZONE_64K].sz_zone = uma_zcreate("gr3:64k", 3173 65536, g_raid3_uma_ctor, g_raid3_uma_dtor, NULL, NULL, 3174 UMA_ALIGN_PTR, 0); 3175 sc->sc_zones[G_RAID3_ZONE_64K].sz_inuse = 0; 3176 sc->sc_zones[G_RAID3_ZONE_64K].sz_max = g_raid3_n64k; 3177 sc->sc_zones[G_RAID3_ZONE_64K].sz_requested = 3178 sc->sc_zones[G_RAID3_ZONE_64K].sz_failed = 0; 3179 sc->sc_zones[G_RAID3_ZONE_16K].sz_zone = uma_zcreate("gr3:16k", 3180 16384, g_raid3_uma_ctor, g_raid3_uma_dtor, NULL, NULL, 3181 UMA_ALIGN_PTR, 0); 3182 sc->sc_zones[G_RAID3_ZONE_16K].sz_inuse = 0; 3183 sc->sc_zones[G_RAID3_ZONE_16K].sz_max = g_raid3_n16k; 3184 sc->sc_zones[G_RAID3_ZONE_16K].sz_requested = 3185 sc->sc_zones[G_RAID3_ZONE_16K].sz_failed = 0; 3186 sc->sc_zones[G_RAID3_ZONE_4K].sz_zone = uma_zcreate("gr3:4k", 3187 4096, g_raid3_uma_ctor, g_raid3_uma_dtor, NULL, NULL, 3188 UMA_ALIGN_PTR, 0); 3189 sc->sc_zones[G_RAID3_ZONE_4K].sz_inuse = 0; 3190 sc->sc_zones[G_RAID3_ZONE_4K].sz_max = g_raid3_n4k; 3191 sc->sc_zones[G_RAID3_ZONE_4K].sz_requested = 3192 sc->sc_zones[G_RAID3_ZONE_4K].sz_failed = 0; 3193 } 3194 3195 error = kproc_create(g_raid3_worker, sc, &sc->sc_worker, 0, 0, 3196 "g_raid3 %s", md->md_name); 3197 if (error != 0) { 3198 G_RAID3_DEBUG(1, "Cannot create kernel thread for %s.", 3199 sc->sc_name); 3200 if (!g_raid3_use_malloc) { 3201 uma_zdestroy(sc->sc_zones[G_RAID3_ZONE_64K].sz_zone); 3202 uma_zdestroy(sc->sc_zones[G_RAID3_ZONE_16K].sz_zone); 3203 uma_zdestroy(sc->sc_zones[G_RAID3_ZONE_4K].sz_zone); 3204 } 3205 g_destroy_geom(sc->sc_sync.ds_geom); 3206 mtx_destroy(&sc->sc_events_mtx); 3207 mtx_destroy(&sc->sc_queue_mtx); 3208 sx_destroy(&sc->sc_lock); 3209 g_destroy_geom(sc->sc_geom); 3210 free(sc->sc_disks, M_RAID3); 3211 free(sc, M_RAID3); 3212 return (NULL); 3213 } 3214 3215 G_RAID3_DEBUG(1, "Device %s created (%u components, id=%u).", 3216 sc->sc_name, sc->sc_ndisks, sc->sc_id); 3217 3218 sc->sc_rootmount = root_mount_hold("GRAID3"); 3219 G_RAID3_DEBUG(1, "root_mount_hold %p", sc->sc_rootmount); 3220 3221 /* 3222 * Run timeout. 3223 */ 3224 timeout = atomic_load_acq_int(&g_raid3_timeout); 3225 callout_reset(&sc->sc_callout, timeout * hz, g_raid3_go, sc); 3226 return (sc->sc_geom); 3227 } 3228 3229 int 3230 g_raid3_destroy(struct g_raid3_softc *sc, int how) 3231 { 3232 struct g_provider *pp; 3233 3234 g_topology_assert_not(); 3235 if (sc == NULL) 3236 return (ENXIO); 3237 sx_assert(&sc->sc_lock, SX_XLOCKED); 3238 3239 pp = sc->sc_provider; 3240 if (pp != NULL && (pp->acr != 0 || pp->acw != 0 || pp->ace != 0)) { 3241 switch (how) { 3242 case G_RAID3_DESTROY_SOFT: 3243 G_RAID3_DEBUG(1, 3244 "Device %s is still open (r%dw%de%d).", pp->name, 3245 pp->acr, pp->acw, pp->ace); 3246 return (EBUSY); 3247 case G_RAID3_DESTROY_DELAYED: 3248 G_RAID3_DEBUG(1, 3249 "Device %s will be destroyed on last close.", 3250 pp->name); 3251 if (sc->sc_syncdisk != NULL) 3252 g_raid3_sync_stop(sc, 1); 3253 sc->sc_flags |= G_RAID3_DEVICE_FLAG_DESTROYING; 3254 return (EBUSY); 3255 case G_RAID3_DESTROY_HARD: 3256 G_RAID3_DEBUG(1, "Device %s is still open, so it " 3257 "can't be definitely removed.", pp->name); 3258 break; 3259 } 3260 } 3261 3262 g_topology_lock(); 3263 if (sc->sc_geom->softc == NULL) { 3264 g_topology_unlock(); 3265 return (0); 3266 } 3267 sc->sc_geom->softc = NULL; 3268 sc->sc_sync.ds_geom->softc = NULL; 3269 g_topology_unlock(); 3270 3271 sc->sc_flags |= G_RAID3_DEVICE_FLAG_DESTROY; 3272 sc->sc_flags |= G_RAID3_DEVICE_FLAG_WAIT; 3273 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, sc); 3274 sx_xunlock(&sc->sc_lock); 3275 mtx_lock(&sc->sc_queue_mtx); 3276 wakeup(sc); 3277 wakeup(&sc->sc_queue); 3278 mtx_unlock(&sc->sc_queue_mtx); 3279 G_RAID3_DEBUG(4, "%s: Sleeping %p.", __func__, &sc->sc_worker); 3280 while (sc->sc_worker != NULL) 3281 tsleep(&sc->sc_worker, PRIBIO, "r3:destroy", hz / 5); 3282 G_RAID3_DEBUG(4, "%s: Woken up %p.", __func__, &sc->sc_worker); 3283 sx_xlock(&sc->sc_lock); 3284 g_raid3_destroy_device(sc); 3285 free(sc->sc_disks, M_RAID3); 3286 free(sc, M_RAID3); 3287 return (0); 3288 } 3289 3290 static void 3291 g_raid3_taste_orphan(struct g_consumer *cp) 3292 { 3293 3294 KASSERT(1 == 0, ("%s called while tasting %s.", __func__, 3295 cp->provider->name)); 3296 } 3297 3298 static struct g_geom * 3299 g_raid3_taste(struct g_class *mp, struct g_provider *pp, int flags __unused) 3300 { 3301 struct g_raid3_metadata md; 3302 struct g_raid3_softc *sc; 3303 struct g_consumer *cp; 3304 struct g_geom *gp; 3305 int error; 3306 3307 g_topology_assert(); 3308 g_trace(G_T_TOPOLOGY, "%s(%s, %s)", __func__, mp->name, pp->name); 3309 G_RAID3_DEBUG(2, "Tasting %s.", pp->name); 3310 3311 gp = g_new_geomf(mp, "raid3:taste"); 3312 /* This orphan function should be never called. */ 3313 gp->orphan = g_raid3_taste_orphan; 3314 cp = g_new_consumer(gp); 3315 g_attach(cp, pp); 3316 error = g_raid3_read_metadata(cp, &md); 3317 g_detach(cp); 3318 g_destroy_consumer(cp); 3319 g_destroy_geom(gp); 3320 if (error != 0) 3321 return (NULL); 3322 gp = NULL; 3323 3324 if (md.md_provider[0] != '\0' && 3325 !g_compare_names(md.md_provider, pp->name)) 3326 return (NULL); 3327 if (md.md_provsize != 0 && md.md_provsize != pp->mediasize) 3328 return (NULL); 3329 if (g_raid3_debug >= 2) 3330 raid3_metadata_dump(&md); 3331 3332 /* 3333 * Let's check if device already exists. 3334 */ 3335 sc = NULL; 3336 LIST_FOREACH(gp, &mp->geom, geom) { 3337 sc = gp->softc; 3338 if (sc == NULL) 3339 continue; 3340 if (sc->sc_sync.ds_geom == gp) 3341 continue; 3342 if (strcmp(md.md_name, sc->sc_name) != 0) 3343 continue; 3344 if (md.md_id != sc->sc_id) { 3345 G_RAID3_DEBUG(0, "Device %s already configured.", 3346 sc->sc_name); 3347 return (NULL); 3348 } 3349 break; 3350 } 3351 if (gp == NULL) { 3352 gp = g_raid3_create(mp, &md); 3353 if (gp == NULL) { 3354 G_RAID3_DEBUG(0, "Cannot create device %s.", 3355 md.md_name); 3356 return (NULL); 3357 } 3358 sc = gp->softc; 3359 } 3360 G_RAID3_DEBUG(1, "Adding disk %s to %s.", pp->name, gp->name); 3361 g_topology_unlock(); 3362 sx_xlock(&sc->sc_lock); 3363 error = g_raid3_add_disk(sc, pp, &md); 3364 if (error != 0) { 3365 G_RAID3_DEBUG(0, "Cannot add disk %s to %s (error=%d).", 3366 pp->name, gp->name, error); 3367 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_NODISK) == 3368 sc->sc_ndisks) { 3369 g_cancel_event(sc); 3370 g_raid3_destroy(sc, G_RAID3_DESTROY_HARD); 3371 g_topology_lock(); 3372 return (NULL); 3373 } 3374 gp = NULL; 3375 } 3376 sx_xunlock(&sc->sc_lock); 3377 g_topology_lock(); 3378 return (gp); 3379 } 3380 3381 static int 3382 g_raid3_destroy_geom(struct gctl_req *req __unused, struct g_class *mp __unused, 3383 struct g_geom *gp) 3384 { 3385 struct g_raid3_softc *sc; 3386 int error; 3387 3388 g_topology_unlock(); 3389 sc = gp->softc; 3390 sx_xlock(&sc->sc_lock); 3391 g_cancel_event(sc); 3392 error = g_raid3_destroy(gp->softc, G_RAID3_DESTROY_SOFT); 3393 if (error != 0) 3394 sx_xunlock(&sc->sc_lock); 3395 g_topology_lock(); 3396 return (error); 3397 } 3398 3399 static void 3400 g_raid3_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, 3401 struct g_consumer *cp, struct g_provider *pp) 3402 { 3403 struct g_raid3_softc *sc; 3404 3405 g_topology_assert(); 3406 3407 sc = gp->softc; 3408 if (sc == NULL) 3409 return; 3410 /* Skip synchronization geom. */ 3411 if (gp == sc->sc_sync.ds_geom) 3412 return; 3413 if (pp != NULL) { 3414 /* Nothing here. */ 3415 } else if (cp != NULL) { 3416 struct g_raid3_disk *disk; 3417 3418 disk = cp->private; 3419 if (disk == NULL) 3420 return; 3421 g_topology_unlock(); 3422 sx_xlock(&sc->sc_lock); 3423 sbuf_printf(sb, "%s<Type>", indent); 3424 if (disk->d_no == sc->sc_ndisks - 1) 3425 sbuf_printf(sb, "PARITY"); 3426 else 3427 sbuf_printf(sb, "DATA"); 3428 sbuf_printf(sb, "</Type>\n"); 3429 sbuf_printf(sb, "%s<Number>%u</Number>\n", indent, 3430 (u_int)disk->d_no); 3431 if (disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 3432 sbuf_printf(sb, "%s<Synchronized>", indent); 3433 if (disk->d_sync.ds_offset == 0) 3434 sbuf_printf(sb, "0%%"); 3435 else { 3436 sbuf_printf(sb, "%u%%", 3437 (u_int)((disk->d_sync.ds_offset * 100) / 3438 (sc->sc_mediasize / (sc->sc_ndisks - 1)))); 3439 } 3440 sbuf_printf(sb, "</Synchronized>\n"); 3441 if (disk->d_sync.ds_offset > 0) { 3442 sbuf_printf(sb, "%s<BytesSynced>%jd" 3443 "</BytesSynced>\n", indent, 3444 (intmax_t)disk->d_sync.ds_offset); 3445 } 3446 } 3447 sbuf_printf(sb, "%s<SyncID>%u</SyncID>\n", indent, 3448 disk->d_sync.ds_syncid); 3449 sbuf_printf(sb, "%s<GenID>%u</GenID>\n", indent, disk->d_genid); 3450 sbuf_printf(sb, "%s<Flags>", indent); 3451 if (disk->d_flags == 0) 3452 sbuf_printf(sb, "NONE"); 3453 else { 3454 int first = 1; 3455 3456 #define ADD_FLAG(flag, name) do { \ 3457 if ((disk->d_flags & (flag)) != 0) { \ 3458 if (!first) \ 3459 sbuf_printf(sb, ", "); \ 3460 else \ 3461 first = 0; \ 3462 sbuf_printf(sb, name); \ 3463 } \ 3464 } while (0) 3465 ADD_FLAG(G_RAID3_DISK_FLAG_DIRTY, "DIRTY"); 3466 ADD_FLAG(G_RAID3_DISK_FLAG_HARDCODED, "HARDCODED"); 3467 ADD_FLAG(G_RAID3_DISK_FLAG_SYNCHRONIZING, 3468 "SYNCHRONIZING"); 3469 ADD_FLAG(G_RAID3_DISK_FLAG_FORCE_SYNC, "FORCE_SYNC"); 3470 ADD_FLAG(G_RAID3_DISK_FLAG_BROKEN, "BROKEN"); 3471 #undef ADD_FLAG 3472 } 3473 sbuf_printf(sb, "</Flags>\n"); 3474 sbuf_printf(sb, "%s<State>%s</State>\n", indent, 3475 g_raid3_disk_state2str(disk->d_state)); 3476 sx_xunlock(&sc->sc_lock); 3477 g_topology_lock(); 3478 } else { 3479 g_topology_unlock(); 3480 sx_xlock(&sc->sc_lock); 3481 if (!g_raid3_use_malloc) { 3482 sbuf_printf(sb, 3483 "%s<Zone4kRequested>%u</Zone4kRequested>\n", indent, 3484 sc->sc_zones[G_RAID3_ZONE_4K].sz_requested); 3485 sbuf_printf(sb, 3486 "%s<Zone4kFailed>%u</Zone4kFailed>\n", indent, 3487 sc->sc_zones[G_RAID3_ZONE_4K].sz_failed); 3488 sbuf_printf(sb, 3489 "%s<Zone16kRequested>%u</Zone16kRequested>\n", indent, 3490 sc->sc_zones[G_RAID3_ZONE_16K].sz_requested); 3491 sbuf_printf(sb, 3492 "%s<Zone16kFailed>%u</Zone16kFailed>\n", indent, 3493 sc->sc_zones[G_RAID3_ZONE_16K].sz_failed); 3494 sbuf_printf(sb, 3495 "%s<Zone64kRequested>%u</Zone64kRequested>\n", indent, 3496 sc->sc_zones[G_RAID3_ZONE_64K].sz_requested); 3497 sbuf_printf(sb, 3498 "%s<Zone64kFailed>%u</Zone64kFailed>\n", indent, 3499 sc->sc_zones[G_RAID3_ZONE_64K].sz_failed); 3500 } 3501 sbuf_printf(sb, "%s<ID>%u</ID>\n", indent, (u_int)sc->sc_id); 3502 sbuf_printf(sb, "%s<SyncID>%u</SyncID>\n", indent, sc->sc_syncid); 3503 sbuf_printf(sb, "%s<GenID>%u</GenID>\n", indent, sc->sc_genid); 3504 sbuf_printf(sb, "%s<Flags>", indent); 3505 if (sc->sc_flags == 0) 3506 sbuf_printf(sb, "NONE"); 3507 else { 3508 int first = 1; 3509 3510 #define ADD_FLAG(flag, name) do { \ 3511 if ((sc->sc_flags & (flag)) != 0) { \ 3512 if (!first) \ 3513 sbuf_printf(sb, ", "); \ 3514 else \ 3515 first = 0; \ 3516 sbuf_printf(sb, name); \ 3517 } \ 3518 } while (0) 3519 ADD_FLAG(G_RAID3_DEVICE_FLAG_NOFAILSYNC, "NOFAILSYNC"); 3520 ADD_FLAG(G_RAID3_DEVICE_FLAG_NOAUTOSYNC, "NOAUTOSYNC"); 3521 ADD_FLAG(G_RAID3_DEVICE_FLAG_ROUND_ROBIN, 3522 "ROUND-ROBIN"); 3523 ADD_FLAG(G_RAID3_DEVICE_FLAG_VERIFY, "VERIFY"); 3524 #undef ADD_FLAG 3525 } 3526 sbuf_printf(sb, "</Flags>\n"); 3527 sbuf_printf(sb, "%s<Components>%u</Components>\n", indent, 3528 sc->sc_ndisks); 3529 sbuf_printf(sb, "%s<State>%s</State>\n", indent, 3530 g_raid3_device_state2str(sc->sc_state)); 3531 sx_xunlock(&sc->sc_lock); 3532 g_topology_lock(); 3533 } 3534 } 3535 3536 static void 3537 g_raid3_shutdown_post_sync(void *arg, int howto) 3538 { 3539 struct g_class *mp; 3540 struct g_geom *gp, *gp2; 3541 struct g_raid3_softc *sc; 3542 int error; 3543 3544 mp = arg; 3545 g_topology_lock(); 3546 g_raid3_shutdown = 1; 3547 LIST_FOREACH_SAFE(gp, &mp->geom, geom, gp2) { 3548 if ((sc = gp->softc) == NULL) 3549 continue; 3550 /* Skip synchronization geom. */ 3551 if (gp == sc->sc_sync.ds_geom) 3552 continue; 3553 g_topology_unlock(); 3554 sx_xlock(&sc->sc_lock); 3555 g_raid3_idle(sc, -1); 3556 g_cancel_event(sc); 3557 error = g_raid3_destroy(sc, G_RAID3_DESTROY_DELAYED); 3558 if (error != 0) 3559 sx_xunlock(&sc->sc_lock); 3560 g_topology_lock(); 3561 } 3562 g_topology_unlock(); 3563 } 3564 3565 static void 3566 g_raid3_init(struct g_class *mp) 3567 { 3568 3569 g_raid3_post_sync = EVENTHANDLER_REGISTER(shutdown_post_sync, 3570 g_raid3_shutdown_post_sync, mp, SHUTDOWN_PRI_FIRST); 3571 if (g_raid3_post_sync == NULL) 3572 G_RAID3_DEBUG(0, "Warning! Cannot register shutdown event."); 3573 } 3574 3575 static void 3576 g_raid3_fini(struct g_class *mp) 3577 { 3578 3579 if (g_raid3_post_sync != NULL) 3580 EVENTHANDLER_DEREGISTER(shutdown_post_sync, g_raid3_post_sync); 3581 } 3582 3583 DECLARE_GEOM_CLASS(g_raid3_class, g_raid3); 3584