]>
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 | |
1d3ba0bf | 9 | * or https://opensource.org/licenses/CDDL-1.0. |
34dc7c2f BB |
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. |
a448a255 | 24 | * Copyright (c) 2012, 2018 by Delphix. All rights reserved. |
5475aada | 25 | * Copyright (c) 2015 by Chunwei Chen. All rights reserved. |
9b7b9cd3 | 26 | * Copyright 2017 Nexenta Systems, Inc. |
34dc7c2f BB |
27 | */ |
28 | ||
29 | /* Portions Copyright 2007 Jeremy Teo */ | |
428870ff | 30 | /* Portions Copyright 2010 Robert Milkowski */ |
34dc7c2f | 31 | |
60101509 | 32 | |
34dc7c2f BB |
33 | #include <sys/types.h> |
34 | #include <sys/param.h> | |
35 | #include <sys/time.h> | |
34dc7c2f | 36 | #include <sys/sysmacros.h> |
34dc7c2f | 37 | #include <sys/vfs.h> |
34dc7c2f BB |
38 | #include <sys/file.h> |
39 | #include <sys/stat.h> | |
40 | #include <sys/kmem.h> | |
41 | #include <sys/taskq.h> | |
42 | #include <sys/uio.h> | |
43 | #include <sys/vmsystm.h> | |
44 | #include <sys/atomic.h> | |
34dc7c2f BB |
45 | #include <sys/pathname.h> |
46 | #include <sys/cmn_err.h> | |
47 | #include <sys/errno.h> | |
34dc7c2f BB |
48 | #include <sys/zfs_dir.h> |
49 | #include <sys/zfs_acl.h> | |
50 | #include <sys/zfs_ioctl.h> | |
51 | #include <sys/fs/zfs.h> | |
52 | #include <sys/dmu.h> | |
428870ff | 53 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
54 | #include <sys/spa.h> |
55 | #include <sys/txg.h> | |
56 | #include <sys/dbuf.h> | |
57 | #include <sys/zap.h> | |
428870ff | 58 | #include <sys/sa.h> |
34dc7c2f BB |
59 | #include <sys/policy.h> |
60 | #include <sys/sunddi.h> | |
b128c09f | 61 | #include <sys/sid.h> |
ebe7e575 | 62 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 63 | #include <sys/zfs_fuid.h> |
957c7aa2 | 64 | #include <sys/zfs_quota.h> |
428870ff | 65 | #include <sys/zfs_sa.h> |
e5c39b95 | 66 | #include <sys/zfs_vnops.h> |
34dc7c2f | 67 | #include <sys/zfs_rlock.h> |
428870ff | 68 | #include <sys/cred.h> |
218b8eaf | 69 | #include <sys/zpl.h> |
1ce23dca | 70 | #include <sys/zil.h> |
9c5167d1 | 71 | #include <sys/sa_impl.h> |
34dc7c2f BB |
72 | |
73 | /* | |
74 | * Programming rules. | |
75 | * | |
76 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
77 | * properly lock its in-core state, create a DMU transaction, do the work, | |
78 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
79 | * and wait for the intent log to commit if it is a synchronous operation. | |
80 | * Moreover, the vnode ops must work in both normal and log replay context. | |
81 | * The ordering of events is important to avoid deadlocks and references | |
82 | * to freed memory. The example below illustrates the following Big Rules: | |
83 | * | |
84 | * (1) A check must be made in each zfs thread for a mounted file system. | |
768eaced CC |
85 | * This is done avoiding races using zfs_enter(zfsvfs). |
86 | * A zfs_exit(zfsvfs) is needed before all returns. Any znodes | |
87 | * must be checked with zfs_verify_zp(zp). Both of these macros | |
34dc7c2f BB |
88 | * can return EIO from the calling function. |
89 | * | |
2921ad6c | 90 | * (2) zrele() should always be the last thing except for zil_commit() (if |
768eaced | 91 | * necessary) and zfs_exit(). This is for 3 reasons: First, if it's the |
2921ad6c PD |
92 | * last reference, the vnode/znode can be freed, so the zp may point to |
93 | * freed memory. Second, the last reference will call zfs_zinactive(), | |
94 | * which may induce a lot of work -- pushing cached pages (which acquires | |
95 | * range locks) and syncing out cached atime changes. Third, | |
96 | * zfs_zinactive() may require a new tx, which could deadlock the system | |
97 | * if you were already holding one. This deadlock occurs because the tx | |
98 | * currently being operated on prevents a txg from syncing, which | |
99 | * prevents the new tx from progressing, resulting in a deadlock. If you | |
100 | * must call zrele() within a tx, use zfs_zrele_async(). Note that iput() | |
101 | * is a synonym for zrele(). | |
34dc7c2f BB |
102 | * |
103 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
104 | * as they can span dmu_tx_assign() calls. | |
105 | * | |
384f8a09 MA |
106 | * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to |
107 | * dmu_tx_assign(). This is critical because we don't want to block | |
108 | * while holding locks. | |
109 | * | |
768eaced | 110 | * If no ZPL locks are held (aside from zfs_enter()), use TXG_WAIT. This |
384f8a09 MA |
111 | * reduces lock contention and CPU usage when we must wait (note that if |
112 | * throughput is constrained by the storage, nearly every transaction | |
113 | * must wait). | |
114 | * | |
115 | * Note, in particular, that if a lock is sometimes acquired before | |
116 | * the tx assigns, and sometimes after (e.g. z_lock), then failing | |
117 | * to use a non-blocking assign can deadlock the system. The scenario: | |
34dc7c2f BB |
118 | * |
119 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
120 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
121 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
122 | * forever, because the previous txg can't quiesce until B's tx commits. | |
123 | * | |
0037b49e | 124 | * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, |
e8b96c60 | 125 | * then drop all locks, call dmu_tx_wait(), and try again. On subsequent |
0735ecb3 | 126 | * calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT, |
e8b96c60 MA |
127 | * to indicate that this operation has already called dmu_tx_wait(). |
128 | * This will ensure that we don't retry forever, waiting a short bit | |
129 | * each time. | |
34dc7c2f BB |
130 | * |
131 | * (5) If the operation succeeded, generate the intent log entry for it | |
132 | * before dropping locks. This ensures that the ordering of events | |
133 | * in the intent log matches the order in which they actually occurred. | |
d3cc8b15 | 134 | * During ZIL replay the zfs_log_* functions will update the sequence |
fb5f0bc8 | 135 | * number to indicate the zil transaction has replayed. |
34dc7c2f BB |
136 | * |
137 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
138 | * regardless of whether there were any errors. | |
139 | * | |
572e2857 | 140 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
141 | * to ensure that synchronous semantics are provided when necessary. |
142 | * | |
143 | * In general, this is how things should be ordered in each vnode op: | |
144 | * | |
768eaced | 145 | * zfs_enter(zfsvfs); // exit if unmounted |
34dc7c2f | 146 | * top: |
3558fd73 | 147 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
148 | * rw_enter(...); // grab any other locks you need |
149 | * tx = dmu_tx_create(...); // get DMU tx | |
150 | * dmu_tx_hold_*(); // hold each object you might modify | |
0735ecb3 | 151 | * error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
152 | * if (error) { |
153 | * rw_exit(...); // drop locks | |
154 | * zfs_dirent_unlock(dl); // unlock directory entry | |
657ce253 | 155 | * zrele(...); // release held znodes |
fb5f0bc8 | 156 | * if (error == ERESTART) { |
e8b96c60 | 157 | * waited = B_TRUE; |
34dc7c2f BB |
158 | * dmu_tx_wait(tx); |
159 | * dmu_tx_abort(tx); | |
160 | * goto top; | |
161 | * } | |
162 | * dmu_tx_abort(tx); // abort DMU tx | |
768eaced | 163 | * zfs_exit(zfsvfs); // finished in zfs |
34dc7c2f BB |
164 | * return (error); // really out of space |
165 | * } | |
166 | * error = do_real_work(); // do whatever this VOP does | |
167 | * if (error == 0) | |
168 | * zfs_log_*(...); // on success, make ZIL entry | |
169 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
170 | * rw_exit(...); // drop locks | |
171 | * zfs_dirent_unlock(dl); // unlock directory entry | |
657ce253 | 172 | * zrele(...); // release held znodes |
572e2857 | 173 | * zil_commit(zilog, foid); // synchronous when necessary |
768eaced | 174 | * zfs_exit(zfsvfs); // finished in zfs |
34dc7c2f BB |
175 | * return (error); // done, report error |
176 | */ | |
126400a1 BB |
177 | int |
178 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
179 | { | |
ef70eff1 | 180 | (void) cr; |
126400a1 | 181 | znode_t *zp = ITOZ(ip); |
0037b49e | 182 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
768eaced | 183 | int error; |
126400a1 | 184 | |
768eaced CC |
185 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
186 | return (error); | |
126400a1 BB |
187 | |
188 | /* Honor ZFS_APPENDONLY file attribute */ | |
43e8f6e3 | 189 | if (blk_mode_is_open_write(mode) && (zp->z_pflags & ZFS_APPENDONLY) && |
126400a1 | 190 | ((flag & O_APPEND) == 0)) { |
768eaced | 191 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 192 | return (SET_ERROR(EPERM)); |
126400a1 BB |
193 | } |
194 | ||
c3773de1 AM |
195 | /* |
196 | * Keep a count of the synchronous opens in the znode. On first | |
197 | * synchronous open we must convert all previous async transactions | |
198 | * into sync to keep correct ordering. | |
199 | */ | |
200 | if (flag & O_SYNC) { | |
201 | if (atomic_inc_32_nv(&zp->z_sync_cnt) == 1) | |
202 | zil_async_to_sync(zfsvfs->z_log, zp->z_id); | |
203 | } | |
126400a1 | 204 | |
768eaced | 205 | zfs_exit(zfsvfs, FTAG); |
126400a1 BB |
206 | return (0); |
207 | } | |
126400a1 | 208 | |
126400a1 BB |
209 | int |
210 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
211 | { | |
ef70eff1 | 212 | (void) cr; |
126400a1 | 213 | znode_t *zp = ITOZ(ip); |
0037b49e | 214 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
768eaced | 215 | int error; |
126400a1 | 216 | |
768eaced CC |
217 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
218 | return (error); | |
126400a1 | 219 | |
7dc71949 | 220 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 221 | if (flag & O_SYNC) |
7dc71949 | 222 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 | 223 | |
768eaced | 224 | zfs_exit(zfsvfs, FTAG); |
8780c539 | 225 | return (0); |
126400a1 | 226 | } |
126400a1 | 227 | |
c0d35759 | 228 | #if defined(_KERNEL) |
89cd2197 BB |
229 | |
230 | static int zfs_fillpage(struct inode *ip, struct page *pp); | |
231 | ||
34dc7c2f BB |
232 | /* |
233 | * When a file is memory mapped, we must keep the IO data synchronized | |
89cd2197 BB |
234 | * between the DMU cache and the memory mapped pages. Update all mapped |
235 | * pages with the contents of the coresponding dmu buffer. | |
34dc7c2f | 236 | */ |
e53d678d | 237 | void |
8a9634e2 | 238 | update_pages(znode_t *zp, int64_t start, int len, objset_t *os) |
34dc7c2f | 239 | { |
89cd2197 BB |
240 | struct address_space *mp = ZTOI(zp)->i_mapping; |
241 | int64_t off = start & (PAGE_SIZE - 1); | |
34dc7c2f | 242 | |
8b1899d3 | 243 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { |
89cd2197 | 244 | uint64_t nbytes = MIN(PAGE_SIZE - off, len); |
34dc7c2f | 245 | |
89cd2197 | 246 | struct page *pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
247 | if (pp) { |
248 | if (mapping_writably_mapped(mp)) | |
249 | flush_dcache_page(pp); | |
34dc7c2f | 250 | |
89cd2197 BB |
251 | void *pb = kmap(pp); |
252 | int error = dmu_read(os, zp->z_id, start + off, | |
253 | nbytes, pb + off, DMU_READ_PREFETCH); | |
c0d35759 BB |
254 | kunmap(pp); |
255 | ||
89cd2197 BB |
256 | if (error) { |
257 | SetPageError(pp); | |
258 | ClearPageUptodate(pp); | |
259 | } else { | |
260 | ClearPageError(pp); | |
261 | SetPageUptodate(pp); | |
262 | ||
263 | if (mapping_writably_mapped(mp)) | |
264 | flush_dcache_page(pp); | |
265 | ||
266 | mark_page_accessed(pp); | |
267 | } | |
c0d35759 | 268 | |
c0d35759 | 269 | unlock_page(pp); |
8b1899d3 | 270 | put_page(pp); |
34dc7c2f | 271 | } |
c0d35759 | 272 | |
d164b209 | 273 | len -= nbytes; |
34dc7c2f | 274 | off = 0; |
34dc7c2f | 275 | } |
34dc7c2f BB |
276 | } |
277 | ||
278 | /* | |
89cd2197 BB |
279 | * When a file is memory mapped, we must keep the I/O data synchronized |
280 | * between the DMU cache and the memory mapped pages. Preferentially read | |
281 | * from memory mapped pages, otherwise fallback to reading through the dmu. | |
34dc7c2f | 282 | */ |
e53d678d | 283 | int |
d0cd9a5c | 284 | mappedread(znode_t *zp, int nbytes, zfs_uio_t *uio) |
34dc7c2f | 285 | { |
e53d678d | 286 | struct inode *ip = ZTOI(zp); |
c0d35759 | 287 | struct address_space *mp = ip->i_mapping; |
89cd2197 BB |
288 | int64_t start = uio->uio_loffset; |
289 | int64_t off = start & (PAGE_SIZE - 1); | |
34dc7c2f BB |
290 | int len = nbytes; |
291 | int error = 0; | |
292 | ||
8b1899d3 | 293 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { |
89cd2197 | 294 | uint64_t bytes = MIN(PAGE_SIZE - off, len); |
c0d35759 | 295 | |
89cd2197 | 296 | struct page *pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 | 297 | if (pp) { |
89cd2197 BB |
298 | /* |
299 | * If filemap_fault() retries there exists a window | |
300 | * where the page will be unlocked and not up to date. | |
301 | * In this case we must try and fill the page. | |
302 | */ | |
303 | if (unlikely(!PageUptodate(pp))) { | |
304 | error = zfs_fillpage(ip, pp); | |
305 | if (error) { | |
306 | unlock_page(pp); | |
307 | put_page(pp); | |
308 | return (error); | |
309 | } | |
310 | } | |
311 | ||
312 | ASSERT(PageUptodate(pp) || PageDirty(pp)); | |
313 | ||
b2ab468d | 314 | unlock_page(pp); |
c0d35759 | 315 | |
89cd2197 | 316 | void *pb = kmap(pp); |
d0cd9a5c | 317 | error = zfs_uiomove(pb + off, bytes, UIO_READ, uio); |
c0d35759 BB |
318 | kunmap(pp); |
319 | ||
320 | if (mapping_writably_mapped(mp)) | |
321 | flush_dcache_page(pp); | |
322 | ||
323 | mark_page_accessed(pp); | |
8b1899d3 | 324 | put_page(pp); |
34dc7c2f | 325 | } else { |
804e0504 MA |
326 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
327 | uio, bytes); | |
34dc7c2f | 328 | } |
c0d35759 | 329 | |
34dc7c2f BB |
330 | len -= bytes; |
331 | off = 0; | |
89cd2197 | 332 | |
34dc7c2f BB |
333 | if (error) |
334 | break; | |
335 | } | |
89cd2197 | 336 | |
34dc7c2f BB |
337 | return (error); |
338 | } | |
c0d35759 | 339 | #endif /* _KERNEL */ |
34dc7c2f | 340 | |
18168da7 | 341 | static unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT; |
34dc7c2f | 342 | |
13a9a6f5 MM |
343 | /* |
344 | * Write the bytes to a file. | |
345 | * | |
346 | * IN: zp - znode of file to be written to | |
347 | * data - bytes to write | |
348 | * len - number of bytes to write | |
349 | * pos - offset to start writing at | |
350 | * | |
351 | * OUT: resid - remaining bytes to write | |
352 | * | |
353 | * RETURN: 0 if success | |
1c2358c1 BB |
354 | * positive error code if failure. EIO is returned |
355 | * for a short write when residp isn't provided. | |
13a9a6f5 MM |
356 | * |
357 | * Timestamps: | |
358 | * zp - ctime|mtime updated if byte count > 0 | |
359 | */ | |
360 | int | |
361 | zfs_write_simple(znode_t *zp, const void *data, size_t len, | |
1c2358c1 | 362 | loff_t pos, size_t *residp) |
13a9a6f5 | 363 | { |
1c2358c1 BB |
364 | fstrans_cookie_t cookie; |
365 | int error; | |
366 | ||
367 | struct iovec iov; | |
368 | iov.iov_base = (void *)data; | |
369 | iov.iov_len = len; | |
370 | ||
d0cd9a5c BA |
371 | zfs_uio_t uio; |
372 | zfs_uio_iovec_init(&uio, &iov, 1, pos, UIO_SYSSPACE, len, 0); | |
13a9a6f5 | 373 | |
1c2358c1 BB |
374 | cookie = spl_fstrans_mark(); |
375 | error = zfs_write(zp, &uio, 0, kcred); | |
376 | spl_fstrans_unmark(cookie); | |
377 | ||
378 | if (error == 0) { | |
379 | if (residp != NULL) | |
d0cd9a5c BA |
380 | *residp = zfs_uio_resid(&uio); |
381 | else if (zfs_uio_resid(&uio) != 0) | |
1c2358c1 | 382 | error = SET_ERROR(EIO); |
13a9a6f5 | 383 | } |
1c2358c1 | 384 | |
13a9a6f5 MM |
385 | return (error); |
386 | } | |
387 | ||
23c13c7e AL |
388 | static void |
389 | zfs_rele_async_task(void *arg) | |
390 | { | |
391 | iput(arg); | |
392 | } | |
393 | ||
0a50679c | 394 | void |
657ce253 | 395 | zfs_zrele_async(znode_t *zp) |
3558fd73 | 396 | { |
657ce253 | 397 | struct inode *ip = ZTOI(zp); |
0a50679c BB |
398 | objset_t *os = ITOZSB(ip)->z_os; |
399 | ||
3558fd73 | 400 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
401 | ASSERT(os != NULL); |
402 | ||
2921ad6c PD |
403 | /* |
404 | * If decrementing the count would put us at 0, we can't do it inline | |
405 | * here, because that would be synchronous. Instead, dispatch an iput | |
406 | * to run later. | |
407 | * | |
408 | * For more information on the dangers of a synchronous iput, see the | |
409 | * header comment of this file. | |
410 | */ | |
411 | if (!atomic_add_unless(&ip->i_count, -1, 1)) { | |
657ce253 | 412 | VERIFY(taskq_dispatch(dsl_pool_zrele_taskq(dmu_objset_pool(os)), |
23c13c7e | 413 | zfs_rele_async_task, ip, TQ_SLEEP) != TASKQID_INVALID); |
2921ad6c | 414 | } |
3558fd73 BB |
415 | } |
416 | ||
34dc7c2f | 417 | |
34dc7c2f BB |
418 | /* |
419 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 420 | * If it exists, return a held inode reference for it. |
34dc7c2f | 421 | * |
657ce253 | 422 | * IN: zdp - znode of directory to search. |
34dc7c2f | 423 | * nm - name of entry to lookup. |
34dc7c2f | 424 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 425 | * cr - credentials of caller. |
34dc7c2f BB |
426 | * direntflags - directory lookup flags |
427 | * realpnp - returned pathname. | |
428 | * | |
657ce253 | 429 | * OUT: zpp - znode of located entry, NULL if not found. |
34dc7c2f | 430 | * |
d3cc8b15 | 431 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
432 | * |
433 | * Timestamps: | |
434 | * NA | |
435 | */ | |
e5c39b95 | 436 | int |
4d55ea81 RM |
437 | zfs_lookup(znode_t *zdp, char *nm, znode_t **zpp, int flags, cred_t *cr, |
438 | int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 439 | { |
657ce253 | 440 | zfsvfs_t *zfsvfs = ZTOZSB(zdp); |
3558fd73 | 441 | int error = 0; |
45d1cae3 | 442 | |
9b7b9cd3 GM |
443 | /* |
444 | * Fast path lookup, however we must skip DNLC lookup | |
445 | * for case folding or normalizing lookups because the | |
446 | * DNLC code only stores the passed in name. This means | |
447 | * creating 'a' and removing 'A' on a case insensitive | |
448 | * file system would work, but DNLC still thinks 'a' | |
449 | * exists and won't let you create it again on the next | |
450 | * pass through fast path. | |
451 | */ | |
45d1cae3 BB |
452 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { |
453 | ||
657ce253 | 454 | if (!S_ISDIR(ZTOI(zdp)->i_mode)) { |
2e528b49 | 455 | return (SET_ERROR(ENOTDIR)); |
428870ff | 456 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 457 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
458 | } |
459 | ||
460 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
461 | error = zfs_fastaccesschk_execute(zdp, cr); | |
462 | if (!error) { | |
657ce253 MM |
463 | *zpp = zdp; |
464 | zhold(*zpp); | |
45d1cae3 BB |
465 | return (0); |
466 | } | |
467 | return (error); | |
45d1cae3 BB |
468 | } |
469 | } | |
470 | ||
768eaced CC |
471 | if ((error = zfs_enter_verify_zp(zfsvfs, zdp, FTAG)) != 0) |
472 | return (error); | |
34dc7c2f | 473 | |
657ce253 | 474 | *zpp = NULL; |
34dc7c2f BB |
475 | |
476 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
477 | /* |
478 | * We don't allow recursive attributes.. | |
479 | * Maybe someday we will. | |
480 | */ | |
428870ff | 481 | if (zdp->z_pflags & ZFS_XATTR) { |
768eaced | 482 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 483 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
484 | } |
485 | ||
657ce253 | 486 | if ((error = zfs_get_xattrdir(zdp, zpp, cr, flags))) { |
768eaced | 487 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
488 | return (error); |
489 | } | |
490 | ||
491 | /* | |
492 | * Do we have permission to get into attribute directory? | |
493 | */ | |
494 | ||
657ce253 | 495 | if ((error = zfs_zaccess(*zpp, ACE_EXECUTE, 0, |
d4dc53da | 496 | B_TRUE, cr, zfs_init_idmap))) { |
657ce253 MM |
497 | zrele(*zpp); |
498 | *zpp = NULL; | |
34dc7c2f BB |
499 | } |
500 | ||
768eaced | 501 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
502 | return (error); |
503 | } | |
504 | ||
657ce253 | 505 | if (!S_ISDIR(ZTOI(zdp)->i_mode)) { |
768eaced | 506 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 507 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
508 | } |
509 | ||
510 | /* | |
511 | * Check accessibility of directory. | |
512 | */ | |
513 | ||
f224eddf | 514 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr, |
d4dc53da | 515 | zfs_init_idmap))) { |
768eaced | 516 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
517 | return (error); |
518 | } | |
519 | ||
0037b49e | 520 | if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 521 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 522 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 523 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
524 | } |
525 | ||
657ce253 MM |
526 | error = zfs_dirlook(zdp, nm, zpp, flags, direntflags, realpnp); |
527 | if ((error == 0) && (*zpp)) | |
fc273894 | 528 | zfs_znode_update_vfs(*zpp); |
34dc7c2f | 529 | |
768eaced | 530 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
531 | return (error); |
532 | } | |
533 | ||
534 | /* | |
535 | * Attempt to create a new entry in a directory. If the entry | |
536 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 537 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 538 | * |
657ce253 | 539 | * IN: dzp - znode of directory to put new file entry in. |
34dc7c2f BB |
540 | * name - name of new file entry. |
541 | * vap - attributes of new file. | |
542 | * excl - flag indicating exclusive or non-exclusive mode. | |
543 | * mode - mode to open file with. | |
544 | * cr - credentials of caller. | |
a35c1207 | 545 | * flag - file flag. |
3558fd73 | 546 | * vsecp - ACL to be set |
2a068a13 | 547 | * mnt_ns - user namespace of the mount |
34dc7c2f | 548 | * |
657ce253 | 549 | * OUT: zpp - znode of created or trunc'd entry. |
34dc7c2f | 550 | * |
d3cc8b15 | 551 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
552 | * |
553 | * Timestamps: | |
657ce253 MM |
554 | * dzp - ctime|mtime updated if new entry created |
555 | * zp - ctime|mtime always, atime if new | |
34dc7c2f | 556 | */ |
e5c39b95 | 557 | int |
657ce253 | 558 | zfs_create(znode_t *dzp, char *name, vattr_t *vap, int excl, |
2a068a13 | 559 | int mode, znode_t **zpp, cred_t *cr, int flag, vsecattr_t *vsecp, |
d4dc53da | 560 | zidmap_t *mnt_ns) |
34dc7c2f | 561 | { |
657ce253 MM |
562 | znode_t *zp; |
563 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); | |
34dc7c2f BB |
564 | zilog_t *zilog; |
565 | objset_t *os; | |
566 | zfs_dirlock_t *dl; | |
567 | dmu_tx_t *tx; | |
568 | int error; | |
b128c09f | 569 | uid_t uid; |
149e873a | 570 | gid_t gid; |
428870ff | 571 | zfs_acl_ids_t acl_ids; |
9babb374 | 572 | boolean_t fuid_dirtied; |
428870ff | 573 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 574 | boolean_t waited = B_FALSE; |
e3785718 | 575 | boolean_t skip_acl = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
34dc7c2f BB |
576 | |
577 | /* | |
578 | * If we have an ephemeral id, ACL, or XVATTR then | |
579 | * make sure file system is at proper version | |
580 | */ | |
581 | ||
149e873a | 582 | gid = crgetgid(cr); |
3558fd73 | 583 | uid = crgetuid(cr); |
b128c09f | 584 | |
0037b49e | 585 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 586 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 587 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 588 | |
32dec7bd | 589 | if (name == NULL) |
590 | return (SET_ERROR(EINVAL)); | |
591 | ||
768eaced CC |
592 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
593 | return (error); | |
0037b49e BB |
594 | os = zfsvfs->z_os; |
595 | zilog = zfsvfs->z_log; | |
34dc7c2f | 596 | |
0037b49e | 597 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 598 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 599 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 600 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
601 | } |
602 | ||
5484965a | 603 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 604 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 605 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
768eaced | 606 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
607 | return (error); |
608 | } | |
609 | } | |
34dc7c2f | 610 | |
3558fd73 | 611 | top: |
657ce253 | 612 | *zpp = NULL; |
34dc7c2f BB |
613 | if (*name == '\0') { |
614 | /* | |
615 | * Null component name refers to the directory itself. | |
616 | */ | |
657ce253 | 617 | zhold(dzp); |
34dc7c2f BB |
618 | zp = dzp; |
619 | dl = NULL; | |
620 | error = 0; | |
621 | } else { | |
3558fd73 | 622 | /* possible igrab(zp) */ |
34dc7c2f BB |
623 | int zflg = 0; |
624 | ||
625 | if (flag & FIGNORECASE) | |
626 | zflg |= ZCILOOK; | |
627 | ||
628 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
629 | NULL, NULL); | |
630 | if (error) { | |
572e2857 BB |
631 | if (have_acl) |
632 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 633 | if (strcmp(name, "..") == 0) |
2e528b49 | 634 | error = SET_ERROR(EISDIR); |
768eaced | 635 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
636 | return (error); |
637 | } | |
638 | } | |
428870ff | 639 | |
34dc7c2f BB |
640 | if (zp == NULL) { |
641 | uint64_t txtype; | |
9c5167d1 | 642 | uint64_t projid = ZFS_DEFAULT_PROJID; |
34dc7c2f BB |
643 | |
644 | /* | |
645 | * Create a new file object and update the directory | |
646 | * to reference it. | |
647 | */ | |
e3785718 | 648 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, skip_acl, cr, |
2a068a13 | 649 | mnt_ns))) { |
572e2857 BB |
650 | if (have_acl) |
651 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
652 | goto out; |
653 | } | |
654 | ||
655 | /* | |
656 | * We only support the creation of regular files in | |
657 | * extended attribute directories. | |
658 | */ | |
428870ff | 659 | |
3558fd73 | 660 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
661 | if (have_acl) |
662 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 663 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
664 | goto out; |
665 | } | |
666 | ||
428870ff | 667 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
2a068a13 | 668 | cr, vsecp, &acl_ids, mnt_ns)) != 0) |
9babb374 | 669 | goto out; |
428870ff BB |
670 | have_acl = B_TRUE; |
671 | ||
9c5167d1 NF |
672 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
673 | projid = zfs_inherit_projid(dzp); | |
674 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
45d1cae3 | 675 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 676 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
677 | goto out; |
678 | } | |
679 | ||
34dc7c2f | 680 | tx = dmu_tx_create(os); |
428870ff BB |
681 | |
682 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
683 | ZFS_SA_BASE_ATTR_SIZE); | |
684 | ||
0037b49e | 685 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 686 | if (fuid_dirtied) |
0037b49e | 687 | zfs_fuid_txhold(zfsvfs, tx); |
34dc7c2f | 688 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 689 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
0037b49e | 690 | if (!zfsvfs->z_use_sa && |
428870ff | 691 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 692 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 693 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 694 | } |
599b8648 | 695 | |
0735ecb3 PS |
696 | error = dmu_tx_assign(tx, |
697 | (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); | |
34dc7c2f BB |
698 | if (error) { |
699 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 700 | if (error == ERESTART) { |
e8b96c60 | 701 | waited = B_TRUE; |
34dc7c2f BB |
702 | dmu_tx_wait(tx); |
703 | dmu_tx_abort(tx); | |
704 | goto top; | |
705 | } | |
428870ff | 706 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 707 | dmu_tx_abort(tx); |
768eaced | 708 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
709 | return (error); |
710 | } | |
428870ff | 711 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 712 | |
599b8648 CC |
713 | error = zfs_link_create(dl, zp, tx, ZNEW); |
714 | if (error != 0) { | |
715 | /* | |
716 | * Since, we failed to add the directory entry for it, | |
717 | * delete the newly created dnode. | |
718 | */ | |
719 | zfs_znode_delete(zp, tx); | |
720 | remove_inode_hash(ZTOI(zp)); | |
721 | zfs_acl_ids_free(&acl_ids); | |
722 | dmu_tx_commit(tx); | |
723 | goto out; | |
724 | } | |
725 | ||
9babb374 | 726 | if (fuid_dirtied) |
0037b49e | 727 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 728 | |
34dc7c2f BB |
729 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); |
730 | if (flag & FIGNORECASE) | |
731 | txtype |= TX_CI; | |
732 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
733 | vsecp, acl_ids.z_fuidp, vap); |
734 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
735 | dmu_tx_commit(tx); |
736 | } else { | |
da92d5cb | 737 | int aflags = (flag & O_APPEND) ? V_APPEND : 0; |
34dc7c2f | 738 | |
572e2857 BB |
739 | if (have_acl) |
740 | zfs_acl_ids_free(&acl_ids); | |
572e2857 | 741 | |
34dc7c2f BB |
742 | /* |
743 | * A directory entry already exists for this name. | |
744 | */ | |
745 | /* | |
746 | * Can't truncate an existing file if in exclusive mode. | |
747 | */ | |
3558fd73 | 748 | if (excl) { |
2e528b49 | 749 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
750 | goto out; |
751 | } | |
752 | /* | |
753 | * Can't open a directory for writing. | |
754 | */ | |
3558fd73 | 755 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 756 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
757 | goto out; |
758 | } | |
759 | /* | |
760 | * Verify requested access to file. | |
761 | */ | |
2a068a13 YY |
762 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr, |
763 | mnt_ns))) { | |
34dc7c2f BB |
764 | goto out; |
765 | } | |
766 | ||
767 | mutex_enter(&dzp->z_lock); | |
768 | dzp->z_seq++; | |
769 | mutex_exit(&dzp->z_lock); | |
770 | ||
771 | /* | |
772 | * Truncate regular files if requested. | |
773 | */ | |
3558fd73 BB |
774 | if (S_ISREG(ZTOI(zp)->i_mode) && |
775 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 776 | /* we can't hold any locks when calling zfs_freesp() */ |
609603a5 B |
777 | if (dl) { |
778 | zfs_dirent_unlock(dl); | |
779 | dl = NULL; | |
780 | } | |
34dc7c2f | 781 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
782 | } |
783 | } | |
784 | out: | |
785 | ||
786 | if (dl) | |
787 | zfs_dirent_unlock(dl); | |
788 | ||
789 | if (error) { | |
790 | if (zp) | |
657ce253 | 791 | zrele(zp); |
34dc7c2f | 792 | } else { |
fc273894 KHN |
793 | zfs_znode_update_vfs(dzp); |
794 | zfs_znode_update_vfs(zp); | |
657ce253 | 795 | *zpp = zp; |
34dc7c2f | 796 | } |
34dc7c2f | 797 | |
0037b49e | 798 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 799 | zil_commit(zilog, 0); |
428870ff | 800 | |
768eaced | 801 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
802 | return (error); |
803 | } | |
804 | ||
ace1eae8 CC |
805 | int |
806 | zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, | |
2a068a13 | 807 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp, |
d4dc53da | 808 | zidmap_t *mnt_ns) |
ace1eae8 | 809 | { |
ef70eff1 | 810 | (void) excl, (void) mode, (void) flag; |
ace1eae8 | 811 | znode_t *zp = NULL, *dzp = ITOZ(dip); |
0037b49e | 812 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
ace1eae8 CC |
813 | objset_t *os; |
814 | dmu_tx_t *tx; | |
815 | int error; | |
816 | uid_t uid; | |
817 | gid_t gid; | |
818 | zfs_acl_ids_t acl_ids; | |
9c5167d1 | 819 | uint64_t projid = ZFS_DEFAULT_PROJID; |
ace1eae8 CC |
820 | boolean_t fuid_dirtied; |
821 | boolean_t have_acl = B_FALSE; | |
822 | boolean_t waited = B_FALSE; | |
823 | ||
824 | /* | |
825 | * If we have an ephemeral id, ACL, or XVATTR then | |
826 | * make sure file system is at proper version | |
827 | */ | |
828 | ||
829 | gid = crgetgid(cr); | |
830 | uid = crgetuid(cr); | |
831 | ||
0037b49e | 832 | if (zfsvfs->z_use_fuids == B_FALSE && |
ace1eae8 CC |
833 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
834 | return (SET_ERROR(EINVAL)); | |
835 | ||
768eaced CC |
836 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
837 | return (error); | |
0037b49e | 838 | os = zfsvfs->z_os; |
ace1eae8 CC |
839 | |
840 | if (vap->va_mask & ATTR_XVATTR) { | |
841 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
842 | crgetuid(cr), cr, vap->va_mode)) != 0) { | |
768eaced | 843 | zfs_exit(zfsvfs, FTAG); |
ace1eae8 CC |
844 | return (error); |
845 | } | |
846 | } | |
847 | ||
848 | top: | |
849 | *ipp = NULL; | |
850 | ||
851 | /* | |
852 | * Create a new file object and update the directory | |
853 | * to reference it. | |
854 | */ | |
2a068a13 | 855 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr, mnt_ns))) { |
ace1eae8 CC |
856 | if (have_acl) |
857 | zfs_acl_ids_free(&acl_ids); | |
858 | goto out; | |
859 | } | |
860 | ||
861 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, | |
2a068a13 | 862 | cr, vsecp, &acl_ids, mnt_ns)) != 0) |
ace1eae8 CC |
863 | goto out; |
864 | have_acl = B_TRUE; | |
865 | ||
9c5167d1 NF |
866 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
867 | projid = zfs_inherit_projid(dzp); | |
868 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
ace1eae8 CC |
869 | zfs_acl_ids_free(&acl_ids); |
870 | error = SET_ERROR(EDQUOT); | |
871 | goto out; | |
872 | } | |
873 | ||
874 | tx = dmu_tx_create(os); | |
875 | ||
876 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
877 | ZFS_SA_BASE_ATTR_SIZE); | |
0037b49e | 878 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 879 | |
0037b49e | 880 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
ace1eae8 | 881 | if (fuid_dirtied) |
0037b49e BB |
882 | zfs_fuid_txhold(zfsvfs, tx); |
883 | if (!zfsvfs->z_use_sa && | |
ace1eae8 CC |
884 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
885 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
886 | 0, acl_ids.z_aclp->z_acl_bytes); | |
887 | } | |
0735ecb3 | 888 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
ace1eae8 CC |
889 | if (error) { |
890 | if (error == ERESTART) { | |
891 | waited = B_TRUE; | |
892 | dmu_tx_wait(tx); | |
893 | dmu_tx_abort(tx); | |
894 | goto top; | |
895 | } | |
896 | zfs_acl_ids_free(&acl_ids); | |
897 | dmu_tx_abort(tx); | |
768eaced | 898 | zfs_exit(zfsvfs, FTAG); |
ace1eae8 CC |
899 | return (error); |
900 | } | |
901 | zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); | |
902 | ||
903 | if (fuid_dirtied) | |
0037b49e | 904 | zfs_fuid_sync(zfsvfs, tx); |
ace1eae8 CC |
905 | |
906 | /* Add to unlinked set */ | |
a43570c5 | 907 | zp->z_unlinked = B_TRUE; |
ace1eae8 CC |
908 | zfs_unlinked_add(zp, tx); |
909 | zfs_acl_ids_free(&acl_ids); | |
910 | dmu_tx_commit(tx); | |
911 | out: | |
912 | ||
913 | if (error) { | |
914 | if (zp) | |
657ce253 | 915 | zrele(zp); |
ace1eae8 | 916 | } else { |
fc273894 KHN |
917 | zfs_znode_update_vfs(dzp); |
918 | zfs_znode_update_vfs(zp); | |
ace1eae8 CC |
919 | *ipp = ZTOI(zp); |
920 | } | |
921 | ||
768eaced | 922 | zfs_exit(zfsvfs, FTAG); |
ace1eae8 CC |
923 | return (error); |
924 | } | |
925 | ||
34dc7c2f BB |
926 | /* |
927 | * Remove an entry from a directory. | |
928 | * | |
657ce253 | 929 | * IN: dzp - znode of directory to remove entry from. |
34dc7c2f BB |
930 | * name - name of entry to remove. |
931 | * cr - credentials of caller. | |
841a7a98 | 932 | * flags - case flags. |
34dc7c2f BB |
933 | * |
934 | * RETURN: 0 if success | |
935 | * error code if failure | |
936 | * | |
937 | * Timestamps: | |
657ce253 | 938 | * dzp - ctime|mtime |
3558fd73 | 939 | * ip - ctime (if nlink > 0) |
34dc7c2f | 940 | */ |
428870ff | 941 | |
51c747de | 942 | static uint64_t null_xattr = 0; |
428870ff | 943 | |
e5c39b95 | 944 | int |
657ce253 | 945 | zfs_remove(znode_t *dzp, char *name, cred_t *cr, int flags) |
34dc7c2f | 946 | { |
657ce253 | 947 | znode_t *zp; |
572e2857 | 948 | znode_t *xzp; |
657ce253 | 949 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f | 950 | zilog_t *zilog; |
a966c564 | 951 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 952 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 953 | uint64_t obj = 0; |
dfbc8630 | 954 | uint64_t links; |
34dc7c2f BB |
955 | zfs_dirlock_t *dl; |
956 | dmu_tx_t *tx; | |
a966c564 K |
957 | boolean_t may_delete_now, delete_now = FALSE; |
958 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
959 | uint64_t txtype; |
960 | pathname_t *realnmp = NULL; | |
961 | pathname_t realnm; | |
962 | int error; | |
963 | int zflg = ZEXISTS; | |
e8b96c60 | 964 | boolean_t waited = B_FALSE; |
34dc7c2f | 965 | |
32dec7bd | 966 | if (name == NULL) |
967 | return (SET_ERROR(EINVAL)); | |
968 | ||
768eaced CC |
969 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
970 | return (error); | |
0037b49e | 971 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
972 | |
973 | if (flags & FIGNORECASE) { | |
974 | zflg |= ZCILOOK; | |
975 | pn_alloc(&realnm); | |
976 | realnmp = &realnm; | |
977 | } | |
978 | ||
979 | top: | |
572e2857 BB |
980 | xattr_obj = 0; |
981 | xzp = NULL; | |
34dc7c2f BB |
982 | /* |
983 | * Attempt to lock directory; fail if entry doesn't exist. | |
984 | */ | |
149e873a BB |
985 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
986 | NULL, realnmp))) { | |
34dc7c2f BB |
987 | if (realnmp) |
988 | pn_free(realnmp); | |
768eaced | 989 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
990 | return (error); |
991 | } | |
992 | ||
d4dc53da | 993 | if ((error = zfs_zaccess_delete(dzp, zp, cr, zfs_init_idmap))) { |
34dc7c2f BB |
994 | goto out; |
995 | } | |
996 | ||
997 | /* | |
998 | * Need to use rmdir for removing directories. | |
999 | */ | |
657ce253 | 1000 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1001 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1002 | goto out; |
1003 | } | |
1004 | ||
19d55079 | 1005 | mutex_enter(&zp->z_lock); |
657ce253 | 1006 | may_delete_now = atomic_read(&ZTOI(zp)->i_count) == 1 && |
3fc92adc | 1007 | !zn_has_cached_data(zp, 0, LLONG_MAX); |
19d55079 MA |
1008 | mutex_exit(&zp->z_lock); |
1009 | ||
34dc7c2f | 1010 | /* |
a966c564 K |
1011 | * We may delete the znode now, or we may put it in the unlinked set; |
1012 | * it depends on whether we're the last link, and on whether there are | |
1013 | * other holds on the inode. So we dmu_tx_hold() the right things to | |
1014 | * allow for either case. | |
34dc7c2f | 1015 | */ |
572e2857 | 1016 | obj = zp->z_id; |
0037b49e | 1017 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 1018 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1019 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1020 | zfs_sa_upgrade_txholds(tx, zp); | |
1021 | zfs_sa_upgrade_txholds(tx, dzp); | |
a966c564 K |
1022 | if (may_delete_now) { |
1023 | toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; | |
1024 | /* if the file is too big, only hold_free a token amount */ | |
1025 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1026 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1027 | } | |
34dc7c2f BB |
1028 | |
1029 | /* are there any extended attributes? */ | |
0037b49e | 1030 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1031 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1032 | if (error == 0 && xattr_obj) { |
0037b49e | 1033 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
c99c9001 | 1034 | ASSERT0(error); |
428870ff BB |
1035 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1036 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1037 | } |
1038 | ||
a966c564 K |
1039 | mutex_enter(&zp->z_lock); |
1040 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
1041 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
1042 | mutex_exit(&zp->z_lock); | |
1043 | ||
34dc7c2f | 1044 | /* charge as an update -- would be nice not to charge at all */ |
0037b49e | 1045 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1046 | |
19d55079 | 1047 | /* |
1a04bab3 | 1048 | * Mark this transaction as typically resulting in a net free of space |
19d55079 | 1049 | */ |
1a04bab3 | 1050 | dmu_tx_mark_netfree(tx); |
19d55079 | 1051 | |
0735ecb3 | 1052 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
1053 | if (error) { |
1054 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1055 | if (error == ERESTART) { |
e8b96c60 | 1056 | waited = B_TRUE; |
34dc7c2f BB |
1057 | dmu_tx_wait(tx); |
1058 | dmu_tx_abort(tx); | |
657ce253 | 1059 | zrele(zp); |
ea7e86d8 | 1060 | if (xzp) |
657ce253 | 1061 | zrele(xzp); |
34dc7c2f BB |
1062 | goto top; |
1063 | } | |
1064 | if (realnmp) | |
1065 | pn_free(realnmp); | |
1066 | dmu_tx_abort(tx); | |
657ce253 | 1067 | zrele(zp); |
ea7e86d8 | 1068 | if (xzp) |
657ce253 | 1069 | zrele(xzp); |
768eaced | 1070 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1071 | return (error); |
1072 | } | |
1073 | ||
1074 | /* | |
1075 | * Remove the directory entry. | |
1076 | */ | |
1077 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1078 | ||
1079 | if (error) { | |
1080 | dmu_tx_commit(tx); | |
1081 | goto out; | |
1082 | } | |
1083 | ||
1084 | if (unlinked) { | |
572e2857 BB |
1085 | /* |
1086 | * Hold z_lock so that we can make sure that the ACL obj | |
1087 | * hasn't changed. Could have been deleted due to | |
1088 | * zfs_sa_upgrade(). | |
1089 | */ | |
1090 | mutex_enter(&zp->z_lock); | |
0037b49e | 1091 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1092 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
a966c564 | 1093 | delete_now = may_delete_now && !toobig && |
657ce253 | 1094 | atomic_read(&ZTOI(zp)->i_count) == 1 && |
3fc92adc BB |
1095 | !zn_has_cached_data(zp, 0, LLONG_MAX) && |
1096 | xattr_obj == xattr_obj_unlinked && | |
657ce253 | 1097 | zfs_external_acl(zp) == acl_obj; |
66a38fd1 | 1098 | VERIFY_IMPLY(xattr_obj_unlinked, xzp); |
a966c564 K |
1099 | } |
1100 | ||
1101 | if (delete_now) { | |
1102 | if (xattr_obj_unlinked) { | |
dfbc8630 | 1103 | ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); |
a966c564 | 1104 | mutex_enter(&xzp->z_lock); |
a43570c5 | 1105 | xzp->z_unlinked = B_TRUE; |
dfbc8630 CD |
1106 | clear_nlink(ZTOI(xzp)); |
1107 | links = 0; | |
0037b49e | 1108 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 1109 | &links, sizeof (links), tx); |
a966c564 K |
1110 | ASSERT3U(error, ==, 0); |
1111 | mutex_exit(&xzp->z_lock); | |
1112 | zfs_unlinked_add(xzp, tx); | |
1113 | ||
1114 | if (zp->z_is_sa) | |
1115 | error = sa_remove(zp->z_sa_hdl, | |
0037b49e | 1116 | SA_ZPL_XATTR(zfsvfs), tx); |
a966c564 K |
1117 | else |
1118 | error = sa_update(zp->z_sa_hdl, | |
0037b49e | 1119 | SA_ZPL_XATTR(zfsvfs), &null_xattr, |
a966c564 K |
1120 | sizeof (uint64_t), tx); |
1121 | ASSERT0(error); | |
1122 | } | |
1123 | /* | |
1124 | * Add to the unlinked set because a new reference could be | |
1125 | * taken concurrently resulting in a deferred destruction. | |
1126 | */ | |
1127 | zfs_unlinked_add(zp, tx); | |
1128 | mutex_exit(&zp->z_lock); | |
a966c564 | 1129 | } else if (unlinked) { |
572e2857 | 1130 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1131 | zfs_unlinked_add(zp, tx); |
1132 | } | |
1133 | ||
1134 | txtype = TX_REMOVE; | |
1135 | if (flags & FIGNORECASE) | |
1136 | txtype |= TX_CI; | |
8e556c5e | 1137 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj, unlinked); |
34dc7c2f BB |
1138 | |
1139 | dmu_tx_commit(tx); | |
1140 | out: | |
1141 | if (realnmp) | |
1142 | pn_free(realnmp); | |
1143 | ||
1144 | zfs_dirent_unlock(dl); | |
fc273894 KHN |
1145 | zfs_znode_update_vfs(dzp); |
1146 | zfs_znode_update_vfs(zp); | |
34dc7c2f | 1147 | |
ea7e86d8 | 1148 | if (delete_now) |
657ce253 | 1149 | zrele(zp); |
ea7e86d8 | 1150 | else |
657ce253 | 1151 | zfs_zrele_async(zp); |
a966c564 K |
1152 | |
1153 | if (xzp) { | |
fc273894 | 1154 | zfs_znode_update_vfs(xzp); |
657ce253 | 1155 | zfs_zrele_async(xzp); |
a966c564 | 1156 | } |
428870ff | 1157 | |
0037b49e | 1158 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1159 | zil_commit(zilog, 0); |
34dc7c2f | 1160 | |
768eaced | 1161 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1162 | return (error); |
1163 | } | |
1164 | ||
1165 | /* | |
657ce253 | 1166 | * Create a new directory and insert it into dzp using the name |
34dc7c2f BB |
1167 | * provided. Return a pointer to the inserted directory. |
1168 | * | |
657ce253 | 1169 | * IN: dzp - znode of directory to add subdir to. |
34dc7c2f BB |
1170 | * dirname - name of new directory. |
1171 | * vap - attributes of new directory. | |
1172 | * cr - credentials of caller. | |
841a7a98 | 1173 | * flags - case flags. |
34dc7c2f | 1174 | * vsecp - ACL to be set |
2a068a13 | 1175 | * mnt_ns - user namespace of the mount |
34dc7c2f | 1176 | * |
657ce253 | 1177 | * OUT: zpp - znode of created directory. |
34dc7c2f BB |
1178 | * |
1179 | * RETURN: 0 if success | |
1180 | * error code if failure | |
1181 | * | |
1182 | * Timestamps: | |
657ce253 MM |
1183 | * dzp - ctime|mtime updated |
1184 | * zpp - ctime|mtime|atime updated | |
34dc7c2f | 1185 | */ |
e5c39b95 | 1186 | int |
657ce253 | 1187 | zfs_mkdir(znode_t *dzp, char *dirname, vattr_t *vap, znode_t **zpp, |
d4dc53da | 1188 | cred_t *cr, int flags, vsecattr_t *vsecp, zidmap_t *mnt_ns) |
34dc7c2f | 1189 | { |
657ce253 MM |
1190 | znode_t *zp; |
1191 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); | |
34dc7c2f BB |
1192 | zilog_t *zilog; |
1193 | zfs_dirlock_t *dl; | |
1194 | uint64_t txtype; | |
1195 | dmu_tx_t *tx; | |
1196 | int error; | |
34dc7c2f | 1197 | int zf = ZNEW; |
b128c09f BB |
1198 | uid_t uid; |
1199 | gid_t gid = crgetgid(cr); | |
428870ff | 1200 | zfs_acl_ids_t acl_ids; |
9babb374 | 1201 | boolean_t fuid_dirtied; |
e8b96c60 | 1202 | boolean_t waited = B_FALSE; |
34dc7c2f | 1203 | |
3558fd73 | 1204 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1205 | |
1206 | /* | |
1207 | * If we have an ephemeral id, ACL, or XVATTR then | |
1208 | * make sure file system is at proper version | |
1209 | */ | |
1210 | ||
3558fd73 | 1211 | uid = crgetuid(cr); |
0037b49e | 1212 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1213 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1214 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1215 | |
32dec7bd | 1216 | if (dirname == NULL) |
1217 | return (SET_ERROR(EINVAL)); | |
1218 | ||
768eaced CC |
1219 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
1220 | return (error); | |
0037b49e | 1221 | zilog = zfsvfs->z_log; |
34dc7c2f | 1222 | |
428870ff | 1223 | if (dzp->z_pflags & ZFS_XATTR) { |
768eaced | 1224 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1225 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1226 | } |
1227 | ||
0037b49e | 1228 | if (zfsvfs->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1229 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 1230 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1231 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1232 | } |
1233 | if (flags & FIGNORECASE) | |
1234 | zf |= ZCILOOK; | |
1235 | ||
5484965a | 1236 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1237 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1238 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
768eaced | 1239 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1240 | return (error); |
1241 | } | |
428870ff | 1242 | } |
34dc7c2f | 1243 | |
428870ff | 1244 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
2a068a13 | 1245 | vsecp, &acl_ids, mnt_ns)) != 0) { |
768eaced | 1246 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
1247 | return (error); |
1248 | } | |
34dc7c2f BB |
1249 | /* |
1250 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1251 | * |
1252 | * Existence is checked first to make sure we don't return | |
1253 | * EACCES instead of EEXIST which can cause some applications | |
1254 | * to fail. | |
34dc7c2f BB |
1255 | */ |
1256 | top: | |
657ce253 | 1257 | *zpp = NULL; |
34dc7c2f | 1258 | |
149e873a BB |
1259 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1260 | NULL, NULL))) { | |
428870ff | 1261 | zfs_acl_ids_free(&acl_ids); |
768eaced | 1262 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1263 | return (error); |
1264 | } | |
1265 | ||
2a068a13 YY |
1266 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr, |
1267 | mnt_ns))) { | |
428870ff | 1268 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1269 | zfs_dirent_unlock(dl); |
768eaced | 1270 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1271 | return (error); |
1272 | } | |
1273 | ||
9c5167d1 | 1274 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) { |
45d1cae3 | 1275 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 1276 | zfs_dirent_unlock(dl); |
768eaced | 1277 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1278 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
1279 | } |
1280 | ||
34dc7c2f BB |
1281 | /* |
1282 | * Add a new entry to the directory. | |
1283 | */ | |
0037b49e | 1284 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f BB |
1285 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
1286 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
0037b49e | 1287 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1288 | if (fuid_dirtied) |
0037b49e BB |
1289 | zfs_fuid_txhold(zfsvfs, tx); |
1290 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
1291 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
1292 | acl_ids.z_aclp->z_acl_bytes); | |
1293 | } | |
1294 | ||
1295 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1296 | ZFS_SA_BASE_ATTR_SIZE); | |
1297 | ||
0735ecb3 | 1298 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
1299 | if (error) { |
1300 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1301 | if (error == ERESTART) { |
e8b96c60 | 1302 | waited = B_TRUE; |
34dc7c2f BB |
1303 | dmu_tx_wait(tx); |
1304 | dmu_tx_abort(tx); | |
1305 | goto top; | |
1306 | } | |
428870ff | 1307 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1308 | dmu_tx_abort(tx); |
768eaced | 1309 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1310 | return (error); |
1311 | } | |
1312 | ||
1313 | /* | |
1314 | * Create new node. | |
1315 | */ | |
428870ff | 1316 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f BB |
1317 | |
1318 | /* | |
1319 | * Now put new name in parent dir. | |
1320 | */ | |
599b8648 CC |
1321 | error = zfs_link_create(dl, zp, tx, ZNEW); |
1322 | if (error != 0) { | |
1323 | zfs_znode_delete(zp, tx); | |
1324 | remove_inode_hash(ZTOI(zp)); | |
1325 | goto out; | |
1326 | } | |
1327 | ||
1328 | if (fuid_dirtied) | |
1329 | zfs_fuid_sync(zfsvfs, tx); | |
34dc7c2f | 1330 | |
657ce253 | 1331 | *zpp = zp; |
34dc7c2f BB |
1332 | |
1333 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
1334 | if (flags & FIGNORECASE) | |
1335 | txtype |= TX_CI; | |
9babb374 BB |
1336 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
1337 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 1338 | |
599b8648 | 1339 | out: |
9babb374 | 1340 | zfs_acl_ids_free(&acl_ids); |
428870ff | 1341 | |
34dc7c2f BB |
1342 | dmu_tx_commit(tx); |
1343 | ||
1344 | zfs_dirent_unlock(dl); | |
1345 | ||
0037b49e | 1346 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1347 | zil_commit(zilog, 0); |
428870ff | 1348 | |
599b8648 | 1349 | if (error != 0) { |
657ce253 | 1350 | zrele(zp); |
599b8648 | 1351 | } else { |
fc273894 KHN |
1352 | zfs_znode_update_vfs(dzp); |
1353 | zfs_znode_update_vfs(zp); | |
599b8648 | 1354 | } |
768eaced | 1355 | zfs_exit(zfsvfs, FTAG); |
599b8648 | 1356 | return (error); |
34dc7c2f BB |
1357 | } |
1358 | ||
1359 | /* | |
1360 | * Remove a directory subdir entry. If the current working | |
1361 | * directory is the same as the subdir to be removed, the | |
1362 | * remove will fail. | |
1363 | * | |
657ce253 | 1364 | * IN: dzp - znode of directory to remove from. |
34dc7c2f | 1365 | * name - name of directory to be removed. |
3558fd73 | 1366 | * cwd - inode of current working directory. |
34dc7c2f | 1367 | * cr - credentials of caller. |
34dc7c2f BB |
1368 | * flags - case flags |
1369 | * | |
d3cc8b15 | 1370 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1371 | * |
1372 | * Timestamps: | |
657ce253 | 1373 | * dzp - ctime|mtime updated |
34dc7c2f | 1374 | */ |
e5c39b95 | 1375 | int |
657ce253 | 1376 | zfs_rmdir(znode_t *dzp, char *name, znode_t *cwd, cred_t *cr, |
3558fd73 | 1377 | int flags) |
34dc7c2f | 1378 | { |
34dc7c2f | 1379 | znode_t *zp; |
657ce253 | 1380 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f BB |
1381 | zilog_t *zilog; |
1382 | zfs_dirlock_t *dl; | |
1383 | dmu_tx_t *tx; | |
1384 | int error; | |
1385 | int zflg = ZEXISTS; | |
e8b96c60 | 1386 | boolean_t waited = B_FALSE; |
34dc7c2f | 1387 | |
32dec7bd | 1388 | if (name == NULL) |
1389 | return (SET_ERROR(EINVAL)); | |
1390 | ||
768eaced CC |
1391 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
1392 | return (error); | |
0037b49e | 1393 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1394 | |
1395 | if (flags & FIGNORECASE) | |
1396 | zflg |= ZCILOOK; | |
1397 | top: | |
1398 | zp = NULL; | |
1399 | ||
1400 | /* | |
1401 | * Attempt to lock directory; fail if entry doesn't exist. | |
1402 | */ | |
149e873a BB |
1403 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1404 | NULL, NULL))) { | |
768eaced | 1405 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1406 | return (error); |
1407 | } | |
1408 | ||
d4dc53da | 1409 | if ((error = zfs_zaccess_delete(dzp, zp, cr, zfs_init_idmap))) { |
34dc7c2f BB |
1410 | goto out; |
1411 | } | |
1412 | ||
657ce253 | 1413 | if (!S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1414 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
1415 | goto out; |
1416 | } | |
1417 | ||
657ce253 | 1418 | if (zp == cwd) { |
2e528b49 | 1419 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1420 | goto out; |
1421 | } | |
1422 | ||
34dc7c2f | 1423 | /* |
4e33ba4c | 1424 | * Grab a lock on the directory to make sure that no one is |
34dc7c2f BB |
1425 | * trying to add (or lookup) entries while we are removing it. |
1426 | */ | |
1427 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
1428 | ||
1429 | /* | |
1430 | * Grab a lock on the parent pointer to make sure we play well | |
1431 | * with the treewalk and directory rename code. | |
1432 | */ | |
1433 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
1434 | ||
0037b49e | 1435 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 1436 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 1437 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
0037b49e | 1438 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
1439 | zfs_sa_upgrade_txholds(tx, zp); |
1440 | zfs_sa_upgrade_txholds(tx, dzp); | |
db707ad0 | 1441 | dmu_tx_mark_netfree(tx); |
0735ecb3 | 1442 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
1443 | if (error) { |
1444 | rw_exit(&zp->z_parent_lock); | |
1445 | rw_exit(&zp->z_name_lock); | |
1446 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1447 | if (error == ERESTART) { |
e8b96c60 | 1448 | waited = B_TRUE; |
34dc7c2f BB |
1449 | dmu_tx_wait(tx); |
1450 | dmu_tx_abort(tx); | |
657ce253 | 1451 | zrele(zp); |
34dc7c2f BB |
1452 | goto top; |
1453 | } | |
1454 | dmu_tx_abort(tx); | |
657ce253 | 1455 | zrele(zp); |
768eaced | 1456 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1457 | return (error); |
1458 | } | |
1459 | ||
1460 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
1461 | ||
1462 | if (error == 0) { | |
1463 | uint64_t txtype = TX_RMDIR; | |
1464 | if (flags & FIGNORECASE) | |
1465 | txtype |= TX_CI; | |
8e556c5e CC |
1466 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT, |
1467 | B_FALSE); | |
34dc7c2f BB |
1468 | } |
1469 | ||
1470 | dmu_tx_commit(tx); | |
1471 | ||
1472 | rw_exit(&zp->z_parent_lock); | |
1473 | rw_exit(&zp->z_name_lock); | |
1474 | out: | |
1475 | zfs_dirent_unlock(dl); | |
1476 | ||
fc273894 KHN |
1477 | zfs_znode_update_vfs(dzp); |
1478 | zfs_znode_update_vfs(zp); | |
657ce253 | 1479 | zrele(zp); |
34dc7c2f | 1480 | |
0037b49e | 1481 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1482 | zil_commit(zilog, 0); |
428870ff | 1483 | |
768eaced | 1484 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1485 | return (error); |
1486 | } | |
1487 | ||
1488 | /* | |
841a7a98 TK |
1489 | * Read directory entries from the given directory cursor position and emit |
1490 | * name and position for each entry. | |
34dc7c2f | 1491 | * |
3558fd73 | 1492 | * IN: ip - inode of directory to read. |
841a7a98 TK |
1493 | * ctx - directory entry context. |
1494 | * cr - credentials of caller. | |
34dc7c2f BB |
1495 | * |
1496 | * RETURN: 0 if success | |
1497 | * error code if failure | |
1498 | * | |
1499 | * Timestamps: | |
3558fd73 | 1500 | * ip - atime updated |
34dc7c2f BB |
1501 | * |
1502 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
1503 | * This allows us to use the low range for "special" directory entries: | |
1504 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
1505 | * we use the offset 2 for the '.zfs' directory. | |
1506 | */ | |
3558fd73 | 1507 | int |
9464b959 | 1508 | zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) |
34dc7c2f | 1509 | { |
ef70eff1 | 1510 | (void) cr; |
3558fd73 | 1511 | znode_t *zp = ITOZ(ip); |
0037b49e | 1512 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 1513 | objset_t *os; |
34dc7c2f BB |
1514 | zap_cursor_t zc; |
1515 | zap_attribute_t zap; | |
34dc7c2f BB |
1516 | int error; |
1517 | uint8_t prefetch; | |
c12e3a59 | 1518 | uint8_t type; |
3558fd73 BB |
1519 | int done = 0; |
1520 | uint64_t parent; | |
c12e3a59 | 1521 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 1522 | |
768eaced CC |
1523 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
1524 | return (error); | |
34dc7c2f | 1525 | |
0037b49e | 1526 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
3558fd73 BB |
1527 | &parent, sizeof (parent))) != 0) |
1528 | goto out; | |
34dc7c2f BB |
1529 | |
1530 | /* | |
1531 | * Quit if directory has been removed (posix) | |
1532 | */ | |
3558fd73 BB |
1533 | if (zp->z_unlinked) |
1534 | goto out; | |
1535 | ||
c12e3a59 | 1536 | error = 0; |
0037b49e | 1537 | os = zfsvfs->z_os; |
c12e3a59 | 1538 | offset = ctx->pos; |
34dc7c2f BB |
1539 | prefetch = zp->z_zn_prefetch; |
1540 | ||
1541 | /* | |
1542 | * Initialize the iterator cursor. | |
1543 | */ | |
c12e3a59 | 1544 | if (offset <= 3) { |
34dc7c2f BB |
1545 | /* |
1546 | * Start iteration from the beginning of the directory. | |
1547 | */ | |
1548 | zap_cursor_init(&zc, os, zp->z_id); | |
1549 | } else { | |
1550 | /* | |
1551 | * The offset is a serialized cursor. | |
1552 | */ | |
c12e3a59 | 1553 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
1554 | } |
1555 | ||
34dc7c2f BB |
1556 | /* |
1557 | * Transform to file-system independent format | |
1558 | */ | |
3558fd73 BB |
1559 | while (!done) { |
1560 | uint64_t objnum; | |
34dc7c2f BB |
1561 | /* |
1562 | * Special case `.', `..', and `.zfs'. | |
1563 | */ | |
c12e3a59 | 1564 | if (offset == 0) { |
34dc7c2f BB |
1565 | (void) strcpy(zap.za_name, "."); |
1566 | zap.za_normalization_conflict = 0; | |
1567 | objnum = zp->z_id; | |
c12e3a59 RY |
1568 | type = DT_DIR; |
1569 | } else if (offset == 1) { | |
34dc7c2f BB |
1570 | (void) strcpy(zap.za_name, ".."); |
1571 | zap.za_normalization_conflict = 0; | |
428870ff | 1572 | objnum = parent; |
c12e3a59 RY |
1573 | type = DT_DIR; |
1574 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
1575 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
1576 | zap.za_normalization_conflict = 0; | |
1577 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 1578 | type = DT_DIR; |
34dc7c2f BB |
1579 | } else { |
1580 | /* | |
1581 | * Grab next entry. | |
1582 | */ | |
3558fd73 BB |
1583 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
1584 | if (error == ENOENT) | |
34dc7c2f BB |
1585 | break; |
1586 | else | |
1587 | goto update; | |
1588 | } | |
1589 | ||
0c5dde49 BB |
1590 | /* |
1591 | * Allow multiple entries provided the first entry is | |
1592 | * the object id. Non-zpl consumers may safely make | |
1593 | * use of the additional space. | |
1594 | * | |
1595 | * XXX: This should be a feature flag for compatibility | |
1596 | */ | |
34dc7c2f | 1597 | if (zap.za_integer_length != 8 || |
0c5dde49 | 1598 | zap.za_num_integers == 0) { |
34dc7c2f | 1599 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
1600 | "entry, obj = %lld, offset = %lld, " |
1601 | "length = %d, num = %lld\n", | |
34dc7c2f | 1602 | (u_longlong_t)zp->z_id, |
c12e3a59 | 1603 | (u_longlong_t)offset, |
0c5dde49 BB |
1604 | zap.za_integer_length, |
1605 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 1606 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
1607 | goto update; |
1608 | } | |
1609 | ||
1610 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 1611 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 1612 | } |
0f37d0c8 | 1613 | |
9464b959 | 1614 | done = !zpl_dir_emit(ctx, zap.za_name, strlen(zap.za_name), |
c12e3a59 | 1615 | objnum, type); |
0f37d0c8 | 1616 | if (done) |
34dc7c2f | 1617 | break; |
34dc7c2f | 1618 | |
6c94e649 AM |
1619 | if (prefetch) |
1620 | dmu_prefetch_dnode(os, objnum, ZIO_PRIORITY_SYNC_READ); | |
34dc7c2f | 1621 | |
c12e3a59 RY |
1622 | /* |
1623 | * Move to the next entry, fill in the previous offset. | |
1624 | */ | |
1625 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 1626 | zap_cursor_advance(&zc); |
c12e3a59 | 1627 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 1628 | } else { |
c12e3a59 | 1629 | offset += 1; |
34dc7c2f | 1630 | } |
c12e3a59 | 1631 | ctx->pos = offset; |
34dc7c2f BB |
1632 | } |
1633 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
1634 | ||
34dc7c2f BB |
1635 | update: |
1636 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
1637 | if (error == ENOENT) |
1638 | error = 0; | |
3558fd73 | 1639 | out: |
768eaced | 1640 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f | 1641 | |
34dc7c2f BB |
1642 | return (error); |
1643 | } | |
1644 | ||
057e8eee BB |
1645 | /* |
1646 | * Get the basic file attributes and place them in the provided kstat | |
1647 | * structure. The inode is assumed to be the authoritative source | |
1648 | * for most of the attributes. However, the znode currently has the | |
1649 | * authoritative atime, blksize, and block count. | |
1650 | * | |
1651 | * IN: ip - inode of file. | |
1652 | * | |
1653 | * OUT: sp - kstat values. | |
1654 | * | |
1655 | * RETURN: 0 (always succeeds) | |
1656 | */ | |
057e8eee | 1657 | int |
01d00dfa CK |
1658 | #ifdef HAVE_GENERIC_FILLATTR_IDMAP_REQMASK |
1659 | zfs_getattr_fast(zidmap_t *user_ns, u32 request_mask, struct inode *ip, | |
1660 | struct kstat *sp) | |
1661 | #else | |
d4dc53da | 1662 | zfs_getattr_fast(zidmap_t *user_ns, struct inode *ip, struct kstat *sp) |
01d00dfa | 1663 | #endif |
057e8eee BB |
1664 | { |
1665 | znode_t *zp = ITOZ(ip); | |
0037b49e | 1666 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
b585bc4a BB |
1667 | uint32_t blksize; |
1668 | u_longlong_t nblocks; | |
768eaced | 1669 | int error; |
057e8eee | 1670 | |
768eaced CC |
1671 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
1672 | return (error); | |
a7b125e9 | 1673 | |
057e8eee BB |
1674 | mutex_enter(&zp->z_lock); |
1675 | ||
01d00dfa CK |
1676 | #ifdef HAVE_GENERIC_FILLATTR_IDMAP_REQMASK |
1677 | zpl_generic_fillattr(user_ns, request_mask, ip, sp); | |
1678 | #else | |
e2a82961 | 1679 | zpl_generic_fillattr(user_ns, ip, sp); |
01d00dfa | 1680 | #endif |
97aa3ba4 TK |
1681 | /* |
1682 | * +1 link count for root inode with visible '.zfs' directory. | |
1683 | */ | |
1684 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) | |
1685 | if (sp->nlink < ZFS_LINK_MAX) | |
1686 | sp->nlink++; | |
057e8eee | 1687 | |
b585bc4a BB |
1688 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
1689 | sp->blksize = blksize; | |
1690 | sp->blocks = nblocks; | |
1691 | ||
057e8eee BB |
1692 | if (unlikely(zp->z_blksz == 0)) { |
1693 | /* | |
1694 | * Block size hasn't been set; suggest maximal I/O transfers. | |
1695 | */ | |
0037b49e | 1696 | sp->blksize = zfsvfs->z_max_blksz; |
057e8eee BB |
1697 | } |
1698 | ||
1699 | mutex_exit(&zp->z_lock); | |
1700 | ||
aa9b2708 AV |
1701 | /* |
1702 | * Required to prevent NFS client from detecting different inode | |
1703 | * numbers of snapshot root dentry before and after snapshot mount. | |
1704 | */ | |
0037b49e | 1705 | if (zfsvfs->z_issnap) { |
aa9b2708 AV |
1706 | if (ip->i_sb->s_root->d_inode == ip) |
1707 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
0037b49e | 1708 | dmu_objset_id(zfsvfs->z_os); |
aa9b2708 AV |
1709 | } |
1710 | ||
768eaced | 1711 | zfs_exit(zfsvfs, FTAG); |
a7b125e9 | 1712 | |
057e8eee BB |
1713 | return (0); |
1714 | } | |
057e8eee | 1715 | |
9c5167d1 NF |
1716 | /* |
1717 | * For the operation of changing file's user/group/project, we need to | |
1718 | * handle not only the main object that is assigned to the file directly, | |
1719 | * but also the ones that are used by the file via hidden xattr directory. | |
1720 | * | |
1721 | * Because the xattr directory may contains many EA entries, as to it may | |
1722 | * be impossible to change all of them via the transaction of changing the | |
1723 | * main object's user/group/project attributes. Then we have to change them | |
1724 | * via other multiple independent transactions one by one. It may be not good | |
1725 | * solution, but we have no better idea yet. | |
1726 | */ | |
1727 | static int | |
1728 | zfs_setattr_dir(znode_t *dzp) | |
1729 | { | |
1730 | struct inode *dxip = ZTOI(dzp); | |
1731 | struct inode *xip = NULL; | |
657ce253 | 1732 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
9c5167d1 NF |
1733 | objset_t *os = zfsvfs->z_os; |
1734 | zap_cursor_t zc; | |
1735 | zap_attribute_t zap; | |
1736 | zfs_dirlock_t *dl; | |
b7ab7ae2 | 1737 | znode_t *zp = NULL; |
9c5167d1 NF |
1738 | dmu_tx_t *tx = NULL; |
1739 | uint64_t uid, gid; | |
1740 | sa_bulk_attr_t bulk[4]; | |
8cb34421 | 1741 | int count; |
9c5167d1 NF |
1742 | int err; |
1743 | ||
1744 | zap_cursor_init(&zc, os, dzp->z_id); | |
1745 | while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) { | |
8cb34421 | 1746 | count = 0; |
9c5167d1 NF |
1747 | if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { |
1748 | err = ENXIO; | |
1749 | break; | |
1750 | } | |
1751 | ||
1752 | err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp, | |
1753 | ZEXISTS, NULL, NULL); | |
1754 | if (err == ENOENT) | |
1755 | goto next; | |
1756 | if (err) | |
1757 | break; | |
1758 | ||
1759 | xip = ZTOI(zp); | |
1760 | if (KUID_TO_SUID(xip->i_uid) == KUID_TO_SUID(dxip->i_uid) && | |
1761 | KGID_TO_SGID(xip->i_gid) == KGID_TO_SGID(dxip->i_gid) && | |
1762 | zp->z_projid == dzp->z_projid) | |
1763 | goto next; | |
1764 | ||
1765 | tx = dmu_tx_create(os); | |
1766 | if (!(zp->z_pflags & ZFS_PROJID)) | |
1767 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
1768 | else | |
1769 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
1770 | ||
1771 | err = dmu_tx_assign(tx, TXG_WAIT); | |
1772 | if (err) | |
1773 | break; | |
1774 | ||
1775 | mutex_enter(&dzp->z_lock); | |
1776 | ||
1777 | if (KUID_TO_SUID(xip->i_uid) != KUID_TO_SUID(dxip->i_uid)) { | |
1778 | xip->i_uid = dxip->i_uid; | |
1779 | uid = zfs_uid_read(dxip); | |
1780 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, | |
1781 | &uid, sizeof (uid)); | |
1782 | } | |
1783 | ||
1784 | if (KGID_TO_SGID(xip->i_gid) != KGID_TO_SGID(dxip->i_gid)) { | |
1785 | xip->i_gid = dxip->i_gid; | |
1786 | gid = zfs_gid_read(dxip); | |
1787 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, | |
1788 | &gid, sizeof (gid)); | |
1789 | } | |
1790 | ||
1791 | if (zp->z_projid != dzp->z_projid) { | |
1792 | if (!(zp->z_pflags & ZFS_PROJID)) { | |
1793 | zp->z_pflags |= ZFS_PROJID; | |
1794 | SA_ADD_BULK_ATTR(bulk, count, | |
1795 | SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, | |
1796 | sizeof (zp->z_pflags)); | |
1797 | } | |
1798 | ||
1799 | zp->z_projid = dzp->z_projid; | |
1800 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs), | |
1801 | NULL, &zp->z_projid, sizeof (zp->z_projid)); | |
1802 | } | |
1803 | ||
1804 | mutex_exit(&dzp->z_lock); | |
1805 | ||
1806 | if (likely(count > 0)) { | |
1807 | err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
1808 | dmu_tx_commit(tx); | |
1809 | } else { | |
1810 | dmu_tx_abort(tx); | |
1811 | } | |
1812 | tx = NULL; | |
1813 | if (err != 0 && err != ENOENT) | |
1814 | break; | |
1815 | ||
1816 | next: | |
657ce253 MM |
1817 | if (zp) { |
1818 | zrele(zp); | |
1819 | zp = NULL; | |
9c5167d1 NF |
1820 | zfs_dirent_unlock(dl); |
1821 | } | |
1822 | zap_cursor_advance(&zc); | |
1823 | } | |
1824 | ||
1825 | if (tx) | |
1826 | dmu_tx_abort(tx); | |
657ce253 MM |
1827 | if (zp) { |
1828 | zrele(zp); | |
9c5167d1 NF |
1829 | zfs_dirent_unlock(dl); |
1830 | } | |
1831 | zap_cursor_fini(&zc); | |
1832 | ||
1833 | return (err == ENOENT ? 0 : err); | |
1834 | } | |
1835 | ||
34dc7c2f BB |
1836 | /* |
1837 | * Set the file attributes to the values contained in the | |
1838 | * vattr structure. | |
1839 | * | |
657ce253 | 1840 | * IN: zp - znode of file to be modified. |
34dc7c2f | 1841 | * vap - new attribute values. |
5484965a | 1842 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
1843 | * flags - ATTR_UTIME set if non-default time values provided. |
1844 | * - ATTR_NOACLCHECK (CIFS context only). | |
1845 | * cr - credentials of caller. | |
2a068a13 | 1846 | * mnt_ns - user namespace of the mount |
34dc7c2f BB |
1847 | * |
1848 | * RETURN: 0 if success | |
1849 | * error code if failure | |
1850 | * | |
1851 | * Timestamps: | |
3558fd73 | 1852 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f | 1853 | */ |
e5c39b95 | 1854 | int |
d4dc53da | 1855 | zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr, zidmap_t *mnt_ns) |
34dc7c2f | 1856 | { |
657ce253 MM |
1857 | struct inode *ip; |
1858 | zfsvfs_t *zfsvfs = ZTOZSB(zp); | |
d8b26866 | 1859 | objset_t *os; |
34dc7c2f BB |
1860 | zilog_t *zilog; |
1861 | dmu_tx_t *tx; | |
1862 | vattr_t oldva; | |
f4ea75d4 | 1863 | xvattr_t *tmpxvattr; |
5484965a | 1864 | uint_t mask = vap->va_mask; |
a117a6d6 | 1865 | uint_t saved_mask = 0; |
34dc7c2f BB |
1866 | int trim_mask = 0; |
1867 | uint64_t new_mode; | |
64aefee1 | 1868 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 1869 | uint64_t xattr_obj; |
0df9673f | 1870 | uint64_t mtime[2], ctime[2], atime[2]; |
9c5167d1 | 1871 | uint64_t projid = ZFS_INVALID_PROJID; |
34dc7c2f BB |
1872 | znode_t *attrzp; |
1873 | int need_policy = FALSE; | |
9c5167d1 | 1874 | int err, err2 = 0; |
34dc7c2f | 1875 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
1876 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
1877 | xoptattr_t *xoap; | |
1878 | zfs_acl_t *aclp; | |
34dc7c2f | 1879 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 1880 | boolean_t fuid_dirtied = B_FALSE; |
9c5167d1 | 1881 | boolean_t handle_eadir = B_FALSE; |
17c37660 | 1882 | sa_bulk_attr_t *bulk, *xattr_bulk; |
9c5167d1 | 1883 | int count = 0, xattr_count = 0, bulks = 8; |
34dc7c2f BB |
1884 | |
1885 | if (mask == 0) | |
1886 | return (0); | |
1887 | ||
768eaced CC |
1888 | if ((err = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
1889 | return (err); | |
657ce253 | 1890 | ip = ZTOI(zp); |
d8b26866 | 1891 | os = zfsvfs->z_os; |
34dc7c2f | 1892 | |
9c5167d1 NF |
1893 | /* |
1894 | * If this is a xvattr_t, then get a pointer to the structure of | |
1895 | * optional attributes. If this is NULL, then we have a vattr_t. | |
1896 | */ | |
1897 | xoap = xva_getxoptattr(xvap); | |
1898 | if (xoap != NULL && (mask & ATTR_XVATTR)) { | |
1899 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
1900 | if (!dmu_objset_projectquota_enabled(os) || | |
1901 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode))) { | |
768eaced | 1902 | zfs_exit(zfsvfs, FTAG); |
9c5167d1 NF |
1903 | return (SET_ERROR(ENOTSUP)); |
1904 | } | |
1905 | ||
1906 | projid = xoap->xoa_projid; | |
1907 | if (unlikely(projid == ZFS_INVALID_PROJID)) { | |
768eaced | 1908 | zfs_exit(zfsvfs, FTAG); |
9c5167d1 NF |
1909 | return (SET_ERROR(EINVAL)); |
1910 | } | |
1911 | ||
1912 | if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID) | |
1913 | projid = ZFS_INVALID_PROJID; | |
1914 | else | |
1915 | need_policy = TRUE; | |
1916 | } | |
1917 | ||
1918 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) && | |
2705ebf0 NF |
1919 | (xoap->xoa_projinherit != |
1920 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) && | |
9c5167d1 NF |
1921 | (!dmu_objset_projectquota_enabled(os) || |
1922 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode)))) { | |
768eaced | 1923 | zfs_exit(zfsvfs, FTAG); |
2705ebf0 | 1924 | return (SET_ERROR(ENOTSUP)); |
9c5167d1 NF |
1925 | } |
1926 | } | |
1927 | ||
0037b49e | 1928 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1929 | |
1930 | /* | |
1931 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
1932 | * that file system is at proper version level | |
1933 | */ | |
5484965a | 1934 | |
0037b49e | 1935 | if (zfsvfs->z_use_fuids == B_FALSE && |
5484965a BB |
1936 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
1937 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
1938 | (mask & ATTR_XVATTR))) { | |
768eaced | 1939 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1940 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1941 | } |
1942 | ||
3558fd73 | 1943 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
768eaced | 1944 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1945 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
1946 | } |
1947 | ||
3558fd73 | 1948 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
768eaced | 1949 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1950 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1951 | } |
1952 | ||
d1d7e268 | 1953 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 1954 | xva_init(tmpxvattr); |
5484965a | 1955 | |
9c5167d1 NF |
1956 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); |
1957 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); | |
17c37660 | 1958 | |
5484965a BB |
1959 | /* |
1960 | * Immutable files can only alter immutable bit and atime | |
1961 | */ | |
1962 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
1963 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
1964 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
ecb2b7dc | 1965 | err = SET_ERROR(EPERM); |
f4ea75d4 | 1966 | goto out3; |
5484965a BB |
1967 | } |
1968 | ||
3558fd73 | 1969 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
ecb2b7dc | 1970 | err = SET_ERROR(EPERM); |
f4ea75d4 | 1971 | goto out3; |
34dc7c2f BB |
1972 | } |
1973 | ||
5484965a BB |
1974 | /* |
1975 | * Verify timestamps doesn't overflow 32 bits. | |
1976 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
1977 | * handle times greater than 2039. This check should be removed | |
1978 | * once large timestamps are fully supported. | |
1979 | */ | |
1980 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
1981 | if (((mask & ATTR_ATIME) && |
1982 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
1983 | ((mask & ATTR_MTIME) && | |
1984 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
ecb2b7dc | 1985 | err = SET_ERROR(EOVERFLOW); |
f4ea75d4 | 1986 | goto out3; |
5484965a BB |
1987 | } |
1988 | } | |
1989 | ||
34dc7c2f BB |
1990 | top: |
1991 | attrzp = NULL; | |
572e2857 | 1992 | aclp = NULL; |
34dc7c2f | 1993 | |
45d1cae3 | 1994 | /* Can this be moved to before the top label? */ |
0037b49e | 1995 | if (zfs_is_readonly(zfsvfs)) { |
ecb2b7dc | 1996 | err = SET_ERROR(EROFS); |
f4ea75d4 | 1997 | goto out3; |
34dc7c2f BB |
1998 | } |
1999 | ||
2000 | /* | |
2001 | * First validate permissions | |
2002 | */ | |
2003 | ||
3558fd73 | 2004 | if (mask & ATTR_SIZE) { |
2a068a13 YY |
2005 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr, |
2006 | mnt_ns); | |
f4ea75d4 BB |
2007 | if (err) |
2008 | goto out3; | |
2009 | ||
34dc7c2f BB |
2010 | /* |
2011 | * XXX - Note, we are not providing any open | |
2012 | * mode flags here (like FNDELAY), so we may | |
2013 | * block if there are locks present... this | |
2014 | * should be addressed in openat(). | |
2015 | */ | |
b128c09f | 2016 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2017 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2018 | if (err) |
2019 | goto out3; | |
428870ff | 2020 | } |
34dc7c2f | 2021 | |
5484965a BB |
2022 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2023 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2024 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2025 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2026 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2027 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2028 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2029 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2030 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2a068a13 | 2031 | skipaclchk, cr, mnt_ns); |
5484965a BB |
2032 | } |
2033 | ||
3558fd73 BB |
2034 | if (mask & (ATTR_UID|ATTR_GID)) { |
2035 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2036 | int take_owner; |
2037 | int take_group; | |
2a068a13 YY |
2038 | uid_t uid; |
2039 | gid_t gid; | |
34dc7c2f BB |
2040 | |
2041 | /* | |
2042 | * NOTE: even if a new mode is being set, | |
2043 | * we may clear S_ISUID/S_ISGID bits. | |
2044 | */ | |
2045 | ||
3558fd73 | 2046 | if (!(mask & ATTR_MODE)) |
5484965a | 2047 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2048 | |
2049 | /* | |
2050 | * Take ownership or chgrp to group we are a member of | |
2051 | */ | |
2052 | ||
d4dc53da YY |
2053 | uid = zfs_uid_to_vfsuid(mnt_ns, zfs_i_user_ns(ip), |
2054 | vap->va_uid); | |
2055 | gid = zfs_gid_to_vfsgid(mnt_ns, zfs_i_user_ns(ip), | |
2056 | vap->va_gid); | |
2a068a13 | 2057 | take_owner = (mask & ATTR_UID) && (uid == crgetuid(cr)); |
3558fd73 | 2058 | take_group = (mask & ATTR_GID) && |
2a068a13 | 2059 | zfs_groupmember(zfsvfs, gid, cr); |
34dc7c2f BB |
2060 | |
2061 | /* | |
5484965a | 2062 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
2063 | * take_group must both be set in order to allow taking |
2064 | * ownership. | |
2065 | * | |
2066 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2067 | * | |
2068 | */ | |
2069 | ||
3558fd73 BB |
2070 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
2071 | take_owner && take_group) || | |
2072 | ((idmask == ATTR_UID) && take_owner) || | |
2073 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f | 2074 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
2a068a13 | 2075 | skipaclchk, cr, mnt_ns) == 0) { |
34dc7c2f BB |
2076 | /* |
2077 | * Remove setuid/setgid for non-privileged users | |
2078 | */ | |
5484965a | 2079 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 2080 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
2081 | } else { |
2082 | need_policy = TRUE; | |
2083 | } | |
2084 | } else { | |
2085 | need_policy = TRUE; | |
2086 | } | |
2087 | } | |
2088 | ||
2089 | mutex_enter(&zp->z_lock); | |
428870ff | 2090 | oldva.va_mode = zp->z_mode; |
572e2857 | 2091 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
2092 | if (mask & ATTR_XVATTR) { |
2093 | /* | |
2094 | * Update xvattr mask to include only those attributes | |
2095 | * that are actually changing. | |
2096 | * | |
2097 | * the bits will be restored prior to actually setting | |
2098 | * the attributes so the caller thinks they were set. | |
2099 | */ | |
2100 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2101 | if (xoap->xoa_appendonly != | |
2102 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
2103 | need_policy = TRUE; | |
2104 | } else { | |
2105 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 2106 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
2107 | } |
2108 | } | |
2109 | ||
9c5167d1 NF |
2110 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { |
2111 | if (xoap->xoa_projinherit != | |
2112 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) { | |
2113 | need_policy = TRUE; | |
2114 | } else { | |
2115 | XVA_CLR_REQ(xvap, XAT_PROJINHERIT); | |
2116 | XVA_SET_REQ(tmpxvattr, XAT_PROJINHERIT); | |
2117 | } | |
2118 | } | |
2119 | ||
5484965a BB |
2120 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { |
2121 | if (xoap->xoa_nounlink != | |
2122 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
2123 | need_policy = TRUE; | |
2124 | } else { | |
2125 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 2126 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
2127 | } |
2128 | } | |
2129 | ||
2130 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2131 | if (xoap->xoa_immutable != | |
2132 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
2133 | need_policy = TRUE; | |
2134 | } else { | |
2135 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 2136 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
2137 | } |
2138 | } | |
2139 | ||
2140 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2141 | if (xoap->xoa_nodump != | |
2142 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
2143 | need_policy = TRUE; | |
2144 | } else { | |
2145 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 2146 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
2147 | } |
2148 | } | |
2149 | ||
2150 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2151 | if (xoap->xoa_av_modified != | |
2152 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
2153 | need_policy = TRUE; | |
2154 | } else { | |
2155 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 2156 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
2157 | } |
2158 | } | |
2159 | ||
2160 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2161 | if ((!S_ISREG(ip->i_mode) && | |
2162 | xoap->xoa_av_quarantined) || | |
2163 | xoap->xoa_av_quarantined != | |
2164 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
2165 | need_policy = TRUE; | |
2166 | } else { | |
2167 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 2168 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
2169 | } |
2170 | } | |
2171 | ||
2172 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2173 | mutex_exit(&zp->z_lock); | |
ecb2b7dc | 2174 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2175 | goto out3; |
5484965a BB |
2176 | } |
2177 | ||
2178 | if (need_policy == FALSE && | |
2179 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2180 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
2181 | need_policy = TRUE; | |
2182 | } | |
2183 | } | |
34dc7c2f BB |
2184 | |
2185 | mutex_exit(&zp->z_lock); | |
2186 | ||
3558fd73 | 2187 | if (mask & ATTR_MODE) { |
2a068a13 YY |
2188 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr, |
2189 | mnt_ns) == 0) { | |
5484965a | 2190 | err = secpolicy_setid_setsticky_clear(ip, vap, |
f224eddf | 2191 | &oldva, cr, mnt_ns, zfs_i_user_ns(ip)); |
f4ea75d4 BB |
2192 | if (err) |
2193 | goto out3; | |
3558fd73 | 2194 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
2195 | } else { |
2196 | need_policy = TRUE; | |
2197 | } | |
2198 | } | |
2199 | ||
2200 | if (need_policy) { | |
2201 | /* | |
2202 | * If trim_mask is set then take ownership | |
2203 | * has been granted or write_acl is present and user | |
2204 | * has the ability to modify mode. In that case remove | |
2205 | * UID|GID and or MODE from mask so that | |
2206 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2207 | */ | |
2208 | ||
2209 | if (trim_mask) { | |
5484965a BB |
2210 | saved_mask = vap->va_mask; |
2211 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 2212 | } |
5484965a | 2213 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
b5459dd3 | 2214 | zfs_zaccess_unix, zp); |
f4ea75d4 BB |
2215 | if (err) |
2216 | goto out3; | |
34dc7c2f BB |
2217 | |
2218 | if (trim_mask) | |
5484965a | 2219 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
2220 | } |
2221 | ||
2222 | /* | |
2223 | * secpolicy_vnode_setattr, or take ownership may have | |
2224 | * changed va_mask | |
2225 | */ | |
5484965a | 2226 | mask = vap->va_mask; |
34dc7c2f | 2227 | |
9c5167d1 NF |
2228 | if ((mask & (ATTR_UID | ATTR_GID)) || projid != ZFS_INVALID_PROJID) { |
2229 | handle_eadir = B_TRUE; | |
0037b49e | 2230 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
572e2857 | 2231 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 2232 | |
572e2857 | 2233 | if (err == 0 && xattr_obj) { |
3558fd73 | 2234 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
2235 | if (err) |
2236 | goto out2; | |
2237 | } | |
3558fd73 | 2238 | if (mask & ATTR_UID) { |
0037b49e | 2239 | new_kuid = zfs_fuid_create(zfsvfs, |
5484965a | 2240 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
64aefee1 | 2241 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
9c5167d1 NF |
2242 | zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT, |
2243 | new_kuid)) { | |
572e2857 | 2244 | if (attrzp) |
657ce253 | 2245 | zrele(attrzp); |
ecb2b7dc | 2246 | err = SET_ERROR(EDQUOT); |
428870ff BB |
2247 | goto out2; |
2248 | } | |
2249 | } | |
2250 | ||
3558fd73 | 2251 | if (mask & ATTR_GID) { |
0037b49e BB |
2252 | new_kgid = zfs_fuid_create(zfsvfs, |
2253 | (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); | |
64aefee1 | 2254 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && |
9c5167d1 NF |
2255 | zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT, |
2256 | new_kgid)) { | |
572e2857 | 2257 | if (attrzp) |
657ce253 | 2258 | zrele(attrzp); |
ecb2b7dc | 2259 | err = SET_ERROR(EDQUOT); |
428870ff BB |
2260 | goto out2; |
2261 | } | |
2262 | } | |
9c5167d1 NF |
2263 | |
2264 | if (projid != ZFS_INVALID_PROJID && | |
2265 | zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) { | |
2266 | if (attrzp) | |
657ce253 | 2267 | zrele(attrzp); |
9c5167d1 NF |
2268 | err = EDQUOT; |
2269 | goto out2; | |
2270 | } | |
428870ff | 2271 | } |
9c5167d1 | 2272 | tx = dmu_tx_create(os); |
34dc7c2f | 2273 | |
3558fd73 | 2274 | if (mask & ATTR_MODE) { |
428870ff | 2275 | uint64_t pmode = zp->z_mode; |
572e2857 | 2276 | uint64_t acl_obj; |
5484965a | 2277 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 2278 | |
7bf3e1fa PH |
2279 | if (ZTOZSB(zp)->z_acl_mode == ZFS_ACL_RESTRICTED && |
2280 | !(zp->z_pflags & ZFS_ACL_TRIVIAL)) { | |
2281 | err = EPERM; | |
2282 | goto out; | |
2283 | } | |
2284 | ||
156f74fc MM |
2285 | if ((err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))) |
2286 | goto out; | |
428870ff | 2287 | |
572e2857 BB |
2288 | mutex_enter(&zp->z_lock); |
2289 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
2290 | /* |
2291 | * Are we upgrading ACL from old V0 format | |
2292 | * to V1 format? | |
2293 | */ | |
0037b49e | 2294 | if (zfsvfs->z_version >= ZPL_VERSION_FUID && |
572e2857 | 2295 | zfs_znode_acl_version(zp) == |
34dc7c2f | 2296 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 2297 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
2298 | DMU_OBJECT_END); |
2299 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
2300 | 0, aclp->z_acl_bytes); | |
2301 | } else { | |
572e2857 | 2302 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
2303 | aclp->z_acl_bytes); |
2304 | } | |
428870ff | 2305 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
2306 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
2307 | 0, aclp->z_acl_bytes); | |
2308 | } | |
572e2857 | 2309 | mutex_exit(&zp->z_lock); |
428870ff BB |
2310 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
2311 | } else { | |
9c5167d1 NF |
2312 | if (((mask & ATTR_XVATTR) && |
2313 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) || | |
2314 | (projid != ZFS_INVALID_PROJID && | |
2315 | !(zp->z_pflags & ZFS_PROJID))) | |
5484965a BB |
2316 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
2317 | else | |
2318 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2319 | } |
2320 | ||
428870ff BB |
2321 | if (attrzp) { |
2322 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2323 | } |
2324 | ||
0037b49e | 2325 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
428870ff | 2326 | if (fuid_dirtied) |
0037b49e | 2327 | zfs_fuid_txhold(zfsvfs, tx); |
428870ff BB |
2328 | |
2329 | zfs_sa_upgrade_txholds(tx, zp); | |
2330 | ||
384f8a09 MA |
2331 | err = dmu_tx_assign(tx, TXG_WAIT); |
2332 | if (err) | |
9babb374 | 2333 | goto out; |
34dc7c2f | 2334 | |
428870ff | 2335 | count = 0; |
34dc7c2f BB |
2336 | /* |
2337 | * Set each attribute requested. | |
2338 | * We group settings according to the locks they need to acquire. | |
2339 | * | |
2340 | * Note: you cannot set ctime directly, although it will be | |
2341 | * updated as a side-effect of calling this function. | |
2342 | */ | |
2343 | ||
9c5167d1 NF |
2344 | if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) { |
2345 | /* | |
2346 | * For the existed object that is upgraded from old system, | |
2347 | * its on-disk layout has no slot for the project ID attribute. | |
2348 | * But quota accounting logic needs to access related slots by | |
2349 | * offset directly. So we need to adjust old objects' layout | |
2350 | * to make the project ID to some unified and fixed offset. | |
2351 | */ | |
2352 | if (attrzp) | |
2353 | err = sa_add_projid(attrzp->z_sa_hdl, tx, projid); | |
2354 | if (err == 0) | |
2355 | err = sa_add_projid(zp->z_sa_hdl, tx, projid); | |
2356 | ||
2357 | if (unlikely(err == EEXIST)) | |
2358 | err = 0; | |
2359 | else if (err != 0) | |
2360 | goto out; | |
2361 | else | |
2362 | projid = ZFS_INVALID_PROJID; | |
2363 | } | |
572e2857 | 2364 | |
3558fd73 | 2365 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2366 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
2367 | mutex_enter(&zp->z_lock); |
2368 | ||
0037b49e | 2369 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
2370 | &zp->z_pflags, sizeof (zp->z_pflags)); |
2371 | ||
2372 | if (attrzp) { | |
3558fd73 | 2373 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2374 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
2375 | mutex_enter(&attrzp->z_lock); |
2376 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 2377 | SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, |
428870ff | 2378 | sizeof (attrzp->z_pflags)); |
9c5167d1 NF |
2379 | if (projid != ZFS_INVALID_PROJID) { |
2380 | attrzp->z_projid = projid; | |
2381 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
2382 | SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid, | |
2383 | sizeof (attrzp->z_projid)); | |
2384 | } | |
428870ff BB |
2385 | } |
2386 | ||
3558fd73 | 2387 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 2388 | |
3558fd73 | 2389 | if (mask & ATTR_UID) { |
64aefee1 NB |
2390 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
2391 | new_uid = zfs_uid_read(ZTOI(zp)); | |
0037b49e | 2392 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 2393 | &new_uid, sizeof (new_uid)); |
428870ff BB |
2394 | if (attrzp) { |
2395 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 2396 | SA_ZPL_UID(zfsvfs), NULL, &new_uid, |
428870ff | 2397 | sizeof (new_uid)); |
2c6abf15 | 2398 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
2399 | } |
2400 | } | |
2401 | ||
3558fd73 | 2402 | if (mask & ATTR_GID) { |
64aefee1 NB |
2403 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
2404 | new_gid = zfs_gid_read(ZTOI(zp)); | |
0037b49e | 2405 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), |
428870ff | 2406 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
2407 | if (attrzp) { |
2408 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 2409 | SA_ZPL_GID(zfsvfs), NULL, &new_gid, |
428870ff | 2410 | sizeof (new_gid)); |
64aefee1 | 2411 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
2412 | } |
2413 | } | |
3558fd73 | 2414 | if (!(mask & ATTR_MODE)) { |
0037b49e | 2415 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), |
428870ff BB |
2416 | NULL, &new_mode, sizeof (new_mode)); |
2417 | new_mode = zp->z_mode; | |
2418 | } | |
2419 | err = zfs_acl_chown_setattr(zp); | |
2420 | ASSERT(err == 0); | |
2421 | if (attrzp) { | |
2422 | err = zfs_acl_chown_setattr(attrzp); | |
2423 | ASSERT(err == 0); | |
2424 | } | |
2425 | } | |
2426 | ||
3558fd73 | 2427 | if (mask & ATTR_MODE) { |
0037b49e | 2428 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 2429 | &new_mode, sizeof (new_mode)); |
12fa7f34 | 2430 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 2431 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 2432 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 2433 | ASSERT0(err); |
572e2857 BB |
2434 | if (zp->z_acl_cached) |
2435 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
2436 | zp->z_acl_cached = aclp; |
2437 | aclp = NULL; | |
34dc7c2f BB |
2438 | } |
2439 | ||
704cd075 | 2440 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
a43570c5 | 2441 | zp->z_atime_dirty = B_FALSE; |
f0bf7a24 | 2442 | inode_timespec_t tmp_atime = zpl_inode_get_atime(ip); |
db4fc559 | 2443 | ZFS_TIME_ENCODE(&tmp_atime, atime); |
0037b49e | 2444 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
0df9673f | 2445 | &atime, sizeof (atime)); |
34dc7c2f BB |
2446 | } |
2447 | ||
99834d19 | 2448 | if (mask & (ATTR_MTIME | ATTR_SIZE)) { |
5484965a | 2449 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
db4fc559 RN |
2450 | zpl_inode_set_mtime_to_ts(ZTOI(zp), |
2451 | zpl_inode_timestamp_truncate(vap->va_mtime, ZTOI(zp))); | |
87f9371a | 2452 | |
0037b49e | 2453 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 2454 | mtime, sizeof (mtime)); |
34dc7c2f BB |
2455 | } |
2456 | ||
99834d19 | 2457 | if (mask & (ATTR_CTIME | ATTR_SIZE)) { |
87f9371a | 2458 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); |
b37f2934 CK |
2459 | zpl_inode_set_ctime_to_ts(ZTOI(zp), |
2460 | zpl_inode_timestamp_truncate(vap->va_ctime, ZTOI(zp))); | |
0037b49e | 2461 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
87f9371a | 2462 | ctime, sizeof (ctime)); |
428870ff | 2463 | } |
87f9371a | 2464 | |
9c5167d1 NF |
2465 | if (projid != ZFS_INVALID_PROJID) { |
2466 | zp->z_projid = projid; | |
2467 | SA_ADD_BULK_ATTR(bulk, count, | |
2468 | SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid, | |
2469 | sizeof (zp->z_projid)); | |
2470 | } | |
2471 | ||
87f9371a NB |
2472 | if (attrzp && mask) { |
2473 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 2474 | SA_ZPL_CTIME(zfsvfs), NULL, &ctime, |
87f9371a NB |
2475 | sizeof (ctime)); |
2476 | } | |
2477 | ||
34dc7c2f BB |
2478 | /* |
2479 | * Do this after setting timestamps to prevent timestamp | |
2480 | * update from toggling bit | |
2481 | */ | |
2482 | ||
5484965a BB |
2483 | if (xoap && (mask & ATTR_XVATTR)) { |
2484 | ||
2485 | /* | |
2486 | * restore trimmed off masks | |
2487 | * so that return masks can be set for caller. | |
2488 | */ | |
2489 | ||
f4ea75d4 | 2490 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
2491 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
2492 | } | |
f4ea75d4 | 2493 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
2494 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
2495 | } | |
f4ea75d4 | 2496 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
2497 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
2498 | } | |
f4ea75d4 | 2499 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
2500 | XVA_SET_REQ(xvap, XAT_NODUMP); |
2501 | } | |
f4ea75d4 | 2502 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
2503 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
2504 | } | |
f4ea75d4 | 2505 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
2506 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
2507 | } | |
9c5167d1 NF |
2508 | if (XVA_ISSET_REQ(tmpxvattr, XAT_PROJINHERIT)) { |
2509 | XVA_SET_REQ(xvap, XAT_PROJINHERIT); | |
2510 | } | |
5484965a BB |
2511 | |
2512 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
2513 | ASSERT(S_ISREG(ip->i_mode)); | |
2514 | ||
2515 | zfs_xvattr_set(zp, xvap, tx); | |
2516 | } | |
2517 | ||
9babb374 | 2518 | if (fuid_dirtied) |
0037b49e | 2519 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 2520 | |
34dc7c2f | 2521 | if (mask != 0) |
5484965a | 2522 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 2523 | |
34dc7c2f | 2524 | mutex_exit(&zp->z_lock); |
3558fd73 | 2525 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2526 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 2527 | |
572e2857 | 2528 | if (attrzp) { |
3558fd73 | 2529 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
2530 | mutex_exit(&attrzp->z_acl_lock); |
2531 | mutex_exit(&attrzp->z_lock); | |
2532 | } | |
9babb374 | 2533 | out: |
9c5167d1 | 2534 | if (err == 0 && xattr_count > 0) { |
428870ff BB |
2535 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, |
2536 | xattr_count, tx); | |
2537 | ASSERT(err2 == 0); | |
2538 | } | |
2539 | ||
45d1cae3 | 2540 | if (aclp) |
9babb374 | 2541 | zfs_acl_free(aclp); |
9babb374 BB |
2542 | |
2543 | if (fuidp) { | |
2544 | zfs_fuid_info_free(fuidp); | |
2545 | fuidp = NULL; | |
2546 | } | |
2547 | ||
428870ff | 2548 | if (err) { |
9babb374 | 2549 | dmu_tx_abort(tx); |
ea7e86d8 | 2550 | if (attrzp) |
657ce253 | 2551 | zrele(attrzp); |
428870ff BB |
2552 | if (err == ERESTART) |
2553 | goto top; | |
2554 | } else { | |
9c5167d1 NF |
2555 | if (count > 0) |
2556 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 2557 | dmu_tx_commit(tx); |
9c5167d1 NF |
2558 | if (attrzp) { |
2559 | if (err2 == 0 && handle_eadir) | |
2e7f664f | 2560 | err = zfs_setattr_dir(attrzp); |
657ce253 | 2561 | zrele(attrzp); |
9c5167d1 | 2562 | } |
fc273894 | 2563 | zfs_znode_update_vfs(zp); |
428870ff BB |
2564 | } |
2565 | ||
428870ff | 2566 | out2: |
9c5167d1 | 2567 | if (os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2568 | zil_commit(zilog, 0); |
34dc7c2f | 2569 | |
f4ea75d4 | 2570 | out3: |
9c5167d1 NF |
2571 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks); |
2572 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * bulks); | |
d1d7e268 | 2573 | kmem_free(tmpxvattr, sizeof (xvattr_t)); |
768eaced | 2574 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
2575 | return (err); |
2576 | } | |
2577 | ||
2578 | typedef struct zfs_zlock { | |
2579 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
2580 | znode_t *zl_znode; /* znode we held */ | |
2581 | struct zfs_zlock *zl_next; /* next in list */ | |
2582 | } zfs_zlock_t; | |
2583 | ||
2584 | /* | |
2585 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
2586 | */ | |
2587 | static void | |
2588 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
2589 | { | |
2590 | zfs_zlock_t *zl; | |
2591 | ||
2592 | while ((zl = *zlpp) != NULL) { | |
2593 | if (zl->zl_znode != NULL) | |
657ce253 | 2594 | zfs_zrele_async(zl->zl_znode); |
34dc7c2f BB |
2595 | rw_exit(zl->zl_rwlock); |
2596 | *zlpp = zl->zl_next; | |
2597 | kmem_free(zl, sizeof (*zl)); | |
2598 | } | |
2599 | } | |
2600 | ||
2601 | /* | |
2602 | * Search back through the directory tree, using the ".." entries. | |
2603 | * Lock each directory in the chain to prevent concurrent renames. | |
2604 | * Fail any attempt to move a directory into one of its own descendants. | |
2605 | * XXX - z_parent_lock can overlap with map or grow locks | |
2606 | */ | |
2607 | static int | |
2608 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
2609 | { | |
2610 | zfs_zlock_t *zl; | |
2611 | znode_t *zp = tdzp; | |
3558fd73 | 2612 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 2613 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
2614 | krwlock_t *rwlp = &szp->z_parent_lock; |
2615 | krw_t rw = RW_WRITER; | |
2616 | ||
2617 | /* | |
2618 | * First pass write-locks szp and compares to zp->z_id. | |
2619 | * Later passes read-lock zp and compare to zp->z_parent. | |
2620 | */ | |
2621 | do { | |
2622 | if (!rw_tryenter(rwlp, rw)) { | |
2623 | /* | |
2624 | * Another thread is renaming in this path. | |
2625 | * Note that if we are a WRITER, we don't have any | |
2626 | * parent_locks held yet. | |
2627 | */ | |
2628 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
2629 | /* | |
2630 | * Drop our locks and restart | |
2631 | */ | |
2632 | zfs_rename_unlock(&zl); | |
2633 | *zlpp = NULL; | |
2634 | zp = tdzp; | |
428870ff | 2635 | oidp = zp->z_id; |
34dc7c2f BB |
2636 | rwlp = &szp->z_parent_lock; |
2637 | rw = RW_WRITER; | |
2638 | continue; | |
2639 | } else { | |
2640 | /* | |
2641 | * Wait for other thread to drop its locks | |
2642 | */ | |
2643 | rw_enter(rwlp, rw); | |
2644 | } | |
2645 | } | |
2646 | ||
2647 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
2648 | zl->zl_rwlock = rwlp; | |
2649 | zl->zl_znode = NULL; | |
2650 | zl->zl_next = *zlpp; | |
2651 | *zlpp = zl; | |
2652 | ||
428870ff | 2653 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 2654 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 2655 | |
428870ff | 2656 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
2657 | return (0); |
2658 | ||
2659 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 2660 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
2661 | if (error) |
2662 | return (error); | |
2663 | zl->zl_znode = zp; | |
2664 | } | |
3558fd73 | 2665 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 2666 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
2667 | rwlp = &zp->z_parent_lock; |
2668 | rw = RW_READER; | |
2669 | ||
2670 | } while (zp->z_id != sdzp->z_id); | |
2671 | ||
2672 | return (0); | |
2673 | } | |
2674 | ||
2675 | /* | |
2676 | * Move an entry from the provided source directory to the target | |
2677 | * directory. Change the entry name as indicated. | |
2678 | * | |
657ce253 | 2679 | * IN: sdzp - Source directory containing the "old entry". |
34dc7c2f | 2680 | * snm - Old entry name. |
657ce253 | 2681 | * tdzp - Target directory to contain the "new entry". |
34dc7c2f BB |
2682 | * tnm - New entry name. |
2683 | * cr - credentials of caller. | |
34dc7c2f | 2684 | * flags - case flags |
dbf6108b AS |
2685 | * rflags - RENAME_* flags |
2686 | * wa_vap - attributes for RENAME_WHITEOUT (must be a char 0:0). | |
2a068a13 | 2687 | * mnt_ns - user namespace of the mount |
34dc7c2f | 2688 | * |
d3cc8b15 | 2689 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
2690 | * |
2691 | * Timestamps: | |
657ce253 | 2692 | * sdzp,tdzp - ctime|mtime updated |
34dc7c2f | 2693 | */ |
e5c39b95 | 2694 | int |
657ce253 | 2695 | zfs_rename(znode_t *sdzp, char *snm, znode_t *tdzp, char *tnm, |
d4dc53da | 2696 | cred_t *cr, int flags, uint64_t rflags, vattr_t *wo_vap, zidmap_t *mnt_ns) |
34dc7c2f | 2697 | { |
657ce253 MM |
2698 | znode_t *szp, *tzp; |
2699 | zfsvfs_t *zfsvfs = ZTOZSB(sdzp); | |
34dc7c2f | 2700 | zilog_t *zilog; |
34dc7c2f BB |
2701 | zfs_dirlock_t *sdl, *tdl; |
2702 | dmu_tx_t *tx; | |
2703 | zfs_zlock_t *zl; | |
2704 | int cmp, serr, terr; | |
2705 | int error = 0; | |
2706 | int zflg = 0; | |
e8b96c60 | 2707 | boolean_t waited = B_FALSE; |
dbf6108b AS |
2708 | /* Needed for whiteout inode creation. */ |
2709 | boolean_t fuid_dirtied; | |
2710 | zfs_acl_ids_t acl_ids; | |
2711 | boolean_t have_acl = B_FALSE; | |
2712 | znode_t *wzp = NULL; | |
2713 | ||
34dc7c2f | 2714 | |
32dec7bd | 2715 | if (snm == NULL || tnm == NULL) |
2716 | return (SET_ERROR(EINVAL)); | |
2717 | ||
dbf6108b AS |
2718 | if (rflags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
2719 | return (SET_ERROR(EINVAL)); | |
2720 | ||
2721 | /* Already checked by Linux VFS, but just to make sure. */ | |
2722 | if (rflags & RENAME_EXCHANGE && | |
2723 | (rflags & (RENAME_NOREPLACE | RENAME_WHITEOUT))) | |
2724 | return (SET_ERROR(EINVAL)); | |
2725 | ||
2726 | /* | |
2727 | * Make sure we only get wo_vap iff. RENAME_WHITEOUT and that it's the | |
2728 | * right kind of vattr_t for the whiteout file. These are set | |
2729 | * internally by ZFS so should never be incorrect. | |
2730 | */ | |
2731 | VERIFY_EQUIV(rflags & RENAME_WHITEOUT, wo_vap != NULL); | |
2732 | VERIFY_IMPLY(wo_vap, wo_vap->va_mode == S_IFCHR); | |
2733 | VERIFY_IMPLY(wo_vap, wo_vap->va_rdev == makedevice(0, 0)); | |
2734 | ||
768eaced CC |
2735 | if ((error = zfs_enter_verify_zp(zfsvfs, sdzp, FTAG)) != 0) |
2736 | return (error); | |
0037b49e | 2737 | zilog = zfsvfs->z_log; |
34dc7c2f | 2738 | |
768eaced CC |
2739 | if ((error = zfs_verify_zp(tdzp)) != 0) { |
2740 | zfs_exit(zfsvfs, FTAG); | |
2741 | return (error); | |
2742 | } | |
812e91a7 MT |
2743 | |
2744 | /* | |
2745 | * We check i_sb because snapshots and the ctldir must have different | |
2746 | * super blocks. | |
2747 | */ | |
657ce253 MM |
2748 | if (ZTOI(tdzp)->i_sb != ZTOI(sdzp)->i_sb || |
2749 | zfsctl_is_node(ZTOI(tdzp))) { | |
768eaced | 2750 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 2751 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
2752 | } |
2753 | ||
0037b49e | 2754 | if (zfsvfs->z_utf8 && u8_validate(tnm, |
34dc7c2f | 2755 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 2756 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 2757 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
2758 | } |
2759 | ||
2760 | if (flags & FIGNORECASE) | |
2761 | zflg |= ZCILOOK; | |
2762 | ||
2763 | top: | |
2764 | szp = NULL; | |
2765 | tzp = NULL; | |
2766 | zl = NULL; | |
2767 | ||
2768 | /* | |
2769 | * This is to prevent the creation of links into attribute space | |
2770 | * by renaming a linked file into/outof an attribute directory. | |
2771 | * See the comment in zfs_link() for why this is considered bad. | |
2772 | */ | |
428870ff | 2773 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
768eaced | 2774 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 2775 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2776 | } |
2777 | ||
2778 | /* | |
2779 | * Lock source and target directory entries. To prevent deadlock, | |
2780 | * a lock ordering must be defined. We lock the directory with | |
2781 | * the smallest object id first, or if it's a tie, the one with | |
2782 | * the lexically first name. | |
2783 | */ | |
2784 | if (sdzp->z_id < tdzp->z_id) { | |
2785 | cmp = -1; | |
2786 | } else if (sdzp->z_id > tdzp->z_id) { | |
2787 | cmp = 1; | |
2788 | } else { | |
2789 | /* | |
2790 | * First compare the two name arguments without | |
2791 | * considering any case folding. | |
2792 | */ | |
0037b49e | 2793 | int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
2794 | |
2795 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
0037b49e | 2796 | ASSERT(error == 0 || !zfsvfs->z_utf8); |
34dc7c2f BB |
2797 | if (cmp == 0) { |
2798 | /* | |
2799 | * POSIX: "If the old argument and the new argument | |
2800 | * both refer to links to the same existing file, | |
2801 | * the rename() function shall return successfully | |
2802 | * and perform no other action." | |
2803 | */ | |
768eaced | 2804 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
2805 | return (0); |
2806 | } | |
2807 | /* | |
2808 | * If the file system is case-folding, then we may | |
2809 | * have some more checking to do. A case-folding file | |
2810 | * system is either supporting mixed case sensitivity | |
2811 | * access or is completely case-insensitive. Note | |
2812 | * that the file system is always case preserving. | |
2813 | * | |
2814 | * In mixed sensitivity mode case sensitive behavior | |
2815 | * is the default. FIGNORECASE must be used to | |
2816 | * explicitly request case insensitive behavior. | |
2817 | * | |
2818 | * If the source and target names provided differ only | |
2819 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
2820 | * we will treat this as a special case in the | |
2821 | * case-insensitive mode: as long as the source name | |
2822 | * is an exact match, we will allow this to proceed as | |
2823 | * a name-change request. | |
2824 | */ | |
0037b49e BB |
2825 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
2826 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 2827 | flags & FIGNORECASE)) && |
0037b49e | 2828 | u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
2829 | &error) == 0) { |
2830 | /* | |
2831 | * case preserving rename request, require exact | |
2832 | * name matches | |
2833 | */ | |
2834 | zflg |= ZCIEXACT; | |
2835 | zflg &= ~ZCILOOK; | |
2836 | } | |
2837 | } | |
2838 | ||
428870ff BB |
2839 | /* |
2840 | * If the source and destination directories are the same, we should | |
2841 | * grab the z_name_lock of that directory only once. | |
2842 | */ | |
2843 | if (sdzp == tdzp) { | |
2844 | zflg |= ZHAVELOCK; | |
2845 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
2846 | } | |
2847 | ||
34dc7c2f BB |
2848 | if (cmp < 0) { |
2849 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
2850 | ZEXISTS | zflg, NULL, NULL); | |
2851 | terr = zfs_dirent_lock(&tdl, | |
2852 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
2853 | } else { | |
2854 | terr = zfs_dirent_lock(&tdl, | |
2855 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
2856 | serr = zfs_dirent_lock(&sdl, | |
2857 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
2858 | NULL, NULL); | |
2859 | } | |
2860 | ||
2861 | if (serr) { | |
2862 | /* | |
2863 | * Source entry invalid or not there. | |
2864 | */ | |
2865 | if (!terr) { | |
2866 | zfs_dirent_unlock(tdl); | |
2867 | if (tzp) | |
657ce253 | 2868 | zrele(tzp); |
34dc7c2f | 2869 | } |
428870ff BB |
2870 | |
2871 | if (sdzp == tdzp) | |
2872 | rw_exit(&sdzp->z_name_lock); | |
2873 | ||
34dc7c2f BB |
2874 | if (strcmp(snm, "..") == 0) |
2875 | serr = EINVAL; | |
768eaced | 2876 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
2877 | return (serr); |
2878 | } | |
2879 | if (terr) { | |
2880 | zfs_dirent_unlock(sdl); | |
657ce253 | 2881 | zrele(szp); |
428870ff BB |
2882 | |
2883 | if (sdzp == tdzp) | |
2884 | rw_exit(&sdzp->z_name_lock); | |
2885 | ||
34dc7c2f BB |
2886 | if (strcmp(tnm, "..") == 0) |
2887 | terr = EINVAL; | |
768eaced | 2888 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
2889 | return (terr); |
2890 | } | |
2891 | ||
9c5167d1 NF |
2892 | /* |
2893 | * If we are using project inheritance, means if the directory has | |
2894 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
2895 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
2896 | * such case, we only allow renames into our tree when the project | |
2897 | * IDs are the same. | |
2898 | */ | |
2899 | if (tdzp->z_pflags & ZFS_PROJINHERIT && | |
2900 | tdzp->z_projid != szp->z_projid) { | |
2901 | error = SET_ERROR(EXDEV); | |
2902 | goto out; | |
2903 | } | |
2904 | ||
34dc7c2f BB |
2905 | /* |
2906 | * Must have write access at the source to remove the old entry | |
2907 | * and write access at the target to create the new entry. | |
2908 | * Note that if target and source are the same, this can be | |
2909 | * done in a single check. | |
2910 | */ | |
2a068a13 | 2911 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr, mnt_ns))) |
34dc7c2f BB |
2912 | goto out; |
2913 | ||
3558fd73 | 2914 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
2915 | /* |
2916 | * Check to make sure rename is valid. | |
2917 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
2918 | */ | |
149e873a | 2919 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
2920 | goto out; |
2921 | } | |
2922 | ||
2923 | /* | |
2924 | * Does target exist? | |
2925 | */ | |
2926 | if (tzp) { | |
dbf6108b AS |
2927 | if (rflags & RENAME_NOREPLACE) { |
2928 | error = SET_ERROR(EEXIST); | |
2929 | goto out; | |
2930 | } | |
34dc7c2f | 2931 | /* |
dbf6108b | 2932 | * Source and target must be the same type (unless exchanging). |
34dc7c2f | 2933 | */ |
dbf6108b AS |
2934 | if (!(rflags & RENAME_EXCHANGE)) { |
2935 | boolean_t s_is_dir = S_ISDIR(ZTOI(szp)->i_mode) != 0; | |
2936 | boolean_t t_is_dir = S_ISDIR(ZTOI(tzp)->i_mode) != 0; | |
e015d6cc | 2937 | |
dbf6108b AS |
2938 | if (s_is_dir != t_is_dir) { |
2939 | error = SET_ERROR(s_is_dir ? ENOTDIR : EISDIR); | |
2940 | goto out; | |
2941 | } | |
34dc7c2f BB |
2942 | } |
2943 | /* | |
2944 | * POSIX dictates that when the source and target | |
2945 | * entries refer to the same file object, rename | |
2946 | * must do nothing and exit without error. | |
2947 | */ | |
2948 | if (szp->z_id == tzp->z_id) { | |
2949 | error = 0; | |
2950 | goto out; | |
2951 | } | |
dbf6108b AS |
2952 | } else if (rflags & RENAME_EXCHANGE) { |
2953 | /* Target must exist for RENAME_EXCHANGE. */ | |
2954 | error = SET_ERROR(ENOENT); | |
2955 | goto out; | |
2956 | } | |
2957 | ||
2958 | /* Set up inode creation for RENAME_WHITEOUT. */ | |
2959 | if (rflags & RENAME_WHITEOUT) { | |
2960 | /* | |
2961 | * Whiteout files are not regular files or directories, so to | |
2962 | * match zfs_create() we do not inherit the project id. | |
2963 | */ | |
2964 | uint64_t wo_projid = ZFS_DEFAULT_PROJID; | |
2965 | ||
2966 | error = zfs_zaccess(sdzp, ACE_ADD_FILE, 0, B_FALSE, cr, mnt_ns); | |
2967 | if (error) | |
2968 | goto out; | |
2969 | ||
2970 | if (!have_acl) { | |
2971 | error = zfs_acl_ids_create(sdzp, 0, wo_vap, cr, NULL, | |
2972 | &acl_ids, mnt_ns); | |
2973 | if (error) | |
2974 | goto out; | |
2975 | have_acl = B_TRUE; | |
2976 | } | |
2977 | ||
2978 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, wo_projid)) { | |
2979 | error = SET_ERROR(EDQUOT); | |
2980 | goto out; | |
2981 | } | |
34dc7c2f BB |
2982 | } |
2983 | ||
0037b49e | 2984 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
2985 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
2986 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
dbf6108b AS |
2987 | dmu_tx_hold_zap(tx, sdzp->z_id, |
2988 | (rflags & RENAME_EXCHANGE) ? TRUE : FALSE, snm); | |
34dc7c2f | 2989 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); |
428870ff BB |
2990 | if (sdzp != tdzp) { |
2991 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
2992 | zfs_sa_upgrade_txholds(tx, tdzp); | |
2993 | } | |
2994 | if (tzp) { | |
2995 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
2996 | zfs_sa_upgrade_txholds(tx, tzp); | |
2997 | } | |
dbf6108b AS |
2998 | if (rflags & RENAME_WHITEOUT) { |
2999 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
3000 | ZFS_SA_BASE_ATTR_SIZE); | |
428870ff | 3001 | |
dbf6108b AS |
3002 | dmu_tx_hold_zap(tx, sdzp->z_id, TRUE, snm); |
3003 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
3004 | if (!zfsvfs->z_use_sa && | |
3005 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
3006 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
3007 | 0, acl_ids.z_aclp->z_acl_bytes); | |
3008 | } | |
3009 | } | |
3010 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
3011 | if (fuid_dirtied) | |
3012 | zfs_fuid_txhold(zfsvfs, tx); | |
428870ff | 3013 | zfs_sa_upgrade_txholds(tx, szp); |
0037b49e | 3014 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
0735ecb3 | 3015 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
3016 | if (error) { |
3017 | if (zl != NULL) | |
3018 | zfs_rename_unlock(&zl); | |
3019 | zfs_dirent_unlock(sdl); | |
3020 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3021 | |
3022 | if (sdzp == tdzp) | |
3023 | rw_exit(&sdzp->z_name_lock); | |
3024 | ||
fb5f0bc8 | 3025 | if (error == ERESTART) { |
e8b96c60 | 3026 | waited = B_TRUE; |
34dc7c2f BB |
3027 | dmu_tx_wait(tx); |
3028 | dmu_tx_abort(tx); | |
657ce253 | 3029 | zrele(szp); |
ea7e86d8 | 3030 | if (tzp) |
657ce253 | 3031 | zrele(tzp); |
34dc7c2f BB |
3032 | goto top; |
3033 | } | |
3034 | dmu_tx_abort(tx); | |
657ce253 | 3035 | zrele(szp); |
ea7e86d8 | 3036 | if (tzp) |
657ce253 | 3037 | zrele(tzp); |
768eaced | 3038 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3039 | return (error); |
3040 | } | |
3041 | ||
e015d6cc AS |
3042 | /* |
3043 | * Unlink the source. | |
3044 | */ | |
3045 | szp->z_pflags |= ZFS_AV_MODIFIED; | |
3046 | if (tdzp->z_pflags & ZFS_PROJINHERIT) | |
3047 | szp->z_pflags |= ZFS_PROJINHERIT; | |
34dc7c2f | 3048 | |
e015d6cc AS |
3049 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), |
3050 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); | |
dbf6108b | 3051 | VERIFY0(error); |
34dc7c2f | 3052 | |
e015d6cc AS |
3053 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3054 | if (error) | |
3055 | goto commit; | |
3056 | ||
3057 | /* | |
3058 | * Unlink the target. | |
3059 | */ | |
3060 | if (tzp) { | |
dbf6108b AS |
3061 | int tzflg = zflg; |
3062 | ||
3063 | if (rflags & RENAME_EXCHANGE) { | |
3064 | /* This inode will be re-linked soon. */ | |
3065 | tzflg |= ZRENAMING; | |
3066 | ||
3067 | tzp->z_pflags |= ZFS_AV_MODIFIED; | |
3068 | if (sdzp->z_pflags & ZFS_PROJINHERIT) | |
3069 | tzp->z_pflags |= ZFS_PROJINHERIT; | |
3070 | ||
3071 | error = sa_update(tzp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), | |
3072 | (void *)&tzp->z_pflags, sizeof (uint64_t), tx); | |
3073 | ASSERT0(error); | |
3074 | } | |
3075 | error = zfs_link_destroy(tdl, tzp, tx, tzflg, NULL); | |
e015d6cc AS |
3076 | if (error) |
3077 | goto commit_link_szp; | |
3078 | } | |
3079 | ||
3080 | /* | |
dbf6108b AS |
3081 | * Create the new target links: |
3082 | * * We always link the target. | |
3083 | * * RENAME_EXCHANGE: Link the old target to the source. | |
3084 | * * RENAME_WHITEOUT: Create a whiteout inode in-place of the source. | |
e015d6cc AS |
3085 | */ |
3086 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3087 | if (error) { | |
3088 | /* | |
3089 | * If we have removed the existing target, a subsequent call to | |
3090 | * zfs_link_create() to add back the same entry, but with a new | |
3091 | * dnode (szp), should not fail. | |
3092 | */ | |
3093 | ASSERT3P(tzp, ==, NULL); | |
3094 | goto commit_link_tzp; | |
34dc7c2f BB |
3095 | } |
3096 | ||
dbf6108b AS |
3097 | switch (rflags & (RENAME_EXCHANGE | RENAME_WHITEOUT)) { |
3098 | case RENAME_EXCHANGE: | |
3099 | error = zfs_link_create(sdl, tzp, tx, ZRENAMING); | |
3100 | /* | |
3101 | * The same argument as zfs_link_create() failing for | |
3102 | * szp applies here, since the source directory must | |
3103 | * have had an entry we are replacing. | |
3104 | */ | |
3105 | ASSERT0(error); | |
3106 | if (error) | |
3107 | goto commit_unlink_td_szp; | |
3108 | break; | |
3109 | case RENAME_WHITEOUT: | |
3110 | zfs_mknode(sdzp, wo_vap, tx, cr, 0, &wzp, &acl_ids); | |
3111 | error = zfs_link_create(sdl, wzp, tx, ZNEW); | |
3112 | if (error) { | |
3113 | zfs_znode_delete(wzp, tx); | |
3114 | remove_inode_hash(ZTOI(wzp)); | |
3115 | goto commit_unlink_td_szp; | |
3116 | } | |
3117 | break; | |
3118 | } | |
3119 | ||
3120 | if (fuid_dirtied) | |
3121 | zfs_fuid_sync(zfsvfs, tx); | |
3122 | ||
3123 | switch (rflags & (RENAME_EXCHANGE | RENAME_WHITEOUT)) { | |
3124 | case RENAME_EXCHANGE: | |
3125 | zfs_log_rename_exchange(zilog, tx, | |
3126 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, sdl->dl_name, | |
3127 | tdzp, tdl->dl_name, szp); | |
3128 | break; | |
3129 | case RENAME_WHITEOUT: | |
3130 | zfs_log_rename_whiteout(zilog, tx, | |
3131 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, sdl->dl_name, | |
3132 | tdzp, tdl->dl_name, szp, wzp); | |
3133 | break; | |
3134 | default: | |
3135 | ASSERT0(rflags & ~RENAME_NOREPLACE); | |
3136 | zfs_log_rename(zilog, tx, (flags & FIGNORECASE ? TX_CI : 0), | |
3137 | sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp); | |
3138 | break; | |
3139 | } | |
e015d6cc AS |
3140 | |
3141 | commit: | |
34dc7c2f BB |
3142 | dmu_tx_commit(tx); |
3143 | out: | |
dbf6108b AS |
3144 | if (have_acl) |
3145 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 3146 | |
fc273894 | 3147 | zfs_znode_update_vfs(sdzp); |
428870ff BB |
3148 | if (sdzp == tdzp) |
3149 | rw_exit(&sdzp->z_name_lock); | |
3150 | ||
960e08fe | 3151 | if (sdzp != tdzp) |
fc273894 | 3152 | zfs_znode_update_vfs(tdzp); |
428870ff | 3153 | |
fc273894 | 3154 | zfs_znode_update_vfs(szp); |
657ce253 | 3155 | zrele(szp); |
dbf6108b AS |
3156 | if (wzp) { |
3157 | zfs_znode_update_vfs(wzp); | |
3158 | zrele(wzp); | |
3159 | } | |
960e08fe | 3160 | if (tzp) { |
fc273894 | 3161 | zfs_znode_update_vfs(tzp); |
657ce253 | 3162 | zrele(tzp); |
960e08fe | 3163 | } |
34dc7c2f | 3164 | |
dbf6108b AS |
3165 | if (zl != NULL) |
3166 | zfs_rename_unlock(&zl); | |
3167 | ||
3168 | zfs_dirent_unlock(sdl); | |
3169 | zfs_dirent_unlock(tdl); | |
3170 | ||
0037b49e | 3171 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3172 | zil_commit(zilog, 0); |
428870ff | 3173 | |
768eaced | 3174 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f | 3175 | return (error); |
e015d6cc AS |
3176 | |
3177 | /* | |
3178 | * Clean-up path for broken link state. | |
3179 | * | |
3180 | * At this point we are in a (very) bad state, so we need to do our | |
dbf6108b AS |
3181 | * best to correct the state. In particular, all of the nlinks are |
3182 | * wrong because we were destroying and creating links with ZRENAMING. | |
3183 | * | |
3184 | * In some form, all of these operations have to resolve the state: | |
3185 | * | |
3186 | * * link_destroy() *must* succeed. Fortunately, this is very likely | |
3187 | * since we only just created it. | |
e015d6cc | 3188 | * |
dbf6108b AS |
3189 | * * link_create()s are allowed to fail (though they shouldn't because |
3190 | * we only just unlinked them and are putting the entries back | |
3191 | * during clean-up). But if they fail, we can just forcefully drop | |
3192 | * the nlink value to (at the very least) avoid broken nlink values | |
3193 | * -- though in the case of non-empty directories we will have to | |
3194 | * panic (otherwise we'd have a leaked directory with a broken ..). | |
e015d6cc | 3195 | */ |
dbf6108b AS |
3196 | commit_unlink_td_szp: |
3197 | VERIFY0(zfs_link_destroy(tdl, szp, tx, ZRENAMING, NULL)); | |
e015d6cc AS |
3198 | commit_link_tzp: |
3199 | if (tzp) { | |
3200 | if (zfs_link_create(tdl, tzp, tx, ZRENAMING)) | |
dbf6108b | 3201 | VERIFY0(zfs_drop_nlink(tzp, tx, NULL)); |
e015d6cc AS |
3202 | } |
3203 | commit_link_szp: | |
3204 | if (zfs_link_create(sdl, szp, tx, ZRENAMING)) | |
dbf6108b | 3205 | VERIFY0(zfs_drop_nlink(szp, tx, NULL)); |
e015d6cc | 3206 | goto commit; |
34dc7c2f BB |
3207 | } |
3208 | ||
3209 | /* | |
3210 | * Insert the indicated symbolic reference entry into the directory. | |
3211 | * | |
657ce253 | 3212 | * IN: dzp - Directory to contain new symbolic link. |
841a7a98 | 3213 | * name - Name of directory entry in dip. |
34dc7c2f | 3214 | * vap - Attributes of new entry. |
841a7a98 | 3215 | * link - Name for new symlink entry. |
34dc7c2f | 3216 | * cr - credentials of caller. |
34dc7c2f | 3217 | * flags - case flags |
2a068a13 | 3218 | * mnt_ns - user namespace of the mount |
34dc7c2f | 3219 | * |
657ce253 | 3220 | * OUT: zpp - Znode for new symbolic link. |
841a7a98 | 3221 | * |
d3cc8b15 | 3222 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3223 | * |
3224 | * Timestamps: | |
3558fd73 | 3225 | * dip - ctime|mtime updated |
34dc7c2f | 3226 | */ |
e5c39b95 | 3227 | int |
657ce253 | 3228 | zfs_symlink(znode_t *dzp, char *name, vattr_t *vap, char *link, |
d4dc53da | 3229 | znode_t **zpp, cred_t *cr, int flags, zidmap_t *mnt_ns) |
34dc7c2f | 3230 | { |
657ce253 | 3231 | znode_t *zp; |
34dc7c2f BB |
3232 | zfs_dirlock_t *dl; |
3233 | dmu_tx_t *tx; | |
657ce253 | 3234 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f | 3235 | zilog_t *zilog; |
428870ff | 3236 | uint64_t len = strlen(link); |
34dc7c2f BB |
3237 | int error; |
3238 | int zflg = ZNEW; | |
9babb374 BB |
3239 | zfs_acl_ids_t acl_ids; |
3240 | boolean_t fuid_dirtied; | |
428870ff | 3241 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 3242 | boolean_t waited = B_FALSE; |
34dc7c2f | 3243 | |
3558fd73 | 3244 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 3245 | |
32dec7bd | 3246 | if (name == NULL) |
3247 | return (SET_ERROR(EINVAL)); | |
3248 | ||
768eaced CC |
3249 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
3250 | return (error); | |
0037b49e | 3251 | zilog = zfsvfs->z_log; |
34dc7c2f | 3252 | |
0037b49e | 3253 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 3254 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 3255 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3256 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3257 | } |
3258 | if (flags & FIGNORECASE) | |
3259 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3260 | |
3261 | if (len > MAXPATHLEN) { | |
768eaced | 3262 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3263 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
3264 | } |
3265 | ||
428870ff | 3266 | if ((error = zfs_acl_ids_create(dzp, 0, |
2a068a13 | 3267 | vap, cr, NULL, &acl_ids, mnt_ns)) != 0) { |
768eaced | 3268 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
3269 | return (error); |
3270 | } | |
3271 | top: | |
657ce253 | 3272 | *zpp = NULL; |
3558fd73 | 3273 | |
34dc7c2f BB |
3274 | /* |
3275 | * Attempt to lock directory; fail if entry already exists. | |
3276 | */ | |
3277 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3278 | if (error) { | |
428870ff | 3279 | zfs_acl_ids_free(&acl_ids); |
768eaced | 3280 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
3281 | return (error); |
3282 | } | |
3283 | ||
2a068a13 | 3284 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr, mnt_ns))) { |
428870ff BB |
3285 | zfs_acl_ids_free(&acl_ids); |
3286 | zfs_dirent_unlock(dl); | |
768eaced | 3287 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3288 | return (error); |
3289 | } | |
3290 | ||
9c5167d1 | 3291 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, ZFS_DEFAULT_PROJID)) { |
9babb374 BB |
3292 | zfs_acl_ids_free(&acl_ids); |
3293 | zfs_dirent_unlock(dl); | |
768eaced | 3294 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3295 | return (SET_ERROR(EDQUOT)); |
9babb374 | 3296 | } |
0037b49e BB |
3297 | tx = dmu_tx_create(zfsvfs->z_os); |
3298 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
34dc7c2f | 3299 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3300 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3301 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3302 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3303 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
0037b49e | 3304 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
3305 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
3306 | acl_ids.z_aclp->z_acl_bytes); | |
3307 | } | |
9babb374 | 3308 | if (fuid_dirtied) |
0037b49e | 3309 | zfs_fuid_txhold(zfsvfs, tx); |
0735ecb3 | 3310 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
3311 | if (error) { |
3312 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3313 | if (error == ERESTART) { |
e8b96c60 | 3314 | waited = B_TRUE; |
34dc7c2f BB |
3315 | dmu_tx_wait(tx); |
3316 | dmu_tx_abort(tx); | |
3317 | goto top; | |
3318 | } | |
428870ff | 3319 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 3320 | dmu_tx_abort(tx); |
768eaced | 3321 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3322 | return (error); |
3323 | } | |
3324 | ||
34dc7c2f BB |
3325 | /* |
3326 | * Create a new object for the symlink. | |
bf169e9f | 3327 | * for version 4 ZPL datasets the symlink will be an SA attribute |
34dc7c2f | 3328 | */ |
428870ff | 3329 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3330 | |
428870ff | 3331 | if (fuid_dirtied) |
0037b49e | 3332 | zfs_fuid_sync(zfsvfs, tx); |
34dc7c2f | 3333 | |
572e2857 | 3334 | mutex_enter(&zp->z_lock); |
428870ff | 3335 | if (zp->z_is_sa) |
0037b49e | 3336 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), |
428870ff BB |
3337 | link, len, tx); |
3338 | else | |
3339 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3340 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3341 | |
428870ff | 3342 | zp->z_size = len; |
0037b49e | 3343 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 3344 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
3345 | /* |
3346 | * Insert the new object into the directory. | |
3347 | */ | |
599b8648 CC |
3348 | error = zfs_link_create(dl, zp, tx, ZNEW); |
3349 | if (error != 0) { | |
3350 | zfs_znode_delete(zp, tx); | |
3351 | remove_inode_hash(ZTOI(zp)); | |
3352 | } else { | |
3353 | if (flags & FIGNORECASE) | |
3354 | txtype |= TX_CI; | |
3355 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
4f301661 | 3356 | |
fc273894 KHN |
3357 | zfs_znode_update_vfs(dzp); |
3358 | zfs_znode_update_vfs(zp); | |
599b8648 | 3359 | } |
960e08fe | 3360 | |
9babb374 | 3361 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3362 | |
3363 | dmu_tx_commit(tx); | |
3364 | ||
3365 | zfs_dirent_unlock(dl); | |
3366 | ||
599b8648 | 3367 | if (error == 0) { |
657ce253 | 3368 | *zpp = zp; |
34dc7c2f | 3369 | |
599b8648 CC |
3370 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
3371 | zil_commit(zilog, 0); | |
3372 | } else { | |
657ce253 | 3373 | zrele(zp); |
599b8648 | 3374 | } |
428870ff | 3375 | |
768eaced | 3376 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3377 | return (error); |
3378 | } | |
3379 | ||
3380 | /* | |
3381 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 3382 | * the symbolic path referred to by ip. |
34dc7c2f | 3383 | * |
8b4f9a2d BB |
3384 | * IN: ip - inode of symbolic link |
3385 | * uio - structure to contain the link path. | |
3386 | * cr - credentials of caller. | |
34dc7c2f BB |
3387 | * |
3388 | * RETURN: 0 if success | |
3389 | * error code if failure | |
3390 | * | |
3391 | * Timestamps: | |
3558fd73 | 3392 | * ip - atime updated |
34dc7c2f | 3393 | */ |
e5c39b95 | 3394 | int |
d0cd9a5c | 3395 | zfs_readlink(struct inode *ip, zfs_uio_t *uio, cred_t *cr) |
34dc7c2f | 3396 | { |
ef70eff1 | 3397 | (void) cr; |
3558fd73 | 3398 | znode_t *zp = ITOZ(ip); |
0037b49e | 3399 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
3400 | int error; |
3401 | ||
768eaced CC |
3402 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
3403 | return (error); | |
34dc7c2f | 3404 | |
572e2857 | 3405 | mutex_enter(&zp->z_lock); |
428870ff | 3406 | if (zp->z_is_sa) |
8b4f9a2d | 3407 | error = sa_lookup_uio(zp->z_sa_hdl, |
0037b49e | 3408 | SA_ZPL_SYMLINK(zfsvfs), uio); |
428870ff | 3409 | else |
8b4f9a2d | 3410 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 3411 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3412 | |
768eaced | 3413 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3414 | return (error); |
3415 | } | |
3416 | ||
3417 | /* | |
657ce253 | 3418 | * Insert a new entry into directory tdzp referencing szp. |
34dc7c2f | 3419 | * |
657ce253 MM |
3420 | * IN: tdzp - Directory to contain new entry. |
3421 | * szp - znode of new entry. | |
34dc7c2f BB |
3422 | * name - name of new entry. |
3423 | * cr - credentials of caller. | |
841a7a98 | 3424 | * flags - case flags. |
34dc7c2f BB |
3425 | * |
3426 | * RETURN: 0 if success | |
3427 | * error code if failure | |
3428 | * | |
3429 | * Timestamps: | |
657ce253 MM |
3430 | * tdzp - ctime|mtime updated |
3431 | * szp - ctime updated | |
34dc7c2f | 3432 | */ |
e5c39b95 | 3433 | int |
657ce253 | 3434 | zfs_link(znode_t *tdzp, znode_t *szp, char *name, cred_t *cr, |
da5e151f | 3435 | int flags) |
34dc7c2f | 3436 | { |
657ce253 MM |
3437 | struct inode *sip = ZTOI(szp); |
3438 | znode_t *tzp; | |
3439 | zfsvfs_t *zfsvfs = ZTOZSB(tdzp); | |
34dc7c2f BB |
3440 | zilog_t *zilog; |
3441 | zfs_dirlock_t *dl; | |
3442 | dmu_tx_t *tx; | |
34dc7c2f BB |
3443 | int error; |
3444 | int zf = ZNEW; | |
428870ff | 3445 | uint64_t parent; |
572e2857 | 3446 | uid_t owner; |
e8b96c60 | 3447 | boolean_t waited = B_FALSE; |
ace1eae8 CC |
3448 | boolean_t is_tmpfile = 0; |
3449 | uint64_t txg; | |
3450 | #ifdef HAVE_TMPFILE | |
3451 | is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); | |
3452 | #endif | |
657ce253 | 3453 | ASSERT(S_ISDIR(ZTOI(tdzp)->i_mode)); |
34dc7c2f | 3454 | |
32dec7bd | 3455 | if (name == NULL) |
3456 | return (SET_ERROR(EINVAL)); | |
3457 | ||
768eaced CC |
3458 | if ((error = zfs_enter_verify_zp(zfsvfs, tdzp, FTAG)) != 0) |
3459 | return (error); | |
0037b49e | 3460 | zilog = zfsvfs->z_log; |
34dc7c2f | 3461 | |
428870ff BB |
3462 | /* |
3463 | * POSIX dictates that we return EPERM here. | |
3464 | * Better choices include ENOTSUP or EISDIR. | |
3465 | */ | |
3558fd73 | 3466 | if (S_ISDIR(sip->i_mode)) { |
768eaced | 3467 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3468 | return (SET_ERROR(EPERM)); |
428870ff BB |
3469 | } |
3470 | ||
768eaced CC |
3471 | if ((error = zfs_verify_zp(szp)) != 0) { |
3472 | zfs_exit(zfsvfs, FTAG); | |
3473 | return (error); | |
3474 | } | |
812e91a7 | 3475 | |
9c5167d1 NF |
3476 | /* |
3477 | * If we are using project inheritance, means if the directory has | |
3478 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
3479 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
3480 | * such case, we only allow hard link creation in our tree when the | |
3481 | * project IDs are the same. | |
3482 | */ | |
657ce253 MM |
3483 | if (tdzp->z_pflags & ZFS_PROJINHERIT && |
3484 | tdzp->z_projid != szp->z_projid) { | |
768eaced | 3485 | zfs_exit(zfsvfs, FTAG); |
9c5167d1 NF |
3486 | return (SET_ERROR(EXDEV)); |
3487 | } | |
3488 | ||
812e91a7 MT |
3489 | /* |
3490 | * We check i_sb because snapshots and the ctldir must have different | |
3491 | * super blocks. | |
3492 | */ | |
657ce253 | 3493 | if (sip->i_sb != ZTOI(tdzp)->i_sb || zfsctl_is_node(sip)) { |
768eaced | 3494 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3495 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 3496 | } |
428870ff | 3497 | |
428870ff BB |
3498 | /* Prevent links to .zfs/shares files */ |
3499 | ||
0037b49e | 3500 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff | 3501 | &parent, sizeof (uint64_t))) != 0) { |
768eaced | 3502 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
3503 | return (error); |
3504 | } | |
0037b49e | 3505 | if (parent == zfsvfs->z_shares_dir) { |
768eaced | 3506 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3507 | return (SET_ERROR(EPERM)); |
428870ff BB |
3508 | } |
3509 | ||
0037b49e | 3510 | if (zfsvfs->z_utf8 && u8_validate(name, |
34dc7c2f | 3511 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 3512 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3513 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3514 | } |
3515 | if (flags & FIGNORECASE) | |
3516 | zf |= ZCILOOK; | |
3517 | ||
34dc7c2f BB |
3518 | /* |
3519 | * We do not support links between attributes and non-attributes | |
3520 | * because of the potential security risk of creating links | |
3521 | * into "normal" file space in order to circumvent restrictions | |
3522 | * imposed in attribute space. | |
3523 | */ | |
657ce253 | 3524 | if ((szp->z_pflags & ZFS_XATTR) != (tdzp->z_pflags & ZFS_XATTR)) { |
768eaced | 3525 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3526 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3527 | } |
3528 | ||
0037b49e BB |
3529 | owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid), |
3530 | cr, ZFS_OWNER); | |
572e2857 | 3531 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
768eaced | 3532 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3533 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
3534 | } |
3535 | ||
f224eddf | 3536 | if ((error = zfs_zaccess(tdzp, ACE_ADD_FILE, 0, B_FALSE, cr, |
d4dc53da | 3537 | zfs_init_idmap))) { |
768eaced | 3538 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3539 | return (error); |
3540 | } | |
3541 | ||
428870ff | 3542 | top: |
34dc7c2f BB |
3543 | /* |
3544 | * Attempt to lock directory; fail if entry already exists. | |
3545 | */ | |
657ce253 | 3546 | error = zfs_dirent_lock(&dl, tdzp, name, &tzp, zf, NULL, NULL); |
34dc7c2f | 3547 | if (error) { |
768eaced | 3548 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3549 | return (error); |
3550 | } | |
3551 | ||
0037b49e | 3552 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 3553 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
657ce253 | 3554 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, name); |
ace1eae8 | 3555 | if (is_tmpfile) |
0037b49e | 3556 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 3557 | |
428870ff | 3558 | zfs_sa_upgrade_txholds(tx, szp); |
657ce253 | 3559 | zfs_sa_upgrade_txholds(tx, tdzp); |
0735ecb3 | 3560 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
3561 | if (error) { |
3562 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3563 | if (error == ERESTART) { |
e8b96c60 | 3564 | waited = B_TRUE; |
34dc7c2f BB |
3565 | dmu_tx_wait(tx); |
3566 | dmu_tx_abort(tx); | |
3567 | goto top; | |
3568 | } | |
3569 | dmu_tx_abort(tx); | |
768eaced | 3570 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3571 | return (error); |
3572 | } | |
ace1eae8 CC |
3573 | /* unmark z_unlinked so zfs_link_create will not reject */ |
3574 | if (is_tmpfile) | |
a43570c5 | 3575 | szp->z_unlinked = B_FALSE; |
34dc7c2f BB |
3576 | error = zfs_link_create(dl, szp, tx, 0); |
3577 | ||
3578 | if (error == 0) { | |
3579 | uint64_t txtype = TX_LINK; | |
ace1eae8 CC |
3580 | /* |
3581 | * tmpfile is created to be in z_unlinkedobj, so remove it. | |
e1cfd73f | 3582 | * Also, we don't log in ZIL, because all previous file |
ace1eae8 CC |
3583 | * operation on the tmpfile are ignored by ZIL. Instead we |
3584 | * always wait for txg to sync to make sure all previous | |
3585 | * operation are sync safe. | |
3586 | */ | |
3587 | if (is_tmpfile) { | |
0037b49e BB |
3588 | VERIFY(zap_remove_int(zfsvfs->z_os, |
3589 | zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); | |
ace1eae8 CC |
3590 | } else { |
3591 | if (flags & FIGNORECASE) | |
3592 | txtype |= TX_CI; | |
657ce253 | 3593 | zfs_log_link(zilog, tx, txtype, tdzp, szp, name); |
ace1eae8 CC |
3594 | } |
3595 | } else if (is_tmpfile) { | |
3596 | /* restore z_unlinked since when linking failed */ | |
a43570c5 | 3597 | szp->z_unlinked = B_TRUE; |
34dc7c2f | 3598 | } |
ace1eae8 | 3599 | txg = dmu_tx_get_txg(tx); |
34dc7c2f BB |
3600 | dmu_tx_commit(tx); |
3601 | ||
3602 | zfs_dirent_unlock(dl); | |
3603 | ||
0037b49e | 3604 | if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3605 | zil_commit(zilog, 0); |
428870ff | 3606 | |
d09dc598 | 3607 | if (is_tmpfile && zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) |
0037b49e | 3608 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg); |
ace1eae8 | 3609 | |
fc273894 KHN |
3610 | zfs_znode_update_vfs(tdzp); |
3611 | zfs_znode_update_vfs(szp); | |
768eaced | 3612 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3613 | return (error); |
3614 | } | |
3615 | ||
3c0e5c0f | 3616 | static void |
411f4a01 | 3617 | zfs_putpage_sync_commit_cb(void *arg) |
3c0e5c0f BB |
3618 | { |
3619 | struct page *pp = arg; | |
3620 | ||
119a394a | 3621 | ClearPageError(pp); |
3c0e5c0f BB |
3622 | end_page_writeback(pp); |
3623 | } | |
3624 | ||
411f4a01 SN |
3625 | static void |
3626 | zfs_putpage_async_commit_cb(void *arg) | |
3627 | { | |
3628 | struct page *pp = arg; | |
3629 | znode_t *zp = ITOZ(pp->mapping->host); | |
3630 | ||
3631 | ClearPageError(pp); | |
3632 | end_page_writeback(pp); | |
3633 | atomic_dec_32(&zp->z_async_writes_cnt); | |
3634 | } | |
3635 | ||
34dc7c2f | 3636 | /* |
3c0e5c0f BB |
3637 | * Push a page out to disk, once the page is on stable storage the |
3638 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 3639 | * |
411f4a01 SN |
3640 | * IN: ip - page mapped for inode. |
3641 | * pp - page to push (page is locked) | |
3642 | * wbc - writeback control data | |
3643 | * for_sync - does the caller intend to wait synchronously for the | |
3644 | * page writeback to complete? | |
34dc7c2f BB |
3645 | * |
3646 | * RETURN: 0 if success | |
3647 | * error code if failure | |
3648 | * | |
3c0e5c0f BB |
3649 | * Timestamps: |
3650 | * ip - ctime|mtime updated | |
34dc7c2f | 3651 | */ |
3c0e5c0f | 3652 | int |
411f4a01 SN |
3653 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc, |
3654 | boolean_t for_sync) | |
34dc7c2f | 3655 | { |
3c0e5c0f | 3656 | znode_t *zp = ITOZ(ip); |
0037b49e | 3657 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
3c0e5c0f BB |
3658 | loff_t offset; |
3659 | loff_t pgoff; | |
4c837f0d | 3660 | unsigned int pglen; |
3c0e5c0f BB |
3661 | dmu_tx_t *tx; |
3662 | caddr_t va; | |
3663 | int err = 0; | |
3664 | uint64_t mtime[2], ctime[2]; | |
db4fc559 | 3665 | inode_timespec_t tmp_ts; |
3c0e5c0f BB |
3666 | sa_bulk_attr_t bulk[3]; |
3667 | int cnt = 0; | |
21a96fb6 | 3668 | struct address_space *mapping; |
3c0e5c0f | 3669 | |
768eaced CC |
3670 | if ((err = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
3671 | return (err); | |
d164b209 | 3672 | |
3c0e5c0f BB |
3673 | ASSERT(PageLocked(pp)); |
3674 | ||
d1d7e268 MK |
3675 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
3676 | offset = i_size_read(ip); /* File length in bytes */ | |
8b1899d3 BB |
3677 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
3678 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
3679 | |
3680 | /* Page is beyond end of file */ | |
3681 | if (pgoff >= offset) { | |
3682 | unlock_page(pp); | |
768eaced | 3683 | zfs_exit(zfsvfs, FTAG); |
3c0e5c0f BB |
3684 | return (0); |
3685 | } | |
3686 | ||
3687 | /* Truncate page length to end of file */ | |
3688 | if (pgoff + pglen > offset) | |
3689 | pglen = offset - pgoff; | |
3690 | ||
3691 | #if 0 | |
34dc7c2f | 3692 | /* |
3c0e5c0f BB |
3693 | * FIXME: Allow mmap writes past its quota. The correct fix |
3694 | * is to register a page_mkwrite() handler to count the page | |
3695 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 3696 | */ |
9c5167d1 NF |
3697 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
3698 | KUID_TO_SUID(ip->i_uid)) || | |
3699 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
3700 | KGID_TO_SGID(ip->i_gid)) || | |
3701 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
3702 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
3703 | zp->z_projid))) { | |
9babb374 | 3704 | err = EDQUOT; |
9babb374 | 3705 | } |
3c0e5c0f BB |
3706 | #endif |
3707 | ||
d958324f BB |
3708 | /* |
3709 | * The ordering here is critical and must adhere to the following | |
3710 | * rules in order to avoid deadlocking in either zfs_read() or | |
3711 | * zfs_free_range() due to a lock inversion. | |
3712 | * | |
3713 | * 1) The page must be unlocked prior to acquiring the range lock. | |
3714 | * This is critical because zfs_read() calls find_lock_page() | |
3715 | * which may block on the page lock while holding the range lock. | |
3716 | * | |
3717 | * 2) Before setting or clearing write back on a page the range lock | |
3718 | * must be held in order to prevent a lock inversion with the | |
3719 | * zfs_free_range() function. | |
21a96fb6 CC |
3720 | * |
3721 | * This presents a problem because upon entering this function the | |
3722 | * page lock is already held. To safely acquire the range lock the | |
3723 | * page lock must be dropped. This creates a window where another | |
3724 | * process could truncate, invalidate, dirty, or write out the page. | |
3725 | * | |
3726 | * Therefore, after successfully reacquiring the range and page locks | |
3727 | * the current page state is checked. In the common case everything | |
3728 | * will be as is expected and it can be written out. However, if | |
3729 | * the page state has changed it must be handled accordingly. | |
d958324f | 3730 | */ |
21a96fb6 CC |
3731 | mapping = pp->mapping; |
3732 | redirty_page_for_writepage(wbc, pp); | |
d958324f | 3733 | unlock_page(pp); |
21a96fb6 | 3734 | |
bd4dde8e | 3735 | zfs_locked_range_t *lr = zfs_rangelock_enter(&zp->z_rangelock, |
5d43cc9a | 3736 | pgoff, pglen, RL_WRITER); |
21a96fb6 CC |
3737 | lock_page(pp); |
3738 | ||
3739 | /* Page mapping changed or it was no longer dirty, we're done */ | |
3740 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
3741 | unlock_page(pp); | |
2cc479d0 | 3742 | zfs_rangelock_exit(lr); |
768eaced | 3743 | zfs_exit(zfsvfs, FTAG); |
21a96fb6 CC |
3744 | return (0); |
3745 | } | |
3746 | ||
3747 | /* Another process started write block if required */ | |
3748 | if (PageWriteback(pp)) { | |
3749 | unlock_page(pp); | |
2cc479d0 | 3750 | zfs_rangelock_exit(lr); |
21a96fb6 | 3751 | |
bfd5a709 | 3752 | if (wbc->sync_mode != WB_SYNC_NONE) { |
411f4a01 SN |
3753 | /* |
3754 | * Speed up any non-sync page writebacks since | |
3755 | * they may take several seconds to complete. | |
3756 | * Refer to the comment in zpl_fsync() (when | |
3757 | * HAVE_FSYNC_RANGE is defined) for details. | |
3758 | */ | |
3759 | if (atomic_load_32(&zp->z_async_writes_cnt) > 0) { | |
3760 | zil_commit(zfsvfs->z_log, zp->z_id); | |
3761 | } | |
3762 | ||
bfd5a709 | 3763 | if (PageWriteback(pp)) |
c0fb44c5 CK |
3764 | #ifdef HAVE_PAGEMAP_FOLIO_WAIT_BIT |
3765 | folio_wait_bit(page_folio(pp), PG_writeback); | |
3766 | #else | |
bfd5a709 | 3767 | wait_on_page_bit(pp, PG_writeback); |
c0fb44c5 | 3768 | #endif |
bfd5a709 | 3769 | } |
21a96fb6 | 3770 | |
768eaced | 3771 | zfs_exit(zfsvfs, FTAG); |
21a96fb6 CC |
3772 | return (0); |
3773 | } | |
3774 | ||
3775 | /* Clear the dirty flag the required locks are held */ | |
3776 | if (!clear_page_dirty_for_io(pp)) { | |
3777 | unlock_page(pp); | |
2cc479d0 | 3778 | zfs_rangelock_exit(lr); |
768eaced | 3779 | zfs_exit(zfsvfs, FTAG); |
21a96fb6 CC |
3780 | return (0); |
3781 | } | |
3782 | ||
3783 | /* | |
3784 | * Counterpart for redirty_page_for_writepage() above. This page | |
3785 | * was in fact not skipped and should not be counted as if it were. | |
3786 | */ | |
3787 | wbc->pages_skipped--; | |
411f4a01 SN |
3788 | if (!for_sync) |
3789 | atomic_inc_32(&zp->z_async_writes_cnt); | |
3c0e5c0f | 3790 | set_page_writeback(pp); |
21a96fb6 | 3791 | unlock_page(pp); |
3c0e5c0f | 3792 | |
0037b49e | 3793 | tx = dmu_tx_create(zfsvfs->z_os); |
3c0e5c0f | 3794 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
3795 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3796 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 3797 | |
102b468b | 3798 | err = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 3799 | if (err != 0) { |
34dc7c2f | 3800 | dmu_tx_abort(tx); |
7dde17e8 SP |
3801 | #ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO |
3802 | filemap_dirty_folio(page_mapping(pp), page_folio(pp)); | |
3803 | #else | |
119a394a | 3804 | __set_page_dirty_nobuffers(pp); |
7dde17e8 | 3805 | #endif |
119a394a ED |
3806 | ClearPageError(pp); |
3807 | end_page_writeback(pp); | |
411f4a01 SN |
3808 | if (!for_sync) |
3809 | atomic_dec_32(&zp->z_async_writes_cnt); | |
2cc479d0 | 3810 | zfs_rangelock_exit(lr); |
768eaced | 3811 | zfs_exit(zfsvfs, FTAG); |
3c0e5c0f | 3812 | return (err); |
34dc7c2f BB |
3813 | } |
3814 | ||
dde471ef | 3815 | va = kmap(pp); |
8b1899d3 | 3816 | ASSERT3U(pglen, <=, PAGE_SIZE); |
0037b49e | 3817 | dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 3818 | kunmap(pp); |
34dc7c2f | 3819 | |
0037b49e BB |
3820 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
3821 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
3822 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, | |
3823 | &zp->z_pflags, 8); | |
428870ff | 3824 | |
d3aa3ea9 | 3825 | /* Preserve the mtime and ctime provided by the inode */ |
db4fc559 RN |
3826 | tmp_ts = zpl_inode_get_mtime(ip); |
3827 | ZFS_TIME_ENCODE(&tmp_ts, mtime); | |
3828 | tmp_ts = zpl_inode_get_ctime(ip); | |
3829 | ZFS_TIME_ENCODE(&tmp_ts, ctime); | |
a43570c5 | 3830 | zp->z_atime_dirty = B_FALSE; |
d3aa3ea9 BB |
3831 | zp->z_seq++; |
3832 | ||
3833 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
3834 | ||
c3773de1 | 3835 | boolean_t commit = B_FALSE; |
119a394a ED |
3836 | if (wbc->sync_mode != WB_SYNC_NONE) { |
3837 | /* | |
3838 | * Note that this is rarely called under writepages(), because | |
3839 | * writepages() normally handles the entire commit for | |
3840 | * performance reasons. | |
3841 | */ | |
c3773de1 | 3842 | commit = B_TRUE; |
411f4a01 SN |
3843 | } else if (!for_sync && atomic_load_32(&zp->z_sync_writes_cnt) > 0) { |
3844 | /* | |
3845 | * If the caller does not intend to wait synchronously | |
3846 | * for this page writeback to complete and there are active | |
3847 | * synchronous calls on this file, do a commit so that | |
3848 | * the latter don't accidentally end up waiting for | |
3849 | * our writeback to complete. Refer to the comment in | |
3850 | * zpl_fsync() (when HAVE_FSYNC_RANGE is defined) for details. | |
3851 | */ | |
c3773de1 | 3852 | commit = B_TRUE; |
2b286136 | 3853 | } |
3c0e5c0f | 3854 | |
c3773de1 AM |
3855 | zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, commit, |
3856 | for_sync ? zfs_putpage_sync_commit_cb : | |
3857 | zfs_putpage_async_commit_cb, pp); | |
3858 | ||
3859 | dmu_tx_commit(tx); | |
3860 | ||
3861 | zfs_rangelock_exit(lr); | |
3862 | ||
3863 | if (commit) | |
3864 | zil_commit(zfsvfs->z_log, zp->z_id); | |
3865 | ||
3819aaaf MB |
3866 | dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, pglen); |
3867 | ||
768eaced | 3868 | zfs_exit(zfsvfs, FTAG); |
3c0e5c0f | 3869 | return (err); |
34dc7c2f BB |
3870 | } |
3871 | ||
8780c539 BB |
3872 | /* |
3873 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 3874 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
3875 | */ |
3876 | int | |
3877 | zfs_dirty_inode(struct inode *ip, int flags) | |
3878 | { | |
3879 | znode_t *zp = ITOZ(ip); | |
0037b49e | 3880 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
8780c539 | 3881 | dmu_tx_t *tx; |
023699cd | 3882 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
db4fc559 | 3883 | inode_timespec_t tmp_ts; |
023699cd | 3884 | sa_bulk_attr_t bulk[4]; |
704cd075 | 3885 | int error = 0; |
8780c539 BB |
3886 | int cnt = 0; |
3887 | ||
0037b49e | 3888 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
c944be5d BB |
3889 | return (0); |
3890 | ||
768eaced CC |
3891 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
3892 | return (error); | |
8780c539 | 3893 | |
704cd075 CC |
3894 | #ifdef I_DIRTY_TIME |
3895 | /* | |
e1cfd73f | 3896 | * This is the lazytime semantic introduced in Linux 4.0 |
704cd075 CC |
3897 | * This flag will only be called from update_time when lazytime is set. |
3898 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
3899 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
3900 | * only need to dirty atime. | |
3901 | */ | |
3902 | if (flags == I_DIRTY_TIME) { | |
a43570c5 | 3903 | zp->z_atime_dirty = B_TRUE; |
704cd075 CC |
3904 | goto out; |
3905 | } | |
3906 | #endif | |
3907 | ||
0037b49e | 3908 | tx = dmu_tx_create(zfsvfs->z_os); |
8780c539 BB |
3909 | |
3910 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
3911 | zfs_sa_upgrade_txholds(tx, zp); | |
3912 | ||
3913 | error = dmu_tx_assign(tx, TXG_WAIT); | |
3914 | if (error) { | |
3915 | dmu_tx_abort(tx); | |
3916 | goto out; | |
3917 | } | |
3918 | ||
3919 | mutex_enter(&zp->z_lock); | |
a43570c5 | 3920 | zp->z_atime_dirty = B_FALSE; |
704cd075 | 3921 | |
0037b49e BB |
3922 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
3923 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
3924 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
3925 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
8780c539 | 3926 | |
023699cd | 3927 | /* Preserve the mode, mtime and ctime provided by the inode */ |
db4fc559 RN |
3928 | tmp_ts = zpl_inode_get_atime(ip); |
3929 | ZFS_TIME_ENCODE(&tmp_ts, atime); | |
3930 | tmp_ts = zpl_inode_get_mtime(ip); | |
3931 | ZFS_TIME_ENCODE(&tmp_ts, mtime); | |
3932 | tmp_ts = zpl_inode_get_ctime(ip); | |
3933 | ZFS_TIME_ENCODE(&tmp_ts, ctime); | |
023699cd MM |
3934 | mode = ip->i_mode; |
3935 | ||
3936 | zp->z_mode = mode; | |
8780c539 BB |
3937 | |
3938 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
3939 | mutex_exit(&zp->z_lock); | |
3940 | ||
3941 | dmu_tx_commit(tx); | |
3942 | out: | |
768eaced | 3943 | zfs_exit(zfsvfs, FTAG); |
8780c539 BB |
3944 | return (error); |
3945 | } | |
8780c539 | 3946 | |
34dc7c2f | 3947 | void |
c0d35759 | 3948 | zfs_inactive(struct inode *ip) |
34dc7c2f | 3949 | { |
c0d35759 | 3950 | znode_t *zp = ITOZ(ip); |
0037b49e | 3951 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
0df9673f | 3952 | uint64_t atime[2]; |
34dc7c2f | 3953 | int error; |
cafbd2ac | 3954 | int need_unlock = 0; |
34dc7c2f | 3955 | |
cafbd2ac | 3956 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
0037b49e | 3957 | if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) { |
cafbd2ac | 3958 | need_unlock = 1; |
0037b49e | 3959 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); |
cafbd2ac | 3960 | } |
c0d35759 | 3961 | if (zp->z_sa_hdl == NULL) { |
cafbd2ac | 3962 | if (need_unlock) |
0037b49e | 3963 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
c0d35759 | 3964 | return; |
34dc7c2f BB |
3965 | } |
3966 | ||
a43570c5 | 3967 | if (zp->z_atime_dirty && zp->z_unlinked == B_FALSE) { |
0037b49e | 3968 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 3969 | |
428870ff BB |
3970 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3971 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
3972 | error = dmu_tx_assign(tx, TXG_WAIT); |
3973 | if (error) { | |
3974 | dmu_tx_abort(tx); | |
3975 | } else { | |
db4fc559 RN |
3976 | inode_timespec_t tmp_atime; |
3977 | tmp_atime = zpl_inode_get_atime(ip); | |
3978 | ZFS_TIME_ENCODE(&tmp_atime, atime); | |
34dc7c2f | 3979 | mutex_enter(&zp->z_lock); |
0037b49e | 3980 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), |
0df9673f | 3981 | (void *)&atime, sizeof (atime), tx); |
a43570c5 | 3982 | zp->z_atime_dirty = B_FALSE; |
34dc7c2f BB |
3983 | mutex_exit(&zp->z_lock); |
3984 | dmu_tx_commit(tx); | |
3985 | } | |
3986 | } | |
3987 | ||
3988 | zfs_zinactive(zp); | |
cafbd2ac | 3989 | if (need_unlock) |
0037b49e | 3990 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
34dc7c2f BB |
3991 | } |
3992 | ||
34dc7c2f | 3993 | /* |
dde471ef | 3994 | * Fill pages with data from the disk. |
34dc7c2f BB |
3995 | */ |
3996 | static int | |
89cd2197 | 3997 | zfs_fillpage(struct inode *ip, struct page *pp) |
34dc7c2f | 3998 | { |
0037b49e | 3999 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
89cd2197 BB |
4000 | loff_t i_size = i_size_read(ip); |
4001 | u_offset_t io_off = page_offset(pp); | |
4002 | size_t io_len = PAGE_SIZE; | |
dde471ef | 4003 | |
dd108f5d RY |
4004 | ASSERT3U(io_off, <, i_size); |
4005 | ||
dde471ef PJ |
4006 | if (io_off + io_len > i_size) |
4007 | io_len = i_size - io_off; | |
34dc7c2f | 4008 | |
89cd2197 BB |
4009 | void *va = kmap(pp); |
4010 | int error = dmu_read(zfsvfs->z_os, ITOZ(ip)->z_id, io_off, | |
dd108f5d RY |
4011 | io_len, va, DMU_READ_PREFETCH); |
4012 | if (io_len != PAGE_SIZE) | |
4013 | memset((char *)va + io_len, 0, PAGE_SIZE - io_len); | |
89cd2197 | 4014 | kunmap(pp); |
d164b209 | 4015 | |
89cd2197 BB |
4016 | if (error) { |
4017 | /* convert checksum errors into IO errors */ | |
4018 | if (error == ECKSUM) | |
4019 | error = SET_ERROR(EIO); | |
4020 | ||
4021 | SetPageError(pp); | |
4022 | ClearPageUptodate(pp); | |
4023 | } else { | |
4024 | ClearPageError(pp); | |
4025 | SetPageUptodate(pp); | |
34dc7c2f | 4026 | } |
d164b209 | 4027 | |
89cd2197 | 4028 | return (error); |
34dc7c2f BB |
4029 | } |
4030 | ||
4031 | /* | |
89cd2197 | 4032 | * Uses zfs_fillpage to read data from the file and fill the page. |
34dc7c2f | 4033 | * |
dde471ef | 4034 | * IN: ip - inode of file to get data from. |
89cd2197 | 4035 | * pp - page to read |
34dc7c2f | 4036 | * |
d3cc8b15 | 4037 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4038 | * |
4039 | * Timestamps: | |
4040 | * vp - atime updated | |
4041 | */ | |
dde471ef | 4042 | int |
89cd2197 | 4043 | zfs_getpage(struct inode *ip, struct page *pp) |
34dc7c2f | 4044 | { |
0037b49e | 4045 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
89cd2197 BB |
4046 | znode_t *zp = ITOZ(ip); |
4047 | int error; | |
34dc7c2f | 4048 | |
89cd2197 BB |
4049 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
4050 | return (error); | |
34dc7c2f | 4051 | |
89cd2197 BB |
4052 | error = zfs_fillpage(ip, pp); |
4053 | if (error == 0) | |
4054 | dataset_kstats_update_read_kstats(&zfsvfs->z_kstat, PAGE_SIZE); | |
3819aaaf | 4055 | |
768eaced | 4056 | zfs_exit(zfsvfs, FTAG); |
89cd2197 BB |
4057 | |
4058 | return (error); | |
34dc7c2f BB |
4059 | } |
4060 | ||
4061 | /* | |
e2e7aa2d | 4062 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4063 | * |
e2e7aa2d BB |
4064 | * IN: ip - inode of the file to mmap |
4065 | * off - file offset | |
4066 | * addrp - start address in memory region | |
4067 | * len - length of memory region | |
4068 | * vm_flags- address flags | |
34dc7c2f | 4069 | * |
e2e7aa2d BB |
4070 | * RETURN: 0 if success |
4071 | * error code if failure | |
34dc7c2f | 4072 | */ |
e2e7aa2d BB |
4073 | int |
4074 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4075 | unsigned long vm_flags) | |
34dc7c2f | 4076 | { |
ef70eff1 | 4077 | (void) addrp; |
e2e7aa2d | 4078 | znode_t *zp = ITOZ(ip); |
0037b49e | 4079 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
768eaced | 4080 | int error; |
34dc7c2f | 4081 | |
768eaced CC |
4082 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
4083 | return (error); | |
34dc7c2f | 4084 | |
a160c153 W |
4085 | if ((vm_flags & VM_WRITE) && (vm_flags & VM_SHARED) && |
4086 | (zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { | |
768eaced | 4087 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4088 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4089 | } |
4090 | ||
e2e7aa2d | 4091 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4092 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
768eaced | 4093 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4094 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4095 | } |
4096 | ||
34dc7c2f | 4097 | if (off < 0 || len > MAXOFFSET_T - off) { |
768eaced | 4098 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4099 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4100 | } |
4101 | ||
768eaced | 4102 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
4103 | return (0); |
4104 | } | |
4105 | ||
4106 | /* | |
4107 | * Free or allocate space in a file. Currently, this function only | |
4108 | * supports the `F_FREESP' command. However, this command is somewhat | |
4109 | * misnamed, as its functionality includes the ability to allocate as | |
4110 | * well as free space. | |
4111 | * | |
657ce253 | 4112 | * IN: zp - znode of file to free data in. |
34dc7c2f BB |
4113 | * cmd - action to take (only F_FREESP supported). |
4114 | * bfp - section of file to free/alloc. | |
4115 | * flag - current file open mode flags. | |
4116 | * offset - current file offset. | |
a35c1207 | 4117 | * cr - credentials of caller. |
34dc7c2f | 4118 | * |
d3cc8b15 | 4119 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4120 | * |
4121 | * Timestamps: | |
657ce253 | 4122 | * zp - ctime|mtime updated |
34dc7c2f | 4123 | */ |
e5c39b95 | 4124 | int |
657ce253 | 4125 | zfs_space(znode_t *zp, int cmd, flock64_t *bfp, int flag, |
3558fd73 | 4126 | offset_t offset, cred_t *cr) |
34dc7c2f | 4127 | { |
ef70eff1 | 4128 | (void) offset; |
657ce253 | 4129 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
4130 | uint64_t off, len; |
4131 | int error; | |
4132 | ||
768eaced CC |
4133 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
4134 | return (error); | |
34dc7c2f | 4135 | |
34dc7c2f | 4136 | if (cmd != F_FREESP) { |
768eaced | 4137 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4138 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4139 | } |
4140 | ||
f3c9dca0 MT |
4141 | /* |
4142 | * Callers might not be able to detect properly that we are read-only, | |
4143 | * so check it explicitly here. | |
4144 | */ | |
0037b49e | 4145 | if (zfs_is_readonly(zfsvfs)) { |
768eaced | 4146 | zfs_exit(zfsvfs, FTAG); |
f3c9dca0 MT |
4147 | return (SET_ERROR(EROFS)); |
4148 | } | |
4149 | ||
34dc7c2f | 4150 | if (bfp->l_len < 0) { |
768eaced | 4151 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4152 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4153 | } |
4154 | ||
aec69371 ED |
4155 | /* |
4156 | * Permissions aren't checked on Solaris because on this OS | |
4157 | * zfs_space() can only be called with an opened file handle. | |
4158 | * On Linux we can get here through truncate_range() which | |
4159 | * operates directly on inodes, so we need to check access rights. | |
4160 | */ | |
f224eddf | 4161 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr, |
d4dc53da | 4162 | zfs_init_idmap))) { |
768eaced | 4163 | zfs_exit(zfsvfs, FTAG); |
aec69371 ED |
4164 | return (error); |
4165 | } | |
4166 | ||
34dc7c2f BB |
4167 | off = bfp->l_start; |
4168 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4169 | ||
b128c09f | 4170 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4171 | |
768eaced | 4172 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
4173 | return (error); |
4174 | } | |
4175 | ||
e5c39b95 | 4176 | int |
3558fd73 | 4177 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4178 | { |
3558fd73 | 4179 | znode_t *zp = ITOZ(ip); |
0037b49e | 4180 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4181 | uint32_t gen; |
428870ff | 4182 | uint64_t gen64; |
34dc7c2f BB |
4183 | uint64_t object = zp->z_id; |
4184 | zfid_short_t *zfid; | |
428870ff | 4185 | int size, i, error; |
34dc7c2f | 4186 | |
768eaced CC |
4187 | if ((error = zfs_enter(zfsvfs, FTAG)) != 0) |
4188 | return (error); | |
c074a7de AR |
4189 | |
4190 | if (fidp->fid_len < SHORT_FID_LEN) { | |
4191 | fidp->fid_len = SHORT_FID_LEN; | |
768eaced | 4192 | zfs_exit(zfsvfs, FTAG); |
c074a7de AR |
4193 | return (SET_ERROR(ENOSPC)); |
4194 | } | |
4195 | ||
768eaced CC |
4196 | if ((error = zfs_verify_zp(zp)) != 0) { |
4197 | zfs_exit(zfsvfs, FTAG); | |
4198 | return (error); | |
4199 | } | |
428870ff | 4200 | |
0037b49e | 4201 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), |
428870ff | 4202 | &gen64, sizeof (uint64_t))) != 0) { |
768eaced | 4203 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
4204 | return (error); |
4205 | } | |
4206 | ||
4207 | gen = (uint32_t)gen64; | |
34dc7c2f | 4208 | |
9b77d1c9 | 4209 | size = SHORT_FID_LEN; |
34dc7c2f BB |
4210 | |
4211 | zfid = (zfid_short_t *)fidp; | |
4212 | ||
4213 | zfid->zf_len = size; | |
4214 | ||
4215 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4216 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4217 | ||
4218 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4219 | if (gen == 0) | |
4220 | gen = 1; | |
4221 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4222 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4223 | ||
768eaced | 4224 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
4225 | return (0); |
4226 | } | |
4227 | ||
93ce2b4c | 4228 | #if defined(_KERNEL) |
f298b24d BB |
4229 | EXPORT_SYMBOL(zfs_open); |
4230 | EXPORT_SYMBOL(zfs_close); | |
f298b24d BB |
4231 | EXPORT_SYMBOL(zfs_lookup); |
4232 | EXPORT_SYMBOL(zfs_create); | |
4233 | EXPORT_SYMBOL(zfs_tmpfile); | |
4234 | EXPORT_SYMBOL(zfs_remove); | |
4235 | EXPORT_SYMBOL(zfs_mkdir); | |
4236 | EXPORT_SYMBOL(zfs_rmdir); | |
4237 | EXPORT_SYMBOL(zfs_readdir); | |
f298b24d BB |
4238 | EXPORT_SYMBOL(zfs_getattr_fast); |
4239 | EXPORT_SYMBOL(zfs_setattr); | |
4240 | EXPORT_SYMBOL(zfs_rename); | |
4241 | EXPORT_SYMBOL(zfs_symlink); | |
4242 | EXPORT_SYMBOL(zfs_readlink); | |
4243 | EXPORT_SYMBOL(zfs_link); | |
4244 | EXPORT_SYMBOL(zfs_inactive); | |
4245 | EXPORT_SYMBOL(zfs_space); | |
4246 | EXPORT_SYMBOL(zfs_fid); | |
f298b24d BB |
4247 | EXPORT_SYMBOL(zfs_getpage); |
4248 | EXPORT_SYMBOL(zfs_putpage); | |
4249 | EXPORT_SYMBOL(zfs_dirty_inode); | |
4250 | EXPORT_SYMBOL(zfs_map); | |
4251 | ||
7ada752a | 4252 | /* CSTYLED */ |
a966c564 K |
4253 | module_param(zfs_delete_blocks, ulong, 0644); |
4254 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
c409e464 | 4255 | #endif |