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 int error;
285
286 /*
287 * Interate over the contents of the unlinked set.
288 */
289 for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj);
290 zap_cursor_retrieve(&zc, &zap) == 0;
291 zap_cursor_advance(&zc)) {
292
293 /*
294 * See what kind of object we have in list
295 */
296
297 error = dmu_object_info(zfsvfs->z_os,
298 zap.za_first_integer, &doi);
299 if (error != 0)
300 continue;
301
302 ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) ||
303 (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS));
304 /*
305 * We need to re-mark these list entries for deletion,
306 * so we pull them back into core and set zp->z_unlinked.
307 */
308 error = zfs_zget(zfsvfs, zap.za_first_integer, &zp);
309
310 /*
311 * We may pick up znodes that are already marked for deletion.
312 * This could happen during the purge of an extended attribute
313 * directory. All we need to do is skip over them, since they
314 * are already in the system marked z_unlinked.
315 */
316 if (error != 0)
317 continue;
318
319 vn_lock(ZTOV(zp), LK_EXCLUSIVE | LK_RETRY);
320 zp->z_unlinked = B_TRUE;
321 vput(ZTOV(zp));
322 }
323 zap_cursor_fini(&zc);
324 }
325
326 /*
327 * Delete the entire contents of a directory. Return a count
328 * of the number of entries that could not be deleted. If we encounter
329 * an error, return a count of at least one so that the directory stays
330 * in the unlinked set.
331 *
332 * NOTE: this function assumes that the directory is inactive,
333 * so there is no need to lock its entries before deletion.
334 * Also, it assumes the directory contents is *only* regular
335 * files.
336 */
337 static int
338 zfs_purgedir(znode_t *dzp)
339 {
340 zap_cursor_t zc;
341 zap_attribute_t zap;
342 znode_t *xzp;
343 dmu_tx_t *tx;
344 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
345 int skipped = 0;
346 int error;
347
348 for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id);
349 (error = zap_cursor_retrieve(&zc, &zap)) == 0;
350 zap_cursor_advance(&zc)) {
351 error = zfs_zget(zfsvfs,
352 ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp);
353 if (error) {
354 skipped += 1;
355 continue;
356 }
357
358 vn_lock(ZTOV(xzp), LK_EXCLUSIVE | LK_RETRY);
359 ASSERT((ZTOV(xzp)->v_type == VREG) ||
360 (ZTOV(xzp)->v_type == VLNK));
361
362 tx = dmu_tx_create(zfsvfs->z_os);
363 dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE);
364 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name);
365 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
366 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
367 /* Is this really needed ? */
368 zfs_sa_upgrade_txholds(tx, xzp);
369 dmu_tx_mark_netfree(tx);
370 error = dmu_tx_assign(tx, TXG_WAIT);
371 if (error) {
372 dmu_tx_abort(tx);
373 vput(ZTOV(xzp));
374 skipped += 1;
375 continue;
376 }
377
378 error = zfs_link_destroy(dzp, zap.za_name, xzp, tx, 0, NULL);
379 if (error)
380 skipped += 1;
381 dmu_tx_commit(tx);
382
383 vput(ZTOV(xzp));
384 }
385 zap_cursor_fini(&zc);
386 if (error != ENOENT)
387 skipped += 1;
388 return (skipped);
389 }
390
391 void
392 zfs_rmnode(znode_t *zp)
393 {
394 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
395 objset_t *os = zfsvfs->z_os;
396 znode_t *xzp = NULL;
397 dmu_tx_t *tx;
398 uint64_t acl_obj;
399 uint64_t xattr_obj;
400 int error;
401
402 ASSERT(zp->z_links == 0);
403 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
404
405 /*
406 * If this is an attribute directory, purge its contents.
407 */
408 if (ZTOV(zp) != NULL && ZTOV(zp)->v_type == VDIR &&
409 (zp->z_pflags & ZFS_XATTR)) {
410 if (zfs_purgedir(zp) != 0) {
411 /*
412 * Not enough space to delete some xattrs.
413 * Leave it in the unlinked set.
414 */
415 zfs_znode_dmu_fini(zp);
416 zfs_znode_free(zp);
417 return;
418 }
419 } else {
420 /*
421 * Free up all the data in the file. We don't do this for
422 * XATTR directories because we need truncate and remove to be
423 * in the same tx, like in zfs_znode_delete(). Otherwise, if
424 * we crash here we'll end up with an inconsistent truncated
425 * zap object in the delete queue. Note a truncated file is
426 * harmless since it only contains user data.
427 */
428 error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
429 if (error) {
430 /*
431 * Not enough space or we were interrupted by unmount.
432 * Leave the file in the unlinked set.
433 */
434 zfs_znode_dmu_fini(zp);
435 zfs_znode_free(zp);
436 return;
437 }
438 }
439
440 /*
441 * If the file has extended attributes, we're going to unlink
442 * the xattr dir.
443 */
444 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
445 &xattr_obj, sizeof (xattr_obj));
446 if (error == 0 && xattr_obj) {
447 error = zfs_zget(zfsvfs, xattr_obj, &xzp);
448 ASSERT3S(error, ==, 0);
449 vn_lock(ZTOV(xzp), LK_EXCLUSIVE | LK_RETRY);
450 }
451
452 acl_obj = zfs_external_acl(zp);
453
454 /*
455 * Set up the final transaction.
456 */
457 tx = dmu_tx_create(os);
458 dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
459 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
460 if (xzp) {
461 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
462 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
463 }
464 if (acl_obj)
465 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
466
467 zfs_sa_upgrade_txholds(tx, zp);
468 error = dmu_tx_assign(tx, TXG_WAIT);
469 if (error) {
470 /*
471 * Not enough space to delete the file. Leave it in the
472 * unlinked set, leaking it until the fs is remounted (at
473 * which point we'll call zfs_unlinked_drain() to process it).
474 */
475 dmu_tx_abort(tx);
476 zfs_znode_dmu_fini(zp);
477 zfs_znode_free(zp);
478 goto out;
479 }
480
481 if (xzp) {
482 ASSERT(error == 0);
483 xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
484 xzp->z_links = 0; /* no more links to it */
485 VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
486 &xzp->z_links, sizeof (xzp->z_links), tx));
487 zfs_unlinked_add(xzp, tx);
488 }
489
490 /* Remove this znode from the unlinked set */
491 VERIFY3U(0, ==,
492 zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
493
494 zfs_znode_delete(zp, tx);
495
496 dmu_tx_commit(tx);
497 out:
498 if (xzp)
499 vput(ZTOV(xzp));
500 }
501
502 static uint64_t
503 zfs_dirent(znode_t *zp, uint64_t mode)
504 {
505 uint64_t de = zp->z_id;
506
507 if (zp->z_zfsvfs->z_version >= ZPL_VERSION_DIRENT_TYPE)
508 de |= IFTODT(mode) << 60;
509 return (de);
510 }
511
512 /*
513 * Link zp into dzp. Can only fail if zp has been unlinked.
514 */
515 int
516 zfs_link_create(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx,
517 int flag)
518 {
519 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
520 vnode_t *vp = ZTOV(zp);
521 uint64_t value;
522 int zp_is_dir = (vp->v_type == VDIR);
523 sa_bulk_attr_t bulk[5];
524 uint64_t mtime[2], ctime[2];
525 int count = 0;
526 int error;
527
528 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
529 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
530 #if 0
531 if (zp_is_dir) {
532 error = 0;
533 if (dzp->z_links >= LINK_MAX)
534 error = SET_ERROR(EMLINK);
535 return (error);
536 }
537 #endif
538 if (!(flag & ZRENAMING)) {
539 if (zp->z_unlinked) { /* no new links to unlinked zp */
540 ASSERT(!(flag & (ZNEW | ZEXISTS)));
541 return (SET_ERROR(ENOENT));
542 }
543 #if 0
544 if (zp->z_links >= LINK_MAX) {
545 return (SET_ERROR(EMLINK));
546 }
547 #endif
548 zp->z_links++;
549 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
550 &zp->z_links, sizeof (zp->z_links));
551
552 } else {
553 ASSERT(zp->z_unlinked == 0);
554 }
555 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
556 &dzp->z_id, sizeof (dzp->z_id));
557 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
558 &zp->z_pflags, sizeof (zp->z_pflags));
559
560 if (!(flag & ZNEW)) {
561 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
562 ctime, sizeof (ctime));
563 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
564 ctime, B_TRUE);
565 }
566 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
567 ASSERT0(error);
568
569 dzp->z_size++;
570 dzp->z_links += zp_is_dir;
571 count = 0;
572 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
573 &dzp->z_size, sizeof (dzp->z_size));
574 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
575 &dzp->z_links, sizeof (dzp->z_links));
576 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
577 mtime, sizeof (mtime));
578 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
579 ctime, sizeof (ctime));
580 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
581 &dzp->z_pflags, sizeof (dzp->z_pflags));
582 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
583 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
584 ASSERT0(error);
585
586 value = zfs_dirent(zp, zp->z_mode);
587 error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, name,
588 8, 1, &value, tx);
589 VERIFY0(error);
590
591 return (0);
592 }
593
594 /*
595 * The match type in the code for this function should conform to:
596 *
597 * ------------------------------------------------------------------------
598 * fs type | z_norm | lookup type | match type
599 * ---------|-------------|-------------|----------------------------------
600 * CS !norm | 0 | 0 | 0 (exact)
601 * CS norm | formX | 0 | MT_NORMALIZE
602 * CI !norm | upper | !ZCIEXACT | MT_NORMALIZE
603 * CI !norm | upper | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE
604 * CI norm | upper|formX | !ZCIEXACT | MT_NORMALIZE
605 * CI norm | upper|formX | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE
606 * CM !norm | upper | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE
607 * CM !norm | upper | ZCILOOK | MT_NORMALIZE
608 * CM norm | upper|formX | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE
609 * CM norm | upper|formX | ZCILOOK | MT_NORMALIZE
610 *
611 * Abbreviations:
612 * CS = Case Sensitive, CI = Case Insensitive, CM = Case Mixed
613 * upper = case folding set by fs type on creation (U8_TEXTPREP_TOUPPER)
614 * formX = unicode normalization form set on fs creation
615 */
616 static int
617 zfs_dropname(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx,
618 int flag)
619 {
620 int error;
621
622 if (zp->z_zfsvfs->z_norm) {
623 matchtype_t mt = MT_NORMALIZE;
624
625 if (zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) {
626 mt |= MT_MATCH_CASE;
627 }
628
629 error = zap_remove_norm(zp->z_zfsvfs->z_os, dzp->z_id,
630 name, mt, tx);
631 } else {
632 error = zap_remove(zp->z_zfsvfs->z_os, dzp->z_id, name, tx);
633 }
634
635 return (error);
636 }
637
638 /*
639 * Unlink zp from dzp, and mark zp for deletion if this was the last link.
640 * Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
641 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
642 * If it's non-NULL, we use it to indicate whether the znode needs deletion,
643 * and it's the caller's job to do it.
644 */
645 int
646 zfs_link_destroy(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx,
647 int flag, boolean_t *unlinkedp)
648 {
649 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
650 vnode_t *vp = ZTOV(zp);
651 int zp_is_dir = (vp->v_type == VDIR);
652 boolean_t unlinked = B_FALSE;
653 sa_bulk_attr_t bulk[5];
654 uint64_t mtime[2], ctime[2];
655 int count = 0;
656 int error;
657
658 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
659 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
660
661 if (!(flag & ZRENAMING)) {
662
663 if (zp_is_dir && !zfs_dirempty(zp)) {
664 #ifdef illumos
665 return (SET_ERROR(EEXIST));
666 #else
667 return (SET_ERROR(ENOTEMPTY));
668 #endif
669 }
670
671 /*
672 * If we get here, we are going to try to remove the object.
673 * First try removing the name from the directory; if that
674 * fails, return the error.
675 */
676 error = zfs_dropname(dzp, name, zp, tx, flag);
677 if (error != 0) {
678 return (error);
679 }
680
681 if (zp->z_links <= zp_is_dir) {
682 zfs_panic_recover("zfs: link count on vnode %p is %u, "
683 "should be at least %u", zp->z_vnode,
684 (int)zp->z_links,
685 zp_is_dir + 1);
686 zp->z_links = zp_is_dir + 1;
687 }
688 if (--zp->z_links == zp_is_dir) {
689 zp->z_unlinked = B_TRUE;
690 zp->z_links = 0;
691 unlinked = B_TRUE;
692 } else {
693 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
694 NULL, &ctime, sizeof (ctime));
695 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
696 NULL, &zp->z_pflags, sizeof (zp->z_pflags));
697 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime,
698 B_TRUE);
699 }
700 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
701 NULL, &zp->z_links, sizeof (zp->z_links));
702 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
703 count = 0;
704 ASSERT0(error);
705 } else {
706 ASSERT(zp->z_unlinked == 0);
707 error = zfs_dropname(dzp, name, zp, tx, flag);
708 if (error != 0)
709 return (error);
710 }
711
712 dzp->z_size--; /* one dirent removed */
713 dzp->z_links -= zp_is_dir; /* ".." link from zp */
714 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
715 NULL, &dzp->z_links, sizeof (dzp->z_links));
716 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
717 NULL, &dzp->z_size, sizeof (dzp->z_size));
718 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
719 NULL, ctime, sizeof (ctime));
720 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
721 NULL, mtime, sizeof (mtime));
722 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
723 NULL, &dzp->z_pflags, sizeof (dzp->z_pflags));
724 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
725 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
726 ASSERT0(error);
727
728 if (unlinkedp != NULL)
729 *unlinkedp = unlinked;
730 else if (unlinked)
731 zfs_unlinked_add(zp, tx);
732
733 return (0);
734 }
735
736 /*
737 * Indicate whether the directory is empty.
738 */
739 boolean_t
740 zfs_dirempty(znode_t *dzp)
741 {
742 return (dzp->z_size == 2);
743 }
744
745 int
746 zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
747 {
748 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
749 znode_t *xzp;
750 dmu_tx_t *tx;
751 int error;
752 zfs_acl_ids_t acl_ids;
753 boolean_t fuid_dirtied;
754 uint64_t parent;
755
756 *xvpp = NULL;
757
758 /*
759 * In FreeBSD, access checking for creating an EA is being done
760 * in zfs_setextattr(),
761 */
762 #ifndef __FreeBSD_kernel__
763 if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
764 return (error);
765 #endif
766
767 if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
768 &acl_ids)) != 0)
769 return (error);
770 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
771 zfs_acl_ids_free(&acl_ids);
772 return (SET_ERROR(EDQUOT));
773 }
774
775 getnewvnode_reserve(1);
776
777 tx = dmu_tx_create(zfsvfs->z_os);
778 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
779 ZFS_SA_BASE_ATTR_SIZE);
780 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
781 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
782 fuid_dirtied = zfsvfs->z_fuid_dirty;
783 if (fuid_dirtied)
784 zfs_fuid_txhold(zfsvfs, tx);
785 error = dmu_tx_assign(tx, TXG_WAIT);
786 if (error) {
787 zfs_acl_ids_free(&acl_ids);
788 dmu_tx_abort(tx);
789 return (error);
790 }
791 zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids);
792
793 if (fuid_dirtied)
794 zfs_fuid_sync(zfsvfs, tx);
795
796 #ifdef DEBUG
797 error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
798 &parent, sizeof (parent));
799 ASSERT(error == 0 && parent == zp->z_id);
800 #endif
801
802 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id,
803 sizeof (xzp->z_id), tx));
804
805 (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
806 xzp, "", NULL, acl_ids.z_fuidp, vap);
807
808 zfs_acl_ids_free(&acl_ids);
809 dmu_tx_commit(tx);
810
811 getnewvnode_drop_reserve();
812
813 *xvpp = ZTOV(xzp);
814
815 return (0);
816 }
817
818 /*
819 * Return a znode for the extended attribute directory for zp.
820 * ** If the directory does not already exist, it is created **
821 *
822 * IN: zp - znode to obtain attribute directory from
823 * cr - credentials of caller
824 * flags - flags from the VOP_LOOKUP call
825 *
826 * OUT: xzpp - pointer to extended attribute znode
827 *
828 * RETURN: 0 on success
829 * error number on failure
830 */
831 int
832 zfs_get_xattrdir(znode_t *zp, vnode_t **xvpp, cred_t *cr, int flags)
833 {
834 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
835 znode_t *xzp;
836 vattr_t va;
837 int error;
838 top:
839 error = zfs_dirent_lookup(zp, "", &xzp, ZXATTR);
840 if (error)
841 return (error);
842
843 if (xzp != NULL) {
844 *xvpp = ZTOV(xzp);
845 return (0);
846 }
847
848
849 if (!(flags & CREATE_XATTR_DIR)) {
850 #ifdef illumos
851 return (SET_ERROR(ENOENT));
852 #else
853 return (SET_ERROR(ENOATTR));
854 #endif
855 }
856
857 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
858 return (SET_ERROR(EROFS));
859 }
860
861 /*
862 * The ability to 'create' files in an attribute
863 * directory comes from the write_xattr permission on the base file.
864 *
865 * The ability to 'search' an attribute directory requires
866 * read_xattr permission on the base file.
867 *
868 * Once in a directory the ability to read/write attributes
869 * is controlled by the permissions on the attribute file.
870 */
871 va.va_mask = AT_TYPE | AT_MODE | AT_UID | AT_GID;
872 va.va_type = VDIR;
873 va.va_mode = S_IFDIR | S_ISVTX | 0777;
874 zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid);
875
876 error = zfs_make_xattrdir(zp, &va, xvpp, cr);
877
878 if (error == ERESTART) {
879 /* NB: we already did dmu_tx_wait() if necessary */
880 goto top;
881 }
882 if (error == 0)
883 VOP_UNLOCK(*xvpp, 0);
884
885 return (error);
886 }
887
888 /*
889 * Decide whether it is okay to remove within a sticky directory.
890 *
891 * In sticky directories, write access is not sufficient;
892 * you can remove entries from a directory only if:
893 *
894 * you own the directory,
895 * you own the entry,
896 * the entry is a plain file and you have write access,
897 * or you are privileged (checked in secpolicy...).
898 *
899 * The function returns 0 if remove access is granted.
900 */
901 int
902 zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr)
903 {
904 uid_t uid;
905 uid_t downer;
906 uid_t fowner;
907 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
908
909 if (zdp->z_zfsvfs->z_replay)
910 return (0);
911
912 if ((zdp->z_mode & S_ISVTX) == 0)
913 return (0);
914
915 downer = zfs_fuid_map_id(zfsvfs, zdp->z_uid, cr, ZFS_OWNER);
916 fowner = zfs_fuid_map_id(zfsvfs, zp->z_uid, cr, ZFS_OWNER);
917
918 if ((uid = crgetuid(cr)) == downer || uid == fowner ||
919 (ZTOV(zp)->v_type == VREG &&
920 zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0))
921 return (0);
922 else
923 return (secpolicy_vnode_remove(ZTOV(zp), cr));
924 }
Cache object: 31eb642df953b16aca817b45b20a7d5b
|