1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 /* 27 * ZFS volume emulation driver. 28 * 29 * Makes a DMU object look like a volume of arbitrary size, up to 2^64 bytes. 30 * Volumes are accessed through the symbolic links named: 31 * 32 * /dev/zvol/dsk/<pool_name>/<dataset_name> 33 * /dev/zvol/rdsk/<pool_name>/<dataset_name> 34 * 35 * These links are created by the ZFS-specific devfsadm link generator. 36 * Volumes are persistent through reboot. No user command needs to be 37 * run before opening and using a device. 38 */ 39 40 #include <sys/types.h> 41 #include <sys/param.h> 42 #include <sys/errno.h> 43 #include <sys/uio.h> 44 #include <sys/buf.h> 45 #include <sys/modctl.h> 46 #include <sys/open.h> 47 #include <sys/kmem.h> 48 #include <sys/conf.h> 49 #include <sys/cmn_err.h> 50 #include <sys/stat.h> 51 #include <sys/zap.h> 52 #include <sys/spa.h> 53 #include <sys/zio.h> 54 #include <sys/dmu_traverse.h> 55 #include <sys/dnode.h> 56 #include <sys/dsl_dataset.h> 57 #include <sys/dsl_prop.h> 58 #include <sys/dkio.h> 59 #include <sys/efi_partition.h> 60 #include <sys/byteorder.h> 61 #include <sys/pathname.h> 62 #include <sys/ddi.h> 63 #include <sys/sunddi.h> 64 #include <sys/crc32.h> 65 #include <sys/dirent.h> 66 #include <sys/policy.h> 67 #include <sys/fs/zfs.h> 68 #include <sys/zfs_ioctl.h> 69 #include <sys/mkdev.h> 70 #include <sys/zil.h> 71 #include <sys/refcount.h> 72 #include <sys/zfs_znode.h> 73 #include <sys/zfs_rlock.h> 74 #include <sys/vdev_disk.h> 75 #include <sys/vdev_impl.h> 76 #include <sys/zvol.h> 77 #include <sys/dumphdr.h> 78 #include <sys/zil_impl.h> 79 80 #include "zfs_namecheck.h" 81 82 static void *zvol_state; 83 84 #define ZVOL_DUMPSIZE "dumpsize" 85 86 /* 87 * This lock protects the zvol_state structure from being modified 88 * while it's being used, e.g. an open that comes in before a create 89 * finishes. It also protects temporary opens of the dataset so that, 90 * e.g., an open doesn't get a spurious EBUSY. 91 */ 92 static kmutex_t zvol_state_lock; 93 static uint32_t zvol_minors; 94 95 typedef struct zvol_extent { 96 list_node_t ze_node; 97 dva_t ze_dva; /* dva associated with this extent */ 98 uint64_t ze_nblks; /* number of blocks in extent */ 99 } zvol_extent_t; 100 101 /* 102 * The in-core state of each volume. 103 */ 104 typedef struct zvol_state { 105 char zv_name[MAXPATHLEN]; /* pool/dd name */ 106 uint64_t zv_volsize; /* amount of space we advertise */ 107 uint64_t zv_volblocksize; /* volume block size */ 108 minor_t zv_minor; /* minor number */ 109 uint8_t zv_min_bs; /* minimum addressable block shift */ 110 uint8_t zv_flags; /* readonly, dumpified, etc. */ 111 objset_t *zv_objset; /* objset handle */ 112 uint32_t zv_mode; /* DS_MODE_* flags at open time */ 113 uint32_t zv_open_count[OTYPCNT]; /* open counts */ 114 uint32_t zv_total_opens; /* total open count */ 115 zilog_t *zv_zilog; /* ZIL handle */ 116 list_t zv_extents; /* List of extents for dump */ 117 znode_t zv_znode; /* for range locking */ 118 } zvol_state_t; 119 120 /* 121 * zvol specific flags 122 */ 123 #define ZVOL_RDONLY 0x1 124 #define ZVOL_DUMPIFIED 0x2 125 #define ZVOL_EXCL 0x4 126 #define ZVOL_WCE 0x8 127 128 /* 129 * zvol maximum transfer in one DMU tx. 130 */ 131 int zvol_maxphys = DMU_MAX_ACCESS/2; 132 133 extern int zfs_set_prop_nvlist(const char *, nvlist_t *); 134 static int zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio); 135 static int zvol_dumpify(zvol_state_t *zv); 136 static int zvol_dump_fini(zvol_state_t *zv); 137 static int zvol_dump_init(zvol_state_t *zv, boolean_t resize); 138 139 static void 140 zvol_size_changed(zvol_state_t *zv, major_t maj) 141 { 142 dev_t dev = makedevice(maj, zv->zv_minor); 143 144 VERIFY(ddi_prop_update_int64(dev, zfs_dip, 145 "Size", zv->zv_volsize) == DDI_SUCCESS); 146 VERIFY(ddi_prop_update_int64(dev, zfs_dip, 147 "Nblocks", lbtodb(zv->zv_volsize)) == DDI_SUCCESS); 148 149 /* Notify specfs to invalidate the cached size */ 150 spec_size_invalidate(dev, VBLK); 151 spec_size_invalidate(dev, VCHR); 152 } 153 154 int 155 zvol_check_volsize(uint64_t volsize, uint64_t blocksize) 156 { 157 if (volsize == 0) 158 return (EINVAL); 159 160 if (volsize % blocksize != 0) 161 return (EINVAL); 162 163 #ifdef _ILP32 164 if (volsize - 1 > SPEC_MAXOFFSET_T) 165 return (EOVERFLOW); 166 #endif 167 return (0); 168 } 169 170 int 171 zvol_check_volblocksize(uint64_t volblocksize) 172 { 173 if (volblocksize < SPA_MINBLOCKSIZE || 174 volblocksize > SPA_MAXBLOCKSIZE || 175 !ISP2(volblocksize)) 176 return (EDOM); 177 178 return (0); 179 } 180 181 static void 182 zvol_readonly_changed_cb(void *arg, uint64_t newval) 183 { 184 zvol_state_t *zv = arg; 185 186 if (newval) 187 zv->zv_flags |= ZVOL_RDONLY; 188 else 189 zv->zv_flags &= ~ZVOL_RDONLY; 190 } 191 192 int 193 zvol_get_stats(objset_t *os, nvlist_t *nv) 194 { 195 int error; 196 dmu_object_info_t doi; 197 uint64_t val; 198 199 200 error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &val); 201 if (error) 202 return (error); 203 204 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLSIZE, val); 205 206 error = dmu_object_info(os, ZVOL_OBJ, &doi); 207 208 if (error == 0) { 209 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLBLOCKSIZE, 210 doi.doi_data_block_size); 211 } 212 213 return (error); 214 } 215 216 /* 217 * Find a free minor number. 218 */ 219 static minor_t 220 zvol_minor_alloc(void) 221 { 222 minor_t minor; 223 224 ASSERT(MUTEX_HELD(&zvol_state_lock)); 225 226 for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) 227 if (ddi_get_soft_state(zvol_state, minor) == NULL) 228 return (minor); 229 230 return (0); 231 } 232 233 static zvol_state_t * 234 zvol_minor_lookup(const char *name) 235 { 236 minor_t minor; 237 zvol_state_t *zv; 238 239 ASSERT(MUTEX_HELD(&zvol_state_lock)); 240 241 for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) { 242 zv = ddi_get_soft_state(zvol_state, minor); 243 if (zv == NULL) 244 continue; 245 if (strcmp(zv->zv_name, name) == 0) 246 break; 247 } 248 249 return (zv); 250 } 251 252 /* extent mapping arg */ 253 struct maparg { 254 zvol_state_t *ma_zv; 255 uint64_t ma_blks; 256 }; 257 258 /*ARGSUSED*/ 259 static int 260 zvol_map_block(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb, 261 const dnode_phys_t *dnp, void *arg) 262 { 263 struct maparg *ma = arg; 264 zvol_extent_t *ze; 265 int bs = ma->ma_zv->zv_volblocksize; 266 267 if (bp == NULL || zb->zb_object != ZVOL_OBJ || zb->zb_level != 0) 268 return (0); 269 270 VERIFY3U(ma->ma_blks, ==, zb->zb_blkid); 271 ma->ma_blks++; 272 273 /* Abort immediately if we have encountered gang blocks */ 274 if (BP_IS_GANG(bp)) 275 return (EFRAGS); 276 277 /* 278 * See if the block is at the end of the previous extent. 279 */ 280 ze = list_tail(&ma->ma_zv->zv_extents); 281 if (ze && 282 DVA_GET_VDEV(BP_IDENTITY(bp)) == DVA_GET_VDEV(&ze->ze_dva) && 283 DVA_GET_OFFSET(BP_IDENTITY(bp)) == 284 DVA_GET_OFFSET(&ze->ze_dva) + ze->ze_nblks * bs) { 285 ze->ze_nblks++; 286 return (0); 287 } 288 289 dprintf_bp(bp, "%s", "next blkptr:"); 290 291 /* start a new extent */ 292 ze = kmem_zalloc(sizeof (zvol_extent_t), KM_SLEEP); 293 ze->ze_dva = bp->blk_dva[0]; /* structure assignment */ 294 ze->ze_nblks = 1; 295 list_insert_tail(&ma->ma_zv->zv_extents, ze); 296 return (0); 297 } 298 299 static void 300 zvol_free_extents(zvol_state_t *zv) 301 { 302 zvol_extent_t *ze; 303 304 while (ze = list_head(&zv->zv_extents)) { 305 list_remove(&zv->zv_extents, ze); 306 kmem_free(ze, sizeof (zvol_extent_t)); 307 } 308 } 309 310 static int 311 zvol_get_lbas(zvol_state_t *zv) 312 { 313 struct maparg ma; 314 int err; 315 316 ma.ma_zv = zv; 317 ma.ma_blks = 0; 318 zvol_free_extents(zv); 319 320 err = traverse_dataset(dmu_objset_ds(zv->zv_objset), 0, 321 TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA, zvol_map_block, &ma); 322 if (err || ma.ma_blks != (zv->zv_volsize / zv->zv_volblocksize)) { 323 zvol_free_extents(zv); 324 return (err ? err : EIO); 325 } 326 327 return (0); 328 } 329 330 /* ARGSUSED */ 331 void 332 zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 333 { 334 zfs_creat_t *zct = arg; 335 nvlist_t *nvprops = zct->zct_props; 336 int error; 337 uint64_t volblocksize, volsize; 338 339 VERIFY(nvlist_lookup_uint64(nvprops, 340 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0); 341 if (nvlist_lookup_uint64(nvprops, 342 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0) 343 volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE); 344 345 /* 346 * These properties must be removed from the list so the generic 347 * property setting step won't apply to them. 348 */ 349 VERIFY(nvlist_remove_all(nvprops, 350 zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0); 351 (void) nvlist_remove_all(nvprops, 352 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE)); 353 354 error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize, 355 DMU_OT_NONE, 0, tx); 356 ASSERT(error == 0); 357 358 error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP, 359 DMU_OT_NONE, 0, tx); 360 ASSERT(error == 0); 361 362 error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx); 363 ASSERT(error == 0); 364 } 365 366 /* 367 * Replay a TX_WRITE ZIL transaction that didn't get committed 368 * after a system failure 369 */ 370 static int 371 zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap) 372 { 373 objset_t *os = zv->zv_objset; 374 char *data = (char *)(lr + 1); /* data follows lr_write_t */ 375 uint64_t off = lr->lr_offset; 376 uint64_t len = lr->lr_length; 377 dmu_tx_t *tx; 378 int error; 379 380 if (byteswap) 381 byteswap_uint64_array(lr, sizeof (*lr)); 382 383 tx = dmu_tx_create(os); 384 dmu_tx_hold_write(tx, ZVOL_OBJ, off, len); 385 error = dmu_tx_assign(tx, TXG_WAIT); 386 if (error) { 387 dmu_tx_abort(tx); 388 } else { 389 dmu_write(os, ZVOL_OBJ, off, len, data, tx); 390 dmu_tx_commit(tx); 391 } 392 393 return (error); 394 } 395 396 /* ARGSUSED */ 397 static int 398 zvol_replay_err(zvol_state_t *zv, lr_t *lr, boolean_t byteswap) 399 { 400 return (ENOTSUP); 401 } 402 403 /* 404 * Callback vectors for replaying records. 405 * Only TX_WRITE is needed for zvol. 406 */ 407 zil_replay_func_t *zvol_replay_vector[TX_MAX_TYPE] = { 408 zvol_replay_err, /* 0 no such transaction type */ 409 zvol_replay_err, /* TX_CREATE */ 410 zvol_replay_err, /* TX_MKDIR */ 411 zvol_replay_err, /* TX_MKXATTR */ 412 zvol_replay_err, /* TX_SYMLINK */ 413 zvol_replay_err, /* TX_REMOVE */ 414 zvol_replay_err, /* TX_RMDIR */ 415 zvol_replay_err, /* TX_LINK */ 416 zvol_replay_err, /* TX_RENAME */ 417 zvol_replay_write, /* TX_WRITE */ 418 zvol_replay_err, /* TX_TRUNCATE */ 419 zvol_replay_err, /* TX_SETATTR */ 420 zvol_replay_err, /* TX_ACL */ 421 }; 422 423 /* 424 * Create a minor node (plus a whole lot more) for the specified volume. 425 */ 426 int 427 zvol_create_minor(const char *name, major_t maj) 428 { 429 zvol_state_t *zv; 430 objset_t *os; 431 dmu_object_info_t doi; 432 uint64_t volsize; 433 minor_t minor = 0; 434 struct pathname linkpath; 435 int ds_mode = DS_MODE_OWNER; 436 vnode_t *vp = NULL; 437 char *devpath; 438 size_t devpathlen = strlen(ZVOL_FULL_DEV_DIR) + strlen(name) + 1; 439 char chrbuf[30], blkbuf[30]; 440 int error; 441 442 mutex_enter(&zvol_state_lock); 443 444 if ((zv = zvol_minor_lookup(name)) != NULL) { 445 mutex_exit(&zvol_state_lock); 446 return (EEXIST); 447 } 448 449 if (strchr(name, '@') != 0) 450 ds_mode |= DS_MODE_READONLY; 451 452 error = dmu_objset_open(name, DMU_OST_ZVOL, ds_mode, &os); 453 454 if (error) { 455 mutex_exit(&zvol_state_lock); 456 return (error); 457 } 458 459 error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); 460 461 if (error) { 462 dmu_objset_close(os); 463 mutex_exit(&zvol_state_lock); 464 return (error); 465 } 466 467 /* 468 * If there's an existing /dev/zvol symlink, try to use the 469 * same minor number we used last time. 470 */ 471 devpath = kmem_alloc(devpathlen, KM_SLEEP); 472 473 (void) sprintf(devpath, "%s%s", ZVOL_FULL_DEV_DIR, name); 474 475 error = lookupname(devpath, UIO_SYSSPACE, NO_FOLLOW, NULL, &vp); 476 477 kmem_free(devpath, devpathlen); 478 479 if (error == 0 && vp->v_type != VLNK) 480 error = EINVAL; 481 482 if (error == 0) { 483 pn_alloc(&linkpath); 484 error = pn_getsymlink(vp, &linkpath, kcred); 485 if (error == 0) { 486 char *ms = strstr(linkpath.pn_path, ZVOL_PSEUDO_DEV); 487 if (ms != NULL) { 488 ms += strlen(ZVOL_PSEUDO_DEV); 489 minor = stoi(&ms); 490 } 491 } 492 pn_free(&linkpath); 493 } 494 495 if (vp != NULL) 496 VN_RELE(vp); 497 498 /* 499 * If we found a minor but it's already in use, we must pick a new one. 500 */ 501 if (minor != 0 && ddi_get_soft_state(zvol_state, minor) != NULL) 502 minor = 0; 503 504 if (minor == 0) 505 minor = zvol_minor_alloc(); 506 507 if (minor == 0) { 508 dmu_objset_close(os); 509 mutex_exit(&zvol_state_lock); 510 return (ENXIO); 511 } 512 513 if (ddi_soft_state_zalloc(zvol_state, minor) != DDI_SUCCESS) { 514 dmu_objset_close(os); 515 mutex_exit(&zvol_state_lock); 516 return (EAGAIN); 517 } 518 519 (void) ddi_prop_update_string(minor, zfs_dip, ZVOL_PROP_NAME, 520 (char *)name); 521 522 (void) sprintf(chrbuf, "%uc,raw", minor); 523 524 if (ddi_create_minor_node(zfs_dip, chrbuf, S_IFCHR, 525 minor, DDI_PSEUDO, 0) == DDI_FAILURE) { 526 ddi_soft_state_free(zvol_state, minor); 527 dmu_objset_close(os); 528 mutex_exit(&zvol_state_lock); 529 return (EAGAIN); 530 } 531 532 (void) sprintf(blkbuf, "%uc", minor); 533 534 if (ddi_create_minor_node(zfs_dip, blkbuf, S_IFBLK, 535 minor, DDI_PSEUDO, 0) == DDI_FAILURE) { 536 ddi_remove_minor_node(zfs_dip, chrbuf); 537 ddi_soft_state_free(zvol_state, minor); 538 dmu_objset_close(os); 539 mutex_exit(&zvol_state_lock); 540 return (EAGAIN); 541 } 542 543 zv = ddi_get_soft_state(zvol_state, minor); 544 545 (void) strcpy(zv->zv_name, name); 546 zv->zv_min_bs = DEV_BSHIFT; 547 zv->zv_minor = minor; 548 zv->zv_volsize = volsize; 549 zv->zv_objset = os; 550 zv->zv_mode = ds_mode; 551 zv->zv_zilog = zil_open(os, zvol_get_data); 552 mutex_init(&zv->zv_znode.z_range_lock, NULL, MUTEX_DEFAULT, NULL); 553 avl_create(&zv->zv_znode.z_range_avl, zfs_range_compare, 554 sizeof (rl_t), offsetof(rl_t, r_node)); 555 list_create(&zv->zv_extents, sizeof (zvol_extent_t), 556 offsetof(zvol_extent_t, ze_node)); 557 /* get and cache the blocksize */ 558 error = dmu_object_info(os, ZVOL_OBJ, &doi); 559 ASSERT(error == 0); 560 zv->zv_volblocksize = doi.doi_data_block_size; 561 562 zil_replay(os, zv, zvol_replay_vector); 563 zvol_size_changed(zv, maj); 564 565 /* XXX this should handle the possible i/o error */ 566 VERIFY(dsl_prop_register(dmu_objset_ds(zv->zv_objset), 567 "readonly", zvol_readonly_changed_cb, zv) == 0); 568 569 zvol_minors++; 570 571 mutex_exit(&zvol_state_lock); 572 573 return (0); 574 } 575 576 /* 577 * Remove minor node for the specified volume. 578 */ 579 int 580 zvol_remove_minor(const char *name) 581 { 582 zvol_state_t *zv; 583 char namebuf[30]; 584 585 mutex_enter(&zvol_state_lock); 586 587 if ((zv = zvol_minor_lookup(name)) == NULL) { 588 mutex_exit(&zvol_state_lock); 589 return (ENXIO); 590 } 591 592 if (zv->zv_total_opens != 0) { 593 mutex_exit(&zvol_state_lock); 594 return (EBUSY); 595 } 596 597 (void) sprintf(namebuf, "%uc,raw", zv->zv_minor); 598 ddi_remove_minor_node(zfs_dip, namebuf); 599 600 (void) sprintf(namebuf, "%uc", zv->zv_minor); 601 ddi_remove_minor_node(zfs_dip, namebuf); 602 603 VERIFY(dsl_prop_unregister(dmu_objset_ds(zv->zv_objset), 604 "readonly", zvol_readonly_changed_cb, zv) == 0); 605 606 zil_close(zv->zv_zilog); 607 zv->zv_zilog = NULL; 608 dmu_objset_close(zv->zv_objset); 609 zv->zv_objset = NULL; 610 avl_destroy(&zv->zv_znode.z_range_avl); 611 mutex_destroy(&zv->zv_znode.z_range_lock); 612 613 ddi_soft_state_free(zvol_state, zv->zv_minor); 614 615 zvol_minors--; 616 617 mutex_exit(&zvol_state_lock); 618 619 return (0); 620 } 621 622 int 623 zvol_prealloc(zvol_state_t *zv) 624 { 625 objset_t *os = zv->zv_objset; 626 dmu_tx_t *tx; 627 uint64_t refd, avail, usedobjs, availobjs; 628 uint64_t resid = zv->zv_volsize; 629 uint64_t off = 0; 630 631 /* Check the space usage before attempting to allocate the space */ 632 dmu_objset_space(os, &refd, &avail, &usedobjs, &availobjs); 633 if (avail < zv->zv_volsize) 634 return (ENOSPC); 635 636 /* Free old extents if they exist */ 637 zvol_free_extents(zv); 638 639 while (resid != 0) { 640 int error; 641 uint64_t bytes = MIN(resid, SPA_MAXBLOCKSIZE); 642 643 tx = dmu_tx_create(os); 644 dmu_tx_hold_write(tx, ZVOL_OBJ, off, bytes); 645 error = dmu_tx_assign(tx, TXG_WAIT); 646 if (error) { 647 dmu_tx_abort(tx); 648 (void) dmu_free_long_range(os, ZVOL_OBJ, 0, off); 649 return (error); 650 } 651 dmu_prealloc(os, ZVOL_OBJ, off, bytes, tx); 652 dmu_tx_commit(tx); 653 off += bytes; 654 resid -= bytes; 655 } 656 txg_wait_synced(dmu_objset_pool(os), 0); 657 658 return (0); 659 } 660 661 int 662 zvol_update_volsize(zvol_state_t *zv, major_t maj, uint64_t volsize) 663 { 664 dmu_tx_t *tx; 665 int error; 666 667 ASSERT(MUTEX_HELD(&zvol_state_lock)); 668 669 tx = dmu_tx_create(zv->zv_objset); 670 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 671 error = dmu_tx_assign(tx, TXG_WAIT); 672 if (error) { 673 dmu_tx_abort(tx); 674 return (error); 675 } 676 677 error = zap_update(zv->zv_objset, ZVOL_ZAP_OBJ, "size", 8, 1, 678 &volsize, tx); 679 dmu_tx_commit(tx); 680 681 if (error == 0) 682 error = dmu_free_long_range(zv->zv_objset, 683 ZVOL_OBJ, volsize, DMU_OBJECT_END); 684 685 /* 686 * If we are using a faked-up state (zv_minor == 0) then don't 687 * try to update the in-core zvol state. 688 */ 689 if (error == 0 && zv->zv_minor) { 690 zv->zv_volsize = volsize; 691 zvol_size_changed(zv, maj); 692 } 693 return (error); 694 } 695 696 int 697 zvol_set_volsize(const char *name, major_t maj, uint64_t volsize) 698 { 699 zvol_state_t *zv; 700 int error; 701 dmu_object_info_t doi; 702 uint64_t old_volsize = 0ULL; 703 zvol_state_t state = { 0 }; 704 705 mutex_enter(&zvol_state_lock); 706 707 if ((zv = zvol_minor_lookup(name)) == NULL) { 708 /* 709 * If we are doing a "zfs clone -o volsize=", then the 710 * minor node won't exist yet. 711 */ 712 error = dmu_objset_open(name, DMU_OST_ZVOL, DS_MODE_OWNER, 713 &state.zv_objset); 714 if (error != 0) 715 goto out; 716 zv = &state; 717 } 718 old_volsize = zv->zv_volsize; 719 720 if ((error = dmu_object_info(zv->zv_objset, ZVOL_OBJ, &doi)) != 0 || 721 (error = zvol_check_volsize(volsize, 722 doi.doi_data_block_size)) != 0) 723 goto out; 724 725 if (zv->zv_flags & ZVOL_RDONLY || (zv->zv_mode & DS_MODE_READONLY)) { 726 error = EROFS; 727 goto out; 728 } 729 730 error = zvol_update_volsize(zv, maj, volsize); 731 732 /* 733 * Reinitialize the dump area to the new size. If we 734 * failed to resize the dump area then restore the it back to 735 * it's original size. 736 */ 737 if (error == 0 && zv->zv_flags & ZVOL_DUMPIFIED) { 738 if ((error = zvol_dumpify(zv)) != 0 || 739 (error = dumpvp_resize()) != 0) { 740 (void) zvol_update_volsize(zv, maj, old_volsize); 741 error = zvol_dumpify(zv); 742 } 743 } 744 745 out: 746 if (state.zv_objset) 747 dmu_objset_close(state.zv_objset); 748 749 mutex_exit(&zvol_state_lock); 750 751 return (error); 752 } 753 754 int 755 zvol_set_volblocksize(const char *name, uint64_t volblocksize) 756 { 757 zvol_state_t *zv; 758 dmu_tx_t *tx; 759 int error; 760 boolean_t needlock; 761 762 /* 763 * The lock may already be held if we are being called from 764 * zvol_dump_init(). 765 */ 766 needlock = !MUTEX_HELD(&zvol_state_lock); 767 if (needlock) 768 mutex_enter(&zvol_state_lock); 769 770 if ((zv = zvol_minor_lookup(name)) == NULL) { 771 if (needlock) 772 mutex_exit(&zvol_state_lock); 773 return (ENXIO); 774 } 775 if (zv->zv_flags & ZVOL_RDONLY || (zv->zv_mode & DS_MODE_READONLY)) { 776 if (needlock) 777 mutex_exit(&zvol_state_lock); 778 return (EROFS); 779 } 780 781 tx = dmu_tx_create(zv->zv_objset); 782 dmu_tx_hold_bonus(tx, ZVOL_OBJ); 783 error = dmu_tx_assign(tx, TXG_WAIT); 784 if (error) { 785 dmu_tx_abort(tx); 786 } else { 787 error = dmu_object_set_blocksize(zv->zv_objset, ZVOL_OBJ, 788 volblocksize, 0, tx); 789 if (error == ENOTSUP) 790 error = EBUSY; 791 dmu_tx_commit(tx); 792 if (error == 0) 793 zv->zv_volblocksize = volblocksize; 794 } 795 796 if (needlock) 797 mutex_exit(&zvol_state_lock); 798 799 return (error); 800 } 801 802 /*ARGSUSED*/ 803 int 804 zvol_open(dev_t *devp, int flag, int otyp, cred_t *cr) 805 { 806 minor_t minor = getminor(*devp); 807 zvol_state_t *zv; 808 809 if (minor == 0) /* This is the control device */ 810 return (0); 811 812 mutex_enter(&zvol_state_lock); 813 814 zv = ddi_get_soft_state(zvol_state, minor); 815 if (zv == NULL) { 816 mutex_exit(&zvol_state_lock); 817 return (ENXIO); 818 } 819 820 ASSERT(zv->zv_objset != NULL); 821 822 if ((flag & FWRITE) && 823 (zv->zv_flags & ZVOL_RDONLY || (zv->zv_mode & DS_MODE_READONLY))) { 824 mutex_exit(&zvol_state_lock); 825 return (EROFS); 826 } 827 if (zv->zv_flags & ZVOL_EXCL) { 828 mutex_exit(&zvol_state_lock); 829 return (EBUSY); 830 } 831 if (flag & FEXCL) { 832 if (zv->zv_total_opens != 0) { 833 mutex_exit(&zvol_state_lock); 834 return (EBUSY); 835 } 836 zv->zv_flags |= ZVOL_EXCL; 837 } 838 839 if (zv->zv_open_count[otyp] == 0 || otyp == OTYP_LYR) { 840 zv->zv_open_count[otyp]++; 841 zv->zv_total_opens++; 842 } 843 844 mutex_exit(&zvol_state_lock); 845 846 return (0); 847 } 848 849 /*ARGSUSED*/ 850 int 851 zvol_close(dev_t dev, int flag, int otyp, cred_t *cr) 852 { 853 minor_t minor = getminor(dev); 854 zvol_state_t *zv; 855 856 if (minor == 0) /* This is the control device */ 857 return (0); 858 859 mutex_enter(&zvol_state_lock); 860 861 zv = ddi_get_soft_state(zvol_state, minor); 862 if (zv == NULL) { 863 mutex_exit(&zvol_state_lock); 864 return (ENXIO); 865 } 866 867 if (zv->zv_flags & ZVOL_EXCL) { 868 ASSERT(zv->zv_total_opens == 1); 869 zv->zv_flags &= ~ZVOL_EXCL; 870 } 871 872 /* 873 * If the open count is zero, this is a spurious close. 874 * That indicates a bug in the kernel / DDI framework. 875 */ 876 ASSERT(zv->zv_open_count[otyp] != 0); 877 ASSERT(zv->zv_total_opens != 0); 878 879 /* 880 * You may get multiple opens, but only one close. 881 */ 882 zv->zv_open_count[otyp]--; 883 zv->zv_total_opens--; 884 885 mutex_exit(&zvol_state_lock); 886 887 return (0); 888 } 889 890 static void 891 zvol_get_done(dmu_buf_t *db, void *vzgd) 892 { 893 zgd_t *zgd = (zgd_t *)vzgd; 894 rl_t *rl = zgd->zgd_rl; 895 896 dmu_buf_rele(db, vzgd); 897 zfs_range_unlock(rl); 898 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); 899 kmem_free(zgd, sizeof (zgd_t)); 900 } 901 902 /* 903 * Get data to generate a TX_WRITE intent log record. 904 */ 905 static int 906 zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) 907 { 908 zvol_state_t *zv = arg; 909 objset_t *os = zv->zv_objset; 910 dmu_buf_t *db; 911 rl_t *rl; 912 zgd_t *zgd; 913 uint64_t boff; /* block starting offset */ 914 int dlen = lr->lr_length; /* length of user data */ 915 int error; 916 917 ASSERT(zio); 918 ASSERT(dlen != 0); 919 920 /* 921 * Write records come in two flavors: immediate and indirect. 922 * For small writes it's cheaper to store the data with the 923 * log record (immediate); for large writes it's cheaper to 924 * sync the data and get a pointer to it (indirect) so that 925 * we don't have to write the data twice. 926 */ 927 if (buf != NULL) /* immediate write */ 928 return (dmu_read(os, ZVOL_OBJ, lr->lr_offset, dlen, buf)); 929 930 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP); 931 zgd->zgd_zilog = zv->zv_zilog; 932 zgd->zgd_bp = &lr->lr_blkptr; 933 934 /* 935 * Lock the range of the block to ensure that when the data is 936 * written out and its checksum is being calculated that no other 937 * thread can change the block. 938 */ 939 boff = P2ALIGN_TYPED(lr->lr_offset, zv->zv_volblocksize, uint64_t); 940 rl = zfs_range_lock(&zv->zv_znode, boff, zv->zv_volblocksize, 941 RL_READER); 942 zgd->zgd_rl = rl; 943 944 VERIFY(0 == dmu_buf_hold(os, ZVOL_OBJ, lr->lr_offset, zgd, &db)); 945 error = dmu_sync(zio, db, &lr->lr_blkptr, 946 lr->lr_common.lrc_txg, zvol_get_done, zgd); 947 if (error == 0) 948 zil_add_block(zv->zv_zilog, &lr->lr_blkptr); 949 /* 950 * If we get EINPROGRESS, then we need to wait for a 951 * write IO initiated by dmu_sync() to complete before 952 * we can release this dbuf. We will finish everything 953 * up in the zvol_get_done() callback. 954 */ 955 if (error == EINPROGRESS) 956 return (0); 957 dmu_buf_rele(db, zgd); 958 zfs_range_unlock(rl); 959 kmem_free(zgd, sizeof (zgd_t)); 960 return (error); 961 } 962 963 /* 964 * zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions. 965 * 966 * We store data in the log buffers if it's small enough. 967 * Otherwise we will later flush the data out via dmu_sync(). 968 */ 969 ssize_t zvol_immediate_write_sz = 32768; 970 971 static void 972 zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, offset_t off, ssize_t len) 973 { 974 uint32_t blocksize = zv->zv_volblocksize; 975 zilog_t *zilog = zv->zv_zilog; 976 lr_write_t *lr; 977 978 if (zilog->zl_replay) { 979 dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); 980 zilog->zl_replayed_seq[dmu_tx_get_txg(tx) & TXG_MASK] = 981 zilog->zl_replaying_seq; 982 return; 983 } 984 985 while (len) { 986 ssize_t nbytes = MIN(len, blocksize - P2PHASE(off, blocksize)); 987 itx_t *itx = zil_itx_create(TX_WRITE, sizeof (*lr)); 988 989 itx->itx_wr_state = 990 len > zvol_immediate_write_sz ? WR_INDIRECT : WR_NEED_COPY; 991 itx->itx_private = zv; 992 lr = (lr_write_t *)&itx->itx_lr; 993 lr->lr_foid = ZVOL_OBJ; 994 lr->lr_offset = off; 995 lr->lr_length = nbytes; 996 lr->lr_blkoff = off - P2ALIGN_TYPED(off, blocksize, uint64_t); 997 BP_ZERO(&lr->lr_blkptr); 998 999 (void) zil_itx_assign(zilog, itx, tx); 1000 len -= nbytes; 1001 off += nbytes; 1002 } 1003 } 1004 1005 static int 1006 zvol_dumpio_vdev(vdev_t *vd, void *addr, uint64_t offset, uint64_t size, 1007 boolean_t doread, boolean_t isdump) 1008 { 1009 vdev_disk_t *dvd; 1010 int c; 1011 int numerrors = 0; 1012 1013 for (c = 0; c < vd->vdev_children; c++) { 1014 ASSERT(vd->vdev_ops == &vdev_mirror_ops); 1015 int err = zvol_dumpio_vdev(vd->vdev_child[c], 1016 addr, offset, size, doread, isdump); 1017 if (err != 0) { 1018 numerrors++; 1019 } else if (doread) { 1020 break; 1021 } 1022 } 1023 1024 if (!vd->vdev_ops->vdev_op_leaf) 1025 return (numerrors < vd->vdev_children ? 0 : EIO); 1026 1027 if (doread && !vdev_readable(vd)) 1028 return (EIO); 1029 else if (!doread && !vdev_writeable(vd)) 1030 return (EIO); 1031 1032 dvd = vd->vdev_tsd; 1033 ASSERT3P(dvd, !=, NULL); 1034 offset += VDEV_LABEL_START_SIZE; 1035 1036 if (ddi_in_panic() || isdump) { 1037 ASSERT(!doread); 1038 if (doread) 1039 return (EIO); 1040 return (ldi_dump(dvd->vd_lh, addr, lbtodb(offset), 1041 lbtodb(size))); 1042 } else { 1043 return (vdev_disk_physio(dvd->vd_lh, addr, size, offset, 1044 doread ? B_READ : B_WRITE)); 1045 } 1046 } 1047 1048 static int 1049 zvol_dumpio(zvol_state_t *zv, void *addr, uint64_t offset, uint64_t size, 1050 boolean_t doread, boolean_t isdump) 1051 { 1052 vdev_t *vd; 1053 int error; 1054 zvol_extent_t *ze; 1055 spa_t *spa = dmu_objset_spa(zv->zv_objset); 1056 1057 /* Must be sector aligned, and not stradle a block boundary. */ 1058 if (P2PHASE(offset, DEV_BSIZE) || P2PHASE(size, DEV_BSIZE) || 1059 P2BOUNDARY(offset, size, zv->zv_volblocksize)) { 1060 return (EINVAL); 1061 } 1062 ASSERT(size <= zv->zv_volblocksize); 1063 1064 /* Locate the extent this belongs to */ 1065 ze = list_head(&zv->zv_extents); 1066 while (offset >= ze->ze_nblks * zv->zv_volblocksize) { 1067 offset -= ze->ze_nblks * zv->zv_volblocksize; 1068 ze = list_next(&zv->zv_extents, ze); 1069 } 1070 spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); 1071 vd = vdev_lookup_top(spa, DVA_GET_VDEV(&ze->ze_dva)); 1072 offset += DVA_GET_OFFSET(&ze->ze_dva); 1073 error = zvol_dumpio_vdev(vd, addr, offset, size, doread, isdump); 1074 spa_config_exit(spa, SCL_STATE, FTAG); 1075 return (error); 1076 } 1077 1078 int 1079 zvol_strategy(buf_t *bp) 1080 { 1081 zvol_state_t *zv = ddi_get_soft_state(zvol_state, getminor(bp->b_edev)); 1082 uint64_t off, volsize; 1083 size_t resid; 1084 char *addr; 1085 objset_t *os; 1086 rl_t *rl; 1087 int error = 0; 1088 boolean_t doread = bp->b_flags & B_READ; 1089 boolean_t is_dump = zv->zv_flags & ZVOL_DUMPIFIED; 1090 1091 if (zv == NULL) { 1092 bioerror(bp, ENXIO); 1093 biodone(bp); 1094 return (0); 1095 } 1096 1097 if (getminor(bp->b_edev) == 0) { 1098 bioerror(bp, EINVAL); 1099 biodone(bp); 1100 return (0); 1101 } 1102 1103 if (!(bp->b_flags & B_READ) && 1104 (zv->zv_flags & ZVOL_RDONLY || 1105 zv->zv_mode & DS_MODE_READONLY)) { 1106 bioerror(bp, EROFS); 1107 biodone(bp); 1108 return (0); 1109 } 1110 1111 off = ldbtob(bp->b_blkno); 1112 volsize = zv->zv_volsize; 1113 1114 os = zv->zv_objset; 1115 ASSERT(os != NULL); 1116 1117 bp_mapin(bp); 1118 addr = bp->b_un.b_addr; 1119 resid = bp->b_bcount; 1120 1121 if (resid > 0 && (off < 0 || off >= volsize)) { 1122 bioerror(bp, EIO); 1123 biodone(bp); 1124 return (0); 1125 } 1126 1127 /* 1128 * There must be no buffer changes when doing a dmu_sync() because 1129 * we can't change the data whilst calculating the checksum. 1130 */ 1131 rl = zfs_range_lock(&zv->zv_znode, off, resid, 1132 doread ? RL_READER : RL_WRITER); 1133 1134 while (resid != 0 && off < volsize) { 1135 size_t size = MIN(resid, zvol_maxphys); 1136 if (is_dump) { 1137 size = MIN(size, P2END(off, zv->zv_volblocksize) - off); 1138 error = zvol_dumpio(zv, addr, off, size, 1139 doread, B_FALSE); 1140 } else if (doread) { 1141 error = dmu_read(os, ZVOL_OBJ, off, size, addr); 1142 } else { 1143 dmu_tx_t *tx = dmu_tx_create(os); 1144 dmu_tx_hold_write(tx, ZVOL_OBJ, off, size); 1145 error = dmu_tx_assign(tx, TXG_WAIT); 1146 if (error) { 1147 dmu_tx_abort(tx); 1148 } else { 1149 dmu_write(os, ZVOL_OBJ, off, size, addr, tx); 1150 zvol_log_write(zv, tx, off, size); 1151 dmu_tx_commit(tx); 1152 } 1153 } 1154 if (error) { 1155 /* convert checksum errors into IO errors */ 1156 if (error == ECKSUM) 1157 error = EIO; 1158 break; 1159 } 1160 off += size; 1161 addr += size; 1162 resid -= size; 1163 } 1164 zfs_range_unlock(rl); 1165 1166 if ((bp->b_resid = resid) == bp->b_bcount) 1167 bioerror(bp, off > volsize ? EINVAL : error); 1168 1169 if (!(bp->b_flags & B_ASYNC) && !doread && !zil_disable && 1170 !is_dump && !(zv->zv_flags & ZVOL_WCE)) 1171 zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ); 1172 biodone(bp); 1173 1174 return (0); 1175 } 1176 1177 /* 1178 * Set the buffer count to the zvol maximum transfer. 1179 * Using our own routine instead of the default minphys() 1180 * means that for larger writes we write bigger buffers on X86 1181 * (128K instead of 56K) and flush the disk write cache less often 1182 * (every zvol_maxphys - currently 1MB) instead of minphys (currently 1183 * 56K on X86 and 128K on sparc). 1184 */ 1185 void 1186 zvol_minphys(struct buf *bp) 1187 { 1188 if (bp->b_bcount > zvol_maxphys) 1189 bp->b_bcount = zvol_maxphys; 1190 } 1191 1192 int 1193 zvol_dump(dev_t dev, caddr_t addr, daddr_t blkno, int nblocks) 1194 { 1195 minor_t minor = getminor(dev); 1196 zvol_state_t *zv; 1197 int error = 0; 1198 uint64_t size; 1199 uint64_t boff; 1200 uint64_t resid; 1201 1202 if (minor == 0) /* This is the control device */ 1203 return (ENXIO); 1204 1205 zv = ddi_get_soft_state(zvol_state, minor); 1206 if (zv == NULL) 1207 return (ENXIO); 1208 1209 boff = ldbtob(blkno); 1210 resid = ldbtob(nblocks); 1211 1212 VERIFY3U(boff + resid, <=, zv->zv_volsize); 1213 1214 while (resid) { 1215 size = MIN(resid, P2END(boff, zv->zv_volblocksize) - boff); 1216 error = zvol_dumpio(zv, addr, boff, size, B_FALSE, B_TRUE); 1217 if (error) 1218 break; 1219 boff += size; 1220 addr += size; 1221 resid -= size; 1222 } 1223 1224 return (error); 1225 } 1226 1227 /*ARGSUSED*/ 1228 int 1229 zvol_read(dev_t dev, uio_t *uio, cred_t *cr) 1230 { 1231 minor_t minor = getminor(dev); 1232 zvol_state_t *zv; 1233 uint64_t volsize; 1234 rl_t *rl; 1235 int error = 0; 1236 1237 if (minor == 0) /* This is the control device */ 1238 return (ENXIO); 1239 1240 zv = ddi_get_soft_state(zvol_state, minor); 1241 if (zv == NULL) 1242 return (ENXIO); 1243 1244 volsize = zv->zv_volsize; 1245 if (uio->uio_resid > 0 && 1246 (uio->uio_loffset < 0 || uio->uio_loffset >= volsize)) 1247 return (EIO); 1248 1249 if (zv->zv_flags & ZVOL_DUMPIFIED) { 1250 error = physio(zvol_strategy, NULL, dev, B_READ, 1251 zvol_minphys, uio); 1252 return (error); 1253 } 1254 1255 rl = zfs_range_lock(&zv->zv_znode, uio->uio_loffset, uio->uio_resid, 1256 RL_READER); 1257 while (uio->uio_resid > 0 && uio->uio_loffset < volsize) { 1258 uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1); 1259 1260 /* don't read past the end */ 1261 if (bytes > volsize - uio->uio_loffset) 1262 bytes = volsize - uio->uio_loffset; 1263 1264 error = dmu_read_uio(zv->zv_objset, ZVOL_OBJ, uio, bytes); 1265 if (error) { 1266 /* convert checksum errors into IO errors */ 1267 if (error == ECKSUM) 1268 error = EIO; 1269 break; 1270 } 1271 } 1272 zfs_range_unlock(rl); 1273 return (error); 1274 } 1275 1276 /*ARGSUSED*/ 1277 int 1278 zvol_write(dev_t dev, uio_t *uio, cred_t *cr) 1279 { 1280 minor_t minor = getminor(dev); 1281 zvol_state_t *zv; 1282 uint64_t volsize; 1283 rl_t *rl; 1284 int error = 0; 1285 1286 if (minor == 0) /* This is the control device */ 1287 return (ENXIO); 1288 1289 zv = ddi_get_soft_state(zvol_state, minor); 1290 if (zv == NULL) 1291 return (ENXIO); 1292 1293 volsize = zv->zv_volsize; 1294 if (uio->uio_resid > 0 && 1295 (uio->uio_loffset < 0 || uio->uio_loffset >= volsize)) 1296 return (EIO); 1297 1298 if (zv->zv_flags & ZVOL_DUMPIFIED) { 1299 error = physio(zvol_strategy, NULL, dev, B_WRITE, 1300 zvol_minphys, uio); 1301 return (error); 1302 } 1303 1304 rl = zfs_range_lock(&zv->zv_znode, uio->uio_loffset, uio->uio_resid, 1305 RL_WRITER); 1306 while (uio->uio_resid > 0 && uio->uio_loffset < volsize) { 1307 uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1); 1308 uint64_t off = uio->uio_loffset; 1309 dmu_tx_t *tx = dmu_tx_create(zv->zv_objset); 1310 1311 if (bytes > volsize - off) /* don't write past the end */ 1312 bytes = volsize - off; 1313 1314 dmu_tx_hold_write(tx, ZVOL_OBJ, off, bytes); 1315 error = dmu_tx_assign(tx, TXG_WAIT); 1316 if (error) { 1317 dmu_tx_abort(tx); 1318 break; 1319 } 1320 error = dmu_write_uio(zv->zv_objset, ZVOL_OBJ, uio, bytes, tx); 1321 if (error == 0) 1322 zvol_log_write(zv, tx, off, bytes); 1323 dmu_tx_commit(tx); 1324 1325 if (error) 1326 break; 1327 } 1328 zfs_range_unlock(rl); 1329 if (!zil_disable && !(zv->zv_flags & ZVOL_WCE)) 1330 zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ); 1331 return (error); 1332 } 1333 1334 int 1335 zvol_getefi(void *arg, int flag, uint64_t vs, uint8_t bs) 1336 { 1337 struct uuid uuid = EFI_RESERVED; 1338 efi_gpe_t gpe = { 0 }; 1339 uint32_t crc; 1340 dk_efi_t efi; 1341 int length; 1342 char *ptr; 1343 1344 if (ddi_copyin(arg, &efi, sizeof (dk_efi_t), flag)) 1345 return (EFAULT); 1346 ptr = (char *)(uintptr_t)efi.dki_data_64; 1347 length = efi.dki_length; 1348 /* 1349 * Some clients may attempt to request a PMBR for the 1350 * zvol. Currently this interface will return EINVAL to 1351 * such requests. These requests could be supported by 1352 * adding a check for lba == 0 and consing up an appropriate 1353 * PMBR. 1354 */ 1355 if (efi.dki_lba < 1 || efi.dki_lba > 2 || length <= 0) 1356 return (EINVAL); 1357 1358 gpe.efi_gpe_StartingLBA = LE_64(34ULL); 1359 gpe.efi_gpe_EndingLBA = LE_64((vs >> bs) - 1); 1360 UUID_LE_CONVERT(gpe.efi_gpe_PartitionTypeGUID, uuid); 1361 1362 if (efi.dki_lba == 1) { 1363 efi_gpt_t gpt = { 0 }; 1364 1365 gpt.efi_gpt_Signature = LE_64(EFI_SIGNATURE); 1366 gpt.efi_gpt_Revision = LE_32(EFI_VERSION_CURRENT); 1367 gpt.efi_gpt_HeaderSize = LE_32(sizeof (gpt)); 1368 gpt.efi_gpt_MyLBA = LE_64(1ULL); 1369 gpt.efi_gpt_FirstUsableLBA = LE_64(34ULL); 1370 gpt.efi_gpt_LastUsableLBA = LE_64((vs >> bs) - 1); 1371 gpt.efi_gpt_PartitionEntryLBA = LE_64(2ULL); 1372 gpt.efi_gpt_NumberOfPartitionEntries = LE_32(1); 1373 gpt.efi_gpt_SizeOfPartitionEntry = 1374 LE_32(sizeof (efi_gpe_t)); 1375 CRC32(crc, &gpe, sizeof (gpe), -1U, crc32_table); 1376 gpt.efi_gpt_PartitionEntryArrayCRC32 = LE_32(~crc); 1377 CRC32(crc, &gpt, sizeof (gpt), -1U, crc32_table); 1378 gpt.efi_gpt_HeaderCRC32 = LE_32(~crc); 1379 if (ddi_copyout(&gpt, ptr, MIN(sizeof (gpt), length), 1380 flag)) 1381 return (EFAULT); 1382 ptr += sizeof (gpt); 1383 length -= sizeof (gpt); 1384 } 1385 if (length > 0 && ddi_copyout(&gpe, ptr, MIN(sizeof (gpe), 1386 length), flag)) 1387 return (EFAULT); 1388 return (0); 1389 } 1390 1391 /* 1392 * Dirtbag ioctls to support mkfs(1M) for UFS filesystems. See dkio(7I). 1393 */ 1394 /*ARGSUSED*/ 1395 int 1396 zvol_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 1397 { 1398 zvol_state_t *zv; 1399 struct dk_cinfo dki; 1400 struct dk_minfo dkm; 1401 struct dk_callback *dkc; 1402 int error = 0; 1403 rl_t *rl; 1404 1405 mutex_enter(&zvol_state_lock); 1406 1407 zv = ddi_get_soft_state(zvol_state, getminor(dev)); 1408 1409 if (zv == NULL) { 1410 mutex_exit(&zvol_state_lock); 1411 return (ENXIO); 1412 } 1413 ASSERT(zv->zv_total_opens > 0); 1414 1415 switch (cmd) { 1416 1417 case DKIOCINFO: 1418 bzero(&dki, sizeof (dki)); 1419 (void) strcpy(dki.dki_cname, "zvol"); 1420 (void) strcpy(dki.dki_dname, "zvol"); 1421 dki.dki_ctype = DKC_UNKNOWN; 1422 dki.dki_maxtransfer = 1 << (SPA_MAXBLOCKSHIFT - zv->zv_min_bs); 1423 mutex_exit(&zvol_state_lock); 1424 if (ddi_copyout(&dki, (void *)arg, sizeof (dki), flag)) 1425 error = EFAULT; 1426 return (error); 1427 1428 case DKIOCGMEDIAINFO: 1429 bzero(&dkm, sizeof (dkm)); 1430 dkm.dki_lbsize = 1U << zv->zv_min_bs; 1431 dkm.dki_capacity = zv->zv_volsize >> zv->zv_min_bs; 1432 dkm.dki_media_type = DK_UNKNOWN; 1433 mutex_exit(&zvol_state_lock); 1434 if (ddi_copyout(&dkm, (void *)arg, sizeof (dkm), flag)) 1435 error = EFAULT; 1436 return (error); 1437 1438 case DKIOCGETEFI: 1439 { 1440 uint64_t vs = zv->zv_volsize; 1441 uint8_t bs = zv->zv_min_bs; 1442 1443 mutex_exit(&zvol_state_lock); 1444 error = zvol_getefi((void *)arg, flag, vs, bs); 1445 return (error); 1446 } 1447 1448 case DKIOCFLUSHWRITECACHE: 1449 dkc = (struct dk_callback *)arg; 1450 mutex_exit(&zvol_state_lock); 1451 zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ); 1452 if ((flag & FKIOCTL) && dkc != NULL && dkc->dkc_callback) { 1453 (*dkc->dkc_callback)(dkc->dkc_cookie, error); 1454 error = 0; 1455 } 1456 return (error); 1457 1458 case DKIOCGETWCE: 1459 { 1460 int wce = (zv->zv_flags & ZVOL_WCE) ? 1 : 0; 1461 if (ddi_copyout(&wce, (void *)arg, sizeof (int), 1462 flag)) 1463 error = EFAULT; 1464 break; 1465 } 1466 case DKIOCSETWCE: 1467 { 1468 int wce; 1469 if (ddi_copyin((void *)arg, &wce, sizeof (int), 1470 flag)) { 1471 error = EFAULT; 1472 break; 1473 } 1474 if (wce) { 1475 zv->zv_flags |= ZVOL_WCE; 1476 mutex_exit(&zvol_state_lock); 1477 } else { 1478 zv->zv_flags &= ~ZVOL_WCE; 1479 mutex_exit(&zvol_state_lock); 1480 zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ); 1481 } 1482 return (0); 1483 } 1484 1485 case DKIOCGGEOM: 1486 case DKIOCGVTOC: 1487 /* 1488 * commands using these (like prtvtoc) expect ENOTSUP 1489 * since we're emulating an EFI label 1490 */ 1491 error = ENOTSUP; 1492 break; 1493 1494 case DKIOCDUMPINIT: 1495 rl = zfs_range_lock(&zv->zv_znode, 0, zv->zv_volsize, 1496 RL_WRITER); 1497 error = zvol_dumpify(zv); 1498 zfs_range_unlock(rl); 1499 break; 1500 1501 case DKIOCDUMPFINI: 1502 if (!(zv->zv_flags & ZVOL_DUMPIFIED)) 1503 break; 1504 rl = zfs_range_lock(&zv->zv_znode, 0, zv->zv_volsize, 1505 RL_WRITER); 1506 error = zvol_dump_fini(zv); 1507 zfs_range_unlock(rl); 1508 break; 1509 1510 default: 1511 error = ENOTTY; 1512 break; 1513 1514 } 1515 mutex_exit(&zvol_state_lock); 1516 return (error); 1517 } 1518 1519 int 1520 zvol_busy(void) 1521 { 1522 return (zvol_minors != 0); 1523 } 1524 1525 void 1526 zvol_init(void) 1527 { 1528 VERIFY(ddi_soft_state_init(&zvol_state, sizeof (zvol_state_t), 1) == 0); 1529 mutex_init(&zvol_state_lock, NULL, MUTEX_DEFAULT, NULL); 1530 } 1531 1532 void 1533 zvol_fini(void) 1534 { 1535 mutex_destroy(&zvol_state_lock); 1536 ddi_soft_state_fini(&zvol_state); 1537 } 1538 1539 static boolean_t 1540 zvol_is_swap(zvol_state_t *zv) 1541 { 1542 vnode_t *vp; 1543 boolean_t ret = B_FALSE; 1544 char *devpath; 1545 size_t devpathlen; 1546 int error; 1547 1548 devpathlen = strlen(ZVOL_FULL_DEV_DIR) + strlen(zv->zv_name) + 1; 1549 devpath = kmem_alloc(devpathlen, KM_SLEEP); 1550 (void) sprintf(devpath, "%s%s", ZVOL_FULL_DEV_DIR, zv->zv_name); 1551 error = lookupname(devpath, UIO_SYSSPACE, FOLLOW, NULLVPP, &vp); 1552 kmem_free(devpath, devpathlen); 1553 1554 ret = !error && IS_SWAPVP(common_specvp(vp)); 1555 1556 if (vp != NULL) 1557 VN_RELE(vp); 1558 1559 return (ret); 1560 } 1561 1562 static int 1563 zvol_dump_init(zvol_state_t *zv, boolean_t resize) 1564 { 1565 dmu_tx_t *tx; 1566 int error = 0; 1567 objset_t *os = zv->zv_objset; 1568 nvlist_t *nv = NULL; 1569 1570 ASSERT(MUTEX_HELD(&zvol_state_lock)); 1571 1572 tx = dmu_tx_create(os); 1573 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 1574 error = dmu_tx_assign(tx, TXG_WAIT); 1575 if (error) { 1576 dmu_tx_abort(tx); 1577 return (error); 1578 } 1579 1580 /* 1581 * If we are resizing the dump device then we only need to 1582 * update the refreservation to match the newly updated 1583 * zvolsize. Otherwise, we save off the original state of the 1584 * zvol so that we can restore them if the zvol is ever undumpified. 1585 */ 1586 if (resize) { 1587 error = zap_update(os, ZVOL_ZAP_OBJ, 1588 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, 1589 &zv->zv_volsize, tx); 1590 } else { 1591 uint64_t checksum, compress, refresrv, vbs; 1592 1593 error = dsl_prop_get_integer(zv->zv_name, 1594 zfs_prop_to_name(ZFS_PROP_COMPRESSION), &compress, NULL); 1595 error = error ? error : dsl_prop_get_integer(zv->zv_name, 1596 zfs_prop_to_name(ZFS_PROP_CHECKSUM), &checksum, NULL); 1597 error = error ? error : dsl_prop_get_integer(zv->zv_name, 1598 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &refresrv, NULL); 1599 error = error ? error : dsl_prop_get_integer(zv->zv_name, 1600 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &vbs, NULL); 1601 1602 error = error ? error : zap_update(os, ZVOL_ZAP_OBJ, 1603 zfs_prop_to_name(ZFS_PROP_COMPRESSION), 8, 1, 1604 &compress, tx); 1605 error = error ? error : zap_update(os, ZVOL_ZAP_OBJ, 1606 zfs_prop_to_name(ZFS_PROP_CHECKSUM), 8, 1, &checksum, tx); 1607 error = error ? error : zap_update(os, ZVOL_ZAP_OBJ, 1608 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, 1609 &refresrv, tx); 1610 error = error ? error : zap_update(os, ZVOL_ZAP_OBJ, 1611 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 8, 1, 1612 &vbs, tx); 1613 } 1614 dmu_tx_commit(tx); 1615 1616 /* Truncate the file */ 1617 if (!error) 1618 error = dmu_free_long_range(zv->zv_objset, 1619 ZVOL_OBJ, 0, DMU_OBJECT_END); 1620 1621 if (error) 1622 return (error); 1623 1624 /* 1625 * We only need update the zvol's property if we are initializing 1626 * the dump area for the first time. 1627 */ 1628 if (!resize) { 1629 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1630 VERIFY(nvlist_add_uint64(nv, 1631 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 0) == 0); 1632 VERIFY(nvlist_add_uint64(nv, 1633 zfs_prop_to_name(ZFS_PROP_COMPRESSION), 1634 ZIO_COMPRESS_OFF) == 0); 1635 VERIFY(nvlist_add_uint64(nv, 1636 zfs_prop_to_name(ZFS_PROP_CHECKSUM), 1637 ZIO_CHECKSUM_OFF) == 0); 1638 VERIFY(nvlist_add_uint64(nv, 1639 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 1640 SPA_MAXBLOCKSIZE) == 0); 1641 1642 error = zfs_set_prop_nvlist(zv->zv_name, nv); 1643 nvlist_free(nv); 1644 1645 if (error) 1646 return (error); 1647 } 1648 1649 /* Allocate the space for the dump */ 1650 error = zvol_prealloc(zv); 1651 return (error); 1652 } 1653 1654 static int 1655 zvol_dumpify(zvol_state_t *zv) 1656 { 1657 int error = 0; 1658 uint64_t dumpsize = 0; 1659 dmu_tx_t *tx; 1660 objset_t *os = zv->zv_objset; 1661 1662 if (zv->zv_flags & ZVOL_RDONLY || (zv->zv_mode & DS_MODE_READONLY)) 1663 return (EROFS); 1664 1665 /* 1666 * We do not support swap devices acting as dump devices. 1667 */ 1668 if (zvol_is_swap(zv)) 1669 return (ENOTSUP); 1670 1671 if (zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, 1672 8, 1, &dumpsize) != 0 || dumpsize != zv->zv_volsize) { 1673 boolean_t resize = (dumpsize > 0) ? B_TRUE : B_FALSE; 1674 1675 if ((error = zvol_dump_init(zv, resize)) != 0) { 1676 (void) zvol_dump_fini(zv); 1677 return (error); 1678 } 1679 } 1680 1681 /* 1682 * Build up our lba mapping. 1683 */ 1684 error = zvol_get_lbas(zv); 1685 if (error) { 1686 (void) zvol_dump_fini(zv); 1687 return (error); 1688 } 1689 1690 tx = dmu_tx_create(os); 1691 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 1692 error = dmu_tx_assign(tx, TXG_WAIT); 1693 if (error) { 1694 dmu_tx_abort(tx); 1695 (void) zvol_dump_fini(zv); 1696 return (error); 1697 } 1698 1699 zv->zv_flags |= ZVOL_DUMPIFIED; 1700 error = zap_update(os, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, 8, 1, 1701 &zv->zv_volsize, tx); 1702 dmu_tx_commit(tx); 1703 1704 if (error) { 1705 (void) zvol_dump_fini(zv); 1706 return (error); 1707 } 1708 1709 txg_wait_synced(dmu_objset_pool(os), 0); 1710 return (0); 1711 } 1712 1713 static int 1714 zvol_dump_fini(zvol_state_t *zv) 1715 { 1716 dmu_tx_t *tx; 1717 objset_t *os = zv->zv_objset; 1718 nvlist_t *nv; 1719 int error = 0; 1720 uint64_t checksum, compress, refresrv, vbs; 1721 1722 /* 1723 * Attempt to restore the zvol back to its pre-dumpified state. 1724 * This is a best-effort attempt as it's possible that not all 1725 * of these properties were initialized during the dumpify process 1726 * (i.e. error during zvol_dump_init). 1727 */ 1728 1729 tx = dmu_tx_create(os); 1730 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 1731 error = dmu_tx_assign(tx, TXG_WAIT); 1732 if (error) { 1733 dmu_tx_abort(tx); 1734 return (error); 1735 } 1736 (void) zap_remove(os, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, tx); 1737 dmu_tx_commit(tx); 1738 1739 (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, 1740 zfs_prop_to_name(ZFS_PROP_CHECKSUM), 8, 1, &checksum); 1741 (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, 1742 zfs_prop_to_name(ZFS_PROP_COMPRESSION), 8, 1, &compress); 1743 (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, 1744 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, &refresrv); 1745 (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, 1746 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 8, 1, &vbs); 1747 1748 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1749 (void) nvlist_add_uint64(nv, 1750 zfs_prop_to_name(ZFS_PROP_CHECKSUM), checksum); 1751 (void) nvlist_add_uint64(nv, 1752 zfs_prop_to_name(ZFS_PROP_COMPRESSION), compress); 1753 (void) nvlist_add_uint64(nv, 1754 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), refresrv); 1755 (void) nvlist_add_uint64(nv, 1756 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), vbs); 1757 (void) zfs_set_prop_nvlist(zv->zv_name, nv); 1758 nvlist_free(nv); 1759 1760 zvol_free_extents(zv); 1761 zv->zv_flags &= ~ZVOL_DUMPIFIED; 1762 (void) dmu_free_long_range(os, ZVOL_OBJ, 0, DMU_OBJECT_END); 1763 1764 return (0); 1765 } 1766