]>
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 | */ | |
21 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
a08ee875 | 23 | * Copyright (c) 2013 by Delphix. All rights reserved. |
34dc7c2f BB |
24 | */ |
25 | ||
60101509 | 26 | |
34dc7c2f BB |
27 | #include <sys/types.h> |
28 | #include <sys/param.h> | |
29 | #include <sys/time.h> | |
30 | #include <sys/systm.h> | |
31 | #include <sys/sysmacros.h> | |
32 | #include <sys/resource.h> | |
33 | #include <sys/vfs.h> | |
34 | #include <sys/vnode.h> | |
35 | #include <sys/file.h> | |
36 | #include <sys/mode.h> | |
37 | #include <sys/kmem.h> | |
38 | #include <sys/uio.h> | |
39 | #include <sys/pathname.h> | |
40 | #include <sys/cmn_err.h> | |
41 | #include <sys/errno.h> | |
42 | #include <sys/stat.h> | |
43 | #include <sys/unistd.h> | |
44 | #include <sys/sunddi.h> | |
45 | #include <sys/random.h> | |
46 | #include <sys/policy.h> | |
47 | #include <sys/zfs_dir.h> | |
48 | #include <sys/zfs_acl.h> | |
ea04106b | 49 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
50 | #include <sys/fs/zfs.h> |
51 | #include "fs/fs_subr.h" | |
52 | #include <sys/zap.h> | |
53 | #include <sys/dmu.h> | |
54 | #include <sys/atomic.h> | |
ebe7e575 | 55 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 56 | #include <sys/zfs_fuid.h> |
428870ff BB |
57 | #include <sys/sa.h> |
58 | #include <sys/zfs_sa.h> | |
34dc7c2f BB |
59 | #include <sys/dnlc.h> |
60 | #include <sys/extdirent.h> | |
61 | ||
62 | /* | |
63 | * zfs_match_find() is used by zfs_dirent_lock() to peform zap lookups | |
64 | * of names after deciding which is the appropriate lookup interface. | |
65 | */ | |
66 | static int | |
3558fd73 | 67 | zfs_match_find(zfs_sb_t *zsb, znode_t *dzp, char *name, boolean_t exact, |
34dc7c2f BB |
68 | boolean_t update, int *deflags, pathname_t *rpnp, uint64_t *zoid) |
69 | { | |
0c5dde49 | 70 | boolean_t conflict = B_FALSE; |
34dc7c2f BB |
71 | int error; |
72 | ||
3558fd73 | 73 | if (zsb->z_norm) { |
34dc7c2f | 74 | matchtype_t mt = MT_FIRST; |
34dc7c2f BB |
75 | size_t bufsz = 0; |
76 | char *buf = NULL; | |
77 | ||
78 | if (rpnp) { | |
79 | buf = rpnp->pn_buf; | |
80 | bufsz = rpnp->pn_bufsize; | |
81 | } | |
82 | if (exact) | |
83 | mt = MT_EXACT; | |
84 | /* | |
85 | * In the non-mixed case we only expect there would ever | |
86 | * be one match, but we need to use the normalizing lookup. | |
87 | */ | |
3558fd73 | 88 | error = zap_lookup_norm(zsb->z_os, dzp->z_id, name, 8, 1, |
34dc7c2f | 89 | zoid, mt, buf, bufsz, &conflict); |
34dc7c2f | 90 | } else { |
3558fd73 | 91 | error = zap_lookup(zsb->z_os, dzp->z_id, name, 8, 1, zoid); |
34dc7c2f | 92 | } |
0c5dde49 BB |
93 | |
94 | /* | |
95 | * Allow multiple entries provided the first entry is | |
96 | * the object id. Non-zpl consumers may safely make | |
97 | * use of the additional space. | |
98 | * | |
99 | * XXX: This should be a feature flag for compatibility | |
100 | */ | |
101 | if (error == EOVERFLOW) | |
102 | error = 0; | |
103 | ||
104 | if (zsb->z_norm && !error && deflags) | |
105 | *deflags = conflict ? ED_CASE_CONFLICT : 0; | |
106 | ||
34dc7c2f BB |
107 | *zoid = ZFS_DIRENT_OBJ(*zoid); |
108 | ||
3558fd73 | 109 | #ifdef HAVE_DNLC |
34dc7c2f | 110 | if (error == ENOENT && update) |
3558fd73 BB |
111 | dnlc_update(ZTOI(dzp), name, DNLC_NO_VNODE); |
112 | #endif /* HAVE_DNLC */ | |
34dc7c2f BB |
113 | |
114 | return (error); | |
115 | } | |
116 | ||
117 | /* | |
118 | * Lock a directory entry. A dirlock on <dzp, name> protects that name | |
119 | * in dzp's directory zap object. As long as you hold a dirlock, you can | |
120 | * assume two things: (1) dzp cannot be reaped, and (2) no other thread | |
121 | * can change the zap entry for (i.e. link or unlink) this name. | |
122 | * | |
123 | * Input arguments: | |
124 | * dzp - znode for directory | |
125 | * name - name of entry to lock | |
126 | * flag - ZNEW: if the entry already exists, fail with EEXIST. | |
127 | * ZEXISTS: if the entry does not exist, fail with ENOENT. | |
128 | * ZSHARED: allow concurrent access with other ZSHARED callers. | |
129 | * ZXATTR: we want dzp's xattr directory | |
130 | * ZCILOOK: On a mixed sensitivity file system, | |
131 | * this lookup should be case-insensitive. | |
132 | * ZCIEXACT: On a purely case-insensitive file system, | |
133 | * this lookup should be case-sensitive. | |
134 | * ZRENAMING: we are locking for renaming, force narrow locks | |
428870ff BB |
135 | * ZHAVELOCK: Don't grab the z_name_lock for this call. The |
136 | * current thread already holds it. | |
34dc7c2f BB |
137 | * |
138 | * Output arguments: | |
139 | * zpp - pointer to the znode for the entry (NULL if there isn't one) | |
140 | * dlpp - pointer to the dirlock for this entry (NULL on error) | |
141 | * direntflags - (case-insensitive lookup only) | |
142 | * flags if multiple case-sensitive matches exist in directory | |
143 | * realpnp - (case-insensitive lookup only) | |
144 | * actual name matched within the directory | |
145 | * | |
146 | * Return value: 0 on success or errno on failure. | |
147 | * | |
148 | * NOTE: Always checks for, and rejects, '.' and '..'. | |
149 | * NOTE: For case-insensitive file systems we take wide locks (see below), | |
150 | * but return znode pointers to a single match. | |
151 | */ | |
152 | int | |
153 | zfs_dirent_lock(zfs_dirlock_t **dlpp, znode_t *dzp, char *name, znode_t **zpp, | |
154 | int flag, int *direntflags, pathname_t *realpnp) | |
155 | { | |
3558fd73 | 156 | zfs_sb_t *zsb = ZTOZSB(dzp); |
34dc7c2f BB |
157 | zfs_dirlock_t *dl; |
158 | boolean_t update; | |
159 | boolean_t exact; | |
160 | uint64_t zoid; | |
3558fd73 | 161 | #ifdef HAVE_DNLC |
34dc7c2f | 162 | vnode_t *vp = NULL; |
3558fd73 | 163 | #endif /* HAVE_DNLC */ |
34dc7c2f BB |
164 | int error = 0; |
165 | int cmpflags; | |
166 | ||
167 | *zpp = NULL; | |
168 | *dlpp = NULL; | |
169 | ||
170 | /* | |
171 | * Verify that we are not trying to lock '.', '..', or '.zfs' | |
172 | */ | |
149e873a BB |
173 | if ((name[0] == '.' && |
174 | (name[1] == '\0' || (name[1] == '.' && name[2] == '\0'))) || | |
175 | (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0)) | |
a08ee875 | 176 | return (SET_ERROR(EEXIST)); |
34dc7c2f BB |
177 | |
178 | /* | |
179 | * Case sensitivity and normalization preferences are set when | |
180 | * the file system is created. These are stored in the | |
3558fd73 | 181 | * zsb->z_case and zsb->z_norm fields. These choices |
34dc7c2f BB |
182 | * affect what vnodes can be cached in the DNLC, how we |
183 | * perform zap lookups, and the "width" of our dirlocks. | |
184 | * | |
185 | * A normal dirlock locks a single name. Note that with | |
186 | * normalization a name can be composed multiple ways, but | |
187 | * when normalized, these names all compare equal. A wide | |
188 | * dirlock locks multiple names. We need these when the file | |
189 | * system is supporting mixed-mode access. It is sometimes | |
190 | * necessary to lock all case permutations of file name at | |
191 | * once so that simultaneous case-insensitive/case-sensitive | |
192 | * behaves as rationally as possible. | |
193 | */ | |
194 | ||
195 | /* | |
196 | * Decide if exact matches should be requested when performing | |
197 | * a zap lookup on file systems supporting case-insensitive | |
198 | * access. | |
199 | */ | |
200 | exact = | |
3558fd73 BB |
201 | ((zsb->z_case == ZFS_CASE_INSENSITIVE) && (flag & ZCIEXACT)) || |
202 | ((zsb->z_case == ZFS_CASE_MIXED) && !(flag & ZCILOOK)); | |
34dc7c2f BB |
203 | |
204 | /* | |
205 | * Only look in or update the DNLC if we are looking for the | |
206 | * name on a file system that does not require normalization | |
207 | * or case folding. We can also look there if we happen to be | |
208 | * on a non-normalizing, mixed sensitivity file system IF we | |
209 | * are looking for the exact name. | |
210 | * | |
211 | * Maybe can add TO-UPPERed version of name to dnlc in ci-only | |
212 | * case for performance improvement? | |
213 | */ | |
3558fd73 BB |
214 | update = !zsb->z_norm || |
215 | ((zsb->z_case == ZFS_CASE_MIXED) && | |
216 | !(zsb->z_norm & ~U8_TEXTPREP_TOUPPER) && !(flag & ZCILOOK)); | |
34dc7c2f BB |
217 | |
218 | /* | |
219 | * ZRENAMING indicates we are in a situation where we should | |
220 | * take narrow locks regardless of the file system's | |
221 | * preferences for normalizing and case folding. This will | |
222 | * prevent us deadlocking trying to grab the same wide lock | |
223 | * twice if the two names happen to be case-insensitive | |
224 | * matches. | |
225 | */ | |
226 | if (flag & ZRENAMING) | |
227 | cmpflags = 0; | |
228 | else | |
3558fd73 | 229 | cmpflags = zsb->z_norm; |
34dc7c2f BB |
230 | |
231 | /* | |
232 | * Wait until there are no locks on this name. | |
428870ff BB |
233 | * |
234 | * Don't grab the the lock if it is already held. However, cannot | |
235 | * have both ZSHARED and ZHAVELOCK together. | |
34dc7c2f | 236 | */ |
428870ff BB |
237 | ASSERT(!(flag & ZSHARED) || !(flag & ZHAVELOCK)); |
238 | if (!(flag & ZHAVELOCK)) | |
239 | rw_enter(&dzp->z_name_lock, RW_READER); | |
240 | ||
34dc7c2f BB |
241 | mutex_enter(&dzp->z_lock); |
242 | for (;;) { | |
243 | if (dzp->z_unlinked) { | |
244 | mutex_exit(&dzp->z_lock); | |
428870ff BB |
245 | if (!(flag & ZHAVELOCK)) |
246 | rw_exit(&dzp->z_name_lock); | |
a08ee875 | 247 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
248 | } |
249 | for (dl = dzp->z_dirlocks; dl != NULL; dl = dl->dl_next) { | |
250 | if ((u8_strcmp(name, dl->dl_name, 0, cmpflags, | |
251 | U8_UNICODE_LATEST, &error) == 0) || error != 0) | |
252 | break; | |
253 | } | |
254 | if (error != 0) { | |
255 | mutex_exit(&dzp->z_lock); | |
428870ff BB |
256 | if (!(flag & ZHAVELOCK)) |
257 | rw_exit(&dzp->z_name_lock); | |
a08ee875 | 258 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
259 | } |
260 | if (dl == NULL) { | |
261 | /* | |
262 | * Allocate a new dirlock and add it to the list. | |
263 | */ | |
264 | dl = kmem_alloc(sizeof (zfs_dirlock_t), KM_SLEEP); | |
265 | cv_init(&dl->dl_cv, NULL, CV_DEFAULT, NULL); | |
266 | dl->dl_name = name; | |
267 | dl->dl_sharecnt = 0; | |
428870ff | 268 | dl->dl_namelock = 0; |
34dc7c2f BB |
269 | dl->dl_namesize = 0; |
270 | dl->dl_dzp = dzp; | |
271 | dl->dl_next = dzp->z_dirlocks; | |
272 | dzp->z_dirlocks = dl; | |
273 | break; | |
274 | } | |
275 | if ((flag & ZSHARED) && dl->dl_sharecnt != 0) | |
276 | break; | |
277 | cv_wait(&dl->dl_cv, &dzp->z_lock); | |
278 | } | |
279 | ||
428870ff BB |
280 | /* |
281 | * If the z_name_lock was NOT held for this dirlock record it. | |
282 | */ | |
283 | if (flag & ZHAVELOCK) | |
284 | dl->dl_namelock = 1; | |
285 | ||
34dc7c2f BB |
286 | if ((flag & ZSHARED) && ++dl->dl_sharecnt > 1 && dl->dl_namesize == 0) { |
287 | /* | |
288 | * We're the second shared reference to dl. Make a copy of | |
289 | * dl_name in case the first thread goes away before we do. | |
290 | * Note that we initialize the new name before storing its | |
291 | * pointer into dl_name, because the first thread may load | |
292 | * dl->dl_name at any time. He'll either see the old value, | |
293 | * which is his, or the new shared copy; either is OK. | |
294 | */ | |
295 | dl->dl_namesize = strlen(dl->dl_name) + 1; | |
296 | name = kmem_alloc(dl->dl_namesize, KM_SLEEP); | |
297 | bcopy(dl->dl_name, name, dl->dl_namesize); | |
298 | dl->dl_name = name; | |
299 | } | |
300 | ||
301 | mutex_exit(&dzp->z_lock); | |
302 | ||
303 | /* | |
304 | * We have a dirlock on the name. (Note that it is the dirlock, | |
305 | * not the dzp's z_lock, that protects the name in the zap object.) | |
306 | * See if there's an object by this name; if so, put a hold on it. | |
307 | */ | |
308 | if (flag & ZXATTR) { | |
3558fd73 | 309 | error = sa_lookup(dzp->z_sa_hdl, SA_ZPL_XATTR(zsb), &zoid, |
428870ff BB |
310 | sizeof (zoid)); |
311 | if (error == 0) | |
a08ee875 | 312 | error = (zoid == 0 ? SET_ERROR(ENOENT) : 0); |
34dc7c2f | 313 | } else { |
3558fd73 | 314 | #ifdef HAVE_DNLC |
34dc7c2f | 315 | if (update) |
3558fd73 | 316 | vp = dnlc_lookup(ZTOI(dzp), name); |
34dc7c2f | 317 | if (vp == DNLC_NO_VNODE) { |
3558fd73 | 318 | iput(vp); |
a08ee875 | 319 | error = SET_ERROR(ENOENT); |
34dc7c2f BB |
320 | } else if (vp) { |
321 | if (flag & ZNEW) { | |
322 | zfs_dirent_unlock(dl); | |
3558fd73 | 323 | iput(vp); |
a08ee875 | 324 | return (SET_ERROR(EEXIST)); |
34dc7c2f BB |
325 | } |
326 | *dlpp = dl; | |
327 | *zpp = VTOZ(vp); | |
328 | return (0); | |
329 | } else { | |
3558fd73 | 330 | error = zfs_match_find(zsb, dzp, name, exact, |
34dc7c2f BB |
331 | update, direntflags, realpnp, &zoid); |
332 | } | |
3558fd73 BB |
333 | #else |
334 | error = zfs_match_find(zsb, dzp, name, exact, | |
335 | update, direntflags, realpnp, &zoid); | |
336 | #endif /* HAVE_DNLC */ | |
34dc7c2f BB |
337 | } |
338 | if (error) { | |
339 | if (error != ENOENT || (flag & ZEXISTS)) { | |
340 | zfs_dirent_unlock(dl); | |
341 | return (error); | |
342 | } | |
343 | } else { | |
344 | if (flag & ZNEW) { | |
345 | zfs_dirent_unlock(dl); | |
a08ee875 | 346 | return (SET_ERROR(EEXIST)); |
34dc7c2f | 347 | } |
3558fd73 | 348 | error = zfs_zget(zsb, zoid, zpp); |
34dc7c2f BB |
349 | if (error) { |
350 | zfs_dirent_unlock(dl); | |
351 | return (error); | |
352 | } | |
3558fd73 | 353 | #ifdef HAVE_DNLC |
34dc7c2f | 354 | if (!(flag & ZXATTR) && update) |
3558fd73 BB |
355 | dnlc_update(ZTOI(dzp), name, ZTOI(*zpp)); |
356 | #endif /* HAVE_DNLC */ | |
34dc7c2f BB |
357 | } |
358 | ||
359 | *dlpp = dl; | |
360 | ||
361 | return (0); | |
362 | } | |
363 | ||
364 | /* | |
365 | * Unlock this directory entry and wake anyone who was waiting for it. | |
366 | */ | |
367 | void | |
368 | zfs_dirent_unlock(zfs_dirlock_t *dl) | |
369 | { | |
370 | znode_t *dzp = dl->dl_dzp; | |
371 | zfs_dirlock_t **prev_dl, *cur_dl; | |
372 | ||
373 | mutex_enter(&dzp->z_lock); | |
428870ff BB |
374 | |
375 | if (!dl->dl_namelock) | |
376 | rw_exit(&dzp->z_name_lock); | |
377 | ||
34dc7c2f BB |
378 | if (dl->dl_sharecnt > 1) { |
379 | dl->dl_sharecnt--; | |
380 | mutex_exit(&dzp->z_lock); | |
381 | return; | |
382 | } | |
383 | prev_dl = &dzp->z_dirlocks; | |
384 | while ((cur_dl = *prev_dl) != dl) | |
385 | prev_dl = &cur_dl->dl_next; | |
386 | *prev_dl = dl->dl_next; | |
387 | cv_broadcast(&dl->dl_cv); | |
388 | mutex_exit(&dzp->z_lock); | |
389 | ||
390 | if (dl->dl_namesize != 0) | |
391 | kmem_free(dl->dl_name, dl->dl_namesize); | |
392 | cv_destroy(&dl->dl_cv); | |
393 | kmem_free(dl, sizeof (*dl)); | |
394 | } | |
395 | ||
396 | /* | |
397 | * Look up an entry in a directory. | |
398 | * | |
399 | * NOTE: '.' and '..' are handled as special cases because | |
400 | * no directory entries are actually stored for them. If this is | |
401 | * the root of a filesystem, then '.zfs' is also treated as a | |
402 | * special pseudo-directory. | |
403 | */ | |
404 | int | |
3558fd73 | 405 | zfs_dirlook(znode_t *dzp, char *name, struct inode **ipp, int flags, |
34dc7c2f BB |
406 | int *deflg, pathname_t *rpnp) |
407 | { | |
408 | zfs_dirlock_t *dl; | |
409 | znode_t *zp; | |
410 | int error = 0; | |
428870ff | 411 | uint64_t parent; |
34dc7c2f BB |
412 | |
413 | if (name[0] == 0 || (name[0] == '.' && name[1] == 0)) { | |
3558fd73 BB |
414 | *ipp = ZTOI(dzp); |
415 | igrab(*ipp); | |
34dc7c2f | 416 | } else if (name[0] == '.' && name[1] == '.' && name[2] == 0) { |
3558fd73 | 417 | zfs_sb_t *zsb = ZTOZSB(dzp); |
428870ff | 418 | |
34dc7c2f BB |
419 | /* |
420 | * If we are a snapshot mounted under .zfs, return | |
ebe7e575 | 421 | * the inode pointer for the snapshot directory. |
34dc7c2f | 422 | */ |
428870ff | 423 | if ((error = sa_lookup(dzp->z_sa_hdl, |
3558fd73 | 424 | SA_ZPL_PARENT(zsb), &parent, sizeof (parent))) != 0) |
428870ff | 425 | return (error); |
ebe7e575 | 426 | |
3558fd73 BB |
427 | if (parent == dzp->z_id && zsb->z_parent != zsb) { |
428 | error = zfsctl_root_lookup(zsb->z_parent->z_ctldir, | |
ebe7e575 | 429 | "snapshot", ipp, 0, kcred, NULL, NULL); |
34dc7c2f BB |
430 | return (error); |
431 | } | |
432 | rw_enter(&dzp->z_parent_lock, RW_READER); | |
3558fd73 | 433 | error = zfs_zget(zsb, parent, &zp); |
34dc7c2f | 434 | if (error == 0) |
3558fd73 | 435 | *ipp = ZTOI(zp); |
34dc7c2f BB |
436 | rw_exit(&dzp->z_parent_lock); |
437 | } else if (zfs_has_ctldir(dzp) && strcmp(name, ZFS_CTLDIR_NAME) == 0) { | |
3558fd73 | 438 | *ipp = zfsctl_root(dzp); |
34dc7c2f BB |
439 | } else { |
440 | int zf; | |
441 | ||
442 | zf = ZEXISTS | ZSHARED; | |
443 | if (flags & FIGNORECASE) | |
444 | zf |= ZCILOOK; | |
445 | ||
446 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zf, deflg, rpnp); | |
447 | if (error == 0) { | |
3558fd73 | 448 | *ipp = ZTOI(zp); |
34dc7c2f BB |
449 | zfs_dirent_unlock(dl); |
450 | dzp->z_zn_prefetch = B_TRUE; /* enable prefetching */ | |
451 | } | |
452 | rpnp = NULL; | |
453 | } | |
454 | ||
455 | if ((flags & FIGNORECASE) && rpnp && !error) | |
456 | (void) strlcpy(rpnp->pn_buf, name, rpnp->pn_bufsize); | |
457 | ||
458 | return (error); | |
459 | } | |
460 | ||
34dc7c2f BB |
461 | /* |
462 | * unlinked Set (formerly known as the "delete queue") Error Handling | |
463 | * | |
464 | * When dealing with the unlinked set, we dmu_tx_hold_zap(), but we | |
465 | * don't specify the name of the entry that we will be manipulating. We | |
466 | * also fib and say that we won't be adding any new entries to the | |
467 | * unlinked set, even though we might (this is to lower the minimum file | |
468 | * size that can be deleted in a full filesystem). So on the small | |
469 | * chance that the nlink list is using a fat zap (ie. has more than | |
470 | * 2000 entries), we *may* not pre-read a block that's needed. | |
471 | * Therefore it is remotely possible for some of the assertions | |
472 | * regarding the unlinked set below to fail due to i/o error. On a | |
473 | * nondebug system, this will result in the space being leaked. | |
474 | */ | |
475 | void | |
476 | zfs_unlinked_add(znode_t *zp, dmu_tx_t *tx) | |
477 | { | |
3558fd73 | 478 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
479 | |
480 | ASSERT(zp->z_unlinked); | |
428870ff | 481 | ASSERT(zp->z_links == 0); |
34dc7c2f | 482 | |
b128c09f | 483 | VERIFY3U(0, ==, |
3558fd73 | 484 | zap_add_int(zsb->z_os, zsb->z_unlinkedobj, zp->z_id, tx)); |
34dc7c2f BB |
485 | } |
486 | ||
ea04106b AX |
487 | /* |
488 | * Clean up any znodes that had no links when we either crashed or | |
489 | * (force) umounted the file system. | |
490 | */ | |
491 | void | |
492 | zfs_unlinked_drain(zfs_sb_t *zsb) | |
493 | { | |
494 | zap_cursor_t zc; | |
495 | zap_attribute_t zap; | |
496 | dmu_object_info_t doi; | |
497 | znode_t *zp; | |
498 | int error; | |
499 | ||
500 | /* | |
501 | * Iterate over the contents of the unlinked set. | |
502 | */ | |
503 | for (zap_cursor_init(&zc, zsb->z_os, zsb->z_unlinkedobj); | |
504 | zap_cursor_retrieve(&zc, &zap) == 0; | |
505 | zap_cursor_advance(&zc)) { | |
506 | ||
507 | /* | |
508 | * See what kind of object we have in list | |
509 | */ | |
510 | ||
511 | error = dmu_object_info(zsb->z_os, zap.za_first_integer, &doi); | |
512 | if (error != 0) | |
513 | continue; | |
514 | ||
515 | ASSERT((doi.doi_type == DMU_OT_PLAIN_FILE_CONTENTS) || | |
516 | (doi.doi_type == DMU_OT_DIRECTORY_CONTENTS)); | |
517 | /* | |
518 | * We need to re-mark these list entries for deletion, | |
519 | * so we pull them back into core and set zp->z_unlinked. | |
520 | */ | |
521 | error = zfs_zget(zsb, zap.za_first_integer, &zp); | |
522 | ||
523 | /* | |
524 | * We may pick up znodes that are already marked for deletion. | |
525 | * This could happen during the purge of an extended attribute | |
526 | * directory. All we need to do is skip over them, since they | |
527 | * are already in the system marked z_unlinked. | |
528 | */ | |
529 | if (error != 0) | |
530 | continue; | |
531 | ||
532 | zp->z_unlinked = B_TRUE; | |
533 | iput(ZTOI(zp)); | |
534 | } | |
535 | zap_cursor_fini(&zc); | |
536 | } | |
537 | ||
34dc7c2f BB |
538 | /* |
539 | * Delete the entire contents of a directory. Return a count | |
540 | * of the number of entries that could not be deleted. If we encounter | |
541 | * an error, return a count of at least one so that the directory stays | |
542 | * in the unlinked set. | |
543 | * | |
544 | * NOTE: this function assumes that the directory is inactive, | |
545 | * so there is no need to lock its entries before deletion. | |
546 | * Also, it assumes the directory contents is *only* regular | |
547 | * files. | |
548 | */ | |
549 | static int | |
550 | zfs_purgedir(znode_t *dzp) | |
551 | { | |
552 | zap_cursor_t zc; | |
553 | zap_attribute_t zap; | |
554 | znode_t *xzp; | |
555 | dmu_tx_t *tx; | |
3558fd73 | 556 | zfs_sb_t *zsb = ZTOZSB(dzp); |
34dc7c2f BB |
557 | zfs_dirlock_t dl; |
558 | int skipped = 0; | |
559 | int error; | |
560 | ||
3558fd73 | 561 | for (zap_cursor_init(&zc, zsb->z_os, dzp->z_id); |
34dc7c2f BB |
562 | (error = zap_cursor_retrieve(&zc, &zap)) == 0; |
563 | zap_cursor_advance(&zc)) { | |
3558fd73 | 564 | error = zfs_zget(zsb, |
34dc7c2f BB |
565 | ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp); |
566 | if (error) { | |
567 | skipped += 1; | |
568 | continue; | |
569 | } | |
570 | ||
ea04106b AX |
571 | ASSERT(S_ISREG(ZTOI(xzp)->i_mode) || |
572 | S_ISLNK(ZTOI(xzp)->i_mode)); | |
34dc7c2f | 573 | |
3558fd73 | 574 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 575 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
34dc7c2f | 576 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name); |
428870ff | 577 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); |
3558fd73 | 578 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
579 | /* Is this really needed ? */ |
580 | zfs_sa_upgrade_txholds(tx, xzp); | |
34dc7c2f BB |
581 | error = dmu_tx_assign(tx, TXG_WAIT); |
582 | if (error) { | |
583 | dmu_tx_abort(tx); | |
ea04106b | 584 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
585 | skipped += 1; |
586 | continue; | |
587 | } | |
588 | bzero(&dl, sizeof (dl)); | |
589 | dl.dl_dzp = dzp; | |
590 | dl.dl_name = zap.za_name; | |
591 | ||
592 | error = zfs_link_destroy(&dl, xzp, tx, 0, NULL); | |
593 | if (error) | |
594 | skipped += 1; | |
595 | dmu_tx_commit(tx); | |
87dac73d | 596 | set_nlink(ZTOI(xzp), xzp->z_links); |
ea04106b | 597 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
598 | } |
599 | zap_cursor_fini(&zc); | |
600 | if (error != ENOENT) | |
601 | skipped += 1; | |
602 | return (skipped); | |
603 | } | |
604 | ||
605 | void | |
606 | zfs_rmnode(znode_t *zp) | |
607 | { | |
3558fd73 BB |
608 | zfs_sb_t *zsb = ZTOZSB(zp); |
609 | objset_t *os = zsb->z_os; | |
34dc7c2f | 610 | znode_t *xzp = NULL; |
34dc7c2f BB |
611 | dmu_tx_t *tx; |
612 | uint64_t acl_obj; | |
428870ff | 613 | uint64_t xattr_obj; |
34dc7c2f BB |
614 | int error; |
615 | ||
428870ff | 616 | ASSERT(zp->z_links == 0); |
3558fd73 | 617 | ASSERT(atomic_read(&ZTOI(zp)->i_count) == 0); |
34dc7c2f BB |
618 | |
619 | /* | |
620 | * If this is an attribute directory, purge its contents. | |
621 | */ | |
3558fd73 | 622 | if (S_ISDIR(ZTOI(zp)->i_mode) && (zp->z_pflags & ZFS_XATTR)) { |
ea04106b | 623 | if (zfs_purgedir(zp) != 0) { |
34dc7c2f | 624 | /* |
ea04106b AX |
625 | * Not enough space to delete some xattrs. |
626 | * Leave it in the unlinked set. | |
34dc7c2f | 627 | */ |
7973e464 | 628 | zfs_znode_dmu_fini(zp); |
ea04106b | 629 | |
34dc7c2f BB |
630 | return; |
631 | } | |
632 | } | |
633 | ||
b128c09f | 634 | /* |
94a40997 AX |
635 | * Free up all the data in the file. We don't do this for directories |
636 | * because we need truncate and remove to be in the same tx, like in | |
637 | * zfs_znode_delete(). Otherwise, if we crash here we'll end up with | |
638 | * an inconsistent truncated zap object in the delete queue. Note a | |
639 | * truncated file is harmless since it only contains user data. | |
b128c09f | 640 | */ |
94a40997 AX |
641 | if (S_ISREG(ZTOI(zp)->i_mode)) { |
642 | error = dmu_free_long_range(os, zp->z_id, 0, DMU_OBJECT_END); | |
643 | if (error) { | |
644 | /* | |
645 | * Not enough space. Leave the file in the unlinked | |
646 | * set. | |
647 | */ | |
648 | zfs_znode_dmu_fini(zp); | |
649 | return; | |
650 | } | |
b128c09f BB |
651 | } |
652 | ||
34dc7c2f BB |
653 | /* |
654 | * If the file has extended attributes, we're going to unlink | |
655 | * the xattr dir. | |
656 | */ | |
3558fd73 | 657 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff BB |
658 | &xattr_obj, sizeof (xattr_obj)); |
659 | if (error == 0 && xattr_obj) { | |
3558fd73 | 660 | error = zfs_zget(zsb, xattr_obj, &xzp); |
34dc7c2f BB |
661 | ASSERT(error == 0); |
662 | } | |
663 | ||
572e2857 | 664 | acl_obj = zfs_external_acl(zp); |
34dc7c2f BB |
665 | |
666 | /* | |
b128c09f | 667 | * Set up the final transaction. |
34dc7c2f BB |
668 | */ |
669 | tx = dmu_tx_create(os); | |
670 | dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END); | |
3558fd73 | 671 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 672 | if (xzp) { |
3558fd73 | 673 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, TRUE, NULL); |
428870ff | 674 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); |
34dc7c2f BB |
675 | } |
676 | if (acl_obj) | |
677 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
428870ff BB |
678 | |
679 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
680 | error = dmu_tx_assign(tx, TXG_WAIT); |
681 | if (error) { | |
682 | /* | |
683 | * Not enough space to delete the file. Leave it in the | |
684 | * unlinked set, leaking it until the fs is remounted (at | |
685 | * which point we'll call zfs_unlinked_drain() to process it). | |
686 | */ | |
687 | dmu_tx_abort(tx); | |
688 | zfs_znode_dmu_fini(zp); | |
34dc7c2f BB |
689 | goto out; |
690 | } | |
691 | ||
692 | if (xzp) { | |
428870ff | 693 | ASSERT(error == 0); |
34dc7c2f BB |
694 | mutex_enter(&xzp->z_lock); |
695 | xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */ | |
428870ff | 696 | xzp->z_links = 0; /* no more links to it */ |
87dac73d | 697 | set_nlink(ZTOI(xzp), 0); /* this will let iput purge us */ |
3558fd73 | 698 | VERIFY(0 == sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zsb), |
428870ff | 699 | &xzp->z_links, sizeof (xzp->z_links), tx)); |
34dc7c2f BB |
700 | mutex_exit(&xzp->z_lock); |
701 | zfs_unlinked_add(xzp, tx); | |
702 | } | |
703 | ||
704 | /* Remove this znode from the unlinked set */ | |
b128c09f | 705 | VERIFY3U(0, ==, |
3558fd73 | 706 | zap_remove_int(zsb->z_os, zsb->z_unlinkedobj, zp->z_id, tx)); |
34dc7c2f BB |
707 | |
708 | zfs_znode_delete(zp, tx); | |
709 | ||
710 | dmu_tx_commit(tx); | |
711 | out: | |
712 | if (xzp) | |
ea04106b | 713 | zfs_iput_async(ZTOI(xzp)); |
34dc7c2f BB |
714 | } |
715 | ||
716 | static uint64_t | |
428870ff | 717 | zfs_dirent(znode_t *zp, uint64_t mode) |
34dc7c2f BB |
718 | { |
719 | uint64_t de = zp->z_id; | |
428870ff | 720 | |
3558fd73 | 721 | if (ZTOZSB(zp)->z_version >= ZPL_VERSION_DIRENT_TYPE) |
428870ff | 722 | de |= IFTODT(mode) << 60; |
34dc7c2f BB |
723 | return (de); |
724 | } | |
725 | ||
726 | /* | |
727 | * Link zp into dl. Can only fail if zp has been unlinked. | |
728 | */ | |
729 | int | |
730 | zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag) | |
731 | { | |
732 | znode_t *dzp = dl->dl_dzp; | |
3558fd73 | 733 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f | 734 | uint64_t value; |
3558fd73 | 735 | int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); |
428870ff BB |
736 | sa_bulk_attr_t bulk[5]; |
737 | uint64_t mtime[2], ctime[2]; | |
738 | int count = 0; | |
34dc7c2f BB |
739 | int error; |
740 | ||
34dc7c2f BB |
741 | mutex_enter(&zp->z_lock); |
742 | ||
743 | if (!(flag & ZRENAMING)) { | |
744 | if (zp->z_unlinked) { /* no new links to unlinked zp */ | |
745 | ASSERT(!(flag & (ZNEW | ZEXISTS))); | |
746 | mutex_exit(&zp->z_lock); | |
a08ee875 | 747 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 748 | } |
428870ff | 749 | zp->z_links++; |
3558fd73 | 750 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, |
428870ff BB |
751 | &zp->z_links, sizeof (zp->z_links)); |
752 | ||
34dc7c2f | 753 | } |
3558fd73 | 754 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, |
428870ff | 755 | &dzp->z_id, sizeof (dzp->z_id)); |
3558fd73 | 756 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff BB |
757 | &zp->z_pflags, sizeof (zp->z_pflags)); |
758 | ||
759 | if (!(flag & ZNEW)) { | |
3558fd73 | 760 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
761 | ctime, sizeof (ctime)); |
762 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, | |
763 | ctime, B_TRUE); | |
764 | } | |
765 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
766 | ASSERT(error == 0); | |
34dc7c2f | 767 | |
34dc7c2f BB |
768 | mutex_exit(&zp->z_lock); |
769 | ||
34dc7c2f | 770 | mutex_enter(&dzp->z_lock); |
428870ff BB |
771 | dzp->z_size++; |
772 | dzp->z_links += zp_is_dir; | |
773 | count = 0; | |
3558fd73 | 774 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, |
428870ff | 775 | &dzp->z_size, sizeof (dzp->z_size)); |
3558fd73 | 776 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, |
428870ff | 777 | &dzp->z_links, sizeof (dzp->z_links)); |
3558fd73 | 778 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, |
428870ff | 779 | mtime, sizeof (mtime)); |
3558fd73 | 780 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
428870ff | 781 | ctime, sizeof (ctime)); |
3558fd73 | 782 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff BB |
783 | &dzp->z_pflags, sizeof (dzp->z_pflags)); |
784 | zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); | |
785 | error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); | |
786 | ASSERT(error == 0); | |
34dc7c2f BB |
787 | mutex_exit(&dzp->z_lock); |
788 | ||
428870ff | 789 | value = zfs_dirent(zp, zp->z_mode); |
3558fd73 | 790 | error = zap_add(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name, |
34dc7c2f BB |
791 | 8, 1, &value, tx); |
792 | ASSERT(error == 0); | |
793 | ||
34dc7c2f BB |
794 | return (0); |
795 | } | |
796 | ||
428870ff BB |
797 | static int |
798 | zfs_dropname(zfs_dirlock_t *dl, znode_t *zp, znode_t *dzp, dmu_tx_t *tx, | |
799 | int flag) | |
800 | { | |
801 | int error; | |
802 | ||
3558fd73 BB |
803 | if (ZTOZSB(zp)->z_norm) { |
804 | if (((ZTOZSB(zp)->z_case == ZFS_CASE_INSENSITIVE) && | |
428870ff | 805 | (flag & ZCIEXACT)) || |
3558fd73 | 806 | ((ZTOZSB(zp)->z_case == ZFS_CASE_MIXED) && |
428870ff | 807 | !(flag & ZCILOOK))) |
3558fd73 | 808 | error = zap_remove_norm(ZTOZSB(zp)->z_os, |
428870ff BB |
809 | dzp->z_id, dl->dl_name, MT_EXACT, tx); |
810 | else | |
3558fd73 | 811 | error = zap_remove_norm(ZTOZSB(zp)->z_os, |
428870ff BB |
812 | dzp->z_id, dl->dl_name, MT_FIRST, tx); |
813 | } else { | |
3558fd73 | 814 | error = zap_remove(ZTOZSB(zp)->z_os, |
428870ff BB |
815 | dzp->z_id, dl->dl_name, tx); |
816 | } | |
817 | ||
818 | return (error); | |
819 | } | |
820 | ||
34dc7c2f | 821 | /* |
cd38ac58 BB |
822 | * Unlink zp from dl, and mark zp for deletion if this was the last link. Can |
823 | * fail if zp is a mount point (EBUSY) or a non-empty directory (ENOTEMPTY). | |
34dc7c2f BB |
824 | * If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list. |
825 | * If it's non-NULL, we use it to indicate whether the znode needs deletion, | |
826 | * and it's the caller's job to do it. | |
827 | */ | |
828 | int | |
829 | zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag, | |
830 | boolean_t *unlinkedp) | |
831 | { | |
832 | znode_t *dzp = dl->dl_dzp; | |
3558fd73 BB |
833 | zfs_sb_t *zsb = ZTOZSB(dzp); |
834 | int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); | |
34dc7c2f | 835 | boolean_t unlinked = B_FALSE; |
428870ff BB |
836 | sa_bulk_attr_t bulk[5]; |
837 | uint64_t mtime[2], ctime[2]; | |
838 | int count = 0; | |
34dc7c2f BB |
839 | int error; |
840 | ||
3558fd73 BB |
841 | #ifdef HAVE_DNLC |
842 | dnlc_remove(ZTOI(dzp), dl->dl_name); | |
843 | #endif /* HAVE_DNLC */ | |
34dc7c2f BB |
844 | |
845 | if (!(flag & ZRENAMING)) { | |
34dc7c2f | 846 | mutex_enter(&zp->z_lock); |
428870ff BB |
847 | |
848 | if (zp_is_dir && !zfs_dirempty(zp)) { | |
34dc7c2f | 849 | mutex_exit(&zp->z_lock); |
a08ee875 | 850 | return (SET_ERROR(ENOTEMPTY)); |
34dc7c2f | 851 | } |
428870ff BB |
852 | |
853 | /* | |
854 | * If we get here, we are going to try to remove the object. | |
855 | * First try removing the name from the directory; if that | |
856 | * fails, return the error. | |
857 | */ | |
858 | error = zfs_dropname(dl, zp, dzp, tx, flag); | |
859 | if (error != 0) { | |
860 | mutex_exit(&zp->z_lock); | |
428870ff BB |
861 | return (error); |
862 | } | |
863 | ||
864 | if (zp->z_links <= zp_is_dir) { | |
3558fd73 BB |
865 | zfs_panic_recover("zfs: link count on %lu is %u, " |
866 | "should be at least %u", zp->z_id, | |
867 | (int)zp->z_links, zp_is_dir + 1); | |
428870ff | 868 | zp->z_links = zp_is_dir + 1; |
34dc7c2f | 869 | } |
428870ff | 870 | if (--zp->z_links == zp_is_dir) { |
34dc7c2f | 871 | zp->z_unlinked = B_TRUE; |
428870ff | 872 | zp->z_links = 0; |
34dc7c2f BB |
873 | unlinked = B_TRUE; |
874 | } else { | |
3558fd73 | 875 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), |
428870ff | 876 | NULL, &ctime, sizeof (ctime)); |
3558fd73 | 877 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), |
428870ff BB |
878 | NULL, &zp->z_pflags, sizeof (zp->z_pflags)); |
879 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, | |
880 | B_TRUE); | |
34dc7c2f | 881 | } |
3558fd73 | 882 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), |
428870ff BB |
883 | NULL, &zp->z_links, sizeof (zp->z_links)); |
884 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
885 | count = 0; | |
886 | ASSERT(error == 0); | |
34dc7c2f | 887 | mutex_exit(&zp->z_lock); |
428870ff BB |
888 | } else { |
889 | error = zfs_dropname(dl, zp, dzp, tx, flag); | |
890 | if (error != 0) | |
891 | return (error); | |
34dc7c2f BB |
892 | } |
893 | ||
34dc7c2f | 894 | mutex_enter(&dzp->z_lock); |
428870ff BB |
895 | dzp->z_size--; /* one dirent removed */ |
896 | dzp->z_links -= zp_is_dir; /* ".." link from zp */ | |
3558fd73 | 897 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), |
428870ff | 898 | NULL, &dzp->z_links, sizeof (dzp->z_links)); |
3558fd73 | 899 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), |
428870ff | 900 | NULL, &dzp->z_size, sizeof (dzp->z_size)); |
3558fd73 | 901 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), |
428870ff | 902 | NULL, ctime, sizeof (ctime)); |
3558fd73 | 903 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), |
428870ff | 904 | NULL, mtime, sizeof (mtime)); |
3558fd73 | 905 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), |
428870ff BB |
906 | NULL, &dzp->z_pflags, sizeof (dzp->z_pflags)); |
907 | zfs_tstamp_update_setup(dzp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); | |
908 | error = sa_bulk_update(dzp->z_sa_hdl, bulk, count, tx); | |
34dc7c2f | 909 | ASSERT(error == 0); |
428870ff | 910 | mutex_exit(&dzp->z_lock); |
34dc7c2f BB |
911 | |
912 | if (unlinkedp != NULL) | |
913 | *unlinkedp = unlinked; | |
914 | else if (unlinked) | |
915 | zfs_unlinked_add(zp, tx); | |
916 | ||
917 | return (0); | |
918 | } | |
919 | ||
920 | /* | |
921 | * Indicate whether the directory is empty. Works with or without z_lock | |
922 | * held, but can only be consider a hint in the latter case. Returns true | |
923 | * if only "." and ".." remain and there's no work in progress. | |
924 | */ | |
925 | boolean_t | |
926 | zfs_dirempty(znode_t *dzp) | |
927 | { | |
428870ff | 928 | return (dzp->z_size == 2 && dzp->z_dirlocks == 0); |
34dc7c2f BB |
929 | } |
930 | ||
931 | int | |
3558fd73 | 932 | zfs_make_xattrdir(znode_t *zp, vattr_t *vap, struct inode **xipp, cred_t *cr) |
34dc7c2f | 933 | { |
3558fd73 | 934 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
935 | znode_t *xzp; |
936 | dmu_tx_t *tx; | |
937 | int error; | |
9babb374 BB |
938 | zfs_acl_ids_t acl_ids; |
939 | boolean_t fuid_dirtied; | |
3558fd73 | 940 | #ifdef DEBUG |
428870ff | 941 | uint64_t parent; |
3558fd73 | 942 | #endif |
34dc7c2f | 943 | |
3558fd73 | 944 | *xipp = NULL; |
34dc7c2f | 945 | |
3558fd73 | 946 | if ((error = zfs_zaccess(zp, ACE_WRITE_NAMED_ATTRS, 0, B_FALSE, cr))) |
34dc7c2f BB |
947 | return (error); |
948 | ||
9babb374 BB |
949 | if ((error = zfs_acl_ids_create(zp, IS_XATTR, vap, cr, NULL, |
950 | &acl_ids)) != 0) | |
951 | return (error); | |
3558fd73 | 952 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
9babb374 | 953 | zfs_acl_ids_free(&acl_ids); |
a08ee875 | 954 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
955 | } |
956 | ||
3558fd73 | 957 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
958 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
959 | ZFS_SA_BASE_ATTR_SIZE); | |
960 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
34dc7c2f | 961 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); |
3558fd73 | 962 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 963 | if (fuid_dirtied) |
3558fd73 | 964 | zfs_fuid_txhold(zsb, tx); |
a08ee875 | 965 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 966 | if (error) { |
428870ff | 967 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
968 | dmu_tx_abort(tx); |
969 | return (error); | |
970 | } | |
428870ff | 971 | zfs_mknode(zp, vap, tx, cr, IS_XATTR, &xzp, &acl_ids); |
9babb374 BB |
972 | |
973 | if (fuid_dirtied) | |
3558fd73 | 974 | zfs_fuid_sync(zsb, tx); |
9babb374 | 975 | |
428870ff | 976 | #ifdef DEBUG |
3558fd73 | 977 | error = sa_lookup(xzp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
428870ff BB |
978 | &parent, sizeof (parent)); |
979 | ASSERT(error == 0 && parent == zp->z_id); | |
980 | #endif | |
981 | ||
3558fd73 | 982 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), &xzp->z_id, |
428870ff | 983 | sizeof (xzp->z_id), tx)); |
34dc7c2f | 984 | |
3558fd73 | 985 | (void) zfs_log_create(zsb->z_log, tx, TX_MKXATTR, zp, |
9babb374 BB |
986 | xzp, "", NULL, acl_ids.z_fuidp, vap); |
987 | ||
988 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
989 | dmu_tx_commit(tx); |
990 | ||
3558fd73 | 991 | *xipp = ZTOI(xzp); |
34dc7c2f BB |
992 | |
993 | return (0); | |
994 | } | |
995 | ||
996 | /* | |
997 | * Return a znode for the extended attribute directory for zp. | |
998 | * ** If the directory does not already exist, it is created ** | |
999 | * | |
1000 | * IN: zp - znode to obtain attribute directory from | |
1001 | * cr - credentials of caller | |
1002 | * flags - flags from the VOP_LOOKUP call | |
1003 | * | |
3558fd73 | 1004 | * OUT: xipp - pointer to extended attribute znode |
34dc7c2f BB |
1005 | * |
1006 | * RETURN: 0 on success | |
1007 | * error number on failure | |
1008 | */ | |
1009 | int | |
3558fd73 | 1010 | zfs_get_xattrdir(znode_t *zp, struct inode **xipp, cred_t *cr, int flags) |
34dc7c2f | 1011 | { |
3558fd73 | 1012 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
1013 | znode_t *xzp; |
1014 | zfs_dirlock_t *dl; | |
1015 | vattr_t va; | |
1016 | int error; | |
1017 | top: | |
1018 | error = zfs_dirent_lock(&dl, zp, "", &xzp, ZXATTR, NULL, NULL); | |
1019 | if (error) | |
1020 | return (error); | |
1021 | ||
1022 | if (xzp != NULL) { | |
3558fd73 | 1023 | *xipp = ZTOI(xzp); |
34dc7c2f BB |
1024 | zfs_dirent_unlock(dl); |
1025 | return (0); | |
1026 | } | |
1027 | ||
34dc7c2f BB |
1028 | if (!(flags & CREATE_XATTR_DIR)) { |
1029 | zfs_dirent_unlock(dl); | |
a08ee875 | 1030 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1031 | } |
1032 | ||
2cf7f52b | 1033 | if (zfs_is_readonly(zsb)) { |
34dc7c2f | 1034 | zfs_dirent_unlock(dl); |
a08ee875 | 1035 | return (SET_ERROR(EROFS)); |
34dc7c2f BB |
1036 | } |
1037 | ||
1038 | /* | |
1039 | * The ability to 'create' files in an attribute | |
1040 | * directory comes from the write_xattr permission on the base file. | |
1041 | * | |
1042 | * The ability to 'search' an attribute directory requires | |
1043 | * read_xattr permission on the base file. | |
1044 | * | |
1045 | * Once in a directory the ability to read/write attributes | |
1046 | * is controlled by the permissions on the attribute file. | |
1047 | */ | |
3558fd73 | 1048 | va.va_mask = ATTR_MODE | ATTR_UID | ATTR_GID; |
34dc7c2f BB |
1049 | va.va_mode = S_IFDIR | S_ISVTX | 0777; |
1050 | zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid); | |
1051 | ||
274b7e79 | 1052 | va.va_dentry = NULL; |
3558fd73 | 1053 | error = zfs_make_xattrdir(zp, &va, xipp, cr); |
34dc7c2f BB |
1054 | zfs_dirent_unlock(dl); |
1055 | ||
fb5f0bc8 | 1056 | if (error == ERESTART) { |
34dc7c2f BB |
1057 | /* NB: we already did dmu_tx_wait() if necessary */ |
1058 | goto top; | |
1059 | } | |
1060 | ||
1061 | return (error); | |
1062 | } | |
1063 | ||
1064 | /* | |
1065 | * Decide whether it is okay to remove within a sticky directory. | |
1066 | * | |
1067 | * In sticky directories, write access is not sufficient; | |
1068 | * you can remove entries from a directory only if: | |
1069 | * | |
1070 | * you own the directory, | |
1071 | * you own the entry, | |
1072 | * the entry is a plain file and you have write access, | |
1073 | * or you are privileged (checked in secpolicy...). | |
1074 | * | |
1075 | * The function returns 0 if remove access is granted. | |
1076 | */ | |
1077 | int | |
1078 | zfs_sticky_remove_access(znode_t *zdp, znode_t *zp, cred_t *cr) | |
1079 | { | |
3558fd73 | 1080 | uid_t uid; |
572e2857 BB |
1081 | uid_t downer; |
1082 | uid_t fowner; | |
3558fd73 | 1083 | zfs_sb_t *zsb = ZTOZSB(zdp); |
34dc7c2f | 1084 | |
3558fd73 | 1085 | if (zsb->z_replay) |
34dc7c2f BB |
1086 | return (0); |
1087 | ||
428870ff | 1088 | if ((zdp->z_mode & S_ISVTX) == 0) |
34dc7c2f BB |
1089 | return (0); |
1090 | ||
3558fd73 BB |
1091 | downer = zfs_fuid_map_id(zsb, zdp->z_uid, cr, ZFS_OWNER); |
1092 | fowner = zfs_fuid_map_id(zsb, zp->z_uid, cr, ZFS_OWNER); | |
572e2857 BB |
1093 | |
1094 | if ((uid = crgetuid(cr)) == downer || uid == fowner || | |
3558fd73 | 1095 | (S_ISDIR(ZTOI(zp)->i_mode) && |
34dc7c2f BB |
1096 | zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr) == 0)) |
1097 | return (0); | |
1098 | else | |
1099 | return (secpolicy_vnode_remove(cr)); | |
1100 | } |