]>
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)); |
dcec0a12 AP |
461 | |
462 | dataset_kstats_update_nunlinks_kstat(&zfsvfs->z_kstat, 1); | |
34dc7c2f BB |
463 | } |
464 | ||
0d5c500d BB |
465 | /* |
466 | * Clean up any znodes that had no links when we either crashed or | |
467 | * (force) umounted the file system. | |
468 | */ | |
dcec0a12 AP |
469 | static void |
470 | zfs_unlinked_drain_task(void *arg) | |
0d5c500d | 471 | { |
dcec0a12 | 472 | zfsvfs_t *zfsvfs = arg; |
0d5c500d BB |
473 | zap_cursor_t zc; |
474 | zap_attribute_t zap; | |
475 | dmu_object_info_t doi; | |
476 | znode_t *zp; | |
477 | int error; | |
478 | ||
dcec0a12 AP |
479 | ASSERT3B(zfsvfs->z_draining, ==, B_TRUE); |
480 | ||
0d5c500d BB |
481 | /* |
482 | * Iterate over the contents of the unlinked set. | |
483 | */ | |
0037b49e | 484 | for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj); |
dcec0a12 | 485 | zap_cursor_retrieve(&zc, &zap) == 0 && !zfsvfs->z_drain_cancel; |
0d5c500d BB |
486 | zap_cursor_advance(&zc)) { |
487 | ||
488 | /* | |
489 | * See what kind of object we have in list | |
490 | */ | |
491 | ||
0037b49e BB |
492 | error = dmu_object_info(zfsvfs->z_os, |
493 | zap.za_first_integer, &doi); | |
0d5c500d BB |
494 | if (error != 0) |
495 | continue; | |
496 | ||
497 | ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) || | |
498 | (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS)); | |
499 | /* | |
500 | * We need to re-mark these list entries for deletion, | |
501 | * so we pull them back into core and set zp->z_unlinked. | |
502 | */ | |
0037b49e | 503 | error = zfs_zget(zfsvfs, zap.za_first_integer, &zp); |
0d5c500d BB |
504 | |
505 | /* | |
506 | * We may pick up znodes that are already marked for deletion. | |
507 | * This could happen during the purge of an extended attribute | |
508 | * directory. All we need to do is skip over them, since they | |
509 | * are already in the system marked z_unlinked. | |
510 | */ | |
511 | if (error != 0) | |
512 | continue; | |
513 | ||
514 | zp->z_unlinked = B_TRUE; | |
dcec0a12 AP |
515 | |
516 | /* | |
517 | * iput() is Linux's equivalent to illumos' VN_RELE(). It will | |
518 | * decrement the inode's ref count and may cause the inode to be | |
519 | * synchronously freed. We interrupt freeing of this inode, by | |
520 | * checking the return value of dmu_objset_zfs_unmounting() in | |
521 | * dmu_free_long_range(), when an unmount is requested. | |
522 | */ | |
0d5c500d | 523 | iput(ZTOI(zp)); |
dcec0a12 | 524 | ASSERT3B(zfsvfs->z_unmounted, ==, B_FALSE); |
0d5c500d BB |
525 | } |
526 | zap_cursor_fini(&zc); | |
dcec0a12 AP |
527 | |
528 | zfsvfs->z_draining = B_FALSE; | |
529 | zfsvfs->z_drain_task = TASKQID_INVALID; | |
530 | } | |
531 | ||
532 | /* | |
533 | * Sets z_draining then tries to dispatch async unlinked drain. | |
534 | * If that fails executes synchronous unlinked drain. | |
535 | */ | |
536 | void | |
537 | zfs_unlinked_drain(zfsvfs_t *zfsvfs) | |
538 | { | |
539 | ASSERT3B(zfsvfs->z_unmounted, ==, B_FALSE); | |
540 | ASSERT3B(zfsvfs->z_draining, ==, B_FALSE); | |
541 | ||
542 | zfsvfs->z_draining = B_TRUE; | |
543 | zfsvfs->z_drain_cancel = B_FALSE; | |
544 | ||
545 | zfsvfs->z_drain_task = taskq_dispatch( | |
546 | dsl_pool_unlinked_drain_taskq(dmu_objset_pool(zfsvfs->z_os)), | |
547 | zfs_unlinked_drain_task, zfsvfs, TQ_SLEEP); | |
548 | if (zfsvfs->z_drain_task == TASKQID_INVALID) { | |
549 | zfs_dbgmsg("async zfs_unlinked_drain dispatch failed"); | |
550 | zfs_unlinked_drain_task(zfsvfs); | |
551 | } | |
552 | } | |
553 | ||
554 | /* | |
555 | * Wait for the unlinked drain taskq task to stop. This will interrupt the | |
556 | * unlinked set processing if it is in progress. | |
557 | */ | |
558 | void | |
559 | zfs_unlinked_drain_stop_wait(zfsvfs_t *zfsvfs) | |
560 | { | |
561 | ASSERT3B(zfsvfs->z_unmounted, ==, B_FALSE); | |
562 | ||
563 | if (zfsvfs->z_draining) { | |
564 | zfsvfs->z_drain_cancel = B_TRUE; | |
565 | taskq_cancel_id(dsl_pool_unlinked_drain_taskq( | |
566 | dmu_objset_pool(zfsvfs->z_os)), zfsvfs->z_drain_task); | |
567 | zfsvfs->z_drain_task = TASKQID_INVALID; | |
568 | zfsvfs->z_draining = B_FALSE; | |
569 | } | |
0d5c500d BB |
570 | } |
571 | ||
34dc7c2f BB |
572 | /* |
573 | * Delete the entire contents of a directory. Return a count | |
574 | * of the number of entries that could not be deleted. If we encounter | |
575 | * an error, return a count of at least one so that the directory stays | |
576 | * in the unlinked set. | |
577 | * | |
578 | * NOTE: this function assumes that the directory is inactive, | |
579 | * so there is no need to lock its entries before deletion. | |
580 | * Also, it assumes the directory contents is *only* regular | |
581 | * files. | |
582 | */ | |
583 | static int | |
584 | zfs_purgedir(znode_t *dzp) | |
585 | { | |
586 | zap_cursor_t zc; | |
587 | zap_attribute_t zap; | |
588 | znode_t *xzp; | |
589 | dmu_tx_t *tx; | |
0037b49e | 590 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f BB |
591 | zfs_dirlock_t dl; |
592 | int skipped = 0; | |
593 | int error; | |
594 | ||
0037b49e | 595 | for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); |
34dc7c2f BB |
596 | (error = zap_cursor_retrieve(&zc, &zap)) == 0; |
597 | zap_cursor_advance(&zc)) { | |
0037b49e | 598 | error = zfs_zget(zfsvfs, |
34dc7c2f BB |
599 | ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp); |
600 | if (error) { | |
601 | skipped += 1; | |
602 | continue; | |
603 | } | |
604 | ||
0d5c500d BB |
605 | ASSERT(S_ISREG(ZTOI(xzp)->i_mode) || |
606 | S_ISLNK(ZTOI(xzp)->i_mode)); | |
34dc7c2f | 607 | |
0037b49e | 608 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 609 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
34dc7c2f | 610 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name); |
428870ff | 611 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); |
0037b49e | 612 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
613 | /* Is this really needed ? */ |
614 | zfs_sa_upgrade_txholds(tx, xzp); | |
19d55079 | 615 | dmu_tx_mark_netfree(tx); |
34dc7c2f BB |
616 | error = dmu_tx_assign(tx, TXG_WAIT); |
617 | if (error) { | |
618 | dmu_tx_abort(tx); | |
0a50679c | 619 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
620 | skipped += 1; |
621 | continue; | |
622 | } | |
623 | bzero(&dl, sizeof (dl)); | |
624 | dl.dl_dzp = dzp; | |
625 | dl.dl_name = zap.za_name; | |
626 | ||
627 | error = zfs_link_destroy(&dl, xzp, tx, 0, NULL); | |
628 | if (error) | |
629 | skipped += 1; | |
630 | dmu_tx_commit(tx); | |
dfbc8630 | 631 | |
0a50679c | 632 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
633 | } |
634 | zap_cursor_fini(&zc); | |
635 | if (error != ENOENT) | |
636 | skipped += 1; | |
637 | return (skipped); | |
638 | } | |
639 | ||
640 | void | |
641 | zfs_rmnode(znode_t *zp) | |
642 | { | |
0037b49e BB |
643 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
644 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 645 | znode_t *xzp = NULL; |
34dc7c2f BB |
646 | dmu_tx_t *tx; |
647 | uint64_t acl_obj; | |
428870ff | 648 | uint64_t xattr_obj; |
dfbc8630 | 649 | uint64_t links; |
34dc7c2f BB |
650 | int error; |
651 | ||
dfbc8630 | 652 | ASSERT(ZTOI(zp)->i_nlink == 0); |
3558fd73 | 653 | ASSERT(atomic_read(&ZTOI(zp)->i_count) == 0); |
34dc7c2f BB |
654 | |
655 | /* | |
656 | * If this is an attribute directory, purge its contents. | |
657 | */ | |
3558fd73 | 658 | if (S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_XATTR)) { |
0d5c500d | 659 | if (zfs_purgedir(zp) != 0) { |
34dc7c2f | 660 | /* |
0d5c500d BB |
661 | * Not enough space to delete some xattrs. |
662 | * Leave it in the unlinked set. | |
34dc7c2f | 663 | */ |
7973e464 | 664 | zfs_znode_dmu_fini(zp); |
0d5c500d | 665 | |
34dc7c2f BB |
666 | return; |
667 | } | |
668 | } | |
669 | ||
b128c09f | 670 | /* |
f5f087eb CC |
671 | * Free up all the data in the file. We don't do this for directories |
672 | * because we need truncate and remove to be in the same tx, like in | |
673 | * zfs_znode_delete(). Otherwise, if we crash here we'll end up with | |
674 | * an inconsistent truncated zap object in the delete queue. Note a | |
675 | * truncated file is harmless since it only contains user data. | |
b128c09f | 676 | */ |
f5f087eb CC |
677 | if (S_ISREG(ZTOI(zp)->i_mode)) { |
678 | error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END); | |
679 | if (error) { | |
680 | /* | |
a08abc1b GM |
681 | * Not enough space or we were interrupted by unmount. |
682 | * Leave the file in the unlinked set. | |
f5f087eb CC |
683 | */ |
684 | zfs_znode_dmu_fini(zp); | |
685 | return; | |
686 | } | |
b128c09f BB |
687 | } |
688 | ||
34dc7c2f BB |
689 | /* |
690 | * If the file has extended attributes, we're going to unlink | |
691 | * the xattr dir. | |
692 | */ | |
0037b49e | 693 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff BB |
694 | &xattr_obj, sizeof (xattr_obj)); |
695 | if (error == 0 && xattr_obj) { | |
0037b49e | 696 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
34dc7c2f BB |
697 | ASSERT(error == 0); |
698 | } | |
699 | ||
572e2857 | 700 | acl_obj = zfs_external_acl(zp); |
34dc7c2f BB |
701 | |
702 | /* | |
b128c09f | 703 | * Set up the final transaction. |
34dc7c2f BB |
704 | */ |
705 | tx = dmu_tx_create(os); | |
706 | dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END); | |
0037b49e | 707 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 708 | if (xzp) { |
0037b49e | 709 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL); |
428870ff | 710 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); |
34dc7c2f BB |
711 | } |
712 | if (acl_obj) | |
713 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
428870ff BB |
714 | |
715 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
716 | error = dmu_tx_assign(tx, TXG_WAIT); |
717 | if (error) { | |
718 | /* | |
719 | * Not enough space to delete the file. Leave it in the | |
720 | * unlinked set, leaking it until the fs is remounted (at | |
721 | * which point we'll call zfs_unlinked_drain() to process it). | |
722 | */ | |
723 | dmu_tx_abort(tx); | |
724 | zfs_znode_dmu_fini(zp); | |
34dc7c2f BB |
725 | goto out; |
726 | } | |
727 | ||
728 | if (xzp) { | |
428870ff | 729 | ASSERT(error == 0); |
34dc7c2f BB |
730 | mutex_enter(&xzp->z_lock); |
731 | xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */ | |
dfbc8630 CD |
732 | clear_nlink(ZTOI(xzp)); /* no more links to it */ |
733 | links = 0; | |
0037b49e | 734 | VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 735 | &links, sizeof (links), tx)); |
34dc7c2f BB |
736 | mutex_exit(&xzp->z_lock); |
737 | zfs_unlinked_add(xzp, tx); | |
738 | } | |
739 | ||
740 | /* Remove this znode from the unlinked set */ | |
b128c09f | 741 | VERIFY3U(0, ==, |
0037b49e | 742 | zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); |
34dc7c2f | 743 | |
dcec0a12 AP |
744 | dataset_kstats_update_nunlinked_kstat(&zfsvfs->z_kstat, 1); |
745 | ||
34dc7c2f BB |
746 | zfs_znode_delete(zp, tx); |
747 | ||
748 | dmu_tx_commit(tx); | |
749 | out: | |
750 | if (xzp) | |
0a50679c | 751 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
752 | } |
753 | ||
754 | static uint64_t | |
428870ff | 755 | zfs_dirent(znode_t *zp, uint64_t mode) |
34dc7c2f BB |
756 | { |
757 | uint64_t de = zp->z_id; | |
428870ff | 758 | |
3558fd73 | 759 | if (ZTOZSB(zp)->z_version >= ZPL_VERSION_DIRENT_TYPE) |
428870ff | 760 | de |= IFTODT(mode) << 60; |
34dc7c2f BB |
761 | return (de); |
762 | } | |
763 | ||
764 | /* | |
599b8648 CC |
765 | * Link zp into dl. Can fail in the following cases : |
766 | * - if zp has been unlinked. | |
767 | * - if the number of entries with the same hash (aka. colliding entries) | |
768 | * exceed the capacity of a leaf-block of fatzap and splitting of the | |
769 | * leaf-block does not help. | |
34dc7c2f BB |
770 | */ |
771 | int | |
772 | zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag) | |
773 | { | |
774 | znode_t *dzp = dl->dl_dzp; | |
0037b49e | 775 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f | 776 | uint64_t value; |
3558fd73 | 777 | int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); |
428870ff BB |
778 | sa_bulk_attr_t bulk[5]; |
779 | uint64_t mtime[2], ctime[2]; | |
dfbc8630 | 780 | uint64_t links; |
428870ff | 781 | int count = 0; |
34dc7c2f BB |
782 | int error; |
783 | ||
34dc7c2f BB |
784 | mutex_enter(&zp->z_lock); |
785 | ||
786 | if (!(flag & ZRENAMING)) { | |
787 | if (zp->z_unlinked) { /* no new links to unlinked zp */ | |
788 | ASSERT(!(flag & (ZNEW | ZEXISTS))); | |
789 | mutex_exit(&zp->z_lock); | |
2e528b49 | 790 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 791 | } |
dfbc8630 CD |
792 | if (!(flag & ZNEW)) { |
793 | /* | |
794 | * ZNEW nodes come from zfs_mknode() where the link | |
795 | * count has already been initialised | |
796 | */ | |
797 | inc_nlink(ZTOI(zp)); | |
798 | links = ZTOI(zp)->i_nlink; | |
0037b49e BB |
799 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), |
800 | NULL, &links, sizeof (links)); | |
dfbc8630 | 801 | } |
34dc7c2f | 802 | } |
599b8648 CC |
803 | |
804 | value = zfs_dirent(zp, zp->z_mode); | |
805 | error = zap_add(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, 8, 1, | |
806 | &value, tx); | |
807 | ||
808 | /* | |
809 | * zap_add could fail to add the entry if it exceeds the capacity of the | |
810 | * leaf-block and zap_leaf_split() failed to help. | |
811 | * The caller of this routine is responsible for failing the transaction | |
812 | * which will rollback the SA updates done above. | |
813 | */ | |
814 | if (error != 0) { | |
815 | if (!(flag & ZRENAMING) && !(flag & ZNEW)) | |
816 | drop_nlink(ZTOI(zp)); | |
817 | mutex_exit(&zp->z_lock); | |
818 | return (error); | |
819 | } | |
820 | ||
0037b49e | 821 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, |
428870ff | 822 | &dzp->z_id, sizeof (dzp->z_id)); |
0037b49e | 823 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
824 | &zp->z_pflags, sizeof (zp->z_pflags)); |
825 | ||
826 | if (!(flag & ZNEW)) { | |
0037b49e | 827 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
428870ff BB |
828 | ctime, sizeof (ctime)); |
829 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, | |
0df9673f | 830 | ctime); |
428870ff BB |
831 | } |
832 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
833 | ASSERT(error == 0); | |
34dc7c2f | 834 | |
34dc7c2f BB |
835 | mutex_exit(&zp->z_lock); |
836 | ||
34dc7c2f | 837 | mutex_enter(&dzp->z_lock); |
428870ff | 838 | dzp->z_size++; |
dfbc8630 CD |
839 | if (zp_is_dir) |
840 | inc_nlink(ZTOI(dzp)); | |
841 | links = ZTOI(dzp)->i_nlink; | |
428870ff | 842 | count = 0; |
0037b49e | 843 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, |
428870ff | 844 | &dzp->z_size, sizeof (dzp->z_size)); |
0037b49e | 845 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, |
dfbc8630 | 846 | &links, sizeof (links)); |
0037b49e | 847 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 848 | mtime, sizeof (mtime)); |
0037b49e | 849 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
428870ff | 850 | ctime, sizeof (ctime)); |
0037b49e | 851 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff | 852 | &dzp->z_pflags, sizeof (dzp->z_pflags)); |
0df9673f | 853 | zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime); |
428870ff BB |
854 | error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); |
855 | ASSERT(error == 0); | |
34dc7c2f BB |
856 | mutex_exit(&dzp->z_lock); |
857 | ||
34dc7c2f BB |
858 | return (0); |
859 | } | |
860 | ||
9b7b9cd3 GM |
861 | /* |
862 | * The match type in the code for this function should conform to: | |
863 | * | |
864 | * ------------------------------------------------------------------------ | |
865 | * fs type | z_norm | lookup type | match type | |
866 | * ---------|-------------|-------------|---------------------------------- | |
867 | * CS !norm | 0 | 0 | 0 (exact) | |
868 | * CS norm | formX | 0 | MT_NORMALIZE | |
869 | * CI !norm | upper | !ZCIEXACT | MT_NORMALIZE | |
870 | * CI !norm | upper | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE | |
871 | * CI norm | upper|formX | !ZCIEXACT | MT_NORMALIZE | |
872 | * CI norm | upper|formX | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE | |
873 | * CM !norm | upper | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE | |
874 | * CM !norm | upper | ZCILOOK | MT_NORMALIZE | |
875 | * CM norm | upper|formX | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE | |
876 | * CM norm | upper|formX | ZCILOOK | MT_NORMALIZE | |
877 | * | |
878 | * Abbreviations: | |
879 | * CS = Case Sensitive, CI = Case Insensitive, CM = Case Mixed | |
880 | * upper = case folding set by fs type on creation (U8_TEXTPREP_TOUPPER) | |
881 | * formX = unicode normalization form set on fs creation | |
882 | */ | |
428870ff BB |
883 | static int |
884 | zfs_dropname(zfs_dirlock_t *dl, znode_t *zp, znode_t *dzp, dmu_tx_t *tx, | |
885 | int flag) | |
886 | { | |
887 | int error; | |
888 | ||
3558fd73 | 889 | if (ZTOZSB(zp)->z_norm) { |
9b7b9cd3 GM |
890 | matchtype_t mt = MT_NORMALIZE; |
891 | ||
892 | if ((ZTOZSB(zp)->z_case == ZFS_CASE_INSENSITIVE && | |
428870ff | 893 | (flag & ZCIEXACT)) || |
9b7b9cd3 GM |
894 | (ZTOZSB(zp)->z_case == ZFS_CASE_MIXED && |
895 | !(flag & ZCILOOK))) { | |
896 | mt |= MT_MATCH_CASE; | |
897 | } | |
898 | ||
899 | error = zap_remove_norm(ZTOZSB(zp)->z_os, dzp->z_id, | |
900 | dl->dl_name, mt, tx); | |
428870ff | 901 | } else { |
9b7b9cd3 GM |
902 | error = zap_remove(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, |
903 | tx); | |
428870ff BB |
904 | } |
905 | ||
906 | return (error); | |
907 | } | |
908 | ||
34dc7c2f | 909 | /* |
cd38ac58 BB |
910 | * Unlink zp from dl, and mark zp for deletion if this was the last link. Can |
911 | * fail if zp is a mount point (EBUSY) or a non-empty directory (ENOTEMPTY). | |
34dc7c2f BB |
912 | * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list. |
913 | * If it's non-NULL, we use it to indicate whether the znode needs deletion, | |
914 | * and it's the caller's job to do it. | |
915 | */ | |
916 | int | |
917 | zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag, | |
e9aa730c | 918 | boolean_t *unlinkedp) |
34dc7c2f BB |
919 | { |
920 | znode_t *dzp = dl->dl_dzp; | |
0037b49e | 921 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
3558fd73 | 922 | int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); |
34dc7c2f | 923 | boolean_t unlinked = B_FALSE; |
428870ff BB |
924 | sa_bulk_attr_t bulk[5]; |
925 | uint64_t mtime[2], ctime[2]; | |
dfbc8630 | 926 | uint64_t links; |
428870ff | 927 | int count = 0; |
34dc7c2f BB |
928 | int error; |
929 | ||
34dc7c2f | 930 | if (!(flag & ZRENAMING)) { |
34dc7c2f | 931 | mutex_enter(&zp->z_lock); |
428870ff BB |
932 | |
933 | if (zp_is_dir && !zfs_dirempty(zp)) { | |
34dc7c2f | 934 | mutex_exit(&zp->z_lock); |
2e528b49 | 935 | return (SET_ERROR(ENOTEMPTY)); |
34dc7c2f | 936 | } |
428870ff BB |
937 | |
938 | /* | |
939 | * If we get here, we are going to try to remove the object. | |
940 | * First try removing the name from the directory; if that | |
941 | * fails, return the error. | |
942 | */ | |
943 | error = zfs_dropname(dl, zp, dzp, tx, flag); | |
944 | if (error != 0) { | |
945 | mutex_exit(&zp->z_lock); | |
428870ff BB |
946 | return (error); |
947 | } | |
948 | ||
dfbc8630 | 949 | if (ZTOI(zp)->i_nlink <= zp_is_dir) { |
3558fd73 BB |
950 | zfs_panic_recover("zfs: link count on %lu is %u, " |
951 | "should be at least %u", zp->z_id, | |
dfbc8630 CD |
952 | (int)ZTOI(zp)->i_nlink, zp_is_dir + 1); |
953 | set_nlink(ZTOI(zp), zp_is_dir + 1); | |
34dc7c2f | 954 | } |
dfbc8630 CD |
955 | drop_nlink(ZTOI(zp)); |
956 | if (ZTOI(zp)->i_nlink == zp_is_dir) { | |
34dc7c2f | 957 | zp->z_unlinked = B_TRUE; |
dfbc8630 | 958 | clear_nlink(ZTOI(zp)); |
34dc7c2f BB |
959 | unlinked = B_TRUE; |
960 | } else { | |
0037b49e | 961 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), |
428870ff | 962 | NULL, &ctime, sizeof (ctime)); |
0037b49e | 963 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 964 | NULL, &zp->z_pflags, sizeof (zp->z_pflags)); |
0df9673f CC |
965 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, |
966 | ctime); | |
34dc7c2f | 967 | } |
dfbc8630 | 968 | links = ZTOI(zp)->i_nlink; |
0037b49e | 969 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 970 | NULL, &links, sizeof (links)); |
428870ff BB |
971 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
972 | count = 0; | |
973 | ASSERT(error == 0); | |
34dc7c2f | 974 | mutex_exit(&zp->z_lock); |
428870ff BB |
975 | } else { |
976 | error = zfs_dropname(dl, zp, dzp, tx, flag); | |
977 | if (error != 0) | |
978 | return (error); | |
34dc7c2f BB |
979 | } |
980 | ||
34dc7c2f | 981 | mutex_enter(&dzp->z_lock); |
428870ff | 982 | dzp->z_size--; /* one dirent removed */ |
dfbc8630 CD |
983 | if (zp_is_dir) |
984 | drop_nlink(ZTOI(dzp)); /* ".." link from zp */ | |
985 | links = ZTOI(dzp)->i_nlink; | |
0037b49e | 986 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 987 | NULL, &links, sizeof (links)); |
0037b49e | 988 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), |
428870ff | 989 | NULL, &dzp->z_size, sizeof (dzp->z_size)); |
0037b49e | 990 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), |
428870ff | 991 | NULL, ctime, sizeof (ctime)); |
0037b49e | 992 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), |
428870ff | 993 | NULL, mtime, sizeof (mtime)); |
0037b49e | 994 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 995 | NULL, &dzp->z_pflags, sizeof (dzp->z_pflags)); |
0df9673f | 996 | zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime); |
428870ff | 997 | error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); |
34dc7c2f | 998 | ASSERT(error == 0); |
428870ff | 999 | mutex_exit(&dzp->z_lock); |
34dc7c2f BB |
1000 | |
1001 | if (unlinkedp != NULL) | |
1002 | *unlinkedp = unlinked; | |
1003 | else if (unlinked) | |
1004 | zfs_unlinked_add(zp, tx); | |
1005 | ||
1006 | return (0); | |
1007 | } | |
1008 | ||
1009 | /* | |
1010 | * Indicate whether the directory is empty. Works with or without z_lock | |
1011 | * held, but can only be consider a hint in the latter case. Returns true | |
1012 | * if only "." and ".." remain and there's no work in progress. | |
3910184d AZ |
1013 | * |
1014 | * The internal ZAP size, rather than zp->z_size, needs to be checked since | |
1015 | * some consumers (Lustre) do not strictly maintain an accurate SA_ZPL_SIZE. | |
34dc7c2f BB |
1016 | */ |
1017 | boolean_t | |
1018 | zfs_dirempty(znode_t *dzp) | |
1019 | { | |
3910184d AZ |
1020 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
1021 | uint64_t count; | |
1022 | int error; | |
1023 | ||
1024 | if (dzp->z_dirlocks != NULL) | |
1025 | return (B_FALSE); | |
1026 | ||
1027 | error = zap_count(zfsvfs->z_os, dzp->z_id, &count); | |
1028 | if (error != 0 || count != 0) | |
1029 | return (B_FALSE); | |
1030 | ||
1031 | return (B_TRUE); | |
34dc7c2f BB |
1032 | } |
1033 | ||
1034 | int | |
3558fd73 | 1035 | zfs_make_xattrdir(znode_t *zp, vattr_t *vap, struct inode **xipp, cred_t *cr) |
34dc7c2f | 1036 | { |
0037b49e | 1037 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
1038 | znode_t *xzp; |
1039 | dmu_tx_t *tx; | |
1040 | int error; | |
9babb374 BB |
1041 | zfs_acl_ids_t acl_ids; |
1042 | boolean_t fuid_dirtied; | |
3558fd73 | 1043 | #ifdef DEBUG |
428870ff | 1044 | uint64_t parent; |
3558fd73 | 1045 | #endif |
34dc7c2f | 1046 | |
3558fd73 | 1047 | *xipp = NULL; |
34dc7c2f | 1048 | |
3558fd73 | 1049 | if ((error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))) |
34dc7c2f BB |
1050 | return (error); |
1051 | ||
9babb374 BB |
1052 | if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL, |
1053 | &acl_ids)) != 0) | |
1054 | return (error); | |
9c5167d1 | 1055 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zp->z_projid)) { |
9babb374 | 1056 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1057 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
1058 | } |
1059 | ||
0037b49e | 1060 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1061 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
1062 | ZFS_SA_BASE_ATTR_SIZE); | |
1063 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
34dc7c2f | 1064 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); |
0037b49e | 1065 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1066 | if (fuid_dirtied) |
0037b49e | 1067 | zfs_fuid_txhold(zfsvfs, tx); |
384f8a09 | 1068 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 1069 | if (error) { |
428870ff | 1070 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
1071 | dmu_tx_abort(tx); |
1072 | return (error); | |
1073 | } | |
428870ff | 1074 | zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids); |
9babb374 BB |
1075 | |
1076 | if (fuid_dirtied) | |
0037b49e | 1077 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 1078 | |
428870ff | 1079 | #ifdef DEBUG |
0037b49e | 1080 | error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff BB |
1081 | &parent, sizeof (parent)); |
1082 | ASSERT(error == 0 && parent == zp->z_id); | |
1083 | #endif | |
1084 | ||
0037b49e | 1085 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id, |
428870ff | 1086 | sizeof (xzp->z_id), tx)); |
34dc7c2f | 1087 | |
98701490 | 1088 | if (!zp->z_unlinked) |
0037b49e | 1089 | (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp, |
98701490 | 1090 | xzp, "", NULL, acl_ids.z_fuidp, vap); |
9babb374 BB |
1091 | |
1092 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1093 | dmu_tx_commit(tx); |
1094 | ||
3558fd73 | 1095 | *xipp = ZTOI(xzp); |
34dc7c2f BB |
1096 | |
1097 | return (0); | |
1098 | } | |
1099 | ||
1100 | /* | |
1101 | * Return a znode for the extended attribute directory for zp. | |
1102 | * ** If the directory does not already exist, it is created ** | |
1103 | * | |
1104 | * IN: zp - znode to obtain attribute directory from | |
1105 | * cr - credentials of caller | |
1106 | * flags - flags from the VOP_LOOKUP call | |
1107 | * | |
3558fd73 | 1108 | * OUT: xipp - pointer to extended attribute znode |
34dc7c2f BB |
1109 | * |
1110 | * RETURN: 0 on success | |
1111 | * error number on failure | |
1112 | */ | |
1113 | int | |
3558fd73 | 1114 | zfs_get_xattrdir(znode_t *zp, struct inode **xipp, cred_t *cr, int flags) |
34dc7c2f | 1115 | { |
0037b49e | 1116 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
1117 | znode_t *xzp; |
1118 | zfs_dirlock_t *dl; | |
1119 | vattr_t va; | |
1120 | int error; | |
1121 | top: | |
1122 | error = zfs_dirent_lock(&dl, zp, "", &xzp, ZXATTR, NULL, NULL); | |
1123 | if (error) | |
1124 | return (error); | |
1125 | ||
1126 | if (xzp != NULL) { | |
3558fd73 | 1127 | *xipp = ZTOI(xzp); |
34dc7c2f BB |
1128 | zfs_dirent_unlock(dl); |
1129 | return (0); | |
1130 | } | |
1131 | ||
34dc7c2f BB |
1132 | if (!(flags & CREATE_XATTR_DIR)) { |
1133 | zfs_dirent_unlock(dl); | |
2e528b49 | 1134 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1135 | } |
1136 | ||
0037b49e | 1137 | if (zfs_is_readonly(zfsvfs)) { |
34dc7c2f | 1138 | zfs_dirent_unlock(dl); |
2e528b49 | 1139 | return (SET_ERROR(EROFS)); |
34dc7c2f BB |
1140 | } |
1141 | ||
1142 | /* | |
1143 | * The ability to 'create' files in an attribute | |
1144 | * directory comes from the write_xattr permission on the base file. | |
1145 | * | |
1146 | * The ability to 'search' an attribute directory requires | |
1147 | * read_xattr permission on the base file. | |
1148 | * | |
1149 | * Once in a directory the ability to read/write attributes | |
1150 | * is controlled by the permissions on the attribute file. | |
1151 | */ | |
3558fd73 | 1152 | va.va_mask = ATTR_MODE | ATTR_UID | ATTR_GID; |
34dc7c2f BB |
1153 | va.va_mode = S_IFDIR | S_ISVTX | 0777; |
1154 | zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid); | |
1155 | ||
274b7e79 | 1156 | va.va_dentry = NULL; |
3558fd73 | 1157 | error = zfs_make_xattrdir(zp, &va, xipp, cr); |
34dc7c2f BB |
1158 | zfs_dirent_unlock(dl); |
1159 | ||
fb5f0bc8 | 1160 | if (error == ERESTART) { |
34dc7c2f BB |
1161 | /* NB: we already did dmu_tx_wait() if necessary */ |
1162 | goto top; | |
1163 | } | |
1164 | ||
1165 | return (error); | |
1166 | } | |
1167 | ||
1168 | /* | |
1169 | * Decide whether it is okay to remove within a sticky directory. | |
1170 | * | |
1171 | * In sticky directories, write access is not sufficient; | |
1172 | * you can remove entries from a directory only if: | |
1173 | * | |
1174 | * you own the directory, | |
1175 | * you own the entry, | |
7b340700 | 1176 | * you have write access to the entry, |
34dc7c2f BB |
1177 | * or you are privileged (checked in secpolicy...). |
1178 | * | |
1179 | * The function returns 0 if remove access is granted. | |
1180 | */ | |
1181 | int | |
1182 | zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr) | |
1183 | { | |
3558fd73 | 1184 | uid_t uid; |
572e2857 BB |
1185 | uid_t downer; |
1186 | uid_t fowner; | |
0037b49e | 1187 | zfsvfs_t *zfsvfs = ZTOZSB(zdp); |
34dc7c2f | 1188 | |
0037b49e | 1189 | if (zfsvfs->z_replay) |
34dc7c2f BB |
1190 | return (0); |
1191 | ||
428870ff | 1192 | if ((zdp->z_mode & S_ISVTX) == 0) |
34dc7c2f BB |
1193 | return (0); |
1194 | ||
0037b49e | 1195 | downer = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(ZTOI(zdp)->i_uid), |
2c6abf15 | 1196 | cr, ZFS_OWNER); |
0037b49e | 1197 | fowner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(ZTOI(zp)->i_uid), |
2c6abf15 | 1198 | cr, ZFS_OWNER); |
572e2857 BB |
1199 | |
1200 | if ((uid = crgetuid(cr)) == downer || uid == fowner || | |
7b340700 | 1201 | zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0) |
34dc7c2f BB |
1202 | return (0); |
1203 | else | |
1204 | return (secpolicy_vnode_remove(cr)); | |
1205 | } |