geom_disk.c revision 226735
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 226735 2011-10-25 14:04:59Z 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 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 113 mtx_lock(&Giant); 114} 115 116static void __inline 117g_disk_unlock_giant(struct disk *dp) 118{ 119 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 120 mtx_unlock(&Giant); 121} 122 123static int 124g_disk_access(struct g_provider *pp, int r, int w, int e) 125{ 126 struct disk *dp; 127 struct g_disk_softc *sc; 128 int error; 129 130 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)", 131 pp->name, r, w, e); 132 g_topology_assert(); 133 sc = pp->geom->softc; 134 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { 135 /* 136 * Allow decreasing access count even if disk is not 137 * avaliable anymore. 138 */ 139 if (r <= 0 && w <= 0 && e <= 0) 140 return (0); 141 return (ENXIO); 142 } 143 r += pp->acr; 144 w += pp->acw; 145 e += pp->ace; 146 error = 0; 147 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) { 148 if (dp->d_open != NULL) { 149 g_disk_lock_giant(dp); 150 error = dp->d_open(dp); 151 if (bootverbose && error != 0) 152 printf("Opened disk %s -> %d\n", 153 pp->name, error); 154 g_disk_unlock_giant(dp); 155 } 156 pp->mediasize = dp->d_mediasize; 157 pp->sectorsize = dp->d_sectorsize; 158 if (dp->d_flags & DISKFLAG_CANDELETE) 159 pp->flags |= G_PF_CANDELETE; 160 else 161 pp->flags &= ~G_PF_CANDELETE; 162 pp->stripeoffset = dp->d_stripeoffset; 163 pp->stripesize = dp->d_stripesize; 164 dp->d_flags |= DISKFLAG_OPEN; 165 if (dp->d_maxsize == 0) { 166 printf("WARNING: Disk drive %s%d has no d_maxsize\n", 167 dp->d_name, dp->d_unit); 168 dp->d_maxsize = DFLTPHYS; 169 } 170 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) { 171 if (dp->d_close != NULL) { 172 g_disk_lock_giant(dp); 173 error = dp->d_close(dp); 174 if (error != 0) 175 printf("Closed disk %s -> %d\n", 176 pp->name, error); 177 g_disk_unlock_giant(dp); 178 } 179 sc->state = G_STATE_ACTIVE; 180 if (sc->led[0] != 0) 181 led_set(sc->led, "0"); 182 dp->d_flags &= ~DISKFLAG_OPEN; 183 } 184 return (error); 185} 186 187static void 188g_disk_kerneldump(struct bio *bp, struct disk *dp) 189{ 190 struct g_kerneldump *gkd; 191 struct g_geom *gp; 192 193 gkd = (struct g_kerneldump*)bp->bio_data; 194 gp = bp->bio_to->geom; 195 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)", 196 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length); 197 if (dp->d_dump == NULL) { 198 g_io_deliver(bp, ENODEV); 199 return; 200 } 201 gkd->di.dumper = dp->d_dump; 202 gkd->di.priv = dp; 203 gkd->di.blocksize = dp->d_sectorsize; 204 gkd->di.maxiosize = dp->d_maxsize; 205 gkd->di.mediaoffset = gkd->offset; 206 if ((gkd->offset + gkd->length) > dp->d_mediasize) 207 gkd->length = dp->d_mediasize - gkd->offset; 208 gkd->di.mediasize = gkd->length; 209 g_io_deliver(bp, 0); 210} 211 212static void 213g_disk_setstate(struct bio *bp, struct g_disk_softc *sc) 214{ 215 const char *cmd; 216 217 memcpy(&sc->state, bp->bio_data, sizeof(sc->state)); 218 if (sc->led[0] != 0) { 219 switch (sc->state) { 220 case G_STATE_FAILED: 221 cmd = "1"; 222 break; 223 case G_STATE_REBUILD: 224 cmd = "f5"; 225 break; 226 case G_STATE_RESYNC: 227 cmd = "f1"; 228 break; 229 default: 230 cmd = "0"; 231 break; 232 } 233 led_set(sc->led, cmd); 234 } 235 g_io_deliver(bp, 0); 236} 237 238static void 239g_disk_done(struct bio *bp) 240{ 241 struct bio *bp2; 242 struct disk *dp; 243 struct g_disk_softc *sc; 244 245 /* See "notes" for why we need a mutex here */ 246 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */ 247 mtx_lock(&g_disk_done_mtx); 248 bp->bio_completed = bp->bio_length - bp->bio_resid; 249 250 bp2 = bp->bio_parent; 251 if (bp2->bio_error == 0) 252 bp2->bio_error = bp->bio_error; 253 bp2->bio_completed += bp->bio_completed; 254 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) && 255 (sc = bp2->bio_to->geom->softc) && 256 (dp = sc->dp)) { 257 devstat_end_transaction_bio(dp->d_devstat, bp); 258 } 259 g_destroy_bio(bp); 260 bp2->bio_inbed++; 261 if (bp2->bio_children == bp2->bio_inbed) { 262 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed; 263 g_io_deliver(bp2, bp2->bio_error); 264 } 265 mtx_unlock(&g_disk_done_mtx); 266} 267 268static int 269g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td) 270{ 271 struct g_geom *gp; 272 struct disk *dp; 273 struct g_disk_softc *sc; 274 int error; 275 276 gp = pp->geom; 277 sc = gp->softc; 278 dp = sc->dp; 279 280 if (dp->d_ioctl == NULL) 281 return (ENOIOCTL); 282 g_disk_lock_giant(dp); 283 error = dp->d_ioctl(dp, cmd, data, fflag, td); 284 g_disk_unlock_giant(dp); 285 return(error); 286} 287 288static void 289g_disk_start(struct bio *bp) 290{ 291 struct bio *bp2, *bp3; 292 struct disk *dp; 293 struct g_disk_softc *sc; 294 int error; 295 off_t off; 296 297 sc = bp->bio_to->geom->softc; 298 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { 299 g_io_deliver(bp, ENXIO); 300 return; 301 } 302 error = EJUSTRETURN; 303 switch(bp->bio_cmd) { 304 case BIO_DELETE: 305 if (!(dp->d_flags & DISKFLAG_CANDELETE)) { 306 error = 0; 307 break; 308 } 309 /* fall-through */ 310 case BIO_READ: 311 case BIO_WRITE: 312 off = 0; 313 bp3 = NULL; 314 bp2 = g_clone_bio(bp); 315 if (bp2 == NULL) { 316 error = ENOMEM; 317 break; 318 } 319 do { 320 bp2->bio_offset += off; 321 bp2->bio_length -= off; 322 bp2->bio_data += off; 323 if (bp2->bio_length > dp->d_maxsize) { 324 /* 325 * XXX: If we have a stripesize we should really 326 * use it here. 327 */ 328 bp2->bio_length = dp->d_maxsize; 329 off += dp->d_maxsize; 330 /* 331 * To avoid a race, we need to grab the next bio 332 * before we schedule this one. See "notes". 333 */ 334 bp3 = g_clone_bio(bp); 335 if (bp3 == NULL) 336 bp->bio_error = ENOMEM; 337 } 338 bp2->bio_done = g_disk_done; 339 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize; 340 bp2->bio_bcount = bp2->bio_length; 341 bp2->bio_disk = dp; 342 devstat_start_transaction_bio(dp->d_devstat, bp2); 343 g_disk_lock_giant(dp); 344 dp->d_strategy(bp2); 345 g_disk_unlock_giant(dp); 346 bp2 = bp3; 347 bp3 = NULL; 348 } while (bp2 != NULL); 349 break; 350 case BIO_GETATTR: 351 /* Give the driver a chance to override */ 352 if (dp->d_getattr != NULL) { 353 if (bp->bio_disk == NULL) 354 bp->bio_disk = dp; 355 error = dp->d_getattr(bp); 356 if (error != -1) 357 break; 358 error = EJUSTRETURN; 359 } 360 if (g_handleattr_int(bp, "GEOM::candelete", 361 (dp->d_flags & DISKFLAG_CANDELETE) != 0)) 362 break; 363 else if (g_handleattr_int(bp, "GEOM::fwsectors", 364 dp->d_fwsectors)) 365 break; 366 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads)) 367 break; 368 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0)) 369 break; 370 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident)) 371 break; 372 else if (g_handleattr(bp, "GEOM::hba_vendor", 373 &dp->d_hba_vendor, 2)) 374 break; 375 else if (g_handleattr(bp, "GEOM::hba_device", 376 &dp->d_hba_device, 2)) 377 break; 378 else if (g_handleattr(bp, "GEOM::hba_subvendor", 379 &dp->d_hba_subvendor, 2)) 380 break; 381 else if (g_handleattr(bp, "GEOM::hba_subdevice", 382 &dp->d_hba_subdevice, 2)) 383 break; 384 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump")) 385 g_disk_kerneldump(bp, dp); 386 else if (!strcmp(bp->bio_attribute, "GEOM::setstate")) 387 g_disk_setstate(bp, sc); 388 else 389 error = ENOIOCTL; 390 break; 391 case BIO_FLUSH: 392 g_trace(G_T_TOPOLOGY, "g_disk_flushcache(%s)", 393 bp->bio_to->name); 394 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) { 395 g_io_deliver(bp, ENODEV); 396 return; 397 } 398 bp2 = g_clone_bio(bp); 399 if (bp2 == NULL) { 400 g_io_deliver(bp, ENOMEM); 401 return; 402 } 403 bp2->bio_done = g_disk_done; 404 bp2->bio_disk = dp; 405 g_disk_lock_giant(dp); 406 dp->d_strategy(bp2); 407 g_disk_unlock_giant(dp); 408 break; 409 default: 410 error = EOPNOTSUPP; 411 break; 412 } 413 if (error != EJUSTRETURN) 414 g_io_deliver(bp, error); 415 return; 416} 417 418static void 419g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) 420{ 421 struct disk *dp; 422 struct g_disk_softc *sc; 423 424 sc = gp->softc; 425 if (sc == NULL || (dp = sc->dp) == NULL) 426 return; 427 if (indent == NULL) { 428 sbuf_printf(sb, " hd %u", dp->d_fwheads); 429 sbuf_printf(sb, " sc %u", dp->d_fwsectors); 430 return; 431 } 432 if (pp != NULL) { 433 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n", 434 indent, dp->d_fwheads); 435 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n", 436 indent, dp->d_fwsectors); 437 sbuf_printf(sb, "%s<ident>%s</ident>\n", indent, dp->d_ident); 438 sbuf_printf(sb, "%s<descr>%s</descr>\n", indent, dp->d_descr); 439 } 440} 441 442static void 443g_disk_create(void *arg, int flag) 444{ 445 struct g_geom *gp; 446 struct g_provider *pp; 447 struct disk *dp; 448 struct g_disk_softc *sc; 449 char tmpstr[80]; 450 451 if (flag == EV_CANCEL) 452 return; 453 g_topology_assert(); 454 dp = arg; 455 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO); 456 sc->dp = dp; 457 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit); 458 gp->softc = sc; 459 pp = g_new_providerf(gp, "%s", gp->name); 460 pp->mediasize = dp->d_mediasize; 461 pp->sectorsize = dp->d_sectorsize; 462 if (dp->d_flags & DISKFLAG_CANDELETE) 463 pp->flags |= G_PF_CANDELETE; 464 pp->stripeoffset = dp->d_stripeoffset; 465 pp->stripesize = dp->d_stripesize; 466 if (bootverbose) 467 printf("GEOM: new disk %s\n", gp->name); 468 sysctl_ctx_init(&sc->sysctl_ctx); 469 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name); 470 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx, 471 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name, 472 CTLFLAG_RD, 0, tmpstr); 473 if (sc->sysctl_tree != NULL) { 474 snprintf(tmpstr, sizeof(tmpstr), 475 "kern.geom.disk.%s.led", gp->name); 476 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led)); 477 SYSCTL_ADD_STRING(&sc->sysctl_ctx, 478 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led", 479 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led), 480 "LED name"); 481 } 482 pp->private = sc; 483 dp->d_geom = gp; 484 g_error_provider(pp, 0); 485} 486 487static void 488g_disk_destroy(void *ptr, int flag) 489{ 490 struct disk *dp; 491 struct g_geom *gp; 492 struct g_disk_softc *sc; 493 494 g_topology_assert(); 495 dp = ptr; 496 gp = dp->d_geom; 497 if (gp != NULL) { 498 sc = gp->softc; 499 if (sc->sysctl_tree != NULL) { 500 sysctl_ctx_free(&sc->sysctl_ctx); 501 sc->sysctl_tree = NULL; 502 } 503 if (sc->led[0] != 0) { 504 led_set(sc->led, "0"); 505 sc->led[0] = 0; 506 } 507 g_free(sc); 508 gp->softc = NULL; 509 g_wither_geom(gp, ENXIO); 510 } 511 g_free(dp); 512} 513 514/* 515 * We only allow printable characters in disk ident, 516 * the rest is converted to 'x<HH>'. 517 */ 518static void 519g_disk_ident_adjust(char *ident, size_t size) 520{ 521 char *p, tmp[4], newid[DISK_IDENT_SIZE]; 522 523 newid[0] = '\0'; 524 for (p = ident; *p != '\0'; p++) { 525 if (isprint(*p)) { 526 tmp[0] = *p; 527 tmp[1] = '\0'; 528 } else { 529 snprintf(tmp, sizeof(tmp), "x%02hhx", 530 *(unsigned char *)p); 531 } 532 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid)) 533 break; 534 } 535 bzero(ident, size); 536 strlcpy(ident, newid, size); 537} 538 539struct disk * 540disk_alloc(void) 541{ 542 543 return (g_malloc(sizeof(struct disk), M_WAITOK | M_ZERO)); 544} 545 546void 547disk_create(struct disk *dp, int version) 548{ 549 if (version != DISK_VERSION_00 && version != DISK_VERSION_01) { 550 printf("WARNING: Attempt to add disk %s%d %s", 551 dp->d_name, dp->d_unit, 552 " using incompatible ABI version of disk(9)\n"); 553 printf("WARNING: Ignoring disk %s%d\n", 554 dp->d_name, dp->d_unit); 555 return; 556 } 557 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy")); 558 KASSERT(dp->d_name != NULL, ("disk_create need d_name")); 559 KASSERT(*dp->d_name != 0, ("disk_create need d_name")); 560 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long")); 561 if (dp->d_devstat == NULL) 562 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit, 563 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED, 564 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); 565 dp->d_geom = NULL; 566 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident)); 567 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL); 568} 569 570void 571disk_destroy(struct disk *dp) 572{ 573 574 g_cancel_event(dp); 575 dp->d_destroyed = 1; 576 if (dp->d_devstat != NULL) 577 devstat_remove_entry(dp->d_devstat); 578 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL); 579} 580 581void 582disk_gone(struct disk *dp) 583{ 584 struct g_geom *gp; 585 struct g_provider *pp; 586 587 gp = dp->d_geom; 588 if (gp != NULL) 589 LIST_FOREACH(pp, &gp->provider, provider) 590 g_wither_provider(pp, ENXIO); 591} 592 593void 594disk_attr_changed(struct disk *dp, const char *attr, int flag) 595{ 596 struct g_geom *gp; 597 struct g_provider *pp; 598 599 gp = dp->d_geom; 600 if (gp != NULL) 601 LIST_FOREACH(pp, &gp->provider, provider) 602 (void)g_attr_changed(pp, attr, flag); 603} 604 605static void 606g_kern_disks(void *p, int flag __unused) 607{ 608 struct sbuf *sb; 609 struct g_geom *gp; 610 char *sp; 611 612 sb = p; 613 sp = ""; 614 g_topology_assert(); 615 LIST_FOREACH(gp, &g_disk_class.geom, geom) { 616 sbuf_printf(sb, "%s%s", sp, gp->name); 617 sp = " "; 618 } 619 sbuf_finish(sb); 620} 621 622static int 623sysctl_disks(SYSCTL_HANDLER_ARGS) 624{ 625 int error; 626 struct sbuf *sb; 627 628 sb = sbuf_new_auto(); 629 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL); 630 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); 631 sbuf_delete(sb); 632 return error; 633} 634 635SYSCTL_PROC(_kern, OID_AUTO, disks, 636 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 637 sysctl_disks, "A", "names of available disks"); 638 639