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 2008 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #include <sys/dmu.h> 27 #include <sys/dmu_impl.h> 28 #include <sys/dmu_tx.h> 29 #include <sys/dbuf.h> 30 #include <sys/dnode.h> 31 #include <sys/zfs_context.h> 32 #include <sys/dmu_objset.h> 33 #include <sys/dmu_traverse.h> 34 #include <sys/dsl_dataset.h> 35 #include <sys/dsl_dir.h> 36 #include <sys/dsl_pool.h> 37 #include <sys/dsl_synctask.h> 38 #include <sys/dsl_prop.h> 39 #include <sys/dmu_zfetch.h> 40 #include <sys/zfs_ioctl.h> 41 #include <sys/zap.h> 42 #include <sys/zio_checksum.h> 43 #ifdef _KERNEL 44 #include <sys/vmsystm.h> 45 #include <sys/zfs_znode.h> 46 #endif 47 48 const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = { 49 { byteswap_uint8_array, TRUE, "unallocated" }, 50 { zap_byteswap, TRUE, "object directory" }, 51 { byteswap_uint64_array, TRUE, "object array" }, 52 { byteswap_uint8_array, TRUE, "packed nvlist" }, 53 { byteswap_uint64_array, TRUE, "packed nvlist size" }, 54 { byteswap_uint64_array, TRUE, "bplist" }, 55 { byteswap_uint64_array, TRUE, "bplist header" }, 56 { byteswap_uint64_array, TRUE, "SPA space map header" }, 57 { byteswap_uint64_array, TRUE, "SPA space map" }, 58 { byteswap_uint64_array, TRUE, "ZIL intent log" }, 59 { dnode_buf_byteswap, TRUE, "DMU dnode" }, 60 { dmu_objset_byteswap, TRUE, "DMU objset" }, 61 { byteswap_uint64_array, TRUE, "DSL directory" }, 62 { zap_byteswap, TRUE, "DSL directory child map"}, 63 { zap_byteswap, TRUE, "DSL dataset snap map" }, 64 { zap_byteswap, TRUE, "DSL props" }, 65 { byteswap_uint64_array, TRUE, "DSL dataset" }, 66 { zfs_znode_byteswap, TRUE, "ZFS znode" }, 67 { zfs_oldacl_byteswap, TRUE, "ZFS V0 ACL" }, 68 { byteswap_uint8_array, FALSE, "ZFS plain file" }, 69 { zap_byteswap, TRUE, "ZFS directory" }, 70 { zap_byteswap, TRUE, "ZFS master node" }, 71 { zap_byteswap, TRUE, "ZFS delete queue" }, 72 { byteswap_uint8_array, FALSE, "zvol object" }, 73 { zap_byteswap, TRUE, "zvol prop" }, 74 { byteswap_uint8_array, FALSE, "other uint8[]" }, 75 { byteswap_uint64_array, FALSE, "other uint64[]" }, 76 { zap_byteswap, TRUE, "other ZAP" }, 77 { zap_byteswap, TRUE, "persistent error log" }, 78 { byteswap_uint8_array, TRUE, "SPA history" }, 79 { byteswap_uint64_array, TRUE, "SPA history offsets" }, 80 { zap_byteswap, TRUE, "Pool properties" }, 81 { zap_byteswap, TRUE, "DSL permissions" }, 82 { zfs_acl_byteswap, TRUE, "ZFS ACL" }, 83 { byteswap_uint8_array, TRUE, "ZFS SYSACL" }, 84 { byteswap_uint8_array, TRUE, "FUID table" }, 85 { byteswap_uint64_array, TRUE, "FUID table size" }, 86 { zap_byteswap, TRUE, "DSL dataset next clones"}, 87 { zap_byteswap, TRUE, "scrub work queue" }, 88 }; 89 90 int 91 dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 92 void *tag, dmu_buf_t **dbp) 93 { 94 dnode_t *dn; 95 uint64_t blkid; 96 dmu_buf_impl_t *db; 97 int err; 98 99 err = dnode_hold(os->os, object, FTAG, &dn); 100 if (err) 101 return (err); 102 blkid = dbuf_whichblock(dn, offset); 103 rw_enter(&dn->dn_struct_rwlock, RW_READER); 104 db = dbuf_hold(dn, blkid, tag); 105 rw_exit(&dn->dn_struct_rwlock); 106 if (db == NULL) { 107 err = EIO; 108 } else { 109 err = dbuf_read(db, NULL, DB_RF_CANFAIL); 110 if (err) { 111 dbuf_rele(db, tag); 112 db = NULL; 113 } 114 } 115 116 dnode_rele(dn, FTAG); 117 *dbp = &db->db; 118 return (err); 119 } 120 121 int 122 dmu_bonus_max(void) 123 { 124 return (DN_MAX_BONUSLEN); 125 } 126 127 int 128 dmu_set_bonus(dmu_buf_t *db, int newsize, dmu_tx_t *tx) 129 { 130 dnode_t *dn = ((dmu_buf_impl_t *)db)->db_dnode; 131 132 if (dn->dn_bonus != (dmu_buf_impl_t *)db) 133 return (EINVAL); 134 if (newsize < 0 || newsize > db->db_size) 135 return (EINVAL); 136 dnode_setbonuslen(dn, newsize, tx); 137 return (0); 138 } 139 140 /* 141 * returns ENOENT, EIO, or 0. 142 */ 143 int 144 dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp) 145 { 146 dnode_t *dn; 147 dmu_buf_impl_t *db; 148 int error; 149 150 error = dnode_hold(os->os, object, FTAG, &dn); 151 if (error) 152 return (error); 153 154 rw_enter(&dn->dn_struct_rwlock, RW_READER); 155 if (dn->dn_bonus == NULL) { 156 rw_exit(&dn->dn_struct_rwlock); 157 rw_enter(&dn->dn_struct_rwlock, RW_WRITER); 158 if (dn->dn_bonus == NULL) 159 dbuf_create_bonus(dn); 160 } 161 db = dn->dn_bonus; 162 rw_exit(&dn->dn_struct_rwlock); 163 164 /* as long as the bonus buf is held, the dnode will be held */ 165 if (refcount_add(&db->db_holds, tag) == 1) 166 VERIFY(dnode_add_ref(dn, db)); 167 168 dnode_rele(dn, FTAG); 169 170 VERIFY(0 == dbuf_read(db, NULL, DB_RF_MUST_SUCCEED)); 171 172 *dbp = &db->db; 173 return (0); 174 } 175 176 /* 177 * Note: longer-term, we should modify all of the dmu_buf_*() interfaces 178 * to take a held dnode rather than <os, object> -- the lookup is wasteful, 179 * and can induce severe lock contention when writing to several files 180 * whose dnodes are in the same block. 181 */ 182 static int 183 dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, 184 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp) 185 { 186 dsl_pool_t *dp = NULL; 187 dmu_buf_t **dbp; 188 uint64_t blkid, nblks, i; 189 uint32_t flags; 190 int err; 191 zio_t *zio; 192 hrtime_t start; 193 194 ASSERT(length <= DMU_MAX_ACCESS); 195 196 flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT; 197 if (length > zfetch_array_rd_sz) 198 flags |= DB_RF_NOPREFETCH; 199 200 rw_enter(&dn->dn_struct_rwlock, RW_READER); 201 if (dn->dn_datablkshift) { 202 int blkshift = dn->dn_datablkshift; 203 nblks = (P2ROUNDUP(offset+length, 1ULL<<blkshift) - 204 P2ALIGN(offset, 1ULL<<blkshift)) >> blkshift; 205 } else { 206 if (offset + length > dn->dn_datablksz) { 207 zfs_panic_recover("zfs: accessing past end of object " 208 "%llx/%llx (size=%u access=%llu+%llu)", 209 (longlong_t)dn->dn_objset-> 210 os_dsl_dataset->ds_object, 211 (longlong_t)dn->dn_object, dn->dn_datablksz, 212 (longlong_t)offset, (longlong_t)length); 213 return (EIO); 214 } 215 nblks = 1; 216 } 217 dbp = kmem_zalloc(sizeof (dmu_buf_t *) * nblks, KM_SLEEP); 218 219 if (dn->dn_objset->os_dsl_dataset) 220 dp = dn->dn_objset->os_dsl_dataset->ds_dir->dd_pool; 221 if (dp && dsl_pool_sync_context(dp)) 222 start = gethrtime(); 223 zio = zio_root(dn->dn_objset->os_spa, NULL, NULL, TRUE); 224 blkid = dbuf_whichblock(dn, offset); 225 for (i = 0; i < nblks; i++) { 226 dmu_buf_impl_t *db = dbuf_hold(dn, blkid+i, tag); 227 if (db == NULL) { 228 rw_exit(&dn->dn_struct_rwlock); 229 dmu_buf_rele_array(dbp, nblks, tag); 230 zio_nowait(zio); 231 return (EIO); 232 } 233 /* initiate async i/o */ 234 if (read) { 235 rw_exit(&dn->dn_struct_rwlock); 236 (void) dbuf_read(db, zio, flags); 237 rw_enter(&dn->dn_struct_rwlock, RW_READER); 238 } 239 dbp[i] = &db->db; 240 } 241 rw_exit(&dn->dn_struct_rwlock); 242 243 /* wait for async i/o */ 244 err = zio_wait(zio); 245 /* track read overhead when we are in sync context */ 246 if (dp && dsl_pool_sync_context(dp)) 247 dp->dp_read_overhead += gethrtime() - start; 248 if (err) { 249 dmu_buf_rele_array(dbp, nblks, tag); 250 return (err); 251 } 252 253 /* wait for other io to complete */ 254 if (read) { 255 for (i = 0; i < nblks; i++) { 256 dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbp[i]; 257 mutex_enter(&db->db_mtx); 258 while (db->db_state == DB_READ || 259 db->db_state == DB_FILL) 260 cv_wait(&db->db_changed, &db->db_mtx); 261 if (db->db_state == DB_UNCACHED) 262 err = EIO; 263 mutex_exit(&db->db_mtx); 264 if (err) { 265 dmu_buf_rele_array(dbp, nblks, tag); 266 return (err); 267 } 268 } 269 } 270 271 *numbufsp = nblks; 272 *dbpp = dbp; 273 return (0); 274 } 275 276 static int 277 dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset, 278 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp) 279 { 280 dnode_t *dn; 281 int err; 282 283 err = dnode_hold(os->os, object, FTAG, &dn); 284 if (err) 285 return (err); 286 287 err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag, 288 numbufsp, dbpp); 289 290 dnode_rele(dn, FTAG); 291 292 return (err); 293 } 294 295 int 296 dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 297 uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp) 298 { 299 dnode_t *dn = ((dmu_buf_impl_t *)db)->db_dnode; 300 int err; 301 302 err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag, 303 numbufsp, dbpp); 304 305 return (err); 306 } 307 308 void 309 dmu_buf_rele_array(dmu_buf_t **dbp_fake, int numbufs, void *tag) 310 { 311 int i; 312 dmu_buf_impl_t **dbp = (dmu_buf_impl_t **)dbp_fake; 313 314 if (numbufs == 0) 315 return; 316 317 for (i = 0; i < numbufs; i++) { 318 if (dbp[i]) 319 dbuf_rele(dbp[i], tag); 320 } 321 322 kmem_free(dbp, sizeof (dmu_buf_t *) * numbufs); 323 } 324 325 void 326 dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset, uint64_t len) 327 { 328 dnode_t *dn; 329 uint64_t blkid; 330 int nblks, i, err; 331 332 if (zfs_prefetch_disable) 333 return; 334 335 if (len == 0) { /* they're interested in the bonus buffer */ 336 dn = os->os->os_meta_dnode; 337 338 if (object == 0 || object >= DN_MAX_OBJECT) 339 return; 340 341 rw_enter(&dn->dn_struct_rwlock, RW_READER); 342 blkid = dbuf_whichblock(dn, object * sizeof (dnode_phys_t)); 343 dbuf_prefetch(dn, blkid); 344 rw_exit(&dn->dn_struct_rwlock); 345 return; 346 } 347 348 /* 349 * XXX - Note, if the dnode for the requested object is not 350 * already cached, we will do a *synchronous* read in the 351 * dnode_hold() call. The same is true for any indirects. 352 */ 353 err = dnode_hold(os->os, object, FTAG, &dn); 354 if (err != 0) 355 return; 356 357 rw_enter(&dn->dn_struct_rwlock, RW_READER); 358 if (dn->dn_datablkshift) { 359 int blkshift = dn->dn_datablkshift; 360 nblks = (P2ROUNDUP(offset+len, 1<<blkshift) - 361 P2ALIGN(offset, 1<<blkshift)) >> blkshift; 362 } else { 363 nblks = (offset < dn->dn_datablksz); 364 } 365 366 if (nblks != 0) { 367 blkid = dbuf_whichblock(dn, offset); 368 for (i = 0; i < nblks; i++) 369 dbuf_prefetch(dn, blkid+i); 370 } 371 372 rw_exit(&dn->dn_struct_rwlock); 373 374 dnode_rele(dn, FTAG); 375 } 376 377 static int 378 get_next_chunk(dnode_t *dn, uint64_t *offset, uint64_t limit) 379 { 380 uint64_t len = *offset - limit; 381 uint64_t chunk_len = dn->dn_datablksz * DMU_MAX_DELETEBLKCNT; 382 uint64_t subchunk = 383 dn->dn_datablksz * EPB(dn->dn_indblkshift, SPA_BLKPTRSHIFT); 384 385 ASSERT(limit <= *offset); 386 387 if (len <= chunk_len) { 388 *offset = limit; 389 return (0); 390 } 391 392 ASSERT(ISP2(subchunk)); 393 394 while (*offset > limit) { 395 uint64_t initial_offset = P2ROUNDUP(*offset, subchunk); 396 uint64_t delta; 397 int err; 398 399 /* skip over allocated data */ 400 err = dnode_next_offset(dn, 401 DNODE_FIND_HOLE|DNODE_FIND_BACKWARDS, offset, 1, 1, 0); 402 if (err == ESRCH) 403 *offset = limit; 404 else if (err) 405 return (err); 406 407 ASSERT3U(*offset, <=, initial_offset); 408 *offset = P2ALIGN(*offset, subchunk); 409 delta = initial_offset - *offset; 410 if (delta >= chunk_len) { 411 *offset += delta - chunk_len; 412 return (0); 413 } 414 chunk_len -= delta; 415 416 /* skip over unallocated data */ 417 err = dnode_next_offset(dn, 418 DNODE_FIND_BACKWARDS, offset, 1, 1, 0); 419 if (err == ESRCH) 420 *offset = limit; 421 else if (err) 422 return (err); 423 424 if (*offset < limit) 425 *offset = limit; 426 ASSERT3U(*offset, <, initial_offset); 427 } 428 return (0); 429 } 430 431 static int 432 dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset, 433 uint64_t length, boolean_t free_dnode) 434 { 435 dmu_tx_t *tx; 436 uint64_t object_size, start, end, len; 437 boolean_t trunc = (length == DMU_OBJECT_END); 438 int align, err; 439 440 align = 1 << dn->dn_datablkshift; 441 ASSERT(align > 0); 442 object_size = align == 1 ? dn->dn_datablksz : 443 (dn->dn_maxblkid + 1) << dn->dn_datablkshift; 444 445 if (trunc || (end = offset + length) > object_size) 446 end = object_size; 447 if (end <= offset) 448 return (0); 449 length = end - offset; 450 451 while (length) { 452 start = end; 453 err = get_next_chunk(dn, &start, offset); 454 if (err) 455 return (err); 456 len = trunc ? DMU_OBJECT_END : end - start; 457 458 tx = dmu_tx_create(os); 459 dmu_tx_hold_free(tx, dn->dn_object, start, len); 460 err = dmu_tx_assign(tx, TXG_WAIT); 461 if (err) { 462 dmu_tx_abort(tx); 463 return (err); 464 } 465 466 dnode_free_range(dn, start, trunc ? -1 : len, tx); 467 468 if (start == 0 && free_dnode) { 469 ASSERT(trunc); 470 dnode_free(dn, tx); 471 } 472 473 length -= end - start; 474 475 dmu_tx_commit(tx); 476 end = start; 477 } 478 return (0); 479 } 480 481 int 482 dmu_free_long_range(objset_t *os, uint64_t object, 483 uint64_t offset, uint64_t length) 484 { 485 dnode_t *dn; 486 int err; 487 488 err = dnode_hold(os->os, object, FTAG, &dn); 489 if (err != 0) 490 return (err); 491 err = dmu_free_long_range_impl(os, dn, offset, length, FALSE); 492 dnode_rele(dn, FTAG); 493 return (err); 494 } 495 496 int 497 dmu_free_object(objset_t *os, uint64_t object) 498 { 499 dnode_t *dn; 500 dmu_tx_t *tx; 501 int err; 502 503 err = dnode_hold_impl(os->os, object, DNODE_MUST_BE_ALLOCATED, 504 FTAG, &dn); 505 if (err != 0) 506 return (err); 507 if (dn->dn_nlevels == 1) { 508 tx = dmu_tx_create(os); 509 dmu_tx_hold_bonus(tx, object); 510 dmu_tx_hold_free(tx, dn->dn_object, 0, DMU_OBJECT_END); 511 err = dmu_tx_assign(tx, TXG_WAIT); 512 if (err == 0) { 513 dnode_free_range(dn, 0, DMU_OBJECT_END, tx); 514 dnode_free(dn, tx); 515 dmu_tx_commit(tx); 516 } else { 517 dmu_tx_abort(tx); 518 } 519 } else { 520 err = dmu_free_long_range_impl(os, dn, 0, DMU_OBJECT_END, TRUE); 521 } 522 dnode_rele(dn, FTAG); 523 return (err); 524 } 525 526 int 527 dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 528 uint64_t size, dmu_tx_t *tx) 529 { 530 dnode_t *dn; 531 int err = dnode_hold(os->os, object, FTAG, &dn); 532 if (err) 533 return (err); 534 ASSERT(offset < UINT64_MAX); 535 ASSERT(size == -1ULL || size <= UINT64_MAX - offset); 536 dnode_free_range(dn, offset, size, tx); 537 dnode_rele(dn, FTAG); 538 return (0); 539 } 540 541 int 542 dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 543 void *buf) 544 { 545 dnode_t *dn; 546 dmu_buf_t **dbp; 547 int numbufs, i, err; 548 549 err = dnode_hold(os->os, object, FTAG, &dn); 550 if (err) 551 return (err); 552 553 /* 554 * Deal with odd block sizes, where there can't be data past the first 555 * block. If we ever do the tail block optimization, we will need to 556 * handle that here as well. 557 */ 558 if (dn->dn_datablkshift == 0) { 559 int newsz = offset > dn->dn_datablksz ? 0 : 560 MIN(size, dn->dn_datablksz - offset); 561 bzero((char *)buf + newsz, size - newsz); 562 size = newsz; 563 } 564 565 while (size > 0) { 566 uint64_t mylen = MIN(size, DMU_MAX_ACCESS / 2); 567 568 /* 569 * NB: we could do this block-at-a-time, but it's nice 570 * to be reading in parallel. 571 */ 572 err = dmu_buf_hold_array_by_dnode(dn, offset, mylen, 573 TRUE, FTAG, &numbufs, &dbp); 574 if (err) 575 break; 576 577 for (i = 0; i < numbufs; i++) { 578 int tocpy; 579 int bufoff; 580 dmu_buf_t *db = dbp[i]; 581 582 ASSERT(size > 0); 583 584 bufoff = offset - db->db_offset; 585 tocpy = (int)MIN(db->db_size - bufoff, size); 586 587 bcopy((char *)db->db_data + bufoff, buf, tocpy); 588 589 offset += tocpy; 590 size -= tocpy; 591 buf = (char *)buf + tocpy; 592 } 593 dmu_buf_rele_array(dbp, numbufs, FTAG); 594 } 595 dnode_rele(dn, FTAG); 596 return (err); 597 } 598 599 void 600 dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 601 const void *buf, dmu_tx_t *tx) 602 { 603 dmu_buf_t **dbp; 604 int numbufs, i; 605 606 if (size == 0) 607 return; 608 609 VERIFY(0 == dmu_buf_hold_array(os, object, offset, size, 610 FALSE, FTAG, &numbufs, &dbp)); 611 612 for (i = 0; i < numbufs; i++) { 613 int tocpy; 614 int bufoff; 615 dmu_buf_t *db = dbp[i]; 616 617 ASSERT(size > 0); 618 619 bufoff = offset - db->db_offset; 620 tocpy = (int)MIN(db->db_size - bufoff, size); 621 622 ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size); 623 624 if (tocpy == db->db_size) 625 dmu_buf_will_fill(db, tx); 626 else 627 dmu_buf_will_dirty(db, tx); 628 629 bcopy(buf, (char *)db->db_data + bufoff, tocpy); 630 631 if (tocpy == db->db_size) 632 dmu_buf_fill_done(db, tx); 633 634 offset += tocpy; 635 size -= tocpy; 636 buf = (char *)buf + tocpy; 637 } 638 dmu_buf_rele_array(dbp, numbufs, FTAG); 639 } 640 641 #ifdef _KERNEL 642 int 643 dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size) 644 { 645 dmu_buf_t **dbp; 646 int numbufs, i, err; 647 648 /* 649 * NB: we could do this block-at-a-time, but it's nice 650 * to be reading in parallel. 651 */ 652 err = dmu_buf_hold_array(os, object, uio->uio_loffset, size, TRUE, FTAG, 653 &numbufs, &dbp); 654 if (err) 655 return (err); 656 657 for (i = 0; i < numbufs; i++) { 658 int tocpy; 659 int bufoff; 660 dmu_buf_t *db = dbp[i]; 661 662 ASSERT(size > 0); 663 664 bufoff = uio->uio_loffset - db->db_offset; 665 tocpy = (int)MIN(db->db_size - bufoff, size); 666 667 err = uiomove((char *)db->db_data + bufoff, tocpy, 668 UIO_READ, uio); 669 if (err) 670 break; 671 672 size -= tocpy; 673 } 674 dmu_buf_rele_array(dbp, numbufs, FTAG); 675 676 return (err); 677 } 678 679 int 680 dmu_write_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size, 681 dmu_tx_t *tx) 682 { 683 dmu_buf_t **dbp; 684 int numbufs, i; 685 int err = 0; 686 687 if (size == 0) 688 return (0); 689 690 err = dmu_buf_hold_array(os, object, uio->uio_loffset, size, 691 FALSE, FTAG, &numbufs, &dbp); 692 if (err) 693 return (err); 694 695 for (i = 0; i < numbufs; i++) { 696 int tocpy; 697 int bufoff; 698 dmu_buf_t *db = dbp[i]; 699 700 ASSERT(size > 0); 701 702 bufoff = uio->uio_loffset - db->db_offset; 703 tocpy = (int)MIN(db->db_size - bufoff, size); 704 705 ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size); 706 707 if (tocpy == db->db_size) 708 dmu_buf_will_fill(db, tx); 709 else 710 dmu_buf_will_dirty(db, tx); 711 712 /* 713 * XXX uiomove could block forever (eg. nfs-backed 714 * pages). There needs to be a uiolockdown() function 715 * to lock the pages in memory, so that uiomove won't 716 * block. 717 */ 718 err = uiomove((char *)db->db_data + bufoff, tocpy, 719 UIO_WRITE, uio); 720 721 if (tocpy == db->db_size) 722 dmu_buf_fill_done(db, tx); 723 724 if (err) 725 break; 726 727 size -= tocpy; 728 } 729 dmu_buf_rele_array(dbp, numbufs, FTAG); 730 return (err); 731 } 732 733 int 734 dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 735 page_t *pp, dmu_tx_t *tx) 736 { 737 dmu_buf_t **dbp; 738 int numbufs, i; 739 int err; 740 741 if (size == 0) 742 return (0); 743 744 err = dmu_buf_hold_array(os, object, offset, size, 745 FALSE, FTAG, &numbufs, &dbp); 746 if (err) 747 return (err); 748 749 for (i = 0; i < numbufs; i++) { 750 int tocpy, copied, thiscpy; 751 int bufoff; 752 dmu_buf_t *db = dbp[i]; 753 caddr_t va; 754 755 ASSERT(size > 0); 756 ASSERT3U(db->db_size, >=, PAGESIZE); 757 758 bufoff = offset - db->db_offset; 759 tocpy = (int)MIN(db->db_size - bufoff, size); 760 761 ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size); 762 763 if (tocpy == db->db_size) 764 dmu_buf_will_fill(db, tx); 765 else 766 dmu_buf_will_dirty(db, tx); 767 768 for (copied = 0; copied < tocpy; copied += PAGESIZE) { 769 ASSERT3U(pp->p_offset, ==, db->db_offset + bufoff); 770 thiscpy = MIN(PAGESIZE, tocpy - copied); 771 va = zfs_map_page(pp, S_READ); 772 bcopy(va, (char *)db->db_data + bufoff, thiscpy); 773 zfs_unmap_page(pp, va); 774 pp = pp->p_next; 775 bufoff += PAGESIZE; 776 } 777 778 if (tocpy == db->db_size) 779 dmu_buf_fill_done(db, tx); 780 781 if (err) 782 break; 783 784 offset += tocpy; 785 size -= tocpy; 786 } 787 dmu_buf_rele_array(dbp, numbufs, FTAG); 788 return (err); 789 } 790 #endif 791 792 typedef struct { 793 dbuf_dirty_record_t *dr; 794 dmu_sync_cb_t *done; 795 void *arg; 796 } dmu_sync_arg_t; 797 798 /* ARGSUSED */ 799 static void 800 dmu_sync_done(zio_t *zio, arc_buf_t *buf, void *varg) 801 { 802 dmu_sync_arg_t *in = varg; 803 dbuf_dirty_record_t *dr = in->dr; 804 dmu_buf_impl_t *db = dr->dr_dbuf; 805 dmu_sync_cb_t *done = in->done; 806 807 if (!BP_IS_HOLE(zio->io_bp)) { 808 zio->io_bp->blk_fill = 1; 809 BP_SET_TYPE(zio->io_bp, db->db_dnode->dn_type); 810 BP_SET_LEVEL(zio->io_bp, 0); 811 } 812 813 mutex_enter(&db->db_mtx); 814 ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC); 815 dr->dt.dl.dr_overridden_by = *zio->io_bp; /* structure assignment */ 816 dr->dt.dl.dr_override_state = DR_OVERRIDDEN; 817 cv_broadcast(&db->db_changed); 818 mutex_exit(&db->db_mtx); 819 820 if (done) 821 done(&(db->db), in->arg); 822 823 kmem_free(in, sizeof (dmu_sync_arg_t)); 824 } 825 826 /* 827 * Intent log support: sync the block associated with db to disk. 828 * N.B. and XXX: the caller is responsible for making sure that the 829 * data isn't changing while dmu_sync() is writing it. 830 * 831 * Return values: 832 * 833 * EEXIST: this txg has already been synced, so there's nothing to to. 834 * The caller should not log the write. 835 * 836 * ENOENT: the block was dbuf_free_range()'d, so there's nothing to do. 837 * The caller should not log the write. 838 * 839 * EALREADY: this block is already in the process of being synced. 840 * The caller should track its progress (somehow). 841 * 842 * EINPROGRESS: the IO has been initiated. 843 * The caller should log this blkptr in the callback. 844 * 845 * 0: completed. Sets *bp to the blkptr just written. 846 * The caller should log this blkptr immediately. 847 */ 848 int 849 dmu_sync(zio_t *pio, dmu_buf_t *db_fake, 850 blkptr_t *bp, uint64_t txg, dmu_sync_cb_t *done, void *arg) 851 { 852 dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; 853 objset_impl_t *os = db->db_objset; 854 dsl_pool_t *dp = os->os_dsl_dataset->ds_dir->dd_pool; 855 tx_state_t *tx = &dp->dp_tx; 856 dbuf_dirty_record_t *dr; 857 dmu_sync_arg_t *in; 858 zbookmark_t zb; 859 writeprops_t wp = { 0 }; 860 zio_t *zio; 861 int zio_flags; 862 int err; 863 864 ASSERT(BP_IS_HOLE(bp)); 865 ASSERT(txg != 0); 866 867 868 dprintf("dmu_sync txg=%llu, s,o,q %llu %llu %llu\n", 869 txg, tx->tx_synced_txg, tx->tx_open_txg, tx->tx_quiesced_txg); 870 871 /* 872 * XXX - would be nice if we could do this without suspending... 873 */ 874 txg_suspend(dp); 875 876 /* 877 * If this txg already synced, there's nothing to do. 878 */ 879 if (txg <= tx->tx_synced_txg) { 880 txg_resume(dp); 881 /* 882 * If we're running ziltest, we need the blkptr regardless. 883 */ 884 if (txg > spa_freeze_txg(dp->dp_spa)) { 885 /* if db_blkptr == NULL, this was an empty write */ 886 if (db->db_blkptr) 887 *bp = *db->db_blkptr; /* structure assignment */ 888 return (0); 889 } 890 return (EEXIST); 891 } 892 893 mutex_enter(&db->db_mtx); 894 895 if (txg == tx->tx_syncing_txg) { 896 while (db->db_data_pending) { 897 /* 898 * IO is in-progress. Wait for it to finish. 899 * XXX - would be nice to be able to somehow "attach" 900 * this zio to the parent zio passed in. 901 */ 902 cv_wait(&db->db_changed, &db->db_mtx); 903 if (!db->db_data_pending && 904 db->db_blkptr && BP_IS_HOLE(db->db_blkptr)) { 905 /* 906 * IO was compressed away 907 */ 908 *bp = *db->db_blkptr; /* structure assignment */ 909 mutex_exit(&db->db_mtx); 910 txg_resume(dp); 911 return (0); 912 } 913 ASSERT(db->db_data_pending || 914 (db->db_blkptr && db->db_blkptr->blk_birth == txg)); 915 } 916 917 if (db->db_blkptr && db->db_blkptr->blk_birth == txg) { 918 /* 919 * IO is already completed. 920 */ 921 *bp = *db->db_blkptr; /* structure assignment */ 922 mutex_exit(&db->db_mtx); 923 txg_resume(dp); 924 return (0); 925 } 926 } 927 928 dr = db->db_last_dirty; 929 while (dr && dr->dr_txg > txg) 930 dr = dr->dr_next; 931 if (dr == NULL || dr->dr_txg < txg) { 932 /* 933 * This dbuf isn't dirty, must have been free_range'd. 934 * There's no need to log writes to freed blocks, so we're done. 935 */ 936 mutex_exit(&db->db_mtx); 937 txg_resume(dp); 938 return (ENOENT); 939 } 940 941 ASSERT(dr->dr_txg == txg); 942 if (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) { 943 /* 944 * We have already issued a sync write for this buffer. 945 */ 946 mutex_exit(&db->db_mtx); 947 txg_resume(dp); 948 return (EALREADY); 949 } else if (dr->dt.dl.dr_override_state == DR_OVERRIDDEN) { 950 /* 951 * This buffer has already been synced. It could not 952 * have been dirtied since, or we would have cleared the state. 953 */ 954 *bp = dr->dt.dl.dr_overridden_by; /* structure assignment */ 955 mutex_exit(&db->db_mtx); 956 txg_resume(dp); 957 return (0); 958 } 959 960 dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC; 961 in = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP); 962 in->dr = dr; 963 in->done = done; 964 in->arg = arg; 965 mutex_exit(&db->db_mtx); 966 txg_resume(dp); 967 968 zb.zb_objset = os->os_dsl_dataset->ds_object; 969 zb.zb_object = db->db.db_object; 970 zb.zb_level = db->db_level; 971 zb.zb_blkid = db->db_blkid; 972 zio_flags = ZIO_FLAG_MUSTSUCCEED; 973 if (dmu_ot[db->db_dnode->dn_type].ot_metadata || zb.zb_level != 0) 974 zio_flags |= ZIO_FLAG_METADATA; 975 wp.wp_type = db->db_dnode->dn_type; 976 wp.wp_copies = os->os_copies; 977 wp.wp_level = db->db_level; 978 wp.wp_dnchecksum = db->db_dnode->dn_checksum; 979 wp.wp_oschecksum = os->os_checksum; 980 wp.wp_dncompress = db->db_dnode->dn_compress; 981 wp.wp_oscompress = os->os_compress; 982 zio = arc_write(pio, os->os_spa, &wp, 983 DBUF_IS_L2CACHEABLE(db), txg, bp, dr->dt.dl.dr_data, NULL, 984 dmu_sync_done, in, ZIO_PRIORITY_SYNC_WRITE, zio_flags, &zb); 985 986 if (pio) { 987 zio_nowait(zio); 988 err = EINPROGRESS; 989 } else { 990 err = zio_wait(zio); 991 ASSERT(err == 0); 992 } 993 return (err); 994 } 995 996 int 997 dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs, 998 dmu_tx_t *tx) 999 { 1000 dnode_t *dn; 1001 int err; 1002 1003 err = dnode_hold(os->os, object, FTAG, &dn); 1004 if (err) 1005 return (err); 1006 err = dnode_set_blksz(dn, size, ibs, tx); 1007 dnode_rele(dn, FTAG); 1008 return (err); 1009 } 1010 1011 void 1012 dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 1013 dmu_tx_t *tx) 1014 { 1015 dnode_t *dn; 1016 1017 /* XXX assumes dnode_hold will not get an i/o error */ 1018 (void) dnode_hold(os->os, object, FTAG, &dn); 1019 ASSERT(checksum < ZIO_CHECKSUM_FUNCTIONS); 1020 dn->dn_checksum = checksum; 1021 dnode_setdirty(dn, tx); 1022 dnode_rele(dn, FTAG); 1023 } 1024 1025 void 1026 dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 1027 dmu_tx_t *tx) 1028 { 1029 dnode_t *dn; 1030 1031 /* XXX assumes dnode_hold will not get an i/o error */ 1032 (void) dnode_hold(os->os, object, FTAG, &dn); 1033 ASSERT(compress < ZIO_COMPRESS_FUNCTIONS); 1034 dn->dn_compress = compress; 1035 dnode_setdirty(dn, tx); 1036 dnode_rele(dn, FTAG); 1037 } 1038 1039 int 1040 dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off) 1041 { 1042 dnode_t *dn; 1043 int i, err; 1044 1045 err = dnode_hold(os->os, object, FTAG, &dn); 1046 if (err) 1047 return (err); 1048 /* 1049 * Sync any current changes before 1050 * we go trundling through the block pointers. 1051 */ 1052 for (i = 0; i < TXG_SIZE; i++) { 1053 if (list_link_active(&dn->dn_dirty_link[i])) 1054 break; 1055 } 1056 if (i != TXG_SIZE) { 1057 dnode_rele(dn, FTAG); 1058 txg_wait_synced(dmu_objset_pool(os), 0); 1059 err = dnode_hold(os->os, object, FTAG, &dn); 1060 if (err) 1061 return (err); 1062 } 1063 1064 err = dnode_next_offset(dn, (hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0); 1065 dnode_rele(dn, FTAG); 1066 1067 return (err); 1068 } 1069 1070 void 1071 dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi) 1072 { 1073 rw_enter(&dn->dn_struct_rwlock, RW_READER); 1074 mutex_enter(&dn->dn_mtx); 1075 1076 doi->doi_data_block_size = dn->dn_datablksz; 1077 doi->doi_metadata_block_size = dn->dn_indblkshift ? 1078 1ULL << dn->dn_indblkshift : 0; 1079 doi->doi_indirection = dn->dn_nlevels; 1080 doi->doi_checksum = dn->dn_checksum; 1081 doi->doi_compress = dn->dn_compress; 1082 doi->doi_physical_blks = (DN_USED_BYTES(dn->dn_phys) + 1083 SPA_MINBLOCKSIZE/2) >> SPA_MINBLOCKSHIFT; 1084 doi->doi_max_block_offset = dn->dn_phys->dn_maxblkid; 1085 doi->doi_type = dn->dn_type; 1086 doi->doi_bonus_size = dn->dn_bonuslen; 1087 doi->doi_bonus_type = dn->dn_bonustype; 1088 1089 mutex_exit(&dn->dn_mtx); 1090 rw_exit(&dn->dn_struct_rwlock); 1091 } 1092 1093 /* 1094 * Get information on a DMU object. 1095 * If doi is NULL, just indicates whether the object exists. 1096 */ 1097 int 1098 dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi) 1099 { 1100 dnode_t *dn; 1101 int err = dnode_hold(os->os, object, FTAG, &dn); 1102 1103 if (err) 1104 return (err); 1105 1106 if (doi != NULL) 1107 dmu_object_info_from_dnode(dn, doi); 1108 1109 dnode_rele(dn, FTAG); 1110 return (0); 1111 } 1112 1113 /* 1114 * As above, but faster; can be used when you have a held dbuf in hand. 1115 */ 1116 void 1117 dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi) 1118 { 1119 dmu_object_info_from_dnode(((dmu_buf_impl_t *)db)->db_dnode, doi); 1120 } 1121 1122 /* 1123 * Faster still when you only care about the size. 1124 * This is specifically optimized for zfs_getattr(). 1125 */ 1126 void 1127 dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, u_longlong_t *nblk512) 1128 { 1129 dnode_t *dn = ((dmu_buf_impl_t *)db)->db_dnode; 1130 1131 *blksize = dn->dn_datablksz; 1132 /* add 1 for dnode space */ 1133 *nblk512 = ((DN_USED_BYTES(dn->dn_phys) + SPA_MINBLOCKSIZE/2) >> 1134 SPA_MINBLOCKSHIFT) + 1; 1135 } 1136 1137 void 1138 byteswap_uint64_array(void *vbuf, size_t size) 1139 { 1140 uint64_t *buf = vbuf; 1141 size_t count = size >> 3; 1142 int i; 1143 1144 ASSERT((size & 7) == 0); 1145 1146 for (i = 0; i < count; i++) 1147 buf[i] = BSWAP_64(buf[i]); 1148 } 1149 1150 void 1151 byteswap_uint32_array(void *vbuf, size_t size) 1152 { 1153 uint32_t *buf = vbuf; 1154 size_t count = size >> 2; 1155 int i; 1156 1157 ASSERT((size & 3) == 0); 1158 1159 for (i = 0; i < count; i++) 1160 buf[i] = BSWAP_32(buf[i]); 1161 } 1162 1163 void 1164 byteswap_uint16_array(void *vbuf, size_t size) 1165 { 1166 uint16_t *buf = vbuf; 1167 size_t count = size >> 1; 1168 int i; 1169 1170 ASSERT((size & 1) == 0); 1171 1172 for (i = 0; i < count; i++) 1173 buf[i] = BSWAP_16(buf[i]); 1174 } 1175 1176 /* ARGSUSED */ 1177 void 1178 byteswap_uint8_array(void *vbuf, size_t size) 1179 { 1180 } 1181 1182 void 1183 dmu_init(void) 1184 { 1185 dbuf_init(); 1186 dnode_init(); 1187 arc_init(); 1188 l2arc_init(); 1189 } 1190 1191 void 1192 dmu_fini(void) 1193 { 1194 arc_fini(); 1195 dnode_fini(); 1196 dbuf_fini(); 1197 l2arc_fini(); 1198 } 1199