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 2006 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #pragma ident "%Z%%M% %I% %E% SMI" 27 28 /* 29 * ZFS volume emulation driver. 30 * 31 * Makes a DMU object look like a volume of arbitrary size, up to 2^64 bytes. 32 * Volumes are accessed through the symbolic links named: 33 * 34 * /dev/zvol/dsk/<pool_name>/<dataset_name> 35 * /dev/zvol/rdsk/<pool_name>/<dataset_name> 36 * 37 * These links are created by the ZFS-specific devfsadm link generator. 38 * Volumes are persistent through reboot. No user command needs to be 39 * run before opening and using a device. 40 */ 41 42 #include <sys/types.h> 43 #include <sys/param.h> 44 #include <sys/errno.h> 45 #include <sys/aio_req.h> 46 #include <sys/uio.h> 47 #include <sys/buf.h> 48 #include <sys/modctl.h> 49 #include <sys/open.h> 50 #include <sys/kmem.h> 51 #include <sys/conf.h> 52 #include <sys/cmn_err.h> 53 #include <sys/stat.h> 54 #include <sys/zap.h> 55 #include <sys/spa.h> 56 #include <sys/zio.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 72 #include "zfs_namecheck.h" 73 74 #define ZVOL_OBJ 1ULL 75 #define ZVOL_ZAP_OBJ 2ULL 76 77 static void *zvol_state; 78 79 /* 80 * This lock protects the zvol_state structure from being modified 81 * while it's being used, e.g. an open that comes in before a create 82 * finishes. It also protects temporary opens of the dataset so that, 83 * e.g., an open doesn't get a spurious EBUSY. 84 */ 85 static kmutex_t zvol_state_lock; 86 static uint32_t zvol_minors; 87 88 /* 89 * The in-core state of each volume. 90 */ 91 typedef struct zvol_state { 92 char zv_name[MAXPATHLEN]; /* pool/dd name */ 93 uint64_t zv_volsize; /* amount of space we advertise */ 94 minor_t zv_minor; /* minor number */ 95 uint8_t zv_min_bs; /* minimum addressable block shift */ 96 uint8_t zv_readonly; /* hard readonly; like write-protect */ 97 objset_t *zv_objset; /* objset handle */ 98 uint32_t zv_mode; /* DS_MODE_* flags at open time */ 99 uint32_t zv_open_count[OTYPCNT]; /* open counts */ 100 uint32_t zv_total_opens; /* total open count */ 101 zilog_t *zv_zilog; /* ZIL handle */ 102 uint64_t zv_txg_assign; /* txg to assign during ZIL replay */ 103 krwlock_t zv_dslock; /* dmu_sync() rwlock */ 104 } zvol_state_t; 105 106 static void 107 zvol_size_changed(zvol_state_t *zv, dev_t dev) 108 { 109 dev = makedevice(getmajor(dev), zv->zv_minor); 110 111 VERIFY(ddi_prop_update_int64(dev, zfs_dip, 112 "Size", zv->zv_volsize) == DDI_SUCCESS); 113 VERIFY(ddi_prop_update_int64(dev, zfs_dip, 114 "Nblocks", lbtodb(zv->zv_volsize)) == DDI_SUCCESS); 115 } 116 117 int 118 zvol_check_volsize(zfs_cmd_t *zc, uint64_t blocksize) 119 { 120 if (zc->zc_volsize == 0) 121 return (EINVAL); 122 123 if (zc->zc_volsize % blocksize != 0) 124 return (EINVAL); 125 126 #ifdef _ILP32 127 if (zc->zc_volsize - 1 > SPEC_MAXOFFSET_T) 128 return (EOVERFLOW); 129 #endif 130 return (0); 131 } 132 133 int 134 zvol_check_volblocksize(zfs_cmd_t *zc) 135 { 136 if (zc->zc_volblocksize < SPA_MINBLOCKSIZE || 137 zc->zc_volblocksize > SPA_MAXBLOCKSIZE || 138 !ISP2(zc->zc_volblocksize)) 139 return (EDOM); 140 141 return (0); 142 } 143 144 static void 145 zvol_readonly_changed_cb(void *arg, uint64_t newval) 146 { 147 zvol_state_t *zv = arg; 148 149 zv->zv_readonly = (uint8_t)newval; 150 } 151 152 int 153 zvol_get_stats(zfs_cmd_t *zc, objset_t *os) 154 { 155 int error; 156 dmu_object_info_t doi; 157 158 error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &zc->zc_volsize); 159 160 if (error) 161 return (error); 162 163 error = dmu_object_info(os, ZVOL_OBJ, &doi); 164 165 if (error == 0) 166 zc->zc_volblocksize = doi.doi_data_block_size; 167 168 return (error); 169 } 170 171 /* 172 * Find a free minor number. 173 */ 174 static minor_t 175 zvol_minor_alloc(void) 176 { 177 minor_t minor; 178 179 ASSERT(MUTEX_HELD(&zvol_state_lock)); 180 181 for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) 182 if (ddi_get_soft_state(zvol_state, minor) == NULL) 183 return (minor); 184 185 return (0); 186 } 187 188 static zvol_state_t * 189 zvol_minor_lookup(char *name) 190 { 191 minor_t minor; 192 zvol_state_t *zv; 193 194 ASSERT(MUTEX_HELD(&zvol_state_lock)); 195 196 for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) { 197 zv = ddi_get_soft_state(zvol_state, minor); 198 if (zv == NULL) 199 continue; 200 if (strcmp(zv->zv_name, name) == 0) 201 break; 202 } 203 204 return (zv); 205 } 206 207 void 208 zvol_create_cb(objset_t *os, void *arg, dmu_tx_t *tx) 209 { 210 zfs_cmd_t *zc = arg; 211 int error; 212 213 error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, zc->zc_volblocksize, 214 DMU_OT_NONE, 0, tx); 215 ASSERT(error == 0); 216 217 error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP, 218 DMU_OT_NONE, 0, tx); 219 ASSERT(error == 0); 220 221 error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &zc->zc_volsize, tx); 222 ASSERT(error == 0); 223 } 224 225 /* 226 * Replay a TX_WRITE ZIL transaction that didn't get committed 227 * after a system failure 228 */ 229 static int 230 zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap) 231 { 232 objset_t *os = zv->zv_objset; 233 char *data = (char *)(lr + 1); /* data follows lr_write_t */ 234 uint64_t off = lr->lr_offset; 235 uint64_t len = lr->lr_length; 236 dmu_tx_t *tx; 237 int error; 238 239 if (byteswap) 240 byteswap_uint64_array(lr, sizeof (*lr)); 241 242 restart: 243 tx = dmu_tx_create(os); 244 dmu_tx_hold_write(tx, ZVOL_OBJ, off, len); 245 error = dmu_tx_assign(tx, zv->zv_txg_assign); 246 if (error) { 247 dmu_tx_abort(tx); 248 if (error == ERESTART && zv->zv_txg_assign == TXG_NOWAIT) { 249 txg_wait_open(dmu_objset_pool(os), 0); 250 goto restart; 251 } 252 } else { 253 dmu_write(os, ZVOL_OBJ, off, len, data, tx); 254 dmu_tx_commit(tx); 255 } 256 257 return (error); 258 } 259 260 /* ARGSUSED */ 261 static int 262 zvol_replay_err(zvol_state_t *zv, lr_t *lr, boolean_t byteswap) 263 { 264 return (ENOTSUP); 265 } 266 267 /* 268 * Callback vectors for replaying records. 269 * Only TX_WRITE is needed for zvol. 270 */ 271 zil_replay_func_t *zvol_replay_vector[TX_MAX_TYPE] = { 272 zvol_replay_err, /* 0 no such transaction type */ 273 zvol_replay_err, /* TX_CREATE */ 274 zvol_replay_err, /* TX_MKDIR */ 275 zvol_replay_err, /* TX_MKXATTR */ 276 zvol_replay_err, /* TX_SYMLINK */ 277 zvol_replay_err, /* TX_REMOVE */ 278 zvol_replay_err, /* TX_RMDIR */ 279 zvol_replay_err, /* TX_LINK */ 280 zvol_replay_err, /* TX_RENAME */ 281 zvol_replay_write, /* TX_WRITE */ 282 zvol_replay_err, /* TX_TRUNCATE */ 283 zvol_replay_err, /* TX_SETATTR */ 284 zvol_replay_err, /* TX_ACL */ 285 }; 286 287 /* 288 * Create a minor node for the specified volume. 289 */ 290 int 291 zvol_create_minor(zfs_cmd_t *zc) 292 { 293 char *name = zc->zc_name; 294 dev_t dev = zc->zc_dev; 295 zvol_state_t *zv; 296 objset_t *os; 297 uint64_t volsize; 298 minor_t minor = 0; 299 struct pathname linkpath; 300 int ds_mode = DS_MODE_PRIMARY; 301 vnode_t *vp = NULL; 302 char *devpath; 303 size_t devpathlen = strlen(ZVOL_FULL_DEV_DIR) + 1 + strlen(name) + 1; 304 char chrbuf[30], blkbuf[30]; 305 int error; 306 307 mutex_enter(&zvol_state_lock); 308 309 if ((zv = zvol_minor_lookup(name)) != NULL) { 310 mutex_exit(&zvol_state_lock); 311 return (EEXIST); 312 } 313 314 if (strchr(name, '@') != 0) 315 ds_mode |= DS_MODE_READONLY; 316 317 error = dmu_objset_open(name, DMU_OST_ZVOL, ds_mode, &os); 318 319 if (error) { 320 mutex_exit(&zvol_state_lock); 321 return (error); 322 } 323 324 error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); 325 326 if (error) { 327 dmu_objset_close(os); 328 mutex_exit(&zvol_state_lock); 329 return (error); 330 } 331 332 /* 333 * If there's an existing /dev/zvol symlink, try to use the 334 * same minor number we used last time. 335 */ 336 devpath = kmem_alloc(devpathlen, KM_SLEEP); 337 338 (void) sprintf(devpath, "%s/%s", ZVOL_FULL_DEV_DIR, name); 339 340 error = lookupname(devpath, UIO_SYSSPACE, NO_FOLLOW, NULL, &vp); 341 342 kmem_free(devpath, devpathlen); 343 344 if (error == 0 && vp->v_type != VLNK) 345 error = EINVAL; 346 347 if (error == 0) { 348 pn_alloc(&linkpath); 349 error = pn_getsymlink(vp, &linkpath, kcred); 350 if (error == 0) { 351 char *ms = strstr(linkpath.pn_path, ZVOL_PSEUDO_DEV); 352 if (ms != NULL) { 353 ms += strlen(ZVOL_PSEUDO_DEV); 354 minor = stoi(&ms); 355 } 356 } 357 pn_free(&linkpath); 358 } 359 360 if (vp != NULL) 361 VN_RELE(vp); 362 363 /* 364 * If we found a minor but it's already in use, we must pick a new one. 365 */ 366 if (minor != 0 && ddi_get_soft_state(zvol_state, minor) != NULL) 367 minor = 0; 368 369 if (minor == 0) 370 minor = zvol_minor_alloc(); 371 372 if (minor == 0) { 373 dmu_objset_close(os); 374 mutex_exit(&zvol_state_lock); 375 return (ENXIO); 376 } 377 378 if (ddi_soft_state_zalloc(zvol_state, minor) != DDI_SUCCESS) { 379 dmu_objset_close(os); 380 mutex_exit(&zvol_state_lock); 381 return (EAGAIN); 382 } 383 384 (void) ddi_prop_update_string(minor, zfs_dip, ZVOL_PROP_NAME, name); 385 386 (void) sprintf(chrbuf, "%uc,raw", minor); 387 388 if (ddi_create_minor_node(zfs_dip, chrbuf, S_IFCHR, 389 minor, DDI_PSEUDO, 0) == DDI_FAILURE) { 390 ddi_soft_state_free(zvol_state, minor); 391 dmu_objset_close(os); 392 mutex_exit(&zvol_state_lock); 393 return (EAGAIN); 394 } 395 396 (void) sprintf(blkbuf, "%uc", minor); 397 398 if (ddi_create_minor_node(zfs_dip, blkbuf, S_IFBLK, 399 minor, DDI_PSEUDO, 0) == DDI_FAILURE) { 400 ddi_remove_minor_node(zfs_dip, chrbuf); 401 ddi_soft_state_free(zvol_state, minor); 402 dmu_objset_close(os); 403 mutex_exit(&zvol_state_lock); 404 return (EAGAIN); 405 } 406 407 zv = ddi_get_soft_state(zvol_state, minor); 408 409 (void) strcpy(zv->zv_name, name); 410 zv->zv_min_bs = DEV_BSHIFT; 411 zv->zv_minor = minor; 412 zv->zv_volsize = volsize; 413 zv->zv_objset = os; 414 zv->zv_mode = ds_mode; 415 zv->zv_zilog = zil_open(os, NULL); 416 417 rw_init(&zv->zv_dslock, NULL, RW_DEFAULT, NULL); 418 419 zil_replay(os, zv, &zv->zv_txg_assign, zvol_replay_vector, NULL); 420 421 zvol_size_changed(zv, dev); 422 423 /* XXX this should handle the possible i/o error */ 424 VERIFY(dsl_prop_register(dmu_objset_ds(zv->zv_objset), 425 "readonly", zvol_readonly_changed_cb, zv) == 0); 426 427 zvol_minors++; 428 429 mutex_exit(&zvol_state_lock); 430 431 return (0); 432 } 433 434 /* 435 * Remove minor node for the specified volume. 436 */ 437 int 438 zvol_remove_minor(zfs_cmd_t *zc) 439 { 440 zvol_state_t *zv; 441 char namebuf[30]; 442 443 mutex_enter(&zvol_state_lock); 444 445 if ((zv = zvol_minor_lookup(zc->zc_name)) == NULL) { 446 mutex_exit(&zvol_state_lock); 447 return (ENXIO); 448 } 449 450 if (zv->zv_total_opens != 0) { 451 mutex_exit(&zvol_state_lock); 452 return (EBUSY); 453 } 454 455 (void) sprintf(namebuf, "%uc,raw", zv->zv_minor); 456 ddi_remove_minor_node(zfs_dip, namebuf); 457 458 (void) sprintf(namebuf, "%uc", zv->zv_minor); 459 ddi_remove_minor_node(zfs_dip, namebuf); 460 461 VERIFY(dsl_prop_unregister(dmu_objset_ds(zv->zv_objset), 462 "readonly", zvol_readonly_changed_cb, zv) == 0); 463 464 zil_close(zv->zv_zilog); 465 zv->zv_zilog = NULL; 466 dmu_objset_close(zv->zv_objset); 467 zv->zv_objset = NULL; 468 469 ddi_soft_state_free(zvol_state, zv->zv_minor); 470 471 zvol_minors--; 472 473 mutex_exit(&zvol_state_lock); 474 475 return (0); 476 } 477 478 int 479 zvol_set_volsize(zfs_cmd_t *zc) 480 { 481 zvol_state_t *zv; 482 dev_t dev = zc->zc_dev; 483 dmu_tx_t *tx; 484 int error; 485 dmu_object_info_t doi; 486 487 mutex_enter(&zvol_state_lock); 488 489 if ((zv = zvol_minor_lookup(zc->zc_name)) == NULL) { 490 mutex_exit(&zvol_state_lock); 491 return (ENXIO); 492 } 493 494 if ((error = dmu_object_info(zv->zv_objset, ZVOL_OBJ, &doi)) != 0 || 495 (error = zvol_check_volsize(zc, doi.doi_data_block_size)) != 0) { 496 mutex_exit(&zvol_state_lock); 497 return (error); 498 } 499 500 if (zv->zv_readonly || (zv->zv_mode & DS_MODE_READONLY)) { 501 mutex_exit(&zvol_state_lock); 502 return (EROFS); 503 } 504 505 tx = dmu_tx_create(zv->zv_objset); 506 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 507 dmu_tx_hold_free(tx, ZVOL_OBJ, zc->zc_volsize, DMU_OBJECT_END); 508 error = dmu_tx_assign(tx, TXG_WAIT); 509 if (error) { 510 dmu_tx_abort(tx); 511 mutex_exit(&zvol_state_lock); 512 return (error); 513 } 514 515 error = zap_update(zv->zv_objset, ZVOL_ZAP_OBJ, "size", 8, 1, 516 &zc->zc_volsize, tx); 517 if (error == 0) { 518 error = dmu_free_range(zv->zv_objset, ZVOL_OBJ, zc->zc_volsize, 519 DMU_OBJECT_END, tx); 520 } 521 522 dmu_tx_commit(tx); 523 524 if (error == 0) { 525 zv->zv_volsize = zc->zc_volsize; 526 zvol_size_changed(zv, dev); 527 } 528 529 mutex_exit(&zvol_state_lock); 530 531 return (error); 532 } 533 534 int 535 zvol_set_volblocksize(zfs_cmd_t *zc) 536 { 537 zvol_state_t *zv; 538 dmu_tx_t *tx; 539 int error; 540 541 mutex_enter(&zvol_state_lock); 542 543 if ((zv = zvol_minor_lookup(zc->zc_name)) == NULL) { 544 mutex_exit(&zvol_state_lock); 545 return (ENXIO); 546 } 547 548 if (zv->zv_readonly || (zv->zv_mode & DS_MODE_READONLY)) { 549 mutex_exit(&zvol_state_lock); 550 return (EROFS); 551 } 552 553 tx = dmu_tx_create(zv->zv_objset); 554 dmu_tx_hold_bonus(tx, ZVOL_OBJ); 555 error = dmu_tx_assign(tx, TXG_WAIT); 556 if (error) { 557 dmu_tx_abort(tx); 558 } else { 559 error = dmu_object_set_blocksize(zv->zv_objset, ZVOL_OBJ, 560 zc->zc_volblocksize, 0, tx); 561 if (error == ENOTSUP) 562 error = EBUSY; 563 dmu_tx_commit(tx); 564 } 565 566 mutex_exit(&zvol_state_lock); 567 568 return (error); 569 } 570 571 /*ARGSUSED*/ 572 int 573 zvol_open(dev_t *devp, int flag, int otyp, cred_t *cr) 574 { 575 minor_t minor = getminor(*devp); 576 zvol_state_t *zv; 577 578 if (minor == 0) /* This is the control device */ 579 return (0); 580 581 mutex_enter(&zvol_state_lock); 582 583 zv = ddi_get_soft_state(zvol_state, minor); 584 if (zv == NULL) { 585 mutex_exit(&zvol_state_lock); 586 return (ENXIO); 587 } 588 589 ASSERT(zv->zv_objset != NULL); 590 591 if ((flag & FWRITE) && 592 (zv->zv_readonly || (zv->zv_mode & DS_MODE_READONLY))) { 593 mutex_exit(&zvol_state_lock); 594 return (EROFS); 595 } 596 597 if (zv->zv_open_count[otyp] == 0 || otyp == OTYP_LYR) { 598 zv->zv_open_count[otyp]++; 599 zv->zv_total_opens++; 600 } 601 602 mutex_exit(&zvol_state_lock); 603 604 return (0); 605 } 606 607 /*ARGSUSED*/ 608 int 609 zvol_close(dev_t dev, int flag, int otyp, cred_t *cr) 610 { 611 minor_t minor = getminor(dev); 612 zvol_state_t *zv; 613 614 if (minor == 0) /* This is the control device */ 615 return (0); 616 617 mutex_enter(&zvol_state_lock); 618 619 zv = ddi_get_soft_state(zvol_state, minor); 620 if (zv == NULL) { 621 mutex_exit(&zvol_state_lock); 622 return (ENXIO); 623 } 624 625 /* 626 * The next statement is a workaround for the following DDI bug: 627 * 6343604 specfs race: multiple "last-close" of the same device 628 */ 629 if (zv->zv_total_opens == 0) { 630 mutex_exit(&zvol_state_lock); 631 return (0); 632 } 633 634 /* 635 * If the open count is zero, this is a spurious close. 636 * That indicates a bug in the kernel / DDI framework. 637 */ 638 ASSERT(zv->zv_open_count[otyp] != 0); 639 ASSERT(zv->zv_total_opens != 0); 640 641 /* 642 * You may get multiple opens, but only one close. 643 */ 644 zv->zv_open_count[otyp]--; 645 zv->zv_total_opens--; 646 647 mutex_exit(&zvol_state_lock); 648 649 return (0); 650 } 651 652 /* 653 * Create and return an immediate write ZIL transaction. 654 */ 655 itx_t * 656 zvol_immediate_itx(offset_t off, ssize_t len, char *addr) 657 { 658 itx_t *itx; 659 lr_write_t *lr; 660 661 itx = zil_itx_create(TX_WRITE, sizeof (*lr) + len); 662 lr = (lr_write_t *)&itx->itx_lr; 663 lr->lr_foid = ZVOL_OBJ; 664 lr->lr_offset = off; 665 lr->lr_length = len; 666 lr->lr_blkoff = 0; 667 BP_ZERO(&lr->lr_blkptr); 668 bcopy(addr, (char *)itx + offsetof(itx_t, itx_lr) + 669 sizeof (*lr), len); 670 itx->itx_wr_state = WR_COPIED; 671 return (itx); 672 } 673 674 /* 675 * zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions. 676 * 677 * We store data in the log buffers if it's small enough. 678 * Otherwise we flush the data out via dmu_sync(). 679 */ 680 ssize_t zvol_immediate_write_sz = 65536; 681 682 int 683 zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, offset_t off, ssize_t len, 684 char *addr) 685 { 686 dmu_object_info_t doi; 687 ssize_t nbytes; 688 itx_t *itx; 689 lr_write_t *lr; 690 objset_t *os; 691 uint64_t txg; 692 int error; 693 uint32_t blocksize; 694 695 /* handle common case */ 696 if (len <= zvol_immediate_write_sz) { 697 itx = zvol_immediate_itx(off, len, addr); 698 (void) zil_itx_assign(zv->zv_zilog, itx, tx); 699 return (0); 700 } 701 702 txg = dmu_tx_get_txg(tx); 703 os = zv->zv_objset; 704 705 /* 706 * We need to dmu_sync() each block in the range. 707 * For this we need the blocksize. 708 */ 709 error = dmu_object_info(os, ZVOL_OBJ, &doi); 710 if (error) 711 return (error); 712 blocksize = doi.doi_data_block_size; 713 714 /* 715 * We need to immediate write or dmu_sync() each block in the range. 716 */ 717 while (len) { 718 nbytes = MIN(len, blocksize - P2PHASE(off, blocksize)); 719 if (nbytes <= zvol_immediate_write_sz) { 720 itx = zvol_immediate_itx(off, nbytes, addr); 721 } else { 722 itx = zil_itx_create(TX_WRITE, sizeof (*lr)); 723 lr = (lr_write_t *)&itx->itx_lr; 724 lr->lr_foid = ZVOL_OBJ; 725 lr->lr_offset = off; 726 lr->lr_length = nbytes; 727 lr->lr_blkoff = 0; 728 BP_ZERO(&lr->lr_blkptr); 729 730 txg_suspend(dmu_objset_pool(os)); 731 error = dmu_sync(os, ZVOL_OBJ, off, &lr->lr_blkoff, 732 &lr->lr_blkptr, txg); 733 txg_resume(dmu_objset_pool(os)); 734 if (error) { 735 kmem_free(itx, offsetof(itx_t, itx_lr)); 736 return (error); 737 } 738 itx->itx_wr_state = WR_COPIED; 739 } 740 (void) zil_itx_assign(zv->zv_zilog, itx, tx); 741 len -= nbytes; 742 off += nbytes; 743 } 744 return (0); 745 } 746 747 int 748 zvol_strategy(buf_t *bp) 749 { 750 zvol_state_t *zv = ddi_get_soft_state(zvol_state, getminor(bp->b_edev)); 751 uint64_t off, volsize; 752 size_t size, resid; 753 char *addr; 754 objset_t *os; 755 int error = 0; 756 int sync; 757 int reading; 758 int txg_sync_needed = B_FALSE; 759 760 if (zv == NULL) { 761 bioerror(bp, ENXIO); 762 biodone(bp); 763 return (0); 764 } 765 766 if (getminor(bp->b_edev) == 0) { 767 bioerror(bp, EINVAL); 768 biodone(bp); 769 return (0); 770 } 771 772 if (zv->zv_readonly && !(bp->b_flags & B_READ)) { 773 bioerror(bp, EROFS); 774 biodone(bp); 775 return (0); 776 } 777 778 off = ldbtob(bp->b_blkno); 779 volsize = zv->zv_volsize; 780 781 os = zv->zv_objset; 782 ASSERT(os != NULL); 783 sync = !(bp->b_flags & B_ASYNC) && !(zil_disable); 784 785 bp_mapin(bp); 786 addr = bp->b_un.b_addr; 787 resid = bp->b_bcount; 788 789 /* 790 * There must be no buffer changes when doing a dmu_sync() because 791 * we can't change the data whilst calculating the checksum. 792 * A better approach than a per zvol rwlock would be to lock ranges. 793 */ 794 reading = bp->b_flags & B_READ; 795 if (reading || resid <= zvol_immediate_write_sz) 796 rw_enter(&zv->zv_dslock, RW_READER); 797 else 798 rw_enter(&zv->zv_dslock, RW_WRITER); 799 800 while (resid != 0 && off < volsize) { 801 802 size = MIN(resid, 1UL << 20); /* cap at 1MB per tx */ 803 804 if (size > volsize - off) /* don't write past the end */ 805 size = volsize - off; 806 807 if (reading) { 808 error = dmu_read(os, ZVOL_OBJ, off, size, addr); 809 } else { 810 dmu_tx_t *tx = dmu_tx_create(os); 811 dmu_tx_hold_write(tx, ZVOL_OBJ, off, size); 812 error = dmu_tx_assign(tx, TXG_WAIT); 813 if (error) { 814 dmu_tx_abort(tx); 815 } else { 816 dmu_write(os, ZVOL_OBJ, off, size, addr, tx); 817 if (sync) { 818 /* use the ZIL to commit this write */ 819 if (zvol_log_write(zv, tx, off, size, 820 addr) != 0) { 821 txg_sync_needed = B_TRUE; 822 } 823 } 824 dmu_tx_commit(tx); 825 } 826 } 827 if (error) 828 break; 829 off += size; 830 addr += size; 831 resid -= size; 832 } 833 rw_exit(&zv->zv_dslock); 834 835 if ((bp->b_resid = resid) == bp->b_bcount) 836 bioerror(bp, off > volsize ? EINVAL : error); 837 838 biodone(bp); 839 840 if (sync) { 841 if (txg_sync_needed) 842 txg_wait_synced(dmu_objset_pool(os), 0); 843 else 844 zil_commit(zv->zv_zilog, UINT64_MAX, FDSYNC); 845 } 846 847 return (0); 848 } 849 850 /*ARGSUSED*/ 851 int 852 zvol_read(dev_t dev, uio_t *uiop, cred_t *cr) 853 { 854 return (physio(zvol_strategy, NULL, dev, B_READ, minphys, uiop)); 855 } 856 857 /*ARGSUSED*/ 858 int 859 zvol_write(dev_t dev, uio_t *uiop, cred_t *cr) 860 { 861 return (physio(zvol_strategy, NULL, dev, B_WRITE, minphys, uiop)); 862 } 863 864 /*ARGSUSED*/ 865 int 866 zvol_aread(dev_t dev, struct aio_req *aio, cred_t *cr) 867 { 868 return (aphysio(zvol_strategy, anocancel, dev, B_READ, minphys, aio)); 869 } 870 871 /*ARGSUSED*/ 872 int 873 zvol_awrite(dev_t dev, struct aio_req *aio, cred_t *cr) 874 { 875 return (aphysio(zvol_strategy, anocancel, dev, B_WRITE, minphys, aio)); 876 } 877 878 /* 879 * Dirtbag ioctls to support mkfs(1M) for UFS filesystems. See dkio(7I). 880 */ 881 /*ARGSUSED*/ 882 int 883 zvol_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 884 { 885 zvol_state_t *zv; 886 struct dk_cinfo dkc; 887 struct dk_minfo dkm; 888 dk_efi_t efi; 889 efi_gpt_t gpt; 890 efi_gpe_t gpe; 891 struct uuid uuid = EFI_RESERVED; 892 uint32_t crc; 893 int error = 0; 894 895 mutex_enter(&zvol_state_lock); 896 897 zv = ddi_get_soft_state(zvol_state, getminor(dev)); 898 899 if (zv == NULL) { 900 mutex_exit(&zvol_state_lock); 901 return (ENXIO); 902 } 903 904 switch (cmd) { 905 906 case DKIOCINFO: 907 bzero(&dkc, sizeof (dkc)); 908 (void) strcpy(dkc.dki_cname, "zvol"); 909 (void) strcpy(dkc.dki_dname, "zvol"); 910 dkc.dki_ctype = DKC_UNKNOWN; 911 dkc.dki_maxtransfer = 1 << (SPA_MAXBLOCKSHIFT - zv->zv_min_bs); 912 mutex_exit(&zvol_state_lock); 913 if (ddi_copyout(&dkc, (void *)arg, sizeof (dkc), flag)) 914 error = EFAULT; 915 return (error); 916 917 case DKIOCGMEDIAINFO: 918 bzero(&dkm, sizeof (dkm)); 919 dkm.dki_lbsize = 1U << zv->zv_min_bs; 920 dkm.dki_capacity = zv->zv_volsize >> zv->zv_min_bs; 921 dkm.dki_media_type = DK_UNKNOWN; 922 mutex_exit(&zvol_state_lock); 923 if (ddi_copyout(&dkm, (void *)arg, sizeof (dkm), flag)) 924 error = EFAULT; 925 return (error); 926 927 case DKIOCGETEFI: 928 if (ddi_copyin((void *)arg, &efi, sizeof (dk_efi_t), flag)) { 929 mutex_exit(&zvol_state_lock); 930 return (EFAULT); 931 } 932 933 bzero(&gpt, sizeof (gpt)); 934 bzero(&gpe, sizeof (gpe)); 935 936 efi.dki_data = (void *)(uintptr_t)efi.dki_data_64; 937 938 if (efi.dki_length < sizeof (gpt) + sizeof (gpe)) { 939 mutex_exit(&zvol_state_lock); 940 return (EINVAL); 941 } 942 943 efi.dki_length = sizeof (gpt) + sizeof (gpe); 944 945 gpt.efi_gpt_Signature = LE_64(EFI_SIGNATURE); 946 gpt.efi_gpt_Revision = LE_32(EFI_VERSION_CURRENT); 947 gpt.efi_gpt_HeaderSize = LE_32(sizeof (gpt)); 948 gpt.efi_gpt_FirstUsableLBA = LE_64(0ULL); 949 gpt.efi_gpt_LastUsableLBA = 950 LE_64((zv->zv_volsize >> zv->zv_min_bs) - 1); 951 gpt.efi_gpt_NumberOfPartitionEntries = LE_32(1); 952 gpt.efi_gpt_SizeOfPartitionEntry = LE_32(sizeof (gpe)); 953 954 UUID_LE_CONVERT(gpe.efi_gpe_PartitionTypeGUID, uuid); 955 gpe.efi_gpe_StartingLBA = gpt.efi_gpt_FirstUsableLBA; 956 gpe.efi_gpe_EndingLBA = gpt.efi_gpt_LastUsableLBA; 957 958 CRC32(crc, &gpe, sizeof (gpe), -1U, crc32_table); 959 gpt.efi_gpt_PartitionEntryArrayCRC32 = LE_32(~crc); 960 961 CRC32(crc, &gpt, sizeof (gpt), -1U, crc32_table); 962 gpt.efi_gpt_HeaderCRC32 = LE_32(~crc); 963 964 mutex_exit(&zvol_state_lock); 965 if (ddi_copyout(&gpt, efi.dki_data, sizeof (gpt), flag) || 966 ddi_copyout(&gpe, efi.dki_data + 1, sizeof (gpe), flag)) 967 error = EFAULT; 968 return (error); 969 970 default: 971 error = ENOTSUP; 972 break; 973 974 } 975 mutex_exit(&zvol_state_lock); 976 return (error); 977 } 978 979 int 980 zvol_busy(void) 981 { 982 return (zvol_minors != 0); 983 } 984 985 void 986 zvol_init(void) 987 { 988 VERIFY(ddi_soft_state_init(&zvol_state, sizeof (zvol_state_t), 1) == 0); 989 mutex_init(&zvol_state_lock, NULL, MUTEX_DEFAULT, NULL); 990 } 991 992 void 993 zvol_fini(void) 994 { 995 mutex_destroy(&zvol_state_lock); 996 ddi_soft_state_fini(&zvol_state); 997 } 998