]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
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 | */ | |
9b7b9cd3 | 21 | |
34dc7c2f | 22 | /* |
428870ff | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
9e2c3bb4 | 24 | * Copyright (c) 2013, 2016 by Delphix. All rights reserved. |
9b7b9cd3 | 25 | * Copyright 2017 Nexenta Systems, Inc. |
34dc7c2f BB |
26 | */ |
27 | ||
34dc7c2f BB |
28 | #include <sys/types.h> |
29 | #include <sys/param.h> | |
30 | #include <sys/time.h> | |
34dc7c2f | 31 | #include <sys/sysmacros.h> |
34dc7c2f BB |
32 | #include <sys/vfs.h> |
33 | #include <sys/vnode.h> | |
34 | #include <sys/file.h> | |
35 | #include <sys/mode.h> | |
36 | #include <sys/kmem.h> | |
37 | #include <sys/uio.h> | |
38 | #include <sys/pathname.h> | |
39 | #include <sys/cmn_err.h> | |
40 | #include <sys/errno.h> | |
41 | #include <sys/stat.h> | |
34dc7c2f BB |
42 | #include <sys/sunddi.h> |
43 | #include <sys/random.h> | |
44 | #include <sys/policy.h> | |
45 | #include <sys/zfs_dir.h> | |
46 | #include <sys/zfs_acl.h> | |
0a50679c | 47 | #include <sys/zfs_vnops.h> |
34dc7c2f | 48 | #include <sys/fs/zfs.h> |
34dc7c2f BB |
49 | #include <sys/zap.h> |
50 | #include <sys/dmu.h> | |
51 | #include <sys/atomic.h> | |
ebe7e575 | 52 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 53 | #include <sys/zfs_fuid.h> |
428870ff BB |
54 | #include <sys/sa.h> |
55 | #include <sys/zfs_sa.h> | |
34dc7c2f BB |
56 | |
57 | /* | |
58 | * zfs_match_find() is used by zfs_dirent_lock() to peform zap lookups | |
59 | * of names after deciding which is the appropriate lookup interface. | |
60 | */ | |
61 | static int | |
0037b49e | 62 | zfs_match_find(zfsvfs_t *zfsvfs, znode_t *dzp, char *name, matchtype_t mt, |
34dc7c2f BB |
63 | boolean_t update, int *deflags, pathname_t *rpnp, uint64_t *zoid) |
64 | { | |
0c5dde49 | 65 | boolean_t conflict = B_FALSE; |
34dc7c2f BB |
66 | int error; |
67 | ||
0037b49e | 68 | if (zfsvfs->z_norm) { |
34dc7c2f BB |
69 | size_t bufsz = 0; |
70 | char *buf = NULL; | |
71 | ||
72 | if (rpnp) { | |
73 | buf = rpnp->pn_buf; | |
74 | bufsz = rpnp->pn_bufsize; | |
75 | } | |
9b7b9cd3 | 76 | |
34dc7c2f BB |
77 | /* |
78 | * In the non-mixed case we only expect there would ever | |
79 | * be one match, but we need to use the normalizing lookup. | |
80 | */ | |
0037b49e | 81 | error = zap_lookup_norm(zfsvfs->z_os, dzp->z_id, name, 8, 1, |
34dc7c2f | 82 | zoid, mt, buf, bufsz, &conflict); |
34dc7c2f | 83 | } else { |
0037b49e | 84 | error = zap_lookup(zfsvfs->z_os, dzp->z_id, name, 8, 1, zoid); |
34dc7c2f | 85 | } |
0c5dde49 BB |
86 | |
87 | /* | |
88 | * Allow multiple entries provided the first entry is | |
89 | * the object id. Non-zpl consumers may safely make | |
90 | * use of the additional space. | |
91 | * | |
92 | * XXX: This should be a feature flag for compatibility | |
93 | */ | |
94 | if (error == EOVERFLOW) | |
95 | error = 0; | |
96 | ||
0037b49e | 97 | if (zfsvfs->z_norm && !error && deflags) |
0c5dde49 BB |
98 | *deflags = conflict ? ED_CASE_CONFLICT : 0; |
99 | ||
34dc7c2f BB |
100 | *zoid = ZFS_DIRENT_OBJ(*zoid); |
101 | ||
34dc7c2f BB |
102 | return (error); |
103 | } | |
104 | ||
105 | /* | |
106 | * Lock a directory entry. A dirlock on <dzp, name> protects that name | |
107 | * in dzp's directory zap object. As long as you hold a dirlock, you can | |
108 | * assume two things: (1) dzp cannot be reaped, and (2) no other thread | |
109 | * can change the zap entry for (i.e. link or unlink) this name. | |
110 | * | |
111 | * Input arguments: | |
112 | * dzp - znode for directory | |
113 | * name - name of entry to lock | |
114 | * flag - ZNEW: if the entry already exists, fail with EEXIST. | |
115 | * ZEXISTS: if the entry does not exist, fail with ENOENT. | |
116 | * ZSHARED: allow concurrent access with other ZSHARED callers. | |
117 | * ZXATTR: we want dzp's xattr directory | |
118 | * ZCILOOK: On a mixed sensitivity file system, | |
119 | * this lookup should be case-insensitive. | |
120 | * ZCIEXACT: On a purely case-insensitive file system, | |
121 | * this lookup should be case-sensitive. | |
122 | * ZRENAMING: we are locking for renaming, force narrow locks | |
428870ff BB |
123 | * ZHAVELOCK: Don't grab the z_name_lock for this call. The |
124 | * current thread already holds it. | |
34dc7c2f BB |
125 | * |
126 | * Output arguments: | |
127 | * zpp - pointer to the znode for the entry (NULL if there isn't one) | |
128 | * dlpp - pointer to the dirlock for this entry (NULL on error) | |
129 | * direntflags - (case-insensitive lookup only) | |
130 | * flags if multiple case-sensitive matches exist in directory | |
131 | * realpnp - (case-insensitive lookup only) | |
132 | * actual name matched within the directory | |
133 | * | |
134 | * Return value: 0 on success or errno on failure. | |
135 | * | |
136 | * NOTE: Always checks for, and rejects, '.' and '..'. | |
137 | * NOTE: For case-insensitive file systems we take wide locks (see below), | |
138 | * but return znode pointers to a single match. | |
139 | */ | |
140 | int | |
141 | zfs_dirent_lock(zfs_dirlock_t **dlpp, znode_t *dzp, char *name, znode_t **zpp, | |
142 | int flag, int *direntflags, pathname_t *realpnp) | |
143 | { | |
0037b49e | 144 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f BB |
145 | zfs_dirlock_t *dl; |
146 | boolean_t update; | |
9b7b9cd3 | 147 | matchtype_t mt = 0; |
34dc7c2f | 148 | uint64_t zoid; |
34dc7c2f BB |
149 | int error = 0; |
150 | int cmpflags; | |
151 | ||
152 | *zpp = NULL; | |
153 | *dlpp = NULL; | |
154 | ||
155 | /* | |
156 | * Verify that we are not trying to lock '.', '..', or '.zfs' | |
157 | */ | |
149e873a BB |
158 | if ((name[0] == '.' && |
159 | (name[1] == '\0' || (name[1] == '.' && name[2] == '\0'))) || | |
160 | (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)) | |
2e528b49 | 161 | return (SET_ERROR(EEXIST)); |
34dc7c2f BB |
162 | |
163 | /* | |
164 | * Case sensitivity and normalization preferences are set when | |
165 | * the file system is created. These are stored in the | |
0037b49e | 166 | * zfsvfs->z_case and zfsvfs->z_norm fields. These choices |
34dc7c2f BB |
167 | * affect what vnodes can be cached in the DNLC, how we |
168 | * perform zap lookups, and the "width" of our dirlocks. | |
169 | * | |
170 | * A normal dirlock locks a single name. Note that with | |
171 | * normalization a name can be composed multiple ways, but | |
172 | * when normalized, these names all compare equal. A wide | |
173 | * dirlock locks multiple names. We need these when the file | |
174 | * system is supporting mixed-mode access. It is sometimes | |
175 | * necessary to lock all case permutations of file name at | |
176 | * once so that simultaneous case-insensitive/case-sensitive | |
177 | * behaves as rationally as possible. | |
178 | */ | |
179 | ||
180 | /* | |
9b7b9cd3 GM |
181 | * When matching we may need to normalize & change case according to |
182 | * FS settings. | |
183 | * | |
184 | * Note that a normalized match is necessary for a case insensitive | |
185 | * filesystem when the lookup request is not exact because normalization | |
186 | * can fold case independent of normalizing code point sequences. | |
187 | * | |
188 | * See the table above zfs_dropname(). | |
34dc7c2f | 189 | */ |
0037b49e | 190 | if (zfsvfs->z_norm != 0) { |
9b7b9cd3 GM |
191 | mt = MT_NORMALIZE; |
192 | ||
193 | /* | |
194 | * Determine if the match needs to honor the case specified in | |
195 | * lookup, and if so keep track of that so that during | |
196 | * normalization we don't fold case. | |
197 | */ | |
0037b49e | 198 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE && |
9b7b9cd3 | 199 | (flag & ZCIEXACT)) || |
0037b49e | 200 | (zfsvfs->z_case == ZFS_CASE_MIXED && !(flag & ZCILOOK))) { |
9b7b9cd3 GM |
201 | mt |= MT_MATCH_CASE; |
202 | } | |
203 | } | |
34dc7c2f BB |
204 | |
205 | /* | |
206 | * Only look in or update the DNLC if we are looking for the | |
207 | * name on a file system that does not require normalization | |
208 | * or case folding. We can also look there if we happen to be | |
209 | * on a non-normalizing, mixed sensitivity file system IF we | |
210 | * are looking for the exact name. | |
211 | * | |
212 | * Maybe can add TO-UPPERed version of name to dnlc in ci-only | |
213 | * case for performance improvement? | |
214 | */ | |
0037b49e BB |
215 | update = !zfsvfs->z_norm || |
216 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
217 | !(zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER) && !(flag & ZCILOOK)); | |
34dc7c2f BB |
218 | |
219 | /* | |
220 | * ZRENAMING indicates we are in a situation where we should | |
221 | * take narrow locks regardless of the file system's | |
222 | * preferences for normalizing and case folding. This will | |
223 | * prevent us deadlocking trying to grab the same wide lock | |
224 | * twice if the two names happen to be case-insensitive | |
225 | * matches. | |
226 | */ | |
227 | if (flag & ZRENAMING) | |
228 | cmpflags = 0; | |
229 | else | |
0037b49e | 230 | cmpflags = zfsvfs->z_norm; |
34dc7c2f BB |
231 | |
232 | /* | |
233 | * Wait until there are no locks on this name. | |
428870ff BB |
234 | * |
235 | * Don't grab the the lock if it is already held. However, cannot | |
236 | * have both ZSHARED and ZHAVELOCK together. | |
34dc7c2f | 237 | */ |
428870ff BB |
238 | ASSERT(!(flag & ZSHARED) || !(flag & ZHAVELOCK)); |
239 | if (!(flag & ZHAVELOCK)) | |
240 | rw_enter(&dzp->z_name_lock, RW_READER); | |
241 | ||
34dc7c2f BB |
242 | mutex_enter(&dzp->z_lock); |
243 | for (;;) { | |
98701490 | 244 | if (dzp->z_unlinked && !(flag & ZXATTR)) { |
34dc7c2f | 245 | mutex_exit(&dzp->z_lock); |
428870ff BB |
246 | if (!(flag & ZHAVELOCK)) |
247 | rw_exit(&dzp->z_name_lock); | |
2e528b49 | 248 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
249 | } |
250 | for (dl = dzp->z_dirlocks; dl != NULL; dl = dl->dl_next) { | |
251 | if ((u8_strcmp(name, dl->dl_name, 0, cmpflags, | |
252 | U8_UNICODE_LATEST, &error) == 0) || error != 0) | |
253 | break; | |
254 | } | |
255 | if (error != 0) { | |
256 | mutex_exit(&dzp->z_lock); | |
428870ff BB |
257 | if (!(flag & ZHAVELOCK)) |
258 | rw_exit(&dzp->z_name_lock); | |
2e528b49 | 259 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
260 | } |
261 | if (dl == NULL) { | |
262 | /* | |
263 | * Allocate a new dirlock and add it to the list. | |
264 | */ | |
265 | dl = kmem_alloc(sizeof (zfs_dirlock_t), KM_SLEEP); | |
266 | cv_init(&dl->dl_cv, NULL, CV_DEFAULT, NULL); | |
267 | dl->dl_name = name; | |
268 | dl->dl_sharecnt = 0; | |
428870ff | 269 | dl->dl_namelock = 0; |
34dc7c2f BB |
270 | dl->dl_namesize = 0; |
271 | dl->dl_dzp = dzp; | |
272 | dl->dl_next = dzp->z_dirlocks; | |
273 | dzp->z_dirlocks = dl; | |
274 | break; | |
275 | } | |
276 | if ((flag & ZSHARED) && dl->dl_sharecnt != 0) | |
277 | break; | |
278 | cv_wait(&dl->dl_cv, &dzp->z_lock); | |
279 | } | |
280 | ||
428870ff BB |
281 | /* |
282 | * If the z_name_lock was NOT held for this dirlock record it. | |
283 | */ | |
284 | if (flag & ZHAVELOCK) | |
285 | dl->dl_namelock = 1; | |
286 | ||
34dc7c2f BB |
287 | if ((flag & ZSHARED) && ++dl->dl_sharecnt > 1 && dl->dl_namesize == 0) { |
288 | /* | |
289 | * We're the second shared reference to dl. Make a copy of | |
290 | * dl_name in case the first thread goes away before we do. | |
291 | * Note that we initialize the new name before storing its | |
292 | * pointer into dl_name, because the first thread may load | |
9e2c3bb4 DH |
293 | * dl->dl_name at any time. It'll either see the old value, |
294 | * which belongs to it, or the new shared copy; either is OK. | |
34dc7c2f BB |
295 | */ |
296 | dl->dl_namesize = strlen(dl->dl_name) + 1; | |
297 | name = kmem_alloc(dl->dl_namesize, KM_SLEEP); | |
298 | bcopy(dl->dl_name, name, dl->dl_namesize); | |
299 | dl->dl_name = name; | |
300 | } | |
301 | ||
302 | mutex_exit(&dzp->z_lock); | |
303 | ||
304 | /* | |
305 | * We have a dirlock on the name. (Note that it is the dirlock, | |
306 | * not the dzp's z_lock, that protects the name in the zap object.) | |
307 | * See if there's an object by this name; if so, put a hold on it. | |
308 | */ | |
309 | if (flag & ZXATTR) { | |
0037b49e | 310 | error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid, |
428870ff BB |
311 | sizeof (zoid)); |
312 | if (error == 0) | |
2e528b49 | 313 | error = (zoid == 0 ? SET_ERROR(ENOENT) : 0); |
34dc7c2f | 314 | } else { |
0037b49e | 315 | error = zfs_match_find(zfsvfs, dzp, name, mt, |
3558fd73 | 316 | update, direntflags, realpnp, &zoid); |
34dc7c2f BB |
317 | } |
318 | if (error) { | |
319 | if (error != ENOENT || (flag & ZEXISTS)) { | |
320 | zfs_dirent_unlock(dl); | |
321 | return (error); | |
322 | } | |
323 | } else { | |
324 | if (flag & ZNEW) { | |
325 | zfs_dirent_unlock(dl); | |
2e528b49 | 326 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 327 | } |
0037b49e | 328 | error = zfs_zget(zfsvfs, zoid, zpp); |
34dc7c2f BB |
329 | if (error) { |
330 | zfs_dirent_unlock(dl); | |
331 | return (error); | |
332 | } | |
34dc7c2f BB |
333 | } |
334 | ||
335 | *dlpp = dl; | |
336 | ||
337 | return (0); | |
338 | } | |
339 | ||
340 | /* | |
341 | * Unlock this directory entry and wake anyone who was waiting for it. | |
342 | */ | |
343 | void | |
344 | zfs_dirent_unlock(zfs_dirlock_t *dl) | |
345 | { | |
346 | znode_t *dzp = dl->dl_dzp; | |
347 | zfs_dirlock_t **prev_dl, *cur_dl; | |
348 | ||
349 | mutex_enter(&dzp->z_lock); | |
428870ff BB |
350 | |
351 | if (!dl->dl_namelock) | |
352 | rw_exit(&dzp->z_name_lock); | |
353 | ||
34dc7c2f BB |
354 | if (dl->dl_sharecnt > 1) { |
355 | dl->dl_sharecnt--; | |
356 | mutex_exit(&dzp->z_lock); | |
357 | return; | |
358 | } | |
359 | prev_dl = &dzp->z_dirlocks; | |
360 | while ((cur_dl = *prev_dl) != dl) | |
361 | prev_dl = &cur_dl->dl_next; | |
362 | *prev_dl = dl->dl_next; | |
363 | cv_broadcast(&dl->dl_cv); | |
364 | mutex_exit(&dzp->z_lock); | |
365 | ||
366 | if (dl->dl_namesize != 0) | |
367 | kmem_free(dl->dl_name, dl->dl_namesize); | |
368 | cv_destroy(&dl->dl_cv); | |
369 | kmem_free(dl, sizeof (*dl)); | |
370 | } | |
371 | ||
372 | /* | |
373 | * Look up an entry in a directory. | |
374 | * | |
375 | * NOTE: '.' and '..' are handled as special cases because | |
376 | * no directory entries are actually stored for them. If this is | |
377 | * the root of a filesystem, then '.zfs' is also treated as a | |
378 | * special pseudo-directory. | |
379 | */ | |
380 | int | |
3558fd73 | 381 | zfs_dirlook(znode_t *dzp, char *name, struct inode **ipp, int flags, |
34dc7c2f BB |
382 | int *deflg, pathname_t *rpnp) |
383 | { | |
384 | zfs_dirlock_t *dl; | |
385 | znode_t *zp; | |
386 | int error = 0; | |
428870ff | 387 | uint64_t parent; |
34dc7c2f BB |
388 | |
389 | if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) { | |
3558fd73 BB |
390 | *ipp = ZTOI(dzp); |
391 | igrab(*ipp); | |
34dc7c2f | 392 | } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) { |
0037b49e | 393 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
428870ff | 394 | |
34dc7c2f BB |
395 | /* |
396 | * If we are a snapshot mounted under .zfs, return | |
ebe7e575 | 397 | * the inode pointer for the snapshot directory. |
34dc7c2f | 398 | */ |
428870ff | 399 | if ((error = sa_lookup(dzp->z_sa_hdl, |
0037b49e | 400 | SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0) |
428870ff | 401 | return (error); |
ebe7e575 | 402 | |
0037b49e BB |
403 | if (parent == dzp->z_id && zfsvfs->z_parent != zfsvfs) { |
404 | error = zfsctl_root_lookup(zfsvfs->z_parent->z_ctldir, | |
ebe7e575 | 405 | "snapshot", ipp, 0, kcred, NULL, NULL); |
34dc7c2f BB |
406 | return (error); |
407 | } | |
408 | rw_enter(&dzp->z_parent_lock, RW_READER); | |
0037b49e | 409 | error = zfs_zget(zfsvfs, parent, &zp); |
34dc7c2f | 410 | if (error == 0) |
3558fd73 | 411 | *ipp = ZTOI(zp); |
34dc7c2f BB |
412 | rw_exit(&dzp->z_parent_lock); |
413 | } else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) { | |
3558fd73 | 414 | *ipp = zfsctl_root(dzp); |
34dc7c2f BB |
415 | } else { |
416 | int zf; | |
417 | ||
418 | zf = ZEXISTS | ZSHARED; | |
419 | if (flags & FIGNORECASE) | |
420 | zf |= ZCILOOK; | |
421 | ||
422 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zf, deflg, rpnp); | |
423 | if (error == 0) { | |
3558fd73 | 424 | *ipp = ZTOI(zp); |
34dc7c2f BB |
425 | zfs_dirent_unlock(dl); |
426 | dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */ | |
427 | } | |
428 | rpnp = NULL; | |
429 | } | |
430 | ||
431 | if ((flags & FIGNORECASE) && rpnp && !error) | |
432 | (void) strlcpy(rpnp->pn_buf, name, rpnp->pn_bufsize); | |
433 | ||
434 | return (error); | |
435 | } | |
436 | ||
34dc7c2f BB |
437 | /* |
438 | * unlinked Set (formerly known as the "delete queue") Error Handling | |
439 | * | |
440 | * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we | |
441 | * don't specify the name of the entry that we will be manipulating. We | |
442 | * also fib and say that we won't be adding any new entries to the | |
443 | * unlinked set, even though we might (this is to lower the minimum file | |
444 | * size that can be deleted in a full filesystem). So on the small | |
445 | * chance that the nlink list is using a fat zap (ie. has more than | |
446 | * 2000 entries), we *may* not pre-read a block that's needed. | |
447 | * Therefore it is remotely possible for some of the assertions | |
448 | * regarding the unlinked set below to fail due to i/o error. On a | |
449 | * nondebug system, this will result in the space being leaked. | |
450 | */ | |
451 | void | |
452 | zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx) | |
453 | { | |
0037b49e | 454 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
455 | |
456 | ASSERT(zp->z_unlinked); | |
dfbc8630 | 457 | ASSERT(ZTOI(zp)->i_nlink == 0); |
34dc7c2f | 458 | |
b128c09f | 459 | VERIFY3U(0, ==, |
0037b49e | 460 | zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); |
34dc7c2f BB |
461 | } |
462 | ||
0d5c500d BB |
463 | /* |
464 | * Clean up any znodes that had no links when we either crashed or | |
465 | * (force) umounted the file system. | |
466 | */ | |
467 | void | |
0037b49e | 468 | zfs_unlinked_drain(zfsvfs_t *zfsvfs) |
0d5c500d BB |
469 | { |
470 | zap_cursor_t zc; | |
471 | zap_attribute_t zap; | |
472 | dmu_object_info_t doi; | |
473 | znode_t *zp; | |
474 | int error; | |
475 | ||
476 | /* | |
477 | * Iterate over the contents of the unlinked set. | |
478 | */ | |
0037b49e | 479 | for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj); |
0d5c500d BB |
480 | zap_cursor_retrieve(&zc, &zap) == 0; |
481 | zap_cursor_advance(&zc)) { | |
482 | ||
483 | /* | |
484 | * See what kind of object we have in list | |
485 | */ | |
486 | ||
0037b49e BB |
487 | error = dmu_object_info(zfsvfs->z_os, |
488 | zap.za_first_integer, &doi); | |
0d5c500d BB |
489 | if (error != 0) |
490 | continue; | |
491 | ||
492 | ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) || | |
493 | (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS)); | |
494 | /* | |
495 | * We need to re-mark these list entries for deletion, | |
496 | * so we pull them back into core and set zp->z_unlinked. | |
497 | */ | |
0037b49e | 498 | error = zfs_zget(zfsvfs, zap.za_first_integer, &zp); |
0d5c500d BB |
499 | |
500 | /* | |
501 | * We may pick up znodes that are already marked for deletion. | |
502 | * This could happen during the purge of an extended attribute | |
503 | * directory. All we need to do is skip over them, since they | |
504 | * are already in the system marked z_unlinked. | |
505 | */ | |
506 | if (error != 0) | |
507 | continue; | |
508 | ||
509 | zp->z_unlinked = B_TRUE; | |
510 | iput(ZTOI(zp)); | |
511 | } | |
512 | zap_cursor_fini(&zc); | |
513 | } | |
514 | ||
34dc7c2f BB |
515 | /* |
516 | * Delete the entire contents of a directory. Return a count | |
517 | * of the number of entries that could not be deleted. If we encounter | |
518 | * an error, return a count of at least one so that the directory stays | |
519 | * in the unlinked set. | |
520 | * | |
521 | * NOTE: this function assumes that the directory is inactive, | |
522 | * so there is no need to lock its entries before deletion. | |
523 | * Also, it assumes the directory contents is *only* regular | |
524 | * files. | |
525 | */ | |
526 | static int | |
527 | zfs_purgedir(znode_t *dzp) | |
528 | { | |
529 | zap_cursor_t zc; | |
530 | zap_attribute_t zap; | |
531 | znode_t *xzp; | |
532 | dmu_tx_t *tx; | |
0037b49e | 533 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f BB |
534 | zfs_dirlock_t dl; |
535 | int skipped = 0; | |
536 | int error; | |
537 | ||
0037b49e | 538 | for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); |
34dc7c2f BB |
539 | (error = zap_cursor_retrieve(&zc, &zap)) == 0; |
540 | zap_cursor_advance(&zc)) { | |
0037b49e | 541 | error = zfs_zget(zfsvfs, |
34dc7c2f BB |
542 | ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp); |
543 | if (error) { | |
544 | skipped += 1; | |
545 | continue; | |
546 | } | |
547 | ||
0d5c500d BB |
548 | ASSERT(S_ISREG(ZTOI(xzp)->i_mode) || |
549 | S_ISLNK(ZTOI(xzp)->i_mode)); | |
34dc7c2f | 550 | |
0037b49e | 551 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 552 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
34dc7c2f | 553 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name); |
428870ff | 554 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); |
0037b49e | 555 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
556 | /* Is this really needed ? */ |
557 | zfs_sa_upgrade_txholds(tx, xzp); | |
19d55079 | 558 | dmu_tx_mark_netfree(tx); |
34dc7c2f BB |
559 | error = dmu_tx_assign(tx, TXG_WAIT); |
560 | if (error) { | |
561 | dmu_tx_abort(tx); | |
0a50679c | 562 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
563 | skipped += 1; |
564 | continue; | |
565 | } | |
566 | bzero(&dl, sizeof (dl)); | |
567 | dl.dl_dzp = dzp; | |
568 | dl.dl_name = zap.za_name; | |
569 | ||
570 | error = zfs_link_destroy(&dl, xzp, tx, 0, NULL); | |
571 | if (error) | |
572 | skipped += 1; | |
573 | dmu_tx_commit(tx); | |
dfbc8630 | 574 | |
0a50679c | 575 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
576 | } |
577 | zap_cursor_fini(&zc); | |
578 | if (error != ENOENT) | |
579 | skipped += 1; | |
580 | return (skipped); | |
581 | } | |
582 | ||
583 | void | |
584 | zfs_rmnode(znode_t *zp) | |
585 | { | |
0037b49e BB |
586 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
587 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 588 | znode_t *xzp = NULL; |
34dc7c2f BB |
589 | dmu_tx_t *tx; |
590 | uint64_t acl_obj; | |
428870ff | 591 | uint64_t xattr_obj; |
dfbc8630 | 592 | uint64_t links; |
34dc7c2f BB |
593 | int error; |
594 | ||
dfbc8630 | 595 | ASSERT(ZTOI(zp)->i_nlink == 0); |
3558fd73 | 596 | ASSERT(atomic_read(&ZTOI(zp)->i_count) == 0); |
34dc7c2f BB |
597 | |
598 | /* | |
599 | * If this is an attribute directory, purge its contents. | |
600 | */ | |
3558fd73 | 601 | if (S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_XATTR)) { |
0d5c500d | 602 | if (zfs_purgedir(zp) != 0) { |
34dc7c2f | 603 | /* |
0d5c500d BB |
604 | * Not enough space to delete some xattrs. |
605 | * Leave it in the unlinked set. | |
34dc7c2f | 606 | */ |
7973e464 | 607 | zfs_znode_dmu_fini(zp); |
0d5c500d | 608 | |
34dc7c2f BB |
609 | return; |
610 | } | |
611 | } | |
612 | ||
b128c09f | 613 | /* |
f5f087eb CC |
614 | * Free up all the data in the file. We don't do this for directories |
615 | * because we need truncate and remove to be in the same tx, like in | |
616 | * zfs_znode_delete(). Otherwise, if we crash here we'll end up with | |
617 | * an inconsistent truncated zap object in the delete queue. Note a | |
618 | * truncated file is harmless since it only contains user data. | |
b128c09f | 619 | */ |
f5f087eb CC |
620 | if (S_ISREG(ZTOI(zp)->i_mode)) { |
621 | error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END); | |
622 | if (error) { | |
623 | /* | |
a08abc1b GM |
624 | * Not enough space or we were interrupted by unmount. |
625 | * Leave the file in the unlinked set. | |
f5f087eb CC |
626 | */ |
627 | zfs_znode_dmu_fini(zp); | |
628 | return; | |
629 | } | |
b128c09f BB |
630 | } |
631 | ||
34dc7c2f BB |
632 | /* |
633 | * If the file has extended attributes, we're going to unlink | |
634 | * the xattr dir. | |
635 | */ | |
0037b49e | 636 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff BB |
637 | &xattr_obj, sizeof (xattr_obj)); |
638 | if (error == 0 && xattr_obj) { | |
0037b49e | 639 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
34dc7c2f BB |
640 | ASSERT(error == 0); |
641 | } | |
642 | ||
572e2857 | 643 | acl_obj = zfs_external_acl(zp); |
34dc7c2f BB |
644 | |
645 | /* | |
b128c09f | 646 | * Set up the final transaction. |
34dc7c2f BB |
647 | */ |
648 | tx = dmu_tx_create(os); | |
649 | dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END); | |
0037b49e | 650 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 651 | if (xzp) { |
0037b49e | 652 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL); |
428870ff | 653 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); |
34dc7c2f BB |
654 | } |
655 | if (acl_obj) | |
656 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
428870ff BB |
657 | |
658 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
659 | error = dmu_tx_assign(tx, TXG_WAIT); |
660 | if (error) { | |
661 | /* | |
662 | * Not enough space to delete the file. Leave it in the | |
663 | * unlinked set, leaking it until the fs is remounted (at | |
664 | * which point we'll call zfs_unlinked_drain() to process it). | |
665 | */ | |
666 | dmu_tx_abort(tx); | |
667 | zfs_znode_dmu_fini(zp); | |
34dc7c2f BB |
668 | goto out; |
669 | } | |
670 | ||
671 | if (xzp) { | |
428870ff | 672 | ASSERT(error == 0); |
34dc7c2f BB |
673 | mutex_enter(&xzp->z_lock); |
674 | xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */ | |
dfbc8630 CD |
675 | clear_nlink(ZTOI(xzp)); /* no more links to it */ |
676 | links = 0; | |
0037b49e | 677 | VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 678 | &links, sizeof (links), tx)); |
34dc7c2f BB |
679 | mutex_exit(&xzp->z_lock); |
680 | zfs_unlinked_add(xzp, tx); | |
681 | } | |
682 | ||
683 | /* Remove this znode from the unlinked set */ | |
b128c09f | 684 | VERIFY3U(0, ==, |
0037b49e | 685 | zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); |
34dc7c2f BB |
686 | |
687 | zfs_znode_delete(zp, tx); | |
688 | ||
689 | dmu_tx_commit(tx); | |
690 | out: | |
691 | if (xzp) | |
0a50679c | 692 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
693 | } |
694 | ||
695 | static uint64_t | |
428870ff | 696 | zfs_dirent(znode_t *zp, uint64_t mode) |
34dc7c2f BB |
697 | { |
698 | uint64_t de = zp->z_id; | |
428870ff | 699 | |
3558fd73 | 700 | if (ZTOZSB(zp)->z_version >= ZPL_VERSION_DIRENT_TYPE) |
428870ff | 701 | de |= IFTODT(mode) << 60; |
34dc7c2f BB |
702 | return (de); |
703 | } | |
704 | ||
705 | /* | |
599b8648 CC |
706 | * Link zp into dl. Can fail in the following cases : |
707 | * - if zp has been unlinked. | |
708 | * - if the number of entries with the same hash (aka. colliding entries) | |
709 | * exceed the capacity of a leaf-block of fatzap and splitting of the | |
710 | * leaf-block does not help. | |
34dc7c2f BB |
711 | */ |
712 | int | |
713 | zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag) | |
714 | { | |
715 | znode_t *dzp = dl->dl_dzp; | |
0037b49e | 716 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f | 717 | uint64_t value; |
3558fd73 | 718 | int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); |
428870ff BB |
719 | sa_bulk_attr_t bulk[5]; |
720 | uint64_t mtime[2], ctime[2]; | |
dfbc8630 | 721 | uint64_t links; |
428870ff | 722 | int count = 0; |
34dc7c2f BB |
723 | int error; |
724 | ||
34dc7c2f BB |
725 | mutex_enter(&zp->z_lock); |
726 | ||
727 | if (!(flag & ZRENAMING)) { | |
728 | if (zp->z_unlinked) { /* no new links to unlinked zp */ | |
729 | ASSERT(!(flag & (ZNEW | ZEXISTS))); | |
730 | mutex_exit(&zp->z_lock); | |
2e528b49 | 731 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 732 | } |
dfbc8630 CD |
733 | if (!(flag & ZNEW)) { |
734 | /* | |
735 | * ZNEW nodes come from zfs_mknode() where the link | |
736 | * count has already been initialised | |
737 | */ | |
738 | inc_nlink(ZTOI(zp)); | |
739 | links = ZTOI(zp)->i_nlink; | |
0037b49e BB |
740 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), |
741 | NULL, &links, sizeof (links)); | |
dfbc8630 | 742 | } |
34dc7c2f | 743 | } |
599b8648 CC |
744 | |
745 | value = zfs_dirent(zp, zp->z_mode); | |
746 | error = zap_add(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, 8, 1, | |
747 | &value, tx); | |
748 | ||
749 | /* | |
750 | * zap_add could fail to add the entry if it exceeds the capacity of the | |
751 | * leaf-block and zap_leaf_split() failed to help. | |
752 | * The caller of this routine is responsible for failing the transaction | |
753 | * which will rollback the SA updates done above. | |
754 | */ | |
755 | if (error != 0) { | |
756 | if (!(flag & ZRENAMING) && !(flag & ZNEW)) | |
757 | drop_nlink(ZTOI(zp)); | |
758 | mutex_exit(&zp->z_lock); | |
759 | return (error); | |
760 | } | |
761 | ||
0037b49e | 762 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, |
428870ff | 763 | &dzp->z_id, sizeof (dzp->z_id)); |
0037b49e | 764 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
765 | &zp->z_pflags, sizeof (zp->z_pflags)); |
766 | ||
767 | if (!(flag & ZNEW)) { | |
0037b49e | 768 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
428870ff BB |
769 | ctime, sizeof (ctime)); |
770 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, | |
0df9673f | 771 | ctime); |
428870ff BB |
772 | } |
773 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
774 | ASSERT(error == 0); | |
34dc7c2f | 775 | |
34dc7c2f BB |
776 | mutex_exit(&zp->z_lock); |
777 | ||
34dc7c2f | 778 | mutex_enter(&dzp->z_lock); |
428870ff | 779 | dzp->z_size++; |
dfbc8630 CD |
780 | if (zp_is_dir) |
781 | inc_nlink(ZTOI(dzp)); | |
782 | links = ZTOI(dzp)->i_nlink; | |
428870ff | 783 | count = 0; |
0037b49e | 784 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, |
428870ff | 785 | &dzp->z_size, sizeof (dzp->z_size)); |
0037b49e | 786 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, |
dfbc8630 | 787 | &links, sizeof (links)); |
0037b49e | 788 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 789 | mtime, sizeof (mtime)); |
0037b49e | 790 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
428870ff | 791 | ctime, sizeof (ctime)); |
0037b49e | 792 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff | 793 | &dzp->z_pflags, sizeof (dzp->z_pflags)); |
0df9673f | 794 | zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime); |
428870ff BB |
795 | error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); |
796 | ASSERT(error == 0); | |
34dc7c2f BB |
797 | mutex_exit(&dzp->z_lock); |
798 | ||
34dc7c2f BB |
799 | return (0); |
800 | } | |
801 | ||
9b7b9cd3 GM |
802 | /* |
803 | * The match type in the code for this function should conform to: | |
804 | * | |
805 | * ------------------------------------------------------------------------ | |
806 | * fs type | z_norm | lookup type | match type | |
807 | * ---------|-------------|-------------|---------------------------------- | |
808 | * CS !norm | 0 | 0 | 0 (exact) | |
809 | * CS norm | formX | 0 | MT_NORMALIZE | |
810 | * CI !norm | upper | !ZCIEXACT | MT_NORMALIZE | |
811 | * CI !norm | upper | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE | |
812 | * CI norm | upper|formX | !ZCIEXACT | MT_NORMALIZE | |
813 | * CI norm | upper|formX | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE | |
814 | * CM !norm | upper | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE | |
815 | * CM !norm | upper | ZCILOOK | MT_NORMALIZE | |
816 | * CM norm | upper|formX | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE | |
817 | * CM norm | upper|formX | ZCILOOK | MT_NORMALIZE | |
818 | * | |
819 | * Abbreviations: | |
820 | * CS = Case Sensitive, CI = Case Insensitive, CM = Case Mixed | |
821 | * upper = case folding set by fs type on creation (U8_TEXTPREP_TOUPPER) | |
822 | * formX = unicode normalization form set on fs creation | |
823 | */ | |
428870ff BB |
824 | static int |
825 | zfs_dropname(zfs_dirlock_t *dl, znode_t *zp, znode_t *dzp, dmu_tx_t *tx, | |
826 | int flag) | |
827 | { | |
828 | int error; | |
829 | ||
3558fd73 | 830 | if (ZTOZSB(zp)->z_norm) { |
9b7b9cd3 GM |
831 | matchtype_t mt = MT_NORMALIZE; |
832 | ||
833 | if ((ZTOZSB(zp)->z_case == ZFS_CASE_INSENSITIVE && | |
428870ff | 834 | (flag & ZCIEXACT)) || |
9b7b9cd3 GM |
835 | (ZTOZSB(zp)->z_case == ZFS_CASE_MIXED && |
836 | !(flag & ZCILOOK))) { | |
837 | mt |= MT_MATCH_CASE; | |
838 | } | |
839 | ||
840 | error = zap_remove_norm(ZTOZSB(zp)->z_os, dzp->z_id, | |
841 | dl->dl_name, mt, tx); | |
428870ff | 842 | } else { |
9b7b9cd3 GM |
843 | error = zap_remove(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, |
844 | tx); | |
428870ff BB |
845 | } |
846 | ||
847 | return (error); | |
848 | } | |
849 | ||
34dc7c2f | 850 | /* |
cd38ac58 BB |
851 | * Unlink zp from dl, and mark zp for deletion if this was the last link. Can |
852 | * fail if zp is a mount point (EBUSY) or a non-empty directory (ENOTEMPTY). | |
34dc7c2f BB |
853 | * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list. |
854 | * If it's non-NULL, we use it to indicate whether the znode needs deletion, | |
855 | * and it's the caller's job to do it. | |
856 | */ | |
857 | int | |
858 | zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag, | |
e9aa730c | 859 | boolean_t *unlinkedp) |
34dc7c2f BB |
860 | { |
861 | znode_t *dzp = dl->dl_dzp; | |
0037b49e | 862 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
3558fd73 | 863 | int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); |
34dc7c2f | 864 | boolean_t unlinked = B_FALSE; |
428870ff BB |
865 | sa_bulk_attr_t bulk[5]; |
866 | uint64_t mtime[2], ctime[2]; | |
dfbc8630 | 867 | uint64_t links; |
428870ff | 868 | int count = 0; |
34dc7c2f BB |
869 | int error; |
870 | ||
34dc7c2f | 871 | if (!(flag & ZRENAMING)) { |
34dc7c2f | 872 | mutex_enter(&zp->z_lock); |
428870ff BB |
873 | |
874 | if (zp_is_dir && !zfs_dirempty(zp)) { | |
34dc7c2f | 875 | mutex_exit(&zp->z_lock); |
2e528b49 | 876 | return (SET_ERROR(ENOTEMPTY)); |
34dc7c2f | 877 | } |
428870ff BB |
878 | |
879 | /* | |
880 | * If we get here, we are going to try to remove the object. | |
881 | * First try removing the name from the directory; if that | |
882 | * fails, return the error. | |
883 | */ | |
884 | error = zfs_dropname(dl, zp, dzp, tx, flag); | |
885 | if (error != 0) { | |
886 | mutex_exit(&zp->z_lock); | |
428870ff BB |
887 | return (error); |
888 | } | |
889 | ||
dfbc8630 | 890 | if (ZTOI(zp)->i_nlink <= zp_is_dir) { |
3558fd73 BB |
891 | zfs_panic_recover("zfs: link count on %lu is %u, " |
892 | "should be at least %u", zp->z_id, | |
dfbc8630 CD |
893 | (int)ZTOI(zp)->i_nlink, zp_is_dir + 1); |
894 | set_nlink(ZTOI(zp), zp_is_dir + 1); | |
34dc7c2f | 895 | } |
dfbc8630 CD |
896 | drop_nlink(ZTOI(zp)); |
897 | if (ZTOI(zp)->i_nlink == zp_is_dir) { | |
34dc7c2f | 898 | zp->z_unlinked = B_TRUE; |
dfbc8630 | 899 | clear_nlink(ZTOI(zp)); |
34dc7c2f BB |
900 | unlinked = B_TRUE; |
901 | } else { | |
0037b49e | 902 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), |
428870ff | 903 | NULL, &ctime, sizeof (ctime)); |
0037b49e | 904 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 905 | NULL, &zp->z_pflags, sizeof (zp->z_pflags)); |
0df9673f CC |
906 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, |
907 | ctime); | |
34dc7c2f | 908 | } |
dfbc8630 | 909 | links = ZTOI(zp)->i_nlink; |
0037b49e | 910 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 911 | NULL, &links, sizeof (links)); |
428870ff BB |
912 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
913 | count = 0; | |
914 | ASSERT(error == 0); | |
34dc7c2f | 915 | mutex_exit(&zp->z_lock); |
428870ff BB |
916 | } else { |
917 | error = zfs_dropname(dl, zp, dzp, tx, flag); | |
918 | if (error != 0) | |
919 | return (error); | |
34dc7c2f BB |
920 | } |
921 | ||
34dc7c2f | 922 | mutex_enter(&dzp->z_lock); |
428870ff | 923 | dzp->z_size--; /* one dirent removed */ |
dfbc8630 CD |
924 | if (zp_is_dir) |
925 | drop_nlink(ZTOI(dzp)); /* ".." link from zp */ | |
926 | links = ZTOI(dzp)->i_nlink; | |
0037b49e | 927 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 928 | NULL, &links, sizeof (links)); |
0037b49e | 929 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), |
428870ff | 930 | NULL, &dzp->z_size, sizeof (dzp->z_size)); |
0037b49e | 931 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), |
428870ff | 932 | NULL, ctime, sizeof (ctime)); |
0037b49e | 933 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), |
428870ff | 934 | NULL, mtime, sizeof (mtime)); |
0037b49e | 935 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 936 | NULL, &dzp->z_pflags, sizeof (dzp->z_pflags)); |
0df9673f | 937 | zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime); |
428870ff | 938 | error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); |
34dc7c2f | 939 | ASSERT(error == 0); |
428870ff | 940 | mutex_exit(&dzp->z_lock); |
34dc7c2f BB |
941 | |
942 | if (unlinkedp != NULL) | |
943 | *unlinkedp = unlinked; | |
944 | else if (unlinked) | |
945 | zfs_unlinked_add(zp, tx); | |
946 | ||
947 | return (0); | |
948 | } | |
949 | ||
950 | /* | |
951 | * Indicate whether the directory is empty. Works with or without z_lock | |
952 | * held, but can only be consider a hint in the latter case. Returns true | |
953 | * if only "." and ".." remain and there's no work in progress. | |
3910184d AZ |
954 | * |
955 | * The internal ZAP size, rather than zp->z_size, needs to be checked since | |
956 | * some consumers (Lustre) do not strictly maintain an accurate SA_ZPL_SIZE. | |
34dc7c2f BB |
957 | */ |
958 | boolean_t | |
959 | zfs_dirempty(znode_t *dzp) | |
960 | { | |
3910184d AZ |
961 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
962 | uint64_t count; | |
963 | int error; | |
964 | ||
965 | if (dzp->z_dirlocks != NULL) | |
966 | return (B_FALSE); | |
967 | ||
968 | error = zap_count(zfsvfs->z_os, dzp->z_id, &count); | |
969 | if (error != 0 || count != 0) | |
970 | return (B_FALSE); | |
971 | ||
972 | return (B_TRUE); | |
34dc7c2f BB |
973 | } |
974 | ||
975 | int | |
3558fd73 | 976 | zfs_make_xattrdir(znode_t *zp, vattr_t *vap, struct inode **xipp, cred_t *cr) |
34dc7c2f | 977 | { |
0037b49e | 978 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
979 | znode_t *xzp; |
980 | dmu_tx_t *tx; | |
981 | int error; | |
9babb374 BB |
982 | zfs_acl_ids_t acl_ids; |
983 | boolean_t fuid_dirtied; | |
3558fd73 | 984 | #ifdef DEBUG |
428870ff | 985 | uint64_t parent; |
3558fd73 | 986 | #endif |
34dc7c2f | 987 | |
3558fd73 | 988 | *xipp = NULL; |
34dc7c2f | 989 | |
3558fd73 | 990 | if ((error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))) |
34dc7c2f BB |
991 | return (error); |
992 | ||
9babb374 BB |
993 | if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL, |
994 | &acl_ids)) != 0) | |
995 | return (error); | |
9c5167d1 | 996 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zp->z_projid)) { |
9babb374 | 997 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 998 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
999 | } |
1000 | ||
0037b49e | 1001 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1002 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
1003 | ZFS_SA_BASE_ATTR_SIZE); | |
1004 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
34dc7c2f | 1005 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); |
0037b49e | 1006 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1007 | if (fuid_dirtied) |
0037b49e | 1008 | zfs_fuid_txhold(zfsvfs, tx); |
384f8a09 | 1009 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 1010 | if (error) { |
428870ff | 1011 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
1012 | dmu_tx_abort(tx); |
1013 | return (error); | |
1014 | } | |
428870ff | 1015 | zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids); |
9babb374 BB |
1016 | |
1017 | if (fuid_dirtied) | |
0037b49e | 1018 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 1019 | |
428870ff | 1020 | #ifdef DEBUG |
0037b49e | 1021 | error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff BB |
1022 | &parent, sizeof (parent)); |
1023 | ASSERT(error == 0 && parent == zp->z_id); | |
1024 | #endif | |
1025 | ||
0037b49e | 1026 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id, |
428870ff | 1027 | sizeof (xzp->z_id), tx)); |
34dc7c2f | 1028 | |
98701490 | 1029 | if (!zp->z_unlinked) |
0037b49e | 1030 | (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp, |
98701490 | 1031 | xzp, "", NULL, acl_ids.z_fuidp, vap); |
9babb374 BB |
1032 | |
1033 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1034 | dmu_tx_commit(tx); |
1035 | ||
3558fd73 | 1036 | *xipp = ZTOI(xzp); |
34dc7c2f BB |
1037 | |
1038 | return (0); | |
1039 | } | |
1040 | ||
1041 | /* | |
1042 | * Return a znode for the extended attribute directory for zp. | |
1043 | * ** If the directory does not already exist, it is created ** | |
1044 | * | |
1045 | * IN: zp - znode to obtain attribute directory from | |
1046 | * cr - credentials of caller | |
1047 | * flags - flags from the VOP_LOOKUP call | |
1048 | * | |
3558fd73 | 1049 | * OUT: xipp - pointer to extended attribute znode |
34dc7c2f BB |
1050 | * |
1051 | * RETURN: 0 on success | |
1052 | * error number on failure | |
1053 | */ | |
1054 | int | |
3558fd73 | 1055 | zfs_get_xattrdir(znode_t *zp, struct inode **xipp, cred_t *cr, int flags) |
34dc7c2f | 1056 | { |
0037b49e | 1057 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
1058 | znode_t *xzp; |
1059 | zfs_dirlock_t *dl; | |
1060 | vattr_t va; | |
1061 | int error; | |
1062 | top: | |
1063 | error = zfs_dirent_lock(&dl, zp, "", &xzp, ZXATTR, NULL, NULL); | |
1064 | if (error) | |
1065 | return (error); | |
1066 | ||
1067 | if (xzp != NULL) { | |
3558fd73 | 1068 | *xipp = ZTOI(xzp); |
34dc7c2f BB |
1069 | zfs_dirent_unlock(dl); |
1070 | return (0); | |
1071 | } | |
1072 | ||
34dc7c2f BB |
1073 | if (!(flags & CREATE_XATTR_DIR)) { |
1074 | zfs_dirent_unlock(dl); | |
2e528b49 | 1075 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1076 | } |
1077 | ||
0037b49e | 1078 | if (zfs_is_readonly(zfsvfs)) { |
34dc7c2f | 1079 | zfs_dirent_unlock(dl); |
2e528b49 | 1080 | return (SET_ERROR(EROFS)); |
34dc7c2f BB |
1081 | } |
1082 | ||
1083 | /* | |
1084 | * The ability to 'create' files in an attribute | |
1085 | * directory comes from the write_xattr permission on the base file. | |
1086 | * | |
1087 | * The ability to 'search' an attribute directory requires | |
1088 | * read_xattr permission on the base file. | |
1089 | * | |
1090 | * Once in a directory the ability to read/write attributes | |
1091 | * is controlled by the permissions on the attribute file. | |
1092 | */ | |
3558fd73 | 1093 | va.va_mask = ATTR_MODE | ATTR_UID | ATTR_GID; |
34dc7c2f BB |
1094 | va.va_mode = S_IFDIR | S_ISVTX | 0777; |
1095 | zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid); | |
1096 | ||
274b7e79 | 1097 | va.va_dentry = NULL; |
3558fd73 | 1098 | error = zfs_make_xattrdir(zp, &va, xipp, cr); |
34dc7c2f BB |
1099 | zfs_dirent_unlock(dl); |
1100 | ||
fb5f0bc8 | 1101 | if (error == ERESTART) { |
34dc7c2f BB |
1102 | /* NB: we already did dmu_tx_wait() if necessary */ |
1103 | goto top; | |
1104 | } | |
1105 | ||
1106 | return (error); | |
1107 | } | |
1108 | ||
1109 | /* | |
1110 | * Decide whether it is okay to remove within a sticky directory. | |
1111 | * | |
1112 | * In sticky directories, write access is not sufficient; | |
1113 | * you can remove entries from a directory only if: | |
1114 | * | |
1115 | * you own the directory, | |
1116 | * you own the entry, | |
7b340700 | 1117 | * you have write access to the entry, |
34dc7c2f BB |
1118 | * or you are privileged (checked in secpolicy...). |
1119 | * | |
1120 | * The function returns 0 if remove access is granted. | |
1121 | */ | |
1122 | int | |
1123 | zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr) | |
1124 | { | |
3558fd73 | 1125 | uid_t uid; |
572e2857 BB |
1126 | uid_t downer; |
1127 | uid_t fowner; | |
0037b49e | 1128 | zfsvfs_t *zfsvfs = ZTOZSB(zdp); |
34dc7c2f | 1129 | |
0037b49e | 1130 | if (zfsvfs->z_replay) |
34dc7c2f BB |
1131 | return (0); |
1132 | ||
428870ff | 1133 | if ((zdp->z_mode & S_ISVTX) == 0) |
34dc7c2f BB |
1134 | return (0); |
1135 | ||
0037b49e | 1136 | downer = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(ZTOI(zdp)->i_uid), |
2c6abf15 | 1137 | cr, ZFS_OWNER); |
0037b49e | 1138 | fowner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(ZTOI(zp)->i_uid), |
2c6abf15 | 1139 | cr, ZFS_OWNER); |
572e2857 BB |
1140 | |
1141 | if ((uid = crgetuid(cr)) == downer || uid == fowner || | |
7b340700 | 1142 | zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0) |
34dc7c2f BB |
1143 | return (0); |
1144 | else | |
1145 | return (secpolicy_vnode_remove(cr)); | |
1146 | } |