1 /*- 2 * Copyright (c) 2004-2005 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/sysctl.h> 39 #include <sys/malloc.h> 40 #include <sys/eventhandler.h> 41 #include <vm/uma.h> 42 #include <geom/geom.h> 43 #include <sys/proc.h> 44 #include <sys/kthread.h> 45 #include <sys/sched.h> 46 #include <geom/raid3/g_raid3.h> 47 48 49 static MALLOC_DEFINE(M_RAID3, "raid3 data", "GEOM_RAID3 Data"); 50 51 SYSCTL_DECL(_kern_geom); 52 SYSCTL_NODE(_kern_geom, OID_AUTO, raid3, CTLFLAG_RW, 0, "GEOM_RAID3 stuff"); 53 u_int g_raid3_debug = 0; 54 TUNABLE_INT("kern.geom.raid3.debug", &g_raid3_debug); 55 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, debug, CTLFLAG_RW, &g_raid3_debug, 0, 56 "Debug level"); 57 static u_int g_raid3_timeout = 4; 58 TUNABLE_INT("kern.geom.raid3.timeout", &g_raid3_timeout); 59 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, timeout, CTLFLAG_RW, &g_raid3_timeout, 60 0, "Time to wait on all raid3 components"); 61 static u_int g_raid3_idletime = 5; 62 TUNABLE_INT("kern.geom.raid3.idletime", &g_raid3_idletime); 63 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, idletime, CTLFLAG_RW, 64 &g_raid3_idletime, 0, "Mark components as clean when idling"); 65 static u_int g_raid3_reqs_per_sync = 5; 66 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, reqs_per_sync, CTLFLAG_RW, 67 &g_raid3_reqs_per_sync, 0, 68 "Number of regular I/O requests per synchronization request"); 69 static u_int g_raid3_syncs_per_sec = 1000; 70 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, syncs_per_sec, CTLFLAG_RW, 71 &g_raid3_syncs_per_sec, 0, 72 "Number of synchronizations requests per second"); 73 74 static u_int g_raid3_n64k = 50; 75 TUNABLE_INT("kern.geom.raid3.n64k", &g_raid3_n64k); 76 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, n64k, CTLFLAG_RD, &g_raid3_n64k, 0, 77 "Maximum number of 64kB allocations"); 78 static u_int g_raid3_n16k = 200; 79 TUNABLE_INT("kern.geom.raid3.n16k", &g_raid3_n16k); 80 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, n16k, CTLFLAG_RD, &g_raid3_n16k, 0, 81 "Maximum number of 16kB allocations"); 82 static u_int g_raid3_n4k = 1200; 83 TUNABLE_INT("kern.geom.raid3.n4k", &g_raid3_n4k); 84 SYSCTL_UINT(_kern_geom_raid3, OID_AUTO, n4k, CTLFLAG_RD, &g_raid3_n4k, 0, 85 "Maximum number of 4kB allocations"); 86 87 SYSCTL_NODE(_kern_geom_raid3, OID_AUTO, stat, CTLFLAG_RW, 0, 88 "GEOM_RAID3 statistics"); 89 static u_int g_raid3_parity_mismatch = 0; 90 SYSCTL_UINT(_kern_geom_raid3_stat, OID_AUTO, parity_mismatch, CTLFLAG_RD, 91 &g_raid3_parity_mismatch, 0, "Number of failures in VERIFY mode"); 92 static u_int g_raid3_64k_requested = 0; 93 SYSCTL_UINT(_kern_geom_raid3_stat, OID_AUTO, 64k_requested, CTLFLAG_RD, 94 &g_raid3_64k_requested, 0, "Number of requested 64kB allocations"); 95 static u_int g_raid3_64k_failed = 0; 96 SYSCTL_UINT(_kern_geom_raid3_stat, OID_AUTO, 64k_failed, CTLFLAG_RD, 97 &g_raid3_64k_failed, 0, "Number of failed 64kB allocations"); 98 static u_int g_raid3_16k_requested = 0; 99 SYSCTL_UINT(_kern_geom_raid3_stat, OID_AUTO, 16k_requested, CTLFLAG_RD, 100 &g_raid3_16k_requested, 0, "Number of requested 16kB allocations"); 101 static u_int g_raid3_16k_failed = 0; 102 SYSCTL_UINT(_kern_geom_raid3_stat, OID_AUTO, 16k_failed, CTLFLAG_RD, 103 &g_raid3_16k_failed, 0, "Number of failed 16kB allocations"); 104 static u_int g_raid3_4k_requested = 0; 105 SYSCTL_UINT(_kern_geom_raid3_stat, OID_AUTO, 4k_requested, CTLFLAG_RD, 106 &g_raid3_4k_requested, 0, "Number of requested 4kB allocations"); 107 static u_int g_raid3_4k_failed = 0; 108 SYSCTL_UINT(_kern_geom_raid3_stat, OID_AUTO, 4k_failed, CTLFLAG_RD, 109 &g_raid3_4k_failed, 0, "Number of failed 4kB allocations"); 110 111 #define MSLEEP(ident, mtx, priority, wmesg, timeout) do { \ 112 G_RAID3_DEBUG(4, "%s: Sleeping %p.", __func__, (ident)); \ 113 msleep((ident), (mtx), (priority), (wmesg), (timeout)); \ 114 G_RAID3_DEBUG(4, "%s: Woken up %p.", __func__, (ident)); \ 115 } while (0) 116 117 static eventhandler_tag g_raid3_ehtag = NULL; 118 119 static int g_raid3_destroy_geom(struct gctl_req *req, struct g_class *mp, 120 struct g_geom *gp); 121 static g_taste_t g_raid3_taste; 122 static void g_raid3_init(struct g_class *mp); 123 static void g_raid3_fini(struct g_class *mp); 124 125 struct g_class g_raid3_class = { 126 .name = G_RAID3_CLASS_NAME, 127 .version = G_VERSION, 128 .ctlreq = g_raid3_config, 129 .taste = g_raid3_taste, 130 .destroy_geom = g_raid3_destroy_geom, 131 .init = g_raid3_init, 132 .fini = g_raid3_fini 133 }; 134 135 136 static void g_raid3_destroy_provider(struct g_raid3_softc *sc); 137 static int g_raid3_update_disk(struct g_raid3_disk *disk, u_int state); 138 static void g_raid3_update_device(struct g_raid3_softc *sc, boolean_t force); 139 static void g_raid3_dumpconf(struct sbuf *sb, const char *indent, 140 struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp); 141 static void g_raid3_sync_stop(struct g_raid3_softc *sc, int type); 142 143 144 static const char * 145 g_raid3_disk_state2str(int state) 146 { 147 148 switch (state) { 149 case G_RAID3_DISK_STATE_NODISK: 150 return ("NODISK"); 151 case G_RAID3_DISK_STATE_NONE: 152 return ("NONE"); 153 case G_RAID3_DISK_STATE_NEW: 154 return ("NEW"); 155 case G_RAID3_DISK_STATE_ACTIVE: 156 return ("ACTIVE"); 157 case G_RAID3_DISK_STATE_STALE: 158 return ("STALE"); 159 case G_RAID3_DISK_STATE_SYNCHRONIZING: 160 return ("SYNCHRONIZING"); 161 case G_RAID3_DISK_STATE_DISCONNECTED: 162 return ("DISCONNECTED"); 163 default: 164 return ("INVALID"); 165 } 166 } 167 168 static const char * 169 g_raid3_device_state2str(int state) 170 { 171 172 switch (state) { 173 case G_RAID3_DEVICE_STATE_STARTING: 174 return ("STARTING"); 175 case G_RAID3_DEVICE_STATE_DEGRADED: 176 return ("DEGRADED"); 177 case G_RAID3_DEVICE_STATE_COMPLETE: 178 return ("COMPLETE"); 179 default: 180 return ("INVALID"); 181 } 182 } 183 184 const char * 185 g_raid3_get_diskname(struct g_raid3_disk *disk) 186 { 187 188 if (disk->d_consumer == NULL || disk->d_consumer->provider == NULL) 189 return ("[unknown]"); 190 return (disk->d_name); 191 } 192 193 #define g_raid3_xor(src1, src2, dst, size) \ 194 _g_raid3_xor((uint64_t *)(src1), (uint64_t *)(src2), \ 195 (uint64_t *)(dst), (size_t)size) 196 static void 197 _g_raid3_xor(uint64_t *src1, uint64_t *src2, uint64_t *dst, size_t size) 198 { 199 200 KASSERT((size % 128) == 0, ("Invalid size: %zu.", size)); 201 for (; size > 0; size -= 128) { 202 *dst++ = (*src1++) ^ (*src2++); 203 *dst++ = (*src1++) ^ (*src2++); 204 *dst++ = (*src1++) ^ (*src2++); 205 *dst++ = (*src1++) ^ (*src2++); 206 *dst++ = (*src1++) ^ (*src2++); 207 *dst++ = (*src1++) ^ (*src2++); 208 *dst++ = (*src1++) ^ (*src2++); 209 *dst++ = (*src1++) ^ (*src2++); 210 *dst++ = (*src1++) ^ (*src2++); 211 *dst++ = (*src1++) ^ (*src2++); 212 *dst++ = (*src1++) ^ (*src2++); 213 *dst++ = (*src1++) ^ (*src2++); 214 *dst++ = (*src1++) ^ (*src2++); 215 *dst++ = (*src1++) ^ (*src2++); 216 *dst++ = (*src1++) ^ (*src2++); 217 *dst++ = (*src1++) ^ (*src2++); 218 } 219 } 220 221 static int 222 g_raid3_is_zero(struct bio *bp) 223 { 224 static const uint64_t zeros[] = { 225 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 226 }; 227 u_char *addr; 228 ssize_t size; 229 230 size = bp->bio_length; 231 addr = (u_char *)bp->bio_data; 232 for (; size > 0; size -= sizeof(zeros), addr += sizeof(zeros)) { 233 if (bcmp(addr, zeros, sizeof(zeros)) != 0) 234 return (0); 235 } 236 return (1); 237 } 238 239 /* 240 * --- Events handling functions --- 241 * Events in geom_raid3 are used to maintain disks and device status 242 * from one thread to simplify locking. 243 */ 244 static void 245 g_raid3_event_free(struct g_raid3_event *ep) 246 { 247 248 free(ep, M_RAID3); 249 } 250 251 int 252 g_raid3_event_send(void *arg, int state, int flags) 253 { 254 struct g_raid3_softc *sc; 255 struct g_raid3_disk *disk; 256 struct g_raid3_event *ep; 257 int error; 258 259 ep = malloc(sizeof(*ep), M_RAID3, M_WAITOK); 260 G_RAID3_DEBUG(4, "%s: Sending event %p.", __func__, ep); 261 if ((flags & G_RAID3_EVENT_DEVICE) != 0) { 262 disk = NULL; 263 sc = arg; 264 } else { 265 disk = arg; 266 sc = disk->d_softc; 267 } 268 ep->e_disk = disk; 269 ep->e_state = state; 270 ep->e_flags = flags; 271 ep->e_error = 0; 272 mtx_lock(&sc->sc_events_mtx); 273 TAILQ_INSERT_TAIL(&sc->sc_events, ep, e_next); 274 mtx_unlock(&sc->sc_events_mtx); 275 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, sc); 276 mtx_lock(&sc->sc_queue_mtx); 277 wakeup(sc); 278 wakeup(&sc->sc_queue); 279 mtx_unlock(&sc->sc_queue_mtx); 280 if ((flags & G_RAID3_EVENT_DONTWAIT) != 0) 281 return (0); 282 g_topology_assert(); 283 G_RAID3_DEBUG(4, "%s: Sleeping %p.", __func__, ep); 284 g_topology_unlock(); 285 while ((ep->e_flags & G_RAID3_EVENT_DONE) == 0) { 286 mtx_lock(&sc->sc_events_mtx); 287 MSLEEP(ep, &sc->sc_events_mtx, PRIBIO | PDROP, "r3:event", 288 hz * 5); 289 } 290 /* Don't even try to use 'sc' here, because it could be already dead. */ 291 g_topology_lock(); 292 error = ep->e_error; 293 g_raid3_event_free(ep); 294 return (error); 295 } 296 297 static struct g_raid3_event * 298 g_raid3_event_get(struct g_raid3_softc *sc) 299 { 300 struct g_raid3_event *ep; 301 302 mtx_lock(&sc->sc_events_mtx); 303 ep = TAILQ_FIRST(&sc->sc_events); 304 mtx_unlock(&sc->sc_events_mtx); 305 return (ep); 306 } 307 308 static void 309 g_raid3_event_remove(struct g_raid3_softc *sc, struct g_raid3_event *ep) 310 { 311 312 mtx_lock(&sc->sc_events_mtx); 313 TAILQ_REMOVE(&sc->sc_events, ep, e_next); 314 mtx_unlock(&sc->sc_events_mtx); 315 } 316 317 static void 318 g_raid3_event_cancel(struct g_raid3_disk *disk) 319 { 320 struct g_raid3_softc *sc; 321 struct g_raid3_event *ep, *tmpep; 322 323 g_topology_assert(); 324 325 sc = disk->d_softc; 326 mtx_lock(&sc->sc_events_mtx); 327 TAILQ_FOREACH_SAFE(ep, &sc->sc_events, e_next, tmpep) { 328 if ((ep->e_flags & G_RAID3_EVENT_DEVICE) != 0) 329 continue; 330 if (ep->e_disk != disk) 331 continue; 332 TAILQ_REMOVE(&sc->sc_events, ep, e_next); 333 if ((ep->e_flags & G_RAID3_EVENT_DONTWAIT) != 0) 334 g_raid3_event_free(ep); 335 else { 336 ep->e_error = ECANCELED; 337 wakeup(ep); 338 } 339 } 340 mtx_unlock(&sc->sc_events_mtx); 341 } 342 343 /* 344 * Return the number of disks in the given state. 345 * If state is equal to -1, count all connected disks. 346 */ 347 u_int 348 g_raid3_ndisks(struct g_raid3_softc *sc, int state) 349 { 350 struct g_raid3_disk *disk; 351 u_int n, ndisks; 352 353 for (n = ndisks = 0; n < sc->sc_ndisks; n++) { 354 disk = &sc->sc_disks[n]; 355 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 356 continue; 357 if (state == -1 || disk->d_state == state) 358 ndisks++; 359 } 360 return (ndisks); 361 } 362 363 static u_int 364 g_raid3_nrequests(struct g_raid3_softc *sc, struct g_consumer *cp) 365 { 366 struct bio *bp; 367 u_int nreqs = 0; 368 369 mtx_lock(&sc->sc_queue_mtx); 370 TAILQ_FOREACH(bp, &sc->sc_queue.queue, bio_queue) { 371 if (bp->bio_from == cp) 372 nreqs++; 373 } 374 mtx_unlock(&sc->sc_queue_mtx); 375 return (nreqs); 376 } 377 378 static int 379 g_raid3_is_busy(struct g_raid3_softc *sc, struct g_consumer *cp) 380 { 381 382 if (cp->index > 0) { 383 G_RAID3_DEBUG(2, 384 "I/O requests for %s exist, can't destroy it now.", 385 cp->provider->name); 386 return (1); 387 } 388 if (g_raid3_nrequests(sc, cp) > 0) { 389 G_RAID3_DEBUG(2, 390 "I/O requests for %s in queue, can't destroy it now.", 391 cp->provider->name); 392 return (1); 393 } 394 return (0); 395 } 396 397 static void 398 g_raid3_destroy_consumer(void *arg, int flags __unused) 399 { 400 struct g_consumer *cp; 401 402 cp = arg; 403 G_RAID3_DEBUG(1, "Consumer %s destroyed.", cp->provider->name); 404 g_detach(cp); 405 g_destroy_consumer(cp); 406 } 407 408 static void 409 g_raid3_kill_consumer(struct g_raid3_softc *sc, struct g_consumer *cp) 410 { 411 struct g_provider *pp; 412 int retaste_wait; 413 414 g_topology_assert(); 415 416 cp->private = NULL; 417 if (g_raid3_is_busy(sc, cp)) 418 return; 419 G_RAID3_DEBUG(2, "Consumer %s destroyed.", cp->provider->name); 420 pp = cp->provider; 421 retaste_wait = 0; 422 if (cp->acw == 1) { 423 if ((pp->geom->flags & G_GEOM_WITHER) == 0) 424 retaste_wait = 1; 425 } 426 G_RAID3_DEBUG(2, "Access %s r%dw%de%d = %d", pp->name, -cp->acr, 427 -cp->acw, -cp->ace, 0); 428 if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0) 429 g_access(cp, -cp->acr, -cp->acw, -cp->ace); 430 if (retaste_wait) { 431 /* 432 * After retaste event was send (inside g_access()), we can send 433 * event to detach and destroy consumer. 434 * A class, which has consumer to the given provider connected 435 * will not receive retaste event for the provider. 436 * This is the way how I ignore retaste events when I close 437 * consumers opened for write: I detach and destroy consumer 438 * after retaste event is sent. 439 */ 440 g_post_event(g_raid3_destroy_consumer, cp, M_WAITOK, NULL); 441 return; 442 } 443 G_RAID3_DEBUG(1, "Consumer %s destroyed.", pp->name); 444 g_detach(cp); 445 g_destroy_consumer(cp); 446 } 447 448 static int 449 g_raid3_connect_disk(struct g_raid3_disk *disk, struct g_provider *pp) 450 { 451 struct g_consumer *cp; 452 int error; 453 454 g_topology_assert(); 455 KASSERT(disk->d_consumer == NULL, 456 ("Disk already connected (device %s).", disk->d_softc->sc_name)); 457 458 cp = g_new_consumer(disk->d_softc->sc_geom); 459 error = g_attach(cp, pp); 460 if (error != 0) { 461 g_destroy_consumer(cp); 462 return (error); 463 } 464 error = g_access(cp, 1, 1, 1); 465 if (error != 0) { 466 g_detach(cp); 467 g_destroy_consumer(cp); 468 G_RAID3_DEBUG(0, "Cannot open consumer %s (error=%d).", 469 pp->name, error); 470 return (error); 471 } 472 disk->d_consumer = cp; 473 disk->d_consumer->private = disk; 474 disk->d_consumer->index = 0; 475 G_RAID3_DEBUG(2, "Disk %s connected.", g_raid3_get_diskname(disk)); 476 return (0); 477 } 478 479 static void 480 g_raid3_disconnect_consumer(struct g_raid3_softc *sc, struct g_consumer *cp) 481 { 482 483 g_topology_assert(); 484 485 if (cp == NULL) 486 return; 487 if (cp->provider != NULL) 488 g_raid3_kill_consumer(sc, cp); 489 else 490 g_destroy_consumer(cp); 491 } 492 493 /* 494 * Initialize disk. This means allocate memory, create consumer, attach it 495 * to the provider and open access (r1w1e1) to it. 496 */ 497 static struct g_raid3_disk * 498 g_raid3_init_disk(struct g_raid3_softc *sc, struct g_provider *pp, 499 struct g_raid3_metadata *md, int *errorp) 500 { 501 struct g_raid3_disk *disk; 502 int error; 503 504 disk = &sc->sc_disks[md->md_no]; 505 error = g_raid3_connect_disk(disk, pp); 506 if (error != 0) { 507 if (errorp != NULL) 508 *errorp = error; 509 return (NULL); 510 } 511 disk->d_state = G_RAID3_DISK_STATE_NONE; 512 disk->d_flags = md->md_dflags; 513 if (md->md_provider[0] != '\0') 514 disk->d_flags |= G_RAID3_DISK_FLAG_HARDCODED; 515 disk->d_sync.ds_consumer = NULL; 516 disk->d_sync.ds_offset = md->md_sync_offset; 517 disk->d_sync.ds_offset_done = md->md_sync_offset; 518 disk->d_sync.ds_resync = -1; 519 disk->d_genid = md->md_genid; 520 disk->d_sync.ds_syncid = md->md_syncid; 521 if (errorp != NULL) 522 *errorp = 0; 523 return (disk); 524 } 525 526 static void 527 g_raid3_destroy_disk(struct g_raid3_disk *disk) 528 { 529 struct g_raid3_softc *sc; 530 531 g_topology_assert(); 532 533 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 534 return; 535 g_raid3_event_cancel(disk); 536 sc = disk->d_softc; 537 switch (disk->d_state) { 538 case G_RAID3_DISK_STATE_SYNCHRONIZING: 539 if (sc->sc_syncdisk != NULL) 540 g_raid3_sync_stop(sc, 1); 541 /* FALLTHROUGH */ 542 case G_RAID3_DISK_STATE_NEW: 543 case G_RAID3_DISK_STATE_STALE: 544 case G_RAID3_DISK_STATE_ACTIVE: 545 g_raid3_disconnect_consumer(sc, disk->d_consumer); 546 disk->d_consumer = NULL; 547 break; 548 default: 549 KASSERT(0 == 1, ("Wrong disk state (%s, %s).", 550 g_raid3_get_diskname(disk), 551 g_raid3_disk_state2str(disk->d_state))); 552 } 553 disk->d_state = G_RAID3_DISK_STATE_NODISK; 554 } 555 556 static void 557 g_raid3_destroy_device(struct g_raid3_softc *sc) 558 { 559 struct g_raid3_event *ep; 560 struct g_raid3_disk *disk; 561 struct g_geom *gp; 562 struct g_consumer *cp; 563 u_int n; 564 565 g_topology_assert(); 566 567 gp = sc->sc_geom; 568 if (sc->sc_provider != NULL) 569 g_raid3_destroy_provider(sc); 570 for (n = 0; n < sc->sc_ndisks; n++) { 571 disk = &sc->sc_disks[n]; 572 if (disk->d_state != G_RAID3_DISK_STATE_NODISK) { 573 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 574 g_raid3_update_metadata(disk); 575 g_raid3_destroy_disk(disk); 576 } 577 } 578 while ((ep = g_raid3_event_get(sc)) != NULL) { 579 g_raid3_event_remove(sc, ep); 580 if ((ep->e_flags & G_RAID3_EVENT_DONTWAIT) != 0) 581 g_raid3_event_free(ep); 582 else { 583 ep->e_error = ECANCELED; 584 ep->e_flags |= G_RAID3_EVENT_DONE; 585 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, ep); 586 mtx_lock(&sc->sc_events_mtx); 587 wakeup(ep); 588 mtx_unlock(&sc->sc_events_mtx); 589 } 590 } 591 callout_drain(&sc->sc_callout); 592 gp->softc = NULL; 593 cp = LIST_FIRST(&sc->sc_sync.ds_geom->consumer); 594 if (cp != NULL) 595 g_raid3_disconnect_consumer(sc, cp); 596 sc->sc_sync.ds_geom->softc = NULL; 597 g_wither_geom(sc->sc_sync.ds_geom, ENXIO); 598 uma_zdestroy(sc->sc_zone_64k); 599 uma_zdestroy(sc->sc_zone_16k); 600 uma_zdestroy(sc->sc_zone_4k); 601 mtx_destroy(&sc->sc_queue_mtx); 602 mtx_destroy(&sc->sc_events_mtx); 603 G_RAID3_DEBUG(0, "Device %s destroyed.", gp->name); 604 g_wither_geom(gp, ENXIO); 605 } 606 607 static void 608 g_raid3_orphan(struct g_consumer *cp) 609 { 610 struct g_raid3_disk *disk; 611 612 g_topology_assert(); 613 614 disk = cp->private; 615 if (disk == NULL) 616 return; 617 disk->d_softc->sc_bump_id = G_RAID3_BUMP_SYNCID; 618 g_raid3_event_send(disk, G_RAID3_DISK_STATE_DISCONNECTED, 619 G_RAID3_EVENT_DONTWAIT); 620 } 621 622 static int 623 g_raid3_write_metadata(struct g_raid3_disk *disk, struct g_raid3_metadata *md) 624 { 625 struct g_raid3_softc *sc; 626 struct g_consumer *cp; 627 off_t offset, length; 628 u_char *sector; 629 int error = 0; 630 631 g_topology_assert(); 632 633 sc = disk->d_softc; 634 cp = disk->d_consumer; 635 KASSERT(cp != NULL, ("NULL consumer (%s).", sc->sc_name)); 636 KASSERT(cp->provider != NULL, ("NULL provider (%s).", sc->sc_name)); 637 KASSERT(cp->acr == 1 && cp->acw == 1 && cp->ace == 1, 638 ("Consumer %s closed? (r%dw%de%d).", cp->provider->name, cp->acr, 639 cp->acw, cp->ace)); 640 length = cp->provider->sectorsize; 641 offset = cp->provider->mediasize - length; 642 sector = malloc((size_t)length, M_RAID3, M_WAITOK | M_ZERO); 643 if (md != NULL) 644 raid3_metadata_encode(md, sector); 645 g_topology_unlock(); 646 error = g_write_data(cp, offset, sector, length); 647 g_topology_lock(); 648 free(sector, M_RAID3); 649 if (error != 0) { 650 disk->d_softc->sc_bump_id = G_RAID3_BUMP_GENID; 651 g_raid3_event_send(disk, G_RAID3_DISK_STATE_DISCONNECTED, 652 G_RAID3_EVENT_DONTWAIT); 653 } 654 return (error); 655 } 656 657 int 658 g_raid3_clear_metadata(struct g_raid3_disk *disk) 659 { 660 int error; 661 662 g_topology_assert(); 663 error = g_raid3_write_metadata(disk, NULL); 664 if (error == 0) { 665 G_RAID3_DEBUG(2, "Metadata on %s cleared.", 666 g_raid3_get_diskname(disk)); 667 } else { 668 G_RAID3_DEBUG(0, 669 "Cannot clear metadata on disk %s (error=%d).", 670 g_raid3_get_diskname(disk), error); 671 } 672 return (error); 673 } 674 675 void 676 g_raid3_fill_metadata(struct g_raid3_disk *disk, struct g_raid3_metadata *md) 677 { 678 struct g_raid3_softc *sc; 679 struct g_provider *pp; 680 681 sc = disk->d_softc; 682 strlcpy(md->md_magic, G_RAID3_MAGIC, sizeof(md->md_magic)); 683 md->md_version = G_RAID3_VERSION; 684 strlcpy(md->md_name, sc->sc_name, sizeof(md->md_name)); 685 md->md_id = sc->sc_id; 686 md->md_all = sc->sc_ndisks; 687 md->md_genid = sc->sc_genid; 688 md->md_mediasize = sc->sc_mediasize; 689 md->md_sectorsize = sc->sc_sectorsize; 690 md->md_mflags = (sc->sc_flags & G_RAID3_DEVICE_FLAG_MASK); 691 md->md_no = disk->d_no; 692 md->md_syncid = disk->d_sync.ds_syncid; 693 md->md_dflags = (disk->d_flags & G_RAID3_DISK_FLAG_MASK); 694 if (disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) 695 md->md_sync_offset = disk->d_sync.ds_offset_done; 696 else 697 md->md_sync_offset = 0; 698 if (disk->d_consumer != NULL && disk->d_consumer->provider != NULL) 699 pp = disk->d_consumer->provider; 700 else 701 pp = NULL; 702 if ((disk->d_flags & G_RAID3_DISK_FLAG_HARDCODED) != 0 && pp != NULL) 703 strlcpy(md->md_provider, pp->name, sizeof(md->md_provider)); 704 else 705 bzero(md->md_provider, sizeof(md->md_provider)); 706 if (pp != NULL) 707 md->md_provsize = pp->mediasize; 708 else 709 md->md_provsize = 0; 710 } 711 712 void 713 g_raid3_update_metadata(struct g_raid3_disk *disk) 714 { 715 struct g_raid3_metadata md; 716 int error; 717 718 g_topology_assert(); 719 g_raid3_fill_metadata(disk, &md); 720 error = g_raid3_write_metadata(disk, &md); 721 if (error == 0) { 722 G_RAID3_DEBUG(2, "Metadata on %s updated.", 723 g_raid3_get_diskname(disk)); 724 } else { 725 G_RAID3_DEBUG(0, 726 "Cannot update metadata on disk %s (error=%d).", 727 g_raid3_get_diskname(disk), error); 728 } 729 } 730 731 static void 732 g_raid3_bump_syncid(struct g_raid3_softc *sc) 733 { 734 struct g_raid3_disk *disk; 735 u_int n; 736 737 g_topology_assert(); 738 KASSERT(g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) > 0, 739 ("%s called with no active disks (device=%s).", __func__, 740 sc->sc_name)); 741 742 sc->sc_syncid++; 743 G_RAID3_DEBUG(1, "Device %s: syncid bumped to %u.", sc->sc_name, 744 sc->sc_syncid); 745 for (n = 0; n < sc->sc_ndisks; n++) { 746 disk = &sc->sc_disks[n]; 747 if (disk->d_state == G_RAID3_DISK_STATE_ACTIVE || 748 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 749 disk->d_sync.ds_syncid = sc->sc_syncid; 750 g_raid3_update_metadata(disk); 751 } 752 } 753 } 754 755 static void 756 g_raid3_bump_genid(struct g_raid3_softc *sc) 757 { 758 struct g_raid3_disk *disk; 759 u_int n; 760 761 g_topology_assert(); 762 KASSERT(g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) > 0, 763 ("%s called with no active disks (device=%s).", __func__, 764 sc->sc_name)); 765 766 sc->sc_genid++; 767 G_RAID3_DEBUG(1, "Device %s: genid bumped to %u.", sc->sc_name, 768 sc->sc_genid); 769 for (n = 0; n < sc->sc_ndisks; n++) { 770 disk = &sc->sc_disks[n]; 771 if (disk->d_state == G_RAID3_DISK_STATE_ACTIVE || 772 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 773 disk->d_genid = sc->sc_genid; 774 g_raid3_update_metadata(disk); 775 } 776 } 777 } 778 779 static void 780 g_raid3_idle(struct g_raid3_softc *sc) 781 { 782 struct g_raid3_disk *disk; 783 u_int i; 784 785 if (sc->sc_provider == NULL || sc->sc_provider->acw == 0) 786 return; 787 sc->sc_idle = 1; 788 g_topology_lock(); 789 for (i = 0; i < sc->sc_ndisks; i++) { 790 disk = &sc->sc_disks[i]; 791 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) 792 continue; 793 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as clean.", 794 g_raid3_get_diskname(disk), sc->sc_name); 795 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 796 g_raid3_update_metadata(disk); 797 } 798 g_topology_unlock(); 799 } 800 801 static void 802 g_raid3_unidle(struct g_raid3_softc *sc) 803 { 804 struct g_raid3_disk *disk; 805 u_int i; 806 807 sc->sc_idle = 0; 808 g_topology_lock(); 809 for (i = 0; i < sc->sc_ndisks; i++) { 810 disk = &sc->sc_disks[i]; 811 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) 812 continue; 813 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as dirty.", 814 g_raid3_get_diskname(disk), sc->sc_name); 815 disk->d_flags |= G_RAID3_DISK_FLAG_DIRTY; 816 g_raid3_update_metadata(disk); 817 } 818 g_topology_unlock(); 819 } 820 821 /* 822 * Return 1 if we should check if RAID3 device is idling. 823 */ 824 static int 825 g_raid3_check_idle(struct g_raid3_softc *sc) 826 { 827 struct g_raid3_disk *disk; 828 u_int i; 829 830 if (sc->sc_idle) 831 return (0); 832 if (sc->sc_provider != NULL && sc->sc_provider->acw == 0) 833 return (0); 834 /* 835 * Check if there are no in-flight requests. 836 */ 837 for (i = 0; i < sc->sc_ndisks; i++) { 838 disk = &sc->sc_disks[i]; 839 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) 840 continue; 841 if (disk->d_consumer->index > 0) 842 return (0); 843 } 844 return (1); 845 } 846 847 /* 848 * Treat bio_driver1 field in parent bio as list head and field bio_caller1 849 * in child bio as pointer to the next element on the list. 850 */ 851 #define G_RAID3_HEAD_BIO(pbp) (pbp)->bio_driver1 852 853 #define G_RAID3_NEXT_BIO(cbp) (cbp)->bio_caller1 854 855 #define G_RAID3_FOREACH_BIO(pbp, bp) \ 856 for ((bp) = G_RAID3_HEAD_BIO(pbp); (bp) != NULL; \ 857 (bp) = G_RAID3_NEXT_BIO(bp)) 858 859 #define G_RAID3_FOREACH_SAFE_BIO(pbp, bp, tmpbp) \ 860 for ((bp) = G_RAID3_HEAD_BIO(pbp); \ 861 (bp) != NULL && ((tmpbp) = G_RAID3_NEXT_BIO(bp), 1); \ 862 (bp) = (tmpbp)) 863 864 static void 865 g_raid3_init_bio(struct bio *pbp) 866 { 867 868 G_RAID3_HEAD_BIO(pbp) = NULL; 869 } 870 871 static void 872 g_raid3_remove_bio(struct bio *cbp) 873 { 874 struct bio *pbp, *bp; 875 876 pbp = cbp->bio_parent; 877 if (G_RAID3_HEAD_BIO(pbp) == cbp) 878 G_RAID3_HEAD_BIO(pbp) = G_RAID3_NEXT_BIO(cbp); 879 else { 880 G_RAID3_FOREACH_BIO(pbp, bp) { 881 if (G_RAID3_NEXT_BIO(bp) == cbp) { 882 G_RAID3_NEXT_BIO(bp) = G_RAID3_NEXT_BIO(cbp); 883 break; 884 } 885 } 886 } 887 G_RAID3_NEXT_BIO(cbp) = NULL; 888 } 889 890 static void 891 g_raid3_replace_bio(struct bio *sbp, struct bio *dbp) 892 { 893 struct bio *pbp, *bp; 894 895 g_raid3_remove_bio(sbp); 896 pbp = dbp->bio_parent; 897 G_RAID3_NEXT_BIO(sbp) = G_RAID3_NEXT_BIO(dbp); 898 if (G_RAID3_HEAD_BIO(pbp) == dbp) 899 G_RAID3_HEAD_BIO(pbp) = sbp; 900 else { 901 G_RAID3_FOREACH_BIO(pbp, bp) { 902 if (G_RAID3_NEXT_BIO(bp) == dbp) { 903 G_RAID3_NEXT_BIO(bp) = sbp; 904 break; 905 } 906 } 907 } 908 G_RAID3_NEXT_BIO(dbp) = NULL; 909 } 910 911 static void 912 g_raid3_destroy_bio(struct g_raid3_softc *sc, struct bio *cbp) 913 { 914 struct bio *bp, *pbp; 915 size_t size; 916 917 pbp = cbp->bio_parent; 918 pbp->bio_children--; 919 KASSERT(cbp->bio_data != NULL, ("NULL bio_data")); 920 size = pbp->bio_length / (sc->sc_ndisks - 1); 921 if (size > 16384) 922 uma_zfree(sc->sc_zone_64k, cbp->bio_data); 923 else if (size > 4096) 924 uma_zfree(sc->sc_zone_16k, cbp->bio_data); 925 else 926 uma_zfree(sc->sc_zone_4k, cbp->bio_data); 927 if (G_RAID3_HEAD_BIO(pbp) == cbp) { 928 G_RAID3_HEAD_BIO(pbp) = G_RAID3_NEXT_BIO(cbp); 929 G_RAID3_NEXT_BIO(cbp) = NULL; 930 g_destroy_bio(cbp); 931 } else { 932 G_RAID3_FOREACH_BIO(pbp, bp) { 933 if (G_RAID3_NEXT_BIO(bp) == cbp) 934 break; 935 } 936 if (bp != NULL) { 937 KASSERT(G_RAID3_NEXT_BIO(bp) != NULL, 938 ("NULL bp->bio_driver1")); 939 G_RAID3_NEXT_BIO(bp) = G_RAID3_NEXT_BIO(cbp); 940 G_RAID3_NEXT_BIO(cbp) = NULL; 941 } 942 g_destroy_bio(cbp); 943 } 944 } 945 946 static struct bio * 947 g_raid3_clone_bio(struct g_raid3_softc *sc, struct bio *pbp) 948 { 949 struct bio *bp, *cbp; 950 size_t size; 951 952 cbp = g_clone_bio(pbp); 953 if (cbp == NULL) 954 return (NULL); 955 size = pbp->bio_length / (sc->sc_ndisks - 1); 956 if (size > 16384) { 957 cbp->bio_data = uma_zalloc(sc->sc_zone_64k, M_NOWAIT); 958 g_raid3_64k_requested++; 959 } else if (size > 4096) { 960 cbp->bio_data = uma_zalloc(sc->sc_zone_16k, M_NOWAIT); 961 g_raid3_16k_requested++; 962 } else { 963 cbp->bio_data = uma_zalloc(sc->sc_zone_4k, M_NOWAIT); 964 g_raid3_4k_requested++; 965 } 966 if (cbp->bio_data == NULL) { 967 if (size > 16384) 968 g_raid3_64k_failed++; 969 if (size > 4096) 970 g_raid3_16k_failed++; 971 else 972 g_raid3_4k_failed++; 973 pbp->bio_children--; 974 g_destroy_bio(cbp); 975 return (NULL); 976 } 977 G_RAID3_NEXT_BIO(cbp) = NULL; 978 if (G_RAID3_HEAD_BIO(pbp) == NULL) 979 G_RAID3_HEAD_BIO(pbp) = cbp; 980 else { 981 G_RAID3_FOREACH_BIO(pbp, bp) { 982 if (G_RAID3_NEXT_BIO(bp) == NULL) { 983 G_RAID3_NEXT_BIO(bp) = cbp; 984 break; 985 } 986 } 987 } 988 return (cbp); 989 } 990 991 static void 992 g_raid3_scatter(struct bio *pbp) 993 { 994 struct g_raid3_softc *sc; 995 struct g_raid3_disk *disk; 996 struct bio *bp, *cbp; 997 off_t atom, cadd, padd, left; 998 999 sc = pbp->bio_to->geom->softc; 1000 bp = NULL; 1001 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_NOPARITY) == 0) { 1002 /* 1003 * Find bio for which we should calculate data. 1004 */ 1005 G_RAID3_FOREACH_BIO(pbp, cbp) { 1006 if ((cbp->bio_cflags & G_RAID3_BIO_CFLAG_PARITY) != 0) { 1007 bp = cbp; 1008 break; 1009 } 1010 } 1011 KASSERT(bp != NULL, ("NULL parity bio.")); 1012 } 1013 atom = sc->sc_sectorsize / (sc->sc_ndisks - 1); 1014 cadd = padd = 0; 1015 for (left = pbp->bio_length; left > 0; left -= sc->sc_sectorsize) { 1016 G_RAID3_FOREACH_BIO(pbp, cbp) { 1017 if (cbp == bp) 1018 continue; 1019 bcopy(pbp->bio_data + padd, cbp->bio_data + cadd, atom); 1020 padd += atom; 1021 } 1022 cadd += atom; 1023 } 1024 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_NOPARITY) == 0) { 1025 struct bio *tmpbp; 1026 1027 /* 1028 * Calculate parity. 1029 */ 1030 bzero(bp->bio_data, bp->bio_length); 1031 G_RAID3_FOREACH_SAFE_BIO(pbp, cbp, tmpbp) { 1032 if (cbp == bp) 1033 continue; 1034 g_raid3_xor(cbp->bio_data, bp->bio_data, bp->bio_data, 1035 bp->bio_length); 1036 if ((cbp->bio_cflags & G_RAID3_BIO_CFLAG_NODISK) != 0) 1037 g_raid3_destroy_bio(sc, cbp); 1038 } 1039 } 1040 G_RAID3_FOREACH_BIO(pbp, cbp) { 1041 struct g_consumer *cp; 1042 1043 disk = cbp->bio_caller2; 1044 cp = disk->d_consumer; 1045 cbp->bio_to = cp->provider; 1046 G_RAID3_LOGREQ(3, cbp, "Sending request."); 1047 KASSERT(cp->acr == 1 && cp->acw == 1 && cp->ace == 1, 1048 ("Consumer %s not opened (r%dw%de%d).", cp->provider->name, 1049 cp->acr, cp->acw, cp->ace)); 1050 cp->index++; 1051 g_io_request(cbp, cp); 1052 } 1053 } 1054 1055 static void 1056 g_raid3_gather(struct bio *pbp) 1057 { 1058 struct g_raid3_softc *sc; 1059 struct g_raid3_disk *disk; 1060 struct bio *xbp, *fbp, *cbp; 1061 off_t atom, cadd, padd, left; 1062 1063 sc = pbp->bio_to->geom->softc; 1064 /* 1065 * Find bio for which we have to calculate data. 1066 * While going through this path, check if all requests 1067 * succeeded, if not, deny whole request. 1068 * If we're in COMPLETE mode, we allow one request to fail, 1069 * so if we find one, we're sending it to the parity consumer. 1070 * If there are more failed requests, we deny whole request. 1071 */ 1072 xbp = fbp = NULL; 1073 G_RAID3_FOREACH_BIO(pbp, cbp) { 1074 if ((cbp->bio_cflags & G_RAID3_BIO_CFLAG_PARITY) != 0) { 1075 KASSERT(xbp == NULL, ("More than one parity bio.")); 1076 xbp = cbp; 1077 } 1078 if (cbp->bio_error == 0) 1079 continue; 1080 /* 1081 * Found failed request. 1082 */ 1083 G_RAID3_LOGREQ(0, cbp, "Request failed."); 1084 disk = cbp->bio_caller2; 1085 if (disk != NULL) { 1086 /* 1087 * Actually this is pointless to bump genid, 1088 * because whole device is fucked up. 1089 */ 1090 sc->sc_bump_id |= G_RAID3_BUMP_GENID; 1091 g_raid3_event_send(disk, 1092 G_RAID3_DISK_STATE_DISCONNECTED, 1093 G_RAID3_EVENT_DONTWAIT); 1094 } 1095 if (fbp == NULL) { 1096 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_DEGRADED) != 0) { 1097 /* 1098 * We are already in degraded mode, so we can't 1099 * accept any failures. 1100 */ 1101 if (pbp->bio_error == 0) 1102 pbp->bio_error = fbp->bio_error; 1103 } else { 1104 fbp = cbp; 1105 } 1106 } else { 1107 /* 1108 * Next failed request, that's too many. 1109 */ 1110 if (pbp->bio_error == 0) 1111 pbp->bio_error = fbp->bio_error; 1112 } 1113 } 1114 if (pbp->bio_error != 0) 1115 goto finish; 1116 if (fbp != NULL && (pbp->bio_pflags & G_RAID3_BIO_PFLAG_VERIFY) != 0) { 1117 pbp->bio_pflags &= ~G_RAID3_BIO_PFLAG_VERIFY; 1118 if (xbp != fbp) 1119 g_raid3_replace_bio(xbp, fbp); 1120 g_raid3_destroy_bio(sc, fbp); 1121 } else if (fbp != NULL) { 1122 struct g_consumer *cp; 1123 1124 /* 1125 * One request failed, so send the same request to 1126 * the parity consumer. 1127 */ 1128 disk = pbp->bio_driver2; 1129 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) { 1130 pbp->bio_error = fbp->bio_error; 1131 goto finish; 1132 } 1133 pbp->bio_pflags |= G_RAID3_BIO_PFLAG_DEGRADED; 1134 pbp->bio_inbed--; 1135 fbp->bio_flags &= ~(BIO_DONE | BIO_ERROR); 1136 if (disk->d_no == sc->sc_ndisks - 1) 1137 fbp->bio_cflags |= G_RAID3_BIO_CFLAG_PARITY; 1138 fbp->bio_error = 0; 1139 fbp->bio_completed = 0; 1140 fbp->bio_children = 0; 1141 fbp->bio_inbed = 0; 1142 cp = disk->d_consumer; 1143 fbp->bio_caller2 = disk; 1144 fbp->bio_to = cp->provider; 1145 G_RAID3_LOGREQ(3, fbp, "Sending request (recover)."); 1146 KASSERT(cp->acr == 1 && cp->acw == 1 && cp->ace == 1, 1147 ("Consumer %s not opened (r%dw%de%d).", cp->provider->name, 1148 cp->acr, cp->acw, cp->ace)); 1149 cp->index++; 1150 g_io_request(fbp, cp); 1151 return; 1152 } 1153 if (xbp != NULL) { 1154 /* 1155 * Calculate parity. 1156 */ 1157 G_RAID3_FOREACH_BIO(pbp, cbp) { 1158 if ((cbp->bio_cflags & G_RAID3_BIO_CFLAG_PARITY) != 0) 1159 continue; 1160 g_raid3_xor(cbp->bio_data, xbp->bio_data, xbp->bio_data, 1161 xbp->bio_length); 1162 } 1163 xbp->bio_cflags &= ~G_RAID3_BIO_CFLAG_PARITY; 1164 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_VERIFY) != 0) { 1165 if (!g_raid3_is_zero(xbp)) { 1166 g_raid3_parity_mismatch++; 1167 pbp->bio_error = EIO; 1168 goto finish; 1169 } 1170 g_raid3_destroy_bio(sc, xbp); 1171 } 1172 } 1173 atom = sc->sc_sectorsize / (sc->sc_ndisks - 1); 1174 cadd = padd = 0; 1175 for (left = pbp->bio_length; left > 0; left -= sc->sc_sectorsize) { 1176 G_RAID3_FOREACH_BIO(pbp, cbp) { 1177 bcopy(cbp->bio_data + cadd, pbp->bio_data + padd, atom); 1178 pbp->bio_completed += atom; 1179 padd += atom; 1180 } 1181 cadd += atom; 1182 } 1183 finish: 1184 if (pbp->bio_error == 0) 1185 G_RAID3_LOGREQ(3, pbp, "Request finished."); 1186 else { 1187 if ((pbp->bio_pflags & G_RAID3_BIO_PFLAG_VERIFY) != 0) 1188 G_RAID3_LOGREQ(1, pbp, "Verification error."); 1189 else 1190 G_RAID3_LOGREQ(0, pbp, "Request failed."); 1191 } 1192 pbp->bio_pflags &= ~G_RAID3_BIO_PFLAG_MASK; 1193 g_io_deliver(pbp, pbp->bio_error); 1194 while ((cbp = G_RAID3_HEAD_BIO(pbp)) != NULL) 1195 g_raid3_destroy_bio(sc, cbp); 1196 } 1197 1198 static void 1199 g_raid3_done(struct bio *bp) 1200 { 1201 struct g_raid3_softc *sc; 1202 1203 sc = bp->bio_from->geom->softc; 1204 bp->bio_cflags |= G_RAID3_BIO_CFLAG_REGULAR; 1205 G_RAID3_LOGREQ(3, bp, "Regular request done (error=%d).", bp->bio_error); 1206 mtx_lock(&sc->sc_queue_mtx); 1207 bioq_insert_head(&sc->sc_queue, bp); 1208 wakeup(sc); 1209 wakeup(&sc->sc_queue); 1210 mtx_unlock(&sc->sc_queue_mtx); 1211 } 1212 1213 static void 1214 g_raid3_regular_request(struct bio *cbp) 1215 { 1216 struct g_raid3_softc *sc; 1217 struct g_raid3_disk *disk; 1218 struct bio *pbp; 1219 1220 g_topology_assert_not(); 1221 1222 cbp->bio_from->index--; 1223 pbp = cbp->bio_parent; 1224 sc = pbp->bio_to->geom->softc; 1225 disk = cbp->bio_from->private; 1226 if (disk == NULL) { 1227 g_topology_lock(); 1228 g_raid3_kill_consumer(sc, cbp->bio_from); 1229 g_topology_unlock(); 1230 } 1231 1232 G_RAID3_LOGREQ(3, cbp, "Request finished."); 1233 pbp->bio_inbed++; 1234 KASSERT(pbp->bio_inbed <= pbp->bio_children, 1235 ("bio_inbed (%u) is bigger than bio_children (%u).", pbp->bio_inbed, 1236 pbp->bio_children)); 1237 if (pbp->bio_inbed != pbp->bio_children) 1238 return; 1239 switch (pbp->bio_cmd) { 1240 case BIO_READ: 1241 g_raid3_gather(pbp); 1242 break; 1243 case BIO_WRITE: 1244 case BIO_DELETE: 1245 { 1246 int error = 0; 1247 1248 pbp->bio_completed = pbp->bio_length; 1249 while ((cbp = G_RAID3_HEAD_BIO(pbp)) != NULL) { 1250 if (cbp->bio_error != 0) { 1251 disk = cbp->bio_caller2; 1252 if (disk != NULL) { 1253 sc->sc_bump_id |= G_RAID3_BUMP_GENID; 1254 g_raid3_event_send(disk, 1255 G_RAID3_DISK_STATE_DISCONNECTED, 1256 G_RAID3_EVENT_DONTWAIT); 1257 } 1258 if (error == 0) 1259 error = cbp->bio_error; 1260 else if (pbp->bio_error == 0) { 1261 /* 1262 * Next failed request, that's too many. 1263 */ 1264 pbp->bio_error = error; 1265 } 1266 } 1267 g_raid3_destroy_bio(sc, cbp); 1268 } 1269 if (pbp->bio_error == 0) 1270 G_RAID3_LOGREQ(3, pbp, "Request finished."); 1271 else 1272 G_RAID3_LOGREQ(0, pbp, "Request failed."); 1273 pbp->bio_pflags &= ~G_RAID3_BIO_PFLAG_DEGRADED; 1274 pbp->bio_pflags &= ~G_RAID3_BIO_PFLAG_NOPARITY; 1275 g_io_deliver(pbp, pbp->bio_error); 1276 break; 1277 } 1278 } 1279 } 1280 1281 static void 1282 g_raid3_sync_done(struct bio *bp) 1283 { 1284 struct g_raid3_softc *sc; 1285 1286 G_RAID3_LOGREQ(3, bp, "Synchronization request delivered."); 1287 sc = bp->bio_from->geom->softc; 1288 bp->bio_cflags |= G_RAID3_BIO_CFLAG_SYNC; 1289 mtx_lock(&sc->sc_queue_mtx); 1290 bioq_insert_head(&sc->sc_queue, bp); 1291 wakeup(sc); 1292 wakeup(&sc->sc_queue); 1293 mtx_unlock(&sc->sc_queue_mtx); 1294 } 1295 1296 static void 1297 g_raid3_start(struct bio *bp) 1298 { 1299 struct g_raid3_softc *sc; 1300 1301 sc = bp->bio_to->geom->softc; 1302 /* 1303 * If sc == NULL or there are no valid disks, provider's error 1304 * should be set and g_raid3_start() should not be called at all. 1305 */ 1306 KASSERT(sc != NULL && (sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 1307 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE), 1308 ("Provider's error should be set (error=%d)(device=%s).", 1309 bp->bio_to->error, bp->bio_to->name)); 1310 G_RAID3_LOGREQ(3, bp, "Request received."); 1311 1312 switch (bp->bio_cmd) { 1313 case BIO_READ: 1314 case BIO_WRITE: 1315 case BIO_DELETE: 1316 break; 1317 case BIO_GETATTR: 1318 default: 1319 g_io_deliver(bp, EOPNOTSUPP); 1320 return; 1321 } 1322 mtx_lock(&sc->sc_queue_mtx); 1323 bioq_insert_tail(&sc->sc_queue, bp); 1324 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, sc); 1325 wakeup(sc); 1326 mtx_unlock(&sc->sc_queue_mtx); 1327 } 1328 1329 /* 1330 * Send one synchronization request. 1331 */ 1332 static void 1333 g_raid3_sync_one(struct g_raid3_softc *sc) 1334 { 1335 struct g_raid3_disk *disk; 1336 struct bio *bp; 1337 1338 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED, 1339 ("Wrong device state (%s, %s).", sc->sc_name, 1340 g_raid3_device_state2str(sc->sc_state))); 1341 disk = sc->sc_syncdisk; 1342 KASSERT(disk != NULL, ("No sync disk (%s).", sc->sc_name)); 1343 KASSERT(disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING, 1344 ("Disk %s is not marked for synchronization.", 1345 g_raid3_get_diskname(disk))); 1346 1347 bp = g_new_bio(); 1348 if (bp == NULL) 1349 return; 1350 bp->bio_parent = NULL; 1351 bp->bio_cmd = BIO_READ; 1352 bp->bio_offset = disk->d_sync.ds_offset * (sc->sc_ndisks - 1); 1353 bp->bio_length = MIN(MAXPHYS, sc->sc_mediasize - bp->bio_offset); 1354 bp->bio_cflags = 0; 1355 bp->bio_done = g_raid3_sync_done; 1356 bp->bio_data = disk->d_sync.ds_data; 1357 if (bp->bio_data == NULL) { 1358 g_destroy_bio(bp); 1359 return; 1360 } 1361 bp->bio_cflags = G_RAID3_BIO_CFLAG_REGSYNC; 1362 disk->d_sync.ds_offset += bp->bio_length / (sc->sc_ndisks - 1); 1363 bp->bio_to = sc->sc_provider; 1364 G_RAID3_LOGREQ(3, bp, "Sending synchronization request."); 1365 disk->d_sync.ds_consumer->index++; 1366 g_io_request(bp, disk->d_sync.ds_consumer); 1367 } 1368 1369 static void 1370 g_raid3_sync_request(struct bio *bp) 1371 { 1372 struct g_raid3_softc *sc; 1373 struct g_raid3_disk *disk; 1374 1375 bp->bio_from->index--; 1376 sc = bp->bio_from->geom->softc; 1377 disk = bp->bio_from->private; 1378 if (disk == NULL) { 1379 g_topology_lock(); 1380 g_raid3_kill_consumer(sc, bp->bio_from); 1381 g_topology_unlock(); 1382 g_destroy_bio(bp); 1383 return; 1384 } 1385 1386 /* 1387 * Synchronization request. 1388 */ 1389 switch (bp->bio_cmd) { 1390 case BIO_READ: 1391 { 1392 struct g_consumer *cp; 1393 u_char *dst, *src; 1394 off_t left; 1395 u_int atom; 1396 1397 if (bp->bio_error != 0) { 1398 G_RAID3_LOGREQ(0, bp, 1399 "Synchronization request failed (error=%d).", 1400 bp->bio_error); 1401 g_destroy_bio(bp); 1402 return; 1403 } 1404 G_RAID3_LOGREQ(3, bp, "Synchronization request finished."); 1405 atom = sc->sc_sectorsize / (sc->sc_ndisks - 1); 1406 dst = src = bp->bio_data; 1407 if (disk->d_no == sc->sc_ndisks - 1) { 1408 u_int n; 1409 1410 /* Parity component. */ 1411 for (left = bp->bio_length; left > 0; 1412 left -= sc->sc_sectorsize) { 1413 bcopy(src, dst, atom); 1414 src += atom; 1415 for (n = 1; n < sc->sc_ndisks - 1; n++) { 1416 g_raid3_xor(src, dst, dst, atom); 1417 src += atom; 1418 } 1419 dst += atom; 1420 } 1421 } else { 1422 /* Regular component. */ 1423 src += atom * disk->d_no; 1424 for (left = bp->bio_length; left > 0; 1425 left -= sc->sc_sectorsize) { 1426 bcopy(src, dst, atom); 1427 src += sc->sc_sectorsize; 1428 dst += atom; 1429 } 1430 } 1431 bp->bio_offset /= sc->sc_ndisks - 1; 1432 bp->bio_length /= sc->sc_ndisks - 1; 1433 bp->bio_cmd = BIO_WRITE; 1434 bp->bio_cflags = 0; 1435 bp->bio_children = bp->bio_inbed = 0; 1436 cp = disk->d_consumer; 1437 KASSERT(cp->acr == 1 && cp->acw == 1 && cp->ace == 1, 1438 ("Consumer %s not opened (r%dw%de%d).", cp->provider->name, 1439 cp->acr, cp->acw, cp->ace)); 1440 cp->index++; 1441 g_io_request(bp, cp); 1442 return; 1443 } 1444 case BIO_WRITE: 1445 { 1446 struct g_raid3_disk_sync *sync; 1447 1448 if (bp->bio_error != 0) { 1449 G_RAID3_LOGREQ(0, bp, 1450 "Synchronization request failed (error=%d).", 1451 bp->bio_error); 1452 g_destroy_bio(bp); 1453 sc->sc_bump_id |= G_RAID3_BUMP_GENID; 1454 g_raid3_event_send(disk, 1455 G_RAID3_DISK_STATE_DISCONNECTED, 1456 G_RAID3_EVENT_DONTWAIT); 1457 return; 1458 } 1459 G_RAID3_LOGREQ(3, bp, "Synchronization request finished."); 1460 sync = &disk->d_sync; 1461 sync->ds_offset_done = bp->bio_offset + bp->bio_length; 1462 g_destroy_bio(bp); 1463 if (sync->ds_resync != -1) 1464 return; 1465 if (sync->ds_offset_done == 1466 sc->sc_mediasize / (sc->sc_ndisks - 1)) { 1467 /* 1468 * Disk up-to-date, activate it. 1469 */ 1470 g_raid3_event_send(disk, G_RAID3_DISK_STATE_ACTIVE, 1471 G_RAID3_EVENT_DONTWAIT); 1472 return; 1473 } else if (sync->ds_offset_done % (MAXPHYS * 100) == 0) { 1474 /* 1475 * Update offset_done on every 100 blocks. 1476 * XXX: This should be configurable. 1477 */ 1478 g_topology_lock(); 1479 g_raid3_update_metadata(disk); 1480 g_topology_unlock(); 1481 } 1482 return; 1483 } 1484 default: 1485 KASSERT(1 == 0, ("Invalid command here: %u (device=%s)", 1486 bp->bio_cmd, sc->sc_name)); 1487 break; 1488 } 1489 } 1490 1491 static int 1492 g_raid3_register_request(struct bio *pbp) 1493 { 1494 struct g_raid3_softc *sc; 1495 struct g_raid3_disk *disk; 1496 struct g_consumer *cp; 1497 struct bio *cbp; 1498 off_t offset, length; 1499 u_int n, ndisks; 1500 int round_robin, verify; 1501 1502 ndisks = 0; 1503 sc = pbp->bio_to->geom->softc; 1504 if ((pbp->bio_cflags & G_RAID3_BIO_CFLAG_REGSYNC) != 0 && 1505 sc->sc_syncdisk == NULL) { 1506 g_io_deliver(pbp, EIO); 1507 return (0); 1508 } 1509 g_raid3_init_bio(pbp); 1510 length = pbp->bio_length / (sc->sc_ndisks - 1); 1511 offset = pbp->bio_offset / (sc->sc_ndisks - 1); 1512 round_robin = verify = 0; 1513 switch (pbp->bio_cmd) { 1514 case BIO_READ: 1515 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_VERIFY) != 0 && 1516 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE) { 1517 pbp->bio_pflags |= G_RAID3_BIO_PFLAG_VERIFY; 1518 verify = 1; 1519 ndisks = sc->sc_ndisks; 1520 } else { 1521 verify = 0; 1522 ndisks = sc->sc_ndisks - 1; 1523 } 1524 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_ROUND_ROBIN) != 0 && 1525 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE) { 1526 round_robin = 1; 1527 } else { 1528 round_robin = 0; 1529 } 1530 KASSERT(!round_robin || !verify, 1531 ("ROUND-ROBIN and VERIFY are mutually exclusive.")); 1532 pbp->bio_driver2 = &sc->sc_disks[sc->sc_ndisks - 1]; 1533 break; 1534 case BIO_WRITE: 1535 case BIO_DELETE: 1536 { 1537 struct g_raid3_disk_sync *sync; 1538 1539 if (sc->sc_idle) 1540 g_raid3_unidle(sc); 1541 1542 ndisks = sc->sc_ndisks; 1543 1544 if (sc->sc_syncdisk == NULL) 1545 break; 1546 sync = &sc->sc_syncdisk->d_sync; 1547 if (offset >= sync->ds_offset) 1548 break; 1549 if (offset + length <= sync->ds_offset_done) 1550 break; 1551 if (offset >= sync->ds_resync && sync->ds_resync != -1) 1552 break; 1553 sync->ds_resync = offset - (offset % MAXPHYS); 1554 break; 1555 } 1556 } 1557 for (n = 0; n < ndisks; n++) { 1558 disk = &sc->sc_disks[n]; 1559 cbp = g_raid3_clone_bio(sc, pbp); 1560 if (cbp == NULL) { 1561 while ((cbp = G_RAID3_HEAD_BIO(pbp)) != NULL) 1562 g_raid3_destroy_bio(sc, cbp); 1563 return (ENOMEM); 1564 } 1565 cbp->bio_offset = offset; 1566 cbp->bio_length = length; 1567 cbp->bio_done = g_raid3_done; 1568 switch (pbp->bio_cmd) { 1569 case BIO_READ: 1570 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) { 1571 /* 1572 * Replace invalid component with the parity 1573 * component. 1574 */ 1575 disk = &sc->sc_disks[sc->sc_ndisks - 1]; 1576 cbp->bio_cflags |= G_RAID3_BIO_CFLAG_PARITY; 1577 pbp->bio_pflags |= G_RAID3_BIO_PFLAG_DEGRADED; 1578 } else if (round_robin && 1579 disk->d_no == sc->sc_round_robin) { 1580 /* 1581 * In round-robin mode skip one data component 1582 * and use parity component when reading. 1583 */ 1584 pbp->bio_driver2 = disk; 1585 disk = &sc->sc_disks[sc->sc_ndisks - 1]; 1586 cbp->bio_cflags |= G_RAID3_BIO_CFLAG_PARITY; 1587 sc->sc_round_robin++; 1588 round_robin = 0; 1589 } else if (verify && disk->d_no == sc->sc_ndisks - 1) { 1590 cbp->bio_cflags |= G_RAID3_BIO_CFLAG_PARITY; 1591 } 1592 break; 1593 case BIO_WRITE: 1594 case BIO_DELETE: 1595 if (disk->d_state == G_RAID3_DISK_STATE_ACTIVE || 1596 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 1597 if (n == ndisks - 1) { 1598 /* 1599 * Active parity component, mark it as such. 1600 */ 1601 cbp->bio_cflags |= 1602 G_RAID3_BIO_CFLAG_PARITY; 1603 } 1604 } else { 1605 pbp->bio_pflags |= G_RAID3_BIO_PFLAG_DEGRADED; 1606 if (n == ndisks - 1) { 1607 /* 1608 * Parity component is not connected, 1609 * so destroy its request. 1610 */ 1611 pbp->bio_pflags |= 1612 G_RAID3_BIO_PFLAG_NOPARITY; 1613 g_raid3_destroy_bio(sc, cbp); 1614 cbp = NULL; 1615 } else { 1616 cbp->bio_cflags |= 1617 G_RAID3_BIO_CFLAG_NODISK; 1618 disk = NULL; 1619 } 1620 } 1621 break; 1622 } 1623 if (cbp != NULL) 1624 cbp->bio_caller2 = disk; 1625 } 1626 switch (pbp->bio_cmd) { 1627 case BIO_READ: 1628 if (round_robin) { 1629 /* 1630 * If we are in round-robin mode and 'round_robin' is 1631 * still 1, it means, that we skipped parity component 1632 * for this read and must reset sc_round_robin field. 1633 */ 1634 sc->sc_round_robin = 0; 1635 } 1636 G_RAID3_FOREACH_BIO(pbp, cbp) { 1637 disk = cbp->bio_caller2; 1638 cp = disk->d_consumer; 1639 cbp->bio_to = cp->provider; 1640 G_RAID3_LOGREQ(3, cbp, "Sending request."); 1641 KASSERT(cp->acr == 1 && cp->acw == 1 && cp->ace == 1, 1642 ("Consumer %s not opened (r%dw%de%d).", 1643 cp->provider->name, cp->acr, cp->acw, cp->ace)); 1644 cp->index++; 1645 g_io_request(cbp, cp); 1646 } 1647 break; 1648 case BIO_WRITE: 1649 case BIO_DELETE: 1650 /* 1651 * Bump syncid on first write. 1652 */ 1653 if ((sc->sc_bump_id & G_RAID3_BUMP_SYNCID) != 0) { 1654 sc->sc_bump_id &= ~G_RAID3_BUMP_SYNCID; 1655 g_topology_lock(); 1656 g_raid3_bump_syncid(sc); 1657 g_topology_unlock(); 1658 } 1659 g_raid3_scatter(pbp); 1660 break; 1661 } 1662 return (0); 1663 } 1664 1665 static int 1666 g_raid3_can_destroy(struct g_raid3_softc *sc) 1667 { 1668 struct g_geom *gp; 1669 struct g_consumer *cp; 1670 1671 g_topology_assert(); 1672 gp = sc->sc_geom; 1673 LIST_FOREACH(cp, &gp->consumer, consumer) { 1674 if (g_raid3_is_busy(sc, cp)) 1675 return (0); 1676 } 1677 gp = sc->sc_sync.ds_geom; 1678 LIST_FOREACH(cp, &gp->consumer, consumer) { 1679 if (g_raid3_is_busy(sc, cp)) 1680 return (0); 1681 } 1682 G_RAID3_DEBUG(2, "No I/O requests for %s, it can be destroyed.", 1683 sc->sc_name); 1684 return (1); 1685 } 1686 1687 static int 1688 g_raid3_try_destroy(struct g_raid3_softc *sc) 1689 { 1690 1691 g_topology_lock(); 1692 if (!g_raid3_can_destroy(sc)) { 1693 g_topology_unlock(); 1694 return (0); 1695 } 1696 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_WAIT) != 0) { 1697 g_topology_unlock(); 1698 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, 1699 &sc->sc_worker); 1700 wakeup(&sc->sc_worker); 1701 sc->sc_worker = NULL; 1702 } else { 1703 g_raid3_destroy_device(sc); 1704 g_topology_unlock(); 1705 free(sc->sc_disks, M_RAID3); 1706 free(sc, M_RAID3); 1707 } 1708 return (1); 1709 } 1710 1711 /* 1712 * Worker thread. 1713 */ 1714 static void 1715 g_raid3_worker(void *arg) 1716 { 1717 struct g_raid3_softc *sc; 1718 struct g_raid3_disk *disk; 1719 struct g_raid3_disk_sync *sync; 1720 struct g_raid3_event *ep; 1721 struct bio *bp; 1722 u_int nreqs; 1723 1724 sc = arg; 1725 mtx_lock_spin(&sched_lock); 1726 sched_prio(curthread, PRIBIO); 1727 mtx_unlock_spin(&sched_lock); 1728 1729 nreqs = 0; 1730 for (;;) { 1731 G_RAID3_DEBUG(5, "%s: Let's see...", __func__); 1732 /* 1733 * First take a look at events. 1734 * This is important to handle events before any I/O requests. 1735 */ 1736 ep = g_raid3_event_get(sc); 1737 if (ep != NULL && g_topology_try_lock()) { 1738 g_raid3_event_remove(sc, ep); 1739 if ((ep->e_flags & G_RAID3_EVENT_DEVICE) != 0) { 1740 /* Update only device status. */ 1741 G_RAID3_DEBUG(3, 1742 "Running event for device %s.", 1743 sc->sc_name); 1744 ep->e_error = 0; 1745 g_raid3_update_device(sc, 1); 1746 } else { 1747 /* Update disk status. */ 1748 G_RAID3_DEBUG(3, "Running event for disk %s.", 1749 g_raid3_get_diskname(ep->e_disk)); 1750 ep->e_error = g_raid3_update_disk(ep->e_disk, 1751 ep->e_state); 1752 if (ep->e_error == 0) 1753 g_raid3_update_device(sc, 0); 1754 } 1755 g_topology_unlock(); 1756 if ((ep->e_flags & G_RAID3_EVENT_DONTWAIT) != 0) { 1757 KASSERT(ep->e_error == 0, 1758 ("Error cannot be handled.")); 1759 g_raid3_event_free(ep); 1760 } else { 1761 ep->e_flags |= G_RAID3_EVENT_DONE; 1762 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, 1763 ep); 1764 mtx_lock(&sc->sc_events_mtx); 1765 wakeup(ep); 1766 mtx_unlock(&sc->sc_events_mtx); 1767 } 1768 if ((sc->sc_flags & 1769 G_RAID3_DEVICE_FLAG_DESTROY) != 0) { 1770 if (g_raid3_try_destroy(sc)) 1771 kthread_exit(0); 1772 } 1773 G_RAID3_DEBUG(5, "%s: I'm here 1.", __func__); 1774 continue; 1775 } 1776 /* 1777 * Now I/O requests. 1778 */ 1779 /* Get first request from the queue. */ 1780 mtx_lock(&sc->sc_queue_mtx); 1781 bp = bioq_first(&sc->sc_queue); 1782 if (bp == NULL) { 1783 if (ep != NULL) { 1784 /* 1785 * No I/O requests and topology lock was 1786 * already held? Try again. 1787 */ 1788 mtx_unlock(&sc->sc_queue_mtx); 1789 tsleep(ep, PRIBIO, "r3:top1", hz / 5); 1790 continue; 1791 } 1792 if ((sc->sc_flags & 1793 G_RAID3_DEVICE_FLAG_DESTROY) != 0) { 1794 mtx_unlock(&sc->sc_queue_mtx); 1795 if (g_raid3_try_destroy(sc)) 1796 kthread_exit(0); 1797 mtx_lock(&sc->sc_queue_mtx); 1798 } 1799 } 1800 if (sc->sc_syncdisk != NULL && 1801 (bp == NULL || nreqs > g_raid3_reqs_per_sync)) { 1802 mtx_unlock(&sc->sc_queue_mtx); 1803 /* 1804 * It is time for synchronization... 1805 */ 1806 nreqs = 0; 1807 disk = sc->sc_syncdisk; 1808 sync = &disk->d_sync; 1809 if (sync->ds_offset < 1810 sc->sc_mediasize / (sc->sc_ndisks - 1) && 1811 sync->ds_offset == sync->ds_offset_done) { 1812 if (sync->ds_resync != -1) { 1813 sync->ds_offset = sync->ds_resync; 1814 sync->ds_offset_done = sync->ds_resync; 1815 sync->ds_resync = -1; 1816 } 1817 g_raid3_sync_one(sc); 1818 } 1819 G_RAID3_DEBUG(5, "%s: I'm here 2.", __func__); 1820 goto sleep; 1821 } 1822 if (bp == NULL) { 1823 if (g_raid3_check_idle(sc)) { 1824 u_int idletime; 1825 1826 idletime = g_raid3_idletime; 1827 if (idletime == 0) 1828 idletime = 1; 1829 idletime *= hz; 1830 if (msleep(sc, &sc->sc_queue_mtx, PRIBIO | PDROP, 1831 "r3:w1", idletime) == EWOULDBLOCK) { 1832 G_RAID3_DEBUG(5, "%s: I'm here 3.", 1833 __func__); 1834 /* 1835 * No I/O requests in 'idletime' 1836 * seconds, so mark components as clean. 1837 */ 1838 g_raid3_idle(sc); 1839 } 1840 G_RAID3_DEBUG(5, "%s: I'm here 4.", __func__); 1841 } else { 1842 MSLEEP(sc, &sc->sc_queue_mtx, PRIBIO | PDROP, 1843 "r3:w2", 0); 1844 G_RAID3_DEBUG(5, "%s: I'm here 5.", __func__); 1845 } 1846 continue; 1847 } 1848 nreqs++; 1849 bioq_remove(&sc->sc_queue, bp); 1850 mtx_unlock(&sc->sc_queue_mtx); 1851 1852 if ((bp->bio_cflags & G_RAID3_BIO_CFLAG_REGULAR) != 0) { 1853 g_raid3_regular_request(bp); 1854 } else if ((bp->bio_cflags & G_RAID3_BIO_CFLAG_SYNC) != 0) { 1855 u_int timeout, sps; 1856 1857 g_raid3_sync_request(bp); 1858 sleep: 1859 sps = atomic_load_acq_int(&g_raid3_syncs_per_sec); 1860 if (sps == 0) { 1861 G_RAID3_DEBUG(5, "%s: I'm here 6.", __func__); 1862 continue; 1863 } 1864 if (ep != NULL) { 1865 /* 1866 * We have some pending events, don't sleep now. 1867 */ 1868 G_RAID3_DEBUG(5, "%s: I'm here 7.", __func__); 1869 tsleep(ep, PRIBIO, "r3:top2", hz / 5); 1870 continue; 1871 } 1872 mtx_lock(&sc->sc_queue_mtx); 1873 if (bioq_first(&sc->sc_queue) != NULL) { 1874 mtx_unlock(&sc->sc_queue_mtx); 1875 G_RAID3_DEBUG(5, "%s: I'm here 8.", __func__); 1876 continue; 1877 } 1878 timeout = hz / sps; 1879 if (timeout == 0) 1880 timeout = 1; 1881 MSLEEP(sc, &sc->sc_queue_mtx, PRIBIO | PDROP, "r3:w2", 1882 timeout); 1883 } else { 1884 if (g_raid3_register_request(bp) != 0) { 1885 mtx_lock(&sc->sc_queue_mtx); 1886 bioq_insert_tail(&sc->sc_queue, bp); 1887 MSLEEP(&sc->sc_queue, &sc->sc_queue_mtx, 1888 PRIBIO | PDROP, "r3:lowmem", hz / 10); 1889 } 1890 } 1891 G_RAID3_DEBUG(5, "%s: I'm here 9.", __func__); 1892 } 1893 } 1894 1895 /* 1896 * Open disk's consumer if needed. 1897 */ 1898 static void 1899 g_raid3_update_access(struct g_raid3_disk *disk) 1900 { 1901 struct g_provider *pp; 1902 1903 g_topology_assert(); 1904 1905 pp = disk->d_softc->sc_provider; 1906 if (pp == NULL) 1907 return; 1908 if (pp->acw > 0) { 1909 if ((disk->d_flags & G_RAID3_DISK_FLAG_DIRTY) == 0) { 1910 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as dirty.", 1911 g_raid3_get_diskname(disk), disk->d_softc->sc_name); 1912 disk->d_flags |= G_RAID3_DISK_FLAG_DIRTY; 1913 } 1914 } else if (pp->acw == 0) { 1915 if ((disk->d_flags & G_RAID3_DISK_FLAG_DIRTY) != 0) { 1916 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as clean.", 1917 g_raid3_get_diskname(disk), disk->d_softc->sc_name); 1918 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 1919 } 1920 } 1921 } 1922 1923 static void 1924 g_raid3_sync_start(struct g_raid3_softc *sc) 1925 { 1926 struct g_raid3_disk *disk; 1927 int error; 1928 u_int n; 1929 1930 g_topology_assert(); 1931 1932 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED, 1933 ("Device not in DEGRADED state (%s, %u).", sc->sc_name, 1934 sc->sc_state)); 1935 KASSERT(sc->sc_syncdisk == NULL, ("Syncdisk is not NULL (%s, %u).", 1936 sc->sc_name, sc->sc_state)); 1937 disk = NULL; 1938 for (n = 0; n < sc->sc_ndisks; n++) { 1939 if (sc->sc_disks[n].d_state != G_RAID3_DISK_STATE_SYNCHRONIZING) 1940 continue; 1941 disk = &sc->sc_disks[n]; 1942 break; 1943 } 1944 if (disk == NULL) 1945 return; 1946 1947 G_RAID3_DEBUG(0, "Device %s: rebuilding provider %s.", sc->sc_name, 1948 g_raid3_get_diskname(disk)); 1949 disk->d_flags |= G_RAID3_DISK_FLAG_DIRTY; 1950 KASSERT(disk->d_sync.ds_consumer == NULL, 1951 ("Sync consumer already exists (device=%s, disk=%s).", 1952 sc->sc_name, g_raid3_get_diskname(disk))); 1953 disk->d_sync.ds_consumer = g_new_consumer(sc->sc_sync.ds_geom); 1954 disk->d_sync.ds_consumer->private = disk; 1955 disk->d_sync.ds_consumer->index = 0; 1956 error = g_attach(disk->d_sync.ds_consumer, disk->d_softc->sc_provider); 1957 KASSERT(error == 0, ("Cannot attach to %s (error=%d).", 1958 disk->d_softc->sc_name, error)); 1959 error = g_access(disk->d_sync.ds_consumer, 1, 0, 0); 1960 KASSERT(error == 0, ("Cannot open %s (error=%d).", 1961 disk->d_softc->sc_name, error)); 1962 disk->d_sync.ds_data = malloc(MAXPHYS, M_RAID3, M_WAITOK); 1963 sc->sc_syncdisk = disk; 1964 } 1965 1966 /* 1967 * Stop synchronization process. 1968 * type: 0 - synchronization finished 1969 * 1 - synchronization stopped 1970 */ 1971 static void 1972 g_raid3_sync_stop(struct g_raid3_softc *sc, int type) 1973 { 1974 struct g_raid3_disk *disk; 1975 1976 g_topology_assert(); 1977 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED, 1978 ("Device not in DEGRADED state (%s, %u).", sc->sc_name, 1979 sc->sc_state)); 1980 disk = sc->sc_syncdisk; 1981 sc->sc_syncdisk = NULL; 1982 KASSERT(disk != NULL, ("No disk was synchronized (%s).", sc->sc_name)); 1983 KASSERT(disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING, 1984 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 1985 g_raid3_disk_state2str(disk->d_state))); 1986 if (disk->d_sync.ds_consumer == NULL) 1987 return; 1988 1989 if (type == 0) { 1990 G_RAID3_DEBUG(0, "Device %s: rebuilding provider %s finished.", 1991 disk->d_softc->sc_name, g_raid3_get_diskname(disk)); 1992 } else /* if (type == 1) */ { 1993 G_RAID3_DEBUG(0, "Device %s: rebuilding provider %s stopped.", 1994 disk->d_softc->sc_name, g_raid3_get_diskname(disk)); 1995 } 1996 g_raid3_kill_consumer(disk->d_softc, disk->d_sync.ds_consumer); 1997 free(disk->d_sync.ds_data, M_RAID3); 1998 disk->d_sync.ds_consumer = NULL; 1999 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 2000 } 2001 2002 static void 2003 g_raid3_launch_provider(struct g_raid3_softc *sc) 2004 { 2005 struct g_provider *pp; 2006 2007 g_topology_assert(); 2008 2009 pp = g_new_providerf(sc->sc_geom, "raid3/%s", sc->sc_name); 2010 pp->mediasize = sc->sc_mediasize; 2011 pp->sectorsize = sc->sc_sectorsize; 2012 sc->sc_provider = pp; 2013 g_error_provider(pp, 0); 2014 G_RAID3_DEBUG(0, "Device %s: provider %s launched.", sc->sc_name, 2015 pp->name); 2016 if (sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED) 2017 g_raid3_sync_start(sc); 2018 } 2019 2020 static void 2021 g_raid3_destroy_provider(struct g_raid3_softc *sc) 2022 { 2023 struct bio *bp; 2024 2025 g_topology_assert(); 2026 KASSERT(sc->sc_provider != NULL, ("NULL provider (device=%s).", 2027 sc->sc_name)); 2028 2029 g_error_provider(sc->sc_provider, ENXIO); 2030 mtx_lock(&sc->sc_queue_mtx); 2031 while ((bp = bioq_first(&sc->sc_queue)) != NULL) { 2032 bioq_remove(&sc->sc_queue, bp); 2033 g_io_deliver(bp, ENXIO); 2034 } 2035 mtx_unlock(&sc->sc_queue_mtx); 2036 G_RAID3_DEBUG(0, "Device %s: provider %s destroyed.", sc->sc_name, 2037 sc->sc_provider->name); 2038 sc->sc_provider->flags |= G_PF_WITHER; 2039 g_orphan_provider(sc->sc_provider, ENXIO); 2040 sc->sc_provider = NULL; 2041 if (sc->sc_syncdisk != NULL) 2042 g_raid3_sync_stop(sc, 1); 2043 } 2044 2045 static void 2046 g_raid3_go(void *arg) 2047 { 2048 struct g_raid3_softc *sc; 2049 2050 sc = arg; 2051 G_RAID3_DEBUG(0, "Force device %s start due to timeout.", sc->sc_name); 2052 g_raid3_event_send(sc, 0, 2053 G_RAID3_EVENT_DONTWAIT | G_RAID3_EVENT_DEVICE); 2054 } 2055 2056 static u_int 2057 g_raid3_determine_state(struct g_raid3_disk *disk) 2058 { 2059 struct g_raid3_softc *sc; 2060 u_int state; 2061 2062 sc = disk->d_softc; 2063 if (sc->sc_syncid == disk->d_sync.ds_syncid) { 2064 if ((disk->d_flags & 2065 G_RAID3_DISK_FLAG_SYNCHRONIZING) == 0) { 2066 /* Disk does not need synchronization. */ 2067 state = G_RAID3_DISK_STATE_ACTIVE; 2068 } else { 2069 if ((sc->sc_flags & 2070 G_RAID3_DEVICE_FLAG_NOAUTOSYNC) == 0 || 2071 (disk->d_flags & 2072 G_RAID3_DISK_FLAG_FORCE_SYNC) != 0) { 2073 /* 2074 * We can start synchronization from 2075 * the stored offset. 2076 */ 2077 state = G_RAID3_DISK_STATE_SYNCHRONIZING; 2078 } else { 2079 state = G_RAID3_DISK_STATE_STALE; 2080 } 2081 } 2082 } else if (disk->d_sync.ds_syncid < sc->sc_syncid) { 2083 /* 2084 * Reset all synchronization data for this disk, 2085 * because if it even was synchronized, it was 2086 * synchronized to disks with different syncid. 2087 */ 2088 disk->d_flags |= G_RAID3_DISK_FLAG_SYNCHRONIZING; 2089 disk->d_sync.ds_offset = 0; 2090 disk->d_sync.ds_offset_done = 0; 2091 disk->d_sync.ds_syncid = sc->sc_syncid; 2092 if ((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) == 0 || 2093 (disk->d_flags & G_RAID3_DISK_FLAG_FORCE_SYNC) != 0) { 2094 state = G_RAID3_DISK_STATE_SYNCHRONIZING; 2095 } else { 2096 state = G_RAID3_DISK_STATE_STALE; 2097 } 2098 } else /* if (sc->sc_syncid < disk->d_sync.ds_syncid) */ { 2099 /* 2100 * Not good, NOT GOOD! 2101 * It means that device was started on stale disks 2102 * and more fresh disk just arrive. 2103 * If there were writes, device is fucked up, sorry. 2104 * I think the best choice here is don't touch 2105 * this disk and inform the user laudly. 2106 */ 2107 G_RAID3_DEBUG(0, "Device %s was started before the freshest " 2108 "disk (%s) arrives!! It will not be connected to the " 2109 "running device.", sc->sc_name, 2110 g_raid3_get_diskname(disk)); 2111 g_raid3_destroy_disk(disk); 2112 state = G_RAID3_DISK_STATE_NONE; 2113 /* Return immediately, because disk was destroyed. */ 2114 return (state); 2115 } 2116 G_RAID3_DEBUG(3, "State for %s disk: %s.", 2117 g_raid3_get_diskname(disk), g_raid3_disk_state2str(state)); 2118 return (state); 2119 } 2120 2121 /* 2122 * Update device state. 2123 */ 2124 static void 2125 g_raid3_update_device(struct g_raid3_softc *sc, boolean_t force) 2126 { 2127 struct g_raid3_disk *disk; 2128 u_int state; 2129 2130 g_topology_assert(); 2131 2132 switch (sc->sc_state) { 2133 case G_RAID3_DEVICE_STATE_STARTING: 2134 { 2135 u_int n, ndirty, ndisks, genid, syncid; 2136 2137 KASSERT(sc->sc_provider == NULL, 2138 ("Non-NULL provider in STARTING state (%s).", sc->sc_name)); 2139 /* 2140 * Are we ready? We are, if all disks are connected or 2141 * one disk is missing and 'force' is true. 2142 */ 2143 if (g_raid3_ndisks(sc, -1) + force == sc->sc_ndisks) { 2144 if (!force) 2145 callout_drain(&sc->sc_callout); 2146 } else { 2147 if (force) { 2148 /* 2149 * Timeout expired, so destroy device. 2150 */ 2151 sc->sc_flags |= G_RAID3_DEVICE_FLAG_DESTROY; 2152 } 2153 return; 2154 } 2155 2156 /* 2157 * Find the biggest genid. 2158 */ 2159 genid = 0; 2160 for (n = 0; n < sc->sc_ndisks; n++) { 2161 disk = &sc->sc_disks[n]; 2162 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 2163 continue; 2164 if (disk->d_genid > genid) 2165 genid = disk->d_genid; 2166 } 2167 sc->sc_genid = genid; 2168 /* 2169 * Remove all disks without the biggest genid. 2170 */ 2171 for (n = 0; n < sc->sc_ndisks; n++) { 2172 disk = &sc->sc_disks[n]; 2173 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 2174 continue; 2175 if (disk->d_genid < genid) { 2176 G_RAID3_DEBUG(0, 2177 "Component %s (device %s) broken, skipping.", 2178 g_raid3_get_diskname(disk), sc->sc_name); 2179 g_raid3_destroy_disk(disk); 2180 } 2181 } 2182 2183 /* 2184 * There must be at least 'sc->sc_ndisks - 1' components 2185 * with the same syncid and without SYNCHRONIZING flag. 2186 */ 2187 2188 /* 2189 * Find the biggest syncid, number of valid components and 2190 * number of dirty components. 2191 */ 2192 ndirty = ndisks = syncid = 0; 2193 for (n = 0; n < sc->sc_ndisks; n++) { 2194 disk = &sc->sc_disks[n]; 2195 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 2196 continue; 2197 if ((disk->d_flags & G_RAID3_DISK_FLAG_DIRTY) != 0) 2198 ndirty++; 2199 if (disk->d_sync.ds_syncid > syncid) { 2200 syncid = disk->d_sync.ds_syncid; 2201 ndisks = 0; 2202 } else if (disk->d_sync.ds_syncid < syncid) { 2203 continue; 2204 } 2205 if ((disk->d_flags & 2206 G_RAID3_DISK_FLAG_SYNCHRONIZING) != 0) { 2207 continue; 2208 } 2209 ndisks++; 2210 } 2211 /* 2212 * Do we have enough valid components? 2213 */ 2214 if (ndisks + 1 < sc->sc_ndisks) { 2215 G_RAID3_DEBUG(0, 2216 "Device %s is broken, too few valid components.", 2217 sc->sc_name); 2218 sc->sc_flags |= G_RAID3_DEVICE_FLAG_DESTROY; 2219 return; 2220 } 2221 /* 2222 * If there is one DIRTY component and all disks are present, 2223 * mark it for synchronization. If there is more than one DIRTY 2224 * component, mark parity component for synchronization. 2225 */ 2226 if (ndisks == sc->sc_ndisks && ndirty == 1) { 2227 for (n = 0; n < sc->sc_ndisks; n++) { 2228 disk = &sc->sc_disks[n]; 2229 if ((disk->d_flags & 2230 G_RAID3_DISK_FLAG_DIRTY) == 0) { 2231 continue; 2232 } 2233 disk->d_flags |= 2234 G_RAID3_DISK_FLAG_SYNCHRONIZING; 2235 } 2236 } else if (ndisks == sc->sc_ndisks && ndirty > 1) { 2237 disk = &sc->sc_disks[sc->sc_ndisks - 1]; 2238 disk->d_flags |= G_RAID3_DISK_FLAG_SYNCHRONIZING; 2239 } 2240 2241 sc->sc_syncid = syncid; 2242 if (force) { 2243 /* Remember to bump syncid on first write. */ 2244 sc->sc_bump_id |= G_RAID3_BUMP_SYNCID; 2245 } 2246 if (ndisks == sc->sc_ndisks) 2247 state = G_RAID3_DEVICE_STATE_COMPLETE; 2248 else /* if (ndisks == sc->sc_ndisks - 1) */ 2249 state = G_RAID3_DEVICE_STATE_DEGRADED; 2250 G_RAID3_DEBUG(1, "Device %s state changed from %s to %s.", 2251 sc->sc_name, g_raid3_device_state2str(sc->sc_state), 2252 g_raid3_device_state2str(state)); 2253 sc->sc_state = state; 2254 for (n = 0; n < sc->sc_ndisks; n++) { 2255 disk = &sc->sc_disks[n]; 2256 if (disk->d_state == G_RAID3_DISK_STATE_NODISK) 2257 continue; 2258 state = g_raid3_determine_state(disk); 2259 g_raid3_event_send(disk, state, G_RAID3_EVENT_DONTWAIT); 2260 if (state == G_RAID3_DISK_STATE_STALE) 2261 sc->sc_bump_id |= G_RAID3_BUMP_SYNCID; 2262 } 2263 break; 2264 } 2265 case G_RAID3_DEVICE_STATE_DEGRADED: 2266 /* 2267 * Genid need to be bumped immediately, so do it here. 2268 */ 2269 if ((sc->sc_bump_id & G_RAID3_BUMP_GENID) != 0) { 2270 sc->sc_bump_id &= ~G_RAID3_BUMP_GENID; 2271 g_raid3_bump_genid(sc); 2272 } 2273 2274 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_NEW) > 0) 2275 return; 2276 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) < 2277 sc->sc_ndisks - 1) { 2278 if (sc->sc_provider != NULL) 2279 g_raid3_destroy_provider(sc); 2280 sc->sc_flags |= G_RAID3_DEVICE_FLAG_DESTROY; 2281 return; 2282 } 2283 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) == 2284 sc->sc_ndisks) { 2285 state = G_RAID3_DEVICE_STATE_COMPLETE; 2286 G_RAID3_DEBUG(1, 2287 "Device %s state changed from %s to %s.", 2288 sc->sc_name, g_raid3_device_state2str(sc->sc_state), 2289 g_raid3_device_state2str(state)); 2290 sc->sc_state = state; 2291 } 2292 if (sc->sc_provider == NULL) 2293 g_raid3_launch_provider(sc); 2294 break; 2295 case G_RAID3_DEVICE_STATE_COMPLETE: 2296 /* 2297 * Genid need to be bumped immediately, so do it here. 2298 */ 2299 if ((sc->sc_bump_id & G_RAID3_BUMP_GENID) != 0) { 2300 sc->sc_bump_id &= ~G_RAID3_BUMP_GENID; 2301 g_raid3_bump_genid(sc); 2302 } 2303 2304 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_NEW) > 0) 2305 return; 2306 KASSERT(g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) >= 2307 sc->sc_ndisks - 1, 2308 ("Too few ACTIVE components in COMPLETE state (device %s).", 2309 sc->sc_name)); 2310 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) == 2311 sc->sc_ndisks - 1) { 2312 state = G_RAID3_DEVICE_STATE_DEGRADED; 2313 G_RAID3_DEBUG(1, 2314 "Device %s state changed from %s to %s.", 2315 sc->sc_name, g_raid3_device_state2str(sc->sc_state), 2316 g_raid3_device_state2str(state)); 2317 sc->sc_state = state; 2318 } 2319 if (sc->sc_provider == NULL) 2320 g_raid3_launch_provider(sc); 2321 break; 2322 default: 2323 KASSERT(1 == 0, ("Wrong device state (%s, %s).", sc->sc_name, 2324 g_raid3_device_state2str(sc->sc_state))); 2325 break; 2326 } 2327 } 2328 2329 /* 2330 * Update disk state and device state if needed. 2331 */ 2332 #define DISK_STATE_CHANGED() G_RAID3_DEBUG(1, \ 2333 "Disk %s state changed from %s to %s (device %s).", \ 2334 g_raid3_get_diskname(disk), \ 2335 g_raid3_disk_state2str(disk->d_state), \ 2336 g_raid3_disk_state2str(state), sc->sc_name) 2337 static int 2338 g_raid3_update_disk(struct g_raid3_disk *disk, u_int state) 2339 { 2340 struct g_raid3_softc *sc; 2341 2342 g_topology_assert(); 2343 2344 sc = disk->d_softc; 2345 again: 2346 G_RAID3_DEBUG(3, "Changing disk %s state from %s to %s.", 2347 g_raid3_get_diskname(disk), g_raid3_disk_state2str(disk->d_state), 2348 g_raid3_disk_state2str(state)); 2349 switch (state) { 2350 case G_RAID3_DISK_STATE_NEW: 2351 /* 2352 * Possible scenarios: 2353 * 1. New disk arrive. 2354 */ 2355 /* Previous state should be NONE. */ 2356 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NONE, 2357 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 2358 g_raid3_disk_state2str(disk->d_state))); 2359 DISK_STATE_CHANGED(); 2360 2361 disk->d_state = state; 2362 G_RAID3_DEBUG(0, "Device %s: provider %s detected.", 2363 sc->sc_name, g_raid3_get_diskname(disk)); 2364 if (sc->sc_state == G_RAID3_DEVICE_STATE_STARTING) 2365 break; 2366 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2367 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE, 2368 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2369 g_raid3_device_state2str(sc->sc_state), 2370 g_raid3_get_diskname(disk), 2371 g_raid3_disk_state2str(disk->d_state))); 2372 state = g_raid3_determine_state(disk); 2373 if (state != G_RAID3_DISK_STATE_NONE) 2374 goto again; 2375 break; 2376 case G_RAID3_DISK_STATE_ACTIVE: 2377 /* 2378 * Possible scenarios: 2379 * 1. New disk does not need synchronization. 2380 * 2. Synchronization process finished successfully. 2381 */ 2382 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2383 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE, 2384 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2385 g_raid3_device_state2str(sc->sc_state), 2386 g_raid3_get_diskname(disk), 2387 g_raid3_disk_state2str(disk->d_state))); 2388 /* Previous state should be NEW or SYNCHRONIZING. */ 2389 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NEW || 2390 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING, 2391 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 2392 g_raid3_disk_state2str(disk->d_state))); 2393 DISK_STATE_CHANGED(); 2394 2395 if (disk->d_state == G_RAID3_DISK_STATE_NEW) 2396 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 2397 else if (disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 2398 disk->d_flags &= ~G_RAID3_DISK_FLAG_SYNCHRONIZING; 2399 disk->d_flags &= ~G_RAID3_DISK_FLAG_FORCE_SYNC; 2400 g_raid3_sync_stop(sc, 0); 2401 } 2402 disk->d_state = state; 2403 disk->d_sync.ds_offset = 0; 2404 disk->d_sync.ds_offset_done = 0; 2405 g_raid3_update_access(disk); 2406 g_raid3_update_metadata(disk); 2407 G_RAID3_DEBUG(0, "Device %s: provider %s activated.", 2408 sc->sc_name, g_raid3_get_diskname(disk)); 2409 break; 2410 case G_RAID3_DISK_STATE_STALE: 2411 /* 2412 * Possible scenarios: 2413 * 1. Stale disk was connected. 2414 */ 2415 /* Previous state should be NEW. */ 2416 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NEW, 2417 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 2418 g_raid3_disk_state2str(disk->d_state))); 2419 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2420 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE, 2421 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2422 g_raid3_device_state2str(sc->sc_state), 2423 g_raid3_get_diskname(disk), 2424 g_raid3_disk_state2str(disk->d_state))); 2425 /* 2426 * STALE state is only possible if device is marked 2427 * NOAUTOSYNC. 2428 */ 2429 KASSERT((sc->sc_flags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0, 2430 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2431 g_raid3_device_state2str(sc->sc_state), 2432 g_raid3_get_diskname(disk), 2433 g_raid3_disk_state2str(disk->d_state))); 2434 DISK_STATE_CHANGED(); 2435 2436 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 2437 disk->d_state = state; 2438 g_raid3_update_metadata(disk); 2439 G_RAID3_DEBUG(0, "Device %s: provider %s is stale.", 2440 sc->sc_name, g_raid3_get_diskname(disk)); 2441 break; 2442 case G_RAID3_DISK_STATE_SYNCHRONIZING: 2443 /* 2444 * Possible scenarios: 2445 * 1. Disk which needs synchronization was connected. 2446 */ 2447 /* Previous state should be NEW. */ 2448 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NEW, 2449 ("Wrong disk state (%s, %s).", g_raid3_get_diskname(disk), 2450 g_raid3_disk_state2str(disk->d_state))); 2451 KASSERT(sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2452 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE, 2453 ("Wrong device state (%s, %s, %s, %s).", sc->sc_name, 2454 g_raid3_device_state2str(sc->sc_state), 2455 g_raid3_get_diskname(disk), 2456 g_raid3_disk_state2str(disk->d_state))); 2457 DISK_STATE_CHANGED(); 2458 2459 if (disk->d_state == G_RAID3_DISK_STATE_NEW) 2460 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 2461 disk->d_state = state; 2462 if (sc->sc_provider != NULL) { 2463 g_raid3_sync_start(sc); 2464 g_raid3_update_metadata(disk); 2465 } 2466 break; 2467 case G_RAID3_DISK_STATE_DISCONNECTED: 2468 /* 2469 * Possible scenarios: 2470 * 1. Device wasn't running yet, but disk disappear. 2471 * 2. Disk was active and disapppear. 2472 * 3. Disk disappear during synchronization process. 2473 */ 2474 if (sc->sc_state == G_RAID3_DEVICE_STATE_DEGRADED || 2475 sc->sc_state == G_RAID3_DEVICE_STATE_COMPLETE) { 2476 /* 2477 * Previous state should be ACTIVE, STALE or 2478 * SYNCHRONIZING. 2479 */ 2480 KASSERT(disk->d_state == G_RAID3_DISK_STATE_ACTIVE || 2481 disk->d_state == G_RAID3_DISK_STATE_STALE || 2482 disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING, 2483 ("Wrong disk state (%s, %s).", 2484 g_raid3_get_diskname(disk), 2485 g_raid3_disk_state2str(disk->d_state))); 2486 } else if (sc->sc_state == G_RAID3_DEVICE_STATE_STARTING) { 2487 /* Previous state should be NEW. */ 2488 KASSERT(disk->d_state == G_RAID3_DISK_STATE_NEW, 2489 ("Wrong disk state (%s, %s).", 2490 g_raid3_get_diskname(disk), 2491 g_raid3_disk_state2str(disk->d_state))); 2492 /* 2493 * Reset bumping syncid if disk disappeared in STARTING 2494 * state. 2495 */ 2496 if ((sc->sc_bump_id & G_RAID3_BUMP_SYNCID) != 0) 2497 sc->sc_bump_id &= ~G_RAID3_BUMP_SYNCID; 2498 #ifdef INVARIANTS 2499 } else { 2500 KASSERT(1 == 0, ("Wrong device state (%s, %s, %s, %s).", 2501 sc->sc_name, 2502 g_raid3_device_state2str(sc->sc_state), 2503 g_raid3_get_diskname(disk), 2504 g_raid3_disk_state2str(disk->d_state))); 2505 #endif 2506 } 2507 DISK_STATE_CHANGED(); 2508 G_RAID3_DEBUG(0, "Device %s: provider %s disconnected.", 2509 sc->sc_name, g_raid3_get_diskname(disk)); 2510 2511 g_raid3_destroy_disk(disk); 2512 break; 2513 default: 2514 KASSERT(1 == 0, ("Unknown state (%u).", state)); 2515 break; 2516 } 2517 return (0); 2518 } 2519 #undef DISK_STATE_CHANGED 2520 2521 int 2522 g_raid3_read_metadata(struct g_consumer *cp, struct g_raid3_metadata *md) 2523 { 2524 struct g_provider *pp; 2525 u_char *buf; 2526 int error; 2527 2528 g_topology_assert(); 2529 2530 error = g_access(cp, 1, 0, 0); 2531 if (error != 0) 2532 return (error); 2533 pp = cp->provider; 2534 g_topology_unlock(); 2535 /* Metadata are stored on last sector. */ 2536 buf = g_read_data(cp, pp->mediasize - pp->sectorsize, pp->sectorsize, 2537 &error); 2538 g_topology_lock(); 2539 g_access(cp, -1, 0, 0); 2540 if (error != 0) { 2541 G_RAID3_DEBUG(1, "Cannot read metadata from %s (error=%d).", 2542 cp->provider->name, error); 2543 if (buf != NULL) 2544 g_free(buf); 2545 return (error); 2546 } 2547 2548 /* Decode metadata. */ 2549 error = raid3_metadata_decode(buf, md); 2550 g_free(buf); 2551 if (strcmp(md->md_magic, G_RAID3_MAGIC) != 0) 2552 return (EINVAL); 2553 if (md->md_version > G_RAID3_VERSION) { 2554 G_RAID3_DEBUG(0, 2555 "Kernel module is too old to handle metadata from %s.", 2556 cp->provider->name); 2557 return (EINVAL); 2558 } 2559 if (error != 0) { 2560 G_RAID3_DEBUG(1, "MD5 metadata hash mismatch for provider %s.", 2561 cp->provider->name); 2562 return (error); 2563 } 2564 2565 return (0); 2566 } 2567 2568 static int 2569 g_raid3_check_metadata(struct g_raid3_softc *sc, struct g_provider *pp, 2570 struct g_raid3_metadata *md) 2571 { 2572 2573 if (md->md_no >= sc->sc_ndisks) { 2574 G_RAID3_DEBUG(1, "Invalid disk %s number (no=%u), skipping.", 2575 pp->name, md->md_no); 2576 return (EINVAL); 2577 } 2578 if (sc->sc_disks[md->md_no].d_state != G_RAID3_DISK_STATE_NODISK) { 2579 G_RAID3_DEBUG(1, "Disk %s (no=%u) already exists, skipping.", 2580 pp->name, md->md_no); 2581 return (EEXIST); 2582 } 2583 if (md->md_all != sc->sc_ndisks) { 2584 G_RAID3_DEBUG(1, 2585 "Invalid '%s' field on disk %s (device %s), skipping.", 2586 "md_all", pp->name, sc->sc_name); 2587 return (EINVAL); 2588 } 2589 if (md->md_mediasize != sc->sc_mediasize) { 2590 G_RAID3_DEBUG(1, 2591 "Invalid '%s' field on disk %s (device %s), skipping.", 2592 "md_mediasize", pp->name, sc->sc_name); 2593 return (EINVAL); 2594 } 2595 if ((md->md_mediasize % (sc->sc_ndisks - 1)) != 0) { 2596 G_RAID3_DEBUG(1, 2597 "Invalid '%s' field on disk %s (device %s), skipping.", 2598 "md_mediasize", pp->name, sc->sc_name); 2599 return (EINVAL); 2600 } 2601 if ((sc->sc_mediasize / (sc->sc_ndisks - 1)) > pp->mediasize) { 2602 G_RAID3_DEBUG(1, 2603 "Invalid size of disk %s (device %s), skipping.", pp->name, 2604 sc->sc_name); 2605 return (EINVAL); 2606 } 2607 if ((md->md_sectorsize / pp->sectorsize) < sc->sc_ndisks - 1) { 2608 G_RAID3_DEBUG(1, 2609 "Invalid '%s' field on disk %s (device %s), skipping.", 2610 "md_sectorsize", pp->name, sc->sc_name); 2611 return (EINVAL); 2612 } 2613 if (md->md_sectorsize != sc->sc_sectorsize) { 2614 G_RAID3_DEBUG(1, 2615 "Invalid '%s' field on disk %s (device %s), skipping.", 2616 "md_sectorsize", pp->name, sc->sc_name); 2617 return (EINVAL); 2618 } 2619 if ((sc->sc_sectorsize % pp->sectorsize) != 0) { 2620 G_RAID3_DEBUG(1, 2621 "Invalid sector size of disk %s (device %s), skipping.", 2622 pp->name, sc->sc_name); 2623 return (EINVAL); 2624 } 2625 if ((md->md_mflags & ~G_RAID3_DEVICE_FLAG_MASK) != 0) { 2626 G_RAID3_DEBUG(1, 2627 "Invalid device flags on disk %s (device %s), skipping.", 2628 pp->name, sc->sc_name); 2629 return (EINVAL); 2630 } 2631 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_VERIFY) != 0 && 2632 (md->md_mflags & G_RAID3_DEVICE_FLAG_ROUND_ROBIN) != 0) { 2633 /* 2634 * VERIFY and ROUND-ROBIN options are mutally exclusive. 2635 */ 2636 G_RAID3_DEBUG(1, "Both VERIFY and ROUND-ROBIN flags exist on " 2637 "disk %s (device %s), skipping.", pp->name, sc->sc_name); 2638 return (EINVAL); 2639 } 2640 if ((md->md_dflags & ~G_RAID3_DISK_FLAG_MASK) != 0) { 2641 G_RAID3_DEBUG(1, 2642 "Invalid disk flags on disk %s (device %s), skipping.", 2643 pp->name, sc->sc_name); 2644 return (EINVAL); 2645 } 2646 return (0); 2647 } 2648 2649 int 2650 g_raid3_add_disk(struct g_raid3_softc *sc, struct g_provider *pp, 2651 struct g_raid3_metadata *md) 2652 { 2653 struct g_raid3_disk *disk; 2654 int error; 2655 2656 g_topology_assert(); 2657 G_RAID3_DEBUG(2, "Adding disk %s.", pp->name); 2658 2659 error = g_raid3_check_metadata(sc, pp, md); 2660 if (error != 0) 2661 return (error); 2662 if (sc->sc_state != G_RAID3_DEVICE_STATE_STARTING && 2663 md->md_genid < sc->sc_genid) { 2664 G_RAID3_DEBUG(0, "Component %s (device %s) broken, skipping.", 2665 pp->name, sc->sc_name); 2666 return (EINVAL); 2667 } 2668 disk = g_raid3_init_disk(sc, pp, md, &error); 2669 if (disk == NULL) 2670 return (error); 2671 error = g_raid3_event_send(disk, G_RAID3_DISK_STATE_NEW, 2672 G_RAID3_EVENT_WAIT); 2673 if (error != 0) 2674 return (error); 2675 if (md->md_version < G_RAID3_VERSION) { 2676 G_RAID3_DEBUG(0, "Upgrading metadata on %s (v%d->v%d).", 2677 pp->name, md->md_version, G_RAID3_VERSION); 2678 g_raid3_update_metadata(disk); 2679 } 2680 return (0); 2681 } 2682 2683 static int 2684 g_raid3_access(struct g_provider *pp, int acr, int acw, int ace) 2685 { 2686 struct g_raid3_softc *sc; 2687 struct g_raid3_disk *disk; 2688 int dcr, dcw, dce; 2689 u_int n; 2690 2691 g_topology_assert(); 2692 G_RAID3_DEBUG(2, "Access request for %s: r%dw%de%d.", pp->name, acr, 2693 acw, ace); 2694 2695 dcr = pp->acr + acr; 2696 dcw = pp->acw + acw; 2697 dce = pp->ace + ace; 2698 2699 sc = pp->geom->softc; 2700 if (sc == NULL || 2701 g_raid3_ndisks(sc, G_RAID3_DISK_STATE_ACTIVE) < sc->sc_ndisks - 1 || 2702 (sc->sc_flags & G_RAID3_DEVICE_FLAG_DESTROY) != 0) { 2703 if (acr <= 0 && acw <= 0 && ace <= 0) 2704 return (0); 2705 else 2706 return (ENXIO); 2707 } 2708 for (n = 0; n < sc->sc_ndisks; n++) { 2709 disk = &sc->sc_disks[n]; 2710 if (disk->d_state != G_RAID3_DISK_STATE_ACTIVE) 2711 continue; 2712 /* 2713 * Mark disk as dirty on open and unmark on close. 2714 */ 2715 if (pp->acw == 0 && dcw > 0) { 2716 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as dirty.", 2717 g_raid3_get_diskname(disk), sc->sc_name); 2718 disk->d_flags |= G_RAID3_DISK_FLAG_DIRTY; 2719 g_raid3_update_metadata(disk); 2720 } else if (pp->acw > 0 && dcw == 0) { 2721 G_RAID3_DEBUG(1, "Disk %s (device %s) marked as clean.", 2722 g_raid3_get_diskname(disk), sc->sc_name); 2723 disk->d_flags &= ~G_RAID3_DISK_FLAG_DIRTY; 2724 g_raid3_update_metadata(disk); 2725 } 2726 } 2727 return (0); 2728 } 2729 2730 static struct g_geom * 2731 g_raid3_create(struct g_class *mp, const struct g_raid3_metadata *md) 2732 { 2733 struct g_raid3_softc *sc; 2734 struct g_geom *gp; 2735 int error, timeout; 2736 u_int n; 2737 2738 g_topology_assert(); 2739 G_RAID3_DEBUG(1, "Creating device %s (id=%u).", md->md_name, md->md_id); 2740 2741 /* One disk is minimum. */ 2742 if (md->md_all < 1) 2743 return (NULL); 2744 /* 2745 * Action geom. 2746 */ 2747 gp = g_new_geomf(mp, "%s", md->md_name); 2748 sc = malloc(sizeof(*sc), M_RAID3, M_WAITOK | M_ZERO); 2749 sc->sc_disks = malloc(sizeof(struct g_raid3_disk) * md->md_all, M_RAID3, 2750 M_WAITOK | M_ZERO); 2751 gp->start = g_raid3_start; 2752 gp->orphan = g_raid3_orphan; 2753 gp->access = g_raid3_access; 2754 gp->dumpconf = g_raid3_dumpconf; 2755 2756 sc->sc_id = md->md_id; 2757 sc->sc_mediasize = md->md_mediasize; 2758 sc->sc_sectorsize = md->md_sectorsize; 2759 sc->sc_ndisks = md->md_all; 2760 sc->sc_round_robin = 0; 2761 sc->sc_flags = md->md_mflags; 2762 sc->sc_bump_id = 0; 2763 sc->sc_idle = 0; 2764 for (n = 0; n < sc->sc_ndisks; n++) { 2765 sc->sc_disks[n].d_softc = sc; 2766 sc->sc_disks[n].d_no = n; 2767 sc->sc_disks[n].d_state = G_RAID3_DISK_STATE_NODISK; 2768 } 2769 bioq_init(&sc->sc_queue); 2770 mtx_init(&sc->sc_queue_mtx, "graid3:queue", NULL, MTX_DEF); 2771 TAILQ_INIT(&sc->sc_events); 2772 mtx_init(&sc->sc_events_mtx, "graid3:events", NULL, MTX_DEF); 2773 callout_init(&sc->sc_callout, CALLOUT_MPSAFE); 2774 sc->sc_state = G_RAID3_DEVICE_STATE_STARTING; 2775 gp->softc = sc; 2776 sc->sc_geom = gp; 2777 sc->sc_provider = NULL; 2778 /* 2779 * Synchronization geom. 2780 */ 2781 gp = g_new_geomf(mp, "%s.sync", md->md_name); 2782 gp->softc = sc; 2783 gp->orphan = g_raid3_orphan; 2784 sc->sc_sync.ds_geom = gp; 2785 sc->sc_zone_64k = uma_zcreate("gr3:64k", 65536, NULL, NULL, NULL, NULL, 2786 UMA_ALIGN_PTR, 0); 2787 uma_zone_set_max(sc->sc_zone_64k, g_raid3_n64k); 2788 sc->sc_zone_16k = uma_zcreate("gr3:16k", 16384, NULL, NULL, NULL, NULL, 2789 UMA_ALIGN_PTR, 0); 2790 uma_zone_set_max(sc->sc_zone_64k, g_raid3_n16k); 2791 sc->sc_zone_4k = uma_zcreate("gr3:4k", 4096, NULL, NULL, NULL, NULL, 2792 UMA_ALIGN_PTR, 0); 2793 uma_zone_set_max(sc->sc_zone_4k, g_raid3_n4k); 2794 error = kthread_create(g_raid3_worker, sc, &sc->sc_worker, 0, 0, 2795 "g_raid3 %s", md->md_name); 2796 if (error != 0) { 2797 G_RAID3_DEBUG(1, "Cannot create kernel thread for %s.", 2798 sc->sc_name); 2799 uma_zdestroy(sc->sc_zone_64k); 2800 uma_zdestroy(sc->sc_zone_16k); 2801 uma_zdestroy(sc->sc_zone_4k); 2802 g_destroy_geom(sc->sc_sync.ds_geom); 2803 mtx_destroy(&sc->sc_events_mtx); 2804 mtx_destroy(&sc->sc_queue_mtx); 2805 g_destroy_geom(sc->sc_geom); 2806 free(sc->sc_disks, M_RAID3); 2807 free(sc, M_RAID3); 2808 return (NULL); 2809 } 2810 2811 G_RAID3_DEBUG(0, "Device %s created (id=%u).", sc->sc_name, sc->sc_id); 2812 2813 /* 2814 * Run timeout. 2815 */ 2816 timeout = atomic_load_acq_int(&g_raid3_timeout); 2817 callout_reset(&sc->sc_callout, timeout * hz, g_raid3_go, sc); 2818 return (sc->sc_geom); 2819 } 2820 2821 int 2822 g_raid3_destroy(struct g_raid3_softc *sc, boolean_t force) 2823 { 2824 struct g_provider *pp; 2825 2826 g_topology_assert(); 2827 2828 if (sc == NULL) 2829 return (ENXIO); 2830 pp = sc->sc_provider; 2831 if (pp != NULL && (pp->acr != 0 || pp->acw != 0 || pp->ace != 0)) { 2832 if (force) { 2833 G_RAID3_DEBUG(1, "Device %s is still open, so it " 2834 "can't be definitely removed.", pp->name); 2835 } else { 2836 G_RAID3_DEBUG(1, 2837 "Device %s is still open (r%dw%de%d).", pp->name, 2838 pp->acr, pp->acw, pp->ace); 2839 return (EBUSY); 2840 } 2841 } 2842 2843 sc->sc_flags |= G_RAID3_DEVICE_FLAG_DESTROY; 2844 sc->sc_flags |= G_RAID3_DEVICE_FLAG_WAIT; 2845 g_topology_unlock(); 2846 G_RAID3_DEBUG(4, "%s: Waking up %p.", __func__, sc); 2847 mtx_lock(&sc->sc_queue_mtx); 2848 wakeup(sc); 2849 wakeup(&sc->sc_queue); 2850 mtx_unlock(&sc->sc_queue_mtx); 2851 G_RAID3_DEBUG(4, "%s: Sleeping %p.", __func__, &sc->sc_worker); 2852 while (sc->sc_worker != NULL) 2853 tsleep(&sc->sc_worker, PRIBIO, "r3:destroy", hz / 5); 2854 G_RAID3_DEBUG(4, "%s: Woken up %p.", __func__, &sc->sc_worker); 2855 g_topology_lock(); 2856 g_raid3_destroy_device(sc); 2857 free(sc->sc_disks, M_RAID3); 2858 free(sc, M_RAID3); 2859 return (0); 2860 } 2861 2862 static void 2863 g_raid3_taste_orphan(struct g_consumer *cp) 2864 { 2865 2866 KASSERT(1 == 0, ("%s called while tasting %s.", __func__, 2867 cp->provider->name)); 2868 } 2869 2870 static struct g_geom * 2871 g_raid3_taste(struct g_class *mp, struct g_provider *pp, int flags __unused) 2872 { 2873 struct g_raid3_metadata md; 2874 struct g_raid3_softc *sc; 2875 struct g_consumer *cp; 2876 struct g_geom *gp; 2877 int error; 2878 2879 g_topology_assert(); 2880 g_trace(G_T_TOPOLOGY, "%s(%s, %s)", __func__, mp->name, pp->name); 2881 G_RAID3_DEBUG(2, "Tasting %s.", pp->name); 2882 2883 gp = g_new_geomf(mp, "raid3:taste"); 2884 /* This orphan function should be never called. */ 2885 gp->orphan = g_raid3_taste_orphan; 2886 cp = g_new_consumer(gp); 2887 g_attach(cp, pp); 2888 error = g_raid3_read_metadata(cp, &md); 2889 g_detach(cp); 2890 g_destroy_consumer(cp); 2891 g_destroy_geom(gp); 2892 if (error != 0) 2893 return (NULL); 2894 gp = NULL; 2895 2896 if (md.md_provider[0] != '\0' && strcmp(md.md_provider, pp->name) != 0) 2897 return (NULL); 2898 if (md.md_provsize != 0 && md.md_provsize != pp->mediasize) 2899 return (NULL); 2900 if (g_raid3_debug >= 2) 2901 raid3_metadata_dump(&md); 2902 2903 /* 2904 * Let's check if device already exists. 2905 */ 2906 sc = NULL; 2907 LIST_FOREACH(gp, &mp->geom, geom) { 2908 sc = gp->softc; 2909 if (sc == NULL) 2910 continue; 2911 if (sc->sc_sync.ds_geom == gp) 2912 continue; 2913 if (strcmp(md.md_name, sc->sc_name) != 0) 2914 continue; 2915 if (md.md_id != sc->sc_id) { 2916 G_RAID3_DEBUG(0, "Device %s already configured.", 2917 sc->sc_name); 2918 return (NULL); 2919 } 2920 break; 2921 } 2922 if (gp == NULL) { 2923 gp = g_raid3_create(mp, &md); 2924 if (gp == NULL) { 2925 G_RAID3_DEBUG(0, "Cannot create device %s.", 2926 md.md_name); 2927 return (NULL); 2928 } 2929 sc = gp->softc; 2930 } 2931 G_RAID3_DEBUG(1, "Adding disk %s to %s.", pp->name, gp->name); 2932 error = g_raid3_add_disk(sc, pp, &md); 2933 if (error != 0) { 2934 G_RAID3_DEBUG(0, "Cannot add disk %s to %s (error=%d).", 2935 pp->name, gp->name, error); 2936 if (g_raid3_ndisks(sc, G_RAID3_DISK_STATE_NODISK) == 2937 sc->sc_ndisks) { 2938 g_raid3_destroy(sc, 1); 2939 } 2940 return (NULL); 2941 } 2942 return (gp); 2943 } 2944 2945 static int 2946 g_raid3_destroy_geom(struct gctl_req *req __unused, struct g_class *mp __unused, 2947 struct g_geom *gp) 2948 { 2949 2950 return (g_raid3_destroy(gp->softc, 0)); 2951 } 2952 2953 static void 2954 g_raid3_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, 2955 struct g_consumer *cp, struct g_provider *pp) 2956 { 2957 struct g_raid3_softc *sc; 2958 2959 g_topology_assert(); 2960 2961 sc = gp->softc; 2962 if (sc == NULL) 2963 return; 2964 /* Skip synchronization geom. */ 2965 if (gp == sc->sc_sync.ds_geom) 2966 return; 2967 if (pp != NULL) { 2968 /* Nothing here. */ 2969 } else if (cp != NULL) { 2970 struct g_raid3_disk *disk; 2971 2972 disk = cp->private; 2973 if (disk == NULL) 2974 return; 2975 sbuf_printf(sb, "%s<Type>", indent); 2976 if (disk->d_no == sc->sc_ndisks - 1) 2977 sbuf_printf(sb, "PARITY"); 2978 else 2979 sbuf_printf(sb, "DATA"); 2980 sbuf_printf(sb, "</Type>\n"); 2981 sbuf_printf(sb, "%s<Number>%u</Number>\n", indent, 2982 (u_int)disk->d_no); 2983 if (disk->d_state == G_RAID3_DISK_STATE_SYNCHRONIZING) { 2984 sbuf_printf(sb, "%s<Synchronized>", indent); 2985 if (disk->d_sync.ds_offset_done == 0) 2986 sbuf_printf(sb, "0%%"); 2987 else { 2988 sbuf_printf(sb, "%u%%", 2989 (u_int)((disk->d_sync.ds_offset_done * 100) / 2990 (sc->sc_mediasize / (sc->sc_ndisks - 1)))); 2991 } 2992 sbuf_printf(sb, "</Synchronized>\n"); 2993 } 2994 sbuf_printf(sb, "%s<SyncID>%u</SyncID>\n", indent, 2995 disk->d_sync.ds_syncid); 2996 sbuf_printf(sb, "%s<GenID>%u</GenID>\n", indent, disk->d_genid); 2997 sbuf_printf(sb, "%s<Flags>", indent); 2998 if (disk->d_flags == 0) 2999 sbuf_printf(sb, "NONE"); 3000 else { 3001 int first = 1; 3002 3003 #define ADD_FLAG(flag, name) do { \ 3004 if ((disk->d_flags & (flag)) != 0) { \ 3005 if (!first) \ 3006 sbuf_printf(sb, ", "); \ 3007 else \ 3008 first = 0; \ 3009 sbuf_printf(sb, name); \ 3010 } \ 3011 } while (0) 3012 ADD_FLAG(G_RAID3_DISK_FLAG_DIRTY, "DIRTY"); 3013 ADD_FLAG(G_RAID3_DISK_FLAG_HARDCODED, "HARDCODED"); 3014 ADD_FLAG(G_RAID3_DISK_FLAG_SYNCHRONIZING, 3015 "SYNCHRONIZING"); 3016 ADD_FLAG(G_RAID3_DISK_FLAG_FORCE_SYNC, "FORCE_SYNC"); 3017 #undef ADD_FLAG 3018 } 3019 sbuf_printf(sb, "</Flags>\n"); 3020 sbuf_printf(sb, "%s<State>%s</State>\n", indent, 3021 g_raid3_disk_state2str(disk->d_state)); 3022 } else { 3023 sbuf_printf(sb, "%s<ID>%u</ID>\n", indent, (u_int)sc->sc_id); 3024 sbuf_printf(sb, "%s<SyncID>%u</SyncID>\n", indent, sc->sc_syncid); 3025 sbuf_printf(sb, "%s<GenID>%u</GenID>\n", indent, sc->sc_genid); 3026 sbuf_printf(sb, "%s<Flags>", indent); 3027 if (sc->sc_flags == 0) 3028 sbuf_printf(sb, "NONE"); 3029 else { 3030 int first = 1; 3031 3032 #define ADD_FLAG(flag, name) do { \ 3033 if ((sc->sc_flags & (flag)) != 0) { \ 3034 if (!first) \ 3035 sbuf_printf(sb, ", "); \ 3036 else \ 3037 first = 0; \ 3038 sbuf_printf(sb, name); \ 3039 } \ 3040 } while (0) 3041 ADD_FLAG(G_RAID3_DEVICE_FLAG_NOAUTOSYNC, "NOAUTOSYNC"); 3042 ADD_FLAG(G_RAID3_DEVICE_FLAG_ROUND_ROBIN, 3043 "ROUND-ROBIN"); 3044 ADD_FLAG(G_RAID3_DEVICE_FLAG_VERIFY, "VERIFY"); 3045 #undef ADD_FLAG 3046 } 3047 sbuf_printf(sb, "</Flags>\n"); 3048 sbuf_printf(sb, "%s<Components>%u</Components>\n", indent, 3049 sc->sc_ndisks); 3050 sbuf_printf(sb, "%s<State>%s</State>\n", indent, 3051 g_raid3_device_state2str(sc->sc_state)); 3052 } 3053 } 3054 3055 static void 3056 g_raid3_shutdown(void *arg, int howto) 3057 { 3058 struct g_class *mp; 3059 struct g_geom *gp, *gp2; 3060 3061 mp = arg; 3062 DROP_GIANT(); 3063 g_topology_lock(); 3064 LIST_FOREACH_SAFE(gp, &mp->geom, geom, gp2) { 3065 if (gp->softc == NULL) 3066 continue; 3067 g_raid3_destroy(gp->softc, 1); 3068 } 3069 g_topology_unlock(); 3070 PICKUP_GIANT(); 3071 #if 0 3072 tsleep(&gp, PRIBIO, "r3:shutdown", hz * 20); 3073 #endif 3074 } 3075 3076 static void 3077 g_raid3_init(struct g_class *mp) 3078 { 3079 3080 g_raid3_ehtag = EVENTHANDLER_REGISTER(shutdown_post_sync, 3081 g_raid3_shutdown, mp, SHUTDOWN_PRI_FIRST); 3082 if (g_raid3_ehtag == NULL) 3083 G_RAID3_DEBUG(0, "Warning! Cannot register shutdown event."); 3084 } 3085 3086 static void 3087 g_raid3_fini(struct g_class *mp) 3088 { 3089 3090 if (g_raid3_ehtag == NULL) 3091 return; 3092 EVENTHANDLER_DEREGISTER(shutdown_post_sync, g_raid3_ehtag); 3093 } 3094 3095 DECLARE_GEOM_CLASS(g_raid3_class, g_raid3); 3096