geom_disk.c revision 226737
1/*- 2 * Copyright (c) 2002 Poul-Henning Kamp 3 * Copyright (c) 2002 Networks Associates Technology, Inc. 4 * All rights reserved. 5 * 6 * This software was developed for the FreeBSD Project by Poul-Henning Kamp 7 * and NAI Labs, the Security Research Division of Network Associates, Inc. 8 * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the 9 * DARPA CHATS research program. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. The names of the authors may not be used to endorse or promote 20 * products derived from this software without specific prior written 21 * permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 */ 35 36#include <sys/cdefs.h> 37__FBSDID("$FreeBSD: head/sys/geom/geom_disk.c 226737 2011-10-25 14:07:17Z pjd $"); 38 39#include "opt_geom.h" 40 41#include <sys/param.h> 42#include <sys/systm.h> 43#include <sys/kernel.h> 44#include <sys/sysctl.h> 45#include <sys/bio.h> 46#include <sys/ctype.h> 47#include <sys/fcntl.h> 48#include <sys/malloc.h> 49#include <sys/sbuf.h> 50#include <sys/sysctl.h> 51#include <sys/devicestat.h> 52#include <machine/md_var.h> 53 54#include <sys/lock.h> 55#include <sys/mutex.h> 56#include <geom/geom.h> 57#include <geom/geom_disk.h> 58#include <geom/geom_int.h> 59 60#include <dev/led/led.h> 61 62struct g_disk_softc { 63 struct disk *dp; 64 struct sysctl_ctx_list sysctl_ctx; 65 struct sysctl_oid *sysctl_tree; 66 char led[64]; 67 uint32_t state; 68}; 69 70static struct mtx g_disk_done_mtx; 71 72static g_access_t g_disk_access; 73static g_init_t g_disk_init; 74static g_fini_t g_disk_fini; 75static g_start_t g_disk_start; 76static g_ioctl_t g_disk_ioctl; 77static g_dumpconf_t g_disk_dumpconf; 78 79static struct g_class g_disk_class = { 80 .name = "DISK", 81 .version = G_VERSION, 82 .init = g_disk_init, 83 .fini = g_disk_fini, 84 .start = g_disk_start, 85 .access = g_disk_access, 86 .ioctl = g_disk_ioctl, 87 .dumpconf = g_disk_dumpconf, 88}; 89 90SYSCTL_DECL(_kern_geom); 91SYSCTL_NODE(_kern_geom, OID_AUTO, disk, CTLFLAG_RW, 0, "GEOM_DISK stuff"); 92 93static void 94g_disk_init(struct g_class *mp __unused) 95{ 96 97 mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF); 98} 99 100static void 101g_disk_fini(struct g_class *mp __unused) 102{ 103 104 mtx_destroy(&g_disk_done_mtx); 105} 106 107DECLARE_GEOM_CLASS(g_disk_class, g_disk); 108 109static void __inline 110g_disk_lock_giant(struct disk *dp) 111{ 112 113 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 114 mtx_lock(&Giant); 115} 116 117static void __inline 118g_disk_unlock_giant(struct disk *dp) 119{ 120 121 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 122 mtx_unlock(&Giant); 123} 124 125static int 126g_disk_access(struct g_provider *pp, int r, int w, int e) 127{ 128 struct disk *dp; 129 struct g_disk_softc *sc; 130 int error; 131 132 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)", 133 pp->name, r, w, e); 134 g_topology_assert(); 135 sc = pp->geom->softc; 136 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { 137 /* 138 * Allow decreasing access count even if disk is not 139 * avaliable anymore. 140 */ 141 if (r <= 0 && w <= 0 && e <= 0) 142 return (0); 143 return (ENXIO); 144 } 145 r += pp->acr; 146 w += pp->acw; 147 e += pp->ace; 148 error = 0; 149 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) { 150 if (dp->d_open != NULL) { 151 g_disk_lock_giant(dp); 152 error = dp->d_open(dp); 153 if (bootverbose && error != 0) 154 printf("Opened disk %s -> %d\n", 155 pp->name, error); 156 g_disk_unlock_giant(dp); 157 } 158 pp->mediasize = dp->d_mediasize; 159 pp->sectorsize = dp->d_sectorsize; 160 if (dp->d_flags & DISKFLAG_CANDELETE) 161 pp->flags |= G_PF_CANDELETE; 162 else 163 pp->flags &= ~G_PF_CANDELETE; 164 pp->stripeoffset = dp->d_stripeoffset; 165 pp->stripesize = dp->d_stripesize; 166 dp->d_flags |= DISKFLAG_OPEN; 167 if (dp->d_maxsize == 0) { 168 printf("WARNING: Disk drive %s%d has no d_maxsize\n", 169 dp->d_name, dp->d_unit); 170 dp->d_maxsize = DFLTPHYS; 171 } 172 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) { 173 if (dp->d_close != NULL) { 174 g_disk_lock_giant(dp); 175 error = dp->d_close(dp); 176 if (error != 0) 177 printf("Closed disk %s -> %d\n", 178 pp->name, error); 179 g_disk_unlock_giant(dp); 180 } 181 sc->state = G_STATE_ACTIVE; 182 if (sc->led[0] != 0) 183 led_set(sc->led, "0"); 184 dp->d_flags &= ~DISKFLAG_OPEN; 185 } 186 return (error); 187} 188 189static void 190g_disk_kerneldump(struct bio *bp, struct disk *dp) 191{ 192 struct g_kerneldump *gkd; 193 struct g_geom *gp; 194 195 gkd = (struct g_kerneldump*)bp->bio_data; 196 gp = bp->bio_to->geom; 197 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)", 198 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length); 199 if (dp->d_dump == NULL) { 200 g_io_deliver(bp, ENODEV); 201 return; 202 } 203 gkd->di.dumper = dp->d_dump; 204 gkd->di.priv = dp; 205 gkd->di.blocksize = dp->d_sectorsize; 206 gkd->di.maxiosize = dp->d_maxsize; 207 gkd->di.mediaoffset = gkd->offset; 208 if ((gkd->offset + gkd->length) > dp->d_mediasize) 209 gkd->length = dp->d_mediasize - gkd->offset; 210 gkd->di.mediasize = gkd->length; 211 g_io_deliver(bp, 0); 212} 213 214static void 215g_disk_setstate(struct bio *bp, struct g_disk_softc *sc) 216{ 217 const char *cmd; 218 219 memcpy(&sc->state, bp->bio_data, sizeof(sc->state)); 220 if (sc->led[0] != 0) { 221 switch (sc->state) { 222 case G_STATE_FAILED: 223 cmd = "1"; 224 break; 225 case G_STATE_REBUILD: 226 cmd = "f5"; 227 break; 228 case G_STATE_RESYNC: 229 cmd = "f1"; 230 break; 231 default: 232 cmd = "0"; 233 break; 234 } 235 led_set(sc->led, cmd); 236 } 237 g_io_deliver(bp, 0); 238} 239 240static void 241g_disk_done(struct bio *bp) 242{ 243 struct bio *bp2; 244 struct disk *dp; 245 struct g_disk_softc *sc; 246 247 /* See "notes" for why we need a mutex here */ 248 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */ 249 mtx_lock(&g_disk_done_mtx); 250 bp->bio_completed = bp->bio_length - bp->bio_resid; 251 252 bp2 = bp->bio_parent; 253 if (bp2->bio_error == 0) 254 bp2->bio_error = bp->bio_error; 255 bp2->bio_completed += bp->bio_completed; 256 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) != 0 && 257 (sc = bp2->bio_to->geom->softc) != NULL && 258 (dp = sc->dp) != NULL) { 259 devstat_end_transaction_bio(dp->d_devstat, bp); 260 } 261 g_destroy_bio(bp); 262 bp2->bio_inbed++; 263 if (bp2->bio_children == bp2->bio_inbed) { 264 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed; 265 g_io_deliver(bp2, bp2->bio_error); 266 } 267 mtx_unlock(&g_disk_done_mtx); 268} 269 270static int 271g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td) 272{ 273 struct g_geom *gp; 274 struct disk *dp; 275 struct g_disk_softc *sc; 276 int error; 277 278 gp = pp->geom; 279 sc = gp->softc; 280 dp = sc->dp; 281 282 if (dp->d_ioctl == NULL) 283 return (ENOIOCTL); 284 g_disk_lock_giant(dp); 285 error = dp->d_ioctl(dp, cmd, data, fflag, td); 286 g_disk_unlock_giant(dp); 287 return (error); 288} 289 290static void 291g_disk_start(struct bio *bp) 292{ 293 struct bio *bp2, *bp3; 294 struct disk *dp; 295 struct g_disk_softc *sc; 296 int error; 297 off_t off; 298 299 sc = bp->bio_to->geom->softc; 300 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { 301 g_io_deliver(bp, ENXIO); 302 return; 303 } 304 error = EJUSTRETURN; 305 switch(bp->bio_cmd) { 306 case BIO_DELETE: 307 if (!(dp->d_flags & DISKFLAG_CANDELETE)) { 308 error = EOPNOTSUPP; 309 break; 310 } 311 /* fall-through */ 312 case BIO_READ: 313 case BIO_WRITE: 314 off = 0; 315 bp3 = NULL; 316 bp2 = g_clone_bio(bp); 317 if (bp2 == NULL) { 318 error = ENOMEM; 319 break; 320 } 321 do { 322 bp2->bio_offset += off; 323 bp2->bio_length -= off; 324 bp2->bio_data += off; 325 if (bp2->bio_length > dp->d_maxsize) { 326 /* 327 * XXX: If we have a stripesize we should really 328 * use it here. 329 */ 330 bp2->bio_length = dp->d_maxsize; 331 off += dp->d_maxsize; 332 /* 333 * To avoid a race, we need to grab the next bio 334 * before we schedule this one. See "notes". 335 */ 336 bp3 = g_clone_bio(bp); 337 if (bp3 == NULL) 338 bp->bio_error = ENOMEM; 339 } 340 bp2->bio_done = g_disk_done; 341 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize; 342 bp2->bio_bcount = bp2->bio_length; 343 bp2->bio_disk = dp; 344 devstat_start_transaction_bio(dp->d_devstat, bp2); 345 g_disk_lock_giant(dp); 346 dp->d_strategy(bp2); 347 g_disk_unlock_giant(dp); 348 bp2 = bp3; 349 bp3 = NULL; 350 } while (bp2 != NULL); 351 break; 352 case BIO_GETATTR: 353 /* Give the driver a chance to override */ 354 if (dp->d_getattr != NULL) { 355 if (bp->bio_disk == NULL) 356 bp->bio_disk = dp; 357 error = dp->d_getattr(bp); 358 if (error != -1) 359 break; 360 error = EJUSTRETURN; 361 } 362 if (g_handleattr_int(bp, "GEOM::candelete", 363 (dp->d_flags & DISKFLAG_CANDELETE) != 0)) 364 break; 365 else if (g_handleattr_int(bp, "GEOM::fwsectors", 366 dp->d_fwsectors)) 367 break; 368 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads)) 369 break; 370 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0)) 371 break; 372 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident)) 373 break; 374 else if (g_handleattr(bp, "GEOM::hba_vendor", 375 &dp->d_hba_vendor, 2)) 376 break; 377 else if (g_handleattr(bp, "GEOM::hba_device", 378 &dp->d_hba_device, 2)) 379 break; 380 else if (g_handleattr(bp, "GEOM::hba_subvendor", 381 &dp->d_hba_subvendor, 2)) 382 break; 383 else if (g_handleattr(bp, "GEOM::hba_subdevice", 384 &dp->d_hba_subdevice, 2)) 385 break; 386 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump")) 387 g_disk_kerneldump(bp, dp); 388 else if (!strcmp(bp->bio_attribute, "GEOM::setstate")) 389 g_disk_setstate(bp, sc); 390 else 391 error = ENOIOCTL; 392 break; 393 case BIO_FLUSH: 394 g_trace(G_T_TOPOLOGY, "g_disk_flushcache(%s)", 395 bp->bio_to->name); 396 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) { 397 error = EOPNOTSUPP; 398 break; 399 } 400 bp2 = g_clone_bio(bp); 401 if (bp2 == NULL) { 402 g_io_deliver(bp, ENOMEM); 403 return; 404 } 405 bp2->bio_done = g_disk_done; 406 bp2->bio_disk = dp; 407 g_disk_lock_giant(dp); 408 dp->d_strategy(bp2); 409 g_disk_unlock_giant(dp); 410 break; 411 default: 412 error = EOPNOTSUPP; 413 break; 414 } 415 if (error != EJUSTRETURN) 416 g_io_deliver(bp, error); 417 return; 418} 419 420static void 421g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) 422{ 423 struct disk *dp; 424 struct g_disk_softc *sc; 425 426 sc = gp->softc; 427 if (sc == NULL || (dp = sc->dp) == NULL) 428 return; 429 if (indent == NULL) { 430 sbuf_printf(sb, " hd %u", dp->d_fwheads); 431 sbuf_printf(sb, " sc %u", dp->d_fwsectors); 432 return; 433 } 434 if (pp != NULL) { 435 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n", 436 indent, dp->d_fwheads); 437 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n", 438 indent, dp->d_fwsectors); 439 sbuf_printf(sb, "%s<ident>%s</ident>\n", indent, dp->d_ident); 440 sbuf_printf(sb, "%s<descr>%s</descr>\n", indent, dp->d_descr); 441 } 442} 443 444static void 445g_disk_create(void *arg, int flag) 446{ 447 struct g_geom *gp; 448 struct g_provider *pp; 449 struct disk *dp; 450 struct g_disk_softc *sc; 451 char tmpstr[80]; 452 453 if (flag == EV_CANCEL) 454 return; 455 g_topology_assert(); 456 dp = arg; 457 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO); 458 sc->dp = dp; 459 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit); 460 gp->softc = sc; 461 pp = g_new_providerf(gp, "%s", gp->name); 462 pp->mediasize = dp->d_mediasize; 463 pp->sectorsize = dp->d_sectorsize; 464 if (dp->d_flags & DISKFLAG_CANDELETE) 465 pp->flags |= G_PF_CANDELETE; 466 pp->stripeoffset = dp->d_stripeoffset; 467 pp->stripesize = dp->d_stripesize; 468 if (bootverbose) 469 printf("GEOM: new disk %s\n", gp->name); 470 sysctl_ctx_init(&sc->sysctl_ctx); 471 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name); 472 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx, 473 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name, 474 CTLFLAG_RD, 0, tmpstr); 475 if (sc->sysctl_tree != NULL) { 476 snprintf(tmpstr, sizeof(tmpstr), 477 "kern.geom.disk.%s.led", gp->name); 478 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led)); 479 SYSCTL_ADD_STRING(&sc->sysctl_ctx, 480 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led", 481 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led), 482 "LED name"); 483 } 484 pp->private = sc; 485 dp->d_geom = gp; 486 g_error_provider(pp, 0); 487} 488 489static void 490g_disk_destroy(void *ptr, int flag) 491{ 492 struct disk *dp; 493 struct g_geom *gp; 494 struct g_disk_softc *sc; 495 496 g_topology_assert(); 497 dp = ptr; 498 gp = dp->d_geom; 499 if (gp != NULL) { 500 sc = gp->softc; 501 if (sc->sysctl_tree != NULL) { 502 sysctl_ctx_free(&sc->sysctl_ctx); 503 sc->sysctl_tree = NULL; 504 } 505 if (sc->led[0] != 0) { 506 led_set(sc->led, "0"); 507 sc->led[0] = 0; 508 } 509 g_free(sc); 510 gp->softc = NULL; 511 g_wither_geom(gp, ENXIO); 512 } 513 g_free(dp); 514} 515 516/* 517 * We only allow printable characters in disk ident, 518 * the rest is converted to 'x<HH>'. 519 */ 520static void 521g_disk_ident_adjust(char *ident, size_t size) 522{ 523 char *p, tmp[4], newid[DISK_IDENT_SIZE]; 524 525 newid[0] = '\0'; 526 for (p = ident; *p != '\0'; p++) { 527 if (isprint(*p)) { 528 tmp[0] = *p; 529 tmp[1] = '\0'; 530 } else { 531 snprintf(tmp, sizeof(tmp), "x%02hhx", 532 *(unsigned char *)p); 533 } 534 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid)) 535 break; 536 } 537 bzero(ident, size); 538 strlcpy(ident, newid, size); 539} 540 541struct disk * 542disk_alloc(void) 543{ 544 545 return (g_malloc(sizeof(struct disk), M_WAITOK | M_ZERO)); 546} 547 548void 549disk_create(struct disk *dp, int version) 550{ 551 552 if (version != DISK_VERSION_00 && version != DISK_VERSION_01) { 553 printf("WARNING: Attempt to add disk %s%d %s", 554 dp->d_name, dp->d_unit, 555 " using incompatible ABI version of disk(9)\n"); 556 printf("WARNING: Ignoring disk %s%d\n", 557 dp->d_name, dp->d_unit); 558 return; 559 } 560 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy")); 561 KASSERT(dp->d_name != NULL, ("disk_create need d_name")); 562 KASSERT(*dp->d_name != 0, ("disk_create need d_name")); 563 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long")); 564 if (dp->d_devstat == NULL) 565 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit, 566 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED, 567 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); 568 dp->d_geom = NULL; 569 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident)); 570 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL); 571} 572 573void 574disk_destroy(struct disk *dp) 575{ 576 577 g_cancel_event(dp); 578 dp->d_destroyed = 1; 579 if (dp->d_devstat != NULL) 580 devstat_remove_entry(dp->d_devstat); 581 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL); 582} 583 584void 585disk_gone(struct disk *dp) 586{ 587 struct g_geom *gp; 588 struct g_provider *pp; 589 590 gp = dp->d_geom; 591 if (gp != NULL) 592 LIST_FOREACH(pp, &gp->provider, provider) 593 g_wither_provider(pp, ENXIO); 594} 595 596void 597disk_attr_changed(struct disk *dp, const char *attr, int flag) 598{ 599 struct g_geom *gp; 600 struct g_provider *pp; 601 602 gp = dp->d_geom; 603 if (gp != NULL) 604 LIST_FOREACH(pp, &gp->provider, provider) 605 (void)g_attr_changed(pp, attr, flag); 606} 607 608static void 609g_kern_disks(void *p, int flag __unused) 610{ 611 struct sbuf *sb; 612 struct g_geom *gp; 613 char *sp; 614 615 sb = p; 616 sp = ""; 617 g_topology_assert(); 618 LIST_FOREACH(gp, &g_disk_class.geom, geom) { 619 sbuf_printf(sb, "%s%s", sp, gp->name); 620 sp = " "; 621 } 622 sbuf_finish(sb); 623} 624 625static int 626sysctl_disks(SYSCTL_HANDLER_ARGS) 627{ 628 int error; 629 struct sbuf *sb; 630 631 sb = sbuf_new_auto(); 632 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL); 633 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); 634 sbuf_delete(sb); 635 return error; 636} 637 638SYSCTL_PROC(_kern, OID_AUTO, disks, 639 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 640 sysctl_disks, "A", "names of available disks"); 641