]>
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 | ||
3558fd73 | 102 | #ifdef HAVE_DNLC |
34dc7c2f | 103 | if (error == ENOENT && update) |
3558fd73 BB |
104 | dnlc_update(ZTOI(dzp), name, DNLC_NO_VNODE); |
105 | #endif /* HAVE_DNLC */ | |
34dc7c2f BB |
106 | |
107 | return (error); | |
108 | } | |
109 | ||
110 | /* | |
111 | * Lock a directory entry. A dirlock on <dzp, name> protects that name | |
112 | * in dzp's directory zap object. As long as you hold a dirlock, you can | |
113 | * assume two things: (1) dzp cannot be reaped, and (2) no other thread | |
114 | * can change the zap entry for (i.e. link or unlink) this name. | |
115 | * | |
116 | * Input arguments: | |
117 | * dzp - znode for directory | |
118 | * name - name of entry to lock | |
119 | * flag - ZNEW: if the entry already exists, fail with EEXIST. | |
120 | * ZEXISTS: if the entry does not exist, fail with ENOENT. | |
121 | * ZSHARED: allow concurrent access with other ZSHARED callers. | |
122 | * ZXATTR: we want dzp's xattr directory | |
123 | * ZCILOOK: On a mixed sensitivity file system, | |
124 | * this lookup should be case-insensitive. | |
125 | * ZCIEXACT: On a purely case-insensitive file system, | |
126 | * this lookup should be case-sensitive. | |
127 | * ZRENAMING: we are locking for renaming, force narrow locks | |
428870ff BB |
128 | * ZHAVELOCK: Don't grab the z_name_lock for this call. The |
129 | * current thread already holds it. | |
34dc7c2f BB |
130 | * |
131 | * Output arguments: | |
132 | * zpp - pointer to the znode for the entry (NULL if there isn't one) | |
133 | * dlpp - pointer to the dirlock for this entry (NULL on error) | |
134 | * direntflags - (case-insensitive lookup only) | |
135 | * flags if multiple case-sensitive matches exist in directory | |
136 | * realpnp - (case-insensitive lookup only) | |
137 | * actual name matched within the directory | |
138 | * | |
139 | * Return value: 0 on success or errno on failure. | |
140 | * | |
141 | * NOTE: Always checks for, and rejects, '.' and '..'. | |
142 | * NOTE: For case-insensitive file systems we take wide locks (see below), | |
143 | * but return znode pointers to a single match. | |
144 | */ | |
145 | int | |
146 | zfs_dirent_lock(zfs_dirlock_t **dlpp, znode_t *dzp, char *name, znode_t **zpp, | |
147 | int flag, int *direntflags, pathname_t *realpnp) | |
148 | { | |
0037b49e | 149 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f BB |
150 | zfs_dirlock_t *dl; |
151 | boolean_t update; | |
9b7b9cd3 | 152 | matchtype_t mt = 0; |
34dc7c2f | 153 | uint64_t zoid; |
3558fd73 | 154 | #ifdef HAVE_DNLC |
34dc7c2f | 155 | vnode_t *vp = NULL; |
3558fd73 | 156 | #endif /* HAVE_DNLC */ |
34dc7c2f BB |
157 | int error = 0; |
158 | int cmpflags; | |
159 | ||
160 | *zpp = NULL; | |
161 | *dlpp = NULL; | |
162 | ||
163 | /* | |
164 | * Verify that we are not trying to lock '.', '..', or '.zfs' | |
165 | */ | |
149e873a BB |
166 | if ((name[0] == '.' && |
167 | (name[1] == '\0' || (name[1] == '.' && name[2] == '\0'))) || | |
168 | (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)) | |
2e528b49 | 169 | return (SET_ERROR(EEXIST)); |
34dc7c2f BB |
170 | |
171 | /* | |
172 | * Case sensitivity and normalization preferences are set when | |
173 | * the file system is created. These are stored in the | |
0037b49e | 174 | * zfsvfs->z_case and zfsvfs->z_norm fields. These choices |
34dc7c2f BB |
175 | * affect what vnodes can be cached in the DNLC, how we |
176 | * perform zap lookups, and the "width" of our dirlocks. | |
177 | * | |
178 | * A normal dirlock locks a single name. Note that with | |
179 | * normalization a name can be composed multiple ways, but | |
180 | * when normalized, these names all compare equal. A wide | |
181 | * dirlock locks multiple names. We need these when the file | |
182 | * system is supporting mixed-mode access. It is sometimes | |
183 | * necessary to lock all case permutations of file name at | |
184 | * once so that simultaneous case-insensitive/case-sensitive | |
185 | * behaves as rationally as possible. | |
186 | */ | |
187 | ||
188 | /* | |
9b7b9cd3 GM |
189 | * When matching we may need to normalize & change case according to |
190 | * FS settings. | |
191 | * | |
192 | * Note that a normalized match is necessary for a case insensitive | |
193 | * filesystem when the lookup request is not exact because normalization | |
194 | * can fold case independent of normalizing code point sequences. | |
195 | * | |
196 | * See the table above zfs_dropname(). | |
34dc7c2f | 197 | */ |
0037b49e | 198 | if (zfsvfs->z_norm != 0) { |
9b7b9cd3 GM |
199 | mt = MT_NORMALIZE; |
200 | ||
201 | /* | |
202 | * Determine if the match needs to honor the case specified in | |
203 | * lookup, and if so keep track of that so that during | |
204 | * normalization we don't fold case. | |
205 | */ | |
0037b49e | 206 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE && |
9b7b9cd3 | 207 | (flag & ZCIEXACT)) || |
0037b49e | 208 | (zfsvfs->z_case == ZFS_CASE_MIXED && !(flag & ZCILOOK))) { |
9b7b9cd3 GM |
209 | mt |= MT_MATCH_CASE; |
210 | } | |
211 | } | |
34dc7c2f BB |
212 | |
213 | /* | |
214 | * Only look in or update the DNLC if we are looking for the | |
215 | * name on a file system that does not require normalization | |
216 | * or case folding. We can also look there if we happen to be | |
217 | * on a non-normalizing, mixed sensitivity file system IF we | |
218 | * are looking for the exact name. | |
219 | * | |
220 | * Maybe can add TO-UPPERed version of name to dnlc in ci-only | |
221 | * case for performance improvement? | |
222 | */ | |
0037b49e BB |
223 | update = !zfsvfs->z_norm || |
224 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
225 | !(zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER) && !(flag & ZCILOOK)); | |
34dc7c2f BB |
226 | |
227 | /* | |
228 | * ZRENAMING indicates we are in a situation where we should | |
229 | * take narrow locks regardless of the file system's | |
230 | * preferences for normalizing and case folding. This will | |
231 | * prevent us deadlocking trying to grab the same wide lock | |
232 | * twice if the two names happen to be case-insensitive | |
233 | * matches. | |
234 | */ | |
235 | if (flag & ZRENAMING) | |
236 | cmpflags = 0; | |
237 | else | |
0037b49e | 238 | cmpflags = zfsvfs->z_norm; |
34dc7c2f BB |
239 | |
240 | /* | |
241 | * Wait until there are no locks on this name. | |
428870ff BB |
242 | * |
243 | * Don't grab the the lock if it is already held. However, cannot | |
244 | * have both ZSHARED and ZHAVELOCK together. | |
34dc7c2f | 245 | */ |
428870ff BB |
246 | ASSERT(!(flag & ZSHARED) || !(flag & ZHAVELOCK)); |
247 | if (!(flag & ZHAVELOCK)) | |
248 | rw_enter(&dzp->z_name_lock, RW_READER); | |
249 | ||
34dc7c2f BB |
250 | mutex_enter(&dzp->z_lock); |
251 | for (;;) { | |
98701490 | 252 | if (dzp->z_unlinked && !(flag & ZXATTR)) { |
34dc7c2f | 253 | mutex_exit(&dzp->z_lock); |
428870ff BB |
254 | if (!(flag & ZHAVELOCK)) |
255 | rw_exit(&dzp->z_name_lock); | |
2e528b49 | 256 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
257 | } |
258 | for (dl = dzp->z_dirlocks; dl != NULL; dl = dl->dl_next) { | |
259 | if ((u8_strcmp(name, dl->dl_name, 0, cmpflags, | |
260 | U8_UNICODE_LATEST, &error) == 0) || error != 0) | |
261 | break; | |
262 | } | |
263 | if (error != 0) { | |
264 | mutex_exit(&dzp->z_lock); | |
428870ff BB |
265 | if (!(flag & ZHAVELOCK)) |
266 | rw_exit(&dzp->z_name_lock); | |
2e528b49 | 267 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
268 | } |
269 | if (dl == NULL) { | |
270 | /* | |
271 | * Allocate a new dirlock and add it to the list. | |
272 | */ | |
273 | dl = kmem_alloc(sizeof (zfs_dirlock_t), KM_SLEEP); | |
274 | cv_init(&dl->dl_cv, NULL, CV_DEFAULT, NULL); | |
275 | dl->dl_name = name; | |
276 | dl->dl_sharecnt = 0; | |
428870ff | 277 | dl->dl_namelock = 0; |
34dc7c2f BB |
278 | dl->dl_namesize = 0; |
279 | dl->dl_dzp = dzp; | |
280 | dl->dl_next = dzp->z_dirlocks; | |
281 | dzp->z_dirlocks = dl; | |
282 | break; | |
283 | } | |
284 | if ((flag & ZSHARED) && dl->dl_sharecnt != 0) | |
285 | break; | |
286 | cv_wait(&dl->dl_cv, &dzp->z_lock); | |
287 | } | |
288 | ||
428870ff BB |
289 | /* |
290 | * If the z_name_lock was NOT held for this dirlock record it. | |
291 | */ | |
292 | if (flag & ZHAVELOCK) | |
293 | dl->dl_namelock = 1; | |
294 | ||
34dc7c2f BB |
295 | if ((flag & ZSHARED) && ++dl->dl_sharecnt > 1 && dl->dl_namesize == 0) { |
296 | /* | |
297 | * We're the second shared reference to dl. Make a copy of | |
298 | * dl_name in case the first thread goes away before we do. | |
299 | * Note that we initialize the new name before storing its | |
300 | * pointer into dl_name, because the first thread may load | |
9e2c3bb4 DH |
301 | * dl->dl_name at any time. It'll either see the old value, |
302 | * which belongs to it, or the new shared copy; either is OK. | |
34dc7c2f BB |
303 | */ |
304 | dl->dl_namesize = strlen(dl->dl_name) + 1; | |
305 | name = kmem_alloc(dl->dl_namesize, KM_SLEEP); | |
306 | bcopy(dl->dl_name, name, dl->dl_namesize); | |
307 | dl->dl_name = name; | |
308 | } | |
309 | ||
310 | mutex_exit(&dzp->z_lock); | |
311 | ||
312 | /* | |
313 | * We have a dirlock on the name. (Note that it is the dirlock, | |
314 | * not the dzp's z_lock, that protects the name in the zap object.) | |
315 | * See if there's an object by this name; if so, put a hold on it. | |
316 | */ | |
317 | if (flag & ZXATTR) { | |
0037b49e | 318 | error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &zoid, |
428870ff BB |
319 | sizeof (zoid)); |
320 | if (error == 0) | |
2e528b49 | 321 | error = (zoid == 0 ? SET_ERROR(ENOENT) : 0); |
34dc7c2f | 322 | } else { |
3558fd73 | 323 | #ifdef HAVE_DNLC |
34dc7c2f | 324 | if (update) |
3558fd73 | 325 | vp = dnlc_lookup(ZTOI(dzp), name); |
34dc7c2f | 326 | if (vp == DNLC_NO_VNODE) { |
3558fd73 | 327 | iput(vp); |
2e528b49 | 328 | error = SET_ERROR(ENOENT); |
34dc7c2f BB |
329 | } else if (vp) { |
330 | if (flag & ZNEW) { | |
331 | zfs_dirent_unlock(dl); | |
3558fd73 | 332 | iput(vp); |
2e528b49 | 333 | return (SET_ERROR(EEXIST)); |
34dc7c2f BB |
334 | } |
335 | *dlpp = dl; | |
336 | *zpp = VTOZ(vp); | |
337 | return (0); | |
338 | } else { | |
0037b49e | 339 | error = zfs_match_find(zfsvfs, dzp, name, mt, |
34dc7c2f BB |
340 | update, direntflags, realpnp, &zoid); |
341 | } | |
3558fd73 | 342 | #else |
0037b49e | 343 | error = zfs_match_find(zfsvfs, dzp, name, mt, |
3558fd73 BB |
344 | update, direntflags, realpnp, &zoid); |
345 | #endif /* HAVE_DNLC */ | |
34dc7c2f BB |
346 | } |
347 | if (error) { | |
348 | if (error != ENOENT || (flag & ZEXISTS)) { | |
349 | zfs_dirent_unlock(dl); | |
350 | return (error); | |
351 | } | |
352 | } else { | |
353 | if (flag & ZNEW) { | |
354 | zfs_dirent_unlock(dl); | |
2e528b49 | 355 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 356 | } |
0037b49e | 357 | error = zfs_zget(zfsvfs, zoid, zpp); |
34dc7c2f BB |
358 | if (error) { |
359 | zfs_dirent_unlock(dl); | |
360 | return (error); | |
361 | } | |
3558fd73 | 362 | #ifdef HAVE_DNLC |
34dc7c2f | 363 | if (!(flag & ZXATTR) && update) |
3558fd73 BB |
364 | dnlc_update(ZTOI(dzp), name, ZTOI(*zpp)); |
365 | #endif /* HAVE_DNLC */ | |
34dc7c2f BB |
366 | } |
367 | ||
368 | *dlpp = dl; | |
369 | ||
370 | return (0); | |
371 | } | |
372 | ||
373 | /* | |
374 | * Unlock this directory entry and wake anyone who was waiting for it. | |
375 | */ | |
376 | void | |
377 | zfs_dirent_unlock(zfs_dirlock_t *dl) | |
378 | { | |
379 | znode_t *dzp = dl->dl_dzp; | |
380 | zfs_dirlock_t **prev_dl, *cur_dl; | |
381 | ||
382 | mutex_enter(&dzp->z_lock); | |
428870ff BB |
383 | |
384 | if (!dl->dl_namelock) | |
385 | rw_exit(&dzp->z_name_lock); | |
386 | ||
34dc7c2f BB |
387 | if (dl->dl_sharecnt > 1) { |
388 | dl->dl_sharecnt--; | |
389 | mutex_exit(&dzp->z_lock); | |
390 | return; | |
391 | } | |
392 | prev_dl = &dzp->z_dirlocks; | |
393 | while ((cur_dl = *prev_dl) != dl) | |
394 | prev_dl = &cur_dl->dl_next; | |
395 | *prev_dl = dl->dl_next; | |
396 | cv_broadcast(&dl->dl_cv); | |
397 | mutex_exit(&dzp->z_lock); | |
398 | ||
399 | if (dl->dl_namesize != 0) | |
400 | kmem_free(dl->dl_name, dl->dl_namesize); | |
401 | cv_destroy(&dl->dl_cv); | |
402 | kmem_free(dl, sizeof (*dl)); | |
403 | } | |
404 | ||
405 | /* | |
406 | * Look up an entry in a directory. | |
407 | * | |
408 | * NOTE: '.' and '..' are handled as special cases because | |
409 | * no directory entries are actually stored for them. If this is | |
410 | * the root of a filesystem, then '.zfs' is also treated as a | |
411 | * special pseudo-directory. | |
412 | */ | |
413 | int | |
3558fd73 | 414 | zfs_dirlook(znode_t *dzp, char *name, struct inode **ipp, int flags, |
34dc7c2f BB |
415 | int *deflg, pathname_t *rpnp) |
416 | { | |
417 | zfs_dirlock_t *dl; | |
418 | znode_t *zp; | |
419 | int error = 0; | |
428870ff | 420 | uint64_t parent; |
34dc7c2f BB |
421 | |
422 | if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) { | |
3558fd73 BB |
423 | *ipp = ZTOI(dzp); |
424 | igrab(*ipp); | |
34dc7c2f | 425 | } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) { |
0037b49e | 426 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
428870ff | 427 | |
34dc7c2f BB |
428 | /* |
429 | * If we are a snapshot mounted under .zfs, return | |
ebe7e575 | 430 | * the inode pointer for the snapshot directory. |
34dc7c2f | 431 | */ |
428870ff | 432 | if ((error = sa_lookup(dzp->z_sa_hdl, |
0037b49e | 433 | SA_ZPL_PARENT(zfsvfs), &parent, sizeof (parent))) != 0) |
428870ff | 434 | return (error); |
ebe7e575 | 435 | |
0037b49e BB |
436 | if (parent == dzp->z_id && zfsvfs->z_parent != zfsvfs) { |
437 | error = zfsctl_root_lookup(zfsvfs->z_parent->z_ctldir, | |
ebe7e575 | 438 | "snapshot", ipp, 0, kcred, NULL, NULL); |
34dc7c2f BB |
439 | return (error); |
440 | } | |
441 | rw_enter(&dzp->z_parent_lock, RW_READER); | |
0037b49e | 442 | error = zfs_zget(zfsvfs, parent, &zp); |
34dc7c2f | 443 | if (error == 0) |
3558fd73 | 444 | *ipp = ZTOI(zp); |
34dc7c2f BB |
445 | rw_exit(&dzp->z_parent_lock); |
446 | } else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) { | |
3558fd73 | 447 | *ipp = zfsctl_root(dzp); |
34dc7c2f BB |
448 | } else { |
449 | int zf; | |
450 | ||
451 | zf = ZEXISTS | ZSHARED; | |
452 | if (flags & FIGNORECASE) | |
453 | zf |= ZCILOOK; | |
454 | ||
455 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zf, deflg, rpnp); | |
456 | if (error == 0) { | |
3558fd73 | 457 | *ipp = ZTOI(zp); |
34dc7c2f BB |
458 | zfs_dirent_unlock(dl); |
459 | dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */ | |
460 | } | |
461 | rpnp = NULL; | |
462 | } | |
463 | ||
464 | if ((flags & FIGNORECASE) && rpnp && !error) | |
465 | (void) strlcpy(rpnp->pn_buf, name, rpnp->pn_bufsize); | |
466 | ||
467 | return (error); | |
468 | } | |
469 | ||
34dc7c2f BB |
470 | /* |
471 | * unlinked Set (formerly known as the "delete queue") Error Handling | |
472 | * | |
473 | * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we | |
474 | * don't specify the name of the entry that we will be manipulating. We | |
475 | * also fib and say that we won't be adding any new entries to the | |
476 | * unlinked set, even though we might (this is to lower the minimum file | |
477 | * size that can be deleted in a full filesystem). So on the small | |
478 | * chance that the nlink list is using a fat zap (ie. has more than | |
479 | * 2000 entries), we *may* not pre-read a block that's needed. | |
480 | * Therefore it is remotely possible for some of the assertions | |
481 | * regarding the unlinked set below to fail due to i/o error. On a | |
482 | * nondebug system, this will result in the space being leaked. | |
483 | */ | |
484 | void | |
485 | zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx) | |
486 | { | |
0037b49e | 487 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
488 | |
489 | ASSERT(zp->z_unlinked); | |
dfbc8630 | 490 | ASSERT(ZTOI(zp)->i_nlink == 0); |
34dc7c2f | 491 | |
b128c09f | 492 | VERIFY3U(0, ==, |
0037b49e | 493 | zap_add_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); |
34dc7c2f BB |
494 | } |
495 | ||
0d5c500d BB |
496 | /* |
497 | * Clean up any znodes that had no links when we either crashed or | |
498 | * (force) umounted the file system. | |
499 | */ | |
500 | void | |
0037b49e | 501 | zfs_unlinked_drain(zfsvfs_t *zfsvfs) |
0d5c500d BB |
502 | { |
503 | zap_cursor_t zc; | |
504 | zap_attribute_t zap; | |
505 | dmu_object_info_t doi; | |
506 | znode_t *zp; | |
507 | int error; | |
508 | ||
509 | /* | |
510 | * Iterate over the contents of the unlinked set. | |
511 | */ | |
0037b49e | 512 | for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj); |
0d5c500d BB |
513 | zap_cursor_retrieve(&zc, &zap) == 0; |
514 | zap_cursor_advance(&zc)) { | |
515 | ||
516 | /* | |
517 | * See what kind of object we have in list | |
518 | */ | |
519 | ||
0037b49e BB |
520 | error = dmu_object_info(zfsvfs->z_os, |
521 | zap.za_first_integer, &doi); | |
0d5c500d BB |
522 | if (error != 0) |
523 | continue; | |
524 | ||
525 | ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) || | |
526 | (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS)); | |
527 | /* | |
528 | * We need to re-mark these list entries for deletion, | |
529 | * so we pull them back into core and set zp->z_unlinked. | |
530 | */ | |
0037b49e | 531 | error = zfs_zget(zfsvfs, zap.za_first_integer, &zp); |
0d5c500d BB |
532 | |
533 | /* | |
534 | * We may pick up znodes that are already marked for deletion. | |
535 | * This could happen during the purge of an extended attribute | |
536 | * directory. All we need to do is skip over them, since they | |
537 | * are already in the system marked z_unlinked. | |
538 | */ | |
539 | if (error != 0) | |
540 | continue; | |
541 | ||
542 | zp->z_unlinked = B_TRUE; | |
543 | iput(ZTOI(zp)); | |
544 | } | |
545 | zap_cursor_fini(&zc); | |
546 | } | |
547 | ||
34dc7c2f BB |
548 | /* |
549 | * Delete the entire contents of a directory. Return a count | |
550 | * of the number of entries that could not be deleted. If we encounter | |
551 | * an error, return a count of at least one so that the directory stays | |
552 | * in the unlinked set. | |
553 | * | |
554 | * NOTE: this function assumes that the directory is inactive, | |
555 | * so there is no need to lock its entries before deletion. | |
556 | * Also, it assumes the directory contents is *only* regular | |
557 | * files. | |
558 | */ | |
559 | static int | |
560 | zfs_purgedir(znode_t *dzp) | |
561 | { | |
562 | zap_cursor_t zc; | |
563 | zap_attribute_t zap; | |
564 | znode_t *xzp; | |
565 | dmu_tx_t *tx; | |
0037b49e | 566 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f BB |
567 | zfs_dirlock_t dl; |
568 | int skipped = 0; | |
569 | int error; | |
570 | ||
0037b49e | 571 | for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); |
34dc7c2f BB |
572 | (error = zap_cursor_retrieve(&zc, &zap)) == 0; |
573 | zap_cursor_advance(&zc)) { | |
0037b49e | 574 | error = zfs_zget(zfsvfs, |
34dc7c2f BB |
575 | ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp); |
576 | if (error) { | |
577 | skipped += 1; | |
578 | continue; | |
579 | } | |
580 | ||
0d5c500d BB |
581 | ASSERT(S_ISREG(ZTOI(xzp)->i_mode) || |
582 | S_ISLNK(ZTOI(xzp)->i_mode)); | |
34dc7c2f | 583 | |
0037b49e | 584 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 585 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
34dc7c2f | 586 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name); |
428870ff | 587 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); |
0037b49e | 588 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
589 | /* Is this really needed ? */ |
590 | zfs_sa_upgrade_txholds(tx, xzp); | |
19d55079 | 591 | dmu_tx_mark_netfree(tx); |
34dc7c2f BB |
592 | error = dmu_tx_assign(tx, TXG_WAIT); |
593 | if (error) { | |
594 | dmu_tx_abort(tx); | |
0a50679c | 595 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
596 | skipped += 1; |
597 | continue; | |
598 | } | |
599 | bzero(&dl, sizeof (dl)); | |
600 | dl.dl_dzp = dzp; | |
601 | dl.dl_name = zap.za_name; | |
602 | ||
603 | error = zfs_link_destroy(&dl, xzp, tx, 0, NULL); | |
604 | if (error) | |
605 | skipped += 1; | |
606 | dmu_tx_commit(tx); | |
dfbc8630 | 607 | |
0a50679c | 608 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
609 | } |
610 | zap_cursor_fini(&zc); | |
611 | if (error != ENOENT) | |
612 | skipped += 1; | |
613 | return (skipped); | |
614 | } | |
615 | ||
616 | void | |
617 | zfs_rmnode(znode_t *zp) | |
618 | { | |
0037b49e BB |
619 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
620 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 621 | znode_t *xzp = NULL; |
34dc7c2f BB |
622 | dmu_tx_t *tx; |
623 | uint64_t acl_obj; | |
428870ff | 624 | uint64_t xattr_obj; |
dfbc8630 | 625 | uint64_t links; |
34dc7c2f BB |
626 | int error; |
627 | ||
dfbc8630 | 628 | ASSERT(ZTOI(zp)->i_nlink == 0); |
3558fd73 | 629 | ASSERT(atomic_read(&ZTOI(zp)->i_count) == 0); |
34dc7c2f BB |
630 | |
631 | /* | |
632 | * If this is an attribute directory, purge its contents. | |
633 | */ | |
3558fd73 | 634 | if (S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_XATTR)) { |
0d5c500d | 635 | if (zfs_purgedir(zp) != 0) { |
34dc7c2f | 636 | /* |
0d5c500d BB |
637 | * Not enough space to delete some xattrs. |
638 | * Leave it in the unlinked set. | |
34dc7c2f | 639 | */ |
7973e464 | 640 | zfs_znode_dmu_fini(zp); |
0d5c500d | 641 | |
34dc7c2f BB |
642 | return; |
643 | } | |
644 | } | |
645 | ||
b128c09f | 646 | /* |
f5f087eb CC |
647 | * Free up all the data in the file. We don't do this for directories |
648 | * because we need truncate and remove to be in the same tx, like in | |
649 | * zfs_znode_delete(). Otherwise, if we crash here we'll end up with | |
650 | * an inconsistent truncated zap object in the delete queue. Note a | |
651 | * truncated file is harmless since it only contains user data. | |
b128c09f | 652 | */ |
f5f087eb CC |
653 | if (S_ISREG(ZTOI(zp)->i_mode)) { |
654 | error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END); | |
655 | if (error) { | |
656 | /* | |
a08abc1b GM |
657 | * Not enough space or we were interrupted by unmount. |
658 | * Leave the file in the unlinked set. | |
f5f087eb CC |
659 | */ |
660 | zfs_znode_dmu_fini(zp); | |
661 | return; | |
662 | } | |
b128c09f BB |
663 | } |
664 | ||
34dc7c2f BB |
665 | /* |
666 | * If the file has extended attributes, we're going to unlink | |
667 | * the xattr dir. | |
668 | */ | |
0037b49e | 669 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff BB |
670 | &xattr_obj, sizeof (xattr_obj)); |
671 | if (error == 0 && xattr_obj) { | |
0037b49e | 672 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
34dc7c2f BB |
673 | ASSERT(error == 0); |
674 | } | |
675 | ||
572e2857 | 676 | acl_obj = zfs_external_acl(zp); |
34dc7c2f BB |
677 | |
678 | /* | |
b128c09f | 679 | * Set up the final transaction. |
34dc7c2f BB |
680 | */ |
681 | tx = dmu_tx_create(os); | |
682 | dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END); | |
0037b49e | 683 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 684 | if (xzp) { |
0037b49e | 685 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL); |
428870ff | 686 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); |
34dc7c2f BB |
687 | } |
688 | if (acl_obj) | |
689 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
428870ff BB |
690 | |
691 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
692 | error = dmu_tx_assign(tx, TXG_WAIT); |
693 | if (error) { | |
694 | /* | |
695 | * Not enough space to delete the file. Leave it in the | |
696 | * unlinked set, leaking it until the fs is remounted (at | |
697 | * which point we'll call zfs_unlinked_drain() to process it). | |
698 | */ | |
699 | dmu_tx_abort(tx); | |
700 | zfs_znode_dmu_fini(zp); | |
34dc7c2f BB |
701 | goto out; |
702 | } | |
703 | ||
704 | if (xzp) { | |
428870ff | 705 | ASSERT(error == 0); |
34dc7c2f BB |
706 | mutex_enter(&xzp->z_lock); |
707 | xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */ | |
dfbc8630 CD |
708 | clear_nlink(ZTOI(xzp)); /* no more links to it */ |
709 | links = 0; | |
0037b49e | 710 | VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 711 | &links, sizeof (links), tx)); |
34dc7c2f BB |
712 | mutex_exit(&xzp->z_lock); |
713 | zfs_unlinked_add(xzp, tx); | |
714 | } | |
715 | ||
716 | /* Remove this znode from the unlinked set */ | |
b128c09f | 717 | VERIFY3U(0, ==, |
0037b49e | 718 | zap_remove_int(zfsvfs->z_os, zfsvfs->z_unlinkedobj, zp->z_id, tx)); |
34dc7c2f BB |
719 | |
720 | zfs_znode_delete(zp, tx); | |
721 | ||
722 | dmu_tx_commit(tx); | |
723 | out: | |
724 | if (xzp) | |
0a50679c | 725 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
726 | } |
727 | ||
728 | static uint64_t | |
428870ff | 729 | zfs_dirent(znode_t *zp, uint64_t mode) |
34dc7c2f BB |
730 | { |
731 | uint64_t de = zp->z_id; | |
428870ff | 732 | |
3558fd73 | 733 | if (ZTOZSB(zp)->z_version >= ZPL_VERSION_DIRENT_TYPE) |
428870ff | 734 | de |= IFTODT(mode) << 60; |
34dc7c2f BB |
735 | return (de); |
736 | } | |
737 | ||
738 | /* | |
599b8648 CC |
739 | * Link zp into dl. Can fail in the following cases : |
740 | * - if zp has been unlinked. | |
741 | * - if the number of entries with the same hash (aka. colliding entries) | |
742 | * exceed the capacity of a leaf-block of fatzap and splitting of the | |
743 | * leaf-block does not help. | |
34dc7c2f BB |
744 | */ |
745 | int | |
746 | zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag) | |
747 | { | |
748 | znode_t *dzp = dl->dl_dzp; | |
0037b49e | 749 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f | 750 | uint64_t value; |
3558fd73 | 751 | int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); |
428870ff BB |
752 | sa_bulk_attr_t bulk[5]; |
753 | uint64_t mtime[2], ctime[2]; | |
dfbc8630 | 754 | uint64_t links; |
428870ff | 755 | int count = 0; |
34dc7c2f BB |
756 | int error; |
757 | ||
34dc7c2f BB |
758 | mutex_enter(&zp->z_lock); |
759 | ||
760 | if (!(flag & ZRENAMING)) { | |
761 | if (zp->z_unlinked) { /* no new links to unlinked zp */ | |
762 | ASSERT(!(flag & (ZNEW | ZEXISTS))); | |
763 | mutex_exit(&zp->z_lock); | |
2e528b49 | 764 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 765 | } |
dfbc8630 CD |
766 | if (!(flag & ZNEW)) { |
767 | /* | |
768 | * ZNEW nodes come from zfs_mknode() where the link | |
769 | * count has already been initialised | |
770 | */ | |
771 | inc_nlink(ZTOI(zp)); | |
772 | links = ZTOI(zp)->i_nlink; | |
0037b49e BB |
773 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), |
774 | NULL, &links, sizeof (links)); | |
dfbc8630 | 775 | } |
34dc7c2f | 776 | } |
599b8648 CC |
777 | |
778 | value = zfs_dirent(zp, zp->z_mode); | |
779 | error = zap_add(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, 8, 1, | |
780 | &value, tx); | |
781 | ||
782 | /* | |
783 | * zap_add could fail to add the entry if it exceeds the capacity of the | |
784 | * leaf-block and zap_leaf_split() failed to help. | |
785 | * The caller of this routine is responsible for failing the transaction | |
786 | * which will rollback the SA updates done above. | |
787 | */ | |
788 | if (error != 0) { | |
789 | if (!(flag & ZRENAMING) && !(flag & ZNEW)) | |
790 | drop_nlink(ZTOI(zp)); | |
791 | mutex_exit(&zp->z_lock); | |
792 | return (error); | |
793 | } | |
794 | ||
0037b49e | 795 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, |
428870ff | 796 | &dzp->z_id, sizeof (dzp->z_id)); |
0037b49e | 797 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
798 | &zp->z_pflags, sizeof (zp->z_pflags)); |
799 | ||
800 | if (!(flag & ZNEW)) { | |
0037b49e | 801 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
428870ff BB |
802 | ctime, sizeof (ctime)); |
803 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, | |
0df9673f | 804 | ctime); |
428870ff BB |
805 | } |
806 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
807 | ASSERT(error == 0); | |
34dc7c2f | 808 | |
34dc7c2f BB |
809 | mutex_exit(&zp->z_lock); |
810 | ||
34dc7c2f | 811 | mutex_enter(&dzp->z_lock); |
428870ff | 812 | dzp->z_size++; |
dfbc8630 CD |
813 | if (zp_is_dir) |
814 | inc_nlink(ZTOI(dzp)); | |
815 | links = ZTOI(dzp)->i_nlink; | |
428870ff | 816 | count = 0; |
0037b49e | 817 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, |
428870ff | 818 | &dzp->z_size, sizeof (dzp->z_size)); |
0037b49e | 819 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, |
dfbc8630 | 820 | &links, sizeof (links)); |
0037b49e | 821 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 822 | mtime, sizeof (mtime)); |
0037b49e | 823 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
428870ff | 824 | ctime, sizeof (ctime)); |
0037b49e | 825 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff | 826 | &dzp->z_pflags, sizeof (dzp->z_pflags)); |
0df9673f | 827 | zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime); |
428870ff BB |
828 | error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); |
829 | ASSERT(error == 0); | |
34dc7c2f BB |
830 | mutex_exit(&dzp->z_lock); |
831 | ||
34dc7c2f BB |
832 | return (0); |
833 | } | |
834 | ||
9b7b9cd3 GM |
835 | /* |
836 | * The match type in the code for this function should conform to: | |
837 | * | |
838 | * ------------------------------------------------------------------------ | |
839 | * fs type | z_norm | lookup type | match type | |
840 | * ---------|-------------|-------------|---------------------------------- | |
841 | * CS !norm | 0 | 0 | 0 (exact) | |
842 | * CS norm | formX | 0 | MT_NORMALIZE | |
843 | * CI !norm | upper | !ZCIEXACT | MT_NORMALIZE | |
844 | * CI !norm | upper | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE | |
845 | * CI norm | upper|formX | !ZCIEXACT | MT_NORMALIZE | |
846 | * CI norm | upper|formX | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE | |
847 | * CM !norm | upper | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE | |
848 | * CM !norm | upper | ZCILOOK | MT_NORMALIZE | |
849 | * CM norm | upper|formX | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE | |
850 | * CM norm | upper|formX | ZCILOOK | MT_NORMALIZE | |
851 | * | |
852 | * Abbreviations: | |
853 | * CS = Case Sensitive, CI = Case Insensitive, CM = Case Mixed | |
854 | * upper = case folding set by fs type on creation (U8_TEXTPREP_TOUPPER) | |
855 | * formX = unicode normalization form set on fs creation | |
856 | */ | |
428870ff BB |
857 | static int |
858 | zfs_dropname(zfs_dirlock_t *dl, znode_t *zp, znode_t *dzp, dmu_tx_t *tx, | |
859 | int flag) | |
860 | { | |
861 | int error; | |
862 | ||
3558fd73 | 863 | if (ZTOZSB(zp)->z_norm) { |
9b7b9cd3 GM |
864 | matchtype_t mt = MT_NORMALIZE; |
865 | ||
866 | if ((ZTOZSB(zp)->z_case == ZFS_CASE_INSENSITIVE && | |
428870ff | 867 | (flag & ZCIEXACT)) || |
9b7b9cd3 GM |
868 | (ZTOZSB(zp)->z_case == ZFS_CASE_MIXED && |
869 | !(flag & ZCILOOK))) { | |
870 | mt |= MT_MATCH_CASE; | |
871 | } | |
872 | ||
873 | error = zap_remove_norm(ZTOZSB(zp)->z_os, dzp->z_id, | |
874 | dl->dl_name, mt, tx); | |
428870ff | 875 | } else { |
9b7b9cd3 GM |
876 | error = zap_remove(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, |
877 | tx); | |
428870ff BB |
878 | } |
879 | ||
880 | return (error); | |
881 | } | |
882 | ||
34dc7c2f | 883 | /* |
cd38ac58 BB |
884 | * Unlink zp from dl, and mark zp for deletion if this was the last link. Can |
885 | * fail if zp is a mount point (EBUSY) or a non-empty directory (ENOTEMPTY). | |
34dc7c2f BB |
886 | * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list. |
887 | * If it's non-NULL, we use it to indicate whether the znode needs deletion, | |
888 | * and it's the caller's job to do it. | |
889 | */ | |
890 | int | |
891 | zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag, | |
e9aa730c | 892 | boolean_t *unlinkedp) |
34dc7c2f BB |
893 | { |
894 | znode_t *dzp = dl->dl_dzp; | |
0037b49e | 895 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
3558fd73 | 896 | int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); |
34dc7c2f | 897 | boolean_t unlinked = B_FALSE; |
428870ff BB |
898 | sa_bulk_attr_t bulk[5]; |
899 | uint64_t mtime[2], ctime[2]; | |
dfbc8630 | 900 | uint64_t links; |
428870ff | 901 | int count = 0; |
34dc7c2f BB |
902 | int error; |
903 | ||
3558fd73 BB |
904 | #ifdef HAVE_DNLC |
905 | dnlc_remove(ZTOI(dzp), dl->dl_name); | |
906 | #endif /* HAVE_DNLC */ | |
34dc7c2f BB |
907 | |
908 | if (!(flag & ZRENAMING)) { | |
34dc7c2f | 909 | mutex_enter(&zp->z_lock); |
428870ff BB |
910 | |
911 | if (zp_is_dir && !zfs_dirempty(zp)) { | |
34dc7c2f | 912 | mutex_exit(&zp->z_lock); |
2e528b49 | 913 | return (SET_ERROR(ENOTEMPTY)); |
34dc7c2f | 914 | } |
428870ff BB |
915 | |
916 | /* | |
917 | * If we get here, we are going to try to remove the object. | |
918 | * First try removing the name from the directory; if that | |
919 | * fails, return the error. | |
920 | */ | |
921 | error = zfs_dropname(dl, zp, dzp, tx, flag); | |
922 | if (error != 0) { | |
923 | mutex_exit(&zp->z_lock); | |
428870ff BB |
924 | return (error); |
925 | } | |
926 | ||
dfbc8630 | 927 | if (ZTOI(zp)->i_nlink <= zp_is_dir) { |
3558fd73 BB |
928 | zfs_panic_recover("zfs: link count on %lu is %u, " |
929 | "should be at least %u", zp->z_id, | |
dfbc8630 CD |
930 | (int)ZTOI(zp)->i_nlink, zp_is_dir + 1); |
931 | set_nlink(ZTOI(zp), zp_is_dir + 1); | |
34dc7c2f | 932 | } |
dfbc8630 CD |
933 | drop_nlink(ZTOI(zp)); |
934 | if (ZTOI(zp)->i_nlink == zp_is_dir) { | |
34dc7c2f | 935 | zp->z_unlinked = B_TRUE; |
dfbc8630 | 936 | clear_nlink(ZTOI(zp)); |
34dc7c2f BB |
937 | unlinked = B_TRUE; |
938 | } else { | |
0037b49e | 939 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), |
428870ff | 940 | NULL, &ctime, sizeof (ctime)); |
0037b49e | 941 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 942 | NULL, &zp->z_pflags, sizeof (zp->z_pflags)); |
0df9673f CC |
943 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, |
944 | ctime); | |
34dc7c2f | 945 | } |
dfbc8630 | 946 | links = ZTOI(zp)->i_nlink; |
0037b49e | 947 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 948 | NULL, &links, sizeof (links)); |
428870ff BB |
949 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
950 | count = 0; | |
951 | ASSERT(error == 0); | |
34dc7c2f | 952 | mutex_exit(&zp->z_lock); |
428870ff BB |
953 | } else { |
954 | error = zfs_dropname(dl, zp, dzp, tx, flag); | |
955 | if (error != 0) | |
956 | return (error); | |
34dc7c2f BB |
957 | } |
958 | ||
34dc7c2f | 959 | mutex_enter(&dzp->z_lock); |
428870ff | 960 | dzp->z_size--; /* one dirent removed */ |
dfbc8630 CD |
961 | if (zp_is_dir) |
962 | drop_nlink(ZTOI(dzp)); /* ".." link from zp */ | |
963 | links = ZTOI(dzp)->i_nlink; | |
0037b49e | 964 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 965 | NULL, &links, sizeof (links)); |
0037b49e | 966 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), |
428870ff | 967 | NULL, &dzp->z_size, sizeof (dzp->z_size)); |
0037b49e | 968 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), |
428870ff | 969 | NULL, ctime, sizeof (ctime)); |
0037b49e | 970 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), |
428870ff | 971 | NULL, mtime, sizeof (mtime)); |
0037b49e | 972 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 973 | NULL, &dzp->z_pflags, sizeof (dzp->z_pflags)); |
0df9673f | 974 | zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime); |
428870ff | 975 | error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); |
34dc7c2f | 976 | ASSERT(error == 0); |
428870ff | 977 | mutex_exit(&dzp->z_lock); |
34dc7c2f BB |
978 | |
979 | if (unlinkedp != NULL) | |
980 | *unlinkedp = unlinked; | |
981 | else if (unlinked) | |
982 | zfs_unlinked_add(zp, tx); | |
983 | ||
984 | return (0); | |
985 | } | |
986 | ||
987 | /* | |
988 | * Indicate whether the directory is empty. Works with or without z_lock | |
989 | * held, but can only be consider a hint in the latter case. Returns true | |
990 | * if only "." and ".." remain and there's no work in progress. | |
3910184d AZ |
991 | * |
992 | * The internal ZAP size, rather than zp->z_size, needs to be checked since | |
993 | * some consumers (Lustre) do not strictly maintain an accurate SA_ZPL_SIZE. | |
34dc7c2f BB |
994 | */ |
995 | boolean_t | |
996 | zfs_dirempty(znode_t *dzp) | |
997 | { | |
3910184d AZ |
998 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
999 | uint64_t count; | |
1000 | int error; | |
1001 | ||
1002 | if (dzp->z_dirlocks != NULL) | |
1003 | return (B_FALSE); | |
1004 | ||
1005 | error = zap_count(zfsvfs->z_os, dzp->z_id, &count); | |
1006 | if (error != 0 || count != 0) | |
1007 | return (B_FALSE); | |
1008 | ||
1009 | return (B_TRUE); | |
34dc7c2f BB |
1010 | } |
1011 | ||
1012 | int | |
3558fd73 | 1013 | zfs_make_xattrdir(znode_t *zp, vattr_t *vap, struct inode **xipp, cred_t *cr) |
34dc7c2f | 1014 | { |
0037b49e | 1015 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
1016 | znode_t *xzp; |
1017 | dmu_tx_t *tx; | |
1018 | int error; | |
9babb374 BB |
1019 | zfs_acl_ids_t acl_ids; |
1020 | boolean_t fuid_dirtied; | |
3558fd73 | 1021 | #ifdef DEBUG |
428870ff | 1022 | uint64_t parent; |
3558fd73 | 1023 | #endif |
34dc7c2f | 1024 | |
3558fd73 | 1025 | *xipp = NULL; |
34dc7c2f | 1026 | |
3558fd73 | 1027 | if ((error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))) |
34dc7c2f BB |
1028 | return (error); |
1029 | ||
9babb374 BB |
1030 | if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL, |
1031 | &acl_ids)) != 0) | |
1032 | return (error); | |
9c5167d1 | 1033 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zp->z_projid)) { |
9babb374 | 1034 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1035 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
1036 | } |
1037 | ||
0037b49e | 1038 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1039 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
1040 | ZFS_SA_BASE_ATTR_SIZE); | |
1041 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
34dc7c2f | 1042 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); |
0037b49e | 1043 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1044 | if (fuid_dirtied) |
0037b49e | 1045 | zfs_fuid_txhold(zfsvfs, tx); |
384f8a09 | 1046 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 1047 | if (error) { |
428870ff | 1048 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
1049 | dmu_tx_abort(tx); |
1050 | return (error); | |
1051 | } | |
428870ff | 1052 | zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids); |
9babb374 BB |
1053 | |
1054 | if (fuid_dirtied) | |
0037b49e | 1055 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 1056 | |
428870ff | 1057 | #ifdef DEBUG |
0037b49e | 1058 | error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff BB |
1059 | &parent, sizeof (parent)); |
1060 | ASSERT(error == 0 && parent == zp->z_id); | |
1061 | #endif | |
1062 | ||
0037b49e | 1063 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), &xzp->z_id, |
428870ff | 1064 | sizeof (xzp->z_id), tx)); |
34dc7c2f | 1065 | |
98701490 | 1066 | if (!zp->z_unlinked) |
0037b49e | 1067 | (void) zfs_log_create(zfsvfs->z_log, tx, TX_MKXATTR, zp, |
98701490 | 1068 | xzp, "", NULL, acl_ids.z_fuidp, vap); |
9babb374 BB |
1069 | |
1070 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1071 | dmu_tx_commit(tx); |
1072 | ||
3558fd73 | 1073 | *xipp = ZTOI(xzp); |
34dc7c2f BB |
1074 | |
1075 | return (0); | |
1076 | } | |
1077 | ||
1078 | /* | |
1079 | * Return a znode for the extended attribute directory for zp. | |
1080 | * ** If the directory does not already exist, it is created ** | |
1081 | * | |
1082 | * IN: zp - znode to obtain attribute directory from | |
1083 | * cr - credentials of caller | |
1084 | * flags - flags from the VOP_LOOKUP call | |
1085 | * | |
3558fd73 | 1086 | * OUT: xipp - pointer to extended attribute znode |
34dc7c2f BB |
1087 | * |
1088 | * RETURN: 0 on success | |
1089 | * error number on failure | |
1090 | */ | |
1091 | int | |
3558fd73 | 1092 | zfs_get_xattrdir(znode_t *zp, struct inode **xipp, cred_t *cr, int flags) |
34dc7c2f | 1093 | { |
0037b49e | 1094 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
1095 | znode_t *xzp; |
1096 | zfs_dirlock_t *dl; | |
1097 | vattr_t va; | |
1098 | int error; | |
1099 | top: | |
1100 | error = zfs_dirent_lock(&dl, zp, "", &xzp, ZXATTR, NULL, NULL); | |
1101 | if (error) | |
1102 | return (error); | |
1103 | ||
1104 | if (xzp != NULL) { | |
3558fd73 | 1105 | *xipp = ZTOI(xzp); |
34dc7c2f BB |
1106 | zfs_dirent_unlock(dl); |
1107 | return (0); | |
1108 | } | |
1109 | ||
34dc7c2f BB |
1110 | if (!(flags & CREATE_XATTR_DIR)) { |
1111 | zfs_dirent_unlock(dl); | |
2e528b49 | 1112 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1113 | } |
1114 | ||
0037b49e | 1115 | if (zfs_is_readonly(zfsvfs)) { |
34dc7c2f | 1116 | zfs_dirent_unlock(dl); |
2e528b49 | 1117 | return (SET_ERROR(EROFS)); |
34dc7c2f BB |
1118 | } |
1119 | ||
1120 | /* | |
1121 | * The ability to 'create' files in an attribute | |
1122 | * directory comes from the write_xattr permission on the base file. | |
1123 | * | |
1124 | * The ability to 'search' an attribute directory requires | |
1125 | * read_xattr permission on the base file. | |
1126 | * | |
1127 | * Once in a directory the ability to read/write attributes | |
1128 | * is controlled by the permissions on the attribute file. | |
1129 | */ | |
3558fd73 | 1130 | va.va_mask = ATTR_MODE | ATTR_UID | ATTR_GID; |
34dc7c2f BB |
1131 | va.va_mode = S_IFDIR | S_ISVTX | 0777; |
1132 | zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid); | |
1133 | ||
274b7e79 | 1134 | va.va_dentry = NULL; |
3558fd73 | 1135 | error = zfs_make_xattrdir(zp, &va, xipp, cr); |
34dc7c2f BB |
1136 | zfs_dirent_unlock(dl); |
1137 | ||
fb5f0bc8 | 1138 | if (error == ERESTART) { |
34dc7c2f BB |
1139 | /* NB: we already did dmu_tx_wait() if necessary */ |
1140 | goto top; | |
1141 | } | |
1142 | ||
1143 | return (error); | |
1144 | } | |
1145 | ||
1146 | /* | |
1147 | * Decide whether it is okay to remove within a sticky directory. | |
1148 | * | |
1149 | * In sticky directories, write access is not sufficient; | |
1150 | * you can remove entries from a directory only if: | |
1151 | * | |
1152 | * you own the directory, | |
1153 | * you own the entry, | |
7b340700 | 1154 | * you have write access to the entry, |
34dc7c2f BB |
1155 | * or you are privileged (checked in secpolicy...). |
1156 | * | |
1157 | * The function returns 0 if remove access is granted. | |
1158 | */ | |
1159 | int | |
1160 | zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr) | |
1161 | { | |
3558fd73 | 1162 | uid_t uid; |
572e2857 BB |
1163 | uid_t downer; |
1164 | uid_t fowner; | |
0037b49e | 1165 | zfsvfs_t *zfsvfs = ZTOZSB(zdp); |
34dc7c2f | 1166 | |
0037b49e | 1167 | if (zfsvfs->z_replay) |
34dc7c2f BB |
1168 | return (0); |
1169 | ||
428870ff | 1170 | if ((zdp->z_mode & S_ISVTX) == 0) |
34dc7c2f BB |
1171 | return (0); |
1172 | ||
0037b49e | 1173 | downer = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(ZTOI(zdp)->i_uid), |
2c6abf15 | 1174 | cr, ZFS_OWNER); |
0037b49e | 1175 | fowner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(ZTOI(zp)->i_uid), |
2c6abf15 | 1176 | cr, ZFS_OWNER); |
572e2857 BB |
1177 | |
1178 | if ((uid = crgetuid(cr)) == downer || uid == fowner || | |
7b340700 | 1179 | zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0) |
34dc7c2f BB |
1180 | return (0); |
1181 | else | |
1182 | return (secpolicy_vnode_remove(cr)); | |
1183 | } |