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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright (c) 2012, 2014 by Delphix. All rights reserved. 24 * Copyright (c) 2014 Integros [integros.com] 25 */ 26 27 /* Portions Copyright 2007 Jeremy Teo */ 28 /* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */ 29 30 #ifdef _KERNEL 31 #include <sys/types.h> 32 #include <sys/param.h> 33 #include <sys/time.h> 34 #include <sys/systm.h> 35 #include <sys/sysmacros.h> 36 #include <sys/resource.h> 37 #include <sys/mntent.h> 38 #include <sys/u8_textprep.h> 39 #include <sys/dsl_dataset.h> 40 #include <sys/vfs.h> 41 #include <sys/vnode.h> 42 #include <sys/file.h> 43 #include <sys/kmem.h> 44 #include <sys/errno.h> 45 #include <sys/unistd.h> 46 #include <sys/atomic.h> 47 #include <sys/zfs_dir.h> 48 #include <sys/zfs_acl.h> 49 #include <sys/zfs_ioctl.h> 50 #include <sys/zfs_rlock.h> 51 #include <sys/zfs_fuid.h> 52 #include <sys/dnode.h> 53 #include <sys/fs/zfs.h> 54 #endif /* _KERNEL */ 55 56 #include <sys/dmu.h> 57 #include <sys/dmu_objset.h> 58 #include <sys/dmu_tx.h> 59 #include <sys/zfs_refcount.h> 60 #include <sys/stat.h> 61 #include <sys/zap.h> 62 #include <sys/zfs_znode.h> 63 #include <sys/sa.h> 64 #include <sys/zfs_sa.h> 65 #include <sys/zfs_stat.h> 66 67 #include "zfs_prop.h" 68 #include "zfs_comutil.h" 69 70 /* Used by fstat(1). */ 71 SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD, 72 SYSCTL_NULL_INT_PTR, sizeof (znode_t), "sizeof(znode_t)"); 73 74 /* 75 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only 76 * turned on when DEBUG is also defined. 77 */ 78 #ifdef ZFS_DEBUG 79 #define ZNODE_STATS 80 #endif /* DEBUG */ 81 82 #ifdef ZNODE_STATS 83 #define ZNODE_STAT_ADD(stat) ((stat)++) 84 #else 85 #define ZNODE_STAT_ADD(stat) /* nothing */ 86 #endif /* ZNODE_STATS */ 87 88 /* 89 * Functions needed for userland (ie: libzpool) are not put under 90 * #ifdef_KERNEL; the rest of the functions have dependencies 91 * (such as VFS logic) that will not compile easily in userland. 92 */ 93 #ifdef _KERNEL 94 #if !defined(KMEM_DEBUG) && __FreeBSD_version >= 1300102 95 #define _ZFS_USE_SMR 96 static uma_zone_t znode_uma_zone; 97 #else 98 static kmem_cache_t *znode_cache = NULL; 99 #endif 100 101 extern struct vop_vector zfs_vnodeops; 102 extern struct vop_vector zfs_fifoops; 103 extern struct vop_vector zfs_shareops; 104 105 106 /* 107 * This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on 108 * z_rangelock. It will modify the offset and length of the lock to reflect 109 * znode-specific information, and convert RL_APPEND to RL_WRITER. This is 110 * called with the rangelock_t's rl_lock held, which avoids races. 111 */ 112 static void 113 zfs_rangelock_cb(zfs_locked_range_t *new, void *arg) 114 { 115 znode_t *zp = arg; 116 117 /* 118 * If in append mode, convert to writer and lock starting at the 119 * current end of file. 120 */ 121 if (new->lr_type == RL_APPEND) { 122 new->lr_offset = zp->z_size; 123 new->lr_type = RL_WRITER; 124 } 125 126 /* 127 * If we need to grow the block size then lock the whole file range. 128 */ 129 uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length); 130 if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) || 131 zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) { 132 new->lr_offset = 0; 133 new->lr_length = UINT64_MAX; 134 } 135 } 136 137 static int 138 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) 139 { 140 znode_t *zp = buf; 141 142 POINTER_INVALIDATE(&zp->z_zfsvfs); 143 144 list_link_init(&zp->z_link_node); 145 146 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); 147 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); 148 rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); 149 150 zfs_rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp); 151 152 zp->z_acl_cached = NULL; 153 zp->z_xattr_cached = NULL; 154 zp->z_xattr_parent = 0; 155 zp->z_vnode = NULL; 156 return (0); 157 } 158 159 static void 160 zfs_znode_cache_destructor(void *buf, void *arg) 161 { 162 (void) arg; 163 znode_t *zp = buf; 164 165 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 166 ASSERT3P(zp->z_vnode, ==, NULL); 167 ASSERT(!list_link_active(&zp->z_link_node)); 168 mutex_destroy(&zp->z_lock); 169 mutex_destroy(&zp->z_acl_lock); 170 rw_destroy(&zp->z_xattr_lock); 171 zfs_rangelock_fini(&zp->z_rangelock); 172 173 ASSERT3P(zp->z_acl_cached, ==, NULL); 174 ASSERT3P(zp->z_xattr_cached, ==, NULL); 175 } 176 177 178 #ifdef _ZFS_USE_SMR 179 VFS_SMR_DECLARE; 180 181 static int 182 zfs_znode_cache_constructor_smr(void *mem, int size __unused, void *private, 183 int flags) 184 { 185 return (zfs_znode_cache_constructor(mem, private, flags)); 186 } 187 188 static void 189 zfs_znode_cache_destructor_smr(void *mem, int size __unused, void *private) 190 { 191 zfs_znode_cache_destructor(mem, private); 192 } 193 194 void 195 zfs_znode_init(void) 196 { 197 /* 198 * Initialize zcache 199 */ 200 ASSERT3P(znode_uma_zone, ==, NULL); 201 znode_uma_zone = uma_zcreate("zfs_znode_cache", 202 sizeof (znode_t), zfs_znode_cache_constructor_smr, 203 zfs_znode_cache_destructor_smr, NULL, NULL, 0, 0); 204 VFS_SMR_ZONE_SET(znode_uma_zone); 205 } 206 207 static znode_t * 208 zfs_znode_alloc_kmem(int flags) 209 { 210 return (uma_zalloc_smr(znode_uma_zone, flags)); 211 } 212 213 static void 214 zfs_znode_free_kmem(znode_t *zp) 215 { 216 if (zp->z_xattr_cached) { 217 nvlist_free(zp->z_xattr_cached); 218 zp->z_xattr_cached = NULL; 219 } 220 uma_zfree_smr(znode_uma_zone, zp); 221 } 222 #else 223 void 224 zfs_znode_init(void) 225 { 226 /* 227 * Initialize zcache 228 */ 229 ASSERT3P(znode_cache, ==, NULL); 230 znode_cache = kmem_cache_create("zfs_znode_cache", 231 sizeof (znode_t), 0, zfs_znode_cache_constructor, 232 zfs_znode_cache_destructor, NULL, NULL, NULL, 0); 233 } 234 235 static znode_t * 236 zfs_znode_alloc_kmem(int flags) 237 { 238 return (kmem_cache_alloc(znode_cache, flags)); 239 } 240 241 static void 242 zfs_znode_free_kmem(znode_t *zp) 243 { 244 if (zp->z_xattr_cached) { 245 nvlist_free(zp->z_xattr_cached); 246 zp->z_xattr_cached = NULL; 247 } 248 kmem_cache_free(znode_cache, zp); 249 } 250 #endif 251 252 void 253 zfs_znode_fini(void) 254 { 255 /* 256 * Cleanup zcache 257 */ 258 #ifdef _ZFS_USE_SMR 259 if (znode_uma_zone) { 260 uma_zdestroy(znode_uma_zone); 261 znode_uma_zone = NULL; 262 } 263 #else 264 if (znode_cache) { 265 kmem_cache_destroy(znode_cache); 266 znode_cache = NULL; 267 } 268 #endif 269 } 270 271 272 static int 273 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) 274 { 275 zfs_acl_ids_t acl_ids; 276 vattr_t vattr; 277 znode_t *sharezp; 278 znode_t *zp; 279 int error; 280 281 vattr.va_mask = AT_MODE|AT_UID|AT_GID; 282 vattr.va_type = VDIR; 283 vattr.va_mode = S_IFDIR|0555; 284 vattr.va_uid = crgetuid(kcred); 285 vattr.va_gid = crgetgid(kcred); 286 287 sharezp = zfs_znode_alloc_kmem(KM_SLEEP); 288 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs)); 289 sharezp->z_unlinked = 0; 290 sharezp->z_atime_dirty = 0; 291 sharezp->z_zfsvfs = zfsvfs; 292 sharezp->z_is_sa = zfsvfs->z_use_sa; 293 294 VERIFY0(zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, 295 kcred, NULL, &acl_ids)); 296 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); 297 ASSERT3P(zp, ==, sharezp); 298 POINTER_INVALIDATE(&sharezp->z_zfsvfs); 299 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, 300 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); 301 zfsvfs->z_shares_dir = sharezp->z_id; 302 303 zfs_acl_ids_free(&acl_ids); 304 sa_handle_destroy(sharezp->z_sa_hdl); 305 zfs_znode_free_kmem(sharezp); 306 307 return (error); 308 } 309 310 /* 311 * define a couple of values we need available 312 * for both 64 and 32 bit environments. 313 */ 314 #ifndef NBITSMINOR64 315 #define NBITSMINOR64 32 316 #endif 317 #ifndef MAXMAJ64 318 #define MAXMAJ64 0xffffffffUL 319 #endif 320 #ifndef MAXMIN64 321 #define MAXMIN64 0xffffffffUL 322 #endif 323 324 /* 325 * Create special expldev for ZFS private use. 326 * Can't use standard expldev since it doesn't do 327 * what we want. The standard expldev() takes a 328 * dev32_t in LP64 and expands it to a long dev_t. 329 * We need an interface that takes a dev32_t in ILP32 330 * and expands it to a long dev_t. 331 */ 332 static uint64_t 333 zfs_expldev(dev_t dev) 334 { 335 return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev)); 336 } 337 /* 338 * Special cmpldev for ZFS private use. 339 * Can't use standard cmpldev since it takes 340 * a long dev_t and compresses it to dev32_t in 341 * LP64. We need to do a compaction of a long dev_t 342 * to a dev32_t in ILP32. 343 */ 344 dev_t 345 zfs_cmpldev(uint64_t dev) 346 { 347 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64))); 348 } 349 350 static void 351 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, 352 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) 353 { 354 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs)); 355 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id))); 356 357 ASSERT3P(zp->z_sa_hdl, ==, NULL); 358 ASSERT3P(zp->z_acl_cached, ==, NULL); 359 if (sa_hdl == NULL) { 360 VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, zp, 361 SA_HDL_SHARED, &zp->z_sa_hdl)); 362 } else { 363 zp->z_sa_hdl = sa_hdl; 364 sa_set_userp(sa_hdl, zp); 365 } 366 367 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; 368 369 /* 370 * Slap on VROOT if we are the root znode unless we are the root 371 * node of a snapshot mounted under .zfs. 372 */ 373 if (zp->z_id == zfsvfs->z_root && zfsvfs->z_parent == zfsvfs) 374 ZTOV(zp)->v_flag |= VROOT; 375 376 vn_exists(ZTOV(zp)); 377 } 378 379 void 380 zfs_znode_dmu_fini(znode_t *zp) 381 { 382 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) || 383 zp->z_unlinked || 384 ZFS_TEARDOWN_INACTIVE_WRITE_HELD(zp->z_zfsvfs)); 385 386 sa_handle_destroy(zp->z_sa_hdl); 387 zp->z_sa_hdl = NULL; 388 } 389 390 static void 391 zfs_vnode_forget(vnode_t *vp) 392 { 393 394 /* copied from insmntque_stddtr */ 395 vp->v_data = NULL; 396 vp->v_op = &dead_vnodeops; 397 vgone(vp); 398 vput(vp); 399 } 400 401 /* 402 * Construct a new znode/vnode and initialize. 403 * 404 * This does not do a call to dmu_set_user() that is 405 * up to the caller to do, in case you don't want to 406 * return the znode 407 */ 408 static znode_t * 409 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, 410 dmu_object_type_t obj_type, sa_handle_t *hdl) 411 { 412 znode_t *zp; 413 vnode_t *vp; 414 uint64_t mode; 415 uint64_t parent; 416 #ifdef notyet 417 uint64_t mtime[2], ctime[2]; 418 #endif 419 uint64_t projid = ZFS_DEFAULT_PROJID; 420 sa_bulk_attr_t bulk[9]; 421 int count = 0; 422 int error; 423 424 zp = zfs_znode_alloc_kmem(KM_SLEEP); 425 426 #ifndef _ZFS_USE_SMR 427 KASSERT((zfsvfs->z_parent->z_vfs->mnt_kern_flag & MNTK_FPLOOKUP) == 0, 428 ("%s: fast path lookup enabled without smr", __func__)); 429 #endif 430 431 #if __FreeBSD_version >= 1300076 432 KASSERT(curthread->td_vp_reserved != NULL, 433 ("zfs_znode_alloc: getnewvnode without any vnodes reserved")); 434 #else 435 KASSERT(curthread->td_vp_reserv > 0, 436 ("zfs_znode_alloc: getnewvnode without any vnodes reserved")); 437 #endif 438 error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp); 439 if (error != 0) { 440 zfs_znode_free_kmem(zp); 441 return (NULL); 442 } 443 zp->z_vnode = vp; 444 vp->v_data = zp; 445 446 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 447 448 zp->z_sa_hdl = NULL; 449 zp->z_unlinked = 0; 450 zp->z_atime_dirty = 0; 451 zp->z_mapcnt = 0; 452 zp->z_id = db->db_object; 453 zp->z_blksz = blksz; 454 zp->z_seq = 0x7A4653; 455 zp->z_sync_cnt = 0; 456 #if __FreeBSD_version >= 1300139 457 atomic_store_ptr(&zp->z_cached_symlink, NULL); 458 #endif 459 460 vp = ZTOV(zp); 461 462 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); 463 464 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); 465 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8); 466 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 467 &zp->z_size, 8); 468 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 469 &zp->z_links, 8); 470 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 471 &zp->z_pflags, 8); 472 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); 473 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, 474 &zp->z_atime, 16); 475 #ifdef notyet 476 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, 477 &mtime, 16); 478 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, 479 &ctime, 16); 480 #endif 481 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, 482 &zp->z_uid, 8); 483 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, 484 &zp->z_gid, 8); 485 486 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0 || 487 (dmu_objset_projectquota_enabled(zfsvfs->z_os) && 488 (zp->z_pflags & ZFS_PROJID) && 489 sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) { 490 if (hdl == NULL) 491 sa_handle_destroy(zp->z_sa_hdl); 492 zfs_vnode_forget(vp); 493 zp->z_vnode = NULL; 494 zfs_znode_free_kmem(zp); 495 return (NULL); 496 } 497 498 zp->z_projid = projid; 499 zp->z_mode = mode; 500 501 /* Cache the xattr parent id */ 502 if (zp->z_pflags & ZFS_XATTR) 503 zp->z_xattr_parent = parent; 504 505 vp->v_type = IFTOVT((mode_t)mode); 506 507 switch (vp->v_type) { 508 case VDIR: 509 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ 510 break; 511 case VFIFO: 512 vp->v_op = &zfs_fifoops; 513 break; 514 case VREG: 515 if (parent == zfsvfs->z_shares_dir) { 516 ASSERT0(zp->z_uid); 517 ASSERT0(zp->z_gid); 518 vp->v_op = &zfs_shareops; 519 } 520 break; 521 default: 522 break; 523 } 524 525 mutex_enter(&zfsvfs->z_znodes_lock); 526 list_insert_tail(&zfsvfs->z_all_znodes, zp); 527 zfsvfs->z_nr_znodes++; 528 zp->z_zfsvfs = zfsvfs; 529 mutex_exit(&zfsvfs->z_znodes_lock); 530 531 /* 532 * Acquire vnode lock before making it available to the world. 533 */ 534 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 535 VN_LOCK_AREC(vp); 536 if (vp->v_type != VFIFO) 537 VN_LOCK_ASHARE(vp); 538 539 return (zp); 540 } 541 542 static uint64_t empty_xattr; 543 static uint64_t pad[4]; 544 static zfs_acl_phys_t acl_phys; 545 /* 546 * Create a new DMU object to hold a zfs znode. 547 * 548 * IN: dzp - parent directory for new znode 549 * vap - file attributes for new znode 550 * tx - dmu transaction id for zap operations 551 * cr - credentials of caller 552 * flag - flags: 553 * IS_ROOT_NODE - new object will be root 554 * IS_XATTR - new object is an attribute 555 * bonuslen - length of bonus buffer 556 * setaclp - File/Dir initial ACL 557 * fuidp - Tracks fuid allocation. 558 * 559 * OUT: zpp - allocated znode 560 * 561 */ 562 void 563 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, 564 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) 565 { 566 uint64_t crtime[2], atime[2], mtime[2], ctime[2]; 567 uint64_t mode, size, links, parent, pflags; 568 uint64_t dzp_pflags = 0; 569 uint64_t rdev = 0; 570 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 571 dmu_buf_t *db; 572 timestruc_t now; 573 uint64_t gen, obj; 574 int bonuslen; 575 int dnodesize; 576 sa_handle_t *sa_hdl; 577 dmu_object_type_t obj_type; 578 sa_bulk_attr_t *sa_attrs; 579 int cnt = 0; 580 zfs_acl_locator_cb_t locate = { 0 }; 581 582 ASSERT3P(vap, !=, NULL); 583 ASSERT3U((vap->va_mask & AT_MODE), ==, AT_MODE); 584 585 if (zfsvfs->z_replay) { 586 obj = vap->va_nodeid; 587 now = vap->va_ctime; /* see zfs_replay_create() */ 588 gen = vap->va_nblocks; /* ditto */ 589 dnodesize = vap->va_fsid; /* ditto */ 590 } else { 591 obj = 0; 592 vfs_timestamp(&now); 593 gen = dmu_tx_get_txg(tx); 594 dnodesize = dmu_objset_dnodesize(zfsvfs->z_os); 595 } 596 597 if (dnodesize == 0) 598 dnodesize = DNODE_MIN_SIZE; 599 600 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; 601 bonuslen = (obj_type == DMU_OT_SA) ? 602 DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE; 603 604 /* 605 * Create a new DMU object. 606 */ 607 /* 608 * There's currently no mechanism for pre-reading the blocks that will 609 * be needed to allocate a new object, so we accept the small chance 610 * that there will be an i/o error and we will fail one of the 611 * assertions below. 612 */ 613 if (vap->va_type == VDIR) { 614 if (zfsvfs->z_replay) { 615 VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj, 616 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 617 obj_type, bonuslen, dnodesize, tx)); 618 } else { 619 obj = zap_create_norm_dnsize(zfsvfs->z_os, 620 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 621 obj_type, bonuslen, dnodesize, tx); 622 } 623 } else { 624 if (zfsvfs->z_replay) { 625 VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj, 626 DMU_OT_PLAIN_FILE_CONTENTS, 0, 627 obj_type, bonuslen, dnodesize, tx)); 628 } else { 629 obj = dmu_object_alloc_dnsize(zfsvfs->z_os, 630 DMU_OT_PLAIN_FILE_CONTENTS, 0, 631 obj_type, bonuslen, dnodesize, tx); 632 } 633 } 634 635 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 636 VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db)); 637 638 /* 639 * If this is the root, fix up the half-initialized parent pointer 640 * to reference the just-allocated physical data area. 641 */ 642 if (flag & IS_ROOT_NODE) { 643 dzp->z_id = obj; 644 } else { 645 dzp_pflags = dzp->z_pflags; 646 } 647 648 /* 649 * If parent is an xattr, so am I. 650 */ 651 if (dzp_pflags & ZFS_XATTR) { 652 flag |= IS_XATTR; 653 } 654 655 if (zfsvfs->z_use_fuids) 656 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; 657 else 658 pflags = 0; 659 660 if (vap->va_type == VDIR) { 661 size = 2; /* contents ("." and "..") */ 662 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; 663 } else { 664 size = links = 0; 665 } 666 667 if (vap->va_type == VBLK || vap->va_type == VCHR) { 668 rdev = zfs_expldev(vap->va_rdev); 669 } 670 671 parent = dzp->z_id; 672 mode = acl_ids->z_mode; 673 if (flag & IS_XATTR) 674 pflags |= ZFS_XATTR; 675 676 /* 677 * No execs denied will be determined when zfs_mode_compute() is called. 678 */ 679 pflags |= acl_ids->z_aclp->z_hints & 680 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| 681 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); 682 683 ZFS_TIME_ENCODE(&now, crtime); 684 ZFS_TIME_ENCODE(&now, ctime); 685 686 if (vap->va_mask & AT_ATIME) { 687 ZFS_TIME_ENCODE(&vap->va_atime, atime); 688 } else { 689 ZFS_TIME_ENCODE(&now, atime); 690 } 691 692 if (vap->va_mask & AT_MTIME) { 693 ZFS_TIME_ENCODE(&vap->va_mtime, mtime); 694 } else { 695 ZFS_TIME_ENCODE(&now, mtime); 696 } 697 698 /* Now add in all of the "SA" attributes */ 699 VERIFY0(sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, 700 &sa_hdl)); 701 702 /* 703 * Setup the array of attributes to be replaced/set on the new file 704 * 705 * order for DMU_OT_ZNODE is critical since it needs to be constructed 706 * in the old znode_phys_t format. Don't change this ordering 707 */ 708 sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); 709 710 if (obj_type == DMU_OT_ZNODE) { 711 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), 712 NULL, &atime, 16); 713 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), 714 NULL, &mtime, 16); 715 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), 716 NULL, &ctime, 16); 717 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), 718 NULL, &crtime, 16); 719 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), 720 NULL, &gen, 8); 721 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), 722 NULL, &mode, 8); 723 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), 724 NULL, &size, 8); 725 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), 726 NULL, &parent, 8); 727 } else { 728 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), 729 NULL, &mode, 8); 730 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), 731 NULL, &size, 8); 732 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), 733 NULL, &gen, 8); 734 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), 735 NULL, &acl_ids->z_fuid, 8); 736 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), 737 NULL, &acl_ids->z_fgid, 8); 738 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), 739 NULL, &parent, 8); 740 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), 741 NULL, &pflags, 8); 742 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), 743 NULL, &atime, 16); 744 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), 745 NULL, &mtime, 16); 746 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), 747 NULL, &ctime, 16); 748 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), 749 NULL, &crtime, 16); 750 } 751 752 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); 753 754 if (obj_type == DMU_OT_ZNODE) { 755 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, 756 &empty_xattr, 8); 757 } 758 if (obj_type == DMU_OT_ZNODE || 759 (vap->va_type == VBLK || vap->va_type == VCHR)) { 760 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs), 761 NULL, &rdev, 8); 762 763 } 764 if (obj_type == DMU_OT_ZNODE) { 765 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), 766 NULL, &pflags, 8); 767 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, 768 &acl_ids->z_fuid, 8); 769 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, 770 &acl_ids->z_fgid, 8); 771 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad, 772 sizeof (uint64_t) * 4); 773 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, 774 &acl_phys, sizeof (zfs_acl_phys_t)); 775 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { 776 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL, 777 &acl_ids->z_aclp->z_acl_count, 8); 778 locate.cb_aclp = acl_ids->z_aclp; 779 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs), 780 zfs_acl_data_locator, &locate, 781 acl_ids->z_aclp->z_acl_bytes); 782 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, 783 acl_ids->z_fuid, acl_ids->z_fgid); 784 } 785 786 VERIFY0(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx)); 787 788 if (!(flag & IS_ROOT_NODE)) { 789 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl); 790 ASSERT3P(*zpp, !=, NULL); 791 } else { 792 /* 793 * If we are creating the root node, the "parent" we 794 * passed in is the znode for the root. 795 */ 796 *zpp = dzp; 797 798 (*zpp)->z_sa_hdl = sa_hdl; 799 } 800 801 (*zpp)->z_pflags = pflags; 802 (*zpp)->z_mode = mode; 803 (*zpp)->z_dnodesize = dnodesize; 804 805 if (vap->va_mask & AT_XVATTR) 806 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx); 807 808 if (obj_type == DMU_OT_ZNODE || 809 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { 810 VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); 811 } 812 if (!(flag & IS_ROOT_NODE)) { 813 vnode_t *vp = ZTOV(*zpp); 814 vp->v_vflag |= VV_FORCEINSMQ; 815 int err = insmntque(vp, zfsvfs->z_vfs); 816 vp->v_vflag &= ~VV_FORCEINSMQ; 817 (void) err; 818 KASSERT(err == 0, ("insmntque() failed: error %d", err)); 819 } 820 kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); 821 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 822 } 823 824 /* 825 * Update in-core attributes. It is assumed the caller will be doing an 826 * sa_bulk_update to push the changes out. 827 */ 828 void 829 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) 830 { 831 xoptattr_t *xoap; 832 833 xoap = xva_getxoptattr(xvap); 834 ASSERT3P(xoap, !=, NULL); 835 836 if (zp->z_zfsvfs->z_replay == B_FALSE) { 837 ASSERT_VOP_IN_SEQC(ZTOV(zp)); 838 } 839 840 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { 841 uint64_t times[2]; 842 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); 843 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs), 844 ×, sizeof (times), tx); 845 XVA_SET_RTN(xvap, XAT_CREATETIME); 846 } 847 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { 848 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, 849 zp->z_pflags, tx); 850 XVA_SET_RTN(xvap, XAT_READONLY); 851 } 852 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { 853 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, 854 zp->z_pflags, tx); 855 XVA_SET_RTN(xvap, XAT_HIDDEN); 856 } 857 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { 858 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, 859 zp->z_pflags, tx); 860 XVA_SET_RTN(xvap, XAT_SYSTEM); 861 } 862 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { 863 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, 864 zp->z_pflags, tx); 865 XVA_SET_RTN(xvap, XAT_ARCHIVE); 866 } 867 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { 868 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, 869 zp->z_pflags, tx); 870 XVA_SET_RTN(xvap, XAT_IMMUTABLE); 871 } 872 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { 873 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, 874 zp->z_pflags, tx); 875 XVA_SET_RTN(xvap, XAT_NOUNLINK); 876 } 877 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { 878 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, 879 zp->z_pflags, tx); 880 XVA_SET_RTN(xvap, XAT_APPENDONLY); 881 } 882 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { 883 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, 884 zp->z_pflags, tx); 885 XVA_SET_RTN(xvap, XAT_NODUMP); 886 } 887 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { 888 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, 889 zp->z_pflags, tx); 890 XVA_SET_RTN(xvap, XAT_OPAQUE); 891 } 892 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { 893 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, 894 xoap->xoa_av_quarantined, zp->z_pflags, tx); 895 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); 896 } 897 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { 898 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, 899 zp->z_pflags, tx); 900 XVA_SET_RTN(xvap, XAT_AV_MODIFIED); 901 } 902 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { 903 zfs_sa_set_scanstamp(zp, xvap, tx); 904 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); 905 } 906 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { 907 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, 908 zp->z_pflags, tx); 909 XVA_SET_RTN(xvap, XAT_REPARSE); 910 } 911 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { 912 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, 913 zp->z_pflags, tx); 914 XVA_SET_RTN(xvap, XAT_OFFLINE); 915 } 916 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { 917 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, 918 zp->z_pflags, tx); 919 XVA_SET_RTN(xvap, XAT_SPARSE); 920 } 921 } 922 923 int 924 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) 925 { 926 dmu_object_info_t doi; 927 dmu_buf_t *db; 928 znode_t *zp; 929 vnode_t *vp; 930 sa_handle_t *hdl; 931 int locked; 932 int err; 933 934 getnewvnode_reserve_(); 935 again: 936 *zpp = NULL; 937 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 938 939 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); 940 if (err) { 941 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 942 getnewvnode_drop_reserve(); 943 return (err); 944 } 945 946 dmu_object_info_from_db(db, &doi); 947 if (doi.doi_bonus_type != DMU_OT_SA && 948 (doi.doi_bonus_type != DMU_OT_ZNODE || 949 (doi.doi_bonus_type == DMU_OT_ZNODE && 950 doi.doi_bonus_size < sizeof (znode_phys_t)))) { 951 sa_buf_rele(db, NULL); 952 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 953 getnewvnode_drop_reserve(); 954 return (SET_ERROR(EINVAL)); 955 } 956 957 hdl = dmu_buf_get_user(db); 958 if (hdl != NULL) { 959 zp = sa_get_userdata(hdl); 960 961 /* 962 * Since "SA" does immediate eviction we 963 * should never find a sa handle that doesn't 964 * know about the znode. 965 */ 966 ASSERT3P(zp, !=, NULL); 967 ASSERT3U(zp->z_id, ==, obj_num); 968 if (zp->z_unlinked) { 969 err = SET_ERROR(ENOENT); 970 } else { 971 vp = ZTOV(zp); 972 /* 973 * Don't let the vnode disappear after 974 * ZFS_OBJ_HOLD_EXIT. 975 */ 976 VN_HOLD(vp); 977 *zpp = zp; 978 err = 0; 979 } 980 981 sa_buf_rele(db, NULL); 982 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 983 984 if (err) { 985 getnewvnode_drop_reserve(); 986 return (err); 987 } 988 989 locked = VOP_ISLOCKED(vp); 990 VI_LOCK(vp); 991 if (VN_IS_DOOMED(vp) && locked != LK_EXCLUSIVE) { 992 /* 993 * The vnode is doomed and this thread doesn't 994 * hold the exclusive lock on it, so the vnode 995 * must be being reclaimed by another thread. 996 * Otherwise the doomed vnode is being reclaimed 997 * by this thread and zfs_zget is called from 998 * ZIL internals. 999 */ 1000 VI_UNLOCK(vp); 1001 1002 /* 1003 * XXX vrele() locks the vnode when the last reference 1004 * is dropped. Although in this case the vnode is 1005 * doomed / dead and so no inactivation is required, 1006 * the vnode lock is still acquired. That could result 1007 * in a LOR with z_teardown_lock if another thread holds 1008 * the vnode's lock and tries to take z_teardown_lock. 1009 * But that is only possible if the other thread peforms 1010 * a ZFS vnode operation on the vnode. That either 1011 * should not happen if the vnode is dead or the thread 1012 * should also have a reference to the vnode and thus 1013 * our reference is not last. 1014 */ 1015 VN_RELE(vp); 1016 goto again; 1017 } 1018 VI_UNLOCK(vp); 1019 getnewvnode_drop_reserve(); 1020 return (err); 1021 } 1022 1023 /* 1024 * Not found create new znode/vnode 1025 * but only if file exists. 1026 * 1027 * There is a small window where zfs_vget() could 1028 * find this object while a file create is still in 1029 * progress. This is checked for in zfs_znode_alloc() 1030 * 1031 * if zfs_znode_alloc() fails it will drop the hold on the 1032 * bonus buffer. 1033 */ 1034 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size, 1035 doi.doi_bonus_type, NULL); 1036 if (zp == NULL) { 1037 err = SET_ERROR(ENOENT); 1038 } else { 1039 *zpp = zp; 1040 } 1041 if (err == 0) { 1042 vnode_t *vp = ZTOV(zp); 1043 1044 err = insmntque(vp, zfsvfs->z_vfs); 1045 if (err == 0) { 1046 vp->v_hash = obj_num; 1047 VOP_UNLOCK1(vp); 1048 } else { 1049 zp->z_vnode = NULL; 1050 zfs_znode_dmu_fini(zp); 1051 zfs_znode_free(zp); 1052 *zpp = NULL; 1053 } 1054 } 1055 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1056 getnewvnode_drop_reserve(); 1057 return (err); 1058 } 1059 1060 int 1061 zfs_rezget(znode_t *zp) 1062 { 1063 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1064 dmu_object_info_t doi; 1065 dmu_buf_t *db; 1066 vnode_t *vp; 1067 uint64_t obj_num = zp->z_id; 1068 uint64_t mode, size; 1069 sa_bulk_attr_t bulk[8]; 1070 int err; 1071 int count = 0; 1072 uint64_t gen; 1073 1074 /* 1075 * Remove cached pages before reloading the znode, so that they are not 1076 * lingering after we run into any error. Ideally, we should vgone() 1077 * the vnode in case of error, but currently we cannot do that 1078 * because of the LOR between the vnode lock and z_teardown_lock. 1079 * So, instead, we have to "doom" the znode in the illumos style. 1080 * 1081 * Ignore invalid pages during the scan. This is to avoid deadlocks 1082 * between page busying and the teardown lock, as pages are busied prior 1083 * to a VOP_GETPAGES operation, which acquires the teardown read lock. 1084 * Such pages will be invalid and can safely be skipped here. 1085 */ 1086 vp = ZTOV(zp); 1087 #if __FreeBSD_version >= 1400042 1088 vn_pages_remove_valid(vp, 0, 0); 1089 #else 1090 vn_pages_remove(vp, 0, 0); 1091 #endif 1092 1093 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 1094 1095 mutex_enter(&zp->z_acl_lock); 1096 if (zp->z_acl_cached) { 1097 zfs_acl_free(zp->z_acl_cached); 1098 zp->z_acl_cached = NULL; 1099 } 1100 mutex_exit(&zp->z_acl_lock); 1101 1102 rw_enter(&zp->z_xattr_lock, RW_WRITER); 1103 if (zp->z_xattr_cached) { 1104 nvlist_free(zp->z_xattr_cached); 1105 zp->z_xattr_cached = NULL; 1106 } 1107 rw_exit(&zp->z_xattr_lock); 1108 1109 ASSERT3P(zp->z_sa_hdl, ==, NULL); 1110 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); 1111 if (err) { 1112 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1113 return (err); 1114 } 1115 1116 dmu_object_info_from_db(db, &doi); 1117 if (doi.doi_bonus_type != DMU_OT_SA && 1118 (doi.doi_bonus_type != DMU_OT_ZNODE || 1119 (doi.doi_bonus_type == DMU_OT_ZNODE && 1120 doi.doi_bonus_size < sizeof (znode_phys_t)))) { 1121 sa_buf_rele(db, NULL); 1122 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1123 return (SET_ERROR(EINVAL)); 1124 } 1125 1126 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL); 1127 size = zp->z_size; 1128 1129 /* reload cached values */ 1130 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, 1131 &gen, sizeof (gen)); 1132 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 1133 &zp->z_size, sizeof (zp->z_size)); 1134 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 1135 &zp->z_links, sizeof (zp->z_links)); 1136 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 1137 &zp->z_pflags, sizeof (zp->z_pflags)); 1138 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, 1139 &zp->z_atime, sizeof (zp->z_atime)); 1140 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, 1141 &zp->z_uid, sizeof (zp->z_uid)); 1142 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, 1143 &zp->z_gid, sizeof (zp->z_gid)); 1144 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, 1145 &mode, sizeof (mode)); 1146 1147 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { 1148 zfs_znode_dmu_fini(zp); 1149 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1150 return (SET_ERROR(EIO)); 1151 } 1152 1153 zp->z_mode = mode; 1154 1155 if (gen != zp->z_gen) { 1156 zfs_znode_dmu_fini(zp); 1157 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1158 return (SET_ERROR(EIO)); 1159 } 1160 1161 /* 1162 * It is highly improbable but still quite possible that two 1163 * objects in different datasets are created with the same 1164 * object numbers and in transaction groups with the same 1165 * numbers. znodes corresponding to those objects would 1166 * have the same z_id and z_gen, but their other attributes 1167 * may be different. 1168 * zfs recv -F may replace one of such objects with the other. 1169 * As a result file properties recorded in the replaced 1170 * object's vnode may no longer match the received object's 1171 * properties. At present the only cached property is the 1172 * files type recorded in v_type. 1173 * So, handle this case by leaving the old vnode and znode 1174 * disassociated from the actual object. A new vnode and a 1175 * znode will be created if the object is accessed 1176 * (e.g. via a look-up). The old vnode and znode will be 1177 * recycled when the last vnode reference is dropped. 1178 */ 1179 if (vp->v_type != IFTOVT((mode_t)zp->z_mode)) { 1180 zfs_znode_dmu_fini(zp); 1181 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1182 return (SET_ERROR(EIO)); 1183 } 1184 1185 /* 1186 * If the file has zero links, then it has been unlinked on the send 1187 * side and it must be in the received unlinked set. 1188 * We call zfs_znode_dmu_fini() now to prevent any accesses to the 1189 * stale data and to prevent automatically removal of the file in 1190 * zfs_zinactive(). The file will be removed either when it is removed 1191 * on the send side and the next incremental stream is received or 1192 * when the unlinked set gets processed. 1193 */ 1194 zp->z_unlinked = (zp->z_links == 0); 1195 if (zp->z_unlinked) { 1196 zfs_znode_dmu_fini(zp); 1197 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1198 return (0); 1199 } 1200 1201 zp->z_blksz = doi.doi_data_block_size; 1202 if (zp->z_size != size) 1203 vnode_pager_setsize(vp, zp->z_size); 1204 1205 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1206 1207 return (0); 1208 } 1209 1210 void 1211 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) 1212 { 1213 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1214 objset_t *os = zfsvfs->z_os; 1215 uint64_t obj = zp->z_id; 1216 uint64_t acl_obj = zfs_external_acl(zp); 1217 1218 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 1219 if (acl_obj) { 1220 VERIFY(!zp->z_is_sa); 1221 VERIFY0(dmu_object_free(os, acl_obj, tx)); 1222 } 1223 VERIFY0(dmu_object_free(os, obj, tx)); 1224 zfs_znode_dmu_fini(zp); 1225 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 1226 zfs_znode_free(zp); 1227 } 1228 1229 void 1230 zfs_zinactive(znode_t *zp) 1231 { 1232 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1233 uint64_t z_id = zp->z_id; 1234 1235 ASSERT3P(zp->z_sa_hdl, !=, NULL); 1236 1237 /* 1238 * Don't allow a zfs_zget() while were trying to release this znode 1239 */ 1240 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id); 1241 1242 /* 1243 * If this was the last reference to a file with no links, remove 1244 * the file from the file system unless the file system is mounted 1245 * read-only. That can happen, for example, if the file system was 1246 * originally read-write, the file was opened, then unlinked and 1247 * the file system was made read-only before the file was finally 1248 * closed. The file will remain in the unlinked set. 1249 */ 1250 if (zp->z_unlinked) { 1251 ASSERT(!zfsvfs->z_issnap); 1252 if ((zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) == 0) { 1253 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1254 zfs_rmnode(zp); 1255 return; 1256 } 1257 } 1258 1259 zfs_znode_dmu_fini(zp); 1260 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1261 zfs_znode_free(zp); 1262 } 1263 1264 void 1265 zfs_znode_free(znode_t *zp) 1266 { 1267 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1268 #if __FreeBSD_version >= 1300139 1269 char *symlink; 1270 #endif 1271 1272 ASSERT3P(zp->z_sa_hdl, ==, NULL); 1273 zp->z_vnode = NULL; 1274 mutex_enter(&zfsvfs->z_znodes_lock); 1275 POINTER_INVALIDATE(&zp->z_zfsvfs); 1276 list_remove(&zfsvfs->z_all_znodes, zp); 1277 zfsvfs->z_nr_znodes--; 1278 mutex_exit(&zfsvfs->z_znodes_lock); 1279 symlink = atomic_load_ptr(&zp->z_cached_symlink); 1280 if (symlink != NULL) { 1281 atomic_store_rel_ptr((uintptr_t *)&zp->z_cached_symlink, (uintptr_t)NULL); 1282 cache_symlink_free(symlink, strlen(symlink) + 1); 1283 } 1284 1285 #if __FreeBSD_version >= 1300139 1286 symlink = atomic_load_ptr(&zp->z_cached_symlink); 1287 if (symlink != NULL) { 1288 atomic_store_rel_ptr((uintptr_t *)&zp->z_cached_symlink, 1289 (uintptr_t)NULL); 1290 cache_symlink_free(symlink, strlen(symlink) + 1); 1291 } 1292 #endif 1293 1294 if (zp->z_acl_cached) { 1295 zfs_acl_free(zp->z_acl_cached); 1296 zp->z_acl_cached = NULL; 1297 } 1298 1299 zfs_znode_free_kmem(zp); 1300 } 1301 1302 void 1303 zfs_tstamp_update_setup_ext(znode_t *zp, uint_t flag, uint64_t mtime[2], 1304 uint64_t ctime[2], boolean_t have_tx) 1305 { 1306 timestruc_t now; 1307 1308 vfs_timestamp(&now); 1309 1310 if (have_tx) { /* will sa_bulk_update happen really soon? */ 1311 zp->z_atime_dirty = 0; 1312 zp->z_seq++; 1313 } else { 1314 zp->z_atime_dirty = 1; 1315 } 1316 1317 if (flag & AT_ATIME) { 1318 ZFS_TIME_ENCODE(&now, zp->z_atime); 1319 } 1320 1321 if (flag & AT_MTIME) { 1322 ZFS_TIME_ENCODE(&now, mtime); 1323 if (zp->z_zfsvfs->z_use_fuids) { 1324 zp->z_pflags |= (ZFS_ARCHIVE | 1325 ZFS_AV_MODIFIED); 1326 } 1327 } 1328 1329 if (flag & AT_CTIME) { 1330 ZFS_TIME_ENCODE(&now, ctime); 1331 if (zp->z_zfsvfs->z_use_fuids) 1332 zp->z_pflags |= ZFS_ARCHIVE; 1333 } 1334 } 1335 1336 1337 void 1338 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], 1339 uint64_t ctime[2]) 1340 { 1341 zfs_tstamp_update_setup_ext(zp, flag, mtime, ctime, B_TRUE); 1342 } 1343 /* 1344 * Grow the block size for a file. 1345 * 1346 * IN: zp - znode of file to free data in. 1347 * size - requested block size 1348 * tx - open transaction. 1349 * 1350 * NOTE: this function assumes that the znode is write locked. 1351 */ 1352 void 1353 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) 1354 { 1355 int error; 1356 u_longlong_t dummy; 1357 1358 if (size <= zp->z_blksz) 1359 return; 1360 /* 1361 * If the file size is already greater than the current blocksize, 1362 * we will not grow. If there is more than one block in a file, 1363 * the blocksize cannot change. 1364 */ 1365 if (zp->z_blksz && zp->z_size > zp->z_blksz) 1366 return; 1367 1368 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id, 1369 size, 0, tx); 1370 1371 if (error == ENOTSUP) 1372 return; 1373 ASSERT0(error); 1374 1375 /* What blocksize did we actually get? */ 1376 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); 1377 } 1378 1379 /* 1380 * Increase the file length 1381 * 1382 * IN: zp - znode of file to free data in. 1383 * end - new end-of-file 1384 * 1385 * RETURN: 0 on success, error code on failure 1386 */ 1387 static int 1388 zfs_extend(znode_t *zp, uint64_t end) 1389 { 1390 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1391 dmu_tx_t *tx; 1392 zfs_locked_range_t *lr; 1393 uint64_t newblksz; 1394 int error; 1395 1396 /* 1397 * We will change zp_size, lock the whole file. 1398 */ 1399 lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); 1400 1401 /* 1402 * Nothing to do if file already at desired length. 1403 */ 1404 if (end <= zp->z_size) { 1405 zfs_rangelock_exit(lr); 1406 return (0); 1407 } 1408 tx = dmu_tx_create(zfsvfs->z_os); 1409 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1410 zfs_sa_upgrade_txholds(tx, zp); 1411 if (end > zp->z_blksz && 1412 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { 1413 /* 1414 * We are growing the file past the current block size. 1415 */ 1416 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) { 1417 /* 1418 * File's blocksize is already larger than the 1419 * "recordsize" property. Only let it grow to 1420 * the next power of 2. 1421 */ 1422 ASSERT(!ISP2(zp->z_blksz)); 1423 newblksz = MIN(end, 1 << highbit64(zp->z_blksz)); 1424 } else { 1425 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz); 1426 } 1427 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); 1428 } else { 1429 newblksz = 0; 1430 } 1431 1432 error = dmu_tx_assign(tx, TXG_WAIT); 1433 if (error) { 1434 dmu_tx_abort(tx); 1435 zfs_rangelock_exit(lr); 1436 return (error); 1437 } 1438 1439 if (newblksz) 1440 zfs_grow_blocksize(zp, newblksz, tx); 1441 1442 zp->z_size = end; 1443 1444 VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs), 1445 &zp->z_size, sizeof (zp->z_size), tx)); 1446 1447 vnode_pager_setsize(ZTOV(zp), end); 1448 1449 zfs_rangelock_exit(lr); 1450 1451 dmu_tx_commit(tx); 1452 1453 return (0); 1454 } 1455 1456 /* 1457 * Free space in a file. 1458 * 1459 * IN: zp - znode of file to free data in. 1460 * off - start of section to free. 1461 * len - length of section to free. 1462 * 1463 * RETURN: 0 on success, error code on failure 1464 */ 1465 static int 1466 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) 1467 { 1468 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1469 zfs_locked_range_t *lr; 1470 int error; 1471 1472 /* 1473 * Lock the range being freed. 1474 */ 1475 lr = zfs_rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER); 1476 1477 /* 1478 * Nothing to do if file already at desired length. 1479 */ 1480 if (off >= zp->z_size) { 1481 zfs_rangelock_exit(lr); 1482 return (0); 1483 } 1484 1485 if (off + len > zp->z_size) 1486 len = zp->z_size - off; 1487 1488 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); 1489 1490 if (error == 0) { 1491 #if __FreeBSD_version >= 1400032 1492 vnode_pager_purge_range(ZTOV(zp), off, off + len); 1493 #else 1494 /* 1495 * Before __FreeBSD_version 1400032 we cannot free block in the 1496 * middle of a file, but only at the end of a file, so this code 1497 * path should never happen. 1498 */ 1499 vnode_pager_setsize(ZTOV(zp), off); 1500 #endif 1501 } 1502 1503 zfs_rangelock_exit(lr); 1504 1505 return (error); 1506 } 1507 1508 /* 1509 * Truncate a file 1510 * 1511 * IN: zp - znode of file to free data in. 1512 * end - new end-of-file. 1513 * 1514 * RETURN: 0 on success, error code on failure 1515 */ 1516 static int 1517 zfs_trunc(znode_t *zp, uint64_t end) 1518 { 1519 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1520 vnode_t *vp = ZTOV(zp); 1521 dmu_tx_t *tx; 1522 zfs_locked_range_t *lr; 1523 int error; 1524 sa_bulk_attr_t bulk[2]; 1525 int count = 0; 1526 1527 /* 1528 * We will change zp_size, lock the whole file. 1529 */ 1530 lr = zfs_rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); 1531 1532 /* 1533 * Nothing to do if file already at desired length. 1534 */ 1535 if (end >= zp->z_size) { 1536 zfs_rangelock_exit(lr); 1537 return (0); 1538 } 1539 1540 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, 1541 DMU_OBJECT_END); 1542 if (error) { 1543 zfs_rangelock_exit(lr); 1544 return (error); 1545 } 1546 tx = dmu_tx_create(zfsvfs->z_os); 1547 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1548 zfs_sa_upgrade_txholds(tx, zp); 1549 dmu_tx_mark_netfree(tx); 1550 error = dmu_tx_assign(tx, TXG_WAIT); 1551 if (error) { 1552 dmu_tx_abort(tx); 1553 zfs_rangelock_exit(lr); 1554 return (error); 1555 } 1556 1557 zp->z_size = end; 1558 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), 1559 NULL, &zp->z_size, sizeof (zp->z_size)); 1560 1561 if (end == 0) { 1562 zp->z_pflags &= ~ZFS_SPARSE; 1563 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 1564 NULL, &zp->z_pflags, 8); 1565 } 1566 VERIFY0(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx)); 1567 1568 dmu_tx_commit(tx); 1569 1570 /* 1571 * Clear any mapped pages in the truncated region. This has to 1572 * happen outside of the transaction to avoid the possibility of 1573 * a deadlock with someone trying to push a page that we are 1574 * about to invalidate. 1575 */ 1576 vnode_pager_setsize(vp, end); 1577 1578 zfs_rangelock_exit(lr); 1579 1580 return (0); 1581 } 1582 1583 /* 1584 * Free space in a file 1585 * 1586 * IN: zp - znode of file to free data in. 1587 * off - start of range 1588 * len - end of range (0 => EOF) 1589 * flag - current file open mode flags. 1590 * log - TRUE if this action should be logged 1591 * 1592 * RETURN: 0 on success, error code on failure 1593 */ 1594 int 1595 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) 1596 { 1597 dmu_tx_t *tx; 1598 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1599 zilog_t *zilog = zfsvfs->z_log; 1600 uint64_t mode; 1601 uint64_t mtime[2], ctime[2]; 1602 sa_bulk_attr_t bulk[3]; 1603 int count = 0; 1604 int error; 1605 1606 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode, 1607 sizeof (mode))) != 0) 1608 return (error); 1609 1610 if (off > zp->z_size) { 1611 error = zfs_extend(zp, off+len); 1612 if (error == 0 && log) 1613 goto log; 1614 else 1615 return (error); 1616 } 1617 1618 if (len == 0) { 1619 error = zfs_trunc(zp, off); 1620 } else { 1621 if ((error = zfs_free_range(zp, off, len)) == 0 && 1622 off + len > zp->z_size) 1623 error = zfs_extend(zp, off+len); 1624 } 1625 if (error || !log) 1626 return (error); 1627 log: 1628 tx = dmu_tx_create(zfsvfs->z_os); 1629 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1630 zfs_sa_upgrade_txholds(tx, zp); 1631 error = dmu_tx_assign(tx, TXG_WAIT); 1632 if (error) { 1633 dmu_tx_abort(tx); 1634 return (error); 1635 } 1636 1637 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16); 1638 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16); 1639 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 1640 NULL, &zp->z_pflags, 8); 1641 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); 1642 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 1643 ASSERT0(error); 1644 1645 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); 1646 1647 dmu_tx_commit(tx); 1648 return (0); 1649 } 1650 1651 void 1652 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) 1653 { 1654 uint64_t moid, obj, sa_obj, version; 1655 uint64_t sense = ZFS_CASE_SENSITIVE; 1656 uint64_t norm = 0; 1657 nvpair_t *elem; 1658 int error; 1659 int i; 1660 znode_t *rootzp = NULL; 1661 zfsvfs_t *zfsvfs; 1662 vattr_t vattr; 1663 znode_t *zp; 1664 zfs_acl_ids_t acl_ids; 1665 1666 /* 1667 * First attempt to create master node. 1668 */ 1669 /* 1670 * In an empty objset, there are no blocks to read and thus 1671 * there can be no i/o errors (which we assert below). 1672 */ 1673 moid = MASTER_NODE_OBJ; 1674 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, 1675 DMU_OT_NONE, 0, tx); 1676 ASSERT0(error); 1677 1678 /* 1679 * Set starting attributes. 1680 */ 1681 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); 1682 elem = NULL; 1683 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { 1684 /* For the moment we expect all zpl props to be uint64_ts */ 1685 uint64_t val; 1686 char *name; 1687 1688 ASSERT3S(nvpair_type(elem), ==, DATA_TYPE_UINT64); 1689 val = fnvpair_value_uint64(elem); 1690 name = nvpair_name(elem); 1691 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { 1692 if (val < version) 1693 version = val; 1694 } else { 1695 error = zap_update(os, moid, name, 8, 1, &val, tx); 1696 } 1697 ASSERT0(error); 1698 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) 1699 norm = val; 1700 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) 1701 sense = val; 1702 } 1703 ASSERT3U(version, !=, 0); 1704 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); 1705 1706 /* 1707 * Create zap object used for SA attribute registration 1708 */ 1709 1710 if (version >= ZPL_VERSION_SA) { 1711 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, 1712 DMU_OT_NONE, 0, tx); 1713 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); 1714 ASSERT0(error); 1715 } else { 1716 sa_obj = 0; 1717 } 1718 /* 1719 * Create a delete queue. 1720 */ 1721 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); 1722 1723 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); 1724 ASSERT0(error); 1725 1726 /* 1727 * Create root znode. Create minimal znode/vnode/zfsvfs 1728 * to allow zfs_mknode to work. 1729 */ 1730 VATTR_NULL(&vattr); 1731 vattr.va_mask = AT_MODE|AT_UID|AT_GID; 1732 vattr.va_type = VDIR; 1733 vattr.va_mode = S_IFDIR|0755; 1734 vattr.va_uid = crgetuid(cr); 1735 vattr.va_gid = crgetgid(cr); 1736 1737 zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); 1738 1739 rootzp = zfs_znode_alloc_kmem(KM_SLEEP); 1740 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs)); 1741 rootzp->z_unlinked = 0; 1742 rootzp->z_atime_dirty = 0; 1743 rootzp->z_is_sa = USE_SA(version, os); 1744 1745 zfsvfs->z_os = os; 1746 zfsvfs->z_parent = zfsvfs; 1747 zfsvfs->z_version = version; 1748 zfsvfs->z_use_fuids = USE_FUIDS(version, os); 1749 zfsvfs->z_use_sa = USE_SA(version, os); 1750 zfsvfs->z_norm = norm; 1751 1752 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, 1753 &zfsvfs->z_attr_table); 1754 1755 ASSERT0(error); 1756 1757 /* 1758 * Fold case on file systems that are always or sometimes case 1759 * insensitive. 1760 */ 1761 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) 1762 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; 1763 1764 mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); 1765 list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), 1766 offsetof(znode_t, z_link_node)); 1767 1768 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1769 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); 1770 1771 rootzp->z_zfsvfs = zfsvfs; 1772 VERIFY0(zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, 1773 cr, NULL, &acl_ids)); 1774 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); 1775 ASSERT3P(zp, ==, rootzp); 1776 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); 1777 ASSERT0(error); 1778 zfs_acl_ids_free(&acl_ids); 1779 POINTER_INVALIDATE(&rootzp->z_zfsvfs); 1780 1781 sa_handle_destroy(rootzp->z_sa_hdl); 1782 zfs_znode_free_kmem(rootzp); 1783 1784 /* 1785 * Create shares directory 1786 */ 1787 1788 error = zfs_create_share_dir(zfsvfs, tx); 1789 1790 ASSERT0(error); 1791 1792 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1793 mutex_destroy(&zfsvfs->z_hold_mtx[i]); 1794 kmem_free(zfsvfs, sizeof (zfsvfs_t)); 1795 } 1796 #endif /* _KERNEL */ 1797 1798 static int 1799 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) 1800 { 1801 uint64_t sa_obj = 0; 1802 int error; 1803 1804 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); 1805 if (error != 0 && error != ENOENT) 1806 return (error); 1807 1808 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); 1809 return (error); 1810 } 1811 1812 static int 1813 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, 1814 dmu_buf_t **db, void *tag) 1815 { 1816 dmu_object_info_t doi; 1817 int error; 1818 1819 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) 1820 return (error); 1821 1822 dmu_object_info_from_db(*db, &doi); 1823 if ((doi.doi_bonus_type != DMU_OT_SA && 1824 doi.doi_bonus_type != DMU_OT_ZNODE) || 1825 (doi.doi_bonus_type == DMU_OT_ZNODE && 1826 doi.doi_bonus_size < sizeof (znode_phys_t))) { 1827 sa_buf_rele(*db, tag); 1828 return (SET_ERROR(ENOTSUP)); 1829 } 1830 1831 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); 1832 if (error != 0) { 1833 sa_buf_rele(*db, tag); 1834 return (error); 1835 } 1836 1837 return (0); 1838 } 1839 1840 static void 1841 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) 1842 { 1843 sa_handle_destroy(hdl); 1844 sa_buf_rele(db, tag); 1845 } 1846 1847 /* 1848 * Given an object number, return its parent object number and whether 1849 * or not the object is an extended attribute directory. 1850 */ 1851 static int 1852 zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, 1853 uint64_t *pobjp, int *is_xattrdir) 1854 { 1855 uint64_t parent; 1856 uint64_t pflags; 1857 uint64_t mode; 1858 uint64_t parent_mode; 1859 sa_bulk_attr_t bulk[3]; 1860 sa_handle_t *sa_hdl; 1861 dmu_buf_t *sa_db; 1862 int count = 0; 1863 int error; 1864 1865 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, 1866 &parent, sizeof (parent)); 1867 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, 1868 &pflags, sizeof (pflags)); 1869 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, 1870 &mode, sizeof (mode)); 1871 1872 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) 1873 return (error); 1874 1875 /* 1876 * When a link is removed its parent pointer is not changed and will 1877 * be invalid. There are two cases where a link is removed but the 1878 * file stays around, when it goes to the delete queue and when there 1879 * are additional links. 1880 */ 1881 error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); 1882 if (error != 0) 1883 return (error); 1884 1885 error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); 1886 zfs_release_sa_handle(sa_hdl, sa_db, FTAG); 1887 if (error != 0) 1888 return (error); 1889 1890 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); 1891 1892 /* 1893 * Extended attributes can be applied to files, directories, etc. 1894 * Otherwise the parent must be a directory. 1895 */ 1896 if (!*is_xattrdir && !S_ISDIR(parent_mode)) 1897 return (SET_ERROR(EINVAL)); 1898 1899 *pobjp = parent; 1900 1901 return (0); 1902 } 1903 1904 /* 1905 * Given an object number, return some zpl level statistics 1906 */ 1907 static int 1908 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, 1909 zfs_stat_t *sb) 1910 { 1911 sa_bulk_attr_t bulk[4]; 1912 int count = 0; 1913 1914 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, 1915 &sb->zs_mode, sizeof (sb->zs_mode)); 1916 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, 1917 &sb->zs_gen, sizeof (sb->zs_gen)); 1918 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, 1919 &sb->zs_links, sizeof (sb->zs_links)); 1920 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, 1921 &sb->zs_ctime, sizeof (sb->zs_ctime)); 1922 1923 return (sa_bulk_lookup(hdl, bulk, count)); 1924 } 1925 1926 static int 1927 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, 1928 sa_attr_type_t *sa_table, char *buf, int len) 1929 { 1930 sa_handle_t *sa_hdl; 1931 sa_handle_t *prevhdl = NULL; 1932 dmu_buf_t *prevdb = NULL; 1933 dmu_buf_t *sa_db = NULL; 1934 char *path = buf + len - 1; 1935 int error; 1936 1937 *path = '\0'; 1938 sa_hdl = hdl; 1939 1940 uint64_t deleteq_obj; 1941 VERIFY0(zap_lookup(osp, MASTER_NODE_OBJ, 1942 ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj)); 1943 error = zap_lookup_int(osp, deleteq_obj, obj); 1944 if (error == 0) { 1945 return (ESTALE); 1946 } else if (error != ENOENT) { 1947 return (error); 1948 } 1949 error = 0; 1950 1951 for (;;) { 1952 uint64_t pobj; 1953 char component[MAXNAMELEN + 2]; 1954 size_t complen; 1955 int is_xattrdir; 1956 1957 if (prevdb) { 1958 ASSERT3P(prevhdl, !=, NULL); 1959 zfs_release_sa_handle(prevhdl, prevdb, FTAG); 1960 } 1961 1962 if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, 1963 &is_xattrdir)) != 0) 1964 break; 1965 1966 if (pobj == obj) { 1967 if (path[0] != '/') 1968 *--path = '/'; 1969 break; 1970 } 1971 1972 component[0] = '/'; 1973 if (is_xattrdir) { 1974 (void) sprintf(component + 1, "<xattrdir>"); 1975 } else { 1976 error = zap_value_search(osp, pobj, obj, 1977 ZFS_DIRENT_OBJ(-1ULL), component + 1); 1978 if (error != 0) 1979 break; 1980 } 1981 1982 complen = strlen(component); 1983 path -= complen; 1984 ASSERT3P(path, >=, buf); 1985 memcpy(path, component, complen); 1986 obj = pobj; 1987 1988 if (sa_hdl != hdl) { 1989 prevhdl = sa_hdl; 1990 prevdb = sa_db; 1991 } 1992 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); 1993 if (error != 0) { 1994 sa_hdl = prevhdl; 1995 sa_db = prevdb; 1996 break; 1997 } 1998 } 1999 2000 if (sa_hdl != NULL && sa_hdl != hdl) { 2001 ASSERT3P(sa_db, !=, NULL); 2002 zfs_release_sa_handle(sa_hdl, sa_db, FTAG); 2003 } 2004 2005 if (error == 0) 2006 (void) memmove(buf, path, buf + len - path); 2007 2008 return (error); 2009 } 2010 2011 int 2012 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) 2013 { 2014 sa_attr_type_t *sa_table; 2015 sa_handle_t *hdl; 2016 dmu_buf_t *db; 2017 int error; 2018 2019 error = zfs_sa_setup(osp, &sa_table); 2020 if (error != 0) 2021 return (error); 2022 2023 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); 2024 if (error != 0) 2025 return (error); 2026 2027 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); 2028 2029 zfs_release_sa_handle(hdl, db, FTAG); 2030 return (error); 2031 } 2032 2033 int 2034 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, 2035 char *buf, int len) 2036 { 2037 char *path = buf + len - 1; 2038 sa_attr_type_t *sa_table; 2039 sa_handle_t *hdl; 2040 dmu_buf_t *db; 2041 int error; 2042 2043 *path = '\0'; 2044 2045 error = zfs_sa_setup(osp, &sa_table); 2046 if (error != 0) 2047 return (error); 2048 2049 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); 2050 if (error != 0) 2051 return (error); 2052 2053 error = zfs_obj_to_stats_impl(hdl, sa_table, sb); 2054 if (error != 0) { 2055 zfs_release_sa_handle(hdl, db, FTAG); 2056 return (error); 2057 } 2058 2059 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); 2060 2061 zfs_release_sa_handle(hdl, db, FTAG); 2062 return (error); 2063 } 2064 2065 2066 void 2067 zfs_znode_update_vfs(znode_t *zp) 2068 { 2069 vm_object_t object; 2070 2071 if ((object = ZTOV(zp)->v_object) == NULL || 2072 zp->z_size == object->un_pager.vnp.vnp_size) 2073 return; 2074 2075 vnode_pager_setsize(ZTOV(zp), zp->z_size); 2076 } 2077 2078 2079 #ifdef _KERNEL 2080 int 2081 zfs_znode_parent_and_name(znode_t *zp, znode_t **dzpp, char *buf) 2082 { 2083 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2084 uint64_t parent; 2085 int is_xattrdir; 2086 int err; 2087 2088 /* Extended attributes should not be visible as regular files. */ 2089 if ((zp->z_pflags & ZFS_XATTR) != 0) 2090 return (SET_ERROR(EINVAL)); 2091 2092 err = zfs_obj_to_pobj(zfsvfs->z_os, zp->z_sa_hdl, zfsvfs->z_attr_table, 2093 &parent, &is_xattrdir); 2094 if (err != 0) 2095 return (err); 2096 ASSERT0(is_xattrdir); 2097 2098 /* No name as this is a root object. */ 2099 if (parent == zp->z_id) 2100 return (SET_ERROR(EINVAL)); 2101 2102 err = zap_value_search(zfsvfs->z_os, parent, zp->z_id, 2103 ZFS_DIRENT_OBJ(-1ULL), buf); 2104 if (err != 0) 2105 return (err); 2106 err = zfs_zget(zfsvfs, parent, dzpp); 2107 return (err); 2108 } 2109 #endif /* _KERNEL */ 2110