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 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright (c) 2013, 2015 by Delphix. All rights reserved. 24 */ 25 26 #include <sys/types.h> 27 #include <sys/param.h> 28 #include <sys/time.h> 29 #include <sys/systm.h> 30 #include <sys/sysmacros.h> 31 #include <sys/resource.h> 32 #include <sys/vfs.h> 33 #include <sys/vnode.h> 34 #include <sys/file.h> 35 #include <sys/kmem.h> 36 #include <sys/uio.h> 37 #include <sys/cmn_err.h> 38 #include <sys/errno.h> 39 #include <sys/stat.h> 40 #include <sys/unistd.h> 41 #include <sys/sunddi.h> 42 #include <sys/random.h> 43 #include <sys/policy.h> 44 #include <sys/kcondvar.h> 45 #include <sys/callb.h> 46 #include <sys/smp.h> 47 #include <sys/zfs_dir.h> 48 #include <sys/zfs_acl.h> 49 #include <sys/fs/zfs.h> 50 #include <sys/zap.h> 51 #include <sys/dmu.h> 52 #include <sys/atomic.h> 53 #include <sys/zfs_ctldir.h> 54 #include <sys/zfs_fuid.h> 55 #include <sys/sa.h> 56 #include <sys/zfs_sa.h> 57 #include <sys/dnlc.h> 58 #include <sys/extdirent.h> 59 60 /* 61 * zfs_match_find() is used by zfs_dirent_lookup() to peform zap lookups 62 * of names after deciding which is the appropriate lookup interface. 63 */ 64 static int 65 zfs_match_find(zfsvfs_t *zfsvfs, znode_t *dzp, const char *name, 66 boolean_t exact, uint64_t *zoid) 67 { 68 int error; 69 70 if (zfsvfs->z_norm) { 71 matchtype_t mt = exact? MT_EXACT : MT_FIRST; 72 73 /* 74 * In the non-mixed case we only expect there would ever 75 * be one match, but we need to use the normalizing lookup. 76 */ 77 error = zap_lookup_norm(zfsvfs->z_os, dzp->z_id, name, 8, 1, 78 zoid, mt, NULL, 0, NULL); 79 } else { 80 error = zap_lookup(zfsvfs->z_os, dzp->z_id, name, 8, 1, zoid); 81 } 82 *zoid = ZFS_DIRENT_OBJ(*zoid); 83 84 return (error); 85 } 86 87 /* 88 * Look up a directory entry under a locked vnode. 89 * dvp being locked gives us a guarantee that there are no concurrent 90 * modification of the directory and, thus, if a node can be found in 91 * the directory, then it must not be unlinked. 92 * 93 * Input arguments: 94 * dzp - znode for directory 95 * name - name of entry to lock 96 * flag - ZNEW: if the entry already exists, fail with EEXIST. 97 * ZEXISTS: if the entry does not exist, fail with ENOENT. 98 * ZXATTR: we want dzp's xattr directory 99 * 100 * Output arguments: 101 * zpp - pointer to the znode for the entry (NULL if there isn't one) 102 * 103 * Return value: 0 on success or errno on failure. 104 * 105 * NOTE: Always checks for, and rejects, '.' and '..'. 106 */ 107 int 108 zfs_dirent_lookup(znode_t *dzp, const char *name, znode_t **zpp, int flag) 109 { 110 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 111 boolean_t exact; 112 uint64_t zoid; 113 vnode_t *vp = NULL; 114 int error = 0; 115 116 ASSERT_VOP_LOCKED(ZTOV(dzp), __func__); 117 118 *zpp = NULL; 119 120 /* 121 * Verify that we are not trying to lock '.', '..', or '.zfs' 122 */ 123 if (name[0] == '.' && 124 (name[1] == '\0' || (name[1] == '.' && name[2] == '\0')) || 125 zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) 126 return (SET_ERROR(EEXIST)); 127 128 /* 129 * Case sensitivity and normalization preferences are set when 130 * the file system is created. These are stored in the 131 * zfsvfs->z_case and zfsvfs->z_norm fields. These choices 132 * affect how we perform zap lookups. 133 * 134 * Decide if exact matches should be requested when performing 135 * a zap lookup on file systems supporting case-insensitive 136 * access. 137 * 138 * NB: we do not need to worry about this flag for ZFS_CASE_SENSITIVE 139 * because in that case MT_EXACT and MT_FIRST should produce exactly 140 * the same result. 141 */ 142 exact = zfsvfs->z_case == ZFS_CASE_MIXED; 143 144 if (dzp->z_unlinked && !(flag & ZXATTR)) 145 return (ENOENT); 146 if (flag & ZXATTR) { 147 error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid, 148 sizeof (zoid)); 149 if (error == 0) 150 error = (zoid == 0 ? ENOENT : 0); 151 } else { 152 error = zfs_match_find(zfsvfs, dzp, name, exact, &zoid); 153 } 154 if (error) { 155 if (error != ENOENT || (flag & ZEXISTS)) { 156 return (error); 157 } 158 } else { 159 if (flag & ZNEW) { 160 return (SET_ERROR(EEXIST)); 161 } 162 error = zfs_zget(zfsvfs, zoid, zpp); 163 if (error) 164 return (error); 165 ASSERT(!(*zpp)->z_unlinked); 166 } 167 168 return (0); 169 } 170 171 static int 172 zfs_dd_lookup(znode_t *dzp, znode_t **zpp) 173 { 174 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 175 znode_t *zp; 176 uint64_t parent; 177 int error; 178 179 ASSERT_VOP_LOCKED(ZTOV(dzp), __func__); 180 ASSERT(RRM_READ_HELD(&zfsvfs->z_teardown_lock)); 181 182 if (dzp->z_unlinked) 183 return (ENOENT); 184 185 if ((error = sa_lookup(dzp->z_sa_hdl, 186 SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0) 187 return (error); 188 189 error = zfs_zget(zfsvfs, parent, &zp); 190 if (error == 0) 191 *zpp = zp; 192 return (error); 193 } 194 195 int 196 zfs_dirlook(znode_t *dzp, const char *name, znode_t **zpp) 197 { 198 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 199 znode_t *zp; 200 int error = 0; 201 202 ASSERT_VOP_LOCKED(ZTOV(dzp), __func__); 203 ASSERT(RRM_READ_HELD(&zfsvfs->z_teardown_lock)); 204 205 if (dzp->z_unlinked) 206 return (SET_ERROR(ENOENT)); 207 208 if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) { 209 *zpp = dzp; 210 } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) { 211 error = zfs_dd_lookup(dzp, zpp); 212 } else { 213 error = zfs_dirent_lookup(dzp, name, &zp, ZEXISTS); 214 if (error == 0) { 215 dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */ 216 *zpp = zp; 217 } 218 } 219 return (error); 220 } 221 222 /* 223 * unlinked Set (formerly known as the "delete queue") Error Handling 224 * 225 * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we 226 * don't specify the name of the entry that we will be manipulating. We 227 * also fib and say that we won't be adding any new entries to the 228 * unlinked set, even though we might (this is to lower the minimum file 229 * size that can be deleted in a full filesystem). So on the small 230 * chance that the nlink list is using a fat zap (ie. has more than 231 * 2000 entries), we *may* not pre-read a block that's needed. 232 * Therefore it is remotely possible for some of the assertions 233 * regarding the unlinked set below to fail due to i/o error. On a 234 * nondebug system, this will result in the space being leaked. 235 */ 236 void 237 zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx) 238 { 239 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 240 241 ASSERT(zp->z_unlinked); 242 ASSERT(zp->z_links == 0); 243 244 VERIFY3U(0, ==, 245 zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); 246 } 247 248 /* 249 * Clean up any znodes that had no links when we either crashed or 250 * (force) umounted the file system. 251 */ 252 void 253 zfs_unlinked_drain(zfsvfs_t *zfsvfs) 254 { 255 zap_cursor_t zc; 256 zap_attribute_t zap; 257 dmu_object_info_t doi; 258 znode_t *zp; 259 int error; 260 261 /* 262 * Interate over the contents of the unlinked set. 263 */ 264 for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj); 265 zap_cursor_retrieve(&zc, &zap) == 0; 266 zap_cursor_advance(&zc)) { 267 268 /* 269 * See what kind of object we have in list 270 */ 271 272 error = dmu_object_info(zfsvfs->z_os, 273 zap.za_first_integer, &doi); 274 if (error != 0) 275 continue; 276 277 ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) || 278 (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS)); 279 /* 280 * We need to re-mark these list entries for deletion, 281 * so we pull them back into core and set zp->z_unlinked. 282 */ 283 error = zfs_zget(zfsvfs, zap.za_first_integer, &zp); 284 285 /* 286 * We may pick up znodes that are already marked for deletion. 287 * This could happen during the purge of an extended attribute 288 * directory. All we need to do is skip over them, since they 289 * are already in the system marked z_unlinked. 290 */ 291 if (error != 0) 292 continue; 293 294 vn_lock(ZTOV(zp), LK_EXCLUSIVE | LK_RETRY); 295 zp->z_unlinked = B_TRUE; 296 vput(ZTOV(zp)); 297 } 298 zap_cursor_fini(&zc); 299 } 300 301 /* 302 * Delete the entire contents of a directory. Return a count 303 * of the number of entries that could not be deleted. If we encounter 304 * an error, return a count of at least one so that the directory stays 305 * in the unlinked set. 306 * 307 * NOTE: this function assumes that the directory is inactive, 308 * so there is no need to lock its entries before deletion. 309 * Also, it assumes the directory contents is *only* regular 310 * files. 311 */ 312 static int 313 zfs_purgedir(znode_t *dzp) 314 { 315 zap_cursor_t zc; 316 zap_attribute_t zap; 317 znode_t *xzp; 318 dmu_tx_t *tx; 319 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 320 int skipped = 0; 321 int error; 322 323 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); 324 (error = zap_cursor_retrieve(&zc, &zap)) == 0; 325 zap_cursor_advance(&zc)) { 326 error = zfs_zget(zfsvfs, 327 ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp); 328 if (error) { 329 skipped += 1; 330 continue; 331 } 332 333 vn_lock(ZTOV(xzp), LK_EXCLUSIVE | LK_RETRY); 334 ASSERT((ZTOV(xzp)->v_type == VREG) || 335 (ZTOV(xzp)->v_type == VLNK)); 336 337 tx = dmu_tx_create(zfsvfs->z_os); 338 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); 339 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name); 340 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); 341 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 342 /* Is this really needed ? */ 343 zfs_sa_upgrade_txholds(tx, xzp); 344 dmu_tx_mark_netfree(tx); 345 error = dmu_tx_assign(tx, TXG_WAIT); 346 if (error) { 347 dmu_tx_abort(tx); 348 vput(ZTOV(xzp)); 349 skipped += 1; 350 continue; 351 } 352 353 error = zfs_link_destroy(dzp, zap.za_name, xzp, tx, 0, NULL); 354 if (error) 355 skipped += 1; 356 dmu_tx_commit(tx); 357 358 vput(ZTOV(xzp)); 359 } 360 zap_cursor_fini(&zc); 361 if (error != ENOENT) 362 skipped += 1; 363 return (skipped); 364 } 365 366 void 367 zfs_rmnode(znode_t *zp) 368 { 369 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 370 objset_t *os = zfsvfs->z_os; 371 znode_t *xzp = NULL; 372 dmu_tx_t *tx; 373 uint64_t acl_obj; 374 uint64_t xattr_obj; 375 int error; 376 377 ASSERT(zp->z_links == 0); 378 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__); 379 380 /* 381 * If this is an attribute directory, purge its contents. 382 */ 383 if (ZTOV(zp) != NULL && ZTOV(zp)->v_type == VDIR && 384 (zp->z_pflags & ZFS_XATTR)) { 385 if (zfs_purgedir(zp) != 0) { 386 /* 387 * Not enough space to delete some xattrs. 388 * Leave it in the unlinked set. 389 */ 390 zfs_znode_dmu_fini(zp); 391 zfs_znode_free(zp); 392 return; 393 } 394 } else { 395 /* 396 * Free up all the data in the file. We don't do this for 397 * XATTR directories because we need truncate and remove to be 398 * in the same tx, like in zfs_znode_delete(). Otherwise, if 399 * we crash here we'll end up with an inconsistent truncated 400 * zap object in the delete queue. Note a truncated file is 401 * harmless since it only contains user data. 402 */ 403 error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END); 404 if (error) { 405 /* 406 * Not enough space. Leave the file in the unlinked 407 * set. 408 */ 409 zfs_znode_dmu_fini(zp); 410 zfs_znode_free(zp); 411 return; 412 } 413 } 414 415 /* 416 * If the file has extended attributes, we're going to unlink 417 * the xattr dir. 418 */ 419 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), 420 &xattr_obj, sizeof (xattr_obj)); 421 if (error == 0 && xattr_obj) { 422 error = zfs_zget(zfsvfs, xattr_obj, &xzp); 423 ASSERT3S(error, ==, 0); 424 vn_lock(ZTOV(xzp), LK_EXCLUSIVE | LK_RETRY); 425 } 426 427 acl_obj = zfs_external_acl(zp); 428 429 /* 430 * Set up the final transaction. 431 */ 432 tx = dmu_tx_create(os); 433 dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END); 434 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 435 if (xzp) { 436 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL); 437 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); 438 } 439 if (acl_obj) 440 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); 441 442 zfs_sa_upgrade_txholds(tx, zp); 443 error = dmu_tx_assign(tx, TXG_WAIT); 444 if (error) { 445 /* 446 * Not enough space to delete the file. Leave it in the 447 * unlinked set, leaking it until the fs is remounted (at 448 * which point we'll call zfs_unlinked_drain() to process it). 449 */ 450 dmu_tx_abort(tx); 451 zfs_znode_dmu_fini(zp); 452 zfs_znode_free(zp); 453 goto out; 454 } 455 456 if (xzp) { 457 ASSERT(error == 0); 458 xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */ 459 xzp->z_links = 0; /* no more links to it */ 460 VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), 461 &xzp->z_links, sizeof (xzp->z_links), tx)); 462 zfs_unlinked_add(xzp, tx); 463 } 464 465 /* Remove this znode from the unlinked set */ 466 VERIFY3U(0, ==, 467 zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); 468 469 zfs_znode_delete(zp, tx); 470 471 dmu_tx_commit(tx); 472 out: 473 if (xzp) 474 vput(ZTOV(xzp)); 475 } 476 477 static uint64_t 478 zfs_dirent(znode_t *zp, uint64_t mode) 479 { 480 uint64_t de = zp->z_id; 481 482 if (zp->z_zfsvfs->z_version >= ZPL_VERSION_DIRENT_TYPE) 483 de |= IFTODT(mode) << 60; 484 return (de); 485 } 486 487 /* 488 * Link zp into dzp. Can only fail if zp has been unlinked. 489 */ 490 int 491 zfs_link_create(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx, 492 int flag) 493 { 494 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 495 vnode_t *vp = ZTOV(zp); 496 uint64_t value; 497 int zp_is_dir = (vp->v_type == VDIR); 498 sa_bulk_attr_t bulk[5]; 499 uint64_t mtime[2], ctime[2]; 500 int count = 0; 501 int error; 502 503 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__); 504 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__); 505 #if 0 506 if (zp_is_dir) { 507 error = 0; 508 if (dzp->z_links >= LINK_MAX) 509 error = SET_ERROR(EMLINK); 510 return (error); 511 } 512 #endif 513 if (!(flag & ZRENAMING)) { 514 if (zp->z_unlinked) { /* no new links to unlinked zp */ 515 ASSERT(!(flag & (ZNEW | ZEXISTS))); 516 return (SET_ERROR(ENOENT)); 517 } 518 #if 0 519 if (zp->z_links >= LINK_MAX) { 520 return (SET_ERROR(EMLINK)); 521 } 522 #endif 523 zp->z_links++; 524 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 525 &zp->z_links, sizeof (zp->z_links)); 526 527 } else { 528 ASSERT(zp->z_unlinked == 0); 529 } 530 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, 531 &dzp->z_id, sizeof (dzp->z_id)); 532 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 533 &zp->z_pflags, sizeof (zp->z_pflags)); 534 535 if (!(flag & ZNEW)) { 536 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, 537 ctime, sizeof (ctime)); 538 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, 539 ctime, B_TRUE); 540 } 541 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 542 ASSERT0(error); 543 544 dzp->z_size++; 545 dzp->z_links += zp_is_dir; 546 count = 0; 547 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 548 &dzp->z_size, sizeof (dzp->z_size)); 549 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 550 &dzp->z_links, sizeof (dzp->z_links)); 551 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, 552 mtime, sizeof (mtime)); 553 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, 554 ctime, sizeof (ctime)); 555 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 556 &dzp->z_pflags, sizeof (dzp->z_pflags)); 557 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); 558 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); 559 ASSERT0(error); 560 561 value = zfs_dirent(zp, zp->z_mode); 562 error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, name, 563 8, 1, &value, tx); 564 VERIFY0(error); 565 566 return (0); 567 } 568 569 static int 570 zfs_dropname(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx, 571 int flag) 572 { 573 int error; 574 575 if (zp->z_zfsvfs->z_norm) { 576 if (zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) 577 error = zap_remove_norm(zp->z_zfsvfs->z_os, 578 dzp->z_id, name, MT_EXACT, tx); 579 else 580 error = zap_remove_norm(zp->z_zfsvfs->z_os, 581 dzp->z_id, name, MT_FIRST, tx); 582 } else { 583 error = zap_remove(zp->z_zfsvfs->z_os, 584 dzp->z_id, name, tx); 585 } 586 587 return (error); 588 } 589 590 /* 591 * Unlink zp from dzp, and mark zp for deletion if this was the last link. 592 * Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST). 593 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list. 594 * If it's non-NULL, we use it to indicate whether the znode needs deletion, 595 * and it's the caller's job to do it. 596 */ 597 int 598 zfs_link_destroy(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx, 599 int flag, boolean_t *unlinkedp) 600 { 601 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 602 vnode_t *vp = ZTOV(zp); 603 int zp_is_dir = (vp->v_type == VDIR); 604 boolean_t unlinked = B_FALSE; 605 sa_bulk_attr_t bulk[5]; 606 uint64_t mtime[2], ctime[2]; 607 int count = 0; 608 int error; 609 610 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__); 611 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__); 612 613 if (!(flag & ZRENAMING)) { 614 615 if (zp_is_dir && !zfs_dirempty(zp)) { 616 #ifdef illumos 617 return (SET_ERROR(EEXIST)); 618 #else 619 return (SET_ERROR(ENOTEMPTY)); 620 #endif 621 } 622 623 /* 624 * If we get here, we are going to try to remove the object. 625 * First try removing the name from the directory; if that 626 * fails, return the error. 627 */ 628 error = zfs_dropname(dzp, name, zp, tx, flag); 629 if (error != 0) { 630 return (error); 631 } 632 633 if (zp->z_links <= zp_is_dir) { 634 zfs_panic_recover("zfs: link count on vnode %p is %u, " 635 "should be at least %u", zp->z_vnode, 636 (int)zp->z_links, 637 zp_is_dir + 1); 638 zp->z_links = zp_is_dir + 1; 639 } 640 if (--zp->z_links == zp_is_dir) { 641 zp->z_unlinked = B_TRUE; 642 zp->z_links = 0; 643 unlinked = B_TRUE; 644 } else { 645 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), 646 NULL, &ctime, sizeof (ctime)); 647 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 648 NULL, &zp->z_pflags, sizeof (zp->z_pflags)); 649 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, 650 B_TRUE); 651 } 652 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), 653 NULL, &zp->z_links, sizeof (zp->z_links)); 654 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 655 count = 0; 656 ASSERT0(error); 657 } else { 658 ASSERT(zp->z_unlinked == 0); 659 error = zfs_dropname(dzp, name, zp, tx, flag); 660 if (error != 0) 661 return (error); 662 } 663 664 dzp->z_size--; /* one dirent removed */ 665 dzp->z_links -= zp_is_dir; /* ".." link from zp */ 666 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), 667 NULL, &dzp->z_links, sizeof (dzp->z_links)); 668 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), 669 NULL, &dzp->z_size, sizeof (dzp->z_size)); 670 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), 671 NULL, ctime, sizeof (ctime)); 672 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), 673 NULL, mtime, sizeof (mtime)); 674 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 675 NULL, &dzp->z_pflags, sizeof (dzp->z_pflags)); 676 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); 677 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); 678 ASSERT0(error); 679 680 if (unlinkedp != NULL) 681 *unlinkedp = unlinked; 682 else if (unlinked) 683 zfs_unlinked_add(zp, tx); 684 685 return (0); 686 } 687 688 /* 689 * Indicate whether the directory is empty. 690 */ 691 boolean_t 692 zfs_dirempty(znode_t *dzp) 693 { 694 return (dzp->z_size == 2); 695 } 696 697 int 698 zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr) 699 { 700 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 701 znode_t *xzp; 702 dmu_tx_t *tx; 703 int error; 704 zfs_acl_ids_t acl_ids; 705 boolean_t fuid_dirtied; 706 uint64_t parent; 707 708 *xvpp = NULL; 709 710 /* 711 * In FreeBSD, access checking for creating an EA is being done 712 * in zfs_setextattr(), 713 */ 714 #ifndef __FreeBSD_kernel__ 715 if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr)) 716 return (error); 717 #endif 718 719 if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL, 720 &acl_ids)) != 0) 721 return (error); 722 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { 723 zfs_acl_ids_free(&acl_ids); 724 return (SET_ERROR(EDQUOT)); 725 } 726 727 getnewvnode_reserve(1); 728 729 tx = dmu_tx_create(zfsvfs->z_os); 730 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + 731 ZFS_SA_BASE_ATTR_SIZE); 732 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); 733 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 734 fuid_dirtied = zfsvfs->z_fuid_dirty; 735 if (fuid_dirtied) 736 zfs_fuid_txhold(zfsvfs, tx); 737 error = dmu_tx_assign(tx, TXG_WAIT); 738 if (error) { 739 zfs_acl_ids_free(&acl_ids); 740 dmu_tx_abort(tx); 741 return (error); 742 } 743 zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids); 744 745 if (fuid_dirtied) 746 zfs_fuid_sync(zfsvfs, tx); 747 748 #ifdef DEBUG 749 error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), 750 &parent, sizeof (parent)); 751 ASSERT(error == 0 && parent == zp->z_id); 752 #endif 753 754 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id, 755 sizeof (xzp->z_id), tx)); 756 757 (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp, 758 xzp, "", NULL, acl_ids.z_fuidp, vap); 759 760 zfs_acl_ids_free(&acl_ids); 761 dmu_tx_commit(tx); 762 763 getnewvnode_drop_reserve(); 764 765 *xvpp = ZTOV(xzp); 766 767 return (0); 768 } 769 770 /* 771 * Return a znode for the extended attribute directory for zp. 772 * ** If the directory does not already exist, it is created ** 773 * 774 * IN: zp - znode to obtain attribute directory from 775 * cr - credentials of caller 776 * flags - flags from the VOP_LOOKUP call 777 * 778 * OUT: xzpp - pointer to extended attribute znode 779 * 780 * RETURN: 0 on success 781 * error number on failure 782 */ 783 int 784 zfs_get_xattrdir(znode_t *zp, vnode_t **xvpp, cred_t *cr, int flags) 785 { 786 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 787 znode_t *xzp; 788 vattr_t va; 789 int error; 790 top: 791 error = zfs_dirent_lookup(zp, "", &xzp, ZXATTR); 792 if (error) 793 return (error); 794 795 if (xzp != NULL) { 796 *xvpp = ZTOV(xzp); 797 return (0); 798 } 799 800 801 if (!(flags & CREATE_XATTR_DIR)) { 802 #ifdef illumos 803 return (SET_ERROR(ENOENT)); 804 #else 805 return (SET_ERROR(ENOATTR)); 806 #endif 807 } 808 809 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) { 810 return (SET_ERROR(EROFS)); 811 } 812 813 /* 814 * The ability to 'create' files in an attribute 815 * directory comes from the write_xattr permission on the base file. 816 * 817 * The ability to 'search' an attribute directory requires 818 * read_xattr permission on the base file. 819 * 820 * Once in a directory the ability to read/write attributes 821 * is controlled by the permissions on the attribute file. 822 */ 823 va.va_mask = AT_TYPE | AT_MODE | AT_UID | AT_GID; 824 va.va_type = VDIR; 825 va.va_mode = S_IFDIR | S_ISVTX | 0777; 826 zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid); 827 828 error = zfs_make_xattrdir(zp, &va, xvpp, cr); 829 830 if (error == ERESTART) { 831 /* NB: we already did dmu_tx_wait() if necessary */ 832 goto top; 833 } 834 if (error == 0) 835 VOP_UNLOCK(*xvpp, 0); 836 837 return (error); 838 } 839 840 /* 841 * Decide whether it is okay to remove within a sticky directory. 842 * 843 * In sticky directories, write access is not sufficient; 844 * you can remove entries from a directory only if: 845 * 846 * you own the directory, 847 * you own the entry, 848 * the entry is a plain file and you have write access, 849 * or you are privileged (checked in secpolicy...). 850 * 851 * The function returns 0 if remove access is granted. 852 */ 853 int 854 zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr) 855 { 856 uid_t uid; 857 uid_t downer; 858 uid_t fowner; 859 zfsvfs_t *zfsvfs = zdp->z_zfsvfs; 860 861 if (zdp->z_zfsvfs->z_replay) 862 return (0); 863 864 if ((zdp->z_mode & S_ISVTX) == 0) 865 return (0); 866 867 downer = zfs_fuid_map_id(zfsvfs, zdp->z_uid, cr, ZFS_OWNER); 868 fowner = zfs_fuid_map_id(zfsvfs, zp->z_uid, cr, ZFS_OWNER); 869 870 if ((uid = crgetuid(cr)) == downer || uid == fowner || 871 (ZTOV(zp)->v_type == VREG && 872 zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0)) 873 return (0); 874 else 875 return (secpolicy_vnode_remove(ZTOV(zp), cr)); 876 }
Cache object: d87c5026becea5e29026a50cda510c28
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]