]>
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 */ | |
189 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
190 | ((flag & O_APPEND) == 0)) { | |
768eaced | 191 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 192 | return (SET_ERROR(EPERM)); |
126400a1 BB |
193 | } |
194 | ||
126400a1 BB |
195 | /* Keep a count of the synchronous opens in the znode */ |
196 | if (flag & O_SYNC) | |
197 | atomic_inc_32(&zp->z_sync_cnt); | |
198 | ||
768eaced | 199 | zfs_exit(zfsvfs, FTAG); |
126400a1 BB |
200 | return (0); |
201 | } | |
126400a1 | 202 | |
126400a1 BB |
203 | int |
204 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
205 | { | |
ef70eff1 | 206 | (void) cr; |
126400a1 | 207 | znode_t *zp = ITOZ(ip); |
0037b49e | 208 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
768eaced | 209 | int error; |
126400a1 | 210 | |
768eaced CC |
211 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
212 | return (error); | |
126400a1 | 213 | |
7dc71949 | 214 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 215 | if (flag & O_SYNC) |
7dc71949 | 216 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 | 217 | |
768eaced | 218 | zfs_exit(zfsvfs, FTAG); |
8780c539 | 219 | return (0); |
126400a1 | 220 | } |
126400a1 | 221 | |
c0d35759 | 222 | #if defined(_KERNEL) |
34dc7c2f BB |
223 | /* |
224 | * When a file is memory mapped, we must keep the IO data synchronized | |
225 | * between the DMU cache and the memory mapped pages. What this means: | |
226 | * | |
227 | * On Write: If we find a memory mapped page, we write to *both* | |
228 | * the page and the dmu buffer. | |
34dc7c2f | 229 | */ |
e53d678d | 230 | void |
8a9634e2 | 231 | update_pages(znode_t *zp, int64_t start, int len, objset_t *os) |
34dc7c2f | 232 | { |
e53d678d | 233 | struct inode *ip = ZTOI(zp); |
c0d35759 BB |
234 | struct address_space *mp = ip->i_mapping; |
235 | struct page *pp; | |
236 | uint64_t nbytes; | |
d164b209 | 237 | int64_t off; |
c0d35759 | 238 | void *pb; |
34dc7c2f | 239 | |
8b1899d3 BB |
240 | off = start & (PAGE_SIZE-1); |
241 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
242 | nbytes = MIN(PAGE_SIZE - off, len); | |
34dc7c2f | 243 | |
8b1899d3 | 244 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
245 | if (pp) { |
246 | if (mapping_writably_mapped(mp)) | |
247 | flush_dcache_page(pp); | |
34dc7c2f | 248 | |
c0d35759 | 249 | pb = kmap(pp); |
8a9634e2 RM |
250 | (void) dmu_read(os, zp->z_id, start + off, nbytes, |
251 | pb + off, DMU_READ_PREFETCH); | |
c0d35759 BB |
252 | kunmap(pp); |
253 | ||
254 | if (mapping_writably_mapped(mp)) | |
255 | flush_dcache_page(pp); | |
256 | ||
257 | mark_page_accessed(pp); | |
258 | SetPageUptodate(pp); | |
259 | ClearPageError(pp); | |
260 | unlock_page(pp); | |
8b1899d3 | 261 | put_page(pp); |
34dc7c2f | 262 | } |
c0d35759 | 263 | |
d164b209 | 264 | len -= nbytes; |
34dc7c2f | 265 | off = 0; |
34dc7c2f | 266 | } |
34dc7c2f BB |
267 | } |
268 | ||
269 | /* | |
270 | * When a file is memory mapped, we must keep the IO data synchronized | |
271 | * between the DMU cache and the memory mapped pages. What this means: | |
272 | * | |
273 | * On Read: We "read" preferentially from memory mapped pages, | |
274 | * else we default from the dmu buffer. | |
275 | * | |
276 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
d3cc8b15 | 277 | * the file is memory mapped. |
34dc7c2f | 278 | */ |
e53d678d | 279 | int |
d0cd9a5c | 280 | mappedread(znode_t *zp, int nbytes, zfs_uio_t *uio) |
34dc7c2f | 281 | { |
e53d678d | 282 | struct inode *ip = ZTOI(zp); |
c0d35759 BB |
283 | struct address_space *mp = ip->i_mapping; |
284 | struct page *pp; | |
34dc7c2f | 285 | int64_t start, off; |
c0d35759 | 286 | uint64_t bytes; |
34dc7c2f BB |
287 | int len = nbytes; |
288 | int error = 0; | |
c0d35759 | 289 | void *pb; |
34dc7c2f BB |
290 | |
291 | start = uio->uio_loffset; | |
8b1899d3 BB |
292 | off = start & (PAGE_SIZE-1); |
293 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
294 | bytes = MIN(PAGE_SIZE - off, len); | |
c0d35759 | 295 | |
8b1899d3 | 296 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
297 | if (pp) { |
298 | ASSERT(PageUptodate(pp)); | |
b2ab468d | 299 | unlock_page(pp); |
c0d35759 BB |
300 | |
301 | pb = kmap(pp); | |
d0cd9a5c | 302 | error = zfs_uiomove(pb + off, bytes, UIO_READ, uio); |
c0d35759 BB |
303 | kunmap(pp); |
304 | ||
305 | if (mapping_writably_mapped(mp)) | |
306 | flush_dcache_page(pp); | |
307 | ||
308 | mark_page_accessed(pp); | |
8b1899d3 | 309 | put_page(pp); |
34dc7c2f | 310 | } else { |
804e0504 MA |
311 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
312 | uio, bytes); | |
34dc7c2f | 313 | } |
c0d35759 | 314 | |
34dc7c2f BB |
315 | len -= bytes; |
316 | off = 0; | |
317 | if (error) | |
318 | break; | |
319 | } | |
320 | return (error); | |
321 | } | |
c0d35759 | 322 | #endif /* _KERNEL */ |
34dc7c2f | 323 | |
18168da7 | 324 | static unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT; |
34dc7c2f | 325 | |
13a9a6f5 MM |
326 | /* |
327 | * Write the bytes to a file. | |
328 | * | |
329 | * IN: zp - znode of file to be written to | |
330 | * data - bytes to write | |
331 | * len - number of bytes to write | |
332 | * pos - offset to start writing at | |
333 | * | |
334 | * OUT: resid - remaining bytes to write | |
335 | * | |
336 | * RETURN: 0 if success | |
1c2358c1 BB |
337 | * positive error code if failure. EIO is returned |
338 | * for a short write when residp isn't provided. | |
13a9a6f5 MM |
339 | * |
340 | * Timestamps: | |
341 | * zp - ctime|mtime updated if byte count > 0 | |
342 | */ | |
343 | int | |
344 | zfs_write_simple(znode_t *zp, const void *data, size_t len, | |
1c2358c1 | 345 | loff_t pos, size_t *residp) |
13a9a6f5 | 346 | { |
1c2358c1 BB |
347 | fstrans_cookie_t cookie; |
348 | int error; | |
349 | ||
350 | struct iovec iov; | |
351 | iov.iov_base = (void *)data; | |
352 | iov.iov_len = len; | |
353 | ||
d0cd9a5c BA |
354 | zfs_uio_t uio; |
355 | zfs_uio_iovec_init(&uio, &iov, 1, pos, UIO_SYSSPACE, len, 0); | |
13a9a6f5 | 356 | |
1c2358c1 BB |
357 | cookie = spl_fstrans_mark(); |
358 | error = zfs_write(zp, &uio, 0, kcred); | |
359 | spl_fstrans_unmark(cookie); | |
360 | ||
361 | if (error == 0) { | |
362 | if (residp != NULL) | |
d0cd9a5c BA |
363 | *residp = zfs_uio_resid(&uio); |
364 | else if (zfs_uio_resid(&uio) != 0) | |
1c2358c1 | 365 | error = SET_ERROR(EIO); |
13a9a6f5 | 366 | } |
1c2358c1 | 367 | |
13a9a6f5 MM |
368 | return (error); |
369 | } | |
370 | ||
23c13c7e AL |
371 | static void |
372 | zfs_rele_async_task(void *arg) | |
373 | { | |
374 | iput(arg); | |
375 | } | |
376 | ||
0a50679c | 377 | void |
657ce253 | 378 | zfs_zrele_async(znode_t *zp) |
3558fd73 | 379 | { |
657ce253 | 380 | struct inode *ip = ZTOI(zp); |
0a50679c BB |
381 | objset_t *os = ITOZSB(ip)->z_os; |
382 | ||
3558fd73 | 383 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
384 | ASSERT(os != NULL); |
385 | ||
2921ad6c PD |
386 | /* |
387 | * If decrementing the count would put us at 0, we can't do it inline | |
388 | * here, because that would be synchronous. Instead, dispatch an iput | |
389 | * to run later. | |
390 | * | |
391 | * For more information on the dangers of a synchronous iput, see the | |
392 | * header comment of this file. | |
393 | */ | |
394 | if (!atomic_add_unless(&ip->i_count, -1, 1)) { | |
657ce253 | 395 | VERIFY(taskq_dispatch(dsl_pool_zrele_taskq(dmu_objset_pool(os)), |
23c13c7e | 396 | zfs_rele_async_task, ip, TQ_SLEEP) != TASKQID_INVALID); |
2921ad6c | 397 | } |
3558fd73 BB |
398 | } |
399 | ||
34dc7c2f | 400 | |
34dc7c2f BB |
401 | /* |
402 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 403 | * If it exists, return a held inode reference for it. |
34dc7c2f | 404 | * |
657ce253 | 405 | * IN: zdp - znode of directory to search. |
34dc7c2f | 406 | * nm - name of entry to lookup. |
34dc7c2f | 407 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 408 | * cr - credentials of caller. |
34dc7c2f BB |
409 | * direntflags - directory lookup flags |
410 | * realpnp - returned pathname. | |
411 | * | |
657ce253 | 412 | * OUT: zpp - znode of located entry, NULL if not found. |
34dc7c2f | 413 | * |
d3cc8b15 | 414 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
415 | * |
416 | * Timestamps: | |
417 | * NA | |
418 | */ | |
e5c39b95 | 419 | int |
4d55ea81 RM |
420 | zfs_lookup(znode_t *zdp, char *nm, znode_t **zpp, int flags, cred_t *cr, |
421 | int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 422 | { |
657ce253 | 423 | zfsvfs_t *zfsvfs = ZTOZSB(zdp); |
3558fd73 | 424 | int error = 0; |
45d1cae3 | 425 | |
9b7b9cd3 GM |
426 | /* |
427 | * Fast path lookup, however we must skip DNLC lookup | |
428 | * for case folding or normalizing lookups because the | |
429 | * DNLC code only stores the passed in name. This means | |
430 | * creating 'a' and removing 'A' on a case insensitive | |
431 | * file system would work, but DNLC still thinks 'a' | |
432 | * exists and won't let you create it again on the next | |
433 | * pass through fast path. | |
434 | */ | |
45d1cae3 BB |
435 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { |
436 | ||
657ce253 | 437 | if (!S_ISDIR(ZTOI(zdp)->i_mode)) { |
2e528b49 | 438 | return (SET_ERROR(ENOTDIR)); |
428870ff | 439 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 440 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
441 | } |
442 | ||
443 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
444 | error = zfs_fastaccesschk_execute(zdp, cr); | |
445 | if (!error) { | |
657ce253 MM |
446 | *zpp = zdp; |
447 | zhold(*zpp); | |
45d1cae3 BB |
448 | return (0); |
449 | } | |
450 | return (error); | |
45d1cae3 BB |
451 | } |
452 | } | |
453 | ||
768eaced CC |
454 | if ((error = zfs_enter_verify_zp(zfsvfs, zdp, FTAG)) != 0) |
455 | return (error); | |
34dc7c2f | 456 | |
657ce253 | 457 | *zpp = NULL; |
34dc7c2f BB |
458 | |
459 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
460 | /* |
461 | * We don't allow recursive attributes.. | |
462 | * Maybe someday we will. | |
463 | */ | |
428870ff | 464 | if (zdp->z_pflags & ZFS_XATTR) { |
768eaced | 465 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 466 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
467 | } |
468 | ||
657ce253 | 469 | if ((error = zfs_get_xattrdir(zdp, zpp, cr, flags))) { |
768eaced | 470 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
471 | return (error); |
472 | } | |
473 | ||
474 | /* | |
475 | * Do we have permission to get into attribute directory? | |
476 | */ | |
477 | ||
657ce253 | 478 | if ((error = zfs_zaccess(*zpp, ACE_EXECUTE, 0, |
2a068a13 | 479 | B_TRUE, cr, NULL))) { |
657ce253 MM |
480 | zrele(*zpp); |
481 | *zpp = NULL; | |
34dc7c2f BB |
482 | } |
483 | ||
768eaced | 484 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
485 | return (error); |
486 | } | |
487 | ||
657ce253 | 488 | if (!S_ISDIR(ZTOI(zdp)->i_mode)) { |
768eaced | 489 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 490 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
491 | } |
492 | ||
493 | /* | |
494 | * Check accessibility of directory. | |
495 | */ | |
496 | ||
2a068a13 | 497 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr, NULL))) { |
768eaced | 498 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
499 | return (error); |
500 | } | |
501 | ||
0037b49e | 502 | if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 503 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 504 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 505 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
506 | } |
507 | ||
657ce253 MM |
508 | error = zfs_dirlook(zdp, nm, zpp, flags, direntflags, realpnp); |
509 | if ((error == 0) && (*zpp)) | |
fc273894 | 510 | zfs_znode_update_vfs(*zpp); |
34dc7c2f | 511 | |
768eaced | 512 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
513 | return (error); |
514 | } | |
515 | ||
516 | /* | |
517 | * Attempt to create a new entry in a directory. If the entry | |
518 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 519 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 520 | * |
657ce253 | 521 | * IN: dzp - znode of directory to put new file entry in. |
34dc7c2f BB |
522 | * name - name of new file entry. |
523 | * vap - attributes of new file. | |
524 | * excl - flag indicating exclusive or non-exclusive mode. | |
525 | * mode - mode to open file with. | |
526 | * cr - credentials of caller. | |
a35c1207 | 527 | * flag - file flag. |
3558fd73 | 528 | * vsecp - ACL to be set |
2a068a13 | 529 | * mnt_ns - user namespace of the mount |
34dc7c2f | 530 | * |
657ce253 | 531 | * OUT: zpp - znode of created or trunc'd entry. |
34dc7c2f | 532 | * |
d3cc8b15 | 533 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
534 | * |
535 | * Timestamps: | |
657ce253 MM |
536 | * dzp - ctime|mtime updated if new entry created |
537 | * zp - ctime|mtime always, atime if new | |
34dc7c2f | 538 | */ |
e5c39b95 | 539 | int |
657ce253 | 540 | zfs_create(znode_t *dzp, char *name, vattr_t *vap, int excl, |
2a068a13 YY |
541 | int mode, znode_t **zpp, cred_t *cr, int flag, vsecattr_t *vsecp, |
542 | zuserns_t *mnt_ns) | |
34dc7c2f | 543 | { |
657ce253 MM |
544 | znode_t *zp; |
545 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); | |
34dc7c2f BB |
546 | zilog_t *zilog; |
547 | objset_t *os; | |
548 | zfs_dirlock_t *dl; | |
549 | dmu_tx_t *tx; | |
550 | int error; | |
b128c09f | 551 | uid_t uid; |
149e873a | 552 | gid_t gid; |
428870ff | 553 | zfs_acl_ids_t acl_ids; |
9babb374 | 554 | boolean_t fuid_dirtied; |
428870ff | 555 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 556 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
557 | |
558 | /* | |
559 | * If we have an ephemeral id, ACL, or XVATTR then | |
560 | * make sure file system is at proper version | |
561 | */ | |
562 | ||
149e873a | 563 | gid = crgetgid(cr); |
3558fd73 | 564 | uid = crgetuid(cr); |
b128c09f | 565 | |
0037b49e | 566 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 567 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 568 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 569 | |
32dec7bd | 570 | if (name == NULL) |
571 | return (SET_ERROR(EINVAL)); | |
572 | ||
768eaced CC |
573 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
574 | return (error); | |
0037b49e BB |
575 | os = zfsvfs->z_os; |
576 | zilog = zfsvfs->z_log; | |
34dc7c2f | 577 | |
0037b49e | 578 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 579 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 580 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 581 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
582 | } |
583 | ||
5484965a | 584 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 585 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 586 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
768eaced | 587 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
588 | return (error); |
589 | } | |
590 | } | |
34dc7c2f | 591 | |
3558fd73 | 592 | top: |
657ce253 | 593 | *zpp = NULL; |
34dc7c2f BB |
594 | if (*name == '\0') { |
595 | /* | |
596 | * Null component name refers to the directory itself. | |
597 | */ | |
657ce253 | 598 | zhold(dzp); |
34dc7c2f BB |
599 | zp = dzp; |
600 | dl = NULL; | |
601 | error = 0; | |
602 | } else { | |
3558fd73 | 603 | /* possible igrab(zp) */ |
34dc7c2f BB |
604 | int zflg = 0; |
605 | ||
606 | if (flag & FIGNORECASE) | |
607 | zflg |= ZCILOOK; | |
608 | ||
609 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
610 | NULL, NULL); | |
611 | if (error) { | |
572e2857 BB |
612 | if (have_acl) |
613 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 614 | if (strcmp(name, "..") == 0) |
2e528b49 | 615 | error = SET_ERROR(EISDIR); |
768eaced | 616 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
617 | return (error); |
618 | } | |
619 | } | |
428870ff | 620 | |
34dc7c2f BB |
621 | if (zp == NULL) { |
622 | uint64_t txtype; | |
9c5167d1 | 623 | uint64_t projid = ZFS_DEFAULT_PROJID; |
34dc7c2f BB |
624 | |
625 | /* | |
626 | * Create a new file object and update the directory | |
627 | * to reference it. | |
628 | */ | |
2a068a13 YY |
629 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr, |
630 | mnt_ns))) { | |
572e2857 BB |
631 | if (have_acl) |
632 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
633 | goto out; |
634 | } | |
635 | ||
636 | /* | |
637 | * We only support the creation of regular files in | |
638 | * extended attribute directories. | |
639 | */ | |
428870ff | 640 | |
3558fd73 | 641 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
642 | if (have_acl) |
643 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 644 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
645 | goto out; |
646 | } | |
647 | ||
428870ff | 648 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
2a068a13 | 649 | cr, vsecp, &acl_ids, mnt_ns)) != 0) |
9babb374 | 650 | goto out; |
428870ff BB |
651 | have_acl = B_TRUE; |
652 | ||
9c5167d1 NF |
653 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
654 | projid = zfs_inherit_projid(dzp); | |
655 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
45d1cae3 | 656 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 657 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
658 | goto out; |
659 | } | |
660 | ||
34dc7c2f | 661 | tx = dmu_tx_create(os); |
428870ff BB |
662 | |
663 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
664 | ZFS_SA_BASE_ATTR_SIZE); | |
665 | ||
0037b49e | 666 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 667 | if (fuid_dirtied) |
0037b49e | 668 | zfs_fuid_txhold(zfsvfs, tx); |
34dc7c2f | 669 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 670 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
0037b49e | 671 | if (!zfsvfs->z_use_sa && |
428870ff | 672 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 673 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 674 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 675 | } |
599b8648 | 676 | |
0735ecb3 PS |
677 | error = dmu_tx_assign(tx, |
678 | (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); | |
34dc7c2f BB |
679 | if (error) { |
680 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 681 | if (error == ERESTART) { |
e8b96c60 | 682 | waited = B_TRUE; |
34dc7c2f BB |
683 | dmu_tx_wait(tx); |
684 | dmu_tx_abort(tx); | |
685 | goto top; | |
686 | } | |
428870ff | 687 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 688 | dmu_tx_abort(tx); |
768eaced | 689 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
690 | return (error); |
691 | } | |
428870ff | 692 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 693 | |
599b8648 CC |
694 | error = zfs_link_create(dl, zp, tx, ZNEW); |
695 | if (error != 0) { | |
696 | /* | |
697 | * Since, we failed to add the directory entry for it, | |
698 | * delete the newly created dnode. | |
699 | */ | |
700 | zfs_znode_delete(zp, tx); | |
701 | remove_inode_hash(ZTOI(zp)); | |
702 | zfs_acl_ids_free(&acl_ids); | |
703 | dmu_tx_commit(tx); | |
704 | goto out; | |
705 | } | |
706 | ||
9babb374 | 707 | if (fuid_dirtied) |
0037b49e | 708 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 709 | |
34dc7c2f BB |
710 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); |
711 | if (flag & FIGNORECASE) | |
712 | txtype |= TX_CI; | |
713 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
714 | vsecp, acl_ids.z_fuidp, vap); |
715 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
716 | dmu_tx_commit(tx); |
717 | } else { | |
da92d5cb | 718 | int aflags = (flag & O_APPEND) ? V_APPEND : 0; |
34dc7c2f | 719 | |
572e2857 BB |
720 | if (have_acl) |
721 | zfs_acl_ids_free(&acl_ids); | |
722 | have_acl = B_FALSE; | |
723 | ||
34dc7c2f BB |
724 | /* |
725 | * A directory entry already exists for this name. | |
726 | */ | |
727 | /* | |
728 | * Can't truncate an existing file if in exclusive mode. | |
729 | */ | |
3558fd73 | 730 | if (excl) { |
2e528b49 | 731 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
732 | goto out; |
733 | } | |
734 | /* | |
735 | * Can't open a directory for writing. | |
736 | */ | |
3558fd73 | 737 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 738 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
739 | goto out; |
740 | } | |
741 | /* | |
742 | * Verify requested access to file. | |
743 | */ | |
2a068a13 YY |
744 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr, |
745 | mnt_ns))) { | |
34dc7c2f BB |
746 | goto out; |
747 | } | |
748 | ||
749 | mutex_enter(&dzp->z_lock); | |
750 | dzp->z_seq++; | |
751 | mutex_exit(&dzp->z_lock); | |
752 | ||
753 | /* | |
754 | * Truncate regular files if requested. | |
755 | */ | |
3558fd73 BB |
756 | if (S_ISREG(ZTOI(zp)->i_mode) && |
757 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 758 | /* we can't hold any locks when calling zfs_freesp() */ |
609603a5 B |
759 | if (dl) { |
760 | zfs_dirent_unlock(dl); | |
761 | dl = NULL; | |
762 | } | |
34dc7c2f | 763 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
764 | } |
765 | } | |
766 | out: | |
767 | ||
768 | if (dl) | |
769 | zfs_dirent_unlock(dl); | |
770 | ||
771 | if (error) { | |
772 | if (zp) | |
657ce253 | 773 | zrele(zp); |
34dc7c2f | 774 | } else { |
fc273894 KHN |
775 | zfs_znode_update_vfs(dzp); |
776 | zfs_znode_update_vfs(zp); | |
657ce253 | 777 | *zpp = zp; |
34dc7c2f | 778 | } |
34dc7c2f | 779 | |
0037b49e | 780 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 781 | zil_commit(zilog, 0); |
428870ff | 782 | |
768eaced | 783 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
784 | return (error); |
785 | } | |
786 | ||
ace1eae8 CC |
787 | int |
788 | zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, | |
2a068a13 YY |
789 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp, |
790 | zuserns_t *mnt_ns) | |
ace1eae8 | 791 | { |
ef70eff1 | 792 | (void) excl, (void) mode, (void) flag; |
ace1eae8 | 793 | znode_t *zp = NULL, *dzp = ITOZ(dip); |
0037b49e | 794 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
ace1eae8 CC |
795 | objset_t *os; |
796 | dmu_tx_t *tx; | |
797 | int error; | |
798 | uid_t uid; | |
799 | gid_t gid; | |
800 | zfs_acl_ids_t acl_ids; | |
9c5167d1 | 801 | uint64_t projid = ZFS_DEFAULT_PROJID; |
ace1eae8 CC |
802 | boolean_t fuid_dirtied; |
803 | boolean_t have_acl = B_FALSE; | |
804 | boolean_t waited = B_FALSE; | |
805 | ||
806 | /* | |
807 | * If we have an ephemeral id, ACL, or XVATTR then | |
808 | * make sure file system is at proper version | |
809 | */ | |
810 | ||
811 | gid = crgetgid(cr); | |
812 | uid = crgetuid(cr); | |
813 | ||
0037b49e | 814 | if (zfsvfs->z_use_fuids == B_FALSE && |
ace1eae8 CC |
815 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
816 | return (SET_ERROR(EINVAL)); | |
817 | ||
768eaced CC |
818 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
819 | return (error); | |
0037b49e | 820 | os = zfsvfs->z_os; |
ace1eae8 CC |
821 | |
822 | if (vap->va_mask & ATTR_XVATTR) { | |
823 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
824 | crgetuid(cr), cr, vap->va_mode)) != 0) { | |
768eaced | 825 | zfs_exit(zfsvfs, FTAG); |
ace1eae8 CC |
826 | return (error); |
827 | } | |
828 | } | |
829 | ||
830 | top: | |
831 | *ipp = NULL; | |
832 | ||
833 | /* | |
834 | * Create a new file object and update the directory | |
835 | * to reference it. | |
836 | */ | |
2a068a13 | 837 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr, mnt_ns))) { |
ace1eae8 CC |
838 | if (have_acl) |
839 | zfs_acl_ids_free(&acl_ids); | |
840 | goto out; | |
841 | } | |
842 | ||
843 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, | |
2a068a13 | 844 | cr, vsecp, &acl_ids, mnt_ns)) != 0) |
ace1eae8 CC |
845 | goto out; |
846 | have_acl = B_TRUE; | |
847 | ||
9c5167d1 NF |
848 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
849 | projid = zfs_inherit_projid(dzp); | |
850 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
ace1eae8 CC |
851 | zfs_acl_ids_free(&acl_ids); |
852 | error = SET_ERROR(EDQUOT); | |
853 | goto out; | |
854 | } | |
855 | ||
856 | tx = dmu_tx_create(os); | |
857 | ||
858 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
859 | ZFS_SA_BASE_ATTR_SIZE); | |
0037b49e | 860 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 861 | |
0037b49e | 862 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
ace1eae8 | 863 | if (fuid_dirtied) |
0037b49e BB |
864 | zfs_fuid_txhold(zfsvfs, tx); |
865 | if (!zfsvfs->z_use_sa && | |
ace1eae8 CC |
866 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
867 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
868 | 0, acl_ids.z_aclp->z_acl_bytes); | |
869 | } | |
0735ecb3 | 870 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
ace1eae8 CC |
871 | if (error) { |
872 | if (error == ERESTART) { | |
873 | waited = B_TRUE; | |
874 | dmu_tx_wait(tx); | |
875 | dmu_tx_abort(tx); | |
876 | goto top; | |
877 | } | |
878 | zfs_acl_ids_free(&acl_ids); | |
879 | dmu_tx_abort(tx); | |
768eaced | 880 | zfs_exit(zfsvfs, FTAG); |
ace1eae8 CC |
881 | return (error); |
882 | } | |
883 | zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); | |
884 | ||
885 | if (fuid_dirtied) | |
0037b49e | 886 | zfs_fuid_sync(zfsvfs, tx); |
ace1eae8 CC |
887 | |
888 | /* Add to unlinked set */ | |
a43570c5 | 889 | zp->z_unlinked = B_TRUE; |
ace1eae8 CC |
890 | zfs_unlinked_add(zp, tx); |
891 | zfs_acl_ids_free(&acl_ids); | |
892 | dmu_tx_commit(tx); | |
893 | out: | |
894 | ||
895 | if (error) { | |
896 | if (zp) | |
657ce253 | 897 | zrele(zp); |
ace1eae8 | 898 | } else { |
fc273894 KHN |
899 | zfs_znode_update_vfs(dzp); |
900 | zfs_znode_update_vfs(zp); | |
ace1eae8 CC |
901 | *ipp = ZTOI(zp); |
902 | } | |
903 | ||
768eaced | 904 | zfs_exit(zfsvfs, FTAG); |
ace1eae8 CC |
905 | return (error); |
906 | } | |
907 | ||
34dc7c2f BB |
908 | /* |
909 | * Remove an entry from a directory. | |
910 | * | |
657ce253 | 911 | * IN: dzp - znode of directory to remove entry from. |
34dc7c2f BB |
912 | * name - name of entry to remove. |
913 | * cr - credentials of caller. | |
841a7a98 | 914 | * flags - case flags. |
34dc7c2f BB |
915 | * |
916 | * RETURN: 0 if success | |
917 | * error code if failure | |
918 | * | |
919 | * Timestamps: | |
657ce253 | 920 | * dzp - ctime|mtime |
3558fd73 | 921 | * ip - ctime (if nlink > 0) |
34dc7c2f | 922 | */ |
428870ff | 923 | |
51c747de | 924 | static uint64_t null_xattr = 0; |
428870ff | 925 | |
e5c39b95 | 926 | int |
657ce253 | 927 | zfs_remove(znode_t *dzp, char *name, cred_t *cr, int flags) |
34dc7c2f | 928 | { |
657ce253 | 929 | znode_t *zp; |
572e2857 | 930 | znode_t *xzp; |
657ce253 | 931 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f | 932 | zilog_t *zilog; |
a966c564 | 933 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 934 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 935 | uint64_t obj = 0; |
dfbc8630 | 936 | uint64_t links; |
34dc7c2f BB |
937 | zfs_dirlock_t *dl; |
938 | dmu_tx_t *tx; | |
a966c564 K |
939 | boolean_t may_delete_now, delete_now = FALSE; |
940 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
941 | uint64_t txtype; |
942 | pathname_t *realnmp = NULL; | |
943 | pathname_t realnm; | |
944 | int error; | |
945 | int zflg = ZEXISTS; | |
e8b96c60 | 946 | boolean_t waited = B_FALSE; |
34dc7c2f | 947 | |
32dec7bd | 948 | if (name == NULL) |
949 | return (SET_ERROR(EINVAL)); | |
950 | ||
768eaced CC |
951 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
952 | return (error); | |
0037b49e | 953 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
954 | |
955 | if (flags & FIGNORECASE) { | |
956 | zflg |= ZCILOOK; | |
957 | pn_alloc(&realnm); | |
958 | realnmp = &realnm; | |
959 | } | |
960 | ||
961 | top: | |
572e2857 BB |
962 | xattr_obj = 0; |
963 | xzp = NULL; | |
34dc7c2f BB |
964 | /* |
965 | * Attempt to lock directory; fail if entry doesn't exist. | |
966 | */ | |
149e873a BB |
967 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
968 | NULL, realnmp))) { | |
34dc7c2f BB |
969 | if (realnmp) |
970 | pn_free(realnmp); | |
768eaced | 971 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
972 | return (error); |
973 | } | |
974 | ||
2a068a13 | 975 | if ((error = zfs_zaccess_delete(dzp, zp, cr, NULL))) { |
34dc7c2f BB |
976 | goto out; |
977 | } | |
978 | ||
979 | /* | |
980 | * Need to use rmdir for removing directories. | |
981 | */ | |
657ce253 | 982 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 983 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
984 | goto out; |
985 | } | |
986 | ||
19d55079 | 987 | mutex_enter(&zp->z_lock); |
657ce253 MM |
988 | may_delete_now = atomic_read(&ZTOI(zp)->i_count) == 1 && |
989 | !(zp->z_is_mapped); | |
19d55079 MA |
990 | mutex_exit(&zp->z_lock); |
991 | ||
34dc7c2f | 992 | /* |
a966c564 K |
993 | * We may delete the znode now, or we may put it in the unlinked set; |
994 | * it depends on whether we're the last link, and on whether there are | |
995 | * other holds on the inode. So we dmu_tx_hold() the right things to | |
996 | * allow for either case. | |
34dc7c2f | 997 | */ |
572e2857 | 998 | obj = zp->z_id; |
0037b49e | 999 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 1000 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1001 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1002 | zfs_sa_upgrade_txholds(tx, zp); | |
1003 | zfs_sa_upgrade_txholds(tx, dzp); | |
a966c564 K |
1004 | if (may_delete_now) { |
1005 | toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; | |
1006 | /* if the file is too big, only hold_free a token amount */ | |
1007 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1008 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1009 | } | |
34dc7c2f BB |
1010 | |
1011 | /* are there any extended attributes? */ | |
0037b49e | 1012 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1013 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1014 | if (error == 0 && xattr_obj) { |
0037b49e | 1015 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
c99c9001 | 1016 | ASSERT0(error); |
428870ff BB |
1017 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1018 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1019 | } |
1020 | ||
a966c564 K |
1021 | mutex_enter(&zp->z_lock); |
1022 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
1023 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
1024 | mutex_exit(&zp->z_lock); | |
1025 | ||
34dc7c2f | 1026 | /* charge as an update -- would be nice not to charge at all */ |
0037b49e | 1027 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1028 | |
19d55079 | 1029 | /* |
1a04bab3 | 1030 | * Mark this transaction as typically resulting in a net free of space |
19d55079 | 1031 | */ |
1a04bab3 | 1032 | dmu_tx_mark_netfree(tx); |
19d55079 | 1033 | |
0735ecb3 | 1034 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
1035 | if (error) { |
1036 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1037 | if (error == ERESTART) { |
e8b96c60 | 1038 | waited = B_TRUE; |
34dc7c2f BB |
1039 | dmu_tx_wait(tx); |
1040 | dmu_tx_abort(tx); | |
657ce253 | 1041 | zrele(zp); |
ea7e86d8 | 1042 | if (xzp) |
657ce253 | 1043 | zrele(xzp); |
34dc7c2f BB |
1044 | goto top; |
1045 | } | |
1046 | if (realnmp) | |
1047 | pn_free(realnmp); | |
1048 | dmu_tx_abort(tx); | |
657ce253 | 1049 | zrele(zp); |
ea7e86d8 | 1050 | if (xzp) |
657ce253 | 1051 | zrele(xzp); |
768eaced | 1052 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1053 | return (error); |
1054 | } | |
1055 | ||
1056 | /* | |
1057 | * Remove the directory entry. | |
1058 | */ | |
1059 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1060 | ||
1061 | if (error) { | |
1062 | dmu_tx_commit(tx); | |
1063 | goto out; | |
1064 | } | |
1065 | ||
1066 | if (unlinked) { | |
572e2857 BB |
1067 | /* |
1068 | * Hold z_lock so that we can make sure that the ACL obj | |
1069 | * hasn't changed. Could have been deleted due to | |
1070 | * zfs_sa_upgrade(). | |
1071 | */ | |
1072 | mutex_enter(&zp->z_lock); | |
0037b49e | 1073 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1074 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
a966c564 | 1075 | delete_now = may_delete_now && !toobig && |
657ce253 MM |
1076 | atomic_read(&ZTOI(zp)->i_count) == 1 && |
1077 | !(zp->z_is_mapped) && xattr_obj == xattr_obj_unlinked && | |
1078 | zfs_external_acl(zp) == acl_obj; | |
a966c564 K |
1079 | } |
1080 | ||
1081 | if (delete_now) { | |
1082 | if (xattr_obj_unlinked) { | |
dfbc8630 | 1083 | ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); |
a966c564 | 1084 | mutex_enter(&xzp->z_lock); |
a43570c5 | 1085 | xzp->z_unlinked = B_TRUE; |
dfbc8630 CD |
1086 | clear_nlink(ZTOI(xzp)); |
1087 | links = 0; | |
0037b49e | 1088 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 1089 | &links, sizeof (links), tx); |
a966c564 K |
1090 | ASSERT3U(error, ==, 0); |
1091 | mutex_exit(&xzp->z_lock); | |
1092 | zfs_unlinked_add(xzp, tx); | |
1093 | ||
1094 | if (zp->z_is_sa) | |
1095 | error = sa_remove(zp->z_sa_hdl, | |
0037b49e | 1096 | SA_ZPL_XATTR(zfsvfs), tx); |
a966c564 K |
1097 | else |
1098 | error = sa_update(zp->z_sa_hdl, | |
0037b49e | 1099 | SA_ZPL_XATTR(zfsvfs), &null_xattr, |
a966c564 K |
1100 | sizeof (uint64_t), tx); |
1101 | ASSERT0(error); | |
1102 | } | |
1103 | /* | |
1104 | * Add to the unlinked set because a new reference could be | |
1105 | * taken concurrently resulting in a deferred destruction. | |
1106 | */ | |
1107 | zfs_unlinked_add(zp, tx); | |
1108 | mutex_exit(&zp->z_lock); | |
a966c564 | 1109 | } else if (unlinked) { |
572e2857 | 1110 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1111 | zfs_unlinked_add(zp, tx); |
1112 | } | |
1113 | ||
1114 | txtype = TX_REMOVE; | |
1115 | if (flags & FIGNORECASE) | |
1116 | txtype |= TX_CI; | |
8e556c5e | 1117 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj, unlinked); |
34dc7c2f BB |
1118 | |
1119 | dmu_tx_commit(tx); | |
1120 | out: | |
1121 | if (realnmp) | |
1122 | pn_free(realnmp); | |
1123 | ||
1124 | zfs_dirent_unlock(dl); | |
fc273894 KHN |
1125 | zfs_znode_update_vfs(dzp); |
1126 | zfs_znode_update_vfs(zp); | |
34dc7c2f | 1127 | |
ea7e86d8 | 1128 | if (delete_now) |
657ce253 | 1129 | zrele(zp); |
ea7e86d8 | 1130 | else |
657ce253 | 1131 | zfs_zrele_async(zp); |
a966c564 K |
1132 | |
1133 | if (xzp) { | |
fc273894 | 1134 | zfs_znode_update_vfs(xzp); |
657ce253 | 1135 | zfs_zrele_async(xzp); |
a966c564 | 1136 | } |
428870ff | 1137 | |
0037b49e | 1138 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1139 | zil_commit(zilog, 0); |
34dc7c2f | 1140 | |
768eaced | 1141 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1142 | return (error); |
1143 | } | |
1144 | ||
1145 | /* | |
657ce253 | 1146 | * Create a new directory and insert it into dzp using the name |
34dc7c2f BB |
1147 | * provided. Return a pointer to the inserted directory. |
1148 | * | |
657ce253 | 1149 | * IN: dzp - znode of directory to add subdir to. |
34dc7c2f BB |
1150 | * dirname - name of new directory. |
1151 | * vap - attributes of new directory. | |
1152 | * cr - credentials of caller. | |
841a7a98 | 1153 | * flags - case flags. |
34dc7c2f | 1154 | * vsecp - ACL to be set |
2a068a13 | 1155 | * mnt_ns - user namespace of the mount |
34dc7c2f | 1156 | * |
657ce253 | 1157 | * OUT: zpp - znode of created directory. |
34dc7c2f BB |
1158 | * |
1159 | * RETURN: 0 if success | |
1160 | * error code if failure | |
1161 | * | |
1162 | * Timestamps: | |
657ce253 MM |
1163 | * dzp - ctime|mtime updated |
1164 | * zpp - ctime|mtime|atime updated | |
34dc7c2f | 1165 | */ |
e5c39b95 | 1166 | int |
657ce253 | 1167 | zfs_mkdir(znode_t *dzp, char *dirname, vattr_t *vap, znode_t **zpp, |
2a068a13 | 1168 | cred_t *cr, int flags, vsecattr_t *vsecp, zuserns_t *mnt_ns) |
34dc7c2f | 1169 | { |
657ce253 MM |
1170 | znode_t *zp; |
1171 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); | |
34dc7c2f BB |
1172 | zilog_t *zilog; |
1173 | zfs_dirlock_t *dl; | |
1174 | uint64_t txtype; | |
1175 | dmu_tx_t *tx; | |
1176 | int error; | |
34dc7c2f | 1177 | int zf = ZNEW; |
b128c09f BB |
1178 | uid_t uid; |
1179 | gid_t gid = crgetgid(cr); | |
428870ff | 1180 | zfs_acl_ids_t acl_ids; |
9babb374 | 1181 | boolean_t fuid_dirtied; |
e8b96c60 | 1182 | boolean_t waited = B_FALSE; |
34dc7c2f | 1183 | |
3558fd73 | 1184 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1185 | |
1186 | /* | |
1187 | * If we have an ephemeral id, ACL, or XVATTR then | |
1188 | * make sure file system is at proper version | |
1189 | */ | |
1190 | ||
3558fd73 | 1191 | uid = crgetuid(cr); |
0037b49e | 1192 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1193 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1194 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1195 | |
32dec7bd | 1196 | if (dirname == NULL) |
1197 | return (SET_ERROR(EINVAL)); | |
1198 | ||
768eaced CC |
1199 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
1200 | return (error); | |
0037b49e | 1201 | zilog = zfsvfs->z_log; |
34dc7c2f | 1202 | |
428870ff | 1203 | if (dzp->z_pflags & ZFS_XATTR) { |
768eaced | 1204 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1205 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1206 | } |
1207 | ||
0037b49e | 1208 | if (zfsvfs->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1209 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 1210 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1211 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1212 | } |
1213 | if (flags & FIGNORECASE) | |
1214 | zf |= ZCILOOK; | |
1215 | ||
5484965a | 1216 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1217 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1218 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
768eaced | 1219 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1220 | return (error); |
1221 | } | |
428870ff | 1222 | } |
34dc7c2f | 1223 | |
428870ff | 1224 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
2a068a13 | 1225 | vsecp, &acl_ids, mnt_ns)) != 0) { |
768eaced | 1226 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
1227 | return (error); |
1228 | } | |
34dc7c2f BB |
1229 | /* |
1230 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1231 | * |
1232 | * Existence is checked first to make sure we don't return | |
1233 | * EACCES instead of EEXIST which can cause some applications | |
1234 | * to fail. | |
34dc7c2f BB |
1235 | */ |
1236 | top: | |
657ce253 | 1237 | *zpp = NULL; |
34dc7c2f | 1238 | |
149e873a BB |
1239 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1240 | NULL, NULL))) { | |
428870ff | 1241 | zfs_acl_ids_free(&acl_ids); |
768eaced | 1242 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1243 | return (error); |
1244 | } | |
1245 | ||
2a068a13 YY |
1246 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr, |
1247 | mnt_ns))) { | |
428870ff | 1248 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1249 | zfs_dirent_unlock(dl); |
768eaced | 1250 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1251 | return (error); |
1252 | } | |
1253 | ||
9c5167d1 | 1254 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) { |
45d1cae3 | 1255 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 1256 | zfs_dirent_unlock(dl); |
768eaced | 1257 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1258 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
1259 | } |
1260 | ||
34dc7c2f BB |
1261 | /* |
1262 | * Add a new entry to the directory. | |
1263 | */ | |
0037b49e | 1264 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f BB |
1265 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
1266 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
0037b49e | 1267 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1268 | if (fuid_dirtied) |
0037b49e BB |
1269 | zfs_fuid_txhold(zfsvfs, tx); |
1270 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
1271 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
1272 | acl_ids.z_aclp->z_acl_bytes); | |
1273 | } | |
1274 | ||
1275 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1276 | ZFS_SA_BASE_ATTR_SIZE); | |
1277 | ||
0735ecb3 | 1278 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
1279 | if (error) { |
1280 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1281 | if (error == ERESTART) { |
e8b96c60 | 1282 | waited = B_TRUE; |
34dc7c2f BB |
1283 | dmu_tx_wait(tx); |
1284 | dmu_tx_abort(tx); | |
1285 | goto top; | |
1286 | } | |
428870ff | 1287 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1288 | dmu_tx_abort(tx); |
768eaced | 1289 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1290 | return (error); |
1291 | } | |
1292 | ||
1293 | /* | |
1294 | * Create new node. | |
1295 | */ | |
428870ff | 1296 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f BB |
1297 | |
1298 | /* | |
1299 | * Now put new name in parent dir. | |
1300 | */ | |
599b8648 CC |
1301 | error = zfs_link_create(dl, zp, tx, ZNEW); |
1302 | if (error != 0) { | |
1303 | zfs_znode_delete(zp, tx); | |
1304 | remove_inode_hash(ZTOI(zp)); | |
1305 | goto out; | |
1306 | } | |
1307 | ||
1308 | if (fuid_dirtied) | |
1309 | zfs_fuid_sync(zfsvfs, tx); | |
34dc7c2f | 1310 | |
657ce253 | 1311 | *zpp = zp; |
34dc7c2f BB |
1312 | |
1313 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
1314 | if (flags & FIGNORECASE) | |
1315 | txtype |= TX_CI; | |
9babb374 BB |
1316 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
1317 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 1318 | |
599b8648 | 1319 | out: |
9babb374 | 1320 | zfs_acl_ids_free(&acl_ids); |
428870ff | 1321 | |
34dc7c2f BB |
1322 | dmu_tx_commit(tx); |
1323 | ||
1324 | zfs_dirent_unlock(dl); | |
1325 | ||
0037b49e | 1326 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1327 | zil_commit(zilog, 0); |
428870ff | 1328 | |
599b8648 | 1329 | if (error != 0) { |
657ce253 | 1330 | zrele(zp); |
599b8648 | 1331 | } else { |
fc273894 KHN |
1332 | zfs_znode_update_vfs(dzp); |
1333 | zfs_znode_update_vfs(zp); | |
599b8648 | 1334 | } |
768eaced | 1335 | zfs_exit(zfsvfs, FTAG); |
599b8648 | 1336 | return (error); |
34dc7c2f BB |
1337 | } |
1338 | ||
1339 | /* | |
1340 | * Remove a directory subdir entry. If the current working | |
1341 | * directory is the same as the subdir to be removed, the | |
1342 | * remove will fail. | |
1343 | * | |
657ce253 | 1344 | * IN: dzp - znode of directory to remove from. |
34dc7c2f | 1345 | * name - name of directory to be removed. |
3558fd73 | 1346 | * cwd - inode of current working directory. |
34dc7c2f | 1347 | * cr - credentials of caller. |
34dc7c2f BB |
1348 | * flags - case flags |
1349 | * | |
d3cc8b15 | 1350 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1351 | * |
1352 | * Timestamps: | |
657ce253 | 1353 | * dzp - ctime|mtime updated |
34dc7c2f | 1354 | */ |
e5c39b95 | 1355 | int |
657ce253 | 1356 | zfs_rmdir(znode_t *dzp, char *name, znode_t *cwd, cred_t *cr, |
3558fd73 | 1357 | int flags) |
34dc7c2f | 1358 | { |
34dc7c2f | 1359 | znode_t *zp; |
657ce253 | 1360 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f BB |
1361 | zilog_t *zilog; |
1362 | zfs_dirlock_t *dl; | |
1363 | dmu_tx_t *tx; | |
1364 | int error; | |
1365 | int zflg = ZEXISTS; | |
e8b96c60 | 1366 | boolean_t waited = B_FALSE; |
34dc7c2f | 1367 | |
32dec7bd | 1368 | if (name == NULL) |
1369 | return (SET_ERROR(EINVAL)); | |
1370 | ||
768eaced CC |
1371 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
1372 | return (error); | |
0037b49e | 1373 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1374 | |
1375 | if (flags & FIGNORECASE) | |
1376 | zflg |= ZCILOOK; | |
1377 | top: | |
1378 | zp = NULL; | |
1379 | ||
1380 | /* | |
1381 | * Attempt to lock directory; fail if entry doesn't exist. | |
1382 | */ | |
149e873a BB |
1383 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1384 | NULL, NULL))) { | |
768eaced | 1385 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1386 | return (error); |
1387 | } | |
1388 | ||
2a068a13 | 1389 | if ((error = zfs_zaccess_delete(dzp, zp, cr, NULL))) { |
34dc7c2f BB |
1390 | goto out; |
1391 | } | |
1392 | ||
657ce253 | 1393 | if (!S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1394 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
1395 | goto out; |
1396 | } | |
1397 | ||
657ce253 | 1398 | if (zp == cwd) { |
2e528b49 | 1399 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1400 | goto out; |
1401 | } | |
1402 | ||
34dc7c2f | 1403 | /* |
4e33ba4c | 1404 | * Grab a lock on the directory to make sure that no one is |
34dc7c2f BB |
1405 | * trying to add (or lookup) entries while we are removing it. |
1406 | */ | |
1407 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
1408 | ||
1409 | /* | |
1410 | * Grab a lock on the parent pointer to make sure we play well | |
1411 | * with the treewalk and directory rename code. | |
1412 | */ | |
1413 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
1414 | ||
0037b49e | 1415 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 1416 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 1417 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
0037b49e | 1418 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
1419 | zfs_sa_upgrade_txholds(tx, zp); |
1420 | zfs_sa_upgrade_txholds(tx, dzp); | |
db707ad0 | 1421 | dmu_tx_mark_netfree(tx); |
0735ecb3 | 1422 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
1423 | if (error) { |
1424 | rw_exit(&zp->z_parent_lock); | |
1425 | rw_exit(&zp->z_name_lock); | |
1426 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1427 | if (error == ERESTART) { |
e8b96c60 | 1428 | waited = B_TRUE; |
34dc7c2f BB |
1429 | dmu_tx_wait(tx); |
1430 | dmu_tx_abort(tx); | |
657ce253 | 1431 | zrele(zp); |
34dc7c2f BB |
1432 | goto top; |
1433 | } | |
1434 | dmu_tx_abort(tx); | |
657ce253 | 1435 | zrele(zp); |
768eaced | 1436 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1437 | return (error); |
1438 | } | |
1439 | ||
1440 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
1441 | ||
1442 | if (error == 0) { | |
1443 | uint64_t txtype = TX_RMDIR; | |
1444 | if (flags & FIGNORECASE) | |
1445 | txtype |= TX_CI; | |
8e556c5e CC |
1446 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT, |
1447 | B_FALSE); | |
34dc7c2f BB |
1448 | } |
1449 | ||
1450 | dmu_tx_commit(tx); | |
1451 | ||
1452 | rw_exit(&zp->z_parent_lock); | |
1453 | rw_exit(&zp->z_name_lock); | |
1454 | out: | |
1455 | zfs_dirent_unlock(dl); | |
1456 | ||
fc273894 KHN |
1457 | zfs_znode_update_vfs(dzp); |
1458 | zfs_znode_update_vfs(zp); | |
657ce253 | 1459 | zrele(zp); |
34dc7c2f | 1460 | |
0037b49e | 1461 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1462 | zil_commit(zilog, 0); |
428870ff | 1463 | |
768eaced | 1464 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1465 | return (error); |
1466 | } | |
1467 | ||
1468 | /* | |
841a7a98 TK |
1469 | * Read directory entries from the given directory cursor position and emit |
1470 | * name and position for each entry. | |
34dc7c2f | 1471 | * |
3558fd73 | 1472 | * IN: ip - inode of directory to read. |
841a7a98 TK |
1473 | * ctx - directory entry context. |
1474 | * cr - credentials of caller. | |
34dc7c2f BB |
1475 | * |
1476 | * RETURN: 0 if success | |
1477 | * error code if failure | |
1478 | * | |
1479 | * Timestamps: | |
3558fd73 | 1480 | * ip - atime updated |
34dc7c2f BB |
1481 | * |
1482 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
1483 | * This allows us to use the low range for "special" directory entries: | |
1484 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
1485 | * we use the offset 2 for the '.zfs' directory. | |
1486 | */ | |
3558fd73 | 1487 | int |
9464b959 | 1488 | zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) |
34dc7c2f | 1489 | { |
ef70eff1 | 1490 | (void) cr; |
3558fd73 | 1491 | znode_t *zp = ITOZ(ip); |
0037b49e | 1492 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 1493 | objset_t *os; |
34dc7c2f BB |
1494 | zap_cursor_t zc; |
1495 | zap_attribute_t zap; | |
34dc7c2f BB |
1496 | int error; |
1497 | uint8_t prefetch; | |
c12e3a59 | 1498 | uint8_t type; |
3558fd73 BB |
1499 | int done = 0; |
1500 | uint64_t parent; | |
c12e3a59 | 1501 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 1502 | |
768eaced CC |
1503 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
1504 | return (error); | |
34dc7c2f | 1505 | |
0037b49e | 1506 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
3558fd73 BB |
1507 | &parent, sizeof (parent))) != 0) |
1508 | goto out; | |
34dc7c2f BB |
1509 | |
1510 | /* | |
1511 | * Quit if directory has been removed (posix) | |
1512 | */ | |
3558fd73 BB |
1513 | if (zp->z_unlinked) |
1514 | goto out; | |
1515 | ||
c12e3a59 | 1516 | error = 0; |
0037b49e | 1517 | os = zfsvfs->z_os; |
c12e3a59 | 1518 | offset = ctx->pos; |
34dc7c2f BB |
1519 | prefetch = zp->z_zn_prefetch; |
1520 | ||
1521 | /* | |
1522 | * Initialize the iterator cursor. | |
1523 | */ | |
c12e3a59 | 1524 | if (offset <= 3) { |
34dc7c2f BB |
1525 | /* |
1526 | * Start iteration from the beginning of the directory. | |
1527 | */ | |
1528 | zap_cursor_init(&zc, os, zp->z_id); | |
1529 | } else { | |
1530 | /* | |
1531 | * The offset is a serialized cursor. | |
1532 | */ | |
c12e3a59 | 1533 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
1534 | } |
1535 | ||
34dc7c2f BB |
1536 | /* |
1537 | * Transform to file-system independent format | |
1538 | */ | |
3558fd73 BB |
1539 | while (!done) { |
1540 | uint64_t objnum; | |
34dc7c2f BB |
1541 | /* |
1542 | * Special case `.', `..', and `.zfs'. | |
1543 | */ | |
c12e3a59 | 1544 | if (offset == 0) { |
34dc7c2f BB |
1545 | (void) strcpy(zap.za_name, "."); |
1546 | zap.za_normalization_conflict = 0; | |
1547 | objnum = zp->z_id; | |
c12e3a59 RY |
1548 | type = DT_DIR; |
1549 | } else if (offset == 1) { | |
34dc7c2f BB |
1550 | (void) strcpy(zap.za_name, ".."); |
1551 | zap.za_normalization_conflict = 0; | |
428870ff | 1552 | objnum = parent; |
c12e3a59 RY |
1553 | type = DT_DIR; |
1554 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
1555 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
1556 | zap.za_normalization_conflict = 0; | |
1557 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 1558 | type = DT_DIR; |
34dc7c2f BB |
1559 | } else { |
1560 | /* | |
1561 | * Grab next entry. | |
1562 | */ | |
3558fd73 BB |
1563 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
1564 | if (error == ENOENT) | |
34dc7c2f BB |
1565 | break; |
1566 | else | |
1567 | goto update; | |
1568 | } | |
1569 | ||
0c5dde49 BB |
1570 | /* |
1571 | * Allow multiple entries provided the first entry is | |
1572 | * the object id. Non-zpl consumers may safely make | |
1573 | * use of the additional space. | |
1574 | * | |
1575 | * XXX: This should be a feature flag for compatibility | |
1576 | */ | |
34dc7c2f | 1577 | if (zap.za_integer_length != 8 || |
0c5dde49 | 1578 | zap.za_num_integers == 0) { |
34dc7c2f | 1579 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
1580 | "entry, obj = %lld, offset = %lld, " |
1581 | "length = %d, num = %lld\n", | |
34dc7c2f | 1582 | (u_longlong_t)zp->z_id, |
c12e3a59 | 1583 | (u_longlong_t)offset, |
0c5dde49 BB |
1584 | zap.za_integer_length, |
1585 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 1586 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
1587 | goto update; |
1588 | } | |
1589 | ||
1590 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 1591 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 1592 | } |
0f37d0c8 | 1593 | |
9464b959 | 1594 | done = !zpl_dir_emit(ctx, zap.za_name, strlen(zap.za_name), |
c12e3a59 | 1595 | objnum, type); |
0f37d0c8 | 1596 | if (done) |
34dc7c2f | 1597 | break; |
34dc7c2f BB |
1598 | |
1599 | /* Prefetch znode */ | |
3558fd73 | 1600 | if (prefetch) { |
fcff0f35 PD |
1601 | dmu_prefetch(os, objnum, 0, 0, 0, |
1602 | ZIO_PRIORITY_SYNC_READ); | |
3558fd73 | 1603 | } |
34dc7c2f | 1604 | |
c12e3a59 RY |
1605 | /* |
1606 | * Move to the next entry, fill in the previous offset. | |
1607 | */ | |
1608 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 1609 | zap_cursor_advance(&zc); |
c12e3a59 | 1610 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 1611 | } else { |
c12e3a59 | 1612 | offset += 1; |
34dc7c2f | 1613 | } |
c12e3a59 | 1614 | ctx->pos = offset; |
34dc7c2f BB |
1615 | } |
1616 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
1617 | ||
34dc7c2f BB |
1618 | update: |
1619 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
1620 | if (error == ENOENT) |
1621 | error = 0; | |
3558fd73 | 1622 | out: |
768eaced | 1623 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f | 1624 | |
34dc7c2f BB |
1625 | return (error); |
1626 | } | |
1627 | ||
057e8eee BB |
1628 | /* |
1629 | * Get the basic file attributes and place them in the provided kstat | |
1630 | * structure. The inode is assumed to be the authoritative source | |
1631 | * for most of the attributes. However, the znode currently has the | |
1632 | * authoritative atime, blksize, and block count. | |
1633 | * | |
1634 | * IN: ip - inode of file. | |
1635 | * | |
1636 | * OUT: sp - kstat values. | |
1637 | * | |
1638 | * RETURN: 0 (always succeeds) | |
1639 | */ | |
057e8eee | 1640 | int |
e2a82961 CK |
1641 | zfs_getattr_fast(struct user_namespace *user_ns, struct inode *ip, |
1642 | struct kstat *sp) | |
057e8eee BB |
1643 | { |
1644 | znode_t *zp = ITOZ(ip); | |
0037b49e | 1645 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
b585bc4a BB |
1646 | uint32_t blksize; |
1647 | u_longlong_t nblocks; | |
768eaced | 1648 | int error; |
057e8eee | 1649 | |
768eaced CC |
1650 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
1651 | return (error); | |
a7b125e9 | 1652 | |
057e8eee BB |
1653 | mutex_enter(&zp->z_lock); |
1654 | ||
e2a82961 | 1655 | zpl_generic_fillattr(user_ns, ip, sp); |
97aa3ba4 TK |
1656 | /* |
1657 | * +1 link count for root inode with visible '.zfs' directory. | |
1658 | */ | |
1659 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) | |
1660 | if (sp->nlink < ZFS_LINK_MAX) | |
1661 | sp->nlink++; | |
057e8eee | 1662 | |
b585bc4a BB |
1663 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
1664 | sp->blksize = blksize; | |
1665 | sp->blocks = nblocks; | |
1666 | ||
057e8eee BB |
1667 | if (unlikely(zp->z_blksz == 0)) { |
1668 | /* | |
1669 | * Block size hasn't been set; suggest maximal I/O transfers. | |
1670 | */ | |
0037b49e | 1671 | sp->blksize = zfsvfs->z_max_blksz; |
057e8eee BB |
1672 | } |
1673 | ||
1674 | mutex_exit(&zp->z_lock); | |
1675 | ||
aa9b2708 AV |
1676 | /* |
1677 | * Required to prevent NFS client from detecting different inode | |
1678 | * numbers of snapshot root dentry before and after snapshot mount. | |
1679 | */ | |
0037b49e | 1680 | if (zfsvfs->z_issnap) { |
aa9b2708 AV |
1681 | if (ip->i_sb->s_root->d_inode == ip) |
1682 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
0037b49e | 1683 | dmu_objset_id(zfsvfs->z_os); |
aa9b2708 AV |
1684 | } |
1685 | ||
768eaced | 1686 | zfs_exit(zfsvfs, FTAG); |
a7b125e9 | 1687 | |
057e8eee BB |
1688 | return (0); |
1689 | } | |
057e8eee | 1690 | |
9c5167d1 NF |
1691 | /* |
1692 | * For the operation of changing file's user/group/project, we need to | |
1693 | * handle not only the main object that is assigned to the file directly, | |
1694 | * but also the ones that are used by the file via hidden xattr directory. | |
1695 | * | |
1696 | * Because the xattr directory may contains many EA entries, as to it may | |
1697 | * be impossible to change all of them via the transaction of changing the | |
1698 | * main object's user/group/project attributes. Then we have to change them | |
1699 | * via other multiple independent transactions one by one. It may be not good | |
1700 | * solution, but we have no better idea yet. | |
1701 | */ | |
1702 | static int | |
1703 | zfs_setattr_dir(znode_t *dzp) | |
1704 | { | |
1705 | struct inode *dxip = ZTOI(dzp); | |
1706 | struct inode *xip = NULL; | |
657ce253 | 1707 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
9c5167d1 NF |
1708 | objset_t *os = zfsvfs->z_os; |
1709 | zap_cursor_t zc; | |
1710 | zap_attribute_t zap; | |
1711 | zfs_dirlock_t *dl; | |
b7ab7ae2 | 1712 | znode_t *zp = NULL; |
9c5167d1 NF |
1713 | dmu_tx_t *tx = NULL; |
1714 | uint64_t uid, gid; | |
1715 | sa_bulk_attr_t bulk[4]; | |
8cb34421 | 1716 | int count; |
9c5167d1 NF |
1717 | int err; |
1718 | ||
1719 | zap_cursor_init(&zc, os, dzp->z_id); | |
1720 | while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) { | |
8cb34421 | 1721 | count = 0; |
9c5167d1 NF |
1722 | if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { |
1723 | err = ENXIO; | |
1724 | break; | |
1725 | } | |
1726 | ||
1727 | err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp, | |
1728 | ZEXISTS, NULL, NULL); | |
1729 | if (err == ENOENT) | |
1730 | goto next; | |
1731 | if (err) | |
1732 | break; | |
1733 | ||
1734 | xip = ZTOI(zp); | |
1735 | if (KUID_TO_SUID(xip->i_uid) == KUID_TO_SUID(dxip->i_uid) && | |
1736 | KGID_TO_SGID(xip->i_gid) == KGID_TO_SGID(dxip->i_gid) && | |
1737 | zp->z_projid == dzp->z_projid) | |
1738 | goto next; | |
1739 | ||
1740 | tx = dmu_tx_create(os); | |
1741 | if (!(zp->z_pflags & ZFS_PROJID)) | |
1742 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
1743 | else | |
1744 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
1745 | ||
1746 | err = dmu_tx_assign(tx, TXG_WAIT); | |
1747 | if (err) | |
1748 | break; | |
1749 | ||
1750 | mutex_enter(&dzp->z_lock); | |
1751 | ||
1752 | if (KUID_TO_SUID(xip->i_uid) != KUID_TO_SUID(dxip->i_uid)) { | |
1753 | xip->i_uid = dxip->i_uid; | |
1754 | uid = zfs_uid_read(dxip); | |
1755 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, | |
1756 | &uid, sizeof (uid)); | |
1757 | } | |
1758 | ||
1759 | if (KGID_TO_SGID(xip->i_gid) != KGID_TO_SGID(dxip->i_gid)) { | |
1760 | xip->i_gid = dxip->i_gid; | |
1761 | gid = zfs_gid_read(dxip); | |
1762 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, | |
1763 | &gid, sizeof (gid)); | |
1764 | } | |
1765 | ||
1766 | if (zp->z_projid != dzp->z_projid) { | |
1767 | if (!(zp->z_pflags & ZFS_PROJID)) { | |
1768 | zp->z_pflags |= ZFS_PROJID; | |
1769 | SA_ADD_BULK_ATTR(bulk, count, | |
1770 | SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, | |
1771 | sizeof (zp->z_pflags)); | |
1772 | } | |
1773 | ||
1774 | zp->z_projid = dzp->z_projid; | |
1775 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs), | |
1776 | NULL, &zp->z_projid, sizeof (zp->z_projid)); | |
1777 | } | |
1778 | ||
1779 | mutex_exit(&dzp->z_lock); | |
1780 | ||
1781 | if (likely(count > 0)) { | |
1782 | err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
1783 | dmu_tx_commit(tx); | |
1784 | } else { | |
1785 | dmu_tx_abort(tx); | |
1786 | } | |
1787 | tx = NULL; | |
1788 | if (err != 0 && err != ENOENT) | |
1789 | break; | |
1790 | ||
1791 | next: | |
657ce253 MM |
1792 | if (zp) { |
1793 | zrele(zp); | |
1794 | zp = NULL; | |
9c5167d1 NF |
1795 | zfs_dirent_unlock(dl); |
1796 | } | |
1797 | zap_cursor_advance(&zc); | |
1798 | } | |
1799 | ||
1800 | if (tx) | |
1801 | dmu_tx_abort(tx); | |
657ce253 MM |
1802 | if (zp) { |
1803 | zrele(zp); | |
9c5167d1 NF |
1804 | zfs_dirent_unlock(dl); |
1805 | } | |
1806 | zap_cursor_fini(&zc); | |
1807 | ||
1808 | return (err == ENOENT ? 0 : err); | |
1809 | } | |
1810 | ||
34dc7c2f BB |
1811 | /* |
1812 | * Set the file attributes to the values contained in the | |
1813 | * vattr structure. | |
1814 | * | |
657ce253 | 1815 | * IN: zp - znode of file to be modified. |
34dc7c2f | 1816 | * vap - new attribute values. |
5484965a | 1817 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
1818 | * flags - ATTR_UTIME set if non-default time values provided. |
1819 | * - ATTR_NOACLCHECK (CIFS context only). | |
1820 | * cr - credentials of caller. | |
2a068a13 | 1821 | * mnt_ns - user namespace of the mount |
34dc7c2f BB |
1822 | * |
1823 | * RETURN: 0 if success | |
1824 | * error code if failure | |
1825 | * | |
1826 | * Timestamps: | |
3558fd73 | 1827 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f | 1828 | */ |
e5c39b95 | 1829 | int |
2a068a13 | 1830 | zfs_setattr(znode_t *zp, vattr_t *vap, int flags, cred_t *cr, zuserns_t *mnt_ns) |
34dc7c2f | 1831 | { |
657ce253 MM |
1832 | struct inode *ip; |
1833 | zfsvfs_t *zfsvfs = ZTOZSB(zp); | |
9c5167d1 | 1834 | objset_t *os = zfsvfs->z_os; |
34dc7c2f BB |
1835 | zilog_t *zilog; |
1836 | dmu_tx_t *tx; | |
1837 | vattr_t oldva; | |
f4ea75d4 | 1838 | xvattr_t *tmpxvattr; |
5484965a | 1839 | uint_t mask = vap->va_mask; |
a117a6d6 | 1840 | uint_t saved_mask = 0; |
34dc7c2f BB |
1841 | int trim_mask = 0; |
1842 | uint64_t new_mode; | |
64aefee1 | 1843 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 1844 | uint64_t xattr_obj; |
0df9673f | 1845 | uint64_t mtime[2], ctime[2], atime[2]; |
9c5167d1 | 1846 | uint64_t projid = ZFS_INVALID_PROJID; |
34dc7c2f BB |
1847 | znode_t *attrzp; |
1848 | int need_policy = FALSE; | |
9c5167d1 | 1849 | int err, err2 = 0; |
34dc7c2f | 1850 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
1851 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
1852 | xoptattr_t *xoap; | |
1853 | zfs_acl_t *aclp; | |
34dc7c2f | 1854 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 1855 | boolean_t fuid_dirtied = B_FALSE; |
9c5167d1 | 1856 | boolean_t handle_eadir = B_FALSE; |
17c37660 | 1857 | sa_bulk_attr_t *bulk, *xattr_bulk; |
9c5167d1 | 1858 | int count = 0, xattr_count = 0, bulks = 8; |
34dc7c2f BB |
1859 | |
1860 | if (mask == 0) | |
1861 | return (0); | |
1862 | ||
768eaced CC |
1863 | if ((err = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
1864 | return (err); | |
657ce253 | 1865 | ip = ZTOI(zp); |
34dc7c2f | 1866 | |
9c5167d1 NF |
1867 | /* |
1868 | * If this is a xvattr_t, then get a pointer to the structure of | |
1869 | * optional attributes. If this is NULL, then we have a vattr_t. | |
1870 | */ | |
1871 | xoap = xva_getxoptattr(xvap); | |
1872 | if (xoap != NULL && (mask & ATTR_XVATTR)) { | |
1873 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
1874 | if (!dmu_objset_projectquota_enabled(os) || | |
1875 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode))) { | |
768eaced | 1876 | zfs_exit(zfsvfs, FTAG); |
9c5167d1 NF |
1877 | return (SET_ERROR(ENOTSUP)); |
1878 | } | |
1879 | ||
1880 | projid = xoap->xoa_projid; | |
1881 | if (unlikely(projid == ZFS_INVALID_PROJID)) { | |
768eaced | 1882 | zfs_exit(zfsvfs, FTAG); |
9c5167d1 NF |
1883 | return (SET_ERROR(EINVAL)); |
1884 | } | |
1885 | ||
1886 | if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID) | |
1887 | projid = ZFS_INVALID_PROJID; | |
1888 | else | |
1889 | need_policy = TRUE; | |
1890 | } | |
1891 | ||
1892 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) && | |
2705ebf0 NF |
1893 | (xoap->xoa_projinherit != |
1894 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) && | |
9c5167d1 NF |
1895 | (!dmu_objset_projectquota_enabled(os) || |
1896 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode)))) { | |
768eaced | 1897 | zfs_exit(zfsvfs, FTAG); |
2705ebf0 | 1898 | return (SET_ERROR(ENOTSUP)); |
9c5167d1 NF |
1899 | } |
1900 | } | |
1901 | ||
0037b49e | 1902 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1903 | |
1904 | /* | |
1905 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
1906 | * that file system is at proper version level | |
1907 | */ | |
5484965a | 1908 | |
0037b49e | 1909 | if (zfsvfs->z_use_fuids == B_FALSE && |
5484965a BB |
1910 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
1911 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
1912 | (mask & ATTR_XVATTR))) { | |
768eaced | 1913 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1914 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1915 | } |
1916 | ||
3558fd73 | 1917 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
768eaced | 1918 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1919 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
1920 | } |
1921 | ||
3558fd73 | 1922 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
768eaced | 1923 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 1924 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1925 | } |
1926 | ||
d1d7e268 | 1927 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 1928 | xva_init(tmpxvattr); |
5484965a | 1929 | |
9c5167d1 NF |
1930 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); |
1931 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); | |
17c37660 | 1932 | |
5484965a BB |
1933 | /* |
1934 | * Immutable files can only alter immutable bit and atime | |
1935 | */ | |
1936 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
1937 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
1938 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
ecb2b7dc | 1939 | err = SET_ERROR(EPERM); |
f4ea75d4 | 1940 | goto out3; |
5484965a BB |
1941 | } |
1942 | ||
3558fd73 | 1943 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
ecb2b7dc | 1944 | err = SET_ERROR(EPERM); |
f4ea75d4 | 1945 | goto out3; |
34dc7c2f BB |
1946 | } |
1947 | ||
5484965a BB |
1948 | /* |
1949 | * Verify timestamps doesn't overflow 32 bits. | |
1950 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
1951 | * handle times greater than 2039. This check should be removed | |
1952 | * once large timestamps are fully supported. | |
1953 | */ | |
1954 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
1955 | if (((mask & ATTR_ATIME) && |
1956 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
1957 | ((mask & ATTR_MTIME) && | |
1958 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
ecb2b7dc | 1959 | err = SET_ERROR(EOVERFLOW); |
f4ea75d4 | 1960 | goto out3; |
5484965a BB |
1961 | } |
1962 | } | |
1963 | ||
34dc7c2f BB |
1964 | top: |
1965 | attrzp = NULL; | |
572e2857 | 1966 | aclp = NULL; |
34dc7c2f | 1967 | |
45d1cae3 | 1968 | /* Can this be moved to before the top label? */ |
0037b49e | 1969 | if (zfs_is_readonly(zfsvfs)) { |
ecb2b7dc | 1970 | err = SET_ERROR(EROFS); |
f4ea75d4 | 1971 | goto out3; |
34dc7c2f BB |
1972 | } |
1973 | ||
1974 | /* | |
1975 | * First validate permissions | |
1976 | */ | |
1977 | ||
3558fd73 | 1978 | if (mask & ATTR_SIZE) { |
2a068a13 YY |
1979 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr, |
1980 | mnt_ns); | |
f4ea75d4 BB |
1981 | if (err) |
1982 | goto out3; | |
1983 | ||
34dc7c2f BB |
1984 | /* |
1985 | * XXX - Note, we are not providing any open | |
1986 | * mode flags here (like FNDELAY), so we may | |
1987 | * block if there are locks present... this | |
1988 | * should be addressed in openat(). | |
1989 | */ | |
b128c09f | 1990 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 1991 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
1992 | if (err) |
1993 | goto out3; | |
428870ff | 1994 | } |
34dc7c2f | 1995 | |
5484965a BB |
1996 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
1997 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
1998 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
1999 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2000 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2001 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2002 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2003 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2004 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2a068a13 | 2005 | skipaclchk, cr, mnt_ns); |
5484965a BB |
2006 | } |
2007 | ||
3558fd73 BB |
2008 | if (mask & (ATTR_UID|ATTR_GID)) { |
2009 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2010 | int take_owner; |
2011 | int take_group; | |
2a068a13 YY |
2012 | uid_t uid; |
2013 | gid_t gid; | |
34dc7c2f BB |
2014 | |
2015 | /* | |
2016 | * NOTE: even if a new mode is being set, | |
2017 | * we may clear S_ISUID/S_ISGID bits. | |
2018 | */ | |
2019 | ||
3558fd73 | 2020 | if (!(mask & ATTR_MODE)) |
5484965a | 2021 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2022 | |
2023 | /* | |
2024 | * Take ownership or chgrp to group we are a member of | |
2025 | */ | |
2026 | ||
2a068a13 YY |
2027 | uid = zfs_uid_into_mnt((struct user_namespace *)mnt_ns, |
2028 | vap->va_uid); | |
2029 | gid = zfs_gid_into_mnt((struct user_namespace *)mnt_ns, | |
2030 | vap->va_gid); | |
2031 | take_owner = (mask & ATTR_UID) && (uid == crgetuid(cr)); | |
3558fd73 | 2032 | take_group = (mask & ATTR_GID) && |
2a068a13 | 2033 | zfs_groupmember(zfsvfs, gid, cr); |
34dc7c2f BB |
2034 | |
2035 | /* | |
5484965a | 2036 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
2037 | * take_group must both be set in order to allow taking |
2038 | * ownership. | |
2039 | * | |
2040 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2041 | * | |
2042 | */ | |
2043 | ||
3558fd73 BB |
2044 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
2045 | take_owner && take_group) || | |
2046 | ((idmask == ATTR_UID) && take_owner) || | |
2047 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f | 2048 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
2a068a13 | 2049 | skipaclchk, cr, mnt_ns) == 0) { |
34dc7c2f BB |
2050 | /* |
2051 | * Remove setuid/setgid for non-privileged users | |
2052 | */ | |
5484965a | 2053 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 2054 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
2055 | } else { |
2056 | need_policy = TRUE; | |
2057 | } | |
2058 | } else { | |
2059 | need_policy = TRUE; | |
2060 | } | |
2061 | } | |
2062 | ||
2063 | mutex_enter(&zp->z_lock); | |
428870ff | 2064 | oldva.va_mode = zp->z_mode; |
572e2857 | 2065 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
2066 | if (mask & ATTR_XVATTR) { |
2067 | /* | |
2068 | * Update xvattr mask to include only those attributes | |
2069 | * that are actually changing. | |
2070 | * | |
2071 | * the bits will be restored prior to actually setting | |
2072 | * the attributes so the caller thinks they were set. | |
2073 | */ | |
2074 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2075 | if (xoap->xoa_appendonly != | |
2076 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
2077 | need_policy = TRUE; | |
2078 | } else { | |
2079 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 2080 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
2081 | } |
2082 | } | |
2083 | ||
9c5167d1 NF |
2084 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { |
2085 | if (xoap->xoa_projinherit != | |
2086 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) { | |
2087 | need_policy = TRUE; | |
2088 | } else { | |
2089 | XVA_CLR_REQ(xvap, XAT_PROJINHERIT); | |
2090 | XVA_SET_REQ(tmpxvattr, XAT_PROJINHERIT); | |
2091 | } | |
2092 | } | |
2093 | ||
5484965a BB |
2094 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { |
2095 | if (xoap->xoa_nounlink != | |
2096 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
2097 | need_policy = TRUE; | |
2098 | } else { | |
2099 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 2100 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
2101 | } |
2102 | } | |
2103 | ||
2104 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2105 | if (xoap->xoa_immutable != | |
2106 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
2107 | need_policy = TRUE; | |
2108 | } else { | |
2109 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 2110 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
2111 | } |
2112 | } | |
2113 | ||
2114 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2115 | if (xoap->xoa_nodump != | |
2116 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
2117 | need_policy = TRUE; | |
2118 | } else { | |
2119 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 2120 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
2121 | } |
2122 | } | |
2123 | ||
2124 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2125 | if (xoap->xoa_av_modified != | |
2126 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
2127 | need_policy = TRUE; | |
2128 | } else { | |
2129 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 2130 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
2131 | } |
2132 | } | |
2133 | ||
2134 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2135 | if ((!S_ISREG(ip->i_mode) && | |
2136 | xoap->xoa_av_quarantined) || | |
2137 | xoap->xoa_av_quarantined != | |
2138 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
2139 | need_policy = TRUE; | |
2140 | } else { | |
2141 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 2142 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
2143 | } |
2144 | } | |
2145 | ||
2146 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2147 | mutex_exit(&zp->z_lock); | |
ecb2b7dc | 2148 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2149 | goto out3; |
5484965a BB |
2150 | } |
2151 | ||
2152 | if (need_policy == FALSE && | |
2153 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2154 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
2155 | need_policy = TRUE; | |
2156 | } | |
2157 | } | |
34dc7c2f BB |
2158 | |
2159 | mutex_exit(&zp->z_lock); | |
2160 | ||
3558fd73 | 2161 | if (mask & ATTR_MODE) { |
2a068a13 YY |
2162 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr, |
2163 | mnt_ns) == 0) { | |
5484965a | 2164 | err = secpolicy_setid_setsticky_clear(ip, vap, |
2a068a13 | 2165 | &oldva, cr, mnt_ns); |
f4ea75d4 BB |
2166 | if (err) |
2167 | goto out3; | |
3558fd73 | 2168 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
2169 | } else { |
2170 | need_policy = TRUE; | |
2171 | } | |
2172 | } | |
2173 | ||
2174 | if (need_policy) { | |
2175 | /* | |
2176 | * If trim_mask is set then take ownership | |
2177 | * has been granted or write_acl is present and user | |
2178 | * has the ability to modify mode. In that case remove | |
2179 | * UID|GID and or MODE from mask so that | |
2180 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2181 | */ | |
2182 | ||
2183 | if (trim_mask) { | |
5484965a BB |
2184 | saved_mask = vap->va_mask; |
2185 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 2186 | } |
5484965a | 2187 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 2188 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
2189 | if (err) |
2190 | goto out3; | |
34dc7c2f BB |
2191 | |
2192 | if (trim_mask) | |
5484965a | 2193 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
2194 | } |
2195 | ||
2196 | /* | |
2197 | * secpolicy_vnode_setattr, or take ownership may have | |
2198 | * changed va_mask | |
2199 | */ | |
5484965a | 2200 | mask = vap->va_mask; |
34dc7c2f | 2201 | |
9c5167d1 NF |
2202 | if ((mask & (ATTR_UID | ATTR_GID)) || projid != ZFS_INVALID_PROJID) { |
2203 | handle_eadir = B_TRUE; | |
0037b49e | 2204 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
572e2857 | 2205 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 2206 | |
572e2857 | 2207 | if (err == 0 && xattr_obj) { |
3558fd73 | 2208 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
2209 | if (err) |
2210 | goto out2; | |
2211 | } | |
3558fd73 | 2212 | if (mask & ATTR_UID) { |
0037b49e | 2213 | new_kuid = zfs_fuid_create(zfsvfs, |
5484965a | 2214 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
64aefee1 | 2215 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
9c5167d1 NF |
2216 | zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT, |
2217 | new_kuid)) { | |
572e2857 | 2218 | if (attrzp) |
657ce253 | 2219 | zrele(attrzp); |
ecb2b7dc | 2220 | err = SET_ERROR(EDQUOT); |
428870ff BB |
2221 | goto out2; |
2222 | } | |
2223 | } | |
2224 | ||
3558fd73 | 2225 | if (mask & ATTR_GID) { |
0037b49e BB |
2226 | new_kgid = zfs_fuid_create(zfsvfs, |
2227 | (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); | |
64aefee1 | 2228 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && |
9c5167d1 NF |
2229 | zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT, |
2230 | new_kgid)) { | |
572e2857 | 2231 | if (attrzp) |
657ce253 | 2232 | zrele(attrzp); |
ecb2b7dc | 2233 | err = SET_ERROR(EDQUOT); |
428870ff BB |
2234 | goto out2; |
2235 | } | |
2236 | } | |
9c5167d1 NF |
2237 | |
2238 | if (projid != ZFS_INVALID_PROJID && | |
2239 | zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) { | |
2240 | if (attrzp) | |
657ce253 | 2241 | zrele(attrzp); |
9c5167d1 NF |
2242 | err = EDQUOT; |
2243 | goto out2; | |
2244 | } | |
428870ff | 2245 | } |
9c5167d1 | 2246 | tx = dmu_tx_create(os); |
34dc7c2f | 2247 | |
3558fd73 | 2248 | if (mask & ATTR_MODE) { |
428870ff | 2249 | uint64_t pmode = zp->z_mode; |
572e2857 | 2250 | uint64_t acl_obj; |
5484965a | 2251 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 2252 | |
7bf3e1fa PH |
2253 | if (ZTOZSB(zp)->z_acl_mode == ZFS_ACL_RESTRICTED && |
2254 | !(zp->z_pflags & ZFS_ACL_TRIVIAL)) { | |
2255 | err = EPERM; | |
2256 | goto out; | |
2257 | } | |
2258 | ||
156f74fc MM |
2259 | if ((err = zfs_acl_chmod_setattr(zp, &aclp, new_mode))) |
2260 | goto out; | |
428870ff | 2261 | |
572e2857 BB |
2262 | mutex_enter(&zp->z_lock); |
2263 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
2264 | /* |
2265 | * Are we upgrading ACL from old V0 format | |
2266 | * to V1 format? | |
2267 | */ | |
0037b49e | 2268 | if (zfsvfs->z_version >= ZPL_VERSION_FUID && |
572e2857 | 2269 | zfs_znode_acl_version(zp) == |
34dc7c2f | 2270 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 2271 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
2272 | DMU_OBJECT_END); |
2273 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
2274 | 0, aclp->z_acl_bytes); | |
2275 | } else { | |
572e2857 | 2276 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
2277 | aclp->z_acl_bytes); |
2278 | } | |
428870ff | 2279 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
2280 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
2281 | 0, aclp->z_acl_bytes); | |
2282 | } | |
572e2857 | 2283 | mutex_exit(&zp->z_lock); |
428870ff BB |
2284 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
2285 | } else { | |
9c5167d1 NF |
2286 | if (((mask & ATTR_XVATTR) && |
2287 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) || | |
2288 | (projid != ZFS_INVALID_PROJID && | |
2289 | !(zp->z_pflags & ZFS_PROJID))) | |
5484965a BB |
2290 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
2291 | else | |
2292 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2293 | } |
2294 | ||
428870ff BB |
2295 | if (attrzp) { |
2296 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2297 | } |
2298 | ||
0037b49e | 2299 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
428870ff | 2300 | if (fuid_dirtied) |
0037b49e | 2301 | zfs_fuid_txhold(zfsvfs, tx); |
428870ff BB |
2302 | |
2303 | zfs_sa_upgrade_txholds(tx, zp); | |
2304 | ||
384f8a09 MA |
2305 | err = dmu_tx_assign(tx, TXG_WAIT); |
2306 | if (err) | |
9babb374 | 2307 | goto out; |
34dc7c2f | 2308 | |
428870ff | 2309 | count = 0; |
34dc7c2f BB |
2310 | /* |
2311 | * Set each attribute requested. | |
2312 | * We group settings according to the locks they need to acquire. | |
2313 | * | |
2314 | * Note: you cannot set ctime directly, although it will be | |
2315 | * updated as a side-effect of calling this function. | |
2316 | */ | |
2317 | ||
9c5167d1 NF |
2318 | if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) { |
2319 | /* | |
2320 | * For the existed object that is upgraded from old system, | |
2321 | * its on-disk layout has no slot for the project ID attribute. | |
2322 | * But quota accounting logic needs to access related slots by | |
2323 | * offset directly. So we need to adjust old objects' layout | |
2324 | * to make the project ID to some unified and fixed offset. | |
2325 | */ | |
2326 | if (attrzp) | |
2327 | err = sa_add_projid(attrzp->z_sa_hdl, tx, projid); | |
2328 | if (err == 0) | |
2329 | err = sa_add_projid(zp->z_sa_hdl, tx, projid); | |
2330 | ||
2331 | if (unlikely(err == EEXIST)) | |
2332 | err = 0; | |
2333 | else if (err != 0) | |
2334 | goto out; | |
2335 | else | |
2336 | projid = ZFS_INVALID_PROJID; | |
2337 | } | |
572e2857 | 2338 | |
3558fd73 | 2339 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2340 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
2341 | mutex_enter(&zp->z_lock); |
2342 | ||
0037b49e | 2343 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
2344 | &zp->z_pflags, sizeof (zp->z_pflags)); |
2345 | ||
2346 | if (attrzp) { | |
3558fd73 | 2347 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2348 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
2349 | mutex_enter(&attrzp->z_lock); |
2350 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 2351 | SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, |
428870ff | 2352 | sizeof (attrzp->z_pflags)); |
9c5167d1 NF |
2353 | if (projid != ZFS_INVALID_PROJID) { |
2354 | attrzp->z_projid = projid; | |
2355 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
2356 | SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid, | |
2357 | sizeof (attrzp->z_projid)); | |
2358 | } | |
428870ff BB |
2359 | } |
2360 | ||
3558fd73 | 2361 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 2362 | |
3558fd73 | 2363 | if (mask & ATTR_UID) { |
64aefee1 NB |
2364 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
2365 | new_uid = zfs_uid_read(ZTOI(zp)); | |
0037b49e | 2366 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 2367 | &new_uid, sizeof (new_uid)); |
428870ff BB |
2368 | if (attrzp) { |
2369 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 2370 | SA_ZPL_UID(zfsvfs), NULL, &new_uid, |
428870ff | 2371 | sizeof (new_uid)); |
2c6abf15 | 2372 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
2373 | } |
2374 | } | |
2375 | ||
3558fd73 | 2376 | if (mask & ATTR_GID) { |
64aefee1 NB |
2377 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
2378 | new_gid = zfs_gid_read(ZTOI(zp)); | |
0037b49e | 2379 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), |
428870ff | 2380 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
2381 | if (attrzp) { |
2382 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 2383 | SA_ZPL_GID(zfsvfs), NULL, &new_gid, |
428870ff | 2384 | sizeof (new_gid)); |
64aefee1 | 2385 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
2386 | } |
2387 | } | |
3558fd73 | 2388 | if (!(mask & ATTR_MODE)) { |
0037b49e | 2389 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), |
428870ff BB |
2390 | NULL, &new_mode, sizeof (new_mode)); |
2391 | new_mode = zp->z_mode; | |
2392 | } | |
2393 | err = zfs_acl_chown_setattr(zp); | |
2394 | ASSERT(err == 0); | |
2395 | if (attrzp) { | |
2396 | err = zfs_acl_chown_setattr(attrzp); | |
2397 | ASSERT(err == 0); | |
2398 | } | |
2399 | } | |
2400 | ||
3558fd73 | 2401 | if (mask & ATTR_MODE) { |
0037b49e | 2402 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 2403 | &new_mode, sizeof (new_mode)); |
12fa7f34 | 2404 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 2405 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 2406 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 2407 | ASSERT0(err); |
572e2857 BB |
2408 | if (zp->z_acl_cached) |
2409 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
2410 | zp->z_acl_cached = aclp; |
2411 | aclp = NULL; | |
34dc7c2f BB |
2412 | } |
2413 | ||
704cd075 | 2414 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
a43570c5 | 2415 | zp->z_atime_dirty = B_FALSE; |
704cd075 | 2416 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
0037b49e | 2417 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
0df9673f | 2418 | &atime, sizeof (atime)); |
34dc7c2f BB |
2419 | } |
2420 | ||
99834d19 | 2421 | if (mask & (ATTR_MTIME | ATTR_SIZE)) { |
5484965a | 2422 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
795699a6 BB |
2423 | ZTOI(zp)->i_mtime = zpl_inode_timestamp_truncate( |
2424 | vap->va_mtime, ZTOI(zp)); | |
87f9371a | 2425 | |
0037b49e | 2426 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 2427 | mtime, sizeof (mtime)); |
34dc7c2f BB |
2428 | } |
2429 | ||
99834d19 | 2430 | if (mask & (ATTR_CTIME | ATTR_SIZE)) { |
87f9371a | 2431 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); |
795699a6 BB |
2432 | ZTOI(zp)->i_ctime = zpl_inode_timestamp_truncate(vap->va_ctime, |
2433 | ZTOI(zp)); | |
0037b49e | 2434 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
87f9371a | 2435 | ctime, sizeof (ctime)); |
428870ff | 2436 | } |
87f9371a | 2437 | |
9c5167d1 NF |
2438 | if (projid != ZFS_INVALID_PROJID) { |
2439 | zp->z_projid = projid; | |
2440 | SA_ADD_BULK_ATTR(bulk, count, | |
2441 | SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid, | |
2442 | sizeof (zp->z_projid)); | |
2443 | } | |
2444 | ||
87f9371a NB |
2445 | if (attrzp && mask) { |
2446 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 2447 | SA_ZPL_CTIME(zfsvfs), NULL, &ctime, |
87f9371a NB |
2448 | sizeof (ctime)); |
2449 | } | |
2450 | ||
34dc7c2f BB |
2451 | /* |
2452 | * Do this after setting timestamps to prevent timestamp | |
2453 | * update from toggling bit | |
2454 | */ | |
2455 | ||
5484965a BB |
2456 | if (xoap && (mask & ATTR_XVATTR)) { |
2457 | ||
2458 | /* | |
2459 | * restore trimmed off masks | |
2460 | * so that return masks can be set for caller. | |
2461 | */ | |
2462 | ||
f4ea75d4 | 2463 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
2464 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
2465 | } | |
f4ea75d4 | 2466 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
2467 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
2468 | } | |
f4ea75d4 | 2469 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
2470 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
2471 | } | |
f4ea75d4 | 2472 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
2473 | XVA_SET_REQ(xvap, XAT_NODUMP); |
2474 | } | |
f4ea75d4 | 2475 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
2476 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
2477 | } | |
f4ea75d4 | 2478 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
2479 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
2480 | } | |
9c5167d1 NF |
2481 | if (XVA_ISSET_REQ(tmpxvattr, XAT_PROJINHERIT)) { |
2482 | XVA_SET_REQ(xvap, XAT_PROJINHERIT); | |
2483 | } | |
5484965a BB |
2484 | |
2485 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
2486 | ASSERT(S_ISREG(ip->i_mode)); | |
2487 | ||
2488 | zfs_xvattr_set(zp, xvap, tx); | |
2489 | } | |
2490 | ||
9babb374 | 2491 | if (fuid_dirtied) |
0037b49e | 2492 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 2493 | |
34dc7c2f | 2494 | if (mask != 0) |
5484965a | 2495 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 2496 | |
34dc7c2f | 2497 | mutex_exit(&zp->z_lock); |
3558fd73 | 2498 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2499 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 2500 | |
572e2857 | 2501 | if (attrzp) { |
3558fd73 | 2502 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
2503 | mutex_exit(&attrzp->z_acl_lock); |
2504 | mutex_exit(&attrzp->z_lock); | |
2505 | } | |
9babb374 | 2506 | out: |
9c5167d1 | 2507 | if (err == 0 && xattr_count > 0) { |
428870ff BB |
2508 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, |
2509 | xattr_count, tx); | |
2510 | ASSERT(err2 == 0); | |
2511 | } | |
2512 | ||
45d1cae3 | 2513 | if (aclp) |
9babb374 | 2514 | zfs_acl_free(aclp); |
9babb374 BB |
2515 | |
2516 | if (fuidp) { | |
2517 | zfs_fuid_info_free(fuidp); | |
2518 | fuidp = NULL; | |
2519 | } | |
2520 | ||
428870ff | 2521 | if (err) { |
9babb374 | 2522 | dmu_tx_abort(tx); |
ea7e86d8 | 2523 | if (attrzp) |
657ce253 | 2524 | zrele(attrzp); |
428870ff BB |
2525 | if (err == ERESTART) |
2526 | goto top; | |
2527 | } else { | |
9c5167d1 NF |
2528 | if (count > 0) |
2529 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 2530 | dmu_tx_commit(tx); |
9c5167d1 NF |
2531 | if (attrzp) { |
2532 | if (err2 == 0 && handle_eadir) | |
2533 | err2 = zfs_setattr_dir(attrzp); | |
657ce253 | 2534 | zrele(attrzp); |
9c5167d1 | 2535 | } |
fc273894 | 2536 | zfs_znode_update_vfs(zp); |
428870ff BB |
2537 | } |
2538 | ||
428870ff | 2539 | out2: |
9c5167d1 | 2540 | if (os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2541 | zil_commit(zilog, 0); |
34dc7c2f | 2542 | |
f4ea75d4 | 2543 | out3: |
9c5167d1 NF |
2544 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks); |
2545 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * bulks); | |
d1d7e268 | 2546 | kmem_free(tmpxvattr, sizeof (xvattr_t)); |
768eaced | 2547 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
2548 | return (err); |
2549 | } | |
2550 | ||
2551 | typedef struct zfs_zlock { | |
2552 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
2553 | znode_t *zl_znode; /* znode we held */ | |
2554 | struct zfs_zlock *zl_next; /* next in list */ | |
2555 | } zfs_zlock_t; | |
2556 | ||
2557 | /* | |
2558 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
2559 | */ | |
2560 | static void | |
2561 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
2562 | { | |
2563 | zfs_zlock_t *zl; | |
2564 | ||
2565 | while ((zl = *zlpp) != NULL) { | |
2566 | if (zl->zl_znode != NULL) | |
657ce253 | 2567 | zfs_zrele_async(zl->zl_znode); |
34dc7c2f BB |
2568 | rw_exit(zl->zl_rwlock); |
2569 | *zlpp = zl->zl_next; | |
2570 | kmem_free(zl, sizeof (*zl)); | |
2571 | } | |
2572 | } | |
2573 | ||
2574 | /* | |
2575 | * Search back through the directory tree, using the ".." entries. | |
2576 | * Lock each directory in the chain to prevent concurrent renames. | |
2577 | * Fail any attempt to move a directory into one of its own descendants. | |
2578 | * XXX - z_parent_lock can overlap with map or grow locks | |
2579 | */ | |
2580 | static int | |
2581 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
2582 | { | |
2583 | zfs_zlock_t *zl; | |
2584 | znode_t *zp = tdzp; | |
3558fd73 | 2585 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 2586 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
2587 | krwlock_t *rwlp = &szp->z_parent_lock; |
2588 | krw_t rw = RW_WRITER; | |
2589 | ||
2590 | /* | |
2591 | * First pass write-locks szp and compares to zp->z_id. | |
2592 | * Later passes read-lock zp and compare to zp->z_parent. | |
2593 | */ | |
2594 | do { | |
2595 | if (!rw_tryenter(rwlp, rw)) { | |
2596 | /* | |
2597 | * Another thread is renaming in this path. | |
2598 | * Note that if we are a WRITER, we don't have any | |
2599 | * parent_locks held yet. | |
2600 | */ | |
2601 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
2602 | /* | |
2603 | * Drop our locks and restart | |
2604 | */ | |
2605 | zfs_rename_unlock(&zl); | |
2606 | *zlpp = NULL; | |
2607 | zp = tdzp; | |
428870ff | 2608 | oidp = zp->z_id; |
34dc7c2f BB |
2609 | rwlp = &szp->z_parent_lock; |
2610 | rw = RW_WRITER; | |
2611 | continue; | |
2612 | } else { | |
2613 | /* | |
2614 | * Wait for other thread to drop its locks | |
2615 | */ | |
2616 | rw_enter(rwlp, rw); | |
2617 | } | |
2618 | } | |
2619 | ||
2620 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
2621 | zl->zl_rwlock = rwlp; | |
2622 | zl->zl_znode = NULL; | |
2623 | zl->zl_next = *zlpp; | |
2624 | *zlpp = zl; | |
2625 | ||
428870ff | 2626 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 2627 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 2628 | |
428870ff | 2629 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
2630 | return (0); |
2631 | ||
2632 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 2633 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
2634 | if (error) |
2635 | return (error); | |
2636 | zl->zl_znode = zp; | |
2637 | } | |
3558fd73 | 2638 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 2639 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
2640 | rwlp = &zp->z_parent_lock; |
2641 | rw = RW_READER; | |
2642 | ||
2643 | } while (zp->z_id != sdzp->z_id); | |
2644 | ||
2645 | return (0); | |
2646 | } | |
2647 | ||
2648 | /* | |
2649 | * Move an entry from the provided source directory to the target | |
2650 | * directory. Change the entry name as indicated. | |
2651 | * | |
657ce253 | 2652 | * IN: sdzp - Source directory containing the "old entry". |
34dc7c2f | 2653 | * snm - Old entry name. |
657ce253 | 2654 | * tdzp - Target directory to contain the "new entry". |
34dc7c2f BB |
2655 | * tnm - New entry name. |
2656 | * cr - credentials of caller. | |
34dc7c2f | 2657 | * flags - case flags |
dbf6108b AS |
2658 | * rflags - RENAME_* flags |
2659 | * wa_vap - attributes for RENAME_WHITEOUT (must be a char 0:0). | |
2a068a13 | 2660 | * mnt_ns - user namespace of the mount |
34dc7c2f | 2661 | * |
d3cc8b15 | 2662 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
2663 | * |
2664 | * Timestamps: | |
657ce253 | 2665 | * sdzp,tdzp - ctime|mtime updated |
34dc7c2f | 2666 | */ |
e5c39b95 | 2667 | int |
657ce253 | 2668 | zfs_rename(znode_t *sdzp, char *snm, znode_t *tdzp, char *tnm, |
dbf6108b | 2669 | cred_t *cr, int flags, uint64_t rflags, vattr_t *wo_vap, zuserns_t *mnt_ns) |
34dc7c2f | 2670 | { |
657ce253 MM |
2671 | znode_t *szp, *tzp; |
2672 | zfsvfs_t *zfsvfs = ZTOZSB(sdzp); | |
34dc7c2f | 2673 | zilog_t *zilog; |
34dc7c2f BB |
2674 | zfs_dirlock_t *sdl, *tdl; |
2675 | dmu_tx_t *tx; | |
2676 | zfs_zlock_t *zl; | |
2677 | int cmp, serr, terr; | |
2678 | int error = 0; | |
2679 | int zflg = 0; | |
e8b96c60 | 2680 | boolean_t waited = B_FALSE; |
dbf6108b AS |
2681 | /* Needed for whiteout inode creation. */ |
2682 | boolean_t fuid_dirtied; | |
2683 | zfs_acl_ids_t acl_ids; | |
2684 | boolean_t have_acl = B_FALSE; | |
2685 | znode_t *wzp = NULL; | |
2686 | ||
34dc7c2f | 2687 | |
32dec7bd | 2688 | if (snm == NULL || tnm == NULL) |
2689 | return (SET_ERROR(EINVAL)); | |
2690 | ||
dbf6108b AS |
2691 | if (rflags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
2692 | return (SET_ERROR(EINVAL)); | |
2693 | ||
2694 | /* Already checked by Linux VFS, but just to make sure. */ | |
2695 | if (rflags & RENAME_EXCHANGE && | |
2696 | (rflags & (RENAME_NOREPLACE | RENAME_WHITEOUT))) | |
2697 | return (SET_ERROR(EINVAL)); | |
2698 | ||
2699 | /* | |
2700 | * Make sure we only get wo_vap iff. RENAME_WHITEOUT and that it's the | |
2701 | * right kind of vattr_t for the whiteout file. These are set | |
2702 | * internally by ZFS so should never be incorrect. | |
2703 | */ | |
2704 | VERIFY_EQUIV(rflags & RENAME_WHITEOUT, wo_vap != NULL); | |
2705 | VERIFY_IMPLY(wo_vap, wo_vap->va_mode == S_IFCHR); | |
2706 | VERIFY_IMPLY(wo_vap, wo_vap->va_rdev == makedevice(0, 0)); | |
2707 | ||
768eaced CC |
2708 | if ((error = zfs_enter_verify_zp(zfsvfs, sdzp, FTAG)) != 0) |
2709 | return (error); | |
0037b49e | 2710 | zilog = zfsvfs->z_log; |
34dc7c2f | 2711 | |
768eaced CC |
2712 | if ((error = zfs_verify_zp(tdzp)) != 0) { |
2713 | zfs_exit(zfsvfs, FTAG); | |
2714 | return (error); | |
2715 | } | |
812e91a7 MT |
2716 | |
2717 | /* | |
2718 | * We check i_sb because snapshots and the ctldir must have different | |
2719 | * super blocks. | |
2720 | */ | |
657ce253 MM |
2721 | if (ZTOI(tdzp)->i_sb != ZTOI(sdzp)->i_sb || |
2722 | zfsctl_is_node(ZTOI(tdzp))) { | |
768eaced | 2723 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 2724 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
2725 | } |
2726 | ||
0037b49e | 2727 | if (zfsvfs->z_utf8 && u8_validate(tnm, |
34dc7c2f | 2728 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 2729 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 2730 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
2731 | } |
2732 | ||
2733 | if (flags & FIGNORECASE) | |
2734 | zflg |= ZCILOOK; | |
2735 | ||
2736 | top: | |
2737 | szp = NULL; | |
2738 | tzp = NULL; | |
2739 | zl = NULL; | |
2740 | ||
2741 | /* | |
2742 | * This is to prevent the creation of links into attribute space | |
2743 | * by renaming a linked file into/outof an attribute directory. | |
2744 | * See the comment in zfs_link() for why this is considered bad. | |
2745 | */ | |
428870ff | 2746 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
768eaced | 2747 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 2748 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2749 | } |
2750 | ||
2751 | /* | |
2752 | * Lock source and target directory entries. To prevent deadlock, | |
2753 | * a lock ordering must be defined. We lock the directory with | |
2754 | * the smallest object id first, or if it's a tie, the one with | |
2755 | * the lexically first name. | |
2756 | */ | |
2757 | if (sdzp->z_id < tdzp->z_id) { | |
2758 | cmp = -1; | |
2759 | } else if (sdzp->z_id > tdzp->z_id) { | |
2760 | cmp = 1; | |
2761 | } else { | |
2762 | /* | |
2763 | * First compare the two name arguments without | |
2764 | * considering any case folding. | |
2765 | */ | |
0037b49e | 2766 | int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
2767 | |
2768 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
0037b49e | 2769 | ASSERT(error == 0 || !zfsvfs->z_utf8); |
34dc7c2f BB |
2770 | if (cmp == 0) { |
2771 | /* | |
2772 | * POSIX: "If the old argument and the new argument | |
2773 | * both refer to links to the same existing file, | |
2774 | * the rename() function shall return successfully | |
2775 | * and perform no other action." | |
2776 | */ | |
768eaced | 2777 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
2778 | return (0); |
2779 | } | |
2780 | /* | |
2781 | * If the file system is case-folding, then we may | |
2782 | * have some more checking to do. A case-folding file | |
2783 | * system is either supporting mixed case sensitivity | |
2784 | * access or is completely case-insensitive. Note | |
2785 | * that the file system is always case preserving. | |
2786 | * | |
2787 | * In mixed sensitivity mode case sensitive behavior | |
2788 | * is the default. FIGNORECASE must be used to | |
2789 | * explicitly request case insensitive behavior. | |
2790 | * | |
2791 | * If the source and target names provided differ only | |
2792 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
2793 | * we will treat this as a special case in the | |
2794 | * case-insensitive mode: as long as the source name | |
2795 | * is an exact match, we will allow this to proceed as | |
2796 | * a name-change request. | |
2797 | */ | |
0037b49e BB |
2798 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
2799 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 2800 | flags & FIGNORECASE)) && |
0037b49e | 2801 | u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
2802 | &error) == 0) { |
2803 | /* | |
2804 | * case preserving rename request, require exact | |
2805 | * name matches | |
2806 | */ | |
2807 | zflg |= ZCIEXACT; | |
2808 | zflg &= ~ZCILOOK; | |
2809 | } | |
2810 | } | |
2811 | ||
428870ff BB |
2812 | /* |
2813 | * If the source and destination directories are the same, we should | |
2814 | * grab the z_name_lock of that directory only once. | |
2815 | */ | |
2816 | if (sdzp == tdzp) { | |
2817 | zflg |= ZHAVELOCK; | |
2818 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
2819 | } | |
2820 | ||
34dc7c2f BB |
2821 | if (cmp < 0) { |
2822 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
2823 | ZEXISTS | zflg, NULL, NULL); | |
2824 | terr = zfs_dirent_lock(&tdl, | |
2825 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
2826 | } else { | |
2827 | terr = zfs_dirent_lock(&tdl, | |
2828 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
2829 | serr = zfs_dirent_lock(&sdl, | |
2830 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
2831 | NULL, NULL); | |
2832 | } | |
2833 | ||
2834 | if (serr) { | |
2835 | /* | |
2836 | * Source entry invalid or not there. | |
2837 | */ | |
2838 | if (!terr) { | |
2839 | zfs_dirent_unlock(tdl); | |
2840 | if (tzp) | |
657ce253 | 2841 | zrele(tzp); |
34dc7c2f | 2842 | } |
428870ff BB |
2843 | |
2844 | if (sdzp == tdzp) | |
2845 | rw_exit(&sdzp->z_name_lock); | |
2846 | ||
34dc7c2f BB |
2847 | if (strcmp(snm, "..") == 0) |
2848 | serr = EINVAL; | |
768eaced | 2849 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
2850 | return (serr); |
2851 | } | |
2852 | if (terr) { | |
2853 | zfs_dirent_unlock(sdl); | |
657ce253 | 2854 | zrele(szp); |
428870ff BB |
2855 | |
2856 | if (sdzp == tdzp) | |
2857 | rw_exit(&sdzp->z_name_lock); | |
2858 | ||
34dc7c2f BB |
2859 | if (strcmp(tnm, "..") == 0) |
2860 | terr = EINVAL; | |
768eaced | 2861 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
2862 | return (terr); |
2863 | } | |
2864 | ||
9c5167d1 NF |
2865 | /* |
2866 | * If we are using project inheritance, means if the directory has | |
2867 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
2868 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
2869 | * such case, we only allow renames into our tree when the project | |
2870 | * IDs are the same. | |
2871 | */ | |
2872 | if (tdzp->z_pflags & ZFS_PROJINHERIT && | |
2873 | tdzp->z_projid != szp->z_projid) { | |
2874 | error = SET_ERROR(EXDEV); | |
2875 | goto out; | |
2876 | } | |
2877 | ||
34dc7c2f BB |
2878 | /* |
2879 | * Must have write access at the source to remove the old entry | |
2880 | * and write access at the target to create the new entry. | |
2881 | * Note that if target and source are the same, this can be | |
2882 | * done in a single check. | |
2883 | */ | |
2a068a13 | 2884 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr, mnt_ns))) |
34dc7c2f BB |
2885 | goto out; |
2886 | ||
3558fd73 | 2887 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
2888 | /* |
2889 | * Check to make sure rename is valid. | |
2890 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
2891 | */ | |
149e873a | 2892 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
2893 | goto out; |
2894 | } | |
2895 | ||
2896 | /* | |
2897 | * Does target exist? | |
2898 | */ | |
2899 | if (tzp) { | |
dbf6108b AS |
2900 | if (rflags & RENAME_NOREPLACE) { |
2901 | error = SET_ERROR(EEXIST); | |
2902 | goto out; | |
2903 | } | |
34dc7c2f | 2904 | /* |
dbf6108b | 2905 | * Source and target must be the same type (unless exchanging). |
34dc7c2f | 2906 | */ |
dbf6108b AS |
2907 | if (!(rflags & RENAME_EXCHANGE)) { |
2908 | boolean_t s_is_dir = S_ISDIR(ZTOI(szp)->i_mode) != 0; | |
2909 | boolean_t t_is_dir = S_ISDIR(ZTOI(tzp)->i_mode) != 0; | |
e015d6cc | 2910 | |
dbf6108b AS |
2911 | if (s_is_dir != t_is_dir) { |
2912 | error = SET_ERROR(s_is_dir ? ENOTDIR : EISDIR); | |
2913 | goto out; | |
2914 | } | |
34dc7c2f BB |
2915 | } |
2916 | /* | |
2917 | * POSIX dictates that when the source and target | |
2918 | * entries refer to the same file object, rename | |
2919 | * must do nothing and exit without error. | |
2920 | */ | |
2921 | if (szp->z_id == tzp->z_id) { | |
2922 | error = 0; | |
2923 | goto out; | |
2924 | } | |
dbf6108b AS |
2925 | } else if (rflags & RENAME_EXCHANGE) { |
2926 | /* Target must exist for RENAME_EXCHANGE. */ | |
2927 | error = SET_ERROR(ENOENT); | |
2928 | goto out; | |
2929 | } | |
2930 | ||
2931 | /* Set up inode creation for RENAME_WHITEOUT. */ | |
2932 | if (rflags & RENAME_WHITEOUT) { | |
2933 | /* | |
2934 | * Whiteout files are not regular files or directories, so to | |
2935 | * match zfs_create() we do not inherit the project id. | |
2936 | */ | |
2937 | uint64_t wo_projid = ZFS_DEFAULT_PROJID; | |
2938 | ||
2939 | error = zfs_zaccess(sdzp, ACE_ADD_FILE, 0, B_FALSE, cr, mnt_ns); | |
2940 | if (error) | |
2941 | goto out; | |
2942 | ||
2943 | if (!have_acl) { | |
2944 | error = zfs_acl_ids_create(sdzp, 0, wo_vap, cr, NULL, | |
2945 | &acl_ids, mnt_ns); | |
2946 | if (error) | |
2947 | goto out; | |
2948 | have_acl = B_TRUE; | |
2949 | } | |
2950 | ||
2951 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, wo_projid)) { | |
2952 | error = SET_ERROR(EDQUOT); | |
2953 | goto out; | |
2954 | } | |
34dc7c2f BB |
2955 | } |
2956 | ||
0037b49e | 2957 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
2958 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
2959 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
dbf6108b AS |
2960 | dmu_tx_hold_zap(tx, sdzp->z_id, |
2961 | (rflags & RENAME_EXCHANGE) ? TRUE : FALSE, snm); | |
34dc7c2f | 2962 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); |
428870ff BB |
2963 | if (sdzp != tdzp) { |
2964 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
2965 | zfs_sa_upgrade_txholds(tx, tdzp); | |
2966 | } | |
2967 | if (tzp) { | |
2968 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
2969 | zfs_sa_upgrade_txholds(tx, tzp); | |
2970 | } | |
dbf6108b AS |
2971 | if (rflags & RENAME_WHITEOUT) { |
2972 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
2973 | ZFS_SA_BASE_ATTR_SIZE); | |
428870ff | 2974 | |
dbf6108b AS |
2975 | dmu_tx_hold_zap(tx, sdzp->z_id, TRUE, snm); |
2976 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
2977 | if (!zfsvfs->z_use_sa && | |
2978 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
2979 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
2980 | 0, acl_ids.z_aclp->z_acl_bytes); | |
2981 | } | |
2982 | } | |
2983 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
2984 | if (fuid_dirtied) | |
2985 | zfs_fuid_txhold(zfsvfs, tx); | |
428870ff | 2986 | zfs_sa_upgrade_txholds(tx, szp); |
0037b49e | 2987 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
0735ecb3 | 2988 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
2989 | if (error) { |
2990 | if (zl != NULL) | |
2991 | zfs_rename_unlock(&zl); | |
2992 | zfs_dirent_unlock(sdl); | |
2993 | zfs_dirent_unlock(tdl); | |
428870ff BB |
2994 | |
2995 | if (sdzp == tdzp) | |
2996 | rw_exit(&sdzp->z_name_lock); | |
2997 | ||
fb5f0bc8 | 2998 | if (error == ERESTART) { |
e8b96c60 | 2999 | waited = B_TRUE; |
34dc7c2f BB |
3000 | dmu_tx_wait(tx); |
3001 | dmu_tx_abort(tx); | |
657ce253 | 3002 | zrele(szp); |
ea7e86d8 | 3003 | if (tzp) |
657ce253 | 3004 | zrele(tzp); |
34dc7c2f BB |
3005 | goto top; |
3006 | } | |
3007 | dmu_tx_abort(tx); | |
657ce253 | 3008 | zrele(szp); |
ea7e86d8 | 3009 | if (tzp) |
657ce253 | 3010 | zrele(tzp); |
768eaced | 3011 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3012 | return (error); |
3013 | } | |
3014 | ||
e015d6cc AS |
3015 | /* |
3016 | * Unlink the source. | |
3017 | */ | |
3018 | szp->z_pflags |= ZFS_AV_MODIFIED; | |
3019 | if (tdzp->z_pflags & ZFS_PROJINHERIT) | |
3020 | szp->z_pflags |= ZFS_PROJINHERIT; | |
34dc7c2f | 3021 | |
e015d6cc AS |
3022 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), |
3023 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); | |
dbf6108b | 3024 | VERIFY0(error); |
34dc7c2f | 3025 | |
e015d6cc AS |
3026 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3027 | if (error) | |
3028 | goto commit; | |
3029 | ||
3030 | /* | |
3031 | * Unlink the target. | |
3032 | */ | |
3033 | if (tzp) { | |
dbf6108b AS |
3034 | int tzflg = zflg; |
3035 | ||
3036 | if (rflags & RENAME_EXCHANGE) { | |
3037 | /* This inode will be re-linked soon. */ | |
3038 | tzflg |= ZRENAMING; | |
3039 | ||
3040 | tzp->z_pflags |= ZFS_AV_MODIFIED; | |
3041 | if (sdzp->z_pflags & ZFS_PROJINHERIT) | |
3042 | tzp->z_pflags |= ZFS_PROJINHERIT; | |
3043 | ||
3044 | error = sa_update(tzp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), | |
3045 | (void *)&tzp->z_pflags, sizeof (uint64_t), tx); | |
3046 | ASSERT0(error); | |
3047 | } | |
3048 | error = zfs_link_destroy(tdl, tzp, tx, tzflg, NULL); | |
e015d6cc AS |
3049 | if (error) |
3050 | goto commit_link_szp; | |
3051 | } | |
3052 | ||
3053 | /* | |
dbf6108b AS |
3054 | * Create the new target links: |
3055 | * * We always link the target. | |
3056 | * * RENAME_EXCHANGE: Link the old target to the source. | |
3057 | * * RENAME_WHITEOUT: Create a whiteout inode in-place of the source. | |
e015d6cc AS |
3058 | */ |
3059 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3060 | if (error) { | |
3061 | /* | |
3062 | * If we have removed the existing target, a subsequent call to | |
3063 | * zfs_link_create() to add back the same entry, but with a new | |
3064 | * dnode (szp), should not fail. | |
3065 | */ | |
3066 | ASSERT3P(tzp, ==, NULL); | |
3067 | goto commit_link_tzp; | |
34dc7c2f BB |
3068 | } |
3069 | ||
dbf6108b AS |
3070 | switch (rflags & (RENAME_EXCHANGE | RENAME_WHITEOUT)) { |
3071 | case RENAME_EXCHANGE: | |
3072 | error = zfs_link_create(sdl, tzp, tx, ZRENAMING); | |
3073 | /* | |
3074 | * The same argument as zfs_link_create() failing for | |
3075 | * szp applies here, since the source directory must | |
3076 | * have had an entry we are replacing. | |
3077 | */ | |
3078 | ASSERT0(error); | |
3079 | if (error) | |
3080 | goto commit_unlink_td_szp; | |
3081 | break; | |
3082 | case RENAME_WHITEOUT: | |
3083 | zfs_mknode(sdzp, wo_vap, tx, cr, 0, &wzp, &acl_ids); | |
3084 | error = zfs_link_create(sdl, wzp, tx, ZNEW); | |
3085 | if (error) { | |
3086 | zfs_znode_delete(wzp, tx); | |
3087 | remove_inode_hash(ZTOI(wzp)); | |
3088 | goto commit_unlink_td_szp; | |
3089 | } | |
3090 | break; | |
3091 | } | |
3092 | ||
3093 | if (fuid_dirtied) | |
3094 | zfs_fuid_sync(zfsvfs, tx); | |
3095 | ||
3096 | switch (rflags & (RENAME_EXCHANGE | RENAME_WHITEOUT)) { | |
3097 | case RENAME_EXCHANGE: | |
3098 | zfs_log_rename_exchange(zilog, tx, | |
3099 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, sdl->dl_name, | |
3100 | tdzp, tdl->dl_name, szp); | |
3101 | break; | |
3102 | case RENAME_WHITEOUT: | |
3103 | zfs_log_rename_whiteout(zilog, tx, | |
3104 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, sdl->dl_name, | |
3105 | tdzp, tdl->dl_name, szp, wzp); | |
3106 | break; | |
3107 | default: | |
3108 | ASSERT0(rflags & ~RENAME_NOREPLACE); | |
3109 | zfs_log_rename(zilog, tx, (flags & FIGNORECASE ? TX_CI : 0), | |
3110 | sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp); | |
3111 | break; | |
3112 | } | |
e015d6cc AS |
3113 | |
3114 | commit: | |
34dc7c2f BB |
3115 | dmu_tx_commit(tx); |
3116 | out: | |
dbf6108b AS |
3117 | if (have_acl) |
3118 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 3119 | |
fc273894 | 3120 | zfs_znode_update_vfs(sdzp); |
428870ff BB |
3121 | if (sdzp == tdzp) |
3122 | rw_exit(&sdzp->z_name_lock); | |
3123 | ||
960e08fe | 3124 | if (sdzp != tdzp) |
fc273894 | 3125 | zfs_znode_update_vfs(tdzp); |
428870ff | 3126 | |
fc273894 | 3127 | zfs_znode_update_vfs(szp); |
657ce253 | 3128 | zrele(szp); |
dbf6108b AS |
3129 | if (wzp) { |
3130 | zfs_znode_update_vfs(wzp); | |
3131 | zrele(wzp); | |
3132 | } | |
960e08fe | 3133 | if (tzp) { |
fc273894 | 3134 | zfs_znode_update_vfs(tzp); |
657ce253 | 3135 | zrele(tzp); |
960e08fe | 3136 | } |
34dc7c2f | 3137 | |
dbf6108b AS |
3138 | if (zl != NULL) |
3139 | zfs_rename_unlock(&zl); | |
3140 | ||
3141 | zfs_dirent_unlock(sdl); | |
3142 | zfs_dirent_unlock(tdl); | |
3143 | ||
0037b49e | 3144 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3145 | zil_commit(zilog, 0); |
428870ff | 3146 | |
768eaced | 3147 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f | 3148 | return (error); |
e015d6cc AS |
3149 | |
3150 | /* | |
3151 | * Clean-up path for broken link state. | |
3152 | * | |
3153 | * At this point we are in a (very) bad state, so we need to do our | |
dbf6108b AS |
3154 | * best to correct the state. In particular, all of the nlinks are |
3155 | * wrong because we were destroying and creating links with ZRENAMING. | |
3156 | * | |
3157 | * In some form, all of these operations have to resolve the state: | |
3158 | * | |
3159 | * * link_destroy() *must* succeed. Fortunately, this is very likely | |
3160 | * since we only just created it. | |
e015d6cc | 3161 | * |
dbf6108b AS |
3162 | * * link_create()s are allowed to fail (though they shouldn't because |
3163 | * we only just unlinked them and are putting the entries back | |
3164 | * during clean-up). But if they fail, we can just forcefully drop | |
3165 | * the nlink value to (at the very least) avoid broken nlink values | |
3166 | * -- though in the case of non-empty directories we will have to | |
3167 | * panic (otherwise we'd have a leaked directory with a broken ..). | |
e015d6cc | 3168 | */ |
dbf6108b AS |
3169 | commit_unlink_td_szp: |
3170 | VERIFY0(zfs_link_destroy(tdl, szp, tx, ZRENAMING, NULL)); | |
e015d6cc AS |
3171 | commit_link_tzp: |
3172 | if (tzp) { | |
3173 | if (zfs_link_create(tdl, tzp, tx, ZRENAMING)) | |
dbf6108b | 3174 | VERIFY0(zfs_drop_nlink(tzp, tx, NULL)); |
e015d6cc AS |
3175 | } |
3176 | commit_link_szp: | |
3177 | if (zfs_link_create(sdl, szp, tx, ZRENAMING)) | |
dbf6108b | 3178 | VERIFY0(zfs_drop_nlink(szp, tx, NULL)); |
e015d6cc | 3179 | goto commit; |
34dc7c2f BB |
3180 | } |
3181 | ||
3182 | /* | |
3183 | * Insert the indicated symbolic reference entry into the directory. | |
3184 | * | |
657ce253 | 3185 | * IN: dzp - Directory to contain new symbolic link. |
841a7a98 | 3186 | * name - Name of directory entry in dip. |
34dc7c2f | 3187 | * vap - Attributes of new entry. |
841a7a98 | 3188 | * link - Name for new symlink entry. |
34dc7c2f | 3189 | * cr - credentials of caller. |
34dc7c2f | 3190 | * flags - case flags |
2a068a13 | 3191 | * mnt_ns - user namespace of the mount |
34dc7c2f | 3192 | * |
657ce253 | 3193 | * OUT: zpp - Znode for new symbolic link. |
841a7a98 | 3194 | * |
d3cc8b15 | 3195 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3196 | * |
3197 | * Timestamps: | |
3558fd73 | 3198 | * dip - ctime|mtime updated |
34dc7c2f | 3199 | */ |
e5c39b95 | 3200 | int |
657ce253 | 3201 | zfs_symlink(znode_t *dzp, char *name, vattr_t *vap, char *link, |
2a068a13 | 3202 | znode_t **zpp, cred_t *cr, int flags, zuserns_t *mnt_ns) |
34dc7c2f | 3203 | { |
657ce253 | 3204 | znode_t *zp; |
34dc7c2f BB |
3205 | zfs_dirlock_t *dl; |
3206 | dmu_tx_t *tx; | |
657ce253 | 3207 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
34dc7c2f | 3208 | zilog_t *zilog; |
428870ff | 3209 | uint64_t len = strlen(link); |
34dc7c2f BB |
3210 | int error; |
3211 | int zflg = ZNEW; | |
9babb374 BB |
3212 | zfs_acl_ids_t acl_ids; |
3213 | boolean_t fuid_dirtied; | |
428870ff | 3214 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 3215 | boolean_t waited = B_FALSE; |
34dc7c2f | 3216 | |
3558fd73 | 3217 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 3218 | |
32dec7bd | 3219 | if (name == NULL) |
3220 | return (SET_ERROR(EINVAL)); | |
3221 | ||
768eaced CC |
3222 | if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) |
3223 | return (error); | |
0037b49e | 3224 | zilog = zfsvfs->z_log; |
34dc7c2f | 3225 | |
0037b49e | 3226 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 3227 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 3228 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3229 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3230 | } |
3231 | if (flags & FIGNORECASE) | |
3232 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3233 | |
3234 | if (len > MAXPATHLEN) { | |
768eaced | 3235 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3236 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
3237 | } |
3238 | ||
428870ff | 3239 | if ((error = zfs_acl_ids_create(dzp, 0, |
2a068a13 | 3240 | vap, cr, NULL, &acl_ids, mnt_ns)) != 0) { |
768eaced | 3241 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
3242 | return (error); |
3243 | } | |
3244 | top: | |
657ce253 | 3245 | *zpp = NULL; |
3558fd73 | 3246 | |
34dc7c2f BB |
3247 | /* |
3248 | * Attempt to lock directory; fail if entry already exists. | |
3249 | */ | |
3250 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3251 | if (error) { | |
428870ff | 3252 | zfs_acl_ids_free(&acl_ids); |
768eaced | 3253 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
3254 | return (error); |
3255 | } | |
3256 | ||
2a068a13 | 3257 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr, mnt_ns))) { |
428870ff BB |
3258 | zfs_acl_ids_free(&acl_ids); |
3259 | zfs_dirent_unlock(dl); | |
768eaced | 3260 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3261 | return (error); |
3262 | } | |
3263 | ||
9c5167d1 | 3264 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, ZFS_DEFAULT_PROJID)) { |
9babb374 BB |
3265 | zfs_acl_ids_free(&acl_ids); |
3266 | zfs_dirent_unlock(dl); | |
768eaced | 3267 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3268 | return (SET_ERROR(EDQUOT)); |
9babb374 | 3269 | } |
0037b49e BB |
3270 | tx = dmu_tx_create(zfsvfs->z_os); |
3271 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
34dc7c2f | 3272 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3273 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3274 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3275 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3276 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
0037b49e | 3277 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
3278 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
3279 | acl_ids.z_aclp->z_acl_bytes); | |
3280 | } | |
9babb374 | 3281 | if (fuid_dirtied) |
0037b49e | 3282 | zfs_fuid_txhold(zfsvfs, tx); |
0735ecb3 | 3283 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
3284 | if (error) { |
3285 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3286 | if (error == ERESTART) { |
e8b96c60 | 3287 | waited = B_TRUE; |
34dc7c2f BB |
3288 | dmu_tx_wait(tx); |
3289 | dmu_tx_abort(tx); | |
3290 | goto top; | |
3291 | } | |
428870ff | 3292 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 3293 | dmu_tx_abort(tx); |
768eaced | 3294 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3295 | return (error); |
3296 | } | |
3297 | ||
34dc7c2f BB |
3298 | /* |
3299 | * Create a new object for the symlink. | |
bf169e9f | 3300 | * for version 4 ZPL datasets the symlink will be an SA attribute |
34dc7c2f | 3301 | */ |
428870ff | 3302 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3303 | |
428870ff | 3304 | if (fuid_dirtied) |
0037b49e | 3305 | zfs_fuid_sync(zfsvfs, tx); |
34dc7c2f | 3306 | |
572e2857 | 3307 | mutex_enter(&zp->z_lock); |
428870ff | 3308 | if (zp->z_is_sa) |
0037b49e | 3309 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), |
428870ff BB |
3310 | link, len, tx); |
3311 | else | |
3312 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3313 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3314 | |
428870ff | 3315 | zp->z_size = len; |
0037b49e | 3316 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 3317 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
3318 | /* |
3319 | * Insert the new object into the directory. | |
3320 | */ | |
599b8648 CC |
3321 | error = zfs_link_create(dl, zp, tx, ZNEW); |
3322 | if (error != 0) { | |
3323 | zfs_znode_delete(zp, tx); | |
3324 | remove_inode_hash(ZTOI(zp)); | |
3325 | } else { | |
3326 | if (flags & FIGNORECASE) | |
3327 | txtype |= TX_CI; | |
3328 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
4f301661 | 3329 | |
fc273894 KHN |
3330 | zfs_znode_update_vfs(dzp); |
3331 | zfs_znode_update_vfs(zp); | |
599b8648 | 3332 | } |
960e08fe | 3333 | |
9babb374 | 3334 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3335 | |
3336 | dmu_tx_commit(tx); | |
3337 | ||
3338 | zfs_dirent_unlock(dl); | |
3339 | ||
599b8648 | 3340 | if (error == 0) { |
657ce253 | 3341 | *zpp = zp; |
34dc7c2f | 3342 | |
599b8648 CC |
3343 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
3344 | zil_commit(zilog, 0); | |
3345 | } else { | |
657ce253 | 3346 | zrele(zp); |
599b8648 | 3347 | } |
428870ff | 3348 | |
768eaced | 3349 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3350 | return (error); |
3351 | } | |
3352 | ||
3353 | /* | |
3354 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 3355 | * the symbolic path referred to by ip. |
34dc7c2f | 3356 | * |
8b4f9a2d BB |
3357 | * IN: ip - inode of symbolic link |
3358 | * uio - structure to contain the link path. | |
3359 | * cr - credentials of caller. | |
34dc7c2f BB |
3360 | * |
3361 | * RETURN: 0 if success | |
3362 | * error code if failure | |
3363 | * | |
3364 | * Timestamps: | |
3558fd73 | 3365 | * ip - atime updated |
34dc7c2f | 3366 | */ |
e5c39b95 | 3367 | int |
d0cd9a5c | 3368 | zfs_readlink(struct inode *ip, zfs_uio_t *uio, cred_t *cr) |
34dc7c2f | 3369 | { |
ef70eff1 | 3370 | (void) cr; |
3558fd73 | 3371 | znode_t *zp = ITOZ(ip); |
0037b49e | 3372 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
3373 | int error; |
3374 | ||
768eaced CC |
3375 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
3376 | return (error); | |
34dc7c2f | 3377 | |
572e2857 | 3378 | mutex_enter(&zp->z_lock); |
428870ff | 3379 | if (zp->z_is_sa) |
8b4f9a2d | 3380 | error = sa_lookup_uio(zp->z_sa_hdl, |
0037b49e | 3381 | SA_ZPL_SYMLINK(zfsvfs), uio); |
428870ff | 3382 | else |
8b4f9a2d | 3383 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 3384 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3385 | |
768eaced | 3386 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3387 | return (error); |
3388 | } | |
3389 | ||
3390 | /* | |
657ce253 | 3391 | * Insert a new entry into directory tdzp referencing szp. |
34dc7c2f | 3392 | * |
657ce253 MM |
3393 | * IN: tdzp - Directory to contain new entry. |
3394 | * szp - znode of new entry. | |
34dc7c2f BB |
3395 | * name - name of new entry. |
3396 | * cr - credentials of caller. | |
841a7a98 | 3397 | * flags - case flags. |
34dc7c2f BB |
3398 | * |
3399 | * RETURN: 0 if success | |
3400 | * error code if failure | |
3401 | * | |
3402 | * Timestamps: | |
657ce253 MM |
3403 | * tdzp - ctime|mtime updated |
3404 | * szp - ctime updated | |
34dc7c2f | 3405 | */ |
e5c39b95 | 3406 | int |
657ce253 | 3407 | zfs_link(znode_t *tdzp, znode_t *szp, char *name, cred_t *cr, |
da5e151f | 3408 | int flags) |
34dc7c2f | 3409 | { |
657ce253 MM |
3410 | struct inode *sip = ZTOI(szp); |
3411 | znode_t *tzp; | |
3412 | zfsvfs_t *zfsvfs = ZTOZSB(tdzp); | |
34dc7c2f BB |
3413 | zilog_t *zilog; |
3414 | zfs_dirlock_t *dl; | |
3415 | dmu_tx_t *tx; | |
34dc7c2f BB |
3416 | int error; |
3417 | int zf = ZNEW; | |
428870ff | 3418 | uint64_t parent; |
572e2857 | 3419 | uid_t owner; |
e8b96c60 | 3420 | boolean_t waited = B_FALSE; |
ace1eae8 CC |
3421 | boolean_t is_tmpfile = 0; |
3422 | uint64_t txg; | |
3423 | #ifdef HAVE_TMPFILE | |
3424 | is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); | |
3425 | #endif | |
657ce253 | 3426 | ASSERT(S_ISDIR(ZTOI(tdzp)->i_mode)); |
34dc7c2f | 3427 | |
32dec7bd | 3428 | if (name == NULL) |
3429 | return (SET_ERROR(EINVAL)); | |
3430 | ||
768eaced CC |
3431 | if ((error = zfs_enter_verify_zp(zfsvfs, tdzp, FTAG)) != 0) |
3432 | return (error); | |
0037b49e | 3433 | zilog = zfsvfs->z_log; |
34dc7c2f | 3434 | |
428870ff BB |
3435 | /* |
3436 | * POSIX dictates that we return EPERM here. | |
3437 | * Better choices include ENOTSUP or EISDIR. | |
3438 | */ | |
3558fd73 | 3439 | if (S_ISDIR(sip->i_mode)) { |
768eaced | 3440 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3441 | return (SET_ERROR(EPERM)); |
428870ff BB |
3442 | } |
3443 | ||
768eaced CC |
3444 | if ((error = zfs_verify_zp(szp)) != 0) { |
3445 | zfs_exit(zfsvfs, FTAG); | |
3446 | return (error); | |
3447 | } | |
812e91a7 | 3448 | |
9c5167d1 NF |
3449 | /* |
3450 | * If we are using project inheritance, means if the directory has | |
3451 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
3452 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
3453 | * such case, we only allow hard link creation in our tree when the | |
3454 | * project IDs are the same. | |
3455 | */ | |
657ce253 MM |
3456 | if (tdzp->z_pflags & ZFS_PROJINHERIT && |
3457 | tdzp->z_projid != szp->z_projid) { | |
768eaced | 3458 | zfs_exit(zfsvfs, FTAG); |
9c5167d1 NF |
3459 | return (SET_ERROR(EXDEV)); |
3460 | } | |
3461 | ||
812e91a7 MT |
3462 | /* |
3463 | * We check i_sb because snapshots and the ctldir must have different | |
3464 | * super blocks. | |
3465 | */ | |
657ce253 | 3466 | if (sip->i_sb != ZTOI(tdzp)->i_sb || zfsctl_is_node(sip)) { |
768eaced | 3467 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3468 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 3469 | } |
428870ff | 3470 | |
428870ff BB |
3471 | /* Prevent links to .zfs/shares files */ |
3472 | ||
0037b49e | 3473 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff | 3474 | &parent, sizeof (uint64_t))) != 0) { |
768eaced | 3475 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
3476 | return (error); |
3477 | } | |
0037b49e | 3478 | if (parent == zfsvfs->z_shares_dir) { |
768eaced | 3479 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3480 | return (SET_ERROR(EPERM)); |
428870ff BB |
3481 | } |
3482 | ||
0037b49e | 3483 | if (zfsvfs->z_utf8 && u8_validate(name, |
34dc7c2f | 3484 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
768eaced | 3485 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3486 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3487 | } |
3488 | if (flags & FIGNORECASE) | |
3489 | zf |= ZCILOOK; | |
3490 | ||
34dc7c2f BB |
3491 | /* |
3492 | * We do not support links between attributes and non-attributes | |
3493 | * because of the potential security risk of creating links | |
3494 | * into "normal" file space in order to circumvent restrictions | |
3495 | * imposed in attribute space. | |
3496 | */ | |
657ce253 | 3497 | if ((szp->z_pflags & ZFS_XATTR) != (tdzp->z_pflags & ZFS_XATTR)) { |
768eaced | 3498 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3499 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3500 | } |
3501 | ||
0037b49e BB |
3502 | owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid), |
3503 | cr, ZFS_OWNER); | |
572e2857 | 3504 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
768eaced | 3505 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 3506 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
3507 | } |
3508 | ||
2a068a13 | 3509 | if ((error = zfs_zaccess(tdzp, ACE_ADD_FILE, 0, B_FALSE, cr, NULL))) { |
768eaced | 3510 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3511 | return (error); |
3512 | } | |
3513 | ||
428870ff | 3514 | top: |
34dc7c2f BB |
3515 | /* |
3516 | * Attempt to lock directory; fail if entry already exists. | |
3517 | */ | |
657ce253 | 3518 | error = zfs_dirent_lock(&dl, tdzp, name, &tzp, zf, NULL, NULL); |
34dc7c2f | 3519 | if (error) { |
768eaced | 3520 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3521 | return (error); |
3522 | } | |
3523 | ||
0037b49e | 3524 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 3525 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
657ce253 | 3526 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, name); |
ace1eae8 | 3527 | if (is_tmpfile) |
0037b49e | 3528 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 3529 | |
428870ff | 3530 | zfs_sa_upgrade_txholds(tx, szp); |
657ce253 | 3531 | zfs_sa_upgrade_txholds(tx, tdzp); |
0735ecb3 | 3532 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
3533 | if (error) { |
3534 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3535 | if (error == ERESTART) { |
e8b96c60 | 3536 | waited = B_TRUE; |
34dc7c2f BB |
3537 | dmu_tx_wait(tx); |
3538 | dmu_tx_abort(tx); | |
3539 | goto top; | |
3540 | } | |
3541 | dmu_tx_abort(tx); | |
768eaced | 3542 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3543 | return (error); |
3544 | } | |
ace1eae8 CC |
3545 | /* unmark z_unlinked so zfs_link_create will not reject */ |
3546 | if (is_tmpfile) | |
a43570c5 | 3547 | szp->z_unlinked = B_FALSE; |
34dc7c2f BB |
3548 | error = zfs_link_create(dl, szp, tx, 0); |
3549 | ||
3550 | if (error == 0) { | |
3551 | uint64_t txtype = TX_LINK; | |
ace1eae8 CC |
3552 | /* |
3553 | * tmpfile is created to be in z_unlinkedobj, so remove it. | |
e1cfd73f | 3554 | * Also, we don't log in ZIL, because all previous file |
ace1eae8 CC |
3555 | * operation on the tmpfile are ignored by ZIL. Instead we |
3556 | * always wait for txg to sync to make sure all previous | |
3557 | * operation are sync safe. | |
3558 | */ | |
3559 | if (is_tmpfile) { | |
0037b49e BB |
3560 | VERIFY(zap_remove_int(zfsvfs->z_os, |
3561 | zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); | |
ace1eae8 CC |
3562 | } else { |
3563 | if (flags & FIGNORECASE) | |
3564 | txtype |= TX_CI; | |
657ce253 | 3565 | zfs_log_link(zilog, tx, txtype, tdzp, szp, name); |
ace1eae8 CC |
3566 | } |
3567 | } else if (is_tmpfile) { | |
3568 | /* restore z_unlinked since when linking failed */ | |
a43570c5 | 3569 | szp->z_unlinked = B_TRUE; |
34dc7c2f | 3570 | } |
ace1eae8 | 3571 | txg = dmu_tx_get_txg(tx); |
34dc7c2f BB |
3572 | dmu_tx_commit(tx); |
3573 | ||
3574 | zfs_dirent_unlock(dl); | |
3575 | ||
0037b49e | 3576 | if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3577 | zil_commit(zilog, 0); |
428870ff | 3578 | |
d09dc598 | 3579 | if (is_tmpfile && zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) |
0037b49e | 3580 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg); |
ace1eae8 | 3581 | |
fc273894 KHN |
3582 | zfs_znode_update_vfs(tdzp); |
3583 | zfs_znode_update_vfs(szp); | |
768eaced | 3584 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
3585 | return (error); |
3586 | } | |
3587 | ||
3c0e5c0f | 3588 | static void |
411f4a01 | 3589 | zfs_putpage_sync_commit_cb(void *arg) |
3c0e5c0f BB |
3590 | { |
3591 | struct page *pp = arg; | |
3592 | ||
119a394a | 3593 | ClearPageError(pp); |
3c0e5c0f BB |
3594 | end_page_writeback(pp); |
3595 | } | |
3596 | ||
411f4a01 SN |
3597 | static void |
3598 | zfs_putpage_async_commit_cb(void *arg) | |
3599 | { | |
3600 | struct page *pp = arg; | |
3601 | znode_t *zp = ITOZ(pp->mapping->host); | |
3602 | ||
3603 | ClearPageError(pp); | |
3604 | end_page_writeback(pp); | |
3605 | atomic_dec_32(&zp->z_async_writes_cnt); | |
3606 | } | |
3607 | ||
34dc7c2f | 3608 | /* |
3c0e5c0f BB |
3609 | * Push a page out to disk, once the page is on stable storage the |
3610 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 3611 | * |
411f4a01 SN |
3612 | * IN: ip - page mapped for inode. |
3613 | * pp - page to push (page is locked) | |
3614 | * wbc - writeback control data | |
3615 | * for_sync - does the caller intend to wait synchronously for the | |
3616 | * page writeback to complete? | |
34dc7c2f BB |
3617 | * |
3618 | * RETURN: 0 if success | |
3619 | * error code if failure | |
3620 | * | |
3c0e5c0f BB |
3621 | * Timestamps: |
3622 | * ip - ctime|mtime updated | |
34dc7c2f | 3623 | */ |
3c0e5c0f | 3624 | int |
411f4a01 SN |
3625 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc, |
3626 | boolean_t for_sync) | |
34dc7c2f | 3627 | { |
3c0e5c0f | 3628 | znode_t *zp = ITOZ(ip); |
0037b49e | 3629 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
3c0e5c0f BB |
3630 | loff_t offset; |
3631 | loff_t pgoff; | |
4c837f0d | 3632 | unsigned int pglen; |
3c0e5c0f BB |
3633 | dmu_tx_t *tx; |
3634 | caddr_t va; | |
3635 | int err = 0; | |
3636 | uint64_t mtime[2], ctime[2]; | |
3637 | sa_bulk_attr_t bulk[3]; | |
3638 | int cnt = 0; | |
21a96fb6 | 3639 | struct address_space *mapping; |
3c0e5c0f | 3640 | |
768eaced CC |
3641 | if ((err = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
3642 | return (err); | |
d164b209 | 3643 | |
3c0e5c0f BB |
3644 | ASSERT(PageLocked(pp)); |
3645 | ||
d1d7e268 MK |
3646 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
3647 | offset = i_size_read(ip); /* File length in bytes */ | |
8b1899d3 BB |
3648 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
3649 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
3650 | |
3651 | /* Page is beyond end of file */ | |
3652 | if (pgoff >= offset) { | |
3653 | unlock_page(pp); | |
768eaced | 3654 | zfs_exit(zfsvfs, FTAG); |
3c0e5c0f BB |
3655 | return (0); |
3656 | } | |
3657 | ||
3658 | /* Truncate page length to end of file */ | |
3659 | if (pgoff + pglen > offset) | |
3660 | pglen = offset - pgoff; | |
3661 | ||
3662 | #if 0 | |
34dc7c2f | 3663 | /* |
3c0e5c0f BB |
3664 | * FIXME: Allow mmap writes past its quota. The correct fix |
3665 | * is to register a page_mkwrite() handler to count the page | |
3666 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 3667 | */ |
9c5167d1 NF |
3668 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
3669 | KUID_TO_SUID(ip->i_uid)) || | |
3670 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
3671 | KGID_TO_SGID(ip->i_gid)) || | |
3672 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
3673 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
3674 | zp->z_projid))) { | |
9babb374 | 3675 | err = EDQUOT; |
9babb374 | 3676 | } |
3c0e5c0f BB |
3677 | #endif |
3678 | ||
d958324f BB |
3679 | /* |
3680 | * The ordering here is critical and must adhere to the following | |
3681 | * rules in order to avoid deadlocking in either zfs_read() or | |
3682 | * zfs_free_range() due to a lock inversion. | |
3683 | * | |
3684 | * 1) The page must be unlocked prior to acquiring the range lock. | |
3685 | * This is critical because zfs_read() calls find_lock_page() | |
3686 | * which may block on the page lock while holding the range lock. | |
3687 | * | |
3688 | * 2) Before setting or clearing write back on a page the range lock | |
3689 | * must be held in order to prevent a lock inversion with the | |
3690 | * zfs_free_range() function. | |
21a96fb6 CC |
3691 | * |
3692 | * This presents a problem because upon entering this function the | |
3693 | * page lock is already held. To safely acquire the range lock the | |
3694 | * page lock must be dropped. This creates a window where another | |
3695 | * process could truncate, invalidate, dirty, or write out the page. | |
3696 | * | |
3697 | * Therefore, after successfully reacquiring the range and page locks | |
3698 | * the current page state is checked. In the common case everything | |
3699 | * will be as is expected and it can be written out. However, if | |
3700 | * the page state has changed it must be handled accordingly. | |
d958324f | 3701 | */ |
21a96fb6 CC |
3702 | mapping = pp->mapping; |
3703 | redirty_page_for_writepage(wbc, pp); | |
d958324f | 3704 | unlock_page(pp); |
21a96fb6 | 3705 | |
bd4dde8e | 3706 | zfs_locked_range_t *lr = zfs_rangelock_enter(&zp->z_rangelock, |
5d43cc9a | 3707 | pgoff, pglen, RL_WRITER); |
21a96fb6 CC |
3708 | lock_page(pp); |
3709 | ||
3710 | /* Page mapping changed or it was no longer dirty, we're done */ | |
3711 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
3712 | unlock_page(pp); | |
2cc479d0 | 3713 | zfs_rangelock_exit(lr); |
768eaced | 3714 | zfs_exit(zfsvfs, FTAG); |
21a96fb6 CC |
3715 | return (0); |
3716 | } | |
3717 | ||
3718 | /* Another process started write block if required */ | |
3719 | if (PageWriteback(pp)) { | |
3720 | unlock_page(pp); | |
2cc479d0 | 3721 | zfs_rangelock_exit(lr); |
21a96fb6 | 3722 | |
bfd5a709 | 3723 | if (wbc->sync_mode != WB_SYNC_NONE) { |
411f4a01 SN |
3724 | /* |
3725 | * Speed up any non-sync page writebacks since | |
3726 | * they may take several seconds to complete. | |
3727 | * Refer to the comment in zpl_fsync() (when | |
3728 | * HAVE_FSYNC_RANGE is defined) for details. | |
3729 | */ | |
3730 | if (atomic_load_32(&zp->z_async_writes_cnt) > 0) { | |
3731 | zil_commit(zfsvfs->z_log, zp->z_id); | |
3732 | } | |
3733 | ||
bfd5a709 | 3734 | if (PageWriteback(pp)) |
c0fb44c5 CK |
3735 | #ifdef HAVE_PAGEMAP_FOLIO_WAIT_BIT |
3736 | folio_wait_bit(page_folio(pp), PG_writeback); | |
3737 | #else | |
bfd5a709 | 3738 | wait_on_page_bit(pp, PG_writeback); |
c0fb44c5 | 3739 | #endif |
bfd5a709 | 3740 | } |
21a96fb6 | 3741 | |
768eaced | 3742 | zfs_exit(zfsvfs, FTAG); |
21a96fb6 CC |
3743 | return (0); |
3744 | } | |
3745 | ||
3746 | /* Clear the dirty flag the required locks are held */ | |
3747 | if (!clear_page_dirty_for_io(pp)) { | |
3748 | unlock_page(pp); | |
2cc479d0 | 3749 | zfs_rangelock_exit(lr); |
768eaced | 3750 | zfs_exit(zfsvfs, FTAG); |
21a96fb6 CC |
3751 | return (0); |
3752 | } | |
3753 | ||
3754 | /* | |
3755 | * Counterpart for redirty_page_for_writepage() above. This page | |
3756 | * was in fact not skipped and should not be counted as if it were. | |
3757 | */ | |
3758 | wbc->pages_skipped--; | |
411f4a01 SN |
3759 | if (!for_sync) |
3760 | atomic_inc_32(&zp->z_async_writes_cnt); | |
3c0e5c0f | 3761 | set_page_writeback(pp); |
21a96fb6 | 3762 | unlock_page(pp); |
3c0e5c0f | 3763 | |
0037b49e | 3764 | tx = dmu_tx_create(zfsvfs->z_os); |
3c0e5c0f | 3765 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
3766 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3767 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 3768 | |
fb5f0bc8 | 3769 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 3770 | if (err != 0) { |
3c0e5c0f | 3771 | if (err == ERESTART) |
34dc7c2f | 3772 | dmu_tx_wait(tx); |
3c0e5c0f | 3773 | |
34dc7c2f | 3774 | dmu_tx_abort(tx); |
7dde17e8 SP |
3775 | #ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO |
3776 | filemap_dirty_folio(page_mapping(pp), page_folio(pp)); | |
3777 | #else | |
119a394a | 3778 | __set_page_dirty_nobuffers(pp); |
7dde17e8 | 3779 | #endif |
119a394a ED |
3780 | ClearPageError(pp); |
3781 | end_page_writeback(pp); | |
411f4a01 SN |
3782 | if (!for_sync) |
3783 | atomic_dec_32(&zp->z_async_writes_cnt); | |
2cc479d0 | 3784 | zfs_rangelock_exit(lr); |
768eaced | 3785 | zfs_exit(zfsvfs, FTAG); |
3c0e5c0f | 3786 | return (err); |
34dc7c2f BB |
3787 | } |
3788 | ||
dde471ef | 3789 | va = kmap(pp); |
8b1899d3 | 3790 | ASSERT3U(pglen, <=, PAGE_SIZE); |
0037b49e | 3791 | dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 3792 | kunmap(pp); |
34dc7c2f | 3793 | |
0037b49e BB |
3794 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
3795 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
3796 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, | |
3797 | &zp->z_pflags, 8); | |
428870ff | 3798 | |
d3aa3ea9 BB |
3799 | /* Preserve the mtime and ctime provided by the inode */ |
3800 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
3801 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
a43570c5 | 3802 | zp->z_atime_dirty = B_FALSE; |
d3aa3ea9 BB |
3803 | zp->z_seq++; |
3804 | ||
3805 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
3806 | ||
0037b49e | 3807 | zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
411f4a01 SN |
3808 | for_sync ? zfs_putpage_sync_commit_cb : |
3809 | zfs_putpage_async_commit_cb, pp); | |
3810 | ||
45d1cae3 | 3811 | dmu_tx_commit(tx); |
d3aa3ea9 | 3812 | |
2cc479d0 | 3813 | zfs_rangelock_exit(lr); |
34dc7c2f | 3814 | |
119a394a ED |
3815 | if (wbc->sync_mode != WB_SYNC_NONE) { |
3816 | /* | |
3817 | * Note that this is rarely called under writepages(), because | |
3818 | * writepages() normally handles the entire commit for | |
3819 | * performance reasons. | |
3820 | */ | |
0037b49e | 3821 | zil_commit(zfsvfs->z_log, zp->z_id); |
411f4a01 SN |
3822 | } else if (!for_sync && atomic_load_32(&zp->z_sync_writes_cnt) > 0) { |
3823 | /* | |
3824 | * If the caller does not intend to wait synchronously | |
3825 | * for this page writeback to complete and there are active | |
3826 | * synchronous calls on this file, do a commit so that | |
3827 | * the latter don't accidentally end up waiting for | |
3828 | * our writeback to complete. Refer to the comment in | |
3829 | * zpl_fsync() (when HAVE_FSYNC_RANGE is defined) for details. | |
3830 | */ | |
3831 | zil_commit(zfsvfs->z_log, zp->z_id); | |
2b286136 | 3832 | } |
3c0e5c0f | 3833 | |
3819aaaf MB |
3834 | dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, pglen); |
3835 | ||
768eaced | 3836 | zfs_exit(zfsvfs, FTAG); |
3c0e5c0f | 3837 | return (err); |
34dc7c2f BB |
3838 | } |
3839 | ||
8780c539 BB |
3840 | /* |
3841 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 3842 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
3843 | */ |
3844 | int | |
3845 | zfs_dirty_inode(struct inode *ip, int flags) | |
3846 | { | |
3847 | znode_t *zp = ITOZ(ip); | |
0037b49e | 3848 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
8780c539 | 3849 | dmu_tx_t *tx; |
023699cd MM |
3850 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
3851 | sa_bulk_attr_t bulk[4]; | |
704cd075 | 3852 | int error = 0; |
8780c539 BB |
3853 | int cnt = 0; |
3854 | ||
0037b49e | 3855 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
c944be5d BB |
3856 | return (0); |
3857 | ||
768eaced CC |
3858 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
3859 | return (error); | |
8780c539 | 3860 | |
704cd075 CC |
3861 | #ifdef I_DIRTY_TIME |
3862 | /* | |
e1cfd73f | 3863 | * This is the lazytime semantic introduced in Linux 4.0 |
704cd075 CC |
3864 | * This flag will only be called from update_time when lazytime is set. |
3865 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
3866 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
3867 | * only need to dirty atime. | |
3868 | */ | |
3869 | if (flags == I_DIRTY_TIME) { | |
a43570c5 | 3870 | zp->z_atime_dirty = B_TRUE; |
704cd075 CC |
3871 | goto out; |
3872 | } | |
3873 | #endif | |
3874 | ||
0037b49e | 3875 | tx = dmu_tx_create(zfsvfs->z_os); |
8780c539 BB |
3876 | |
3877 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
3878 | zfs_sa_upgrade_txholds(tx, zp); | |
3879 | ||
3880 | error = dmu_tx_assign(tx, TXG_WAIT); | |
3881 | if (error) { | |
3882 | dmu_tx_abort(tx); | |
3883 | goto out; | |
3884 | } | |
3885 | ||
3886 | mutex_enter(&zp->z_lock); | |
a43570c5 | 3887 | zp->z_atime_dirty = B_FALSE; |
704cd075 | 3888 | |
0037b49e BB |
3889 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
3890 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
3891 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
3892 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
8780c539 | 3893 | |
023699cd | 3894 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
3895 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
3896 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
3897 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
023699cd MM |
3898 | mode = ip->i_mode; |
3899 | ||
3900 | zp->z_mode = mode; | |
8780c539 BB |
3901 | |
3902 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
3903 | mutex_exit(&zp->z_lock); | |
3904 | ||
3905 | dmu_tx_commit(tx); | |
3906 | out: | |
768eaced | 3907 | zfs_exit(zfsvfs, FTAG); |
8780c539 BB |
3908 | return (error); |
3909 | } | |
8780c539 | 3910 | |
34dc7c2f | 3911 | void |
c0d35759 | 3912 | zfs_inactive(struct inode *ip) |
34dc7c2f | 3913 | { |
c0d35759 | 3914 | znode_t *zp = ITOZ(ip); |
0037b49e | 3915 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
0df9673f | 3916 | uint64_t atime[2]; |
34dc7c2f | 3917 | int error; |
cafbd2ac | 3918 | int need_unlock = 0; |
34dc7c2f | 3919 | |
cafbd2ac | 3920 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
0037b49e | 3921 | if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) { |
cafbd2ac | 3922 | need_unlock = 1; |
0037b49e | 3923 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); |
cafbd2ac | 3924 | } |
c0d35759 | 3925 | if (zp->z_sa_hdl == NULL) { |
cafbd2ac | 3926 | if (need_unlock) |
0037b49e | 3927 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
c0d35759 | 3928 | return; |
34dc7c2f BB |
3929 | } |
3930 | ||
a43570c5 | 3931 | if (zp->z_atime_dirty && zp->z_unlinked == B_FALSE) { |
0037b49e | 3932 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 3933 | |
428870ff BB |
3934 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3935 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
3936 | error = dmu_tx_assign(tx, TXG_WAIT); |
3937 | if (error) { | |
3938 | dmu_tx_abort(tx); | |
3939 | } else { | |
0df9673f | 3940 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
34dc7c2f | 3941 | mutex_enter(&zp->z_lock); |
0037b49e | 3942 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), |
0df9673f | 3943 | (void *)&atime, sizeof (atime), tx); |
a43570c5 | 3944 | zp->z_atime_dirty = B_FALSE; |
34dc7c2f BB |
3945 | mutex_exit(&zp->z_lock); |
3946 | dmu_tx_commit(tx); | |
3947 | } | |
3948 | } | |
3949 | ||
3950 | zfs_zinactive(zp); | |
cafbd2ac | 3951 | if (need_unlock) |
0037b49e | 3952 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
34dc7c2f BB |
3953 | } |
3954 | ||
34dc7c2f | 3955 | /* |
dde471ef | 3956 | * Fill pages with data from the disk. |
34dc7c2f BB |
3957 | */ |
3958 | static int | |
dde471ef | 3959 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 3960 | { |
d1d7e268 | 3961 | znode_t *zp = ITOZ(ip); |
0037b49e | 3962 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
d1d7e268 | 3963 | objset_t *os; |
dde471ef | 3964 | struct page *cur_pp; |
d1d7e268 MK |
3965 | u_offset_t io_off, total; |
3966 | size_t io_len; | |
3967 | loff_t i_size; | |
3968 | unsigned page_idx; | |
3969 | int err; | |
34dc7c2f | 3970 | |
0037b49e | 3971 | os = zfsvfs->z_os; |
8b1899d3 | 3972 | io_len = nr_pages << PAGE_SHIFT; |
dde471ef PJ |
3973 | i_size = i_size_read(ip); |
3974 | io_off = page_offset(pl[0]); | |
3975 | ||
3976 | if (io_off + io_len > i_size) | |
3977 | io_len = i_size - io_off; | |
34dc7c2f BB |
3978 | |
3979 | /* | |
dde471ef | 3980 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 3981 | */ |
dde471ef | 3982 | page_idx = 0; |
34dc7c2f | 3983 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
3984 | caddr_t va; |
3985 | ||
540c3927 | 3986 | cur_pp = pl[page_idx++]; |
dde471ef | 3987 | va = kmap(cur_pp); |
9babb374 BB |
3988 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
3989 | DMU_READ_PREFETCH); | |
dde471ef | 3990 | kunmap(cur_pp); |
34dc7c2f | 3991 | if (err) { |
b128c09f BB |
3992 | /* convert checksum errors into IO errors */ |
3993 | if (err == ECKSUM) | |
2e528b49 | 3994 | err = SET_ERROR(EIO); |
34dc7c2f BB |
3995 | return (err); |
3996 | } | |
34dc7c2f | 3997 | } |
d164b209 | 3998 | |
34dc7c2f BB |
3999 | return (0); |
4000 | } | |
4001 | ||
4002 | /* | |
dde471ef | 4003 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4004 | * |
dde471ef PJ |
4005 | * IN: ip - inode of file to get data from. |
4006 | * pl - list of pages to read | |
4007 | * nr_pages - number of pages to read | |
34dc7c2f | 4008 | * |
d3cc8b15 | 4009 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4010 | * |
4011 | * Timestamps: | |
4012 | * vp - atime updated | |
4013 | */ | |
dde471ef PJ |
4014 | int |
4015 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4016 | { |
dde471ef | 4017 | znode_t *zp = ITOZ(ip); |
0037b49e | 4018 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
dde471ef | 4019 | int err; |
d164b209 | 4020 | |
d164b209 BB |
4021 | if (pl == NULL) |
4022 | return (0); | |
34dc7c2f | 4023 | |
768eaced CC |
4024 | if ((err = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
4025 | return (err); | |
34dc7c2f | 4026 | |
dde471ef | 4027 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4028 | |
3819aaaf MB |
4029 | dataset_kstats_update_read_kstats(&zfsvfs->z_kstat, nr_pages*PAGESIZE); |
4030 | ||
768eaced | 4031 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
4032 | return (err); |
4033 | } | |
4034 | ||
4035 | /* | |
e2e7aa2d | 4036 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4037 | * |
e2e7aa2d BB |
4038 | * IN: ip - inode of the file to mmap |
4039 | * off - file offset | |
4040 | * addrp - start address in memory region | |
4041 | * len - length of memory region | |
4042 | * vm_flags- address flags | |
34dc7c2f | 4043 | * |
e2e7aa2d BB |
4044 | * RETURN: 0 if success |
4045 | * error code if failure | |
34dc7c2f | 4046 | */ |
e2e7aa2d BB |
4047 | int |
4048 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4049 | unsigned long vm_flags) | |
34dc7c2f | 4050 | { |
ef70eff1 | 4051 | (void) addrp; |
e2e7aa2d | 4052 | znode_t *zp = ITOZ(ip); |
0037b49e | 4053 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
768eaced | 4054 | int error; |
34dc7c2f | 4055 | |
768eaced CC |
4056 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
4057 | return (error); | |
34dc7c2f | 4058 | |
e2e7aa2d | 4059 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4060 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
768eaced | 4061 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4062 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4063 | } |
4064 | ||
e2e7aa2d | 4065 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4066 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
768eaced | 4067 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4068 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4069 | } |
4070 | ||
34dc7c2f | 4071 | if (off < 0 || len > MAXOFFSET_T - off) { |
768eaced | 4072 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4073 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4074 | } |
4075 | ||
768eaced | 4076 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
4077 | return (0); |
4078 | } | |
4079 | ||
4080 | /* | |
4081 | * Free or allocate space in a file. Currently, this function only | |
4082 | * supports the `F_FREESP' command. However, this command is somewhat | |
4083 | * misnamed, as its functionality includes the ability to allocate as | |
4084 | * well as free space. | |
4085 | * | |
657ce253 | 4086 | * IN: zp - znode of file to free data in. |
34dc7c2f BB |
4087 | * cmd - action to take (only F_FREESP supported). |
4088 | * bfp - section of file to free/alloc. | |
4089 | * flag - current file open mode flags. | |
4090 | * offset - current file offset. | |
a35c1207 | 4091 | * cr - credentials of caller. |
34dc7c2f | 4092 | * |
d3cc8b15 | 4093 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4094 | * |
4095 | * Timestamps: | |
657ce253 | 4096 | * zp - ctime|mtime updated |
34dc7c2f | 4097 | */ |
e5c39b95 | 4098 | int |
657ce253 | 4099 | zfs_space(znode_t *zp, int cmd, flock64_t *bfp, int flag, |
3558fd73 | 4100 | offset_t offset, cred_t *cr) |
34dc7c2f | 4101 | { |
ef70eff1 | 4102 | (void) offset; |
657ce253 | 4103 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
4104 | uint64_t off, len; |
4105 | int error; | |
4106 | ||
768eaced CC |
4107 | if ((error = zfs_enter_verify_zp(zfsvfs, zp, FTAG)) != 0) |
4108 | return (error); | |
34dc7c2f | 4109 | |
34dc7c2f | 4110 | if (cmd != F_FREESP) { |
768eaced | 4111 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4112 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4113 | } |
4114 | ||
f3c9dca0 MT |
4115 | /* |
4116 | * Callers might not be able to detect properly that we are read-only, | |
4117 | * so check it explicitly here. | |
4118 | */ | |
0037b49e | 4119 | if (zfs_is_readonly(zfsvfs)) { |
768eaced | 4120 | zfs_exit(zfsvfs, FTAG); |
f3c9dca0 MT |
4121 | return (SET_ERROR(EROFS)); |
4122 | } | |
4123 | ||
34dc7c2f | 4124 | if (bfp->l_len < 0) { |
768eaced | 4125 | zfs_exit(zfsvfs, FTAG); |
2e528b49 | 4126 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4127 | } |
4128 | ||
aec69371 ED |
4129 | /* |
4130 | * Permissions aren't checked on Solaris because on this OS | |
4131 | * zfs_space() can only be called with an opened file handle. | |
4132 | * On Linux we can get here through truncate_range() which | |
4133 | * operates directly on inodes, so we need to check access rights. | |
4134 | */ | |
2a068a13 | 4135 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr, NULL))) { |
768eaced | 4136 | zfs_exit(zfsvfs, FTAG); |
aec69371 ED |
4137 | return (error); |
4138 | } | |
4139 | ||
34dc7c2f BB |
4140 | off = bfp->l_start; |
4141 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4142 | ||
b128c09f | 4143 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4144 | |
768eaced | 4145 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
4146 | return (error); |
4147 | } | |
4148 | ||
e5c39b95 | 4149 | int |
3558fd73 | 4150 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4151 | { |
3558fd73 | 4152 | znode_t *zp = ITOZ(ip); |
0037b49e | 4153 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4154 | uint32_t gen; |
428870ff | 4155 | uint64_t gen64; |
34dc7c2f BB |
4156 | uint64_t object = zp->z_id; |
4157 | zfid_short_t *zfid; | |
428870ff | 4158 | int size, i, error; |
34dc7c2f | 4159 | |
768eaced CC |
4160 | if ((error = zfs_enter(zfsvfs, FTAG)) != 0) |
4161 | return (error); | |
c074a7de AR |
4162 | |
4163 | if (fidp->fid_len < SHORT_FID_LEN) { | |
4164 | fidp->fid_len = SHORT_FID_LEN; | |
768eaced | 4165 | zfs_exit(zfsvfs, FTAG); |
c074a7de AR |
4166 | return (SET_ERROR(ENOSPC)); |
4167 | } | |
4168 | ||
768eaced CC |
4169 | if ((error = zfs_verify_zp(zp)) != 0) { |
4170 | zfs_exit(zfsvfs, FTAG); | |
4171 | return (error); | |
4172 | } | |
428870ff | 4173 | |
0037b49e | 4174 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), |
428870ff | 4175 | &gen64, sizeof (uint64_t))) != 0) { |
768eaced | 4176 | zfs_exit(zfsvfs, FTAG); |
428870ff BB |
4177 | return (error); |
4178 | } | |
4179 | ||
4180 | gen = (uint32_t)gen64; | |
34dc7c2f | 4181 | |
9b77d1c9 | 4182 | size = SHORT_FID_LEN; |
34dc7c2f BB |
4183 | |
4184 | zfid = (zfid_short_t *)fidp; | |
4185 | ||
4186 | zfid->zf_len = size; | |
4187 | ||
4188 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4189 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4190 | ||
4191 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4192 | if (gen == 0) | |
4193 | gen = 1; | |
4194 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4195 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4196 | ||
768eaced | 4197 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
4198 | return (0); |
4199 | } | |
4200 | ||
93ce2b4c | 4201 | #if defined(_KERNEL) |
f298b24d BB |
4202 | EXPORT_SYMBOL(zfs_open); |
4203 | EXPORT_SYMBOL(zfs_close); | |
f298b24d BB |
4204 | EXPORT_SYMBOL(zfs_lookup); |
4205 | EXPORT_SYMBOL(zfs_create); | |
4206 | EXPORT_SYMBOL(zfs_tmpfile); | |
4207 | EXPORT_SYMBOL(zfs_remove); | |
4208 | EXPORT_SYMBOL(zfs_mkdir); | |
4209 | EXPORT_SYMBOL(zfs_rmdir); | |
4210 | EXPORT_SYMBOL(zfs_readdir); | |
f298b24d BB |
4211 | EXPORT_SYMBOL(zfs_getattr_fast); |
4212 | EXPORT_SYMBOL(zfs_setattr); | |
4213 | EXPORT_SYMBOL(zfs_rename); | |
4214 | EXPORT_SYMBOL(zfs_symlink); | |
4215 | EXPORT_SYMBOL(zfs_readlink); | |
4216 | EXPORT_SYMBOL(zfs_link); | |
4217 | EXPORT_SYMBOL(zfs_inactive); | |
4218 | EXPORT_SYMBOL(zfs_space); | |
4219 | EXPORT_SYMBOL(zfs_fid); | |
f298b24d BB |
4220 | EXPORT_SYMBOL(zfs_getpage); |
4221 | EXPORT_SYMBOL(zfs_putpage); | |
4222 | EXPORT_SYMBOL(zfs_dirty_inode); | |
4223 | EXPORT_SYMBOL(zfs_map); | |
4224 | ||
7ada752a | 4225 | /* CSTYLED */ |
a966c564 K |
4226 | module_param(zfs_delete_blocks, ulong, 0644); |
4227 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
09276fde | 4228 | |
c409e464 | 4229 | #endif |