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
zfs_match_find(zfsvfs_t * zfsvfs,znode_t * dzp,const char * name,boolean_t exact,uint64_t * zoid)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
zfs_dirent_lookup(znode_t * dzp,const char * name,znode_t ** zpp,int flag)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
zfs_dd_lookup(znode_t * dzp,znode_t ** zpp)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
zfs_dirlook(znode_t * dzp,const char * name,znode_t ** zpp)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
zfs_unlinked_add(znode_t * zp,dmu_tx_t * tx)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
zfs_unlinked_drain(zfsvfs_t * zfsvfs)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
zfs_purgedir(znode_t * dzp)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
zfs_rmnode(znode_t * zp)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 }
395
396 /*
397 * Free up all the data in the file.
398 */
399 error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END);
400 if (error) {
401 /*
402 * Not enough space. Leave the file in the unlinked set.
403 */
404 zfs_znode_dmu_fini(zp);
405 zfs_znode_free(zp);
406 return;
407 }
408
409 /*
410 * If the file has extended attributes, we're going to unlink
411 * the xattr dir.
412 */
413 error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs),
414 &xattr_obj, sizeof (xattr_obj));
415 if (error == 0 && xattr_obj) {
416 error = zfs_zget(zfsvfs, xattr_obj, &xzp);
417 ASSERT3S(error, ==, 0);
418 vn_lock(ZTOV(xzp), LK_EXCLUSIVE | LK_RETRY);
419 }
420
421 acl_obj = zfs_external_acl(zp);
422
423 /*
424 * Set up the final transaction.
425 */
426 tx = dmu_tx_create(os);
427 dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
428 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
429 if (xzp) {
430 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
431 dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE);
432 }
433 if (acl_obj)
434 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
435
436 zfs_sa_upgrade_txholds(tx, zp);
437 error = dmu_tx_assign(tx, TXG_WAIT);
438 if (error) {
439 /*
440 * Not enough space to delete the file. Leave it in the
441 * unlinked set, leaking it until the fs is remounted (at
442 * which point we'll call zfs_unlinked_drain() to process it).
443 */
444 dmu_tx_abort(tx);
445 zfs_znode_dmu_fini(zp);
446 zfs_znode_free(zp);
447 goto out;
448 }
449
450 if (xzp) {
451 ASSERT(error == 0);
452 xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
453 xzp->z_links = 0; /* no more links to it */
454 VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs),
455 &xzp->z_links, sizeof (xzp->z_links), tx));
456 zfs_unlinked_add(xzp, tx);
457 }
458
459 /* Remove this znode from the unlinked set */
460 VERIFY3U(0, ==,
461 zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx));
462
463 zfs_znode_delete(zp, tx);
464
465 dmu_tx_commit(tx);
466 out:
467 if (xzp)
468 vput(ZTOV(xzp));
469 }
470
471 static uint64_t
zfs_dirent(znode_t * zp,uint64_t mode)472 zfs_dirent(znode_t *zp, uint64_t mode)
473 {
474 uint64_t de = zp->z_id;
475
476 if (zp->z_zfsvfs->z_version >= ZPL_VERSION_DIRENT_TYPE)
477 de |= IFTODT(mode) << 60;
478 return (de);
479 }
480
481 /*
482 * Link zp into dzp. Can only fail if zp has been unlinked.
483 */
484 int
zfs_link_create(znode_t * dzp,const char * name,znode_t * zp,dmu_tx_t * tx,int flag)485 zfs_link_create(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx,
486 int flag)
487 {
488 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
489 vnode_t *vp = ZTOV(zp);
490 uint64_t value;
491 int zp_is_dir = (vp->v_type == VDIR);
492 sa_bulk_attr_t bulk[5];
493 uint64_t mtime[2], ctime[2];
494 int count = 0;
495 int error;
496
497 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
498 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
499 #if 0
500 if (zp_is_dir) {
501 error = 0;
502 if (dzp->z_links >= LINK_MAX)
503 error = SET_ERROR(EMLINK);
504 return (error);
505 }
506 #endif
507 if (!(flag & ZRENAMING)) {
508 if (zp->z_unlinked) { /* no new links to unlinked zp */
509 ASSERT(!(flag & (ZNEW | ZEXISTS)));
510 return (SET_ERROR(ENOENT));
511 }
512 #if 0
513 if (zp->z_links >= LINK_MAX) {
514 return (SET_ERROR(EMLINK));
515 }
516 #endif
517 zp->z_links++;
518 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
519 &zp->z_links, sizeof (zp->z_links));
520
521 } else {
522 ASSERT(zp->z_unlinked == 0);
523 }
524 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
525 &dzp->z_id, sizeof (dzp->z_id));
526 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
527 &zp->z_pflags, sizeof (zp->z_pflags));
528
529 if (!(flag & ZNEW)) {
530 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
531 ctime, sizeof (ctime));
532 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime,
533 ctime, B_TRUE);
534 }
535 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
536 ASSERT0(error);
537
538 dzp->z_size++;
539 dzp->z_links += zp_is_dir;
540 count = 0;
541 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
542 &dzp->z_size, sizeof (dzp->z_size));
543 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
544 &dzp->z_links, sizeof (dzp->z_links));
545 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL,
546 mtime, sizeof (mtime));
547 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL,
548 ctime, sizeof (ctime));
549 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
550 &dzp->z_pflags, sizeof (dzp->z_pflags));
551 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
552 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
553 ASSERT0(error);
554
555 value = zfs_dirent(zp, zp->z_mode);
556 error = zap_add(zp->z_zfsvfs->z_os, dzp->z_id, name,
557 8, 1, &value, tx);
558 VERIFY0(error);
559
560 return (0);
561 }
562
563 static int
zfs_dropname(znode_t * dzp,const char * name,znode_t * zp,dmu_tx_t * tx,int flag)564 zfs_dropname(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx,
565 int flag)
566 {
567 int error;
568
569 if (zp->z_zfsvfs->z_norm) {
570 if (zp->z_zfsvfs->z_case == ZFS_CASE_MIXED)
571 error = zap_remove_norm(zp->z_zfsvfs->z_os,
572 dzp->z_id, name, MT_EXACT, tx);
573 else
574 error = zap_remove_norm(zp->z_zfsvfs->z_os,
575 dzp->z_id, name, MT_FIRST, tx);
576 } else {
577 error = zap_remove(zp->z_zfsvfs->z_os,
578 dzp->z_id, name, tx);
579 }
580
581 return (error);
582 }
583
584 /*
585 * Unlink zp from dzp, and mark zp for deletion if this was the last link.
586 * Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
587 * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
588 * If it's non-NULL, we use it to indicate whether the znode needs deletion,
589 * and it's the caller's job to do it.
590 */
591 int
zfs_link_destroy(znode_t * dzp,const char * name,znode_t * zp,dmu_tx_t * tx,int flag,boolean_t * unlinkedp)592 zfs_link_destroy(znode_t *dzp, const char *name, znode_t *zp, dmu_tx_t *tx,
593 int flag, boolean_t *unlinkedp)
594 {
595 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
596 vnode_t *vp = ZTOV(zp);
597 int zp_is_dir = (vp->v_type == VDIR);
598 boolean_t unlinked = B_FALSE;
599 sa_bulk_attr_t bulk[5];
600 uint64_t mtime[2], ctime[2];
601 int count = 0;
602 int error;
603
604 ASSERT_VOP_ELOCKED(ZTOV(dzp), __func__);
605 ASSERT_VOP_ELOCKED(ZTOV(zp), __func__);
606
607 if (!(flag & ZRENAMING)) {
608
609 if (zp_is_dir && !zfs_dirempty(zp)) {
610 #ifdef illumos
611 return (SET_ERROR(EEXIST));
612 #else
613 return (SET_ERROR(ENOTEMPTY));
614 #endif
615 }
616
617 /*
618 * If we get here, we are going to try to remove the object.
619 * First try removing the name from the directory; if that
620 * fails, return the error.
621 */
622 error = zfs_dropname(dzp, name, zp, tx, flag);
623 if (error != 0) {
624 return (error);
625 }
626
627 if (zp->z_links <= zp_is_dir) {
628 zfs_panic_recover("zfs: link count on vnode %p is %u, "
629 "should be at least %u", zp->z_vnode,
630 (int)zp->z_links,
631 zp_is_dir + 1);
632 zp->z_links = zp_is_dir + 1;
633 }
634 if (--zp->z_links == zp_is_dir) {
635 zp->z_unlinked = B_TRUE;
636 zp->z_links = 0;
637 unlinked = B_TRUE;
638 } else {
639 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
640 NULL, &ctime, sizeof (ctime));
641 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
642 NULL, &zp->z_pflags, sizeof (zp->z_pflags));
643 zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime,
644 B_TRUE);
645 }
646 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
647 NULL, &zp->z_links, sizeof (zp->z_links));
648 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
649 count = 0;
650 ASSERT0(error);
651 } else {
652 ASSERT(zp->z_unlinked == 0);
653 error = zfs_dropname(dzp, name, zp, tx, flag);
654 if (error != 0)
655 return (error);
656 }
657
658 dzp->z_size--; /* one dirent removed */
659 dzp->z_links -= zp_is_dir; /* ".." link from zp */
660 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs),
661 NULL, &dzp->z_links, sizeof (dzp->z_links));
662 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
663 NULL, &dzp->z_size, sizeof (dzp->z_size));
664 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs),
665 NULL, ctime, sizeof (ctime));
666 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs),
667 NULL, mtime, sizeof (mtime));
668 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
669 NULL, &dzp->z_pflags, sizeof (dzp->z_pflags));
670 zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
671 error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx);
672 ASSERT0(error);
673
674 if (unlinkedp != NULL)
675 *unlinkedp = unlinked;
676 else if (unlinked)
677 zfs_unlinked_add(zp, tx);
678
679 return (0);
680 }
681
682 /*
683 * Indicate whether the directory is empty.
684 */
685 boolean_t
zfs_dirempty(znode_t * dzp)686 zfs_dirempty(znode_t *dzp)
687 {
688 return (dzp->z_size == 2);
689 }
690
691 int
zfs_make_xattrdir(znode_t * zp,vattr_t * vap,vnode_t ** xvpp,cred_t * cr)692 zfs_make_xattrdir(znode_t *zp, vattr_t *vap, vnode_t **xvpp, cred_t *cr)
693 {
694 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
695 znode_t *xzp;
696 dmu_tx_t *tx;
697 int error;
698 zfs_acl_ids_t acl_ids;
699 boolean_t fuid_dirtied;
700 uint64_t parent;
701
702 *xvpp = NULL;
703
704 /*
705 * In FreeBSD, access checking for creating an EA is being done
706 * in zfs_setextattr(),
707 */
708 #ifndef __FreeBSD_kernel__
709 if (error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))
710 return (error);
711 #endif
712
713 if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL,
714 &acl_ids)) != 0)
715 return (error);
716 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) {
717 zfs_acl_ids_free(&acl_ids);
718 return (SET_ERROR(EDQUOT));
719 }
720
721 getnewvnode_reserve(1);
722
723 tx = dmu_tx_create(zfsvfs->z_os);
724 dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes +
725 ZFS_SA_BASE_ATTR_SIZE);
726 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE);
727 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
728 fuid_dirtied = zfsvfs->z_fuid_dirty;
729 if (fuid_dirtied)
730 zfs_fuid_txhold(zfsvfs, tx);
731 error = dmu_tx_assign(tx, TXG_WAIT);
732 if (error) {
733 zfs_acl_ids_free(&acl_ids);
734 dmu_tx_abort(tx);
735 return (error);
736 }
737 zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids);
738
739 if (fuid_dirtied)
740 zfs_fuid_sync(zfsvfs, tx);
741
742 #ifdef DEBUG
743 error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs),
744 &parent, sizeof (parent));
745 ASSERT(error == 0 && parent == zp->z_id);
746 #endif
747
748 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id,
749 sizeof (xzp->z_id), tx));
750
751 (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp,
752 xzp, "", NULL, acl_ids.z_fuidp, vap);
753
754 zfs_acl_ids_free(&acl_ids);
755 dmu_tx_commit(tx);
756
757 getnewvnode_drop_reserve();
758
759 *xvpp = ZTOV(xzp);
760
761 return (0);
762 }
763
764 /*
765 * Return a znode for the extended attribute directory for zp.
766 * ** If the directory does not already exist, it is created **
767 *
768 * IN: zp - znode to obtain attribute directory from
769 * cr - credentials of caller
770 * flags - flags from the VOP_LOOKUP call
771 *
772 * OUT: xzpp - pointer to extended attribute znode
773 *
774 * RETURN: 0 on success
775 * error number on failure
776 */
777 int
zfs_get_xattrdir(znode_t * zp,vnode_t ** xvpp,cred_t * cr,int flags)778 zfs_get_xattrdir(znode_t *zp, vnode_t **xvpp, cred_t *cr, int flags)
779 {
780 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
781 znode_t *xzp;
782 vattr_t va;
783 int error;
784 top:
785 error = zfs_dirent_lookup(zp, "", &xzp, ZXATTR);
786 if (error)
787 return (error);
788
789 if (xzp != NULL) {
790 *xvpp = ZTOV(xzp);
791 return (0);
792 }
793
794
795 if (!(flags & CREATE_XATTR_DIR)) {
796 #ifdef illumos
797 return (SET_ERROR(ENOENT));
798 #else
799 return (SET_ERROR(ENOATTR));
800 #endif
801 }
802
803 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) {
804 return (SET_ERROR(EROFS));
805 }
806
807 /*
808 * The ability to 'create' files in an attribute
809 * directory comes from the write_xattr permission on the base file.
810 *
811 * The ability to 'search' an attribute directory requires
812 * read_xattr permission on the base file.
813 *
814 * Once in a directory the ability to read/write attributes
815 * is controlled by the permissions on the attribute file.
816 */
817 va.va_mask = AT_TYPE | AT_MODE | AT_UID | AT_GID;
818 va.va_type = VDIR;
819 va.va_mode = S_IFDIR | S_ISVTX | 0777;
820 zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid);
821
822 error = zfs_make_xattrdir(zp, &va, xvpp, cr);
823
824 if (error == ERESTART) {
825 /* NB: we already did dmu_tx_wait() if necessary */
826 goto top;
827 }
828 if (error == 0)
829 VOP_UNLOCK(*xvpp, 0);
830
831 return (error);
832 }
833
834 /*
835 * Decide whether it is okay to remove within a sticky directory.
836 *
837 * In sticky directories, write access is not sufficient;
838 * you can remove entries from a directory only if:
839 *
840 * you own the directory,
841 * you own the entry,
842 * the entry is a plain file and you have write access,
843 * or you are privileged (checked in secpolicy...).
844 *
845 * The function returns 0 if remove access is granted.
846 */
847 int
zfs_sticky_remove_access(znode_t * zdp,znode_t * zp,cred_t * cr)848 zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr)
849 {
850 uid_t uid;
851 uid_t downer;
852 uid_t fowner;
853 zfsvfs_t *zfsvfs = zdp->z_zfsvfs;
854
855 if (zdp->z_zfsvfs->z_replay)
856 return (0);
857
858 if ((zdp->z_mode & S_ISVTX) == 0)
859 return (0);
860
861 downer = zfs_fuid_map_id(zfsvfs, zdp->z_uid, cr, ZFS_OWNER);
862 fowner = zfs_fuid_map_id(zfsvfs, zp->z_uid, cr, ZFS_OWNER);
863
864 if ((uid = crgetuid(cr)) == downer || uid == fowner ||
865 (ZTOV(zp)->v_type == VREG &&
866 zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0))
867 return (0);
868 else
869 return (secpolicy_vnode_remove(ZTOV(zp), cr));
870 }
871