geom_disk.c revision 300214
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: stable/10/sys/geom/geom_disk.c 300214 2016-05-19 16:04:20Z asomers $"); 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/bus.h> 47#include <sys/ctype.h> 48#include <sys/fcntl.h> 49#include <sys/malloc.h> 50#include <sys/sbuf.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 62#include <machine/bus.h> 63 64struct g_disk_softc { 65 struct mtx done_mtx; 66 struct disk *dp; 67 struct sysctl_ctx_list sysctl_ctx; 68 struct sysctl_oid *sysctl_tree; 69 char led[64]; 70 uint32_t state; 71 struct mtx start_mtx; 72}; 73 74static g_access_t g_disk_access; 75static g_start_t g_disk_start; 76static g_ioctl_t g_disk_ioctl; 77static g_dumpconf_t g_disk_dumpconf; 78static g_provgone_t g_disk_providergone; 79 80static struct g_class g_disk_class = { 81 .name = G_DISK_CLASS_NAME, 82 .version = G_VERSION, 83 .start = g_disk_start, 84 .access = g_disk_access, 85 .ioctl = g_disk_ioctl, 86 .providergone = g_disk_providergone, 87 .dumpconf = g_disk_dumpconf, 88}; 89 90SYSCTL_DECL(_kern_geom); 91static SYSCTL_NODE(_kern_geom, OID_AUTO, disk, CTLFLAG_RW, 0, 92 "GEOM_DISK stuff"); 93 94DECLARE_GEOM_CLASS(g_disk_class, g_disk); 95 96static void __inline 97g_disk_lock_giant(struct disk *dp) 98{ 99 100 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 101 mtx_lock(&Giant); 102} 103 104static void __inline 105g_disk_unlock_giant(struct disk *dp) 106{ 107 108 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 109 mtx_unlock(&Giant); 110} 111 112static int 113g_disk_access(struct g_provider *pp, int r, int w, int e) 114{ 115 struct disk *dp; 116 struct g_disk_softc *sc; 117 int error; 118 119 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)", 120 pp->name, r, w, e); 121 g_topology_assert(); 122 sc = pp->private; 123 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { 124 /* 125 * Allow decreasing access count even if disk is not 126 * avaliable anymore. 127 */ 128 if (r <= 0 && w <= 0 && e <= 0) 129 return (0); 130 return (ENXIO); 131 } 132 r += pp->acr; 133 w += pp->acw; 134 e += pp->ace; 135 error = 0; 136 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) { 137 if (dp->d_open != NULL) { 138 g_disk_lock_giant(dp); 139 error = dp->d_open(dp); 140 if (bootverbose && error != 0) 141 printf("Opened disk %s -> %d\n", 142 pp->name, error); 143 g_disk_unlock_giant(dp); 144 if (error != 0) 145 return (error); 146 } 147 pp->mediasize = dp->d_mediasize; 148 pp->sectorsize = dp->d_sectorsize; 149 if (dp->d_maxsize == 0) { 150 printf("WARNING: Disk drive %s%d has no d_maxsize\n", 151 dp->d_name, dp->d_unit); 152 dp->d_maxsize = DFLTPHYS; 153 } 154 if (dp->d_delmaxsize == 0) { 155 if (bootverbose && dp->d_flags & DISKFLAG_CANDELETE) { 156 printf("WARNING: Disk drive %s%d has no " 157 "d_delmaxsize\n", dp->d_name, dp->d_unit); 158 } 159 dp->d_delmaxsize = dp->d_maxsize; 160 } 161 pp->stripeoffset = dp->d_stripeoffset; 162 pp->stripesize = dp->d_stripesize; 163 dp->d_flags |= DISKFLAG_OPEN; 164 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) { 165 if (dp->d_close != NULL) { 166 g_disk_lock_giant(dp); 167 error = dp->d_close(dp); 168 if (error != 0) 169 printf("Closed disk %s -> %d\n", 170 pp->name, error); 171 g_disk_unlock_giant(dp); 172 } 173 sc->state = G_STATE_ACTIVE; 174 if (sc->led[0] != 0) 175 led_set(sc->led, "0"); 176 dp->d_flags &= ~DISKFLAG_OPEN; 177 } 178 return (error); 179} 180 181static void 182g_disk_kerneldump(struct bio *bp, struct disk *dp) 183{ 184 struct g_kerneldump *gkd; 185 struct g_geom *gp; 186 187 gkd = (struct g_kerneldump*)bp->bio_data; 188 gp = bp->bio_to->geom; 189 g_trace(G_T_TOPOLOGY, "g_disk_kerneldump(%s, %jd, %jd)", 190 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length); 191 if (dp->d_dump == NULL) { 192 g_io_deliver(bp, ENODEV); 193 return; 194 } 195 gkd->di.dumper = dp->d_dump; 196 gkd->di.priv = dp; 197 gkd->di.blocksize = dp->d_sectorsize; 198 gkd->di.maxiosize = dp->d_maxsize; 199 gkd->di.mediaoffset = gkd->offset; 200 if ((gkd->offset + gkd->length) > dp->d_mediasize) 201 gkd->length = dp->d_mediasize - gkd->offset; 202 gkd->di.mediasize = gkd->length; 203 g_io_deliver(bp, 0); 204} 205 206static void 207g_disk_setstate(struct bio *bp, struct g_disk_softc *sc) 208{ 209 const char *cmd; 210 211 memcpy(&sc->state, bp->bio_data, sizeof(sc->state)); 212 if (sc->led[0] != 0) { 213 switch (sc->state) { 214 case G_STATE_FAILED: 215 cmd = "1"; 216 break; 217 case G_STATE_REBUILD: 218 cmd = "f5"; 219 break; 220 case G_STATE_RESYNC: 221 cmd = "f1"; 222 break; 223 default: 224 cmd = "0"; 225 break; 226 } 227 led_set(sc->led, cmd); 228 } 229 g_io_deliver(bp, 0); 230} 231 232static void 233g_disk_done(struct bio *bp) 234{ 235 struct bintime now; 236 struct bio *bp2; 237 struct g_disk_softc *sc; 238 239 /* See "notes" for why we need a mutex here */ 240 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */ 241 bp2 = bp->bio_parent; 242 sc = bp2->bio_to->private; 243 bp->bio_completed = bp->bio_length - bp->bio_resid; 244 binuptime(&now); 245 mtx_lock(&sc->done_mtx); 246 if (bp2->bio_error == 0) 247 bp2->bio_error = bp->bio_error; 248 bp2->bio_completed += bp->bio_completed; 249 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE|BIO_FLUSH)) != 0) 250 devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now); 251 bp2->bio_inbed++; 252 if (bp2->bio_children == bp2->bio_inbed) { 253 mtx_unlock(&sc->done_mtx); 254 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed; 255 g_io_deliver(bp2, bp2->bio_error); 256 } else 257 mtx_unlock(&sc->done_mtx); 258 g_destroy_bio(bp); 259} 260 261static int 262g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td) 263{ 264 struct disk *dp; 265 struct g_disk_softc *sc; 266 int error; 267 268 sc = pp->private; 269 dp = sc->dp; 270 271 if (dp->d_ioctl == NULL) 272 return (ENOIOCTL); 273 g_disk_lock_giant(dp); 274 error = dp->d_ioctl(dp, cmd, data, fflag, td); 275 g_disk_unlock_giant(dp); 276 return (error); 277} 278 279static off_t 280g_disk_maxsize(struct disk *dp, struct bio *bp) 281{ 282 if (bp->bio_cmd == BIO_DELETE) 283 return (dp->d_delmaxsize); 284 return (dp->d_maxsize); 285} 286 287static int 288g_disk_maxsegs(struct disk *dp, struct bio *bp) 289{ 290 return ((g_disk_maxsize(dp, bp) / PAGE_SIZE) + 1); 291} 292 293static void 294g_disk_advance(struct disk *dp, struct bio *bp, off_t off) 295{ 296 297 bp->bio_offset += off; 298 bp->bio_length -= off; 299 300 if ((bp->bio_flags & BIO_VLIST) != 0) { 301 bus_dma_segment_t *seg, *end; 302 303 seg = (bus_dma_segment_t *)bp->bio_data; 304 end = (bus_dma_segment_t *)bp->bio_data + bp->bio_ma_n; 305 off += bp->bio_ma_offset; 306 while (off >= seg->ds_len) { 307 KASSERT((seg != end), 308 ("vlist request runs off the end")); 309 off -= seg->ds_len; 310 seg++; 311 } 312 bp->bio_ma_offset = off; 313 bp->bio_ma_n = end - seg; 314 bp->bio_data = (void *)seg; 315 } else if ((bp->bio_flags & BIO_UNMAPPED) != 0) { 316 bp->bio_ma += off / PAGE_SIZE; 317 bp->bio_ma_offset += off; 318 bp->bio_ma_offset %= PAGE_SIZE; 319 bp->bio_ma_n -= off / PAGE_SIZE; 320 } else { 321 bp->bio_data += off; 322 } 323} 324 325static void 326g_disk_seg_limit(bus_dma_segment_t *seg, off_t *poffset, 327 off_t *plength, int *ppages) 328{ 329 uintptr_t seg_page_base; 330 uintptr_t seg_page_end; 331 off_t offset; 332 off_t length; 333 int seg_pages; 334 335 offset = *poffset; 336 length = *plength; 337 338 if (length > seg->ds_len - offset) 339 length = seg->ds_len - offset; 340 341 seg_page_base = trunc_page(seg->ds_addr + offset); 342 seg_page_end = round_page(seg->ds_addr + offset + length); 343 seg_pages = (seg_page_end - seg_page_base) >> PAGE_SHIFT; 344 345 if (seg_pages > *ppages) { 346 seg_pages = *ppages; 347 length = (seg_page_base + (seg_pages << PAGE_SHIFT)) - 348 (seg->ds_addr + offset); 349 } 350 351 *poffset = 0; 352 *plength -= length; 353 *ppages -= seg_pages; 354} 355 356static off_t 357g_disk_vlist_limit(struct disk *dp, struct bio *bp, bus_dma_segment_t **pendseg) 358{ 359 bus_dma_segment_t *seg, *end; 360 off_t residual; 361 off_t offset; 362 int pages; 363 364 seg = (bus_dma_segment_t *)bp->bio_data; 365 end = (bus_dma_segment_t *)bp->bio_data + bp->bio_ma_n; 366 residual = bp->bio_length; 367 offset = bp->bio_ma_offset; 368 pages = g_disk_maxsegs(dp, bp); 369 while (residual != 0 && pages != 0) { 370 KASSERT((seg != end), 371 ("vlist limit runs off the end")); 372 g_disk_seg_limit(seg, &offset, &residual, &pages); 373 seg++; 374 } 375 if (pendseg != NULL) 376 *pendseg = seg; 377 return (residual); 378} 379 380static bool 381g_disk_limit(struct disk *dp, struct bio *bp) 382{ 383 bool limited = false; 384 off_t maxsz; 385 386 maxsz = g_disk_maxsize(dp, bp); 387 388 /* 389 * XXX: If we have a stripesize we should really use it here. 390 * Care should be taken in the delete case if this is done 391 * as deletes can be very sensitive to size given how they 392 * are processed. 393 */ 394 if (bp->bio_length > maxsz) { 395 bp->bio_length = maxsz; 396 limited = true; 397 } 398 399 if ((bp->bio_flags & BIO_VLIST) != 0) { 400 bus_dma_segment_t *firstseg, *endseg; 401 off_t residual; 402 403 firstseg = (bus_dma_segment_t*)bp->bio_data; 404 residual = g_disk_vlist_limit(dp, bp, &endseg); 405 if (residual != 0) { 406 bp->bio_ma_n = endseg - firstseg; 407 bp->bio_length -= residual; 408 limited = true; 409 } 410 } else if ((bp->bio_flags & BIO_UNMAPPED) != 0) { 411 bp->bio_ma_n = 412 howmany(bp->bio_ma_offset + bp->bio_length, PAGE_SIZE); 413 } 414 415 return (limited); 416} 417 418static void 419g_disk_start(struct bio *bp) 420{ 421 struct bio *bp2, *bp3; 422 struct disk *dp; 423 struct g_disk_softc *sc; 424 int error; 425 off_t off; 426 427 sc = bp->bio_to->private; 428 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) { 429 g_io_deliver(bp, ENXIO); 430 return; 431 } 432 error = EJUSTRETURN; 433 switch(bp->bio_cmd) { 434 case BIO_DELETE: 435 if (!(dp->d_flags & DISKFLAG_CANDELETE)) { 436 error = EOPNOTSUPP; 437 break; 438 } 439 /* fall-through */ 440 case BIO_READ: 441 case BIO_WRITE: 442 KASSERT((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0 || 443 (bp->bio_flags & BIO_UNMAPPED) == 0, 444 ("unmapped bio not supported by disk %s", dp->d_name)); 445 off = 0; 446 bp3 = NULL; 447 bp2 = g_clone_bio(bp); 448 if (bp2 == NULL) { 449 error = ENOMEM; 450 break; 451 } 452 for (;;) { 453 if (g_disk_limit(dp, bp2)) { 454 off += bp2->bio_length; 455 456 /* 457 * To avoid a race, we need to grab the next bio 458 * before we schedule this one. See "notes". 459 */ 460 bp3 = g_clone_bio(bp); 461 if (bp3 == NULL) 462 bp->bio_error = ENOMEM; 463 } 464 bp2->bio_done = g_disk_done; 465 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize; 466 bp2->bio_bcount = bp2->bio_length; 467 bp2->bio_disk = dp; 468 mtx_lock(&sc->start_mtx); 469 devstat_start_transaction_bio(dp->d_devstat, bp2); 470 mtx_unlock(&sc->start_mtx); 471 g_disk_lock_giant(dp); 472 dp->d_strategy(bp2); 473 g_disk_unlock_giant(dp); 474 475 if (bp3 == NULL) 476 break; 477 478 bp2 = bp3; 479 bp3 = NULL; 480 g_disk_advance(dp, bp2, off); 481 } 482 break; 483 case BIO_GETATTR: 484 /* Give the driver a chance to override */ 485 if (dp->d_getattr != NULL) { 486 if (bp->bio_disk == NULL) 487 bp->bio_disk = dp; 488 error = dp->d_getattr(bp); 489 if (error != -1) 490 break; 491 error = EJUSTRETURN; 492 } 493 if (g_handleattr_int(bp, "GEOM::candelete", 494 (dp->d_flags & DISKFLAG_CANDELETE) != 0)) 495 break; 496 else if (g_handleattr_int(bp, "GEOM::fwsectors", 497 dp->d_fwsectors)) 498 break; 499 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads)) 500 break; 501 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0)) 502 break; 503 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident)) 504 break; 505 else if (g_handleattr_uint16_t(bp, "GEOM::hba_vendor", 506 dp->d_hba_vendor)) 507 break; 508 else if (g_handleattr_uint16_t(bp, "GEOM::hba_device", 509 dp->d_hba_device)) 510 break; 511 else if (g_handleattr_uint16_t(bp, "GEOM::hba_subvendor", 512 dp->d_hba_subvendor)) 513 break; 514 else if (g_handleattr_uint16_t(bp, "GEOM::hba_subdevice", 515 dp->d_hba_subdevice)) 516 break; 517 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump")) 518 g_disk_kerneldump(bp, dp); 519 else if (!strcmp(bp->bio_attribute, "GEOM::setstate")) 520 g_disk_setstate(bp, sc); 521 else if (!strcmp(bp->bio_attribute, "GEOM::rotation_rate")) { 522 uint64_t v; 523 524 if ((dp->d_flags & DISKFLAG_LACKS_ROTRATE) == 0) 525 v = dp->d_rotation_rate; 526 else 527 v = 0; /* rate unknown */ 528 g_handleattr_uint16_t(bp, "GEOM::rotation_rate", v); 529 break; 530 } else 531 error = ENOIOCTL; 532 break; 533 case BIO_FLUSH: 534 g_trace(G_T_BIO, "g_disk_flushcache(%s)", 535 bp->bio_to->name); 536 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) { 537 error = EOPNOTSUPP; 538 break; 539 } 540 bp2 = g_clone_bio(bp); 541 if (bp2 == NULL) { 542 g_io_deliver(bp, ENOMEM); 543 return; 544 } 545 bp2->bio_done = g_disk_done; 546 bp2->bio_disk = dp; 547 mtx_lock(&sc->start_mtx); 548 devstat_start_transaction_bio(dp->d_devstat, bp2); 549 mtx_unlock(&sc->start_mtx); 550 g_disk_lock_giant(dp); 551 dp->d_strategy(bp2); 552 g_disk_unlock_giant(dp); 553 break; 554 default: 555 error = EOPNOTSUPP; 556 break; 557 } 558 if (error != EJUSTRETURN) 559 g_io_deliver(bp, error); 560 return; 561} 562 563static void 564g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) 565{ 566 struct bio *bp; 567 struct disk *dp; 568 struct g_disk_softc *sc; 569 char *buf; 570 int res = 0; 571 572 sc = gp->softc; 573 if (sc == NULL || (dp = sc->dp) == NULL) 574 return; 575 if (indent == NULL) { 576 sbuf_printf(sb, " hd %u", dp->d_fwheads); 577 sbuf_printf(sb, " sc %u", dp->d_fwsectors); 578 return; 579 } 580 if (pp != NULL) { 581 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n", 582 indent, dp->d_fwheads); 583 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n", 584 indent, dp->d_fwsectors); 585 586 /* 587 * "rotationrate" is a little complicated, because the value 588 * returned by the drive might not be the RPM; 0 and 1 are 589 * special cases, and there's also a valid range. 590 */ 591 sbuf_printf(sb, "%s<rotationrate>", indent); 592 if (dp->d_rotation_rate == 0) /* Old drives don't */ 593 sbuf_printf(sb, "unknown"); /* report RPM. */ 594 else if (dp->d_rotation_rate == 1) /* Since 0 is used */ 595 sbuf_printf(sb, "0"); /* above, SSDs use 1. */ 596 else if ((dp->d_rotation_rate >= 0x041) && 597 (dp->d_rotation_rate <= 0xfffe)) 598 sbuf_printf(sb, "%u", dp->d_rotation_rate); 599 else 600 sbuf_printf(sb, "invalid"); 601 sbuf_printf(sb, "</rotationrate>\n"); 602 if (dp->d_getattr != NULL) { 603 buf = g_malloc(DISK_IDENT_SIZE, M_WAITOK); 604 bp = g_alloc_bio(); 605 bp->bio_disk = dp; 606 bp->bio_attribute = "GEOM::ident"; 607 bp->bio_length = DISK_IDENT_SIZE; 608 bp->bio_data = buf; 609 res = dp->d_getattr(bp); 610 sbuf_printf(sb, "%s<ident>", indent); 611 g_conf_printf_escaped(sb, "%s", 612 res == 0 ? buf: dp->d_ident); 613 sbuf_printf(sb, "</ident>\n"); 614 bp->bio_attribute = "GEOM::lunid"; 615 bp->bio_length = DISK_IDENT_SIZE; 616 bp->bio_data = buf; 617 if (dp->d_getattr(bp) == 0) { 618 sbuf_printf(sb, "%s<lunid>", indent); 619 g_conf_printf_escaped(sb, "%s", buf); 620 sbuf_printf(sb, "</lunid>\n"); 621 } 622 bp->bio_attribute = "GEOM::lunname"; 623 bp->bio_length = DISK_IDENT_SIZE; 624 bp->bio_data = buf; 625 if (dp->d_getattr(bp) == 0) { 626 sbuf_printf(sb, "%s<lunname>", indent); 627 g_conf_printf_escaped(sb, "%s", buf); 628 sbuf_printf(sb, "</lunname>\n"); 629 } 630 g_destroy_bio(bp); 631 g_free(buf); 632 } else { 633 sbuf_printf(sb, "%s<ident>", indent); 634 g_conf_printf_escaped(sb, "%s", dp->d_ident); 635 sbuf_printf(sb, "</ident>\n"); 636 } 637 sbuf_printf(sb, "%s<descr>", indent); 638 g_conf_printf_escaped(sb, "%s", dp->d_descr); 639 sbuf_printf(sb, "</descr>\n"); 640 } 641} 642 643static void 644g_disk_resize(void *ptr, int flag) 645{ 646 struct disk *dp; 647 struct g_geom *gp; 648 struct g_provider *pp; 649 650 if (flag == EV_CANCEL) 651 return; 652 g_topology_assert(); 653 654 dp = ptr; 655 gp = dp->d_geom; 656 657 if (dp->d_destroyed || gp == NULL) 658 return; 659 660 LIST_FOREACH(pp, &gp->provider, provider) { 661 if (pp->sectorsize != 0 && 662 pp->sectorsize != dp->d_sectorsize) 663 g_wither_provider(pp, ENXIO); 664 else 665 g_resize_provider(pp, dp->d_mediasize); 666 } 667} 668 669static void 670g_disk_create(void *arg, int flag) 671{ 672 struct g_geom *gp; 673 struct g_provider *pp; 674 struct disk *dp; 675 struct g_disk_softc *sc; 676 char tmpstr[80]; 677 678 if (flag == EV_CANCEL) 679 return; 680 g_topology_assert(); 681 dp = arg; 682 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO); 683 mtx_init(&sc->start_mtx, "g_disk_start", NULL, MTX_DEF); 684 mtx_init(&sc->done_mtx, "g_disk_done", NULL, MTX_DEF); 685 sc->dp = dp; 686 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit); 687 gp->softc = sc; 688 pp = g_new_providerf(gp, "%s", gp->name); 689 devstat_remove_entry(pp->stat); 690 pp->stat = NULL; 691 dp->d_devstat->id = pp; 692 pp->mediasize = dp->d_mediasize; 693 pp->sectorsize = dp->d_sectorsize; 694 pp->stripeoffset = dp->d_stripeoffset; 695 pp->stripesize = dp->d_stripesize; 696 if ((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0) 697 pp->flags |= G_PF_ACCEPT_UNMAPPED; 698 if ((dp->d_flags & DISKFLAG_DIRECT_COMPLETION) != 0) 699 pp->flags |= G_PF_DIRECT_SEND; 700 pp->flags |= G_PF_DIRECT_RECEIVE; 701 if (bootverbose) 702 printf("GEOM: new disk %s\n", gp->name); 703 sysctl_ctx_init(&sc->sysctl_ctx); 704 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name); 705 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx, 706 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name, 707 CTLFLAG_RD, 0, tmpstr); 708 if (sc->sysctl_tree != NULL) { 709 snprintf(tmpstr, sizeof(tmpstr), 710 "kern.geom.disk.%s.led", gp->name); 711 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led)); 712 SYSCTL_ADD_STRING(&sc->sysctl_ctx, 713 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led", 714 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led), 715 "LED name"); 716 } 717 pp->private = sc; 718 dp->d_geom = gp; 719 g_error_provider(pp, 0); 720} 721 722/* 723 * We get this callback after all of the consumers have gone away, and just 724 * before the provider is freed. If the disk driver provided a d_gone 725 * callback, let them know that it is okay to free resources -- they won't 726 * be getting any more accesses from GEOM. 727 */ 728static void 729g_disk_providergone(struct g_provider *pp) 730{ 731 struct disk *dp; 732 struct g_disk_softc *sc; 733 734 sc = (struct g_disk_softc *)pp->private; 735 dp = sc->dp; 736 if (dp != NULL && dp->d_gone != NULL) 737 dp->d_gone(dp); 738 if (sc->sysctl_tree != NULL) { 739 sysctl_ctx_free(&sc->sysctl_ctx); 740 sc->sysctl_tree = NULL; 741 } 742 if (sc->led[0] != 0) { 743 led_set(sc->led, "0"); 744 sc->led[0] = 0; 745 } 746 pp->private = NULL; 747 pp->geom->softc = NULL; 748 mtx_destroy(&sc->done_mtx); 749 mtx_destroy(&sc->start_mtx); 750 g_free(sc); 751} 752 753static void 754g_disk_destroy(void *ptr, int flag) 755{ 756 struct disk *dp; 757 struct g_geom *gp; 758 struct g_disk_softc *sc; 759 760 g_topology_assert(); 761 dp = ptr; 762 gp = dp->d_geom; 763 if (gp != NULL) { 764 sc = gp->softc; 765 if (sc != NULL) 766 sc->dp = NULL; 767 dp->d_geom = NULL; 768 g_wither_geom(gp, ENXIO); 769 } 770 g_free(dp); 771} 772 773/* 774 * We only allow printable characters in disk ident, 775 * the rest is converted to 'x<HH>'. 776 */ 777static void 778g_disk_ident_adjust(char *ident, size_t size) 779{ 780 char *p, tmp[4], newid[DISK_IDENT_SIZE]; 781 782 newid[0] = '\0'; 783 for (p = ident; *p != '\0'; p++) { 784 if (isprint(*p)) { 785 tmp[0] = *p; 786 tmp[1] = '\0'; 787 } else { 788 snprintf(tmp, sizeof(tmp), "x%02hhx", 789 *(unsigned char *)p); 790 } 791 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid)) 792 break; 793 } 794 bzero(ident, size); 795 strlcpy(ident, newid, size); 796} 797 798struct disk * 799disk_alloc(void) 800{ 801 802 return (g_malloc(sizeof(struct disk), M_WAITOK | M_ZERO)); 803} 804 805void 806disk_create(struct disk *dp, int version) 807{ 808 809 if (version != DISK_VERSION) { 810 printf("WARNING: Attempt to add disk %s%d %s", 811 dp->d_name, dp->d_unit, 812 " using incompatible ABI version of disk(9)\n"); 813 printf("WARNING: Ignoring disk %s%d\n", 814 dp->d_name, dp->d_unit); 815 return; 816 } 817 if (version < DISK_VERSION_04) 818 dp->d_flags |= DISKFLAG_LACKS_ROTRATE; 819 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy")); 820 KASSERT(dp->d_name != NULL, ("disk_create need d_name")); 821 KASSERT(*dp->d_name != 0, ("disk_create need d_name")); 822 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long")); 823 if (dp->d_devstat == NULL) 824 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit, 825 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED, 826 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); 827 dp->d_geom = NULL; 828 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident)); 829 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL); 830} 831 832void 833disk_destroy(struct disk *dp) 834{ 835 836 g_cancel_event(dp); 837 dp->d_destroyed = 1; 838 if (dp->d_devstat != NULL) 839 devstat_remove_entry(dp->d_devstat); 840 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL); 841} 842 843void 844disk_gone(struct disk *dp) 845{ 846 struct g_geom *gp; 847 struct g_provider *pp; 848 849 gp = dp->d_geom; 850 if (gp != NULL) { 851 pp = LIST_FIRST(&gp->provider); 852 if (pp != NULL) { 853 KASSERT(LIST_NEXT(pp, provider) == NULL, 854 ("geom %p has more than one provider", gp)); 855 g_wither_provider(pp, ENXIO); 856 } 857 } 858} 859 860void 861disk_attr_changed(struct disk *dp, const char *attr, int flag) 862{ 863 struct g_geom *gp; 864 struct g_provider *pp; 865 char devnamebuf[128]; 866 867 gp = dp->d_geom; 868 if (gp != NULL) 869 LIST_FOREACH(pp, &gp->provider, provider) 870 (void)g_attr_changed(pp, attr, flag); 871 snprintf(devnamebuf, sizeof(devnamebuf), "devname=%s%d", dp->d_name, 872 dp->d_unit); 873 devctl_notify("GEOM", "disk", attr, devnamebuf); 874} 875 876void 877disk_media_changed(struct disk *dp, int flag) 878{ 879 struct g_geom *gp; 880 struct g_provider *pp; 881 882 gp = dp->d_geom; 883 if (gp != NULL) { 884 pp = LIST_FIRST(&gp->provider); 885 if (pp != NULL) { 886 KASSERT(LIST_NEXT(pp, provider) == NULL, 887 ("geom %p has more than one provider", gp)); 888 g_media_changed(pp, flag); 889 } 890 } 891} 892 893void 894disk_media_gone(struct disk *dp, int flag) 895{ 896 struct g_geom *gp; 897 struct g_provider *pp; 898 899 gp = dp->d_geom; 900 if (gp != NULL) { 901 pp = LIST_FIRST(&gp->provider); 902 if (pp != NULL) { 903 KASSERT(LIST_NEXT(pp, provider) == NULL, 904 ("geom %p has more than one provider", gp)); 905 g_media_gone(pp, flag); 906 } 907 } 908} 909 910int 911disk_resize(struct disk *dp, int flag) 912{ 913 914 if (dp->d_destroyed || dp->d_geom == NULL) 915 return (0); 916 917 return (g_post_event(g_disk_resize, dp, flag, NULL)); 918} 919 920static void 921g_kern_disks(void *p, int flag __unused) 922{ 923 struct sbuf *sb; 924 struct g_geom *gp; 925 char *sp; 926 927 sb = p; 928 sp = ""; 929 g_topology_assert(); 930 LIST_FOREACH(gp, &g_disk_class.geom, geom) { 931 sbuf_printf(sb, "%s%s", sp, gp->name); 932 sp = " "; 933 } 934 sbuf_finish(sb); 935} 936 937static int 938sysctl_disks(SYSCTL_HANDLER_ARGS) 939{ 940 int error; 941 struct sbuf *sb; 942 943 sb = sbuf_new_auto(); 944 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL); 945 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); 946 sbuf_delete(sb); 947 return error; 948} 949 950SYSCTL_PROC(_kern, OID_AUTO, disks, 951 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, 952 sysctl_disks, "A", "names of available disks"); 953