geom_disk.c revision 222652
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 222652 2011-06-03 13:49:18Z mav $"); 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/sysctl.h> 50#include <sys/devicestat.h> 51#include <machine/md_var.h> 52 53#include <sys/lock.h> 54#include <sys/mutex.h> 55#include <geom/geom.h> 56#include <geom/geom_disk.h> 57#include <geom/geom_int.h> 58 59#include <dev/led/led.h> 60 61struct g_disk_softc { 62 struct disk *dp; 63 struct sysctl_ctx_list sysctl_ctx; 64 struct sysctl_oid *sysctl_tree; 65 char led[64]; 66 uint32_t state; 67}; 68 69static struct mtx g_disk_done_mtx; 70 71static g_access_t g_disk_access; 72static g_init_t g_disk_init; 73static g_fini_t g_disk_fini; 74static g_start_t g_disk_start; 75static g_ioctl_t g_disk_ioctl; 76static g_dumpconf_t g_disk_dumpconf; 77 78static struct g_class g_disk_class = { 79 .name = "DISK", 80 .version = G_VERSION, 81 .init = g_disk_init, 82 .fini = g_disk_fini, 83 .start = g_disk_start, 84 .access = g_disk_access, 85 .ioctl = g_disk_ioctl, 86 .dumpconf = g_disk_dumpconf, 87}; 88 89SYSCTL_DECL(_kern_geom); 90SYSCTL_NODE(_kern_geom, OID_AUTO, disk, CTLFLAG_RW, 0, "GEOM_DISK stuff"); 91 92static void 93g_disk_init(struct g_class *mp __unused) 94{ 95 96 mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF); 97} 98 99static void 100g_disk_fini(struct g_class *mp __unused) 101{ 102 103 mtx_destroy(&g_disk_done_mtx); 104} 105 106DECLARE_GEOM_CLASS(g_disk_class, g_disk); 107 108static void __inline 109g_disk_lock_giant(struct disk *dp) 110{ 111 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 112 mtx_lock(&Giant); 113} 114 115static void __inline 116g_disk_unlock_giant(struct disk *dp) 117{ 118 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 119 mtx_unlock(&Giant); 120} 121 122static int 123g_disk_access(struct g_provider *pp, int r, int w, int e) 124{ 125 struct disk *dp; 126 struct g_disk_softc *sc; 127 int error; 128 129 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)", 130 pp->name, r, w, e); 131 g_topology_assert(); 132 sc = pp->geom->softc; 133 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { 134 /* 135 * Allow decreasing access count even if disk is not 136 * avaliable anymore. 137 */ 138 if (r <= 0 && w <= 0 && e <= 0) 139 return (0); 140 return (ENXIO); 141 } 142 r += pp->acr; 143 w += pp->acw; 144 e += pp->ace; 145 error = 0; 146 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) { 147 if (dp->d_open != NULL) { 148 g_disk_lock_giant(dp); 149 error = dp->d_open(dp); 150 if (bootverbose && error != 0) 151 printf("Opened disk %s -> %d\n", 152 pp->name, error); 153 g_disk_unlock_giant(dp); 154 } 155 pp->mediasize = dp->d_mediasize; 156 pp->sectorsize = dp->d_sectorsize; 157 if (dp->d_flags & DISKFLAG_CANDELETE) 158 pp->flags |= G_PF_CANDELETE; 159 else 160 pp->flags &= ~G_PF_CANDELETE; 161 pp->stripeoffset = dp->d_stripeoffset; 162 pp->stripesize = dp->d_stripesize; 163 dp->d_flags |= DISKFLAG_OPEN; 164 if (dp->d_maxsize == 0) { 165 printf("WARNING: Disk drive %s%d has no d_maxsize\n", 166 dp->d_name, dp->d_unit); 167 dp->d_maxsize = DFLTPHYS; 168 } 169 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) { 170 if (dp->d_close != NULL) { 171 g_disk_lock_giant(dp); 172 error = dp->d_close(dp); 173 if (error != 0) 174 printf("Closed disk %s -> %d\n", 175 pp->name, error); 176 g_disk_unlock_giant(dp); 177 } 178 sc->state = G_STATE_ACTIVE; 179 if (sc->led[0] != 0) 180 led_set(sc->led, "0"); 181 dp->d_flags &= ~DISKFLAG_OPEN; 182 } 183 return (error); 184} 185 186static void 187g_disk_kerneldump(struct bio *bp, struct disk *dp) 188{ 189 struct g_kerneldump *gkd; 190 struct g_geom *gp; 191 192 gkd = (struct g_kerneldump*)bp->bio_data; 193 gp = bp->bio_to->geom; 194 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)", 195 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length); 196 if (dp->d_dump == NULL) { 197 g_io_deliver(bp, ENODEV); 198 return; 199 } 200 gkd->di.dumper = dp->d_dump; 201 gkd->di.priv = dp; 202 gkd->di.blocksize = dp->d_sectorsize; 203 gkd->di.maxiosize = dp->d_maxsize; 204 gkd->di.mediaoffset = gkd->offset; 205 if ((gkd->offset + gkd->length) > dp->d_mediasize) 206 gkd->length = dp->d_mediasize - gkd->offset; 207 gkd->di.mediasize = gkd->length; 208 g_io_deliver(bp, 0); 209} 210 211static void 212g_disk_setstate(struct bio *bp, struct g_disk_softc *sc) 213{ 214 const char *cmd; 215 216 memcpy(&sc->state, bp->bio_data, sizeof(sc->state)); 217 if (sc->led[0] != 0) { 218 switch (sc->state) { 219 case G_STATE_FAILED: 220 cmd = "1"; 221 break; 222 case G_STATE_REBUILD: 223 cmd = "f5"; 224 break; 225 case G_STATE_RESYNC: 226 cmd = "f1"; 227 break; 228 default: 229 cmd = "0"; 230 break; 231 } 232 led_set(sc->led, cmd); 233 } 234 g_io_deliver(bp, 0); 235} 236 237static void 238g_disk_done(struct bio *bp) 239{ 240 struct bio *bp2; 241 struct disk *dp; 242 struct g_disk_softc *sc; 243 244 /* See "notes" for why we need a mutex here */ 245 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */ 246 mtx_lock(&g_disk_done_mtx); 247 bp->bio_completed = bp->bio_length - bp->bio_resid; 248 249 bp2 = bp->bio_parent; 250 if (bp2->bio_error == 0) 251 bp2->bio_error = bp->bio_error; 252 bp2->bio_completed += bp->bio_completed; 253 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) && 254 (sc = bp2->bio_to->geom->softc) && 255 (dp = sc->dp)) { 256 devstat_end_transaction_bio(dp->d_devstat, bp); 257 } 258 g_destroy_bio(bp); 259 bp2->bio_inbed++; 260 if (bp2->bio_children == bp2->bio_inbed) { 261 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed; 262 g_io_deliver(bp2, bp2->bio_error); 263 } 264 mtx_unlock(&g_disk_done_mtx); 265} 266 267static int 268g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td) 269{ 270 struct g_geom *gp; 271 struct disk *dp; 272 struct g_disk_softc *sc; 273 int error; 274 275 gp = pp->geom; 276 sc = gp->softc; 277 dp = sc->dp; 278 279 if (dp->d_ioctl == NULL) 280 return (ENOIOCTL); 281 g_disk_lock_giant(dp); 282 error = dp->d_ioctl(dp, cmd, data, fflag, td); 283 g_disk_unlock_giant(dp); 284 return(error); 285} 286 287static void 288g_disk_start(struct bio *bp) 289{ 290 struct bio *bp2, *bp3; 291 struct disk *dp; 292 struct g_disk_softc *sc; 293 int error; 294 off_t off; 295 296 sc = bp->bio_to->geom->softc; 297 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { 298 g_io_deliver(bp, ENXIO); 299 return; 300 } 301 error = EJUSTRETURN; 302 switch(bp->bio_cmd) { 303 case BIO_DELETE: 304 if (!(dp->d_flags & DISKFLAG_CANDELETE)) { 305 error = 0; 306 break; 307 } 308 /* fall-through */ 309 case BIO_READ: 310 case BIO_WRITE: 311 off = 0; 312 bp3 = NULL; 313 bp2 = g_clone_bio(bp); 314 if (bp2 == NULL) { 315 error = ENOMEM; 316 break; 317 } 318 do { 319 bp2->bio_offset += off; 320 bp2->bio_length -= off; 321 bp2->bio_data += off; 322 if (bp2->bio_length > dp->d_maxsize) { 323 /* 324 * XXX: If we have a stripesize we should really 325 * use it here. 326 */ 327 bp2->bio_length = dp->d_maxsize; 328 off += dp->d_maxsize; 329 /* 330 * To avoid a race, we need to grab the next bio 331 * before we schedule this one. See "notes". 332 */ 333 bp3 = g_clone_bio(bp); 334 if (bp3 == NULL) 335 bp->bio_error = ENOMEM; 336 } 337 bp2->bio_done = g_disk_done; 338 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize; 339 bp2->bio_bcount = bp2->bio_length; 340 bp2->bio_disk = dp; 341 devstat_start_transaction_bio(dp->d_devstat, bp2); 342 g_disk_lock_giant(dp); 343 dp->d_strategy(bp2); 344 g_disk_unlock_giant(dp); 345 bp2 = bp3; 346 bp3 = NULL; 347 } while (bp2 != NULL); 348 break; 349 case BIO_GETATTR: 350 if (g_handleattr_int(bp, "GEOM::candelete", 351 (dp->d_flags & DISKFLAG_CANDELETE) != 0)) 352 break; 353 else if (g_handleattr_int(bp, "GEOM::fwsectors", 354 dp->d_fwsectors)) 355 break; 356 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads)) 357 break; 358 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0)) 359 break; 360 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident)) 361 break; 362 else if (g_handleattr(bp, "GEOM::hba_vendor", 363 &dp->d_hba_vendor, 2)) 364 break; 365 else if (g_handleattr(bp, "GEOM::hba_device", 366 &dp->d_hba_device, 2)) 367 break; 368 else if (g_handleattr(bp, "GEOM::hba_subvendor", 369 &dp->d_hba_subvendor, 2)) 370 break; 371 else if (g_handleattr(bp, "GEOM::hba_subdevice", 372 &dp->d_hba_subdevice, 2)) 373 break; 374 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump")) 375 g_disk_kerneldump(bp, dp); 376 else if (!strcmp(bp->bio_attribute, "GEOM::setstate")) 377 g_disk_setstate(bp, sc); 378 else 379 error = ENOIOCTL; 380 break; 381 case BIO_FLUSH: 382 g_trace(G_T_TOPOLOGY, "g_disk_flushcache(%s)", 383 bp->bio_to->name); 384 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) { 385 g_io_deliver(bp, ENODEV); 386 return; 387 } 388 bp2 = g_clone_bio(bp); 389 if (bp2 == NULL) { 390 g_io_deliver(bp, ENOMEM); 391 return; 392 } 393 bp2->bio_done = g_disk_done; 394 bp2->bio_disk = dp; 395 g_disk_lock_giant(dp); 396 dp->d_strategy(bp2); 397 g_disk_unlock_giant(dp); 398 break; 399 default: 400 error = EOPNOTSUPP; 401 break; 402 } 403 if (error != EJUSTRETURN) 404 g_io_deliver(bp, error); 405 return; 406} 407 408static void 409g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) 410{ 411 struct disk *dp; 412 struct g_disk_softc *sc; 413 414 sc = gp->softc; 415 if (sc == NULL || (dp = sc->dp) == NULL) 416 return; 417 if (indent == NULL) { 418 sbuf_printf(sb, " hd %u", dp->d_fwheads); 419 sbuf_printf(sb, " sc %u", dp->d_fwsectors); 420 return; 421 } 422 if (pp != NULL) { 423 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n", 424 indent, dp->d_fwheads); 425 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n", 426 indent, dp->d_fwsectors); 427 sbuf_printf(sb, "%s<ident>%s</ident>\n", indent, dp->d_ident); 428 sbuf_printf(sb, "%s<descr>%s</descr>\n", indent, dp->d_descr); 429 } 430} 431 432static void 433g_disk_create(void *arg, int flag) 434{ 435 struct g_geom *gp; 436 struct g_provider *pp; 437 struct disk *dp; 438 struct g_disk_softc *sc; 439 char tmpstr[80]; 440 441 if (flag == EV_CANCEL) 442 return; 443 g_topology_assert(); 444 dp = arg; 445 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO); 446 sc->dp = dp; 447 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit); 448 gp->softc = sc; 449 pp = g_new_providerf(gp, "%s", gp->name); 450 pp->mediasize = dp->d_mediasize; 451 pp->sectorsize = dp->d_sectorsize; 452 if (dp->d_flags & DISKFLAG_CANDELETE) 453 pp->flags |= G_PF_CANDELETE; 454 pp->stripeoffset = dp->d_stripeoffset; 455 pp->stripesize = dp->d_stripesize; 456 if (bootverbose) 457 printf("GEOM: new disk %s\n", gp->name); 458 sysctl_ctx_init(&sc->sysctl_ctx); 459 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name); 460 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx, 461 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name, 462 CTLFLAG_RD, 0, tmpstr); 463 if (sc->sysctl_tree != NULL) { 464 snprintf(tmpstr, sizeof(tmpstr), 465 "kern.geom.disk.%s.led", gp->name); 466 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led)); 467 SYSCTL_ADD_STRING(&sc->sysctl_ctx, 468 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led", 469 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led), 470 "LED name"); 471 } 472 pp->private = sc; 473 dp->d_geom = gp; 474 g_error_provider(pp, 0); 475} 476 477static void 478g_disk_destroy(void *ptr, int flag) 479{ 480 struct disk *dp; 481 struct g_geom *gp; 482 struct g_disk_softc *sc; 483 484 g_topology_assert(); 485 dp = ptr; 486 gp = dp->d_geom; 487 if (gp != NULL) { 488 sc = gp->softc; 489 if (sc->sysctl_tree != NULL) { 490 sysctl_ctx_free(&sc->sysctl_ctx); 491 sc->sysctl_tree = NULL; 492 } 493 if (sc->led[0] != 0) { 494 led_set(sc->led, "0"); 495 sc->led[0] = 0; 496 } 497 g_free(sc); 498 gp->softc = NULL; 499 g_wither_geom(gp, ENXIO); 500 } 501 g_free(dp); 502} 503 504/* 505 * We only allow printable characters in disk ident, 506 * the rest is converted to 'x<HH>'. 507 */ 508static void 509g_disk_ident_adjust(char *ident, size_t size) 510{ 511 char *p, tmp[4], newid[DISK_IDENT_SIZE]; 512 513 newid[0] = '\0'; 514 for (p = ident; *p != '\0'; p++) { 515 if (isprint(*p)) { 516 tmp[0] = *p; 517 tmp[1] = '\0'; 518 } else { 519 snprintf(tmp, sizeof(tmp), "x%02hhx", 520 *(unsigned char *)p); 521 } 522 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid)) 523 break; 524 } 525 bzero(ident, size); 526 strlcpy(ident, newid, size); 527} 528 529struct disk * 530disk_alloc() 531{ 532 struct disk *dp; 533 534 dp = g_malloc(sizeof *dp, M_WAITOK | M_ZERO); 535 return (dp); 536} 537 538void 539disk_create(struct disk *dp, int version) 540{ 541 if (version != DISK_VERSION_00 && version != DISK_VERSION_01) { 542 printf("WARNING: Attempt to add disk %s%d %s", 543 dp->d_name, dp->d_unit, 544 " using incompatible ABI version of disk(9)\n"); 545 printf("WARNING: Ignoring disk %s%d\n", 546 dp->d_name, dp->d_unit); 547 return; 548 } 549 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy")); 550 KASSERT(dp->d_name != NULL, ("disk_create need d_name")); 551 KASSERT(*dp->d_name != 0, ("disk_create need d_name")); 552 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long")); 553 if (dp->d_devstat == NULL) 554 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit, 555 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED, 556 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); 557 dp->d_geom = NULL; 558 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident)); 559 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL); 560} 561 562void 563disk_destroy(struct disk *dp) 564{ 565 566 g_cancel_event(dp); 567 dp->d_destroyed = 1; 568 if (dp->d_devstat != NULL) 569 devstat_remove_entry(dp->d_devstat); 570 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL); 571} 572 573void 574disk_gone(struct disk *dp) 575{ 576 struct g_geom *gp; 577 struct g_provider *pp; 578 579 gp = dp->d_geom; 580 if (gp != NULL) 581 LIST_FOREACH(pp, &gp->provider, provider) 582 g_wither_provider(pp, ENXIO); 583} 584 585static void 586g_kern_disks(void *p, int flag __unused) 587{ 588 struct sbuf *sb; 589 struct g_geom *gp; 590 char *sp; 591 592 sb = p; 593 sp = ""; 594 g_topology_assert(); 595 LIST_FOREACH(gp, &g_disk_class.geom, geom) { 596 sbuf_printf(sb, "%s%s", sp, gp->name); 597 sp = " "; 598 } 599 sbuf_finish(sb); 600} 601 602static int 603sysctl_disks(SYSCTL_HANDLER_ARGS) 604{ 605 int error; 606 struct sbuf *sb; 607 608 sb = sbuf_new_auto(); 609 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL); 610 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); 611 sbuf_delete(sb); 612 return error; 613} 614 615SYSCTL_PROC(_kern, OID_AUTO, disks, 616 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 617 sysctl_disks, "A", "names of available disks"); 618 619