Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_dir.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
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 /* 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24 * Copyright (c) 2013, 2016 by Delphix. All rights reserved. 25 * Copyright 2017 Nexenta Systems, Inc. 26 */ 27 28 #include <sys/types.h> 29 #include <sys/param.h> 30 #include <sys/time.h> 31 #include <sys/systm.h> 32 #include <sys/sysmacros.h> 33 #include <sys/resource.h> 34 #include <sys/vfs.h> 35 #include <sys/vnode.h> 36 #include <sys/file.h> 37 #include <sys/kmem.h> 38 #include <sys/uio.h> 39 #include <sys/cmn_err.h> 40 #include <sys/errno.h> 41 #include <sys/stat.h> 42 #include <sys/unistd.h> 43 #include <sys/sunddi.h> 44 #include <sys/random.h> 45 #include <sys/policy.h> 46 #include <sys/kcondvar.h> 47 #include <sys/callb.h> 48 #include <sys/smp.h> 49 #include <sys/zfs_dir.h> 50 #include <sys/zfs_acl.h> 51 #include <sys/fs/zfs.h> 52 #include <sys/zap.h> 53 #include <sys/dmu.h> 54 #include <sys/atomic.h> 55 #include <sys/zfs_ctldir.h> 56 #include <sys/zfs_fuid.h> 57 #include <sys/sa.h> 58 #include <sys/zfs_sa.h> 59 #include <sys/dnlc.h> 60 #include <sys/extdirent.h> 61 62 /* 63 * zfs_match_find() is used by zfs_dirent_lookup() to peform zap lookups 64 * of names after deciding which is the appropriate lookup interface. 65 */ 66 static int 67 zfs_match_find(zfsvfs_t *zfsvfs, znode_t *dzp, const char *name, 68 matchtype_t mt, uint64_t *zoid) 69 { 70 int error; 71 72 if (zfsvfs->z_norm) { 73 74 /* 75 * In the non-mixed case we only expect there would ever 76 * be one match, but we need to use the normalizing lookup. 77 */ 78 error = zap_lookup_norm(zfsvfs->z_os, dzp->z_id, name, 8, 1, 79 zoid, mt, NULL, 0, NULL); 80 } else { 81 error = zap_lookup(zfsvfs->z_os, dzp->z_id, name, 8, 1, zoid); 82 } 83 *zoid = ZFS_DIRENT_OBJ(*zoid); 84 85 return (error); 86 } 87 88 /* 89 * Look up a directory entry under a locked vnode. 90 * dvp being locked gives us a guarantee that there are no concurrent 91 * modification of the directory and, thus, if a node can be found in 92 * the directory, then it must not be unlinked. 93 * 94 * Input arguments: 95 * dzp - znode for directory 96 * name - name of entry to lock 97 * flag - ZNEW: if the entry already exists, fail with EEXIST. 98 * ZEXISTS: if the entry does not exist, fail with ENOENT. 99 * ZXATTR: we want dzp's xattr directory 100 * 101 * Output arguments: 102 * zpp - pointer to the znode for the entry (NULL if there isn't one) 103 * 104 * Return value: 0 on success or errno on failure. 105 * 106 * NOTE: Always checks for, and rejects, '.' and '..'. 107 */ 108 int 109 zfs_dirent_lookup(znode_t *dzp, const char *name, znode_t **zpp, int flag) 110 { 111 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 112 matchtype_t mt = 0; 113 uint64_t zoid; 114 vnode_t *vp = NULL; 115 int error = 0; 116 117 ASSERT_VOP_LOCKED(ZTOV(dzp), __func__); 118 119 *zpp = NULL; 120 121 /* 122 * Verify that we are not trying to lock '.', '..', or '.zfs' 123 */ 124 if (name[0] == '.' && 125 (name[1] == '\0' || (name[1] == '.' && name[2] == '\0')) || 126 zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) 127 return (SET_ERROR(EEXIST)); 128 129 /* 130 * Case sensitivity and normalization preferences are set when 131 * the file system is created. These are stored in the 132 * zfsvfs->z_case and zfsvfs->z_norm fields. These choices 133 * affect how we perform zap lookups. 134 * 135 * When matching we may need to normalize & change case according to 136 * FS settings. 137 * 138 * Note that a normalized match is necessary for a case insensitive 139 * filesystem when the lookup request is not exact because normalization 140 * can fold case independent of normalizing code point sequences. 141 * 142 * See the table above zfs_dropname(). 143 */ 144 if (zfsvfs->z_norm != 0) { 145 mt = MT_NORMALIZE; 146 147 /* 148 * Determine if the match needs to honor the case specified in 149 * lookup, and if so keep track of that so that during 150 * normalization we don't fold case. 151 */ 152 if (zfsvfs->z_case == ZFS_CASE_MIXED) { 153 mt |= MT_MATCH_CASE; 154 } 155 } 156 157 /* 158 * Only look in or update the DNLC if we are looking for the 159 * name on a file system that does not require normalization 160 * or case folding. We can also look there if we happen to be 161 * on a non-normalizing, mixed sensitivity file system IF we 162 * are looking for the exact name. 163 * 164 * NB: we do not need to worry about this flag for ZFS_CASE_SENSITIVE 165 * because in that case MT_EXACT and MT_FIRST should produce exactly 166 * the same result. 167 */ 168 169 if (dzp->z_unlinked && !(flag & ZXATTR)) 170 return (ENOENT); 171 if (flag & ZXATTR) { 172 error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid, 173 sizeof (zoid)); 174 if (error == 0) 175 error = (zoid == 0 ? ENOENT : 0); 176 } else { 177 error = zfs_match_find(zfsvfs, dzp, name, mt, &zoid); 178 } 179 if (error) { 180 if (error != ENOENT || (flag & ZEXISTS)) { 181 return (error); 182 } 183 } else { 184 if (flag & ZNEW) { 185 return (SET_ERROR(EEXIST)); 186 } 187 error = zfs_zget(zfsvfs, zoid, zpp); 188 if (error) 189 return (error); 190 ASSERT(!(*zpp)->z_unlinked); 191 } 192 193 return (0); 194 } 195 196 static int 197 zfs_dd_lookup(znode_t *dzp, znode_t **zpp) 198 { 199 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 200 znode_t *zp; 201 uint64_t parent; 202 int error; 203 204 ASSERT_VOP_LOCKED(ZTOV(dzp), __func__); 205 ASSERT(RRM_READ_HELD(&zfsvfs->z_teardown_lock)); 206 207 if (dzp->z_unlinked) 208 return (ENOENT); 209 210 if ((error = sa_lookup(dzp->z_sa_hdl, 211 SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0) 212 return (error); 213 214 error = zfs_zget(zfsvfs, parent, &zp); 215 if (error == 0) 216 *zpp = zp; 217 return (error); 218 } 219 220 int 221 zfs_dirlook(znode_t *dzp, const char *name, znode_t **zpp) 222 { 223 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 224 znode_t *zp; 225 int error = 0; 226 227 ASSERT_VOP_LOCKED(ZTOV(dzp), __func__); 228 ASSERT(RRM_READ_HELD(&zfsvfs->z_teardown_lock)); 229 230 if (dzp->z_unlinked) 231 return (SET_ERROR(ENOENT)); 232 233 if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) { 234 *zpp = dzp; 235 } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) { 236 error = zfs_dd_lookup(dzp, zpp); 237 } else { 238 error = zfs_dirent_lookup(dzp, name, &zp, ZEXISTS); 239 if (error == 0) { 240 dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */ 241 *zpp = zp; 242 } 243 } 244 return (error); 245 } 246 247 /* 248 * unlinked Set (formerly known as the "delete queue") Error Handling 249 * 250 * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we 251 * don't specify the name of the entry that we will be manipulating. We 252 * also fib and say that we won't be adding any new entries to the 253 * unlinked set, even though we might (this is to lower the minimum file 254 * size that can be deleted in a full filesystem). So on the small 255 * chance that the nlink list is using a fat zap (ie. has more than 256 * 2000 entries), we *may* not pre-read a block that's needed. 257 * Therefore it is remotely possible for some of the assertions 258 * regarding the unlinked set below to fail due to i/o error. On a 259 * nondebug system, this will result in the space being leaked. 260 */ 261 void 262 zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx) 263 { 264 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 265 266 ASSERT(zp->z_unlinked); 267 ASSERT(zp->z_links == 0); 268 269 VERIFY3U(0, ==, 270 zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); 271 } 272 273 /* 274 * Clean up any znodes that had no links when we either crashed or 275 * (force) umounted the file system. 276 */ 277 void 278 zfs_unlinked_drain(zfsvfs_t *zfsvfs) 279 { 280 zap_cursor_t zc; 281 zap_attribute_t zap; 282 dmu_object_info_t doi; 283 znode_t *zp; 284 dmu_tx_t *tx; 285 int error; 286 287 /* 288 * Interate over the contents of the unlinked set. 289 */ 290 for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj); 291 zap_cursor_retrieve(&zc, &zap) == 0; 292 zap_cursor_advance(&zc)) { 293 294 /* 295 * See what kind of object we have in list 296 */ 297 298 error = dmu_object_info(zfsvfs->z_os, 299 zap.za_first_integer, &doi); 300 if (error != 0) 301 continue; 302 303 ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) || 304 (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS)); 305 /* 306 * We need to re-mark these list entries for deletion, 307 * so we pull them back into core and set zp->z_unlinked. 308 */ 309 error = zfs_zget(zfsvfs, zap.za_first_integer, &zp); 310 311 /* 312 * We may pick up znodes that are already marked for deletion. 313 * This could happen during the purge of an extended attribute 314 * directory. All we need to do is skip over them, since they 315 * are already in the system marked z_unlinked. 316 */ 317 if (error != 0) 318 continue; 319 320 vn_lock(ZTOV(zp), LK_EXCLUSIVE | LK_RETRY); 321 #if defined(__FreeBSD__) 322 /* 323 * Due to changes in zfs_rmnode we need to make sure the 324 * link count is set to zero here. 325 */ 326 if (zp->z_links != 0) { 327 tx = dmu_tx_create(zfsvfs->z_os); 328 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 329 error = dmu_tx_assign(tx, TXG_WAIT); 330 if (error != 0) { 331 dmu_tx_abort(tx); 332 vput(ZTOV(zp)); 333 continue; 334 } 335 zp->z_links = 0; 336 VERIFY0(sa_update(zp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), 337 &zp->z_links, sizeof (zp->z_links), tx)); 338 dmu_tx_commit(tx); 339 } 340 #endif 341 zp->z_unlinked = B_TRUE; 342 vput(ZTOV(zp)); 343 } 344 zap_cursor_fini(&zc); 345 } 346 347 /* 348 * Delete the entire contents of a directory. Return a count 349 * of the number of entries that could not be deleted. If we encounter 350 * an error, return a count of at least one so that the directory stays 351 * in the unlinked set. 352 * 353 * NOTE: this function assumes that the directory is inactive, 354 * so there is no need to lock its entries before deletion. 355 * Also, it assumes the directory contents is *only* regular 356 * files. 357 */ 358 static int 359 zfs_purgedir(znode_t *dzp) 360 { 361 zap_cursor_t zc; 362 zap_attribute_t zap; 363 znode_t *xzp; 364 dmu_tx_t *tx; 365 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 366 int skipped = 0; 367 int error; 368 369 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 370 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 371 zap_cursor_advance(&zc)) { 372 error = zfs_zget(zfsvfs, 373 ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp); 374 if (error) { 375 skipped += 1; 376 continue; 377 } 378 379 vn_lock(ZTOV(xzp), LK_EXCLUSIVE | LK_RETRY); 380 ASSERT((ZTOV(xzp)->v_type == VREG) || 381 (ZTOV(xzp)->v_type == VLNK)); 382 383 tx = dmu_tx_create(zfsvfs->z_os); 384 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); 385 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name); 386 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); 387 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 388 /* Is this really needed ? */ 389 zfs_sa_upgrade_txholds(tx, xzp); 390 dmu_tx_mark_netfree(tx); 391 error = dmu_tx_assign(tx, TXG_WAIT); 392 if (error) { 393 dmu_tx_abort(tx); 394 vput(ZTOV(xzp)); 395 skipped += 1; 396 continue; 397 } 398 399 error = zfs_link_destroy(dzp, zap.za_name, xzp, tx, 0, NULL); 400 if (error) 401 skipped += 1; 402 dmu_tx_commit(tx); 403 404 vput(ZTOV(xzp)); 405 } 406 zap_cursor_fini(&zc); 407 if (error != ENOENT) 408 skipped += 1; 409 return (skipped); 410 } 411 412 #if defined(__FreeBSD__) 413 extern taskq_t *zfsvfs_taskq; 414 #endif 415 416 void 417 zfs_rmnode(znode_t *zp) 418 { 419 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 420 objset_t *os = zfsvfs->z_os; 421 dmu_tx_t *tx; 422 uint64_t acl_obj; 423 uint64_t xattr_obj; 424 int error; 425 426 ASSERT(zp->z_links == 0); 427 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__); 428 429 /* 430 * If this is an attribute directory, purge its contents. 431 */ 432 if (ZTOV(zp) != NULL && ZTOV(zp)->v_type == VDIR && 433 (zp->z_pflags & ZFS_XATTR)) { 434 if (zfs_purgedir(zp) != 0) { 435 /* 436 * Not enough space to delete some xattrs. 437 * Leave it in the unlinked set. 438 */ 439 zfs_znode_dmu_fini(zp); 440 zfs_znode_free(zp); 441 return; 442 } 443 } else { 444 /* 445 * Free up all the data in the file. We don't do this for 446 * XATTR directories because we need truncate and remove to be 447 * in the same tx, like in zfs_znode_delete(). Otherwise, if 448 * we crash here we'll end up with an inconsistent truncated 449 * zap object in the delete queue. Note a truncated file is 450 * harmless since it only contains user data. 451 */ 452 error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END); 453 if (error) { 454 /* 455 * Not enough space or we were interrupted by unmount. 456 * Leave the file in the unlinked set. 457 */ 458 zfs_znode_dmu_fini(zp); 459 zfs_znode_free(zp); 460 return; 461 } 462 } 463 464 /* 465 * If the file has extended attributes, we're going to unlink 466 * the xattr dir. 467 */ 468 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), 469 &xattr_obj, sizeof (xattr_obj)); 470 if (error) 471 xattr_obj = 0; 472 473 acl_obj = zfs_external_acl(zp); 474 475 /* 476 * Set up the final transaction. 477 */ 478 tx = dmu_tx_create(os); 479 dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END); 480 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 481 if (xattr_obj) 482 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL); 483 if (acl_obj) 484 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); 485 486 zfs_sa_upgrade_txholds(tx, zp); 487 error = dmu_tx_assign(tx, TXG_WAIT); 488 if (error) { 489 /* 490 * Not enough space to delete the file. Leave it in the 491 * unlinked set, leaking it until the fs is remounted (at 492 * which point we'll call zfs_unlinked_drain() to process it). 493 */ 494 dmu_tx_abort(tx); 495 zfs_znode_dmu_fini(zp); 496 zfs_znode_free(zp); 497 return; 498 } 499 500 #if defined(__FreeBSD__) 501 /* 502 * FreeBSD's implemention of zfs_zget requires a vnode to back it. 503 * This means that we could end up calling into getnewvnode while 504 * calling zfs_rmnode as a result of a prior call to getnewvnode 505 * trying to clear vnodes out of the cache. If this repeats we can 506 * recurse enough that we overflow our stack. To avoid this, we 507 * avoid calling zfs_zget on the xattr znode and instead simply add 508 * it to the unlinked set and schedule a call to zfs_unlinked_drain. 509 */ 510 if (xattr_obj) { 511 /* Add extended attribute directory to the unlinked set. */ 512 VERIFY3U(0, ==, 513 zap_add_int(os, zfsvfs->z_unlinkedobj, xattr_obj, tx)); 514 } 515 #else 516 if (xzp) { 517 ASSERT(error == 0); 518 xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */ 519 xzp->z_links = 0; /* no more links to it */ 520 VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), 521 &xzp->z_links, sizeof (xzp->z_links), tx)); 522 zfs_unlinked_add(xzp, tx); 523 } 524 #endif 525 526 /* Remove this znode from the unlinked set */ 527 VERIFY3U(0, ==, 528 zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); 529 530 zfs_znode_delete(zp, tx); 531 532 dmu_tx_commit(tx); 533 534 #if defined(__FreeBSD__) 535 if (xattr_obj) { 536 /* 537 * We're using the FreeBSD taskqueue API here instead of 538 * the Solaris taskq API since the FreeBSD API allows for a 539 * task to be enqueued multiple times but executed once. 540 */ 541 taskqueue_enqueue(zfsvfs_taskq->tq_queue, 542 &zfsvfs->z_unlinked_drain_task); 543 } 544 #endif 545 } 546 547 static uint64_t 548 zfs_dirent(znode_t *zp, uint64_t mode) 549 { 550 uint64_t de = zp->z_id; 551 552 if (zp->z_zfsvfs->z_version >= ZPL_VERSION_DIRENT_TYPE) 553 de |= IFTODT(mode) << 60; 554 return (de); 555 } 556 557 /* 558 * Link zp into dzp. Can only fail if zp has been unlinked. 559 */ 560 int 561 zfs_link_create(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx, 562 int flag) 563 { 564 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 565 vnode_t *vp = ZTOV(zp); 566 uint64_t value; 567 int zp_is_dir = (vp->v_type == VDIR); 568 sa_bulk_attr_t bulk[5]; 569 uint64_t mtime[2], ctime[2]; 570 int count = 0; 571 int error; 572 573 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__); 574 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__); 575 #ifdef __FreeBSD__ 576 if (zp_is_dir) { 577 if (dzp->z_links >= ZFS_LINK_MAX) 578 return (SET_ERROR(EMLINK)); 579 } 580 #endif 581 if (!(flag & ZRENAMING)) { 582 if (zp->z_unlinked) { /* no new links to unlinked zp */ 583 ASSERT(!(flag & (ZNEW | ZEXISTS))); 584 return (SET_ERROR(ENOENT)); 585 } 586 #ifdef __FreeBSD__ 587 if (zp->z_links >= ZFS_LINK_MAX - zp_is_dir) { 588 return (SET_ERROR(EMLINK)); 589 } 590 #endif 591 zp->z_links++; 592 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 593 &zp->z_links, sizeof (zp->z_links)); 594 595 } else { 596 ASSERT(zp->z_unlinked == 0); 597 } 598 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, 599 &dzp->z_id, sizeof (dzp->z_id)); 600 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 601 &zp->z_pflags, sizeof (zp->z_pflags)); 602 603 if (!(flag & ZNEW)) { 604 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, 605 ctime, sizeof (ctime)); 606 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, 607 ctime, B_TRUE); 608 } 609 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 610 ASSERT0(error); 611 612 dzp->z_size++; 613 dzp->z_links += zp_is_dir; 614 count = 0; 615 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 616 &dzp->z_size, sizeof (dzp->z_size)); 617 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 618 &dzp->z_links, sizeof (dzp->z_links)); 619 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, 620 mtime, sizeof (mtime)); 621 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, 622 ctime, sizeof (ctime)); 623 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 624 &dzp->z_pflags, sizeof (dzp->z_pflags)); 625 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); 626 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); 627 ASSERT0(error); 628 629 value = zfs_dirent(zp, zp->z_mode); 630 error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, name, 631 8, 1, &value, tx); 632 VERIFY0(error); 633 634 return (0); 635 } 636 637 /* 638 * The match type in the code for this function should conform to: 639 * 640 * ------------------------------------------------------------------------ 641 * fs type | z_norm | lookup type | match type 642 * ---------|-------------|-------------|---------------------------------- 643 * CS !norm | 0 | 0 | 0 (exact) 644 * CS norm | formX | 0 | MT_NORMALIZE 645 * CI !norm | upper | !ZCIEXACT | MT_NORMALIZE 646 * CI !norm | upper | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE 647 * CI norm | upper|formX | !ZCIEXACT | MT_NORMALIZE 648 * CI norm | upper|formX | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE 649 * CM !norm | upper | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE 650 * CM !norm | upper | ZCILOOK | MT_NORMALIZE 651 * CM norm | upper|formX | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE 652 * CM norm | upper|formX | ZCILOOK | MT_NORMALIZE 653 * 654 * Abbreviations: 655 * CS = Case Sensitive, CI = Case Insensitive, CM = Case Mixed 656 * upper = case folding set by fs type on creation (U8_TEXTPREP_TOUPPER) 657 * formX = unicode normalization form set on fs creation 658 */ 659 static int 660 zfs_dropname(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx, 661 int flag) 662 { 663 int error; 664 665 if (zp->z_zfsvfs->z_norm) { 666 matchtype_t mt = MT_NORMALIZE; 667 668 if (zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) { 669 mt |= MT_MATCH_CASE; 670 } 671 672 error = zap_remove_norm(zp->z_zfsvfs->z_os, dzp->z_id, 673 name, mt, tx); 674 } else { 675 error = zap_remove(zp->z_zfsvfs->z_os, dzp->z_id, name, tx); 676 } 677 678 return (error); 679 } 680 681 /* 682 * Unlink zp from dzp, and mark zp for deletion if this was the last link. 683 * Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST). 684 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list. 685 * If it's non-NULL, we use it to indicate whether the znode needs deletion, 686 * and it's the caller's job to do it. 687 */ 688 int 689 zfs_link_destroy(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx, 690 int flag, boolean_t *unlinkedp) 691 { 692 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 693 vnode_t *vp = ZTOV(zp); 694 int zp_is_dir = (vp->v_type == VDIR); 695 boolean_t unlinked = B_FALSE; 696 sa_bulk_attr_t bulk[5]; 697 uint64_t mtime[2], ctime[2]; 698 int count = 0; 699 int error; 700 701 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__); 702 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__); 703 704 if (!(flag & ZRENAMING)) { 705 706 if (zp_is_dir && !zfs_dirempty(zp)) { 707 #ifdef illumos 708 return (SET_ERROR(EEXIST)); 709 #else 710 return (SET_ERROR(ENOTEMPTY)); 711 #endif 712 } 713 714 /* 715 * If we get here, we are going to try to remove the object. 716 * First try removing the name from the directory; if that 717 * fails, return the error. 718 */ 719 error = zfs_dropname(dzp, name, zp, tx, flag); 720 if (error != 0) { 721 return (error); 722 } 723 724 if (zp->z_links <= zp_is_dir) { 725 zfs_panic_recover("zfs: link count on vnode %p is %u, " 726 "should be at least %u", zp->z_vnode, 727 (int)zp->z_links, 728 zp_is_dir + 1); 729 zp->z_links = zp_is_dir + 1; 730 } 731 if (--zp->z_links == zp_is_dir) { 732 zp->z_unlinked = B_TRUE; 733 zp->z_links = 0; 734 unlinked = B_TRUE; 735 } else { 736 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), 737 NULL, &ctime, sizeof (ctime)); 738 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 739 NULL, &zp->z_pflags, sizeof (zp->z_pflags)); 740 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, 741 B_TRUE); 742 } 743 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), 744 NULL, &zp->z_links, sizeof (zp->z_links)); 745 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 746 count = 0; 747 ASSERT0(error); 748 } else { 749 ASSERT(zp->z_unlinked == 0); 750 error = zfs_dropname(dzp, name, zp, tx, flag); 751 if (error != 0) 752 return (error); 753 } 754 755 dzp->z_size--; /* one dirent removed */ 756 dzp->z_links -= zp_is_dir; /* ".." link from zp */ 757 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), 758 NULL, &dzp->z_links, sizeof (dzp->z_links)); 759 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), 760 NULL, &dzp->z_size, sizeof (dzp->z_size)); 761 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), 762 NULL, ctime, sizeof (ctime)); 763 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), 764 NULL, mtime, sizeof (mtime)); 765 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 766 NULL, &dzp->z_pflags, sizeof (dzp->z_pflags)); 767 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); 768 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); 769 ASSERT0(error); 770 771 if (unlinkedp != NULL) 772 *unlinkedp = unlinked; 773 else if (unlinked) 774 zfs_unlinked_add(zp, tx); 775 776 return (0); 777 } 778 779 /* 780 * Indicate whether the directory is empty. 781 */ 782 boolean_t 783 zfs_dirempty(znode_t *dzp) 784 { 785 return (dzp->z_size == 2); 786 } 787 788 int 789 zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr) 790 { 791 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 792 znode_t *xzp; 793 dmu_tx_t *tx; 794 int error; 795 zfs_acl_ids_t acl_ids; 796 boolean_t fuid_dirtied; 797 uint64_t parent; 798 799 *xvpp = NULL; 800 801 /* 802 * In FreeBSD, access checking for creating an EA is being done 803 * in zfs_setextattr(), 804 */ 805 #ifndef __FreeBSD_kernel__ 806 if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr)) 807 return (error); 808 #endif 809 810 if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL, 811 &acl_ids)) != 0) 812 return (error); 813 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { 814 zfs_acl_ids_free(&acl_ids); 815 return (SET_ERROR(EDQUOT)); 816 } 817 818 getnewvnode_reserve(1); 819 820 tx = dmu_tx_create(zfsvfs->z_os); 821 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + 822 ZFS_SA_BASE_ATTR_SIZE); 823 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); 824 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 825 fuid_dirtied = zfsvfs->z_fuid_dirty; 826 if (fuid_dirtied) 827 zfs_fuid_txhold(zfsvfs, tx); 828 error = dmu_tx_assign(tx, TXG_WAIT); 829 if (error) { 830 zfs_acl_ids_free(&acl_ids); 831 dmu_tx_abort(tx); 832 return (error); 833 } 834 zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids); 835 836 if (fuid_dirtied) 837 zfs_fuid_sync(zfsvfs, tx); 838 839 #ifdef DEBUG 840 error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), 841 &parent, sizeof (parent)); 842 ASSERT(error == 0 && parent == zp->z_id); 843 #endif 844 845 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id, 846 sizeof (xzp->z_id), tx)); 847 848 (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp, 849 xzp, "", NULL, acl_ids.z_fuidp, vap); 850 851 zfs_acl_ids_free(&acl_ids); 852 dmu_tx_commit(tx); 853 854 getnewvnode_drop_reserve(); 855 856 *xvpp = ZTOV(xzp); 857 858 return (0); 859 } 860 861 /* 862 * Return a znode for the extended attribute directory for zp. 863 * ** If the directory does not already exist, it is created ** 864 * 865 * IN: zp - znode to obtain attribute directory from 866 * cr - credentials of caller 867 * flags - flags from the VOP_LOOKUP call 868 * 869 * OUT: xzpp - pointer to extended attribute znode 870 * 871 * RETURN: 0 on success 872 * error number on failure 873 */ 874 int 875 zfs_get_xattrdir(znode_t *zp, vnode_t **xvpp, cred_t *cr, int flags) 876 { 877 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 878 znode_t *xzp; 879 vattr_t va; 880 int error; 881 top: 882 error = zfs_dirent_lookup(zp, "", &xzp, ZXATTR); 883 if (error) 884 return (error); 885 886 if (xzp != NULL) { 887 *xvpp = ZTOV(xzp); 888 return (0); 889 } 890 891 892 if (!(flags & CREATE_XATTR_DIR)) { 893 #ifdef illumos 894 return (SET_ERROR(ENOENT)); 895 #else 896 return (SET_ERROR(ENOATTR)); 897 #endif 898 } 899 900 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) { 901 return (SET_ERROR(EROFS)); 902 } 903 904 /* 905 * The ability to 'create' files in an attribute 906 * directory comes from the write_xattr permission on the base file. 907 * 908 * The ability to 'search' an attribute directory requires 909 * read_xattr permission on the base file. 910 * 911 * Once in a directory the ability to read/write attributes 912 * is controlled by the permissions on the attribute file. 913 */ 914 va.va_mask = AT_TYPE | AT_MODE | AT_UID | AT_GID; 915 va.va_type = VDIR; 916 va.va_mode = S_IFDIR | S_ISVTX | 0777; 917 zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid); 918 919 error = zfs_make_xattrdir(zp, &va, xvpp, cr); 920 921 if (error == ERESTART) { 922 /* NB: we already did dmu_tx_wait() if necessary */ 923 goto top; 924 } 925 if (error == 0) 926 VOP_UNLOCK(*xvpp, 0); 927 928 return (error); 929 } 930 931 /* 932 * Decide whether it is okay to remove within a sticky directory. 933 * 934 * In sticky directories, write access is not sufficient; 935 * you can remove entries from a directory only if: 936 * 937 * you own the directory, 938 * you own the entry, 939 * the entry is a plain file and you have write access, 940 * or you are privileged (checked in secpolicy...). 941 * 942 * The function returns 0 if remove access is granted. 943 */ 944 int 945 zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr) 946 { 947 uid_t uid; 948 uid_t downer; 949 uid_t fowner; 950 zfsvfs_t *zfsvfs = zdp->z_zfsvfs; 951 952 if (zdp->z_zfsvfs->z_replay) 953 return (0); 954 955 if ((zdp->z_mode & S_ISVTX) == 0) 956 return (0); 957 958 downer = zfs_fuid_map_id(zfsvfs, zdp->z_uid, cr, ZFS_OWNER); 959 fowner = zfs_fuid_map_id(zfsvfs, zp->z_uid, cr, ZFS_OWNER); 960 961 if ((uid = crgetuid(cr)) == downer || uid == fowner || 962 (ZTOV(zp)->v_type == VREG && 963 zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0)) 964 return (0); 965 else 966 return (secpolicy_vnode_remove(ZTOV(zp), cr)); 967 }
Cache object: 63b0ce32c9c21fcf1d79b30a90de18cd
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]