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 #include <sys/refcount.h> 72 73 #include "zfs_namecheck.h" 74 75 #define ZVOL_OBJ 1ULL 76 #define ZVOL_ZAP_OBJ 2ULL 77 78 static void *zvol_state; 79 80 /* 81 * This lock protects the zvol_state structure from being modified 82 * while it's being used, e.g. an open that comes in before a create 83 * finishes. It also protects temporary opens of the dataset so that, 84 * e.g., an open doesn't get a spurious EBUSY. 85 */ 86 static kmutex_t zvol_state_lock; 87 static uint32_t zvol_minors; 88 89 /* 90 * The in-core state of each volume. 91 */ 92 typedef struct zvol_state { 93 char zv_name[MAXPATHLEN]; /* pool/dd name */ 94 uint64_t zv_volsize; /* amount of space we advertise */ 95 minor_t zv_minor; /* minor number */ 96 uint8_t zv_min_bs; /* minimum addressable block shift */ 97 uint8_t zv_readonly; /* hard readonly; like write-protect */ 98 objset_t *zv_objset; /* objset handle */ 99 uint32_t zv_mode; /* DS_MODE_* flags at open time */ 100 uint32_t zv_open_count[OTYPCNT]; /* open counts */ 101 uint32_t zv_total_opens; /* total open count */ 102 zilog_t *zv_zilog; /* ZIL handle */ 103 uint64_t zv_txg_assign; /* txg to assign during ZIL replay */ 104 krwlock_t zv_dslock; /* dmu_sync() rwlock */ 105 } zvol_state_t; 106 107 static void 108 zvol_size_changed(zvol_state_t *zv, dev_t dev) 109 { 110 dev = makedevice(getmajor(dev), zv->zv_minor); 111 112 VERIFY(ddi_prop_update_int64(dev, zfs_dip, 113 "Size", zv->zv_volsize) == DDI_SUCCESS); 114 VERIFY(ddi_prop_update_int64(dev, zfs_dip, 115 "Nblocks", lbtodb(zv->zv_volsize)) == DDI_SUCCESS); 116 } 117 118 int 119 zvol_check_volsize(zfs_cmd_t *zc, uint64_t blocksize) 120 { 121 if (zc->zc_volsize == 0) 122 return (EINVAL); 123 124 if (zc->zc_volsize % blocksize != 0) 125 return (EINVAL); 126 127 #ifdef _ILP32 128 if (zc->zc_volsize - 1 > SPEC_MAXOFFSET_T) 129 return (EOVERFLOW); 130 #endif 131 return (0); 132 } 133 134 int 135 zvol_check_volblocksize(zfs_cmd_t *zc) 136 { 137 if (zc->zc_volblocksize < SPA_MINBLOCKSIZE || 138 zc->zc_volblocksize > SPA_MAXBLOCKSIZE || 139 !ISP2(zc->zc_volblocksize)) 140 return (EDOM); 141 142 return (0); 143 } 144 145 static void 146 zvol_readonly_changed_cb(void *arg, uint64_t newval) 147 { 148 zvol_state_t *zv = arg; 149 150 zv->zv_readonly = (uint8_t)newval; 151 } 152 153 int 154 zvol_get_stats(zfs_cmd_t *zc, objset_t *os) 155 { 156 int error; 157 dmu_object_info_t doi; 158 159 error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &zc->zc_volsize); 160 161 if (error) 162 return (error); 163 164 error = dmu_object_info(os, ZVOL_OBJ, &doi); 165 166 if (error == 0) 167 zc->zc_volblocksize = doi.doi_data_block_size; 168 169 return (error); 170 } 171 172 /* 173 * Find a free minor number. 174 */ 175 static minor_t 176 zvol_minor_alloc(void) 177 { 178 minor_t minor; 179 180 ASSERT(MUTEX_HELD(&zvol_state_lock)); 181 182 for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) 183 if (ddi_get_soft_state(zvol_state, minor) == NULL) 184 return (minor); 185 186 return (0); 187 } 188 189 static zvol_state_t * 190 zvol_minor_lookup(char *name) 191 { 192 minor_t minor; 193 zvol_state_t *zv; 194 195 ASSERT(MUTEX_HELD(&zvol_state_lock)); 196 197 for (minor = 1; minor <= ZVOL_MAX_MINOR; minor++) { 198 zv = ddi_get_soft_state(zvol_state, minor); 199 if (zv == NULL) 200 continue; 201 if (strcmp(zv->zv_name, name) == 0) 202 break; 203 } 204 205 return (zv); 206 } 207 208 void 209 zvol_create_cb(objset_t *os, void *arg, dmu_tx_t *tx) 210 { 211 zfs_cmd_t *zc = arg; 212 int error; 213 214 error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, zc->zc_volblocksize, 215 DMU_OT_NONE, 0, tx); 216 ASSERT(error == 0); 217 218 error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP, 219 DMU_OT_NONE, 0, tx); 220 ASSERT(error == 0); 221 222 error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &zc->zc_volsize, tx); 223 ASSERT(error == 0); 224 } 225 226 /* 227 * Replay a TX_WRITE ZIL transaction that didn't get committed 228 * after a system failure 229 */ 230 static int 231 zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap) 232 { 233 objset_t *os = zv->zv_objset; 234 char *data = (char *)(lr + 1); /* data follows lr_write_t */ 235 uint64_t off = lr->lr_offset; 236 uint64_t len = lr->lr_length; 237 dmu_tx_t *tx; 238 int error; 239 240 if (byteswap) 241 byteswap_uint64_array(lr, sizeof (*lr)); 242 243 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 } else { 249 dmu_write(os, ZVOL_OBJ, off, len, data, tx); 250 dmu_tx_commit(tx); 251 } 252 253 return (error); 254 } 255 256 /* ARGSUSED */ 257 static int 258 zvol_replay_err(zvol_state_t *zv, lr_t *lr, boolean_t byteswap) 259 { 260 return (ENOTSUP); 261 } 262 263 /* 264 * Callback vectors for replaying records. 265 * Only TX_WRITE is needed for zvol. 266 */ 267 zil_replay_func_t *zvol_replay_vector[TX_MAX_TYPE] = { 268 zvol_replay_err, /* 0 no such transaction type */ 269 zvol_replay_err, /* TX_CREATE */ 270 zvol_replay_err, /* TX_MKDIR */ 271 zvol_replay_err, /* TX_MKXATTR */ 272 zvol_replay_err, /* TX_SYMLINK */ 273 zvol_replay_err, /* TX_REMOVE */ 274 zvol_replay_err, /* TX_RMDIR */ 275 zvol_replay_err, /* TX_LINK */ 276 zvol_replay_err, /* TX_RENAME */ 277 zvol_replay_write, /* TX_WRITE */ 278 zvol_replay_err, /* TX_TRUNCATE */ 279 zvol_replay_err, /* TX_SETATTR */ 280 zvol_replay_err, /* TX_ACL */ 281 }; 282 283 /* 284 * Create a minor node for the specified volume. 285 */ 286 int 287 zvol_create_minor(zfs_cmd_t *zc) 288 { 289 char *name = zc->zc_name; 290 dev_t dev = zc->zc_dev; 291 zvol_state_t *zv; 292 objset_t *os; 293 uint64_t volsize; 294 minor_t minor = 0; 295 struct pathname linkpath; 296 int ds_mode = DS_MODE_PRIMARY; 297 vnode_t *vp = NULL; 298 char *devpath; 299 size_t devpathlen = strlen(ZVOL_FULL_DEV_DIR) + 1 + strlen(name) + 1; 300 char chrbuf[30], blkbuf[30]; 301 int error; 302 303 mutex_enter(&zvol_state_lock); 304 305 if ((zv = zvol_minor_lookup(name)) != NULL) { 306 mutex_exit(&zvol_state_lock); 307 return (EEXIST); 308 } 309 310 if (strchr(name, '@') != 0) 311 ds_mode |= DS_MODE_READONLY; 312 313 error = dmu_objset_open(name, DMU_OST_ZVOL, ds_mode, &os); 314 315 if (error) { 316 mutex_exit(&zvol_state_lock); 317 return (error); 318 } 319 320 error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); 321 322 if (error) { 323 dmu_objset_close(os); 324 mutex_exit(&zvol_state_lock); 325 return (error); 326 } 327 328 /* 329 * If there's an existing /dev/zvol symlink, try to use the 330 * same minor number we used last time. 331 */ 332 devpath = kmem_alloc(devpathlen, KM_SLEEP); 333 334 (void) sprintf(devpath, "%s/%s", ZVOL_FULL_DEV_DIR, name); 335 336 error = lookupname(devpath, UIO_SYSSPACE, NO_FOLLOW, NULL, &vp); 337 338 kmem_free(devpath, devpathlen); 339 340 if (error == 0 && vp->v_type != VLNK) 341 error = EINVAL; 342 343 if (error == 0) { 344 pn_alloc(&linkpath); 345 error = pn_getsymlink(vp, &linkpath, kcred); 346 if (error == 0) { 347 char *ms = strstr(linkpath.pn_path, ZVOL_PSEUDO_DEV); 348 if (ms != NULL) { 349 ms += strlen(ZVOL_PSEUDO_DEV); 350 minor = stoi(&ms); 351 } 352 } 353 pn_free(&linkpath); 354 } 355 356 if (vp != NULL) 357 VN_RELE(vp); 358 359 /* 360 * If we found a minor but it's already in use, we must pick a new one. 361 */ 362 if (minor != 0 && ddi_get_soft_state(zvol_state, minor) != NULL) 363 minor = 0; 364 365 if (minor == 0) 366 minor = zvol_minor_alloc(); 367 368 if (minor == 0) { 369 dmu_objset_close(os); 370 mutex_exit(&zvol_state_lock); 371 return (ENXIO); 372 } 373 374 if (ddi_soft_state_zalloc(zvol_state, minor) != DDI_SUCCESS) { 375 dmu_objset_close(os); 376 mutex_exit(&zvol_state_lock); 377 return (EAGAIN); 378 } 379 380 (void) ddi_prop_update_string(minor, zfs_dip, ZVOL_PROP_NAME, name); 381 382 (void) sprintf(chrbuf, "%uc,raw", minor); 383 384 if (ddi_create_minor_node(zfs_dip, chrbuf, S_IFCHR, 385 minor, DDI_PSEUDO, 0) == DDI_FAILURE) { 386 ddi_soft_state_free(zvol_state, minor); 387 dmu_objset_close(os); 388 mutex_exit(&zvol_state_lock); 389 return (EAGAIN); 390 } 391 392 (void) sprintf(blkbuf, "%uc", minor); 393 394 if (ddi_create_minor_node(zfs_dip, blkbuf, S_IFBLK, 395 minor, DDI_PSEUDO, 0) == DDI_FAILURE) { 396 ddi_remove_minor_node(zfs_dip, chrbuf); 397 ddi_soft_state_free(zvol_state, minor); 398 dmu_objset_close(os); 399 mutex_exit(&zvol_state_lock); 400 return (EAGAIN); 401 } 402 403 zv = ddi_get_soft_state(zvol_state, minor); 404 405 (void) strcpy(zv->zv_name, name); 406 zv->zv_min_bs = DEV_BSHIFT; 407 zv->zv_minor = minor; 408 zv->zv_volsize = volsize; 409 zv->zv_objset = os; 410 zv->zv_mode = ds_mode; 411 zv->zv_zilog = zil_open(os, NULL); 412 413 rw_init(&zv->zv_dslock, NULL, RW_DEFAULT, NULL); 414 415 zil_replay(os, zv, &zv->zv_txg_assign, zvol_replay_vector, NULL); 416 417 zvol_size_changed(zv, dev); 418 419 /* XXX this should handle the possible i/o error */ 420 VERIFY(dsl_prop_register(dmu_objset_ds(zv->zv_objset), 421 "readonly", zvol_readonly_changed_cb, zv) == 0); 422 423 zvol_minors++; 424 425 mutex_exit(&zvol_state_lock); 426 427 return (0); 428 } 429 430 /* 431 * Remove minor node for the specified volume. 432 */ 433 int 434 zvol_remove_minor(zfs_cmd_t *zc) 435 { 436 zvol_state_t *zv; 437 char namebuf[30]; 438 439 mutex_enter(&zvol_state_lock); 440 441 if ((zv = zvol_minor_lookup(zc->zc_name)) == NULL) { 442 mutex_exit(&zvol_state_lock); 443 return (ENXIO); 444 } 445 446 if (zv->zv_total_opens != 0) { 447 mutex_exit(&zvol_state_lock); 448 return (EBUSY); 449 } 450 451 (void) sprintf(namebuf, "%uc,raw", zv->zv_minor); 452 ddi_remove_minor_node(zfs_dip, namebuf); 453 454 (void) sprintf(namebuf, "%uc", zv->zv_minor); 455 ddi_remove_minor_node(zfs_dip, namebuf); 456 457 VERIFY(dsl_prop_unregister(dmu_objset_ds(zv->zv_objset), 458 "readonly", zvol_readonly_changed_cb, zv) == 0); 459 460 zil_close(zv->zv_zilog); 461 zv->zv_zilog = NULL; 462 dmu_objset_close(zv->zv_objset); 463 zv->zv_objset = NULL; 464 465 ddi_soft_state_free(zvol_state, zv->zv_minor); 466 467 zvol_minors--; 468 469 mutex_exit(&zvol_state_lock); 470 471 return (0); 472 } 473 474 int 475 zvol_set_volsize(zfs_cmd_t *zc) 476 { 477 zvol_state_t *zv; 478 dev_t dev = zc->zc_dev; 479 dmu_tx_t *tx; 480 int error; 481 dmu_object_info_t doi; 482 483 mutex_enter(&zvol_state_lock); 484 485 if ((zv = zvol_minor_lookup(zc->zc_name)) == NULL) { 486 mutex_exit(&zvol_state_lock); 487 return (ENXIO); 488 } 489 490 if ((error = dmu_object_info(zv->zv_objset, ZVOL_OBJ, &doi)) != 0 || 491 (error = zvol_check_volsize(zc, doi.doi_data_block_size)) != 0) { 492 mutex_exit(&zvol_state_lock); 493 return (error); 494 } 495 496 if (zv->zv_readonly || (zv->zv_mode & DS_MODE_READONLY)) { 497 mutex_exit(&zvol_state_lock); 498 return (EROFS); 499 } 500 501 tx = dmu_tx_create(zv->zv_objset); 502 dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); 503 dmu_tx_hold_free(tx, ZVOL_OBJ, zc->zc_volsize, DMU_OBJECT_END); 504 error = dmu_tx_assign(tx, TXG_WAIT); 505 if (error) { 506 dmu_tx_abort(tx); 507 mutex_exit(&zvol_state_lock); 508 return (error); 509 } 510 511 error = zap_update(zv->zv_objset, ZVOL_ZAP_OBJ, "size", 8, 1, 512 &zc->zc_volsize, tx); 513 if (error == 0) { 514 error = dmu_free_range(zv->zv_objset, ZVOL_OBJ, zc->zc_volsize, 515 DMU_OBJECT_END, tx); 516 } 517 518 dmu_tx_commit(tx); 519 520 if (error == 0) { 521 zv->zv_volsize = zc->zc_volsize; 522 zvol_size_changed(zv, dev); 523 } 524 525 mutex_exit(&zvol_state_lock); 526 527 return (error); 528 } 529 530 int 531 zvol_set_volblocksize(zfs_cmd_t *zc) 532 { 533 zvol_state_t *zv; 534 dmu_tx_t *tx; 535 int error; 536 537 mutex_enter(&zvol_state_lock); 538 539 if ((zv = zvol_minor_lookup(zc->zc_name)) == NULL) { 540 mutex_exit(&zvol_state_lock); 541 return (ENXIO); 542 } 543 544 if (zv->zv_readonly || (zv->zv_mode & DS_MODE_READONLY)) { 545 mutex_exit(&zvol_state_lock); 546 return (EROFS); 547 } 548 549 tx = dmu_tx_create(zv->zv_objset); 550 dmu_tx_hold_bonus(tx, ZVOL_OBJ); 551 error = dmu_tx_assign(tx, TXG_WAIT); 552 if (error) { 553 dmu_tx_abort(tx); 554 } else { 555 error = dmu_object_set_blocksize(zv->zv_objset, ZVOL_OBJ, 556 zc->zc_volblocksize, 0, tx); 557 if (error == ENOTSUP) 558 error = EBUSY; 559 dmu_tx_commit(tx); 560 } 561 562 mutex_exit(&zvol_state_lock); 563 564 return (error); 565 } 566 567 /*ARGSUSED*/ 568 int 569 zvol_open(dev_t *devp, int flag, int otyp, cred_t *cr) 570 { 571 minor_t minor = getminor(*devp); 572 zvol_state_t *zv; 573 574 if (minor == 0) /* This is the control device */ 575 return (0); 576 577 mutex_enter(&zvol_state_lock); 578 579 zv = ddi_get_soft_state(zvol_state, minor); 580 if (zv == NULL) { 581 mutex_exit(&zvol_state_lock); 582 return (ENXIO); 583 } 584 585 ASSERT(zv->zv_objset != NULL); 586 587 if ((flag & FWRITE) && 588 (zv->zv_readonly || (zv->zv_mode & DS_MODE_READONLY))) { 589 mutex_exit(&zvol_state_lock); 590 return (EROFS); 591 } 592 593 if (zv->zv_open_count[otyp] == 0 || otyp == OTYP_LYR) { 594 zv->zv_open_count[otyp]++; 595 zv->zv_total_opens++; 596 } 597 598 mutex_exit(&zvol_state_lock); 599 600 return (0); 601 } 602 603 /*ARGSUSED*/ 604 int 605 zvol_close(dev_t dev, int flag, int otyp, cred_t *cr) 606 { 607 minor_t minor = getminor(dev); 608 zvol_state_t *zv; 609 610 if (minor == 0) /* This is the control device */ 611 return (0); 612 613 mutex_enter(&zvol_state_lock); 614 615 zv = ddi_get_soft_state(zvol_state, minor); 616 if (zv == NULL) { 617 mutex_exit(&zvol_state_lock); 618 return (ENXIO); 619 } 620 621 /* 622 * The next statement is a workaround for the following DDI bug: 623 * 6343604 specfs race: multiple "last-close" of the same device 624 */ 625 if (zv->zv_total_opens == 0) { 626 mutex_exit(&zvol_state_lock); 627 return (0); 628 } 629 630 /* 631 * If the open count is zero, this is a spurious close. 632 * That indicates a bug in the kernel / DDI framework. 633 */ 634 ASSERT(zv->zv_open_count[otyp] != 0); 635 ASSERT(zv->zv_total_opens != 0); 636 637 /* 638 * You may get multiple opens, but only one close. 639 */ 640 zv->zv_open_count[otyp]--; 641 zv->zv_total_opens--; 642 643 mutex_exit(&zvol_state_lock); 644 645 return (0); 646 } 647 648 /* 649 * Create and return an immediate write ZIL transaction. 650 */ 651 itx_t * 652 zvol_immediate_itx(offset_t off, ssize_t len, char *addr) 653 { 654 itx_t *itx; 655 lr_write_t *lr; 656 657 itx = zil_itx_create(TX_WRITE, sizeof (*lr) + len); 658 lr = (lr_write_t *)&itx->itx_lr; 659 lr->lr_foid = ZVOL_OBJ; 660 lr->lr_offset = off; 661 lr->lr_length = len; 662 lr->lr_blkoff = 0; 663 BP_ZERO(&lr->lr_blkptr); 664 bcopy(addr, (char *)itx + offsetof(itx_t, itx_lr) + 665 sizeof (*lr), len); 666 itx->itx_wr_state = WR_COPIED; 667 return (itx); 668 } 669 670 /* 671 * zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions. 672 * 673 * We store data in the log buffers if it's small enough. 674 * Otherwise we flush the data out via dmu_sync(). 675 */ 676 ssize_t zvol_immediate_write_sz = 65536; 677 678 int 679 zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, offset_t off, ssize_t len, 680 char *addr) 681 { 682 dmu_object_info_t doi; 683 ssize_t nbytes; 684 itx_t *itx; 685 lr_write_t *lr; 686 objset_t *os; 687 dmu_buf_t *db; 688 uint64_t txg; 689 uint64_t boff; 690 int error; 691 uint32_t blocksize; 692 693 /* handle common case */ 694 if (len <= zvol_immediate_write_sz) { 695 itx = zvol_immediate_itx(off, len, addr); 696 (void) zil_itx_assign(zv->zv_zilog, itx, tx); 697 return (0); 698 } 699 700 txg = dmu_tx_get_txg(tx); 701 os = zv->zv_objset; 702 703 /* 704 * We need to dmu_sync() each block in the range. 705 * For this we need the blocksize. 706 */ 707 error = dmu_object_info(os, ZVOL_OBJ, &doi); 708 if (error) 709 return (error); 710 blocksize = doi.doi_data_block_size; 711 712 /* 713 * We need to immediate write or dmu_sync() each block in the range. 714 */ 715 while (len) { 716 nbytes = MIN(len, blocksize - P2PHASE(off, blocksize)); 717 if (nbytes <= zvol_immediate_write_sz) { 718 itx = zvol_immediate_itx(off, nbytes, addr); 719 } else { 720 boff = P2ALIGN_TYPED(off, blocksize, uint64_t); 721 itx = zil_itx_create(TX_WRITE, sizeof (*lr)); 722 lr = (lr_write_t *)&itx->itx_lr; 723 lr->lr_foid = ZVOL_OBJ; 724 lr->lr_offset = off; 725 lr->lr_length = nbytes; 726 lr->lr_blkoff = off - boff; 727 BP_ZERO(&lr->lr_blkptr); 728 729 /* XXX - we should do these IOs in parallel */ 730 VERIFY(0 == dmu_buf_hold(os, ZVOL_OBJ, boff, 731 FTAG, &db)); 732 ASSERT(boff == db->db_offset); 733 error = dmu_sync(NULL, db, &lr->lr_blkptr, 734 txg, NULL, NULL); 735 dmu_buf_rele(db, FTAG); 736 if (error) { 737 kmem_free(itx, offsetof(itx_t, itx_lr)); 738 return (error); 739 } 740 itx->itx_wr_state = WR_COPIED; 741 } 742 (void) zil_itx_assign(zv->zv_zilog, itx, tx); 743 len -= nbytes; 744 off += nbytes; 745 } 746 return (0); 747 } 748 749 int 750 zvol_strategy(buf_t *bp) 751 { 752 zvol_state_t *zv = ddi_get_soft_state(zvol_state, getminor(bp->b_edev)); 753 uint64_t off, volsize; 754 size_t size, resid; 755 char *addr; 756 objset_t *os; 757 int error = 0; 758 int sync; 759 int reading; 760 int txg_sync_needed = B_FALSE; 761 762 if (zv == NULL) { 763 bioerror(bp, ENXIO); 764 biodone(bp); 765 return (0); 766 } 767 768 if (getminor(bp->b_edev) == 0) { 769 bioerror(bp, EINVAL); 770 biodone(bp); 771 return (0); 772 } 773 774 if (zv->zv_readonly && !(bp->b_flags & B_READ)) { 775 bioerror(bp, EROFS); 776 biodone(bp); 777 return (0); 778 } 779 780 off = ldbtob(bp->b_blkno); 781 volsize = zv->zv_volsize; 782 783 os = zv->zv_objset; 784 ASSERT(os != NULL); 785 sync = !(bp->b_flags & B_ASYNC) && !(zil_disable); 786 787 bp_mapin(bp); 788 addr = bp->b_un.b_addr; 789 resid = bp->b_bcount; 790 791 /* 792 * There must be no buffer changes when doing a dmu_sync() because 793 * we can't change the data whilst calculating the checksum. 794 * A better approach than a per zvol rwlock would be to lock ranges. 795 */ 796 reading = bp->b_flags & B_READ; 797 if (reading || resid <= zvol_immediate_write_sz) 798 rw_enter(&zv->zv_dslock, RW_READER); 799 else 800 rw_enter(&zv->zv_dslock, RW_WRITER); 801 802 while (resid != 0 && off < volsize) { 803 804 size = MIN(resid, 1UL << 20); /* cap at 1MB per tx */ 805 806 if (size > volsize - off) /* don't write past the end */ 807 size = volsize - off; 808 809 if (reading) { 810 error = dmu_read(os, ZVOL_OBJ, off, size, addr); 811 } else { 812 dmu_tx_t *tx = dmu_tx_create(os); 813 dmu_tx_hold_write(tx, ZVOL_OBJ, off, size); 814 error = dmu_tx_assign(tx, TXG_WAIT); 815 if (error) { 816 dmu_tx_abort(tx); 817 } else { 818 dmu_write(os, ZVOL_OBJ, off, size, addr, tx); 819 if (sync) { 820 /* use the ZIL to commit this write */ 821 if (zvol_log_write(zv, tx, off, size, 822 addr) != 0) { 823 txg_sync_needed = B_TRUE; 824 } 825 } 826 dmu_tx_commit(tx); 827 } 828 } 829 if (error) 830 break; 831 off += size; 832 addr += size; 833 resid -= size; 834 } 835 rw_exit(&zv->zv_dslock); 836 837 if ((bp->b_resid = resid) == bp->b_bcount) 838 bioerror(bp, off > volsize ? EINVAL : error); 839 840 biodone(bp); 841 842 if (sync) { 843 if (txg_sync_needed) 844 txg_wait_synced(dmu_objset_pool(os), 0); 845 else 846 zil_commit(zv->zv_zilog, UINT64_MAX, FDSYNC); 847 } 848 849 return (0); 850 } 851 852 /*ARGSUSED*/ 853 int 854 zvol_read(dev_t dev, uio_t *uiop, cred_t *cr) 855 { 856 return (physio(zvol_strategy, NULL, dev, B_READ, minphys, uiop)); 857 } 858 859 /*ARGSUSED*/ 860 int 861 zvol_write(dev_t dev, uio_t *uiop, cred_t *cr) 862 { 863 return (physio(zvol_strategy, NULL, dev, B_WRITE, minphys, uiop)); 864 } 865 866 /*ARGSUSED*/ 867 int 868 zvol_aread(dev_t dev, struct aio_req *aio, cred_t *cr) 869 { 870 return (aphysio(zvol_strategy, anocancel, dev, B_READ, minphys, aio)); 871 } 872 873 /*ARGSUSED*/ 874 int 875 zvol_awrite(dev_t dev, struct aio_req *aio, cred_t *cr) 876 { 877 return (aphysio(zvol_strategy, anocancel, dev, B_WRITE, minphys, aio)); 878 } 879 880 /* 881 * Dirtbag ioctls to support mkfs(1M) for UFS filesystems. See dkio(7I). 882 */ 883 /*ARGSUSED*/ 884 int 885 zvol_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) 886 { 887 zvol_state_t *zv; 888 struct dk_cinfo dkc; 889 struct dk_minfo dkm; 890 dk_efi_t efi; 891 efi_gpt_t gpt; 892 efi_gpe_t gpe; 893 struct uuid uuid = EFI_RESERVED; 894 uint32_t crc; 895 int error = 0; 896 897 mutex_enter(&zvol_state_lock); 898 899 zv = ddi_get_soft_state(zvol_state, getminor(dev)); 900 901 if (zv == NULL) { 902 mutex_exit(&zvol_state_lock); 903 return (ENXIO); 904 } 905 906 switch (cmd) { 907 908 case DKIOCINFO: 909 bzero(&dkc, sizeof (dkc)); 910 (void) strcpy(dkc.dki_cname, "zvol"); 911 (void) strcpy(dkc.dki_dname, "zvol"); 912 dkc.dki_ctype = DKC_UNKNOWN; 913 dkc.dki_maxtransfer = 1 << (SPA_MAXBLOCKSHIFT - zv->zv_min_bs); 914 mutex_exit(&zvol_state_lock); 915 if (ddi_copyout(&dkc, (void *)arg, sizeof (dkc), flag)) 916 error = EFAULT; 917 return (error); 918 919 case DKIOCGMEDIAINFO: 920 bzero(&dkm, sizeof (dkm)); 921 dkm.dki_lbsize = 1U << zv->zv_min_bs; 922 dkm.dki_capacity = zv->zv_volsize >> zv->zv_min_bs; 923 dkm.dki_media_type = DK_UNKNOWN; 924 mutex_exit(&zvol_state_lock); 925 if (ddi_copyout(&dkm, (void *)arg, sizeof (dkm), flag)) 926 error = EFAULT; 927 return (error); 928 929 case DKIOCGETEFI: 930 if (ddi_copyin((void *)arg, &efi, sizeof (dk_efi_t), flag)) { 931 mutex_exit(&zvol_state_lock); 932 return (EFAULT); 933 } 934 935 bzero(&gpt, sizeof (gpt)); 936 bzero(&gpe, sizeof (gpe)); 937 938 efi.dki_data = (void *)(uintptr_t)efi.dki_data_64; 939 940 if (efi.dki_length < sizeof (gpt) + sizeof (gpe)) { 941 mutex_exit(&zvol_state_lock); 942 return (EINVAL); 943 } 944 945 efi.dki_length = sizeof (gpt) + sizeof (gpe); 946 947 gpt.efi_gpt_Signature = LE_64(EFI_SIGNATURE); 948 gpt.efi_gpt_Revision = LE_32(EFI_VERSION_CURRENT); 949 gpt.efi_gpt_HeaderSize = LE_32(sizeof (gpt)); 950 gpt.efi_gpt_FirstUsableLBA = LE_64(0ULL); 951 gpt.efi_gpt_LastUsableLBA = 952 LE_64((zv->zv_volsize >> zv->zv_min_bs) - 1); 953 gpt.efi_gpt_NumberOfPartitionEntries = LE_32(1); 954 gpt.efi_gpt_SizeOfPartitionEntry = LE_32(sizeof (gpe)); 955 956 UUID_LE_CONVERT(gpe.efi_gpe_PartitionTypeGUID, uuid); 957 gpe.efi_gpe_StartingLBA = gpt.efi_gpt_FirstUsableLBA; 958 gpe.efi_gpe_EndingLBA = gpt.efi_gpt_LastUsableLBA; 959 960 CRC32(crc, &gpe, sizeof (gpe), -1U, crc32_table); 961 gpt.efi_gpt_PartitionEntryArrayCRC32 = LE_32(~crc); 962 963 CRC32(crc, &gpt, sizeof (gpt), -1U, crc32_table); 964 gpt.efi_gpt_HeaderCRC32 = LE_32(~crc); 965 966 mutex_exit(&zvol_state_lock); 967 if (ddi_copyout(&gpt, efi.dki_data, sizeof (gpt), flag) || 968 ddi_copyout(&gpe, efi.dki_data + 1, sizeof (gpe), flag)) 969 error = EFAULT; 970 return (error); 971 972 default: 973 error = ENOTSUP; 974 break; 975 976 } 977 mutex_exit(&zvol_state_lock); 978 return (error); 979 } 980 981 int 982 zvol_busy(void) 983 { 984 return (zvol_minors != 0); 985 } 986 987 void 988 zvol_init(void) 989 { 990 VERIFY(ddi_soft_state_init(&zvol_state, sizeof (zvol_state_t), 1) == 0); 991 mutex_init(&zvol_state_lock, NULL, MUTEX_DEFAULT, NULL); 992 } 993 994 void 995 zvol_fini(void) 996 { 997 mutex_destroy(&zvol_state_lock); 998 ddi_soft_state_fini(&zvol_state); 999 } 1000