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 2010 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 /dev filesystem (sdev_zvolops.c). 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/disk.h> 78 #include <sys/dkio.h> 79 #include <sys/disklabel.h> 80 81 #ifdef __NetBSD__ 82 #include <prop/proplib.h> 83 #endif 84 #include <sys/zil_impl.h> 85 86 #include "zfs_namecheck.h" 87 88 static void *zvol_state; 89 static char *zvol_tag = "zvol_tag"; 90 91 #define ZVOL_DUMPSIZE "dumpsize" 92 93 void zvol_minphys(struct buf *); 94 95 static struct dkdriver zvol_dkdriver = { zvol_strategy, zvol_minphys }; 96 97 /* 98 * This lock protects the zvol_state structure from being modified 99 * while it's being used, e.g. an open that comes in before a create 100 * finishes. It also protects temporary opens of the dataset so that, 101 * e.g., an open doesn't get a spurious EBUSY. 102 */ 103 static kmutex_t zvol_state_lock; 104 static uint32_t zvol_minors; 105 106 typedef struct zvol_extent { 107 list_node_t ze_node; 108 dva_t ze_dva; /* dva associated with this extent */ 109 uint64_t ze_nblks; /* number of blocks in extent */ 110 } zvol_extent_t; 111 112 /* 113 * The in-core state of each volume. 114 */ 115 typedef struct zvol_state { 116 char zv_name[MAXPATHLEN]; /* pool/dd name */ 117 uint64_t zv_volsize; /* amount of space we advertise */ 118 uint64_t zv_volblocksize; /* volume block size */ 119 minor_t zv_minor; /* minor number */ 120 uint8_t zv_min_bs; /* minimum addressable block shift */ 121 uint8_t zv_flags; /* readonly, dumpified, etc. */ 122 objset_t *zv_objset; /* objset handle */ 123 uint32_t zv_open_count[OTYPCNT]; /* open counts */ 124 uint32_t zv_total_opens; /* total open count */ 125 zilog_t *zv_zilog; /* ZIL handle */ 126 list_t zv_extents; /* List of extents for dump */ 127 znode_t zv_znode; /* for range locking */ 128 #ifdef __NetBSD__ 129 struct disk zv_dk; /* disk statistics */ 130 kmutex_t zv_dklock; /* disk statistics */ 131 #endif 132 } zvol_state_t; 133 134 /* 135 * zvol specific flags 136 */ 137 #define ZVOL_RDONLY 0x1 138 #define ZVOL_DUMPIFIED 0x2 139 #define ZVOL_EXCL 0x4 140 #define ZVOL_WCE 0x8 141 142 /* 143 * zvol maximum transfer in one DMU tx. 144 */ 145 int zvol_maxphys = DMU_MAX_ACCESS/2; 146 147 extern int zfs_set_prop_nvlist(const char *, zprop_source_t, 148 nvlist_t *, nvlist_t **); 149 static int zvol_remove_zv(zvol_state_t *); 150 static int zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio); 151 static int zvol_dumpify(zvol_state_t *zv); 152 static int zvol_dump_fini(zvol_state_t *zv); 153 static int zvol_dump_init(zvol_state_t *zv, boolean_t resize); 154 155 static void 156 zvol_size_changed(zvol_state_t *zv) 157 { 158 prop_dictionary_t disk_info, odisk_info, geom; 159 struct disk *disk; 160 161 disk = &zv->zv_dk; 162 163 disk_info = prop_dictionary_create(); 164 geom = prop_dictionary_create(); 165 166 prop_dictionary_set_cstring_nocopy(disk_info, "type", "ESDI"); 167 prop_dictionary_set_uint64(geom, "sectors-per-unit", zv->zv_volsize); 168 prop_dictionary_set_uint32(geom, "sector-size", 169 DEV_BSIZE /* XXX 512? */); 170 prop_dictionary_set_uint32(geom, "sectors-per-track", 32); 171 prop_dictionary_set_uint32(geom, "tracks-per-cylinder", 64); 172 prop_dictionary_set_uint32(geom, "cylinders-per-unit", zv->zv_volsize / 2048); 173 prop_dictionary_set(disk_info, "geometry", geom); 174 prop_object_release(geom); 175 176 odisk_info = disk->dk_info; 177 disk->dk_info = disk_info; 178 179 if (odisk_info != NULL) 180 prop_object_release(odisk_info); 181 } 182 183 int 184 zvol_check_volsize(uint64_t volsize, uint64_t blocksize) 185 { 186 if (volsize == 0) 187 return (EINVAL); 188 189 if (volsize % blocksize != 0) 190 return (EINVAL); 191 192 #ifdef _ILP32 193 if (volsize - 1 > SPEC_MAXOFFSET_T) 194 return (EOVERFLOW); 195 #endif 196 return (0); 197 } 198 199 int 200 zvol_check_volblocksize(uint64_t volblocksize) 201 { 202 if (volblocksize < SPA_MINBLOCKSIZE || 203 volblocksize > SPA_MAXBLOCKSIZE || 204 !ISP2(volblocksize)) 205 return (EDOM); 206 207 return (0); 208 } 209 210 int 211 zvol_get_stats(objset_t *os, nvlist_t *nv) 212 { 213 int error; 214 dmu_object_info_t doi; 215 uint64_t val; 216 217 error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &val); 218 if (error) 219 return (error); 220 221 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLSIZE, val); 222 223 error = dmu_object_info(os, ZVOL_OBJ, &doi); 224 225 if (error == 0) { 226 dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLBLOCKSIZE, 227 doi.doi_data_block_size); 228 } 229 230 return (error); 231 } 232 233 /* 234 * Find a free minor number. 235 */ 236 static minor_t 237 zvol_minor_alloc(void) 238 { 239 minor_t minor; 240 241 ASSERT(MUTEX_HELD(&zvol_state_lock)); 242 243 for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) 244 if (ddi_get_soft_state(zvol_state, minor) == NULL) 245 return (minor); 246 247 return (0); 248 } 249 250 static zvol_state_t * 251 zvol_minor_lookup(const char *name) 252 { 253 minor_t minor; 254 zvol_state_t *zv; 255 256 ASSERT(MUTEX_HELD(&zvol_state_lock)); 257 258 for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) { 259 zv = ddi_get_soft_state(zvol_state, minor); 260 if (zv == NULL) 261 continue; 262 if (strcmp(zv->zv_name, name) == 0) 263 break; 264 } 265 266 return (zv); 267 } 268 269 /* extent mapping arg */ 270 struct maparg { 271 zvol_state_t *ma_zv; 272 uint64_t ma_blks; 273 }; 274 275 /*ARGSUSED*/ 276 static int 277 zvol_map_block(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, 278 const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg) 279 { 280 struct maparg *ma = arg; 281 zvol_extent_t *ze; 282 int bs = ma->ma_zv->zv_volblocksize; 283 284 if (bp == NULL || zb->zb_object != ZVOL_OBJ || zb->zb_level != 0) 285 return (0); 286 287 VERIFY3U(ma->ma_blks, ==, zb->zb_blkid); 288 ma->ma_blks++; 289 290 /* Abort immediately if we have encountered gang blocks */ 291 if (BP_IS_GANG(bp)) 292 return (EFRAGS); 293 294 /* 295 * See if the block is at the end of the previous extent. 296 */ 297 ze = list_tail(&ma->ma_zv->zv_extents); 298 if (ze && 299 DVA_GET_VDEV(BP_IDENTITY(bp)) == DVA_GET_VDEV(&ze->ze_dva) && 300 DVA_GET_OFFSET(BP_IDENTITY(bp)) == 301 DVA_GET_OFFSET(&ze->ze_dva) + ze->ze_nblks * bs) { 302 ze->ze_nblks++; 303 return (0); 304 } 305 306 dprintf_bp(bp, "%s", "next blkptr:"); 307 308 /* start a new extent */ 309 ze = kmem_zalloc(sizeof (zvol_extent_t), KM_SLEEP); 310 ze->ze_dva = bp->blk_dva[0]; /* structure assignment */ 311 ze->ze_nblks = 1; 312 list_insert_tail(&ma->ma_zv->zv_extents, ze); 313 return (0); 314 } 315 316 static void 317 zvol_free_extents(zvol_state_t *zv) 318 { 319 zvol_extent_t *ze; 320 321 while (ze = list_head(&zv->zv_extents)) { 322 list_remove(&zv->zv_extents, ze); 323 kmem_free(ze, sizeof (zvol_extent_t)); 324 } 325 } 326 327 static int 328 zvol_get_lbas(zvol_state_t *zv) 329 { 330 objset_t *os = zv->zv_objset; 331 struct maparg ma; 332 int err; 333 334 ma.ma_zv = zv; 335 ma.ma_blks = 0; 336 zvol_free_extents(zv); 337 338 /* commit any in-flight changes before traversing the dataset */ 339 txg_wait_synced(dmu_objset_pool(os), 0); 340 err = traverse_dataset(dmu_objset_ds(os), 0, 341 TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA, zvol_map_block, &ma); 342 if (err || ma.ma_blks != (zv->zv_volsize / zv->zv_volblocksize)) { 343 zvol_free_extents(zv); 344 return (err ? err : EIO); 345 } 346 347 return (0); 348 } 349 350 /* ARGSUSED */ 351 void 352 zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) 353 { 354 zfs_creat_t *zct = arg; 355 nvlist_t *nvprops = zct->zct_props; 356 int error; 357 uint64_t volblocksize, volsize; 358 359 VERIFY(nvlist_lookup_uint64(nvprops, 360 zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0); 361 if (nvlist_lookup_uint64(nvprops, 362 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0) 363 volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE); 364 365 /* 366 * These properties must be removed from the list so the generic 367 * property setting step won't apply to them. 368 */ 369 VERIFY(nvlist_remove_all(nvprops, 370 zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0); 371 (void) nvlist_remove_all(nvprops, 372 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE)); 373 374 error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize, 375 DMU_OT_NONE, 0, tx); 376 ASSERT(error == 0); 377 378 error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP, 379 DMU_OT_NONE, 0, tx); 380 ASSERT(error == 0); 381 382 error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx); 383 ASSERT(error == 0); 384 } 385 386 /* 387 * Replay a TX_WRITE ZIL transaction that didn't get committed 388 * after a system failure 389 */ 390 static int 391 zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap) 392 { 393 objset_t *os = zv->zv_objset; 394 char *data = (char *)(lr + 1); /* data follows lr_write_t */ 395 uint64_t offset, length; 396 dmu_tx_t *tx; 397 int error; 398 399 if (byteswap) 400 byteswap_uint64_array(lr, sizeof (*lr)); 401 402 offset = lr->lr_offset; 403 length = lr->lr_length; 404 405 /* If it's a dmu_sync() block, write the whole block */ 406 if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { 407 uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); 408 if (length < blocksize) { 409 offset -= offset % blocksize; 410 length = blocksize; 411 } 412 } 413 414 tx = dmu_tx_create(os); 415 dmu_tx_hold_write(tx, ZVOL_OBJ, offset, length); 416 error = dmu_tx_assign(tx, TXG_WAIT); 417 if (error) { 418 dmu_tx_abort(tx); 419 } else { 420 dmu_write(os, ZVOL_OBJ, offset, length, data, tx); 421 dmu_tx_commit(tx); 422 } 423 424 return (error); 425 } 426 427 /* ARGSUSED */ 428 static int 429 zvol_replay_err(zvol_state_t *zv, lr_t *lr, boolean_t byteswap) 430 { 431 return (ENOTSUP); 432 } 433 434 /* 435 * Callback vectors for replaying records. 436 * Only TX_WRITE is needed for zvol. 437 */ 438 zil_replay_func_t *zvol_replay_vector[TX_MAX_TYPE] = { 439 zvol_replay_err, /* 0 no such transaction type */ 440 zvol_replay_err, /* TX_CREATE */ 441 zvol_replay_err, /* TX_MKDIR */ 442 zvol_replay_err, /* TX_MKXATTR */ 443 zvol_replay_err, /* TX_SYMLINK */ 444 zvol_replay_err, /* TX_REMOVE */ 445 zvol_replay_err, /* TX_RMDIR */ 446 zvol_replay_err, /* TX_LINK */ 447 zvol_replay_err, /* TX_RENAME */ 448 zvol_replay_write, /* TX_WRITE */ 449 zvol_replay_err, /* TX_TRUNCATE */ 450 zvol_replay_err, /* TX_SETATTR */ 451 zvol_replay_err, /* TX_ACL */ 452 zvol_replay_err, /* TX_CREATE_ACL */ 453 zvol_replay_err, /* TX_CREATE_ATTR */ 454 zvol_replay_err, /* TX_CREATE_ACL_ATTR */ 455 zvol_replay_err, /* TX_MKDIR_ACL */ 456 zvol_replay_err, /* TX_MKDIR_ATTR */ 457 zvol_replay_err, /* TX_MKDIR_ACL_ATTR */ 458 zvol_replay_err, /* TX_WRITE2 */ 459 }; 460 461 int 462 zvol_name2minor(const char *name, minor_t *minor) 463 { 464 zvol_state_t *zv; 465 466 mutex_enter(&zvol_state_lock); 467 zv = zvol_minor_lookup(name); 468 if (minor && zv) 469 *minor = zv->zv_minor; 470 mutex_exit(&zvol_state_lock); 471 return (zv ? 0 : -1); 472 } 473 474 /* 475 * Create a minor node (plus a whole lot more) for the specified volume. 476 */ 477 int 478 zvol_create_minor(const char *name) 479 { 480 zvol_state_t *zv; 481 objset_t *os; 482 dmu_object_info_t doi; 483 minor_t minor = 0; 484 vnode_t *vp = NULL; 485 char *devpath; 486 size_t devpathlen = strlen(ZVOL_FULL_DEV_DIR) + strlen(name) + 1; 487 488 int error; 489 490 mutex_enter(&zvol_state_lock); 491 492 if (zvol_minor_lookup(name) != NULL) { 493 mutex_exit(&zvol_state_lock); 494 return (EEXIST); 495 } 496 497 /* lie and say we're read-only */ 498 error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os); 499 500 if (error) { 501 mutex_exit(&zvol_state_lock); 502 return (error); 503 } 504 505 /* 506 * If there's an existing /dev/zvol symlink, try to use the 507 * same minor number we used last time. 508 */ 509 devpath = kmem_alloc(devpathlen, KM_SLEEP); 510 511 /* Get full path to ZFS volume disk device */ 512 (void) sprintf(devpath, "%s/%s", ZVOL_FULL_DEV_DIR, name); 513 514 error = lookupname(devpath, UIO_SYSSPACE, NULL, &vp); 515 516 if (error == 0 && vp->v_type != VBLK) { 517 error = EINVAL; 518 } 519 520 if (error == 0) { 521 struct stat sb; 522 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 523 error = vn_stat(vp, &sb); 524 VOP_UNLOCK(vp); 525 if (error == 0) { 526 minor = getminor(sb.st_rdev); 527 } 528 } 529 530 if (vp != NULL) 531 VN_RELE(vp); 532 533 /* 534 * If we found a minor but it's already in use, we must pick a new one. 535 */ 536 if (minor != 0 && ddi_get_soft_state(zvol_state, minor) != NULL) 537 minor = 0; 538 539 if (minor == 0) 540 minor = zvol_minor_alloc(); 541 542 if (minor == 0) { 543 dmu_objset_disown(os, zvol_tag); 544 mutex_exit(&zvol_state_lock); 545 kmem_free(devpath, devpathlen); 546 return (ENXIO); 547 } 548 549 if (ddi_soft_state_zalloc(zvol_state, minor) != DDI_SUCCESS) { 550 dmu_objset_disown(os, zvol_tag); 551 mutex_exit(&zvol_state_lock); 552 kmem_free(devpath, devpathlen); 553 return (EAGAIN); 554 } 555 (void) ddi_prop_update_string(minor, zfs_dip, ZVOL_PROP_NAME, 556 (char *)name); 557 558 if (ddi_create_minor_node(zfs_dip, (char *)name, S_IFCHR, 559 minor, DDI_PSEUDO, 0) == DDI_FAILURE) { 560 ddi_soft_state_free(zvol_state, minor); 561 dmu_objset_disown(os, zvol_tag); 562 mutex_exit(&zvol_state_lock); 563 kmem_free(devpath, devpathlen); 564 return (EAGAIN); 565 } 566 567 if (ddi_create_minor_node(zfs_dip, (char *)name, S_IFBLK, 568 minor, DDI_PSEUDO, 0) == DDI_FAILURE) { 569 ddi_remove_minor_node(zfs_dip, (char *)name); 570 ddi_soft_state_free(zvol_state, minor); 571 dmu_objset_disown(os, zvol_tag); 572 mutex_exit(&zvol_state_lock); 573 kmem_free(devpath, devpathlen); 574 return (EAGAIN); 575 } 576 zv = ddi_get_soft_state(zvol_state, minor); 577 578 (void) strlcpy(zv->zv_name, name, MAXPATHLEN); 579 zv->zv_min_bs = DEV_BSHIFT; 580 zv->zv_minor = minor; 581 zv->zv_objset = os; 582 if (dmu_objset_is_snapshot(os)) 583 zv->zv_flags |= ZVOL_RDONLY; 584 mutex_init(&zv->zv_znode.z_range_lock, NULL, MUTEX_DEFAULT, NULL); 585 avl_create(&zv->zv_znode.z_range_avl, zfs_range_compare, 586 sizeof (rl_t), offsetof(rl_t, r_node)); 587 list_create(&zv->zv_extents, sizeof (zvol_extent_t), 588 offsetof(zvol_extent_t, ze_node)); 589 /* get and cache the blocksize */ 590 error = dmu_object_info(os, ZVOL_OBJ, &doi); 591 ASSERT(error == 0); 592 zv->zv_volblocksize = doi.doi_data_block_size; 593 594 disk_init(&zv->zv_dk, name, &zvol_dkdriver); 595 disk_attach(&zv->zv_dk); 596 mutex_init(&zv->zv_dklock, NULL, MUTEX_DEFAULT, NULL); 597 598 zil_replay(os, zv, zvol_replay_vector); 599 dmu_objset_disown(os, zvol_tag); 600 zv->zv_objset = NULL; 601 602 zvol_size_changed(zv); 603 604 zvol_minors++; 605 606 mutex_exit(&zvol_state_lock); 607 608 // kmem_free(devpath, devpathlen); 609 610 return (0); 611 } 612 613 /* 614 * Remove minor node for the specified volume. 615 */ 616 static int 617 zvol_remove_zv(zvol_state_t *zv) 618 { 619 char nmbuf[20]; 620 621 ASSERT(MUTEX_HELD(&zvol_state_lock)); 622 if (zv->zv_total_opens != 0) 623 return (EBUSY); 624 625 (void) snprintf(nmbuf, sizeof (nmbuf), "%u,raw", zv->zv_minor); 626 ddi_remove_minor_node(zfs_dip, nmbuf); 627 628 (void) snprintf(nmbuf, sizeof (nmbuf), "%u", zv->zv_minor); 629 ddi_remove_minor_node(zfs_dip, nmbuf); 630 631 avl_destroy(&zv->zv_znode.z_range_avl); 632 mutex_destroy(&zv->zv_znode.z_range_lock); 633 634 ddi_soft_state_free(zvol_state, zv->zv_minor); 635 636 zvol_minors--; 637 return (0); 638 } 639 640 int 641 zvol_remove_minor(const char *name) 642 { 643 zvol_state_t *zv; 644 int rc; 645 646 mutex_enter(&zvol_state_lock); 647 if ((zv = zvol_minor_lookup(name)) == NULL) { 648 mutex_exit(&zvol_state_lock); 649 return (ENXIO); 650 } 651 rc = zvol_remove_zv(zv); 652 mutex_exit(&zvol_state_lock); 653 return (rc); 654 } 655 656 int 657 zvol_first_open(zvol_state_t *zv) 658 { 659 objset_t *os; 660 uint64_t volsize; 661 int error; 662 uint64_t readonly; 663 664 /* lie and say we're read-only */ 665 error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, B_TRUE, 666 zvol_tag, &os); 667 if (error) 668 return (error); 669 670 error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); 671 if (error) { 672 ASSERT(error == 0); 673 dmu_objset_disown(os, zvol_tag); 674 return (error); 675 } 676 zv->zv_objset = os; 677 zv->zv_volsize = volsize; 678 zv->zv_zilog = zil_open(os, zvol_get_data); 679 zvol_size_changed(zv); 680 681 VERIFY(dsl_prop_get_integer(zv->zv_name, "readonly", &readonly, 682 NULL) == 0); 683 if (readonly || dmu_objset_is_snapshot(os)) 684 zv->zv_flags |= ZVOL_RDONLY; 685 else 686 zv->zv_flags &= ~ZVOL_RDONLY; 687 return (error); 688 } 689 690 void 691 zvol_last_close(zvol_state_t *zv) 692 { 693 zil_close(zv->zv_zilog); 694 zv->zv_zilog = NULL; 695 dmu_objset_disown(zv->zv_objset, zvol_tag); 696 zv->zv_objset = NULL; 697 #ifdef __NetBSD__ 698 disk_detach(&zv->zv_dk); 699 disk_destroy(&zv->zv_dk); 700 mutex_destroy(&zv->zv_dklock); 701 #endif 702 return; 703 } 704 705 int 706 zvol_prealloc(zvol_state_t *zv) 707 { 708 objset_t *os = zv->zv_objset; 709 dmu_tx_t *tx; 710 uint64_t refd, avail, usedobjs, availobjs; 711 uint64_t resid = zv->zv_volsize; 712 uint64_t off = 0; 713 714 /* Check the space usage before attempting to allocate the space */ 715 dmu_objset_space(os, &refd, &avail, &usedobjs, &availobjs); 716 if (avail < zv->zv_volsize) 717 return (ENOSPC); 718 719 /* Free old extents if they exist */ 720 zvol_free_extents(zv); 721 722 while (resid != 0) { 723 int error; 724 uint64_t bytes = MIN(resid, SPA_MAXBLOCKSIZE); 725 726 tx = dmu_tx_create(os); 727 dmu_tx_hold_write(tx, ZVOL_OBJ, off, bytes); 728 error = dmu_tx_assign(tx, TXG_WAIT); 729 if (error) { 730 dmu_tx_abort(tx); 731 (void) dmu_free_long_range(os, ZVOL_OBJ, 0, off); 732 return (error); 733 } 734 dmu_prealloc(os, ZVOL_OBJ, off, bytes, tx); 735 dmu_tx_commit(tx); 736 off += bytes; 737 resid -= bytes; 738 } 739 txg_wait_synced(dmu_objset_pool(os), 0); 740 741 return (0); 742 } 743 744 int 745 zvol_update_volsize(objset_t *os, uint64_t volsize) 746 { 747 dmu_tx_t *tx; 748 int error; 749 750 ASSERT(MUTEX_HELD(&zvol_state_lock)); 751 752 tx = dmu_tx_create(os); 753 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 754 error = dmu_tx_assign(tx, TXG_WAIT); 755 if (error) { 756 dmu_tx_abort(tx); 757 return (error); 758 } 759 760 error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, 761 &volsize, tx); 762 dmu_tx_commit(tx); 763 764 if (error == 0) 765 error = dmu_free_long_range(os, 766 ZVOL_OBJ, volsize, DMU_OBJECT_END); 767 return (error); 768 } 769 770 void 771 zvol_remove_minors(const char *name) 772 { 773 zvol_state_t *zv; 774 char *namebuf; 775 minor_t minor; 776 777 namebuf = kmem_zalloc(strlen(name) + 2, KM_SLEEP); 778 (void) strncpy(namebuf, name, strlen(name)); 779 (void) strcat(namebuf, "/"); 780 mutex_enter(&zvol_state_lock); 781 for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) { 782 783 zv = ddi_get_soft_state(zvol_state, minor); 784 if (zv == NULL) 785 continue; 786 if (strncmp(namebuf, zv->zv_name, strlen(namebuf)) == 0) 787 (void) zvol_remove_zv(zv); 788 } 789 kmem_free(namebuf, strlen(name) + 2); 790 791 mutex_exit(&zvol_state_lock); 792 } 793 794 int 795 zvol_set_volsize(const char *name, major_t maj, uint64_t volsize) 796 { 797 zvol_state_t *zv = NULL; 798 objset_t *os; 799 int error; 800 dmu_object_info_t doi; 801 uint64_t old_volsize = 0ULL; 802 uint64_t readonly; 803 804 mutex_enter(&zvol_state_lock); 805 zv = zvol_minor_lookup(name); 806 if ((error = dmu_objset_hold(name, FTAG, &os)) != 0) { 807 mutex_exit(&zvol_state_lock); 808 return (error); 809 } 810 811 if ((error = dmu_object_info(os, ZVOL_OBJ, &doi)) != 0 || 812 (error = zvol_check_volsize(volsize, 813 doi.doi_data_block_size)) != 0) 814 goto out; 815 816 VERIFY(dsl_prop_get_integer(name, "readonly", &readonly, 817 NULL) == 0); 818 if (readonly) { 819 error = EROFS; 820 goto out; 821 } 822 823 error = zvol_update_volsize(os, volsize); 824 825 #ifndef __NetBSD__ 826 /* 827 * Reinitialize the dump area to the new size. If we 828 * failed to resize the dump area then restore it back to 829 * its original size. 830 */ 831 if (zv && error == 0) { 832 if (zv->zv_flags & ZVOL_DUMPIFIED) { 833 old_volsize = zv->zv_volsize; 834 zv->zv_volsize = volsize; 835 if ((error = zvol_dumpify(zv)) != 0 || 836 (error = dumpvp_resize()) != 0) { 837 (void) zvol_update_volsize(os, old_volsize); 838 zv->zv_volsize = old_volsize; 839 error = zvol_dumpify(zv); 840 } 841 } 842 if (error == 0) { 843 zv->zv_volsize = volsize; 844 zvol_size_changed(volsize, maj, zv->zv_minor); 845 } 846 } 847 #endif 848 849 /* 850 * Generate a LUN expansion event. 851 */ 852 if (zv && error == 0) { 853 sysevent_id_t eid; 854 nvlist_t *attr; 855 char *physpath = kmem_zalloc(MAXPATHLEN, KM_SLEEP); 856 857 (void) snprintf(physpath, MAXPATHLEN, "%s%u", ZVOL_PSEUDO_DEV, 858 zv->zv_minor); 859 860 VERIFY(nvlist_alloc(&attr, NV_UNIQUE_NAME, KM_SLEEP) == 0); 861 VERIFY(nvlist_add_string(attr, DEV_PHYS_PATH, physpath) == 0); 862 863 (void) ddi_log_sysevent(zfs_dip, SUNW_VENDOR, EC_DEV_STATUS, 864 ESC_DEV_DLE, attr, &eid, DDI_SLEEP); 865 866 nvlist_free(attr); 867 kmem_free(physpath, MAXPATHLEN); 868 } 869 870 out: 871 dmu_objset_rele(os, FTAG); 872 873 mutex_exit(&zvol_state_lock); 874 875 return (error); 876 } 877 878 /*ARGSUSED*/ 879 int 880 zvol_open(dev_t *devp, int flag, int otyp, cred_t *cr) 881 { 882 minor_t minor = getminor(*devp); 883 zvol_state_t *zv; 884 int err = 0; 885 886 if (minor == 0) /* This is the control device */ 887 return (0); 888 889 mutex_enter(&zvol_state_lock); 890 891 zv = ddi_get_soft_state(zvol_state, minor); 892 if (zv == NULL) { 893 mutex_exit(&zvol_state_lock); 894 return (ENXIO); 895 } 896 897 if (zv->zv_total_opens == 0) 898 err = zvol_first_open(zv); 899 if (err) { 900 mutex_exit(&zvol_state_lock); 901 return (err); 902 } 903 if ((flag & FWRITE) && (zv->zv_flags & ZVOL_RDONLY)) { 904 err = EROFS; 905 goto out; 906 } 907 if (zv->zv_flags & ZVOL_EXCL) { 908 err = EBUSY; 909 goto out; 910 } 911 if (flag & FEXCL) { 912 if (zv->zv_total_opens != 0) { 913 err = EBUSY; 914 goto out; 915 } 916 zv->zv_flags |= ZVOL_EXCL; 917 } 918 919 if (zv->zv_open_count[otyp] == 0 || otyp == OTYP_LYR) { 920 zv->zv_open_count[otyp]++; 921 zv->zv_total_opens++; 922 } 923 mutex_exit(&zvol_state_lock); 924 925 return (err); 926 out: 927 if (zv->zv_total_opens == 0) 928 zvol_last_close(zv); 929 mutex_exit(&zvol_state_lock); 930 return (err); 931 } 932 933 /*ARGSUSED*/ 934 int 935 zvol_close(dev_t dev, int flag, int otyp, cred_t *cr) 936 { 937 minor_t minor = getminor(dev); 938 zvol_state_t *zv; 939 int error = 0; 940 941 if (minor == 0) /* This is the control device */ 942 return (0); 943 944 mutex_enter(&zvol_state_lock); 945 946 zv = ddi_get_soft_state(zvol_state, minor); 947 if (zv == NULL) { 948 mutex_exit(&zvol_state_lock); 949 return (ENXIO); 950 } 951 952 if (zv->zv_flags & ZVOL_EXCL) { 953 ASSERT(zv->zv_total_opens == 1); 954 zv->zv_flags &= ~ZVOL_EXCL; 955 } 956 957 /* 958 * If the open count is zero, this is a spurious close. 959 * That indicates a bug in the kernel / DDI framework. 960 */ 961 ASSERT(zv->zv_open_count[otyp] != 0); 962 ASSERT(zv->zv_total_opens != 0); 963 964 /* 965 * You may get multiple opens, but only one close. 966 */ 967 zv->zv_open_count[otyp]--; 968 zv->zv_total_opens--; 969 970 if (zv->zv_total_opens == 0) 971 zvol_last_close(zv); 972 973 mutex_exit(&zvol_state_lock); 974 return (error); 975 } 976 977 static void 978 zvol_get_done(zgd_t *zgd, int error) 979 { 980 if (zgd->zgd_db) 981 dmu_buf_rele(zgd->zgd_db, zgd); 982 983 zfs_range_unlock(zgd->zgd_rl); 984 985 if (error == 0 && zgd->zgd_bp) 986 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); 987 988 kmem_free(zgd, sizeof (zgd_t)); 989 } 990 991 /* 992 * Get data to generate a TX_WRITE intent log record. 993 */ 994 static int 995 zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) 996 { 997 zvol_state_t *zv = arg; 998 objset_t *os = zv->zv_objset; 999 uint64_t object = ZVOL_OBJ; 1000 uint64_t offset = lr->lr_offset; 1001 uint64_t size = lr->lr_length; /* length of user data */ 1002 blkptr_t *bp = &lr->lr_blkptr; 1003 dmu_buf_t *db; 1004 zgd_t *zgd; 1005 int error; 1006 1007 ASSERT(zio != NULL); 1008 ASSERT(size != 0); 1009 1010 zgd = kmem_zalloc(sizeof (zgd_t), KM_SLEEP); 1011 zgd->zgd_zilog = zv->zv_zilog; 1012 zgd->zgd_rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_READER); 1013 1014 /* 1015 * Write records come in two flavors: immediate and indirect. 1016 * For small writes it's cheaper to store the data with the 1017 * log record (immediate); for large writes it's cheaper to 1018 * sync the data and get a pointer to it (indirect) so that 1019 * we don't have to write the data twice. 1020 */ 1021 if (buf != NULL) { /* immediate write */ 1022 error = dmu_read(os, object, offset, size, buf, 1023 DMU_READ_NO_PREFETCH); 1024 } else { 1025 size = zv->zv_volblocksize; 1026 offset = P2ALIGN(offset, size); 1027 error = dmu_buf_hold(os, object, offset, zgd, &db); 1028 if (error == 0) { 1029 zgd->zgd_db = db; 1030 zgd->zgd_bp = bp; 1031 1032 ASSERT(db->db_offset == offset); 1033 ASSERT(db->db_size == size); 1034 1035 error = dmu_sync(zio, lr->lr_common.lrc_txg, 1036 zvol_get_done, zgd); 1037 1038 if (error == 0) 1039 return (0); 1040 } 1041 } 1042 1043 zvol_get_done(zgd, error); 1044 1045 return (error); 1046 } 1047 1048 /* 1049 * zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions. 1050 * 1051 * We store data in the log buffers if it's small enough. 1052 * Otherwise we will later flush the data out via dmu_sync(). 1053 */ 1054 ssize_t zvol_immediate_write_sz = 32768; 1055 1056 static void 1057 zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, offset_t off, ssize_t resid, 1058 boolean_t sync) 1059 { 1060 uint32_t blocksize = zv->zv_volblocksize; 1061 zilog_t *zilog = zv->zv_zilog; 1062 boolean_t slogging; 1063 ssize_t immediate_write_sz; 1064 1065 if (zil_disable) 1066 return; 1067 1068 if (zil_replaying(zilog, tx)) 1069 return; 1070 1071 immediate_write_sz = (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT) 1072 ? 0 : zvol_immediate_write_sz; 1073 1074 slogging = spa_has_slogs(zilog->zl_spa) && 1075 (zilog->zl_logbias == ZFS_LOGBIAS_LATENCY); 1076 1077 while (resid) { 1078 itx_t *itx; 1079 lr_write_t *lr; 1080 ssize_t len; 1081 itx_wr_state_t write_state; 1082 1083 /* 1084 * Unlike zfs_log_write() we can be called with 1085 * upto DMU_MAX_ACCESS/2 (5MB) writes. 1086 */ 1087 if (blocksize > immediate_write_sz && !slogging && 1088 resid >= blocksize && off % blocksize == 0) { 1089 write_state = WR_INDIRECT; /* uses dmu_sync */ 1090 len = blocksize; 1091 } else if (sync) { 1092 write_state = WR_COPIED; 1093 len = MIN(ZIL_MAX_LOG_DATA, resid); 1094 } else { 1095 write_state = WR_NEED_COPY; 1096 len = MIN(ZIL_MAX_LOG_DATA, resid); 1097 } 1098 1099 itx = zil_itx_create(TX_WRITE, sizeof (*lr) + 1100 (write_state == WR_COPIED ? len : 0)); 1101 lr = (lr_write_t *)&itx->itx_lr; 1102 if (write_state == WR_COPIED && dmu_read(zv->zv_objset, 1103 ZVOL_OBJ, off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) { 1104 zil_itx_destroy(itx); 1105 itx = zil_itx_create(TX_WRITE, sizeof (*lr)); 1106 lr = (lr_write_t *)&itx->itx_lr; 1107 write_state = WR_NEED_COPY; 1108 } 1109 1110 itx->itx_wr_state = write_state; 1111 if (write_state == WR_NEED_COPY) 1112 itx->itx_sod += len; 1113 lr->lr_foid = ZVOL_OBJ; 1114 lr->lr_offset = off; 1115 lr->lr_length = len; 1116 lr->lr_blkoff = 0; 1117 BP_ZERO(&lr->lr_blkptr); 1118 1119 itx->itx_private = zv; 1120 itx->itx_sync = sync; 1121 1122 (void) zil_itx_assign(zilog, itx, tx); 1123 1124 off += len; 1125 resid -= len; 1126 } 1127 } 1128 1129 #ifndef __NetBSD__ 1130 static int 1131 zvol_dumpio_vdev(vdev_t *vd, void *addr, uint64_t offset, uint64_t size, 1132 boolean_t doread, boolean_t isdump) 1133 { 1134 vdev_disk_t *dvd; 1135 int c; 1136 int numerrors = 0; 1137 1138 for (c = 0; c < vd->vdev_children; c++) { 1139 ASSERT(vd->vdev_ops == &vdev_mirror_ops || 1140 vd->vdev_ops == &vdev_replacing_ops || 1141 vd->vdev_ops == &vdev_spare_ops); 1142 int err = zvol_dumpio_vdev(vd->vdev_child[c], 1143 addr, offset, size, doread, isdump); 1144 if (err != 0) { 1145 numerrors++; 1146 } else if (doread) { 1147 break; 1148 } 1149 } 1150 1151 if (!vd->vdev_ops->vdev_op_leaf) 1152 return (numerrors < vd->vdev_children ? 0 : EIO); 1153 1154 if (doread && !vdev_readable(vd)) 1155 return (EIO); 1156 else if (!doread && !vdev_writeable(vd)) 1157 return (EIO); 1158 1159 dvd = vd->vdev_tsd; 1160 ASSERT3P(dvd, !=, NULL); 1161 offset += VDEV_LABEL_START_SIZE; 1162 1163 if (ddi_in_panic() || isdump) { 1164 ASSERT(!doread); 1165 if (doread) 1166 return (EIO); 1167 return (ldi_dump(dvd->vd_lh, addr, lbtodb(offset), 1168 lbtodb(size))); 1169 } else { 1170 return (vdev_disk_physio(dvd->vd_lh, addr, size, offset, 1171 doread ? B_READ : B_WRITE)); 1172 } 1173 } 1174 1175 static int 1176 zvol_dumpio(zvol_state_t *zv, void *addr, uint64_t offset, uint64_t size, 1177 boolean_t doread, boolean_t isdump) 1178 { 1179 vdev_t *vd; 1180 int error; 1181 zvol_extent_t *ze; 1182 spa_t *spa = dmu_objset_spa(zv->zv_objset); 1183 1184 /* Must be sector aligned, and not stradle a block boundary. */ 1185 if (P2PHASE(offset, DEV_BSIZE) || P2PHASE(size, DEV_BSIZE) || 1186 P2BOUNDARY(offset, size, zv->zv_volblocksize)) { 1187 return (EINVAL); 1188 } 1189 ASSERT(size <= zv->zv_volblocksize); 1190 1191 /* Locate the extent this belongs to */ 1192 ze = list_head(&zv->zv_extents); 1193 while (offset >= ze->ze_nblks * zv->zv_volblocksize) { 1194 offset -= ze->ze_nblks * zv->zv_volblocksize; 1195 ze = list_next(&zv->zv_extents, ze); 1196 } 1197 spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); 1198 vd = vdev_lookup_top(spa, DVA_GET_VDEV(&ze->ze_dva)); 1199 offset += DVA_GET_OFFSET(&ze->ze_dva); 1200 error = zvol_dumpio_vdev(vd, addr, offset, size, doread, isdump); 1201 spa_config_exit(spa, SCL_STATE, FTAG); 1202 return (error); 1203 } 1204 #endif /* __NetBSD__ */ 1205 1206 void 1207 zvol_strategy(buf_t *bp) 1208 { 1209 zvol_state_t *zv = ddi_get_soft_state(zvol_state, getminor(bp->b_edev)); 1210 uint64_t off, volsize; 1211 size_t resid; 1212 char *addr; 1213 objset_t *os; 1214 rl_t *rl; 1215 int error = 0; 1216 boolean_t doread = bp->b_flags & B_READ; 1217 boolean_t is_dump = zv->zv_flags & ZVOL_DUMPIFIED; 1218 boolean_t sync; 1219 1220 if (zv == NULL) { 1221 bioerror(bp, ENXIO); 1222 biodone(bp); 1223 return; 1224 } 1225 1226 if (getminor(bp->b_edev) == 0) { 1227 bioerror(bp, EINVAL); 1228 biodone(bp); 1229 return; 1230 } 1231 1232 if (!(bp->b_flags & B_READ) && (zv->zv_flags & ZVOL_RDONLY)) { 1233 bioerror(bp, EROFS); 1234 biodone(bp); 1235 return; 1236 } 1237 1238 off = (uint64_t)bp->b_blkno * DEV_BSIZE; 1239 volsize = zv->zv_volsize; 1240 1241 os = zv->zv_objset; 1242 ASSERT(os != NULL); 1243 1244 addr = bp->b_data; 1245 resid = bp->b_bcount; 1246 1247 if (resid > 0 && (off < 0 || off >= volsize)) { 1248 bioerror(bp, EIO); 1249 biodone(bp); 1250 return; 1251 } 1252 1253 sync = !(bp->b_flags & B_ASYNC) && !doread && !is_dump && 1254 !(zv->zv_flags & ZVOL_WCE) && !zil_disable; 1255 1256 /* 1257 * There must be no buffer changes when doing a dmu_sync() because 1258 * we can't change the data whilst calculating the checksum. 1259 */ 1260 mutex_enter(&zv->zv_dklock); 1261 disk_busy(&zv->zv_dk); 1262 mutex_exit(&zv->zv_dklock); 1263 1264 rl = zfs_range_lock(&zv->zv_znode, off, resid, 1265 doread ? RL_READER : RL_WRITER); 1266 1267 while (resid != 0 && off < volsize) { 1268 size_t size = MIN(resid, zvol_maxphys); 1269 if (is_dump) { 1270 #ifdef __NetBSD__ 1271 printf("XXXNETBSD zvol_strategy: how?"); 1272 #else 1273 size = MIN(size, P2END(off, zv->zv_volblocksize) - off); 1274 error = zvol_dumpio(zv, addr, off, size, 1275 doread, B_FALSE); 1276 #endif 1277 } else if (doread) { 1278 error = dmu_read(os, ZVOL_OBJ, off, size, addr, 1279 DMU_READ_PREFETCH); 1280 } else { 1281 dmu_tx_t *tx = dmu_tx_create(os); 1282 dmu_tx_hold_write(tx, ZVOL_OBJ, off, size); 1283 error = dmu_tx_assign(tx, TXG_WAIT); 1284 if (error) { 1285 dmu_tx_abort(tx); 1286 } else { 1287 dmu_write(os, ZVOL_OBJ, off, size, addr, tx); 1288 zvol_log_write(zv, tx, off, size, sync); 1289 dmu_tx_commit(tx); 1290 } 1291 } 1292 if (error) { 1293 /* convert checksum errors into IO errors */ 1294 if (error == ECKSUM) 1295 error = EIO; 1296 break; 1297 } 1298 off += size; 1299 addr += size; 1300 resid -= size; 1301 } 1302 zfs_range_unlock(rl); 1303 1304 if ((bp->b_resid = resid) == bp->b_bcount) 1305 bioerror(bp, off > volsize ? EINVAL : error); 1306 1307 if (sync) 1308 zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ); 1309 mutex_enter(&zv->zv_dklock); 1310 disk_unbusy(&zv->zv_dk, bp->b_bcount - bp->b_resid, doread); 1311 mutex_exit(&zv->zv_dklock); 1312 biodone(bp); 1313 1314 return; 1315 } 1316 1317 /* 1318 * Set the buffer count to the zvol maximum transfer. 1319 * Using our own routine instead of the default minphys() 1320 * means that for larger writes we write bigger buffers on X86 1321 * (128K instead of 56K) and flush the disk write cache less often 1322 * (every zvol_maxphys - currently 1MB) instead of minphys (currently 1323 * 56K on X86 and 128K on sparc). 1324 */ 1325 void 1326 zvol_minphys(struct buf *bp) 1327 { 1328 if (bp->b_bcount > zvol_maxphys) 1329 bp->b_bcount = zvol_maxphys; 1330 } 1331 1332 #ifndef __NetBSD__ 1333 int 1334 zvol_dump(dev_t dev, caddr_t addr, daddr_t blkno, int nblocks) 1335 { 1336 minor_t minor = getminor(dev); 1337 zvol_state_t *zv; 1338 int error = 0; 1339 uint64_t size; 1340 uint64_t boff; 1341 uint64_t resid; 1342 1343 if (minor == 0) /* This is the control device */ 1344 return (ENXIO); 1345 1346 zv = ddi_get_soft_state(zvol_state, minor); 1347 if (zv == NULL) 1348 return (ENXIO); 1349 1350 boff = ldbtob(blkno); 1351 resid = ldbtob(nblocks); 1352 1353 VERIFY3U(boff + resid, <=, zv->zv_volsize); 1354 1355 while (resid) { 1356 size = MIN(resid, P2END(boff, zv->zv_volblocksize) - boff); 1357 error = zvol_dumpio(zv, addr, boff, size, B_FALSE, B_TRUE); 1358 if (error) 1359 break; 1360 boff += size; 1361 addr += size; 1362 resid -= size; 1363 } 1364 1365 return (error); 1366 } 1367 #endif /* !__NetBSD__ */ 1368 1369 /*ARGSUSED*/ 1370 int 1371 zvol_read(dev_t dev, uio_t *uio, cred_t *cr) 1372 { 1373 minor_t minor = getminor(dev); 1374 zvol_state_t *zv; 1375 uint64_t volsize; 1376 rl_t *rl; 1377 int error = 0; 1378 1379 if (minor == 0) /* This is the control device */ 1380 return (ENXIO); 1381 1382 zv = ddi_get_soft_state(zvol_state, minor); 1383 if (zv == NULL) 1384 return (ENXIO); 1385 1386 volsize = zv->zv_volsize; 1387 if (uio->uio_resid > 0 && 1388 (uio->uio_loffset < 0 || uio->uio_loffset >= volsize)) 1389 return (EIO); 1390 1391 if (zv->zv_flags & ZVOL_DUMPIFIED) { 1392 error = physio(zvol_strategy, NULL, dev, B_READ, 1393 zvol_minphys, uio); 1394 return (error); 1395 } 1396 1397 rl = zfs_range_lock(&zv->zv_znode, uio->uio_loffset, uio->uio_resid, 1398 RL_READER); 1399 while (uio->uio_resid > 0 && uio->uio_loffset < volsize) { 1400 uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1); 1401 1402 /* don't read past the end */ 1403 if (bytes > volsize - uio->uio_loffset) 1404 bytes = volsize - uio->uio_loffset; 1405 1406 error = dmu_read_uio(zv->zv_objset, ZVOL_OBJ, uio, bytes); 1407 if (error) { 1408 /* convert checksum errors into IO errors */ 1409 if (error == ECKSUM) 1410 error = EIO; 1411 break; 1412 } 1413 } 1414 zfs_range_unlock(rl); 1415 return (error); 1416 } 1417 1418 /*ARGSUSED*/ 1419 int 1420 zvol_write(dev_t dev, uio_t *uio, cred_t *cr) 1421 { 1422 minor_t minor = getminor(dev); 1423 zvol_state_t *zv; 1424 uint64_t volsize; 1425 rl_t *rl; 1426 int error = 0; 1427 boolean_t sync; 1428 1429 if (minor == 0) /* This is the control device */ 1430 return (ENXIO); 1431 1432 zv = ddi_get_soft_state(zvol_state, minor); 1433 if (zv == NULL) 1434 return (ENXIO); 1435 1436 volsize = zv->zv_volsize; 1437 if (uio->uio_resid > 0 && 1438 (uio->uio_loffset < 0 || uio->uio_loffset >= volsize)) 1439 return (EIO); 1440 1441 if (zv->zv_flags & ZVOL_DUMPIFIED) { 1442 error = physio(zvol_strategy, NULL, dev, B_WRITE, 1443 zvol_minphys, uio); 1444 return (error); 1445 } 1446 1447 sync = !(zv->zv_flags & ZVOL_WCE) && !zil_disable; 1448 1449 rl = zfs_range_lock(&zv->zv_znode, uio->uio_loffset, uio->uio_resid, 1450 RL_WRITER); 1451 while (uio->uio_resid > 0 && uio->uio_loffset < volsize) { 1452 uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1); 1453 uint64_t off = uio->uio_loffset; 1454 dmu_tx_t *tx = dmu_tx_create(zv->zv_objset); 1455 1456 if (bytes > volsize - off) /* don't write past the end */ 1457 bytes = volsize - off; 1458 1459 dmu_tx_hold_write(tx, ZVOL_OBJ, off, bytes); 1460 error = dmu_tx_assign(tx, TXG_WAIT); 1461 if (error) { 1462 dmu_tx_abort(tx); 1463 break; 1464 } 1465 error = dmu_write_uio(zv->zv_objset, ZVOL_OBJ, uio, bytes, tx); 1466 if (error == 0) 1467 zvol_log_write(zv, tx, off, bytes, sync); 1468 dmu_tx_commit(tx); 1469 1470 if (error) 1471 break; 1472 } 1473 zfs_range_unlock(rl); 1474 if (sync) 1475 zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ); 1476 return (error); 1477 } 1478 1479 #ifdef __NetBSD__ 1480 1481 /* 1482 * Dirtbag ioctls to support newfs(1) for UFS filesystems. 1483 */ 1484 /*ARGSUSED*/ 1485 int 1486 zvol_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 1487 { 1488 zvol_state_t *zv; 1489 int error = 0; 1490 1491 mutex_enter(&zvol_state_lock); 1492 1493 zv = ddi_get_soft_state(zvol_state, getminor(dev)); 1494 1495 if (zv == NULL) { 1496 mutex_exit(&zvol_state_lock); 1497 return (ENXIO); 1498 } 1499 1500 switch(cmd) { 1501 case DIOCGWEDGEINFO: 1502 { 1503 struct dkwedge_info *dkw = (void *) arg; 1504 1505 strlcpy(dkw->dkw_devname, zv->zv_name, 16); 1506 strlcpy(dkw->dkw_wname, zv->zv_name, MAXPATHLEN); 1507 strlcpy(dkw->dkw_parent, zv->zv_name, 16); 1508 1509 dkw->dkw_offset = 0; 1510 /* XXX NetBSD supports only DEV_BSIZE device block 1511 size zv_volblocksize >> DEV_BSIZE*/ 1512 dkw->dkw_size = (zv->zv_volsize / DEV_BSIZE); 1513 dprintf("dkw %"PRIu64" volsize %"PRIu64" volblock %"PRIu64" \n", 1514 dkw->dkw_size, zv->zv_volsize, zv->zv_volblocksize); 1515 strcpy(dkw->dkw_ptype, DKW_PTYPE_FFS); 1516 1517 break; 1518 } 1519 1520 case DIOCGDISKINFO: 1521 { 1522 struct plistref *pref = (struct plistref *) arg; 1523 1524 if (zv->zv_dk.dk_info == NULL) { 1525 mutex_exit(&zvol_state_lock); 1526 return ENOTSUP; 1527 } else 1528 prop_dictionary_copyout_ioctl(pref, cmd, 1529 zv->zv_dk.dk_info); 1530 1531 break; 1532 } 1533 1534 default: 1535 aprint_debug("unknown disk_ioctl called\n"); 1536 error = ENOTTY; 1537 break; 1538 } 1539 1540 mutex_exit(&zvol_state_lock); 1541 return (error); 1542 } 1543 1544 #else /* __NetBSD__ */ 1545 1546 int 1547 zvol_getefi(void *arg, int flag, uint64_t vs, uint8_t bs) 1548 { 1549 struct uuid uuid = EFI_RESERVED; 1550 efi_gpe_t gpe = { 0 }; 1551 uint32_t crc; 1552 dk_efi_t efi; 1553 int length; 1554 char *ptr; 1555 1556 if (ddi_copyin(arg, &efi, sizeof (dk_efi_t), flag)) 1557 return (EFAULT); 1558 ptr = (char *)(uintptr_t)efi.dki_data_64; 1559 length = efi.dki_length; 1560 /* 1561 * Some clients may attempt to request a PMBR for the 1562 * zvol. Currently this interface will return EINVAL to 1563 * such requests. These requests could be supported by 1564 * adding a check for lba == 0 and consing up an appropriate 1565 * PMBR. 1566 */ 1567 if (efi.dki_lba < 1 || efi.dki_lba > 2 || length <= 0) 1568 return (EINVAL); 1569 1570 gpe.efi_gpe_StartingLBA = LE_64(34ULL); 1571 gpe.efi_gpe_EndingLBA = LE_64((vs >> bs) - 1); 1572 UUID_LE_CONVERT(gpe.efi_gpe_PartitionTypeGUID, uuid); 1573 1574 if (efi.dki_lba == 1) { 1575 efi_gpt_t gpt = { 0 }; 1576 1577 gpt.efi_gpt_Signature = LE_64(EFI_SIGNATURE); 1578 gpt.efi_gpt_Revision = LE_32(EFI_VERSION_CURRENT); 1579 gpt.efi_gpt_HeaderSize = LE_32(sizeof (gpt)); 1580 gpt.efi_gpt_MyLBA = LE_64(1ULL); 1581 gpt.efi_gpt_FirstUsableLBA = LE_64(34ULL); 1582 gpt.efi_gpt_LastUsableLBA = LE_64((vs >> bs) - 1); 1583 gpt.efi_gpt_PartitionEntryLBA = LE_64(2ULL); 1584 gpt.efi_gpt_NumberOfPartitionEntries = LE_32(1); 1585 gpt.efi_gpt_SizeOfPartitionEntry = 1586 LE_32(sizeof (efi_gpe_t)); 1587 CRC32(crc, &gpe, sizeof (gpe), -1U, crc32_table); 1588 gpt.efi_gpt_PartitionEntryArrayCRC32 = LE_32(~crc); 1589 CRC32(crc, &gpt, sizeof (gpt), -1U, crc32_table); 1590 gpt.efi_gpt_HeaderCRC32 = LE_32(~crc); 1591 if (ddi_copyout(&gpt, ptr, MIN(sizeof (gpt), length), 1592 flag)) 1593 return (EFAULT); 1594 ptr += sizeof (gpt); 1595 length -= sizeof (gpt); 1596 } 1597 if (length > 0 && ddi_copyout(&gpe, ptr, MIN(sizeof (gpe), 1598 length), flag)) 1599 return (EFAULT); 1600 return (0); 1601 } 1602 1603 /* 1604 * Dirtbag ioctls to support mkfs(1M) for UFS filesystems. See dkio(7I). 1605 */ 1606 /*ARGSUSED*/ 1607 int 1608 zvol_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 1609 { 1610 zvol_state_t *zv; 1611 struct dk_cinfo dki; 1612 struct dk_minfo dkm; 1613 struct dk_callback *dkc; 1614 int error = 0; 1615 rl_t *rl; 1616 1617 mutex_enter(&zvol_state_lock); 1618 1619 zv = ddi_get_soft_state(zvol_state, getminor(dev)); 1620 1621 if (zv == NULL) { 1622 mutex_exit(&zvol_state_lock); 1623 return (ENXIO); 1624 } 1625 ASSERT(zv->zv_total_opens > 0); 1626 1627 switch (cmd) { 1628 1629 case DKIOCINFO: 1630 bzero(&dki, sizeof (dki)); 1631 (void) strcpy(dki.dki_cname, "zvol"); 1632 (void) strcpy(dki.dki_dname, "zvol"); 1633 dki.dki_ctype = DKC_UNKNOWN; 1634 dki.dki_unit = getminor(dev); 1635 dki.dki_maxtransfer = 1 << (SPA_MAXBLOCKSHIFT - zv->zv_min_bs); 1636 mutex_exit(&zvol_state_lock); 1637 if (ddi_copyout(&dki, (void *)arg, sizeof (dki), flag)) 1638 error = EFAULT; 1639 return (error); 1640 1641 case DKIOCGMEDIAINFO: 1642 bzero(&dkm, sizeof (dkm)); 1643 dkm.dki_lbsize = 1U << zv->zv_min_bs; 1644 dkm.dki_capacity = zv->zv_volsize >> zv->zv_min_bs; 1645 dkm.dki_media_type = DK_UNKNOWN; 1646 mutex_exit(&zvol_state_lock); 1647 if (ddi_copyout(&dkm, (void *)arg, sizeof (dkm), flag)) 1648 error = EFAULT; 1649 return (error); 1650 1651 case DKIOCGETEFI: 1652 { 1653 uint64_t vs = zv->zv_volsize; 1654 uint8_t bs = zv->zv_min_bs; 1655 1656 mutex_exit(&zvol_state_lock); 1657 error = zvol_getefi((void *)arg, flag, vs, bs); 1658 return (error); 1659 } 1660 1661 case DKIOCFLUSHWRITECACHE: 1662 dkc = (struct dk_callback *)arg; 1663 mutex_exit(&zvol_state_lock); 1664 zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ); 1665 if ((flag & FKIOCTL) && dkc != NULL && dkc->dkc_callback) { 1666 (*dkc->dkc_callback)(dkc->dkc_cookie, error); 1667 error = 0; 1668 } 1669 return (error); 1670 1671 case DKIOCGETWCE: 1672 { 1673 int wce = (zv->zv_flags & ZVOL_WCE) ? 1 : 0; 1674 if (ddi_copyout(&wce, (void *)arg, sizeof (int), 1675 flag)) 1676 error = EFAULT; 1677 break; 1678 } 1679 case DKIOCSETWCE: 1680 { 1681 int wce; 1682 if (ddi_copyin((void *)arg, &wce, sizeof (int), 1683 flag)) { 1684 error = EFAULT; 1685 break; 1686 } 1687 if (wce) { 1688 zv->zv_flags |= ZVOL_WCE; 1689 mutex_exit(&zvol_state_lock); 1690 } else { 1691 zv->zv_flags &= ~ZVOL_WCE; 1692 mutex_exit(&zvol_state_lock); 1693 zil_commit(zv->zv_zilog, UINT64_MAX, ZVOL_OBJ); 1694 } 1695 return (0); 1696 } 1697 1698 case DKIOCGGEOM: 1699 case DKIOCGVTOC: 1700 /* 1701 * commands using these (like prtvtoc) expect ENOTSUP 1702 * since we're emulating an EFI label 1703 */ 1704 error = ENOTSUP; 1705 break; 1706 1707 case DKIOCDUMPINIT: 1708 rl = zfs_range_lock(&zv->zv_znode, 0, zv->zv_volsize, 1709 RL_WRITER); 1710 error = zvol_dumpify(zv); 1711 zfs_range_unlock(rl); 1712 break; 1713 1714 case DKIOCDUMPFINI: 1715 if (!(zv->zv_flags & ZVOL_DUMPIFIED)) 1716 break; 1717 rl = zfs_range_lock(&zv->zv_znode, 0, zv->zv_volsize, 1718 RL_WRITER); 1719 error = zvol_dump_fini(zv); 1720 zfs_range_unlock(rl); 1721 break; 1722 1723 default: 1724 error = ENOTTY; 1725 break; 1726 1727 } 1728 mutex_exit(&zvol_state_lock); 1729 return (error); 1730 } 1731 1732 #endif /* __NetBSD__ */ 1733 1734 int 1735 zvol_busy(void) 1736 { 1737 return (zvol_minors != 0); 1738 } 1739 1740 void 1741 zvol_init(void) 1742 { 1743 VERIFY(ddi_soft_state_init(&zvol_state, sizeof (zvol_state_t), 1) == 0); 1744 mutex_init(&zvol_state_lock, NULL, MUTEX_DEFAULT, NULL); 1745 } 1746 1747 void 1748 zvol_fini(void) 1749 { 1750 mutex_destroy(&zvol_state_lock); 1751 ddi_soft_state_fini(&zvol_state); 1752 } 1753 1754 #ifndef __NetBSD__ 1755 static int 1756 zvol_dump_init(zvol_state_t *zv, boolean_t resize) 1757 { 1758 dmu_tx_t *tx; 1759 int error = 0; 1760 objset_t *os = zv->zv_objset; 1761 nvlist_t *nv = NULL; 1762 uint64_t version = spa_version(dmu_objset_spa(zv->zv_objset)); 1763 1764 ASSERT(MUTEX_HELD(&zvol_state_lock)); 1765 error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, 0, 1766 DMU_OBJECT_END); 1767 /* wait for dmu_free_long_range to actually free the blocks */ 1768 txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0); 1769 1770 tx = dmu_tx_create(os); 1771 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 1772 dmu_tx_hold_bonus(tx, ZVOL_OBJ); 1773 error = dmu_tx_assign(tx, TXG_WAIT); 1774 if (error) { 1775 dmu_tx_abort(tx); 1776 return (error); 1777 } 1778 1779 /* 1780 * If we are resizing the dump device then we only need to 1781 * update the refreservation to match the newly updated 1782 * zvolsize. Otherwise, we save off the original state of the 1783 * zvol so that we can restore them if the zvol is ever undumpified. 1784 */ 1785 if (resize) { 1786 error = zap_update(os, ZVOL_ZAP_OBJ, 1787 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, 1788 &zv->zv_volsize, tx); 1789 } else { 1790 uint64_t checksum, compress, refresrv, vbs, dedup; 1791 1792 error = dsl_prop_get_integer(zv->zv_name, 1793 zfs_prop_to_name(ZFS_PROP_COMPRESSION), &compress, NULL); 1794 error = error ? error : dsl_prop_get_integer(zv->zv_name, 1795 zfs_prop_to_name(ZFS_PROP_CHECKSUM), &checksum, NULL); 1796 error = error ? error : dsl_prop_get_integer(zv->zv_name, 1797 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &refresrv, NULL); 1798 error = error ? error : dsl_prop_get_integer(zv->zv_name, 1799 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &vbs, NULL); 1800 if (version >= SPA_VERSION_DEDUP) { 1801 error = error ? error : 1802 dsl_prop_get_integer(zv->zv_name, 1803 zfs_prop_to_name(ZFS_PROP_DEDUP), &dedup, NULL); 1804 } 1805 1806 error = error ? error : zap_update(os, ZVOL_ZAP_OBJ, 1807 zfs_prop_to_name(ZFS_PROP_COMPRESSION), 8, 1, 1808 &compress, tx); 1809 error = error ? error : zap_update(os, ZVOL_ZAP_OBJ, 1810 zfs_prop_to_name(ZFS_PROP_CHECKSUM), 8, 1, &checksum, tx); 1811 error = error ? error : zap_update(os, ZVOL_ZAP_OBJ, 1812 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, 1813 &refresrv, tx); 1814 error = error ? error : zap_update(os, ZVOL_ZAP_OBJ, 1815 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 8, 1, 1816 &vbs, tx); 1817 error = error ? error : dmu_object_set_blocksize( 1818 os, ZVOL_OBJ, SPA_MAXBLOCKSIZE, 0, tx); 1819 if (version >= SPA_VERSION_DEDUP) { 1820 error = error ? error : zap_update(os, ZVOL_ZAP_OBJ, 1821 zfs_prop_to_name(ZFS_PROP_DEDUP), 8, 1, 1822 &dedup, tx); 1823 } 1824 if (error == 0) 1825 zv->zv_volblocksize = SPA_MAXBLOCKSIZE; 1826 } 1827 dmu_tx_commit(tx); 1828 1829 /* 1830 * We only need update the zvol's property if we are initializing 1831 * the dump area for the first time. 1832 */ 1833 if (!resize) { 1834 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1835 VERIFY(nvlist_add_uint64(nv, 1836 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 0) == 0); 1837 VERIFY(nvlist_add_uint64(nv, 1838 zfs_prop_to_name(ZFS_PROP_COMPRESSION), 1839 ZIO_COMPRESS_OFF) == 0); 1840 VERIFY(nvlist_add_uint64(nv, 1841 zfs_prop_to_name(ZFS_PROP_CHECKSUM), 1842 ZIO_CHECKSUM_OFF) == 0); 1843 if (version >= SPA_VERSION_DEDUP) { 1844 VERIFY(nvlist_add_uint64(nv, 1845 zfs_prop_to_name(ZFS_PROP_DEDUP), 1846 ZIO_CHECKSUM_OFF) == 0); 1847 } 1848 1849 error = zfs_set_prop_nvlist(zv->zv_name, ZPROP_SRC_LOCAL, 1850 nv, NULL); 1851 nvlist_free(nv); 1852 1853 if (error) 1854 return (error); 1855 } 1856 1857 /* Allocate the space for the dump */ 1858 error = zvol_prealloc(zv); 1859 return (error); 1860 } 1861 1862 static int 1863 zvol_dumpify(zvol_state_t *zv) 1864 { 1865 int error = 0; 1866 uint64_t dumpsize = 0; 1867 dmu_tx_t *tx; 1868 objset_t *os = zv->zv_objset; 1869 1870 if (zv->zv_flags & ZVOL_RDONLY) 1871 return (EROFS); 1872 1873 if (zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, 1874 8, 1, &dumpsize) != 0 || dumpsize != zv->zv_volsize) { 1875 boolean_t resize = (dumpsize > 0) ? B_TRUE : B_FALSE; 1876 1877 if ((error = zvol_dump_init(zv, resize)) != 0) { 1878 (void) zvol_dump_fini(zv); 1879 return (error); 1880 } 1881 } 1882 1883 /* 1884 * Build up our lba mapping. 1885 */ 1886 error = zvol_get_lbas(zv); 1887 if (error) { 1888 (void) zvol_dump_fini(zv); 1889 return (error); 1890 } 1891 1892 tx = dmu_tx_create(os); 1893 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 1894 error = dmu_tx_assign(tx, TXG_WAIT); 1895 if (error) { 1896 dmu_tx_abort(tx); 1897 (void) zvol_dump_fini(zv); 1898 return (error); 1899 } 1900 1901 zv->zv_flags |= ZVOL_DUMPIFIED; 1902 error = zap_update(os, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, 8, 1, 1903 &zv->zv_volsize, tx); 1904 dmu_tx_commit(tx); 1905 1906 if (error) { 1907 (void) zvol_dump_fini(zv); 1908 return (error); 1909 } 1910 1911 txg_wait_synced(dmu_objset_pool(os), 0); 1912 return (0); 1913 } 1914 1915 static int 1916 zvol_dump_fini(zvol_state_t *zv) 1917 { 1918 dmu_tx_t *tx; 1919 objset_t *os = zv->zv_objset; 1920 nvlist_t *nv; 1921 int error = 0; 1922 uint64_t checksum, compress, refresrv, vbs, dedup; 1923 uint64_t version = spa_version(dmu_objset_spa(zv->zv_objset)); 1924 1925 /* 1926 * Attempt to restore the zvol back to its pre-dumpified state. 1927 * This is a best-effort attempt as it's possible that not all 1928 * of these properties were initialized during the dumpify process 1929 * (i.e. error during zvol_dump_init). 1930 */ 1931 1932 tx = dmu_tx_create(os); 1933 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 1934 error = dmu_tx_assign(tx, TXG_WAIT); 1935 if (error) { 1936 dmu_tx_abort(tx); 1937 return (error); 1938 } 1939 (void) zap_remove(os, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, tx); 1940 dmu_tx_commit(tx); 1941 1942 (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, 1943 zfs_prop_to_name(ZFS_PROP_CHECKSUM), 8, 1, &checksum); 1944 (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, 1945 zfs_prop_to_name(ZFS_PROP_COMPRESSION), 8, 1, &compress); 1946 (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, 1947 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, &refresrv); 1948 (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, 1949 zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 8, 1, &vbs); 1950 1951 VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); 1952 (void) nvlist_add_uint64(nv, 1953 zfs_prop_to_name(ZFS_PROP_CHECKSUM), checksum); 1954 (void) nvlist_add_uint64(nv, 1955 zfs_prop_to_name(ZFS_PROP_COMPRESSION), compress); 1956 (void) nvlist_add_uint64(nv, 1957 zfs_prop_to_name(ZFS_PROP_REFRESERVATION), refresrv); 1958 if (version >= SPA_VERSION_DEDUP && 1959 zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, 1960 zfs_prop_to_name(ZFS_PROP_DEDUP), 8, 1, &dedup) == 0) { 1961 (void) nvlist_add_uint64(nv, 1962 zfs_prop_to_name(ZFS_PROP_DEDUP), dedup); 1963 } 1964 (void) zfs_set_prop_nvlist(zv->zv_name, ZPROP_SRC_LOCAL, 1965 nv, NULL); 1966 nvlist_free(nv); 1967 1968 zvol_free_extents(zv); 1969 zv->zv_flags &= ~ZVOL_DUMPIFIED; 1970 (void) dmu_free_long_range(os, ZVOL_OBJ, 0, DMU_OBJECT_END); 1971 /* wait for dmu_free_long_range to actually free the blocks */ 1972 txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0); 1973 tx = dmu_tx_create(os); 1974 dmu_tx_hold_bonus(tx, ZVOL_OBJ); 1975 error = dmu_tx_assign(tx, TXG_WAIT); 1976 if (error) { 1977 dmu_tx_abort(tx); 1978 return (error); 1979 } 1980 if (dmu_object_set_blocksize(os, ZVOL_OBJ, vbs, 0, tx) == 0) 1981 zv->zv_volblocksize = vbs; 1982 dmu_tx_commit(tx); 1983 1984 return (0); 1985 } 1986 #endif /* !__NetBSD__ */ 1987