]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
cae5b340 | 21 | |
34dc7c2f | 22 | /* |
428870ff | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
cae5b340 | 24 | * Copyright (c) 2012, 2015 by Delphix. All rights reserved. |
e10b0808 | 25 | * Copyright (c) 2015 by Chunwei Chen. All rights reserved. |
cae5b340 | 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> | |
36 | #include <sys/systm.h> | |
37 | #include <sys/sysmacros.h> | |
38 | #include <sys/resource.h> | |
39 | #include <sys/vfs.h> | |
40 | #include <sys/vfs_opreg.h> | |
34dc7c2f BB |
41 | #include <sys/file.h> |
42 | #include <sys/stat.h> | |
43 | #include <sys/kmem.h> | |
44 | #include <sys/taskq.h> | |
45 | #include <sys/uio.h> | |
46 | #include <sys/vmsystm.h> | |
47 | #include <sys/atomic.h> | |
34dc7c2f | 48 | #include <vm/pvn.h> |
34dc7c2f BB |
49 | #include <sys/pathname.h> |
50 | #include <sys/cmn_err.h> | |
51 | #include <sys/errno.h> | |
52 | #include <sys/unistd.h> | |
53 | #include <sys/zfs_dir.h> | |
54 | #include <sys/zfs_acl.h> | |
55 | #include <sys/zfs_ioctl.h> | |
56 | #include <sys/fs/zfs.h> | |
57 | #include <sys/dmu.h> | |
428870ff | 58 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
59 | #include <sys/spa.h> |
60 | #include <sys/txg.h> | |
61 | #include <sys/dbuf.h> | |
62 | #include <sys/zap.h> | |
428870ff | 63 | #include <sys/sa.h> |
34dc7c2f BB |
64 | #include <sys/dirent.h> |
65 | #include <sys/policy.h> | |
66 | #include <sys/sunddi.h> | |
b128c09f | 67 | #include <sys/sid.h> |
bcf30822 | 68 | #include <sys/mode.h> |
34dc7c2f | 69 | #include "fs/fs_subr.h" |
ebe7e575 | 70 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 71 | #include <sys/zfs_fuid.h> |
428870ff | 72 | #include <sys/zfs_sa.h> |
e5c39b95 | 73 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
74 | #include <sys/dnlc.h> |
75 | #include <sys/zfs_rlock.h> | |
76 | #include <sys/extdirent.h> | |
77 | #include <sys/kidmap.h> | |
428870ff | 78 | #include <sys/cred.h> |
34dc7c2f | 79 | #include <sys/attr.h> |
218b8eaf | 80 | #include <sys/zpl.h> |
34dc7c2f BB |
81 | |
82 | /* | |
83 | * Programming rules. | |
84 | * | |
85 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
86 | * properly lock its in-core state, create a DMU transaction, do the work, | |
87 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
88 | * and wait for the intent log to commit if it is a synchronous operation. | |
89 | * Moreover, the vnode ops must work in both normal and log replay context. | |
90 | * The ordering of events is important to avoid deadlocks and references | |
91 | * to freed memory. The example below illustrates the following Big Rules: | |
92 | * | |
93 | * (1) A check must be made in each zfs thread for a mounted file system. | |
cae5b340 AX |
94 | * This is done avoiding races using ZFS_ENTER(zfsvfs). |
95 | * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes | |
34dc7c2f BB |
96 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros |
97 | * can return EIO from the calling function. | |
98 | * | |
3558fd73 | 99 | * (2) iput() should always be the last thing except for zil_commit() |
34dc7c2f BB |
100 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: |
101 | * First, if it's the last reference, the vnode/znode | |
102 | * can be freed, so the zp may point to freed memory. Second, the last | |
103 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
104 | * pushing cached pages (which acquires range locks) and syncing out | |
105 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
106 | * which could deadlock the system if you were already holding one. | |
ea04106b | 107 | * If you must call iput() within a tx then use zfs_iput_async(). |
34dc7c2f BB |
108 | * |
109 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
110 | * as they can span dmu_tx_assign() calls. | |
111 | * | |
a08ee875 LG |
112 | * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to |
113 | * dmu_tx_assign(). This is critical because we don't want to block | |
114 | * while holding locks. | |
115 | * | |
116 | * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This | |
117 | * reduces lock contention and CPU usage when we must wait (note that if | |
118 | * throughput is constrained by the storage, nearly every transaction | |
119 | * must wait). | |
120 | * | |
121 | * Note, in particular, that if a lock is sometimes acquired before | |
122 | * the tx assigns, and sometimes after (e.g. z_lock), then failing | |
123 | * to use a non-blocking assign can deadlock the system. The scenario: | |
34dc7c2f BB |
124 | * |
125 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
126 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
127 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
128 | * forever, because the previous txg can't quiesce until B's tx commits. | |
129 | * | |
cae5b340 | 130 | * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, |
a08ee875 LG |
131 | * then drop all locks, call dmu_tx_wait(), and try again. On subsequent |
132 | * calls to dmu_tx_assign(), pass TXG_WAITED rather than TXG_NOWAIT, | |
133 | * to indicate that this operation has already called dmu_tx_wait(). | |
134 | * This will ensure that we don't retry forever, waiting a short bit | |
135 | * each time. | |
34dc7c2f BB |
136 | * |
137 | * (5) If the operation succeeded, generate the intent log entry for it | |
138 | * before dropping locks. This ensures that the ordering of events | |
139 | * in the intent log matches the order in which they actually occurred. | |
a08ee875 | 140 | * During ZIL replay the zfs_log_* functions will update the sequence |
fb5f0bc8 | 141 | * number to indicate the zil transaction has replayed. |
34dc7c2f BB |
142 | * |
143 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
144 | * regardless of whether there were any errors. | |
145 | * | |
572e2857 | 146 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
147 | * to ensure that synchronous semantics are provided when necessary. |
148 | * | |
149 | * In general, this is how things should be ordered in each vnode op: | |
150 | * | |
cae5b340 | 151 | * ZFS_ENTER(zfsvfs); // exit if unmounted |
34dc7c2f | 152 | * top: |
3558fd73 | 153 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
154 | * rw_enter(...); // grab any other locks you need |
155 | * tx = dmu_tx_create(...); // get DMU tx | |
156 | * dmu_tx_hold_*(); // hold each object you might modify | |
a08ee875 | 157 | * error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
158 | * if (error) { |
159 | * rw_exit(...); // drop locks | |
160 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 161 | * iput(...); // release held vnodes |
fb5f0bc8 | 162 | * if (error == ERESTART) { |
a08ee875 | 163 | * waited = B_TRUE; |
34dc7c2f BB |
164 | * dmu_tx_wait(tx); |
165 | * dmu_tx_abort(tx); | |
166 | * goto top; | |
167 | * } | |
168 | * dmu_tx_abort(tx); // abort DMU tx | |
cae5b340 | 169 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
170 | * return (error); // really out of space |
171 | * } | |
172 | * error = do_real_work(); // do whatever this VOP does | |
173 | * if (error == 0) | |
174 | * zfs_log_*(...); // on success, make ZIL entry | |
175 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
176 | * rw_exit(...); // drop locks | |
177 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 178 | * iput(...); // release held vnodes |
572e2857 | 179 | * zil_commit(zilog, foid); // synchronous when necessary |
cae5b340 | 180 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
181 | * return (error); // done, report error |
182 | */ | |
183 | ||
126400a1 BB |
184 | /* |
185 | * Virus scanning is unsupported. It would be possible to add a hook | |
186 | * here to performance the required virus scan. This could be done | |
187 | * entirely in the kernel or potentially as an update to invoke a | |
188 | * scanning utility. | |
189 | */ | |
190 | static int | |
191 | zfs_vscan(struct inode *ip, cred_t *cr, int async) | |
192 | { | |
193 | return (0); | |
194 | } | |
195 | ||
196 | /* ARGSUSED */ | |
197 | int | |
198 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
199 | { | |
200 | znode_t *zp = ITOZ(ip); | |
cae5b340 | 201 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 202 | |
cae5b340 | 203 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
204 | ZFS_VERIFY_ZP(zp); |
205 | ||
206 | /* Honor ZFS_APPENDONLY file attribute */ | |
207 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
208 | ((flag & O_APPEND) == 0)) { | |
cae5b340 | 209 | ZFS_EXIT(zfsvfs); |
a08ee875 | 210 | return (SET_ERROR(EPERM)); |
126400a1 BB |
211 | } |
212 | ||
213 | /* Virus scan eligible files on open */ | |
cae5b340 | 214 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
215 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { |
216 | if (zfs_vscan(ip, cr, 0) != 0) { | |
cae5b340 | 217 | ZFS_EXIT(zfsvfs); |
a08ee875 | 218 | return (SET_ERROR(EACCES)); |
126400a1 BB |
219 | } |
220 | } | |
221 | ||
222 | /* Keep a count of the synchronous opens in the znode */ | |
223 | if (flag & O_SYNC) | |
224 | atomic_inc_32(&zp->z_sync_cnt); | |
225 | ||
cae5b340 | 226 | ZFS_EXIT(zfsvfs); |
126400a1 BB |
227 | return (0); |
228 | } | |
126400a1 BB |
229 | |
230 | /* ARGSUSED */ | |
231 | int | |
232 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
233 | { | |
234 | znode_t *zp = ITOZ(ip); | |
cae5b340 | 235 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 236 | |
cae5b340 | 237 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
238 | ZFS_VERIFY_ZP(zp); |
239 | ||
a08ee875 | 240 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 241 | if (flag & O_SYNC) |
a08ee875 | 242 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 | 243 | |
cae5b340 | 244 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
245 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) |
246 | VERIFY(zfs_vscan(ip, cr, 1) == 0); | |
247 | ||
cae5b340 | 248 | ZFS_EXIT(zfsvfs); |
8780c539 | 249 | return (0); |
126400a1 | 250 | } |
126400a1 | 251 | |
c06d4368 AX |
252 | #if defined(SEEK_HOLE) && defined(SEEK_DATA) |
253 | /* | |
254 | * Lseek support for finding holes (cmd == SEEK_HOLE) and | |
255 | * data (cmd == SEEK_DATA). "off" is an in/out parameter. | |
256 | */ | |
257 | static int | |
258 | zfs_holey_common(struct inode *ip, int cmd, loff_t *off) | |
259 | { | |
260 | znode_t *zp = ITOZ(ip); | |
261 | uint64_t noff = (uint64_t)*off; /* new offset */ | |
262 | uint64_t file_sz; | |
263 | int error; | |
264 | boolean_t hole; | |
265 | ||
266 | file_sz = zp->z_size; | |
267 | if (noff >= file_sz) { | |
a08ee875 | 268 | return (SET_ERROR(ENXIO)); |
c06d4368 AX |
269 | } |
270 | ||
271 | if (cmd == SEEK_HOLE) | |
272 | hole = B_TRUE; | |
273 | else | |
274 | hole = B_FALSE; | |
275 | ||
276 | error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff); | |
277 | ||
ea04106b | 278 | if (error == ESRCH) |
a08ee875 | 279 | return (SET_ERROR(ENXIO)); |
ea04106b | 280 | |
cae5b340 AX |
281 | /* file was dirty, so fall back to using generic logic */ |
282 | if (error == EBUSY) { | |
283 | if (hole) | |
284 | *off = file_sz; | |
285 | ||
286 | return (0); | |
287 | } | |
288 | ||
ea04106b AX |
289 | /* |
290 | * We could find a hole that begins after the logical end-of-file, | |
291 | * because dmu_offset_next() only works on whole blocks. If the | |
292 | * EOF falls mid-block, then indicate that the "virtual hole" | |
293 | * at the end of the file begins at the logical EOF, rather than | |
294 | * at the end of the last block. | |
295 | */ | |
296 | if (noff > file_sz) { | |
297 | ASSERT(hole); | |
298 | noff = file_sz; | |
c06d4368 AX |
299 | } |
300 | ||
301 | if (noff < *off) | |
302 | return (error); | |
303 | *off = noff; | |
304 | return (error); | |
305 | } | |
306 | ||
307 | int | |
308 | zfs_holey(struct inode *ip, int cmd, loff_t *off) | |
309 | { | |
310 | znode_t *zp = ITOZ(ip); | |
cae5b340 | 311 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
c06d4368 AX |
312 | int error; |
313 | ||
cae5b340 | 314 | ZFS_ENTER(zfsvfs); |
c06d4368 AX |
315 | ZFS_VERIFY_ZP(zp); |
316 | ||
317 | error = zfs_holey_common(ip, cmd, off); | |
318 | ||
cae5b340 | 319 | ZFS_EXIT(zfsvfs); |
c06d4368 AX |
320 | return (error); |
321 | } | |
c06d4368 AX |
322 | #endif /* SEEK_HOLE && SEEK_DATA */ |
323 | ||
c0d35759 | 324 | #if defined(_KERNEL) |
34dc7c2f BB |
325 | /* |
326 | * When a file is memory mapped, we must keep the IO data synchronized | |
327 | * between the DMU cache and the memory mapped pages. What this means: | |
328 | * | |
329 | * On Write: If we find a memory mapped page, we write to *both* | |
330 | * the page and the dmu buffer. | |
34dc7c2f | 331 | */ |
d164b209 | 332 | static void |
c0d35759 BB |
333 | update_pages(struct inode *ip, int64_t start, int len, |
334 | objset_t *os, uint64_t oid) | |
34dc7c2f | 335 | { |
c0d35759 BB |
336 | struct address_space *mp = ip->i_mapping; |
337 | struct page *pp; | |
338 | uint64_t nbytes; | |
d164b209 | 339 | int64_t off; |
c0d35759 | 340 | void *pb; |
34dc7c2f | 341 | |
5eacc075 AX |
342 | off = start & (PAGE_SIZE-1); |
343 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
344 | nbytes = MIN(PAGE_SIZE - off, len); | |
34dc7c2f | 345 | |
5eacc075 | 346 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
347 | if (pp) { |
348 | if (mapping_writably_mapped(mp)) | |
349 | flush_dcache_page(pp); | |
34dc7c2f | 350 | |
c0d35759 BB |
351 | pb = kmap(pp); |
352 | (void) dmu_read(os, oid, start+off, nbytes, pb+off, | |
9babb374 | 353 | DMU_READ_PREFETCH); |
c0d35759 BB |
354 | kunmap(pp); |
355 | ||
356 | if (mapping_writably_mapped(mp)) | |
357 | flush_dcache_page(pp); | |
358 | ||
359 | mark_page_accessed(pp); | |
360 | SetPageUptodate(pp); | |
361 | ClearPageError(pp); | |
362 | unlock_page(pp); | |
5eacc075 | 363 | put_page(pp); |
34dc7c2f | 364 | } |
c0d35759 | 365 | |
d164b209 | 366 | len -= nbytes; |
34dc7c2f | 367 | off = 0; |
34dc7c2f | 368 | } |
34dc7c2f BB |
369 | } |
370 | ||
371 | /* | |
372 | * When a file is memory mapped, we must keep the IO data synchronized | |
373 | * between the DMU cache and the memory mapped pages. What this means: | |
374 | * | |
375 | * On Read: We "read" preferentially from memory mapped pages, | |
376 | * else we default from the dmu buffer. | |
377 | * | |
378 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
a08ee875 | 379 | * the file is memory mapped. |
34dc7c2f BB |
380 | */ |
381 | static int | |
3558fd73 | 382 | mappedread(struct inode *ip, int nbytes, uio_t *uio) |
34dc7c2f | 383 | { |
c0d35759 BB |
384 | struct address_space *mp = ip->i_mapping; |
385 | struct page *pp; | |
3558fd73 | 386 | znode_t *zp = ITOZ(ip); |
34dc7c2f | 387 | int64_t start, off; |
c0d35759 | 388 | uint64_t bytes; |
34dc7c2f BB |
389 | int len = nbytes; |
390 | int error = 0; | |
c0d35759 | 391 | void *pb; |
34dc7c2f BB |
392 | |
393 | start = uio->uio_loffset; | |
5eacc075 AX |
394 | off = start & (PAGE_SIZE-1); |
395 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
396 | bytes = MIN(PAGE_SIZE - off, len); | |
c0d35759 | 397 | |
5eacc075 | 398 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
399 | if (pp) { |
400 | ASSERT(PageUptodate(pp)); | |
401 | ||
402 | pb = kmap(pp); | |
403 | error = uiomove(pb + off, bytes, UIO_READ, uio); | |
404 | kunmap(pp); | |
405 | ||
406 | if (mapping_writably_mapped(mp)) | |
407 | flush_dcache_page(pp); | |
408 | ||
409 | mark_page_accessed(pp); | |
410 | unlock_page(pp); | |
5eacc075 | 411 | put_page(pp); |
34dc7c2f | 412 | } else { |
e10b0808 AX |
413 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
414 | uio, bytes); | |
34dc7c2f | 415 | } |
c0d35759 | 416 | |
34dc7c2f BB |
417 | len -= bytes; |
418 | off = 0; | |
419 | if (error) | |
420 | break; | |
421 | } | |
422 | return (error); | |
423 | } | |
c0d35759 | 424 | #endif /* _KERNEL */ |
34dc7c2f | 425 | |
c409e464 | 426 | unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ |
cae5b340 | 427 | unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT; |
34dc7c2f BB |
428 | |
429 | /* | |
430 | * Read bytes from specified file into supplied buffer. | |
431 | * | |
3558fd73 | 432 | * IN: ip - inode of file to be read from. |
34dc7c2f BB |
433 | * uio - structure supplying read location, range info, |
434 | * and return buffer. | |
c0d35759 BB |
435 | * ioflag - FSYNC flags; used to provide FRSYNC semantics. |
436 | * O_DIRECT flag; used to bypass page cache. | |
34dc7c2f | 437 | * cr - credentials of caller. |
34dc7c2f BB |
438 | * |
439 | * OUT: uio - updated offset and range, buffer filled. | |
440 | * | |
a08ee875 | 441 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
442 | * |
443 | * Side Effects: | |
3558fd73 | 444 | * inode - atime updated if byte count > 0 |
34dc7c2f BB |
445 | */ |
446 | /* ARGSUSED */ | |
e5c39b95 | 447 | int |
3558fd73 | 448 | zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 449 | { |
3558fd73 | 450 | znode_t *zp = ITOZ(ip); |
cae5b340 | 451 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 452 | ssize_t n, nbytes; |
149e873a | 453 | int error = 0; |
34dc7c2f | 454 | rl_t *rl; |
3558fd73 | 455 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 456 | xuio_t *xuio = NULL; |
3558fd73 | 457 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f | 458 | |
cae5b340 | 459 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 460 | ZFS_VERIFY_ZP(zp); |
34dc7c2f | 461 | |
428870ff | 462 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
cae5b340 | 463 | ZFS_EXIT(zfsvfs); |
a08ee875 | 464 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
465 | } |
466 | ||
467 | /* | |
468 | * Validate file offset | |
469 | */ | |
470 | if (uio->uio_loffset < (offset_t)0) { | |
cae5b340 | 471 | ZFS_EXIT(zfsvfs); |
a08ee875 | 472 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
473 | } |
474 | ||
475 | /* | |
476 | * Fasttrack empty reads | |
477 | */ | |
478 | if (uio->uio_resid == 0) { | |
cae5b340 | 479 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
480 | return (0); |
481 | } | |
482 | ||
34dc7c2f BB |
483 | /* |
484 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
cae5b340 | 485 | * Only do this for non-snapshots. |
34dc7c2f | 486 | */ |
cae5b340 AX |
487 | if (zfsvfs->z_log && |
488 | (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)) | |
489 | zil_commit(zfsvfs->z_log, zp->z_id); | |
34dc7c2f BB |
490 | |
491 | /* | |
492 | * Lock the range against changes. | |
493 | */ | |
87dac73d AX |
494 | rl = zfs_range_lock(&zp->z_range_lock, uio->uio_loffset, uio->uio_resid, |
495 | RL_READER); | |
34dc7c2f BB |
496 | |
497 | /* | |
498 | * If we are reading past end-of-file we can skip | |
499 | * to the end; but we might still need to set atime. | |
500 | */ | |
428870ff | 501 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
502 | error = 0; |
503 | goto out; | |
504 | } | |
505 | ||
428870ff BB |
506 | ASSERT(uio->uio_loffset < zp->z_size); |
507 | n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); | |
508 | ||
3558fd73 | 509 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
510 | if ((uio->uio_extflg == UIO_XUIO) && |
511 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
512 | int nblk; | |
513 | int blksz = zp->z_blksz; | |
514 | uint64_t offset = uio->uio_loffset; | |
515 | ||
516 | xuio = (xuio_t *)uio; | |
517 | if ((ISP2(blksz))) { | |
518 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
519 | blksz)) / blksz; | |
520 | } else { | |
521 | ASSERT(offset + n <= blksz); | |
522 | nblk = 1; | |
523 | } | |
524 | (void) dmu_xuio_init(xuio, nblk); | |
525 | ||
3558fd73 | 526 | if (vn_has_cached_data(ip)) { |
428870ff BB |
527 | /* |
528 | * For simplicity, we always allocate a full buffer | |
529 | * even if we only expect to read a portion of a block. | |
530 | */ | |
531 | while (--nblk >= 0) { | |
532 | (void) dmu_xuio_add(xuio, | |
533 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
534 | blksz), 0, blksz); | |
535 | } | |
536 | } | |
537 | } | |
3558fd73 | 538 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
539 | |
540 | while (n > 0) { | |
541 | nbytes = MIN(n, zfs_read_chunk_size - | |
542 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); | |
543 | ||
e10b0808 | 544 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
3558fd73 | 545 | error = mappedread(ip, nbytes, uio); |
e10b0808 AX |
546 | } else { |
547 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), | |
548 | uio, nbytes); | |
549 | } | |
c0d35759 | 550 | |
b128c09f BB |
551 | if (error) { |
552 | /* convert checksum errors into IO errors */ | |
553 | if (error == ECKSUM) | |
a08ee875 | 554 | error = SET_ERROR(EIO); |
34dc7c2f | 555 | break; |
b128c09f | 556 | } |
34dc7c2f BB |
557 | |
558 | n -= nbytes; | |
559 | } | |
34dc7c2f BB |
560 | out: |
561 | zfs_range_unlock(rl); | |
562 | ||
cae5b340 | 563 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
564 | return (error); |
565 | } | |
566 | ||
34dc7c2f BB |
567 | /* |
568 | * Write the bytes to a file. | |
569 | * | |
3558fd73 | 570 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
571 | * uio - structure supplying write location, range info, |
572 | * and data buffer. | |
573 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 574 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 575 | * cr - credentials of caller. |
34dc7c2f BB |
576 | * |
577 | * OUT: uio - updated offset and range. | |
578 | * | |
579 | * RETURN: 0 if success | |
580 | * error code if failure | |
581 | * | |
582 | * Timestamps: | |
3558fd73 | 583 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 584 | */ |
428870ff | 585 | |
34dc7c2f | 586 | /* ARGSUSED */ |
e5c39b95 | 587 | int |
3558fd73 | 588 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 589 | { |
3558fd73 BB |
590 | znode_t *zp = ITOZ(ip); |
591 | rlim64_t limit = uio->uio_limit; | |
34dc7c2f BB |
592 | ssize_t start_resid = uio->uio_resid; |
593 | ssize_t tx_bytes; | |
594 | uint64_t end_size; | |
595 | dmu_tx_t *tx; | |
cae5b340 | 596 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
597 | zilog_t *zilog; |
598 | offset_t woff; | |
599 | ssize_t n, nbytes; | |
600 | rl_t *rl; | |
cae5b340 | 601 | int max_blksz = zfsvfs->z_max_blksz; |
3558fd73 | 602 | int error = 0; |
9babb374 | 603 | arc_buf_t *abuf; |
e10b0808 | 604 | const iovec_t *aiov = NULL; |
428870ff | 605 | xuio_t *xuio = NULL; |
428870ff BB |
606 | int write_eof; |
607 | int count = 0; | |
608 | sa_bulk_attr_t bulk[4]; | |
609 | uint64_t mtime[2], ctime[2]; | |
cae5b340 AX |
610 | uint32_t uid; |
611 | #ifdef HAVE_UIO_ZEROCOPY | |
612 | int i_iov = 0; | |
613 | const iovec_t *iovp = uio->uio_iov; | |
3558fd73 | 614 | ASSERTV(int iovcnt = uio->uio_iovcnt); |
cae5b340 | 615 | #endif |
34dc7c2f | 616 | |
34dc7c2f BB |
617 | /* |
618 | * Fasttrack empty write | |
619 | */ | |
620 | n = start_resid; | |
621 | if (n == 0) | |
622 | return (0); | |
623 | ||
624 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) | |
625 | limit = MAXOFFSET_T; | |
626 | ||
cae5b340 | 627 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 628 | ZFS_VERIFY_ZP(zp); |
b128c09f | 629 | |
cae5b340 AX |
630 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
631 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
632 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
633 | &zp->z_size, 8); | |
634 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
428870ff BB |
635 | &zp->z_pflags, 8); |
636 | ||
cae5b340 AX |
637 | /* |
638 | * Callers might not be able to detect properly that we are read-only, | |
639 | * so check it explicitly here. | |
640 | */ | |
641 | if (zfs_is_readonly(zfsvfs)) { | |
642 | ZFS_EXIT(zfsvfs); | |
643 | return (SET_ERROR(EROFS)); | |
644 | } | |
645 | ||
b128c09f BB |
646 | /* |
647 | * If immutable or not appending then return EPERM | |
648 | */ | |
428870ff BB |
649 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
650 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
651 | (uio->uio_loffset < zp->z_size))) { | |
cae5b340 | 652 | ZFS_EXIT(zfsvfs); |
a08ee875 | 653 | return (SET_ERROR(EPERM)); |
b128c09f BB |
654 | } |
655 | ||
cae5b340 | 656 | zilog = zfsvfs->z_log; |
34dc7c2f | 657 | |
428870ff BB |
658 | /* |
659 | * Validate file offset | |
660 | */ | |
661 | woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; | |
662 | if (woff < 0) { | |
cae5b340 | 663 | ZFS_EXIT(zfsvfs); |
a08ee875 | 664 | return (SET_ERROR(EINVAL)); |
428870ff BB |
665 | } |
666 | ||
34dc7c2f BB |
667 | /* |
668 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
669 | * don't hold up txg. | |
428870ff | 670 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 671 | */ |
a08ee875 | 672 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
673 | if ((uio->uio_extflg == UIO_XUIO) && |
674 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
675 | xuio = (xuio_t *)uio; | |
676 | else | |
a08ee875 | 677 | #endif |
572e2857 | 678 | uio_prefaultpages(MIN(n, max_blksz), uio); |
34dc7c2f BB |
679 | |
680 | /* | |
681 | * If in append mode, set the io offset pointer to eof. | |
682 | */ | |
683 | if (ioflag & FAPPEND) { | |
684 | /* | |
428870ff BB |
685 | * Obtain an appending range lock to guarantee file append |
686 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f | 687 | */ |
87dac73d | 688 | rl = zfs_range_lock(&zp->z_range_lock, 0, n, RL_APPEND); |
428870ff | 689 | woff = rl->r_off; |
34dc7c2f | 690 | if (rl->r_len == UINT64_MAX) { |
428870ff BB |
691 | /* |
692 | * We overlocked the file because this write will cause | |
693 | * the file block size to increase. | |
694 | * Note that zp_size cannot change with this lock held. | |
695 | */ | |
696 | woff = zp->z_size; | |
34dc7c2f | 697 | } |
428870ff | 698 | uio->uio_loffset = woff; |
34dc7c2f | 699 | } else { |
34dc7c2f | 700 | /* |
428870ff BB |
701 | * Note that if the file block size will change as a result of |
702 | * this write, then this range lock will lock the entire file | |
703 | * so that we can re-write the block safely. | |
34dc7c2f | 704 | */ |
87dac73d | 705 | rl = zfs_range_lock(&zp->z_range_lock, woff, n, RL_WRITER); |
34dc7c2f BB |
706 | } |
707 | ||
708 | if (woff >= limit) { | |
709 | zfs_range_unlock(rl); | |
cae5b340 | 710 | ZFS_EXIT(zfsvfs); |
a08ee875 | 711 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
712 | } |
713 | ||
714 | if ((woff + n) > limit || woff > (limit - n)) | |
715 | n = limit - woff; | |
716 | ||
428870ff BB |
717 | /* Will this write extend the file length? */ |
718 | write_eof = (woff + n > zp->z_size); | |
719 | ||
720 | end_size = MAX(zp->z_size, woff + n); | |
34dc7c2f BB |
721 | |
722 | /* | |
723 | * Write the file in reasonable size chunks. Each chunk is written | |
724 | * in a separate transaction; this keeps the intent log records small | |
725 | * and allows us to do more fine-grained space accounting. | |
726 | */ | |
727 | while (n > 0) { | |
9babb374 BB |
728 | abuf = NULL; |
729 | woff = uio->uio_loffset; | |
cae5b340 AX |
730 | if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || |
731 | zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { | |
9babb374 BB |
732 | if (abuf != NULL) |
733 | dmu_return_arcbuf(abuf); | |
a08ee875 | 734 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
735 | break; |
736 | } | |
737 | ||
428870ff | 738 | if (xuio && abuf == NULL) { |
cae5b340 | 739 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 740 | ASSERT(i_iov < iovcnt); |
e10b0808 | 741 | ASSERT3U(uio->uio_segflg, !=, UIO_BVEC); |
428870ff BB |
742 | aiov = &iovp[i_iov]; |
743 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
744 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
745 | ASSERT((aiov->iov_base == abuf->b_data) || |
746 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
747 | aiov->iov_len == arc_buf_size(abuf))); | |
748 | i_iov++; | |
cae5b340 | 749 | #endif |
428870ff BB |
750 | } else if (abuf == NULL && n >= max_blksz && |
751 | woff >= zp->z_size && | |
9babb374 BB |
752 | P2PHASE(woff, max_blksz) == 0 && |
753 | zp->z_blksz == max_blksz) { | |
428870ff BB |
754 | /* |
755 | * This write covers a full block. "Borrow" a buffer | |
756 | * from the dmu so that we can fill it before we enter | |
757 | * a transaction. This avoids the possibility of | |
758 | * holding up the transaction if the data copy hangs | |
759 | * up on a pagefault (e.g., from an NFS server mapping). | |
760 | */ | |
9babb374 BB |
761 | size_t cbytes; |
762 | ||
428870ff BB |
763 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
764 | max_blksz); | |
9babb374 BB |
765 | ASSERT(abuf != NULL); |
766 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
767 | if ((error = uiocopy(abuf->b_data, max_blksz, |
768 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
769 | dmu_return_arcbuf(abuf); |
770 | break; | |
771 | } | |
772 | ASSERT(cbytes == max_blksz); | |
773 | } | |
774 | ||
34dc7c2f BB |
775 | /* |
776 | * Start a transaction. | |
777 | */ | |
cae5b340 | 778 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 779 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 780 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 781 | zfs_sa_upgrade_txholds(tx, zp); |
a08ee875 | 782 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 783 | if (error) { |
34dc7c2f | 784 | dmu_tx_abort(tx); |
9babb374 BB |
785 | if (abuf != NULL) |
786 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
787 | break; |
788 | } | |
789 | ||
790 | /* | |
791 | * If zfs_range_lock() over-locked we grow the blocksize | |
792 | * and then reduce the lock range. This will only happen | |
793 | * on the first iteration since zfs_range_reduce() will | |
794 | * shrink down r_len to the appropriate size. | |
795 | */ | |
796 | if (rl->r_len == UINT64_MAX) { | |
797 | uint64_t new_blksz; | |
798 | ||
799 | if (zp->z_blksz > max_blksz) { | |
e10b0808 AX |
800 | /* |
801 | * File's blocksize is already larger than the | |
802 | * "recordsize" property. Only let it grow to | |
803 | * the next power of 2. | |
804 | */ | |
34dc7c2f | 805 | ASSERT(!ISP2(zp->z_blksz)); |
e10b0808 AX |
806 | new_blksz = MIN(end_size, |
807 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
808 | } else { |
809 | new_blksz = MIN(end_size, max_blksz); | |
810 | } | |
811 | zfs_grow_blocksize(zp, new_blksz, tx); | |
812 | zfs_range_reduce(rl, woff, n); | |
813 | } | |
814 | ||
815 | /* | |
816 | * XXX - should we really limit each write to z_max_blksz? | |
817 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
818 | */ | |
819 | nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); | |
34dc7c2f | 820 | |
9babb374 BB |
821 | if (abuf == NULL) { |
822 | tx_bytes = uio->uio_resid; | |
428870ff BB |
823 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
824 | uio, nbytes, tx); | |
9babb374 BB |
825 | tx_bytes -= uio->uio_resid; |
826 | } else { | |
827 | tx_bytes = nbytes; | |
428870ff BB |
828 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
829 | /* | |
830 | * If this is not a full block write, but we are | |
831 | * extending the file past EOF and this data starts | |
832 | * block-aligned, use assign_arcbuf(). Otherwise, | |
833 | * write via dmu_write(). | |
834 | */ | |
835 | if (tx_bytes < max_blksz && (!write_eof || | |
836 | aiov->iov_base != abuf->b_data)) { | |
837 | ASSERT(xuio); | |
cae5b340 | 838 | dmu_write(zfsvfs->z_os, zp->z_id, woff, |
047218e2 | 839 | /* cppcheck-suppress nullPointer */ |
428870ff BB |
840 | aiov->iov_len, aiov->iov_base, tx); |
841 | dmu_return_arcbuf(abuf); | |
842 | xuio_stat_wbuf_copied(); | |
843 | } else { | |
844 | ASSERT(xuio || tx_bytes == max_blksz); | |
845 | dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl), | |
846 | woff, abuf, tx); | |
847 | } | |
9babb374 BB |
848 | ASSERT(tx_bytes <= uio->uio_resid); |
849 | uioskip(uio, tx_bytes); | |
850 | } | |
cae5b340 AX |
851 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
852 | update_pages(ip, woff, | |
853 | tx_bytes, zfsvfs->z_os, zp->z_id); | |
854 | } | |
34dc7c2f BB |
855 | |
856 | /* | |
857 | * If we made no progress, we're done. If we made even | |
858 | * partial progress, update the znode and ZIL accordingly. | |
859 | */ | |
860 | if (tx_bytes == 0) { | |
cae5b340 | 861 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 862 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
863 | dmu_tx_commit(tx); |
864 | ASSERT(error != 0); | |
865 | break; | |
866 | } | |
867 | ||
868 | /* | |
869 | * Clear Set-UID/Set-GID bits on successful write if not | |
cae5b340 | 870 | * privileged and at least one of the execute bits is set. |
34dc7c2f BB |
871 | * |
872 | * It would be nice to to this after all writes have | |
873 | * been done, but that would still expose the ISUID/ISGID | |
874 | * to another app after the partial write is committed. | |
875 | * | |
572e2857 BB |
876 | * Note: we don't call zfs_fuid_map_id() here because |
877 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
878 | */ |
879 | mutex_enter(&zp->z_acl_lock); | |
cae5b340 | 880 | uid = KUID_TO_SUID(ip->i_uid); |
428870ff | 881 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 882 | (S_IXUSR >> 6))) != 0 && |
428870ff | 883 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 884 | secpolicy_vnode_setid_retain(cr, |
cae5b340 | 885 | ((zp->z_mode & S_ISUID) != 0 && uid == 0)) != 0) { |
428870ff BB |
886 | uint64_t newmode; |
887 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
cae5b340 AX |
888 | ip->i_mode = newmode = zp->z_mode; |
889 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), | |
428870ff | 890 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
891 | } |
892 | mutex_exit(&zp->z_acl_lock); | |
893 | ||
68d83c55 | 894 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
34dc7c2f BB |
895 | |
896 | /* | |
897 | * Update the file size (zp_size) if it has changed; | |
898 | * account for possible concurrent updates. | |
899 | */ | |
428870ff BB |
900 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
901 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 902 | uio->uio_loffset); |
428870ff BB |
903 | ASSERT(error == 0); |
904 | } | |
572e2857 BB |
905 | /* |
906 | * If we are replaying and eof is non zero then force | |
907 | * the file size to the specified eof. Note, there's no | |
908 | * concurrency during replay. | |
909 | */ | |
cae5b340 AX |
910 | if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) |
911 | zp->z_size = zfsvfs->z_replay_eof; | |
572e2857 | 912 | |
428870ff BB |
913 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
914 | ||
a08ee875 LG |
915 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
916 | NULL, NULL); | |
34dc7c2f BB |
917 | dmu_tx_commit(tx); |
918 | ||
919 | if (error != 0) | |
920 | break; | |
921 | ASSERT(tx_bytes == nbytes); | |
922 | n -= nbytes; | |
572e2857 BB |
923 | |
924 | if (!xuio && n > 0) | |
925 | uio_prefaultpages(MIN(n, max_blksz), uio); | |
34dc7c2f BB |
926 | } |
927 | ||
e10b0808 | 928 | zfs_inode_update(zp); |
34dc7c2f BB |
929 | zfs_range_unlock(rl); |
930 | ||
931 | /* | |
932 | * If we're in replay mode, or we made no progress, return error. | |
933 | * Otherwise, it's at least a partial write, so it's successful. | |
934 | */ | |
cae5b340 AX |
935 | if (zfsvfs->z_replay || uio->uio_resid == start_resid) { |
936 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
937 | return (error); |
938 | } | |
939 | ||
428870ff | 940 | if (ioflag & (FSYNC | FDSYNC) || |
cae5b340 | 941 | zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 942 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 943 | |
cae5b340 | 944 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
945 | return (0); |
946 | } | |
947 | ||
cae5b340 AX |
948 | /* |
949 | * Drop a reference on the passed inode asynchronously. This ensures | |
950 | * that the caller will never drop the last reference on an inode in | |
951 | * the current context. Doing so while holding open a tx could result | |
952 | * in a deadlock if iput_final() re-enters the filesystem code. | |
953 | */ | |
ea04106b AX |
954 | void |
955 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 956 | { |
ea04106b AX |
957 | objset_t *os = ITOZSB(ip)->z_os; |
958 | ||
3558fd73 | 959 | ASSERT(atomic_read(&ip->i_count) > 0); |
ea04106b AX |
960 | ASSERT(os != NULL); |
961 | ||
3558fd73 | 962 | if (atomic_read(&ip->i_count) == 1) |
cae5b340 AX |
963 | VERIFY(taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
964 | (task_func_t *)iput, ip, TQ_SLEEP) != TASKQID_INVALID); | |
3558fd73 BB |
965 | else |
966 | iput(ip); | |
967 | } | |
968 | ||
34dc7c2f | 969 | void |
428870ff | 970 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 971 | { |
428870ff | 972 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
973 | |
974 | if (zgd->zgd_db) | |
975 | dmu_buf_rele(zgd->zgd_db, zgd); | |
976 | ||
977 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f | 978 | |
9babb374 BB |
979 | /* |
980 | * Release the vnode asynchronously as we currently have the | |
981 | * txg stopped from syncing. | |
982 | */ | |
ea04106b | 983 | zfs_iput_async(ZTOI(zp)); |
428870ff BB |
984 | |
985 | if (error == 0 && zgd->zgd_bp) | |
986 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
987 | ||
34dc7c2f BB |
988 | kmem_free(zgd, sizeof (zgd_t)); |
989 | } | |
990 | ||
45d1cae3 BB |
991 | #ifdef DEBUG |
992 | static int zil_fault_io = 0; | |
993 | #endif | |
994 | ||
34dc7c2f BB |
995 | /* |
996 | * Get data to generate a TX_WRITE intent log record. | |
997 | */ | |
998 | int | |
999 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
1000 | { | |
cae5b340 AX |
1001 | zfsvfs_t *zfsvfs = arg; |
1002 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 1003 | znode_t *zp; |
428870ff BB |
1004 | uint64_t object = lr->lr_foid; |
1005 | uint64_t offset = lr->lr_offset; | |
1006 | uint64_t size = lr->lr_length; | |
34dc7c2f | 1007 | dmu_buf_t *db; |
34dc7c2f | 1008 | zgd_t *zgd; |
34dc7c2f BB |
1009 | int error = 0; |
1010 | ||
428870ff BB |
1011 | ASSERT(zio != NULL); |
1012 | ASSERT(size != 0); | |
34dc7c2f BB |
1013 | |
1014 | /* | |
1015 | * Nothing to do if the file has been removed | |
1016 | */ | |
cae5b340 | 1017 | if (zfs_zget(zfsvfs, object, &zp) != 0) |
a08ee875 | 1018 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 1019 | if (zp->z_unlinked) { |
9babb374 BB |
1020 | /* |
1021 | * Release the vnode asynchronously as we currently have the | |
1022 | * txg stopped from syncing. | |
1023 | */ | |
ea04106b | 1024 | zfs_iput_async(ZTOI(zp)); |
a08ee875 | 1025 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1026 | } |
1027 | ||
ea04106b | 1028 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
cae5b340 | 1029 | zgd->zgd_zilog = zfsvfs->z_log; |
428870ff BB |
1030 | zgd->zgd_private = zp; |
1031 | ||
34dc7c2f BB |
1032 | /* |
1033 | * Write records come in two flavors: immediate and indirect. | |
1034 | * For small writes it's cheaper to store the data with the | |
1035 | * log record (immediate); for large writes it's cheaper to | |
1036 | * sync the data and get a pointer to it (indirect) so that | |
1037 | * we don't have to write the data twice. | |
1038 | */ | |
1039 | if (buf != NULL) { /* immediate write */ | |
87dac73d AX |
1040 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, size, |
1041 | RL_READER); | |
34dc7c2f | 1042 | /* test for truncation needs to be done while range locked */ |
428870ff | 1043 | if (offset >= zp->z_size) { |
a08ee875 | 1044 | error = SET_ERROR(ENOENT); |
428870ff BB |
1045 | } else { |
1046 | error = dmu_read(os, object, offset, size, buf, | |
1047 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1048 | } |
428870ff | 1049 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1050 | } else { /* indirect write */ |
34dc7c2f BB |
1051 | /* |
1052 | * Have to lock the whole block to ensure when it's | |
cae5b340 | 1053 | * written out and its checksum is being calculated |
34dc7c2f BB |
1054 | * that no one can change the data. We need to re-check |
1055 | * blocksize after we get the lock in case it's changed! | |
1056 | */ | |
1057 | for (;;) { | |
428870ff BB |
1058 | uint64_t blkoff; |
1059 | size = zp->z_blksz; | |
1060 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1061 | offset -= blkoff; | |
87dac73d AX |
1062 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, |
1063 | size, RL_READER); | |
428870ff | 1064 | if (zp->z_blksz == size) |
34dc7c2f | 1065 | break; |
428870ff BB |
1066 | offset += blkoff; |
1067 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f BB |
1068 | } |
1069 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1070 | if (lr->lr_offset >= zp->z_size) |
a08ee875 | 1071 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1072 | #ifdef DEBUG |
1073 | if (zil_fault_io) { | |
a08ee875 | 1074 | error = SET_ERROR(EIO); |
45d1cae3 | 1075 | zil_fault_io = 0; |
45d1cae3 | 1076 | } |
45d1cae3 | 1077 | #endif |
34dc7c2f | 1078 | if (error == 0) |
428870ff BB |
1079 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1080 | DMU_READ_NO_PREFETCH); | |
1081 | ||
1082 | if (error == 0) { | |
cae5b340 | 1083 | blkptr_t *bp = &lr->lr_blkptr; |
a08ee875 | 1084 | |
428870ff BB |
1085 | zgd->zgd_db = db; |
1086 | zgd->zgd_bp = bp; | |
1087 | ||
1088 | ASSERT(db->db_offset == offset); | |
1089 | ASSERT(db->db_size == size); | |
1090 | ||
1091 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1092 | zfs_get_done, zgd); | |
cae5b340 | 1093 | ASSERT(error || lr->lr_length <= size); |
428870ff BB |
1094 | |
1095 | /* | |
1096 | * On success, we need to wait for the write I/O | |
1097 | * initiated by dmu_sync() to complete before we can | |
1098 | * release this dbuf. We will finish everything up | |
1099 | * in the zfs_get_done() callback. | |
1100 | */ | |
1101 | if (error == 0) | |
1102 | return (0); | |
1103 | ||
1104 | if (error == EALREADY) { | |
1105 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
1106 | error = 0; | |
1107 | } | |
1108 | } | |
34dc7c2f | 1109 | } |
428870ff BB |
1110 | |
1111 | zfs_get_done(zgd, error); | |
1112 | ||
34dc7c2f BB |
1113 | return (error); |
1114 | } | |
1115 | ||
1116 | /*ARGSUSED*/ | |
3558fd73 BB |
1117 | int |
1118 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1119 | { |
3558fd73 | 1120 | znode_t *zp = ITOZ(ip); |
cae5b340 | 1121 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
1122 | int error; |
1123 | ||
cae5b340 | 1124 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1125 | ZFS_VERIFY_ZP(zp); |
1126 | ||
1127 | if (flag & V_ACE_MASK) | |
1128 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1129 | else | |
1130 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1131 | ||
cae5b340 | 1132 | ZFS_EXIT(zfsvfs); |
45d1cae3 BB |
1133 | return (error); |
1134 | } | |
45d1cae3 | 1135 | |
34dc7c2f BB |
1136 | /* |
1137 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1138 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1139 | * |
3558fd73 | 1140 | * IN: dip - inode of directory to search. |
34dc7c2f | 1141 | * nm - name of entry to lookup. |
34dc7c2f | 1142 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1143 | * cr - credentials of caller. |
34dc7c2f BB |
1144 | * direntflags - directory lookup flags |
1145 | * realpnp - returned pathname. | |
1146 | * | |
3558fd73 | 1147 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1148 | * |
a08ee875 | 1149 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1150 | * |
1151 | * Timestamps: | |
1152 | * NA | |
1153 | */ | |
1154 | /* ARGSUSED */ | |
e5c39b95 | 1155 | int |
3558fd73 BB |
1156 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1157 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1158 | { |
3558fd73 | 1159 | znode_t *zdp = ITOZ(dip); |
cae5b340 | 1160 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
3558fd73 | 1161 | int error = 0; |
45d1cae3 | 1162 | |
cae5b340 AX |
1163 | /* |
1164 | * Fast path lookup, however we must skip DNLC lookup | |
1165 | * for case folding or normalizing lookups because the | |
1166 | * DNLC code only stores the passed in name. This means | |
1167 | * creating 'a' and removing 'A' on a case insensitive | |
1168 | * file system would work, but DNLC still thinks 'a' | |
1169 | * exists and won't let you create it again on the next | |
1170 | * pass through fast path. | |
1171 | */ | |
45d1cae3 BB |
1172 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { |
1173 | ||
3558fd73 | 1174 | if (!S_ISDIR(dip->i_mode)) { |
a08ee875 | 1175 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1176 | } else if (zdp->z_sa_hdl == NULL) { |
a08ee875 | 1177 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1178 | } |
1179 | ||
1180 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1181 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1182 | if (!error) { | |
3558fd73 BB |
1183 | *ipp = dip; |
1184 | igrab(*ipp); | |
45d1cae3 BB |
1185 | return (0); |
1186 | } | |
1187 | return (error); | |
3558fd73 | 1188 | #ifdef HAVE_DNLC |
cae5b340 AX |
1189 | } else if (!zdp->z_zfsvfs->z_norm && |
1190 | (zdp->z_zfsvfs->z_case == ZFS_CASE_SENSITIVE)) { | |
1191 | ||
45d1cae3 BB |
1192 | vnode_t *tvp = dnlc_lookup(dvp, nm); |
1193 | ||
1194 | if (tvp) { | |
1195 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1196 | if (error) { | |
3558fd73 | 1197 | iput(tvp); |
45d1cae3 BB |
1198 | return (error); |
1199 | } | |
1200 | if (tvp == DNLC_NO_VNODE) { | |
3558fd73 | 1201 | iput(tvp); |
a08ee875 | 1202 | return (SET_ERROR(ENOENT)); |
45d1cae3 BB |
1203 | } else { |
1204 | *vpp = tvp; | |
1205 | return (specvp_check(vpp, cr)); | |
1206 | } | |
1207 | } | |
3558fd73 | 1208 | #endif /* HAVE_DNLC */ |
45d1cae3 BB |
1209 | } |
1210 | } | |
1211 | ||
cae5b340 | 1212 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1213 | ZFS_VERIFY_ZP(zdp); |
1214 | ||
3558fd73 | 1215 | *ipp = NULL; |
34dc7c2f BB |
1216 | |
1217 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1218 | /* |
1219 | * We don't allow recursive attributes.. | |
1220 | * Maybe someday we will. | |
1221 | */ | |
428870ff | 1222 | if (zdp->z_pflags & ZFS_XATTR) { |
cae5b340 | 1223 | ZFS_EXIT(zfsvfs); |
a08ee875 | 1224 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1225 | } |
1226 | ||
3558fd73 | 1227 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
cae5b340 | 1228 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1229 | return (error); |
1230 | } | |
1231 | ||
1232 | /* | |
1233 | * Do we have permission to get into attribute directory? | |
1234 | */ | |
1235 | ||
3558fd73 | 1236 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1237 | B_FALSE, cr))) { |
3558fd73 BB |
1238 | iput(*ipp); |
1239 | *ipp = NULL; | |
34dc7c2f BB |
1240 | } |
1241 | ||
cae5b340 | 1242 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1243 | return (error); |
1244 | } | |
1245 | ||
3558fd73 | 1246 | if (!S_ISDIR(dip->i_mode)) { |
cae5b340 | 1247 | ZFS_EXIT(zfsvfs); |
a08ee875 | 1248 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1249 | } |
1250 | ||
1251 | /* | |
1252 | * Check accessibility of directory. | |
1253 | */ | |
1254 | ||
149e873a | 1255 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
cae5b340 | 1256 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1257 | return (error); |
1258 | } | |
1259 | ||
cae5b340 | 1260 | if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1261 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
cae5b340 | 1262 | ZFS_EXIT(zfsvfs); |
a08ee875 | 1263 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1264 | } |
1265 | ||
3558fd73 BB |
1266 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1267 | if ((error == 0) && (*ipp)) | |
1268 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1269 | |
cae5b340 | 1270 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1271 | return (error); |
1272 | } | |
1273 | ||
1274 | /* | |
1275 | * Attempt to create a new entry in a directory. If the entry | |
1276 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1277 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1278 | * |
3558fd73 | 1279 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1280 | * name - name of new file entry. |
1281 | * vap - attributes of new file. | |
1282 | * excl - flag indicating exclusive or non-exclusive mode. | |
1283 | * mode - mode to open file with. | |
1284 | * cr - credentials of caller. | |
1285 | * flag - large file flag [UNUSED]. | |
3558fd73 | 1286 | * vsecp - ACL to be set |
34dc7c2f | 1287 | * |
3558fd73 | 1288 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1289 | * |
a08ee875 | 1290 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1291 | * |
1292 | * Timestamps: | |
3558fd73 BB |
1293 | * dip - ctime|mtime updated if new entry created |
1294 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1295 | */ |
1296 | ||
1297 | /* ARGSUSED */ | |
e5c39b95 | 1298 | int |
3558fd73 BB |
1299 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1300 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1301 | { |
3558fd73 | 1302 | znode_t *zp, *dzp = ITOZ(dip); |
cae5b340 | 1303 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1304 | zilog_t *zilog; |
1305 | objset_t *os; | |
1306 | zfs_dirlock_t *dl; | |
1307 | dmu_tx_t *tx; | |
1308 | int error; | |
b128c09f | 1309 | uid_t uid; |
149e873a | 1310 | gid_t gid; |
428870ff | 1311 | zfs_acl_ids_t acl_ids; |
9babb374 | 1312 | boolean_t fuid_dirtied; |
428870ff | 1313 | boolean_t have_acl = B_FALSE; |
a08ee875 | 1314 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1315 | |
1316 | /* | |
1317 | * If we have an ephemeral id, ACL, or XVATTR then | |
1318 | * make sure file system is at proper version | |
1319 | */ | |
1320 | ||
149e873a | 1321 | gid = crgetgid(cr); |
3558fd73 | 1322 | uid = crgetuid(cr); |
b128c09f | 1323 | |
cae5b340 | 1324 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1325 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
a08ee875 | 1326 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1327 | |
cae5b340 AX |
1328 | if (name == NULL) |
1329 | return (SET_ERROR(EINVAL)); | |
1330 | ||
1331 | ZFS_ENTER(zfsvfs); | |
34dc7c2f | 1332 | ZFS_VERIFY_ZP(dzp); |
cae5b340 AX |
1333 | os = zfsvfs->z_os; |
1334 | zilog = zfsvfs->z_log; | |
34dc7c2f | 1335 | |
cae5b340 | 1336 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1337 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
cae5b340 | 1338 | ZFS_EXIT(zfsvfs); |
a08ee875 | 1339 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1340 | } |
1341 | ||
5484965a | 1342 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1343 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1344 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
cae5b340 | 1345 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1346 | return (error); |
1347 | } | |
1348 | } | |
34dc7c2f | 1349 | |
3558fd73 BB |
1350 | top: |
1351 | *ipp = NULL; | |
34dc7c2f BB |
1352 | if (*name == '\0') { |
1353 | /* | |
1354 | * Null component name refers to the directory itself. | |
1355 | */ | |
3558fd73 | 1356 | igrab(dip); |
34dc7c2f BB |
1357 | zp = dzp; |
1358 | dl = NULL; | |
1359 | error = 0; | |
1360 | } else { | |
3558fd73 | 1361 | /* possible igrab(zp) */ |
34dc7c2f BB |
1362 | int zflg = 0; |
1363 | ||
1364 | if (flag & FIGNORECASE) | |
1365 | zflg |= ZCILOOK; | |
1366 | ||
1367 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1368 | NULL, NULL); | |
1369 | if (error) { | |
572e2857 BB |
1370 | if (have_acl) |
1371 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1372 | if (strcmp(name, "..") == 0) |
a08ee875 | 1373 | error = SET_ERROR(EISDIR); |
cae5b340 | 1374 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1375 | return (error); |
1376 | } | |
1377 | } | |
428870ff | 1378 | |
34dc7c2f BB |
1379 | if (zp == NULL) { |
1380 | uint64_t txtype; | |
1381 | ||
1382 | /* | |
1383 | * Create a new file object and update the directory | |
1384 | * to reference it. | |
1385 | */ | |
149e873a | 1386 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1387 | if (have_acl) |
1388 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1389 | goto out; |
1390 | } | |
1391 | ||
1392 | /* | |
1393 | * We only support the creation of regular files in | |
1394 | * extended attribute directories. | |
1395 | */ | |
428870ff | 1396 | |
3558fd73 | 1397 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1398 | if (have_acl) |
1399 | zfs_acl_ids_free(&acl_ids); | |
a08ee875 | 1400 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1401 | goto out; |
1402 | } | |
1403 | ||
428870ff BB |
1404 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1405 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1406 | goto out; |
428870ff BB |
1407 | have_acl = B_TRUE; |
1408 | ||
cae5b340 | 1409 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
45d1cae3 | 1410 | zfs_acl_ids_free(&acl_ids); |
a08ee875 | 1411 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1412 | goto out; |
1413 | } | |
1414 | ||
34dc7c2f | 1415 | tx = dmu_tx_create(os); |
428870ff BB |
1416 | |
1417 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1418 | ZFS_SA_BASE_ATTR_SIZE); | |
1419 | ||
cae5b340 | 1420 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1421 | if (fuid_dirtied) |
cae5b340 | 1422 | zfs_fuid_txhold(zfsvfs, tx); |
34dc7c2f | 1423 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1424 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
cae5b340 | 1425 | if (!zfsvfs->z_use_sa && |
428870ff | 1426 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1427 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1428 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1429 | } |
a08ee875 | 1430 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1431 | if (error) { |
1432 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1433 | if (error == ERESTART) { |
a08ee875 | 1434 | waited = B_TRUE; |
34dc7c2f BB |
1435 | dmu_tx_wait(tx); |
1436 | dmu_tx_abort(tx); | |
1437 | goto top; | |
1438 | } | |
428870ff | 1439 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1440 | dmu_tx_abort(tx); |
cae5b340 | 1441 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1442 | return (error); |
1443 | } | |
428870ff | 1444 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 BB |
1445 | |
1446 | if (fuid_dirtied) | |
cae5b340 | 1447 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 1448 | |
34dc7c2f BB |
1449 | (void) zfs_link_create(dl, zp, tx, ZNEW); |
1450 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); | |
1451 | if (flag & FIGNORECASE) | |
1452 | txtype |= TX_CI; | |
1453 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1454 | vsecp, acl_ids.z_fuidp, vap); |
1455 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1456 | dmu_tx_commit(tx); |
1457 | } else { | |
1458 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1459 | ||
572e2857 BB |
1460 | if (have_acl) |
1461 | zfs_acl_ids_free(&acl_ids); | |
1462 | have_acl = B_FALSE; | |
1463 | ||
34dc7c2f BB |
1464 | /* |
1465 | * A directory entry already exists for this name. | |
1466 | */ | |
1467 | /* | |
1468 | * Can't truncate an existing file if in exclusive mode. | |
1469 | */ | |
3558fd73 | 1470 | if (excl) { |
a08ee875 | 1471 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1472 | goto out; |
1473 | } | |
1474 | /* | |
1475 | * Can't open a directory for writing. | |
1476 | */ | |
3558fd73 | 1477 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
a08ee875 | 1478 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1479 | goto out; |
1480 | } | |
1481 | /* | |
1482 | * Verify requested access to file. | |
1483 | */ | |
1484 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1485 | goto out; | |
1486 | } | |
1487 | ||
1488 | mutex_enter(&dzp->z_lock); | |
1489 | dzp->z_seq++; | |
1490 | mutex_exit(&dzp->z_lock); | |
1491 | ||
1492 | /* | |
1493 | * Truncate regular files if requested. | |
1494 | */ | |
3558fd73 BB |
1495 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1496 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 1497 | /* we can't hold any locks when calling zfs_freesp() */ |
cae5b340 AX |
1498 | if (dl) { |
1499 | zfs_dirent_unlock(dl); | |
1500 | dl = NULL; | |
1501 | } | |
34dc7c2f | 1502 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1503 | } |
1504 | } | |
1505 | out: | |
1506 | ||
1507 | if (dl) | |
1508 | zfs_dirent_unlock(dl); | |
1509 | ||
1510 | if (error) { | |
1511 | if (zp) | |
3558fd73 | 1512 | iput(ZTOI(zp)); |
34dc7c2f | 1513 | } else { |
960e08fe BB |
1514 | zfs_inode_update(dzp); |
1515 | zfs_inode_update(zp); | |
3558fd73 | 1516 | *ipp = ZTOI(zp); |
34dc7c2f | 1517 | } |
34dc7c2f | 1518 | |
cae5b340 | 1519 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1520 | zil_commit(zilog, 0); |
428870ff | 1521 | |
cae5b340 AX |
1522 | ZFS_EXIT(zfsvfs); |
1523 | return (error); | |
1524 | } | |
1525 | ||
1526 | /* ARGSUSED */ | |
1527 | int | |
1528 | zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, | |
1529 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
1530 | { | |
1531 | znode_t *zp = NULL, *dzp = ITOZ(dip); | |
1532 | zfsvfs_t *zfsvfs = ITOZSB(dip); | |
1533 | objset_t *os; | |
1534 | dmu_tx_t *tx; | |
1535 | int error; | |
1536 | uid_t uid; | |
1537 | gid_t gid; | |
1538 | zfs_acl_ids_t acl_ids; | |
1539 | boolean_t fuid_dirtied; | |
1540 | boolean_t have_acl = B_FALSE; | |
1541 | boolean_t waited = B_FALSE; | |
1542 | ||
1543 | /* | |
1544 | * If we have an ephemeral id, ACL, or XVATTR then | |
1545 | * make sure file system is at proper version | |
1546 | */ | |
1547 | ||
1548 | gid = crgetgid(cr); | |
1549 | uid = crgetuid(cr); | |
1550 | ||
1551 | if (zfsvfs->z_use_fuids == B_FALSE && | |
1552 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
1553 | return (SET_ERROR(EINVAL)); | |
1554 | ||
1555 | ZFS_ENTER(zfsvfs); | |
1556 | ZFS_VERIFY_ZP(dzp); | |
1557 | os = zfsvfs->z_os; | |
1558 | ||
1559 | if (vap->va_mask & ATTR_XVATTR) { | |
1560 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
1561 | crgetuid(cr), cr, vap->va_mode)) != 0) { | |
1562 | ZFS_EXIT(zfsvfs); | |
1563 | return (error); | |
1564 | } | |
1565 | } | |
1566 | ||
1567 | top: | |
1568 | *ipp = NULL; | |
1569 | ||
1570 | /* | |
1571 | * Create a new file object and update the directory | |
1572 | * to reference it. | |
1573 | */ | |
1574 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { | |
1575 | if (have_acl) | |
1576 | zfs_acl_ids_free(&acl_ids); | |
1577 | goto out; | |
1578 | } | |
1579 | ||
1580 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, | |
1581 | cr, vsecp, &acl_ids)) != 0) | |
1582 | goto out; | |
1583 | have_acl = B_TRUE; | |
1584 | ||
1585 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { | |
1586 | zfs_acl_ids_free(&acl_ids); | |
1587 | error = SET_ERROR(EDQUOT); | |
1588 | goto out; | |
1589 | } | |
1590 | ||
1591 | tx = dmu_tx_create(os); | |
1592 | ||
1593 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1594 | ZFS_SA_BASE_ATTR_SIZE); | |
1595 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); | |
1596 | ||
1597 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
1598 | if (fuid_dirtied) | |
1599 | zfs_fuid_txhold(zfsvfs, tx); | |
1600 | if (!zfsvfs->z_use_sa && | |
1601 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
1602 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
1603 | 0, acl_ids.z_aclp->z_acl_bytes); | |
1604 | } | |
1605 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); | |
1606 | if (error) { | |
1607 | if (error == ERESTART) { | |
1608 | waited = B_TRUE; | |
1609 | dmu_tx_wait(tx); | |
1610 | dmu_tx_abort(tx); | |
1611 | goto top; | |
1612 | } | |
1613 | zfs_acl_ids_free(&acl_ids); | |
1614 | dmu_tx_abort(tx); | |
1615 | ZFS_EXIT(zfsvfs); | |
1616 | return (error); | |
1617 | } | |
1618 | zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); | |
1619 | ||
1620 | if (fuid_dirtied) | |
1621 | zfs_fuid_sync(zfsvfs, tx); | |
1622 | ||
1623 | /* Add to unlinked set */ | |
1624 | zp->z_unlinked = 1; | |
1625 | zfs_unlinked_add(zp, tx); | |
1626 | zfs_acl_ids_free(&acl_ids); | |
1627 | dmu_tx_commit(tx); | |
1628 | out: | |
1629 | ||
1630 | if (error) { | |
1631 | if (zp) | |
1632 | iput(ZTOI(zp)); | |
1633 | } else { | |
1634 | zfs_inode_update(dzp); | |
1635 | zfs_inode_update(zp); | |
1636 | *ipp = ZTOI(zp); | |
1637 | } | |
1638 | ||
1639 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
1640 | return (error); |
1641 | } | |
1642 | ||
1643 | /* | |
1644 | * Remove an entry from a directory. | |
1645 | * | |
3558fd73 | 1646 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1647 | * name - name of entry to remove. |
1648 | * cr - credentials of caller. | |
34dc7c2f BB |
1649 | * |
1650 | * RETURN: 0 if success | |
1651 | * error code if failure | |
1652 | * | |
1653 | * Timestamps: | |
3558fd73 BB |
1654 | * dip - ctime|mtime |
1655 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1656 | */ |
428870ff BB |
1657 | |
1658 | uint64_t null_xattr = 0; | |
1659 | ||
34dc7c2f | 1660 | /*ARGSUSED*/ |
e5c39b95 | 1661 | int |
cae5b340 | 1662 | zfs_remove(struct inode *dip, char *name, cred_t *cr, int flags) |
34dc7c2f | 1663 | { |
3558fd73 | 1664 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1665 | znode_t *xzp; |
3558fd73 | 1666 | struct inode *ip; |
cae5b340 | 1667 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 1668 | zilog_t *zilog; |
cae5b340 | 1669 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 1670 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 1671 | uint64_t obj = 0; |
cae5b340 | 1672 | uint64_t links; |
34dc7c2f BB |
1673 | zfs_dirlock_t *dl; |
1674 | dmu_tx_t *tx; | |
cae5b340 AX |
1675 | boolean_t may_delete_now, delete_now = FALSE; |
1676 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
1677 | uint64_t txtype; |
1678 | pathname_t *realnmp = NULL; | |
1679 | pathname_t realnm; | |
1680 | int error; | |
1681 | int zflg = ZEXISTS; | |
a08ee875 | 1682 | boolean_t waited = B_FALSE; |
34dc7c2f | 1683 | |
cae5b340 AX |
1684 | if (name == NULL) |
1685 | return (SET_ERROR(EINVAL)); | |
1686 | ||
1687 | ZFS_ENTER(zfsvfs); | |
34dc7c2f | 1688 | ZFS_VERIFY_ZP(dzp); |
cae5b340 | 1689 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1690 | |
1691 | if (flags & FIGNORECASE) { | |
1692 | zflg |= ZCILOOK; | |
1693 | pn_alloc(&realnm); | |
1694 | realnmp = &realnm; | |
1695 | } | |
1696 | ||
1697 | top: | |
572e2857 BB |
1698 | xattr_obj = 0; |
1699 | xzp = NULL; | |
34dc7c2f BB |
1700 | /* |
1701 | * Attempt to lock directory; fail if entry doesn't exist. | |
1702 | */ | |
149e873a BB |
1703 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1704 | NULL, realnmp))) { | |
34dc7c2f BB |
1705 | if (realnmp) |
1706 | pn_free(realnmp); | |
cae5b340 | 1707 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1708 | return (error); |
1709 | } | |
1710 | ||
3558fd73 | 1711 | ip = ZTOI(zp); |
34dc7c2f | 1712 | |
149e873a | 1713 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1714 | goto out; |
1715 | } | |
1716 | ||
1717 | /* | |
1718 | * Need to use rmdir for removing directories. | |
1719 | */ | |
3558fd73 | 1720 | if (S_ISDIR(ip->i_mode)) { |
a08ee875 | 1721 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1722 | goto out; |
1723 | } | |
1724 | ||
3558fd73 | 1725 | #ifdef HAVE_DNLC |
34dc7c2f BB |
1726 | if (realnmp) |
1727 | dnlc_remove(dvp, realnmp->pn_buf); | |
1728 | else | |
1729 | dnlc_remove(dvp, name); | |
3558fd73 | 1730 | #endif /* HAVE_DNLC */ |
34dc7c2f | 1731 | |
cae5b340 AX |
1732 | mutex_enter(&zp->z_lock); |
1733 | may_delete_now = atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped); | |
1734 | mutex_exit(&zp->z_lock); | |
1735 | ||
34dc7c2f | 1736 | /* |
cae5b340 AX |
1737 | * We may delete the znode now, or we may put it in the unlinked set; |
1738 | * it depends on whether we're the last link, and on whether there are | |
1739 | * other holds on the inode. So we dmu_tx_hold() the right things to | |
1740 | * allow for either case. | |
34dc7c2f | 1741 | */ |
572e2857 | 1742 | obj = zp->z_id; |
cae5b340 | 1743 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 1744 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1745 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1746 | zfs_sa_upgrade_txholds(tx, zp); | |
1747 | zfs_sa_upgrade_txholds(tx, dzp); | |
cae5b340 AX |
1748 | if (may_delete_now) { |
1749 | toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; | |
1750 | /* if the file is too big, only hold_free a token amount */ | |
1751 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1752 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1753 | } | |
34dc7c2f BB |
1754 | |
1755 | /* are there any extended attributes? */ | |
cae5b340 | 1756 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1757 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1758 | if (error == 0 && xattr_obj) { |
cae5b340 | 1759 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
c06d4368 | 1760 | ASSERT0(error); |
428870ff BB |
1761 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1762 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1763 | } |
1764 | ||
cae5b340 AX |
1765 | mutex_enter(&zp->z_lock); |
1766 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
1767 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
1768 | mutex_exit(&zp->z_lock); | |
1769 | ||
34dc7c2f | 1770 | /* charge as an update -- would be nice not to charge at all */ |
cae5b340 AX |
1771 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
1772 | ||
1773 | /* | |
1774 | * Mark this transaction as typically resulting in a net free of space | |
1775 | */ | |
1776 | dmu_tx_mark_netfree(tx); | |
34dc7c2f | 1777 | |
a08ee875 | 1778 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1779 | if (error) { |
1780 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1781 | if (error == ERESTART) { |
a08ee875 | 1782 | waited = B_TRUE; |
34dc7c2f BB |
1783 | dmu_tx_wait(tx); |
1784 | dmu_tx_abort(tx); | |
22929307 AX |
1785 | iput(ip); |
1786 | if (xzp) | |
1787 | iput(ZTOI(xzp)); | |
34dc7c2f BB |
1788 | goto top; |
1789 | } | |
1790 | if (realnmp) | |
1791 | pn_free(realnmp); | |
1792 | dmu_tx_abort(tx); | |
22929307 AX |
1793 | iput(ip); |
1794 | if (xzp) | |
1795 | iput(ZTOI(xzp)); | |
cae5b340 | 1796 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1797 | return (error); |
1798 | } | |
1799 | ||
1800 | /* | |
1801 | * Remove the directory entry. | |
1802 | */ | |
1803 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1804 | ||
1805 | if (error) { | |
1806 | dmu_tx_commit(tx); | |
1807 | goto out; | |
1808 | } | |
1809 | ||
1810 | if (unlinked) { | |
572e2857 BB |
1811 | /* |
1812 | * Hold z_lock so that we can make sure that the ACL obj | |
1813 | * hasn't changed. Could have been deleted due to | |
1814 | * zfs_sa_upgrade(). | |
1815 | */ | |
1816 | mutex_enter(&zp->z_lock); | |
cae5b340 | 1817 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1818 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
cae5b340 AX |
1819 | delete_now = may_delete_now && !toobig && |
1820 | atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped) && | |
1821 | xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == | |
1822 | acl_obj; | |
1823 | } | |
1824 | ||
1825 | if (delete_now) { | |
1826 | if (xattr_obj_unlinked) { | |
1827 | ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); | |
1828 | mutex_enter(&xzp->z_lock); | |
1829 | xzp->z_unlinked = 1; | |
1830 | clear_nlink(ZTOI(xzp)); | |
1831 | links = 0; | |
1832 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), | |
1833 | &links, sizeof (links), tx); | |
1834 | ASSERT3U(error, ==, 0); | |
1835 | mutex_exit(&xzp->z_lock); | |
1836 | zfs_unlinked_add(xzp, tx); | |
1837 | ||
1838 | if (zp->z_is_sa) | |
1839 | error = sa_remove(zp->z_sa_hdl, | |
1840 | SA_ZPL_XATTR(zfsvfs), tx); | |
1841 | else | |
1842 | error = sa_update(zp->z_sa_hdl, | |
1843 | SA_ZPL_XATTR(zfsvfs), &null_xattr, | |
1844 | sizeof (uint64_t), tx); | |
1845 | ASSERT0(error); | |
1846 | } | |
1847 | /* | |
1848 | * Add to the unlinked set because a new reference could be | |
1849 | * taken concurrently resulting in a deferred destruction. | |
1850 | */ | |
1851 | zfs_unlinked_add(zp, tx); | |
1852 | mutex_exit(&zp->z_lock); | |
1853 | } else if (unlinked) { | |
572e2857 | 1854 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1855 | zfs_unlinked_add(zp, tx); |
1856 | } | |
1857 | ||
1858 | txtype = TX_REMOVE; | |
1859 | if (flags & FIGNORECASE) | |
1860 | txtype |= TX_CI; | |
572e2857 | 1861 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1862 | |
1863 | dmu_tx_commit(tx); | |
1864 | out: | |
1865 | if (realnmp) | |
1866 | pn_free(realnmp); | |
1867 | ||
1868 | zfs_dirent_unlock(dl); | |
960e08fe BB |
1869 | zfs_inode_update(dzp); |
1870 | zfs_inode_update(zp); | |
34dc7c2f | 1871 | |
cae5b340 AX |
1872 | if (delete_now) |
1873 | iput(ip); | |
1874 | else | |
1875 | zfs_iput_async(ip); | |
1876 | ||
1877 | if (xzp) { | |
1878 | zfs_inode_update(xzp); | |
1879 | zfs_iput_async(ZTOI(xzp)); | |
1880 | } | |
428870ff | 1881 | |
cae5b340 | 1882 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1883 | zil_commit(zilog, 0); |
34dc7c2f | 1884 | |
cae5b340 | 1885 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1886 | return (error); |
1887 | } | |
1888 | ||
1889 | /* | |
3558fd73 | 1890 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1891 | * provided. Return a pointer to the inserted directory. |
1892 | * | |
3558fd73 | 1893 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1894 | * dirname - name of new directory. |
1895 | * vap - attributes of new directory. | |
1896 | * cr - credentials of caller. | |
34dc7c2f BB |
1897 | * vsecp - ACL to be set |
1898 | * | |
3558fd73 | 1899 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1900 | * |
1901 | * RETURN: 0 if success | |
1902 | * error code if failure | |
1903 | * | |
1904 | * Timestamps: | |
3558fd73 BB |
1905 | * dip - ctime|mtime updated |
1906 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1907 | */ |
1908 | /*ARGSUSED*/ | |
e5c39b95 | 1909 | int |
3558fd73 BB |
1910 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1911 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1912 | { |
3558fd73 | 1913 | znode_t *zp, *dzp = ITOZ(dip); |
cae5b340 | 1914 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1915 | zilog_t *zilog; |
1916 | zfs_dirlock_t *dl; | |
1917 | uint64_t txtype; | |
1918 | dmu_tx_t *tx; | |
1919 | int error; | |
34dc7c2f | 1920 | int zf = ZNEW; |
b128c09f BB |
1921 | uid_t uid; |
1922 | gid_t gid = crgetgid(cr); | |
428870ff | 1923 | zfs_acl_ids_t acl_ids; |
9babb374 | 1924 | boolean_t fuid_dirtied; |
a08ee875 | 1925 | boolean_t waited = B_FALSE; |
34dc7c2f | 1926 | |
3558fd73 | 1927 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1928 | |
1929 | /* | |
1930 | * If we have an ephemeral id, ACL, or XVATTR then | |
1931 | * make sure file system is at proper version | |
1932 | */ | |
1933 | ||
3558fd73 | 1934 | uid = crgetuid(cr); |
cae5b340 | 1935 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1936 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
a08ee875 | 1937 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1938 | |
cae5b340 AX |
1939 | if (dirname == NULL) |
1940 | return (SET_ERROR(EINVAL)); | |
1941 | ||
1942 | ZFS_ENTER(zfsvfs); | |
34dc7c2f | 1943 | ZFS_VERIFY_ZP(dzp); |
cae5b340 | 1944 | zilog = zfsvfs->z_log; |
34dc7c2f | 1945 | |
428870ff | 1946 | if (dzp->z_pflags & ZFS_XATTR) { |
cae5b340 | 1947 | ZFS_EXIT(zfsvfs); |
a08ee875 | 1948 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1949 | } |
1950 | ||
cae5b340 | 1951 | if (zfsvfs->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1952 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
cae5b340 | 1953 | ZFS_EXIT(zfsvfs); |
a08ee875 | 1954 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1955 | } |
1956 | if (flags & FIGNORECASE) | |
1957 | zf |= ZCILOOK; | |
1958 | ||
5484965a | 1959 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1960 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1961 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
cae5b340 | 1962 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1963 | return (error); |
1964 | } | |
428870ff | 1965 | } |
34dc7c2f | 1966 | |
428870ff BB |
1967 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1968 | vsecp, &acl_ids)) != 0) { | |
cae5b340 | 1969 | ZFS_EXIT(zfsvfs); |
428870ff BB |
1970 | return (error); |
1971 | } | |
34dc7c2f BB |
1972 | /* |
1973 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1974 | * |
1975 | * Existence is checked first to make sure we don't return | |
1976 | * EACCES instead of EEXIST which can cause some applications | |
1977 | * to fail. | |
34dc7c2f BB |
1978 | */ |
1979 | top: | |
3558fd73 | 1980 | *ipp = NULL; |
34dc7c2f | 1981 | |
149e873a BB |
1982 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1983 | NULL, NULL))) { | |
428870ff | 1984 | zfs_acl_ids_free(&acl_ids); |
cae5b340 | 1985 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1986 | return (error); |
1987 | } | |
1988 | ||
149e873a | 1989 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 1990 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1991 | zfs_dirent_unlock(dl); |
cae5b340 | 1992 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1993 | return (error); |
1994 | } | |
1995 | ||
cae5b340 | 1996 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
45d1cae3 | 1997 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 1998 | zfs_dirent_unlock(dl); |
cae5b340 | 1999 | ZFS_EXIT(zfsvfs); |
a08ee875 | 2000 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
2001 | } |
2002 | ||
34dc7c2f BB |
2003 | /* |
2004 | * Add a new entry to the directory. | |
2005 | */ | |
cae5b340 | 2006 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f BB |
2007 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
2008 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
cae5b340 | 2009 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 2010 | if (fuid_dirtied) |
cae5b340 AX |
2011 | zfs_fuid_txhold(zfsvfs, tx); |
2012 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
2013 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
2014 | acl_ids.z_aclp->z_acl_bytes); | |
2015 | } | |
2016 | ||
2017 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
2018 | ZFS_SA_BASE_ATTR_SIZE); | |
2019 | ||
a08ee875 | 2020 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
2021 | if (error) { |
2022 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2023 | if (error == ERESTART) { |
a08ee875 | 2024 | waited = B_TRUE; |
34dc7c2f BB |
2025 | dmu_tx_wait(tx); |
2026 | dmu_tx_abort(tx); | |
2027 | goto top; | |
2028 | } | |
428870ff | 2029 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2030 | dmu_tx_abort(tx); |
cae5b340 | 2031 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2032 | return (error); |
2033 | } | |
2034 | ||
2035 | /* | |
2036 | * Create new node. | |
2037 | */ | |
428870ff | 2038 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f | 2039 | |
9babb374 | 2040 | if (fuid_dirtied) |
cae5b340 | 2041 | zfs_fuid_sync(zfsvfs, tx); |
428870ff | 2042 | |
34dc7c2f BB |
2043 | /* |
2044 | * Now put new name in parent dir. | |
2045 | */ | |
2046 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
2047 | ||
3558fd73 | 2048 | *ipp = ZTOI(zp); |
34dc7c2f BB |
2049 | |
2050 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
2051 | if (flags & FIGNORECASE) | |
2052 | txtype |= TX_CI; | |
9babb374 BB |
2053 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
2054 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 2055 | |
9babb374 | 2056 | zfs_acl_ids_free(&acl_ids); |
428870ff | 2057 | |
34dc7c2f BB |
2058 | dmu_tx_commit(tx); |
2059 | ||
2060 | zfs_dirent_unlock(dl); | |
2061 | ||
cae5b340 | 2062 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2063 | zil_commit(zilog, 0); |
428870ff | 2064 | |
960e08fe BB |
2065 | zfs_inode_update(dzp); |
2066 | zfs_inode_update(zp); | |
cae5b340 | 2067 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2068 | return (0); |
2069 | } | |
2070 | ||
2071 | /* | |
2072 | * Remove a directory subdir entry. If the current working | |
2073 | * directory is the same as the subdir to be removed, the | |
2074 | * remove will fail. | |
2075 | * | |
3558fd73 | 2076 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 2077 | * name - name of directory to be removed. |
3558fd73 | 2078 | * cwd - inode of current working directory. |
34dc7c2f | 2079 | * cr - credentials of caller. |
34dc7c2f BB |
2080 | * flags - case flags |
2081 | * | |
a08ee875 | 2082 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
2083 | * |
2084 | * Timestamps: | |
3558fd73 | 2085 | * dip - ctime|mtime updated |
34dc7c2f BB |
2086 | */ |
2087 | /*ARGSUSED*/ | |
e5c39b95 | 2088 | int |
3558fd73 BB |
2089 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
2090 | int flags) | |
34dc7c2f | 2091 | { |
3558fd73 | 2092 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 2093 | znode_t *zp; |
3558fd73 | 2094 | struct inode *ip; |
cae5b340 | 2095 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
2096 | zilog_t *zilog; |
2097 | zfs_dirlock_t *dl; | |
2098 | dmu_tx_t *tx; | |
2099 | int error; | |
2100 | int zflg = ZEXISTS; | |
a08ee875 | 2101 | boolean_t waited = B_FALSE; |
34dc7c2f | 2102 | |
cae5b340 AX |
2103 | if (name == NULL) |
2104 | return (SET_ERROR(EINVAL)); | |
2105 | ||
2106 | ZFS_ENTER(zfsvfs); | |
34dc7c2f | 2107 | ZFS_VERIFY_ZP(dzp); |
cae5b340 | 2108 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2109 | |
2110 | if (flags & FIGNORECASE) | |
2111 | zflg |= ZCILOOK; | |
2112 | top: | |
2113 | zp = NULL; | |
2114 | ||
2115 | /* | |
2116 | * Attempt to lock directory; fail if entry doesn't exist. | |
2117 | */ | |
149e873a BB |
2118 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
2119 | NULL, NULL))) { | |
cae5b340 | 2120 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2121 | return (error); |
2122 | } | |
2123 | ||
3558fd73 | 2124 | ip = ZTOI(zp); |
34dc7c2f | 2125 | |
149e873a | 2126 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
2127 | goto out; |
2128 | } | |
2129 | ||
3558fd73 | 2130 | if (!S_ISDIR(ip->i_mode)) { |
a08ee875 | 2131 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
2132 | goto out; |
2133 | } | |
2134 | ||
3558fd73 | 2135 | if (ip == cwd) { |
a08ee875 | 2136 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
2137 | goto out; |
2138 | } | |
2139 | ||
34dc7c2f | 2140 | /* |
cae5b340 | 2141 | * Grab a lock on the directory to make sure that no one is |
34dc7c2f BB |
2142 | * trying to add (or lookup) entries while we are removing it. |
2143 | */ | |
2144 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
2145 | ||
2146 | /* | |
2147 | * Grab a lock on the parent pointer to make sure we play well | |
2148 | * with the treewalk and directory rename code. | |
2149 | */ | |
2150 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
2151 | ||
cae5b340 | 2152 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 2153 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 2154 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
cae5b340 | 2155 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
2156 | zfs_sa_upgrade_txholds(tx, zp); |
2157 | zfs_sa_upgrade_txholds(tx, dzp); | |
cae5b340 | 2158 | dmu_tx_mark_netfree(tx); |
a08ee875 | 2159 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
2160 | if (error) { |
2161 | rw_exit(&zp->z_parent_lock); | |
2162 | rw_exit(&zp->z_name_lock); | |
2163 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2164 | if (error == ERESTART) { |
a08ee875 | 2165 | waited = B_TRUE; |
34dc7c2f BB |
2166 | dmu_tx_wait(tx); |
2167 | dmu_tx_abort(tx); | |
22929307 | 2168 | iput(ip); |
34dc7c2f BB |
2169 | goto top; |
2170 | } | |
2171 | dmu_tx_abort(tx); | |
22929307 | 2172 | iput(ip); |
cae5b340 | 2173 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2174 | return (error); |
2175 | } | |
2176 | ||
2177 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
2178 | ||
2179 | if (error == 0) { | |
2180 | uint64_t txtype = TX_RMDIR; | |
2181 | if (flags & FIGNORECASE) | |
2182 | txtype |= TX_CI; | |
572e2857 | 2183 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
2184 | } |
2185 | ||
2186 | dmu_tx_commit(tx); | |
2187 | ||
2188 | rw_exit(&zp->z_parent_lock); | |
2189 | rw_exit(&zp->z_name_lock); | |
2190 | out: | |
2191 | zfs_dirent_unlock(dl); | |
2192 | ||
59157910 BB |
2193 | zfs_inode_update(dzp); |
2194 | zfs_inode_update(zp); | |
3558fd73 | 2195 | iput(ip); |
34dc7c2f | 2196 | |
cae5b340 | 2197 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2198 | zil_commit(zilog, 0); |
428870ff | 2199 | |
cae5b340 | 2200 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2201 | return (error); |
2202 | } | |
2203 | ||
2204 | /* | |
2205 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 2206 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 2207 | * |
3558fd73 BB |
2208 | * IN: ip - inode of directory to read. |
2209 | * dirent - buffer for directory entries. | |
34dc7c2f | 2210 | * |
3558fd73 | 2211 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
2212 | * |
2213 | * RETURN: 0 if success | |
2214 | * error code if failure | |
2215 | * | |
2216 | * Timestamps: | |
3558fd73 | 2217 | * ip - atime updated |
34dc7c2f BB |
2218 | * |
2219 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2220 | * This allows us to use the low range for "special" directory entries: | |
2221 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2222 | * we use the offset 2 for the '.zfs' directory. | |
2223 | */ | |
2224 | /* ARGSUSED */ | |
3558fd73 | 2225 | int |
c06d4368 | 2226 | zfs_readdir(struct inode *ip, struct dir_context *ctx, cred_t *cr) |
34dc7c2f | 2227 | { |
3558fd73 | 2228 | znode_t *zp = ITOZ(ip); |
cae5b340 | 2229 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2230 | objset_t *os; |
34dc7c2f BB |
2231 | zap_cursor_t zc; |
2232 | zap_attribute_t zap; | |
34dc7c2f BB |
2233 | int error; |
2234 | uint8_t prefetch; | |
a08ee875 | 2235 | uint8_t type; |
3558fd73 BB |
2236 | int done = 0; |
2237 | uint64_t parent; | |
a08ee875 | 2238 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2239 | |
cae5b340 | 2240 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2241 | ZFS_VERIFY_ZP(zp); |
2242 | ||
cae5b340 | 2243 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
3558fd73 BB |
2244 | &parent, sizeof (parent))) != 0) |
2245 | goto out; | |
34dc7c2f BB |
2246 | |
2247 | /* | |
2248 | * Quit if directory has been removed (posix) | |
2249 | */ | |
3558fd73 BB |
2250 | if (zp->z_unlinked) |
2251 | goto out; | |
2252 | ||
a08ee875 | 2253 | error = 0; |
cae5b340 | 2254 | os = zfsvfs->z_os; |
a08ee875 | 2255 | offset = ctx->pos; |
34dc7c2f BB |
2256 | prefetch = zp->z_zn_prefetch; |
2257 | ||
2258 | /* | |
2259 | * Initialize the iterator cursor. | |
2260 | */ | |
a08ee875 | 2261 | if (offset <= 3) { |
34dc7c2f BB |
2262 | /* |
2263 | * Start iteration from the beginning of the directory. | |
2264 | */ | |
2265 | zap_cursor_init(&zc, os, zp->z_id); | |
2266 | } else { | |
2267 | /* | |
2268 | * The offset is a serialized cursor. | |
2269 | */ | |
a08ee875 | 2270 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2271 | } |
2272 | ||
34dc7c2f BB |
2273 | /* |
2274 | * Transform to file-system independent format | |
2275 | */ | |
3558fd73 BB |
2276 | while (!done) { |
2277 | uint64_t objnum; | |
34dc7c2f BB |
2278 | /* |
2279 | * Special case `.', `..', and `.zfs'. | |
2280 | */ | |
a08ee875 | 2281 | if (offset == 0) { |
34dc7c2f BB |
2282 | (void) strcpy(zap.za_name, "."); |
2283 | zap.za_normalization_conflict = 0; | |
2284 | objnum = zp->z_id; | |
a08ee875 LG |
2285 | type = DT_DIR; |
2286 | } else if (offset == 1) { | |
34dc7c2f BB |
2287 | (void) strcpy(zap.za_name, ".."); |
2288 | zap.za_normalization_conflict = 0; | |
428870ff | 2289 | objnum = parent; |
a08ee875 LG |
2290 | type = DT_DIR; |
2291 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2292 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2293 | zap.za_normalization_conflict = 0; | |
2294 | objnum = ZFSCTL_INO_ROOT; | |
a08ee875 | 2295 | type = DT_DIR; |
34dc7c2f BB |
2296 | } else { |
2297 | /* | |
2298 | * Grab next entry. | |
2299 | */ | |
3558fd73 BB |
2300 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2301 | if (error == ENOENT) | |
34dc7c2f BB |
2302 | break; |
2303 | else | |
2304 | goto update; | |
2305 | } | |
2306 | ||
0c5dde49 BB |
2307 | /* |
2308 | * Allow multiple entries provided the first entry is | |
2309 | * the object id. Non-zpl consumers may safely make | |
2310 | * use of the additional space. | |
2311 | * | |
2312 | * XXX: This should be a feature flag for compatibility | |
2313 | */ | |
34dc7c2f | 2314 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2315 | zap.za_num_integers == 0) { |
34dc7c2f | 2316 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2317 | "entry, obj = %lld, offset = %lld, " |
2318 | "length = %d, num = %lld\n", | |
34dc7c2f | 2319 | (u_longlong_t)zp->z_id, |
a08ee875 | 2320 | (u_longlong_t)offset, |
0c5dde49 BB |
2321 | zap.za_integer_length, |
2322 | (u_longlong_t)zap.za_num_integers); | |
a08ee875 | 2323 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2324 | goto update; |
2325 | } | |
2326 | ||
2327 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
a08ee875 | 2328 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2329 | } |
c06d4368 AX |
2330 | |
2331 | done = !dir_emit(ctx, zap.za_name, strlen(zap.za_name), | |
a08ee875 | 2332 | objnum, type); |
c06d4368 | 2333 | if (done) |
34dc7c2f | 2334 | break; |
34dc7c2f BB |
2335 | |
2336 | /* Prefetch znode */ | |
3558fd73 | 2337 | if (prefetch) { |
cae5b340 AX |
2338 | dmu_prefetch(os, objnum, 0, 0, 0, |
2339 | ZIO_PRIORITY_SYNC_READ); | |
3558fd73 | 2340 | } |
34dc7c2f | 2341 | |
a08ee875 LG |
2342 | /* |
2343 | * Move to the next entry, fill in the previous offset. | |
2344 | */ | |
2345 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2346 | zap_cursor_advance(&zc); |
a08ee875 | 2347 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2348 | } else { |
a08ee875 | 2349 | offset += 1; |
34dc7c2f | 2350 | } |
a08ee875 | 2351 | ctx->pos = offset; |
34dc7c2f BB |
2352 | } |
2353 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2354 | ||
34dc7c2f BB |
2355 | update: |
2356 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2357 | if (error == ENOENT) |
2358 | error = 0; | |
3558fd73 | 2359 | out: |
cae5b340 | 2360 | ZFS_EXIT(zfsvfs); |
34dc7c2f | 2361 | |
34dc7c2f BB |
2362 | return (error); |
2363 | } | |
2364 | ||
d5446cfc BB |
2365 | ulong_t zfs_fsync_sync_cnt = 4; |
2366 | ||
e5c39b95 | 2367 | int |
3558fd73 | 2368 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2369 | { |
3558fd73 | 2370 | znode_t *zp = ITOZ(ip); |
cae5b340 | 2371 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2372 | |
d5446cfc BB |
2373 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2374 | ||
cae5b340 AX |
2375 | if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2376 | ZFS_ENTER(zfsvfs); | |
428870ff | 2377 | ZFS_VERIFY_ZP(zp); |
cae5b340 AX |
2378 | zil_commit(zfsvfs->z_log, zp->z_id); |
2379 | ZFS_EXIT(zfsvfs); | |
428870ff | 2380 | } |
e10b0808 AX |
2381 | tsd_set(zfs_fsyncer_key, NULL); |
2382 | ||
34dc7c2f BB |
2383 | return (0); |
2384 | } | |
2385 | ||
2386 | ||
2387 | /* | |
2388 | * Get the requested file attributes and place them in the provided | |
2389 | * vattr structure. | |
2390 | * | |
3558fd73 | 2391 | * IN: ip - inode of file. |
5484965a BB |
2392 | * vap - va_mask identifies requested attributes. |
2393 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2394 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2395 | * cr - credentials of caller. | |
34dc7c2f | 2396 | * |
5484965a BB |
2397 | * OUT: vap - attribute values. |
2398 | * | |
2399 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2400 | */ |
2401 | /* ARGSUSED */ | |
e5c39b95 | 2402 | int |
5484965a | 2403 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2404 | { |
3558fd73 | 2405 | znode_t *zp = ITOZ(ip); |
cae5b340 | 2406 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
2407 | int error = 0; |
2408 | uint64_t links; | |
68d83c55 | 2409 | uint64_t atime[2], mtime[2], ctime[2]; |
5484965a BB |
2410 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2411 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2412 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
68d83c55 | 2413 | sa_bulk_attr_t bulk[3]; |
428870ff | 2414 | int count = 0; |
34dc7c2f | 2415 | |
cae5b340 | 2416 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2417 | ZFS_VERIFY_ZP(zp); |
428870ff | 2418 | |
5484965a | 2419 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2420 | |
cae5b340 AX |
2421 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); |
2422 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
2423 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
428870ff BB |
2424 | |
2425 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
cae5b340 | 2426 | ZFS_EXIT(zfsvfs); |
428870ff BB |
2427 | return (error); |
2428 | } | |
34dc7c2f | 2429 | |
34dc7c2f BB |
2430 | /* |
2431 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2432 | * Also, if we are the owner don't bother, since owner should | |
2433 | * always be allowed to read basic attributes of file. | |
2434 | */ | |
572e2857 | 2435 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2436 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2437 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2438 | skipaclchk, cr))) { | |
cae5b340 | 2439 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2440 | return (error); |
2441 | } | |
2442 | } | |
2443 | ||
2444 | /* | |
2445 | * Return all attributes. It's cheaper to provide the answer | |
2446 | * than to determine whether we were asked the question. | |
2447 | */ | |
2448 | ||
9babb374 | 2449 | mutex_enter(&zp->z_lock); |
5484965a BB |
2450 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2451 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2452 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2453 | vap->va_nodeid = zp->z_id; |
cae5b340 AX |
2454 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) |
2455 | links = ZTOI(zp)->i_nlink + 1; | |
34dc7c2f | 2456 | else |
cae5b340 | 2457 | links = ZTOI(zp)->i_nlink; |
5484965a BB |
2458 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2459 | vap->va_size = i_size_read(ip); | |
2460 | vap->va_rdev = ip->i_rdev; | |
2461 | vap->va_seq = ip->i_generation; | |
2462 | ||
2463 | /* | |
2464 | * Add in any requested optional attributes and the create time. | |
2465 | * Also set the corresponding bits in the returned attribute bitmap. | |
2466 | */ | |
cae5b340 | 2467 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { |
5484965a BB |
2468 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { |
2469 | xoap->xoa_archive = | |
2470 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2471 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2472 | } | |
2473 | ||
2474 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2475 | xoap->xoa_readonly = | |
2476 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2477 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2478 | } | |
2479 | ||
2480 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2481 | xoap->xoa_system = | |
2482 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2483 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2484 | } | |
2485 | ||
2486 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2487 | xoap->xoa_hidden = | |
2488 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2489 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2490 | } | |
2491 | ||
2492 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2493 | xoap->xoa_nounlink = | |
2494 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2495 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2496 | } | |
2497 | ||
2498 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2499 | xoap->xoa_immutable = | |
2500 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2501 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2502 | } | |
2503 | ||
2504 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2505 | xoap->xoa_appendonly = | |
2506 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2507 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2508 | } | |
2509 | ||
2510 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2511 | xoap->xoa_nodump = | |
2512 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2513 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2514 | } | |
2515 | ||
2516 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2517 | xoap->xoa_opaque = | |
2518 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2519 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2520 | } | |
2521 | ||
2522 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2523 | xoap->xoa_av_quarantined = | |
2524 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2525 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2526 | } | |
2527 | ||
2528 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2529 | xoap->xoa_av_modified = | |
2530 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2531 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2532 | } | |
2533 | ||
2534 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2535 | S_ISREG(ip->i_mode)) { | |
2536 | zfs_sa_get_scanstamp(zp, xvap); | |
2537 | } | |
34dc7c2f | 2538 | |
5484965a BB |
2539 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2540 | uint64_t times[2]; | |
2541 | ||
cae5b340 | 2542 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), |
5484965a BB |
2543 | times, sizeof (times)); |
2544 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2545 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2546 | } | |
2547 | ||
2548 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2549 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2550 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2551 | } | |
2552 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
cae5b340 | 2553 | xoap->xoa_generation = ip->i_generation; |
5484965a BB |
2554 | XVA_SET_RTN(xvap, XAT_GEN); |
2555 | } | |
2556 | ||
2557 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2558 | xoap->xoa_offline = | |
2559 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2560 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2561 | } | |
2562 | ||
2563 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2564 | xoap->xoa_sparse = | |
2565 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2566 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2567 | } | |
2568 | } | |
2569 | ||
68d83c55 | 2570 | ZFS_TIME_DECODE(&vap->va_atime, atime); |
5484965a BB |
2571 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); |
2572 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2573 | |
2574 | mutex_exit(&zp->z_lock); | |
2575 | ||
5484965a | 2576 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2577 | |
2578 | if (zp->z_blksz == 0) { | |
2579 | /* | |
2580 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2581 | */ | |
cae5b340 | 2582 | vap->va_blksize = zfsvfs->z_max_blksz; |
34dc7c2f BB |
2583 | } |
2584 | ||
cae5b340 | 2585 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2586 | return (0); |
2587 | } | |
2588 | ||
057e8eee BB |
2589 | /* |
2590 | * Get the basic file attributes and place them in the provided kstat | |
2591 | * structure. The inode is assumed to be the authoritative source | |
2592 | * for most of the attributes. However, the znode currently has the | |
2593 | * authoritative atime, blksize, and block count. | |
2594 | * | |
2595 | * IN: ip - inode of file. | |
2596 | * | |
2597 | * OUT: sp - kstat values. | |
2598 | * | |
2599 | * RETURN: 0 (always succeeds) | |
2600 | */ | |
2601 | /* ARGSUSED */ | |
2602 | int | |
2603 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2604 | { | |
2605 | znode_t *zp = ITOZ(ip); | |
cae5b340 | 2606 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
a08ee875 LG |
2607 | uint32_t blksize; |
2608 | u_longlong_t nblocks; | |
057e8eee | 2609 | |
cae5b340 | 2610 | ZFS_ENTER(zfsvfs); |
a7b125e9 GB |
2611 | ZFS_VERIFY_ZP(zp); |
2612 | ||
057e8eee BB |
2613 | mutex_enter(&zp->z_lock); |
2614 | ||
2615 | generic_fillattr(ip, sp); | |
057e8eee | 2616 | |
a08ee875 LG |
2617 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2618 | sp->blksize = blksize; | |
2619 | sp->blocks = nblocks; | |
2620 | ||
057e8eee BB |
2621 | if (unlikely(zp->z_blksz == 0)) { |
2622 | /* | |
2623 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2624 | */ | |
cae5b340 | 2625 | sp->blksize = zfsvfs->z_max_blksz; |
057e8eee BB |
2626 | } |
2627 | ||
2628 | mutex_exit(&zp->z_lock); | |
2629 | ||
e10b0808 AX |
2630 | /* |
2631 | * Required to prevent NFS client from detecting different inode | |
2632 | * numbers of snapshot root dentry before and after snapshot mount. | |
2633 | */ | |
cae5b340 | 2634 | if (zfsvfs->z_issnap) { |
e10b0808 AX |
2635 | if (ip->i_sb->s_root->d_inode == ip) |
2636 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
cae5b340 | 2637 | dmu_objset_id(zfsvfs->z_os); |
e10b0808 AX |
2638 | } |
2639 | ||
cae5b340 | 2640 | ZFS_EXIT(zfsvfs); |
a7b125e9 | 2641 | |
057e8eee BB |
2642 | return (0); |
2643 | } | |
057e8eee | 2644 | |
34dc7c2f BB |
2645 | /* |
2646 | * Set the file attributes to the values contained in the | |
2647 | * vattr structure. | |
2648 | * | |
3558fd73 | 2649 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2650 | * vap - new attribute values. |
5484965a | 2651 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2652 | * flags - ATTR_UTIME set if non-default time values provided. |
2653 | * - ATTR_NOACLCHECK (CIFS context only). | |
2654 | * cr - credentials of caller. | |
34dc7c2f BB |
2655 | * |
2656 | * RETURN: 0 if success | |
2657 | * error code if failure | |
2658 | * | |
2659 | * Timestamps: | |
3558fd73 | 2660 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2661 | */ |
2662 | /* ARGSUSED */ | |
e5c39b95 | 2663 | int |
5484965a | 2664 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2665 | { |
3558fd73 | 2666 | znode_t *zp = ITOZ(ip); |
cae5b340 | 2667 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
2668 | zilog_t *zilog; |
2669 | dmu_tx_t *tx; | |
2670 | vattr_t oldva; | |
f4ea75d4 | 2671 | xvattr_t *tmpxvattr; |
5484965a | 2672 | uint_t mask = vap->va_mask; |
a08ee875 | 2673 | uint_t saved_mask = 0; |
34dc7c2f BB |
2674 | int trim_mask = 0; |
2675 | uint64_t new_mode; | |
cae5b340 | 2676 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 2677 | uint64_t xattr_obj; |
68d83c55 | 2678 | uint64_t mtime[2], ctime[2], atime[2]; |
34dc7c2f BB |
2679 | znode_t *attrzp; |
2680 | int need_policy = FALSE; | |
428870ff | 2681 | int err, err2; |
34dc7c2f | 2682 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2683 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2684 | xoptattr_t *xoap; | |
2685 | zfs_acl_t *aclp; | |
34dc7c2f | 2686 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2687 | boolean_t fuid_dirtied = B_FALSE; |
17c37660 | 2688 | sa_bulk_attr_t *bulk, *xattr_bulk; |
428870ff | 2689 | int count = 0, xattr_count = 0; |
34dc7c2f BB |
2690 | |
2691 | if (mask == 0) | |
2692 | return (0); | |
2693 | ||
cae5b340 | 2694 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2695 | ZFS_VERIFY_ZP(zp); |
2696 | ||
cae5b340 | 2697 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2698 | |
2699 | /* | |
2700 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2701 | * that file system is at proper version level | |
2702 | */ | |
5484965a | 2703 | |
cae5b340 | 2704 | if (zfsvfs->z_use_fuids == B_FALSE && |
5484965a BB |
2705 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2706 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2707 | (mask & ATTR_XVATTR))) { | |
cae5b340 | 2708 | ZFS_EXIT(zfsvfs); |
a08ee875 | 2709 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2710 | } |
2711 | ||
3558fd73 | 2712 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
cae5b340 | 2713 | ZFS_EXIT(zfsvfs); |
a08ee875 | 2714 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2715 | } |
2716 | ||
3558fd73 | 2717 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
cae5b340 | 2718 | ZFS_EXIT(zfsvfs); |
a08ee875 | 2719 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2720 | } |
2721 | ||
5484965a BB |
2722 | /* |
2723 | * If this is an xvattr_t, then get a pointer to the structure of | |
2724 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2725 | */ | |
2726 | xoap = xva_getxoptattr(xvap); | |
2727 | ||
a08ee875 | 2728 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2729 | xva_init(tmpxvattr); |
5484965a | 2730 | |
a08ee875 LG |
2731 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); |
2732 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); | |
17c37660 | 2733 | |
5484965a BB |
2734 | /* |
2735 | * Immutable files can only alter immutable bit and atime | |
2736 | */ | |
2737 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2738 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2739 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
f4ea75d4 BB |
2740 | err = EPERM; |
2741 | goto out3; | |
5484965a BB |
2742 | } |
2743 | ||
3558fd73 | 2744 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
f4ea75d4 BB |
2745 | err = EPERM; |
2746 | goto out3; | |
34dc7c2f BB |
2747 | } |
2748 | ||
5484965a BB |
2749 | /* |
2750 | * Verify timestamps doesn't overflow 32 bits. | |
2751 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2752 | * handle times greater than 2039. This check should be removed | |
2753 | * once large timestamps are fully supported. | |
2754 | */ | |
2755 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
a08ee875 LG |
2756 | if (((mask & ATTR_ATIME) && |
2757 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2758 | ((mask & ATTR_MTIME) && | |
2759 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
f4ea75d4 BB |
2760 | err = EOVERFLOW; |
2761 | goto out3; | |
5484965a BB |
2762 | } |
2763 | } | |
2764 | ||
34dc7c2f BB |
2765 | top: |
2766 | attrzp = NULL; | |
572e2857 | 2767 | aclp = NULL; |
34dc7c2f | 2768 | |
45d1cae3 | 2769 | /* Can this be moved to before the top label? */ |
cae5b340 | 2770 | if (zfs_is_readonly(zfsvfs)) { |
f4ea75d4 BB |
2771 | err = EROFS; |
2772 | goto out3; | |
34dc7c2f BB |
2773 | } |
2774 | ||
2775 | /* | |
2776 | * First validate permissions | |
2777 | */ | |
2778 | ||
3558fd73 | 2779 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2780 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2781 | if (err) |
2782 | goto out3; | |
2783 | ||
34dc7c2f BB |
2784 | /* |
2785 | * XXX - Note, we are not providing any open | |
2786 | * mode flags here (like FNDELAY), so we may | |
2787 | * block if there are locks present... this | |
2788 | * should be addressed in openat(). | |
2789 | */ | |
b128c09f | 2790 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2791 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2792 | if (err) |
2793 | goto out3; | |
428870ff | 2794 | } |
34dc7c2f | 2795 | |
5484965a BB |
2796 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2797 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2798 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2799 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2800 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2801 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2802 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2803 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2804 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2805 | skipaclchk, cr); | |
2806 | } | |
2807 | ||
3558fd73 BB |
2808 | if (mask & (ATTR_UID|ATTR_GID)) { |
2809 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2810 | int take_owner; |
2811 | int take_group; | |
2812 | ||
2813 | /* | |
2814 | * NOTE: even if a new mode is being set, | |
2815 | * we may clear S_ISUID/S_ISGID bits. | |
2816 | */ | |
2817 | ||
3558fd73 | 2818 | if (!(mask & ATTR_MODE)) |
5484965a | 2819 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2820 | |
2821 | /* | |
2822 | * Take ownership or chgrp to group we are a member of | |
2823 | */ | |
2824 | ||
5484965a | 2825 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 2826 | take_group = (mask & ATTR_GID) && |
cae5b340 | 2827 | zfs_groupmember(zfsvfs, vap->va_gid, cr); |
34dc7c2f BB |
2828 | |
2829 | /* | |
5484965a | 2830 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
2831 | * take_group must both be set in order to allow taking |
2832 | * ownership. | |
2833 | * | |
2834 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2835 | * | |
2836 | */ | |
2837 | ||
3558fd73 BB |
2838 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
2839 | take_owner && take_group) || | |
2840 | ((idmask == ATTR_UID) && take_owner) || | |
2841 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
2842 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
2843 | skipaclchk, cr) == 0) { | |
2844 | /* | |
2845 | * Remove setuid/setgid for non-privileged users | |
2846 | */ | |
5484965a | 2847 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 2848 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
2849 | } else { |
2850 | need_policy = TRUE; | |
2851 | } | |
2852 | } else { | |
2853 | need_policy = TRUE; | |
2854 | } | |
2855 | } | |
2856 | ||
2857 | mutex_enter(&zp->z_lock); | |
428870ff | 2858 | oldva.va_mode = zp->z_mode; |
572e2857 | 2859 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
2860 | if (mask & ATTR_XVATTR) { |
2861 | /* | |
2862 | * Update xvattr mask to include only those attributes | |
2863 | * that are actually changing. | |
2864 | * | |
2865 | * the bits will be restored prior to actually setting | |
2866 | * the attributes so the caller thinks they were set. | |
2867 | */ | |
2868 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2869 | if (xoap->xoa_appendonly != | |
2870 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
2871 | need_policy = TRUE; | |
2872 | } else { | |
2873 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 2874 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
2875 | } |
2876 | } | |
2877 | ||
2878 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2879 | if (xoap->xoa_nounlink != | |
2880 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
2881 | need_policy = TRUE; | |
2882 | } else { | |
2883 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 2884 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
2885 | } |
2886 | } | |
2887 | ||
2888 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2889 | if (xoap->xoa_immutable != | |
2890 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
2891 | need_policy = TRUE; | |
2892 | } else { | |
2893 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 2894 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
2895 | } |
2896 | } | |
2897 | ||
2898 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2899 | if (xoap->xoa_nodump != | |
2900 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
2901 | need_policy = TRUE; | |
2902 | } else { | |
2903 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 2904 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
2905 | } |
2906 | } | |
2907 | ||
2908 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2909 | if (xoap->xoa_av_modified != | |
2910 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
2911 | need_policy = TRUE; | |
2912 | } else { | |
2913 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 2914 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
2915 | } |
2916 | } | |
2917 | ||
2918 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2919 | if ((!S_ISREG(ip->i_mode) && | |
2920 | xoap->xoa_av_quarantined) || | |
2921 | xoap->xoa_av_quarantined != | |
2922 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
2923 | need_policy = TRUE; | |
2924 | } else { | |
2925 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 2926 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
2927 | } |
2928 | } | |
2929 | ||
2930 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2931 | mutex_exit(&zp->z_lock); | |
f4ea75d4 BB |
2932 | err = EPERM; |
2933 | goto out3; | |
5484965a BB |
2934 | } |
2935 | ||
2936 | if (need_policy == FALSE && | |
2937 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2938 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
2939 | need_policy = TRUE; | |
2940 | } | |
2941 | } | |
34dc7c2f BB |
2942 | |
2943 | mutex_exit(&zp->z_lock); | |
2944 | ||
3558fd73 | 2945 | if (mask & ATTR_MODE) { |
34dc7c2f | 2946 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 2947 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 2948 | &oldva, cr); |
f4ea75d4 BB |
2949 | if (err) |
2950 | goto out3; | |
2951 | ||
3558fd73 | 2952 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
2953 | } else { |
2954 | need_policy = TRUE; | |
2955 | } | |
2956 | } | |
2957 | ||
2958 | if (need_policy) { | |
2959 | /* | |
2960 | * If trim_mask is set then take ownership | |
2961 | * has been granted or write_acl is present and user | |
2962 | * has the ability to modify mode. In that case remove | |
2963 | * UID|GID and or MODE from mask so that | |
2964 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2965 | */ | |
2966 | ||
2967 | if (trim_mask) { | |
5484965a BB |
2968 | saved_mask = vap->va_mask; |
2969 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 2970 | } |
5484965a | 2971 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 2972 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
2973 | if (err) |
2974 | goto out3; | |
34dc7c2f BB |
2975 | |
2976 | if (trim_mask) | |
5484965a | 2977 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
2978 | } |
2979 | ||
2980 | /* | |
2981 | * secpolicy_vnode_setattr, or take ownership may have | |
2982 | * changed va_mask | |
2983 | */ | |
5484965a | 2984 | mask = vap->va_mask; |
34dc7c2f | 2985 | |
3558fd73 | 2986 | if ((mask & (ATTR_UID | ATTR_GID))) { |
cae5b340 | 2987 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
572e2857 | 2988 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 2989 | |
572e2857 | 2990 | if (err == 0 && xattr_obj) { |
3558fd73 | 2991 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
2992 | if (err) |
2993 | goto out2; | |
2994 | } | |
3558fd73 | 2995 | if (mask & ATTR_UID) { |
cae5b340 | 2996 | new_kuid = zfs_fuid_create(zfsvfs, |
5484965a | 2997 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
cae5b340 AX |
2998 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
2999 | zfs_fuid_overquota(zfsvfs, B_FALSE, new_kuid)) { | |
572e2857 | 3000 | if (attrzp) |
3558fd73 | 3001 | iput(ZTOI(attrzp)); |
428870ff BB |
3002 | err = EDQUOT; |
3003 | goto out2; | |
3004 | } | |
3005 | } | |
3006 | ||
3558fd73 | 3007 | if (mask & ATTR_GID) { |
cae5b340 AX |
3008 | new_kgid = zfs_fuid_create(zfsvfs, |
3009 | (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); | |
3010 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && | |
3011 | zfs_fuid_overquota(zfsvfs, B_TRUE, new_kgid)) { | |
572e2857 | 3012 | if (attrzp) |
3558fd73 | 3013 | iput(ZTOI(attrzp)); |
428870ff BB |
3014 | err = EDQUOT; |
3015 | goto out2; | |
3016 | } | |
3017 | } | |
3018 | } | |
cae5b340 | 3019 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 3020 | |
3558fd73 | 3021 | if (mask & ATTR_MODE) { |
428870ff | 3022 | uint64_t pmode = zp->z_mode; |
572e2857 | 3023 | uint64_t acl_obj; |
5484965a | 3024 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 3025 | |
572e2857 | 3026 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 3027 | |
572e2857 BB |
3028 | mutex_enter(&zp->z_lock); |
3029 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
3030 | /* |
3031 | * Are we upgrading ACL from old V0 format | |
3032 | * to V1 format? | |
3033 | */ | |
cae5b340 | 3034 | if (zfsvfs->z_version >= ZPL_VERSION_FUID && |
572e2857 | 3035 | zfs_znode_acl_version(zp) == |
34dc7c2f | 3036 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 3037 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
3038 | DMU_OBJECT_END); |
3039 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
3040 | 0, aclp->z_acl_bytes); | |
3041 | } else { | |
572e2857 | 3042 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
3043 | aclp->z_acl_bytes); |
3044 | } | |
428870ff | 3045 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
3046 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
3047 | 0, aclp->z_acl_bytes); | |
3048 | } | |
572e2857 | 3049 | mutex_exit(&zp->z_lock); |
428870ff BB |
3050 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3051 | } else { | |
5484965a BB |
3052 | if ((mask & ATTR_XVATTR) && |
3053 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3054 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
3055 | else | |
3056 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3057 | } |
3058 | ||
428870ff BB |
3059 | if (attrzp) { |
3060 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3061 | } |
3062 | ||
cae5b340 | 3063 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
428870ff | 3064 | if (fuid_dirtied) |
cae5b340 | 3065 | zfs_fuid_txhold(zfsvfs, tx); |
428870ff BB |
3066 | |
3067 | zfs_sa_upgrade_txholds(tx, zp); | |
3068 | ||
a08ee875 LG |
3069 | err = dmu_tx_assign(tx, TXG_WAIT); |
3070 | if (err) | |
9babb374 | 3071 | goto out; |
34dc7c2f | 3072 | |
428870ff | 3073 | count = 0; |
34dc7c2f BB |
3074 | /* |
3075 | * Set each attribute requested. | |
3076 | * We group settings according to the locks they need to acquire. | |
3077 | * | |
3078 | * Note: you cannot set ctime directly, although it will be | |
3079 | * updated as a side-effect of calling this function. | |
3080 | */ | |
3081 | ||
572e2857 | 3082 | |
3558fd73 | 3083 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3084 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
3085 | mutex_enter(&zp->z_lock); |
3086 | ||
cae5b340 | 3087 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
3088 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3089 | ||
3090 | if (attrzp) { | |
3558fd73 | 3091 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3092 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
3093 | mutex_enter(&attrzp->z_lock); |
3094 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
cae5b340 | 3095 | SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, |
428870ff BB |
3096 | sizeof (attrzp->z_pflags)); |
3097 | } | |
3098 | ||
3558fd73 | 3099 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 3100 | |
3558fd73 | 3101 | if (mask & ATTR_UID) { |
cae5b340 AX |
3102 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
3103 | new_uid = zfs_uid_read(ZTOI(zp)); | |
3104 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, | |
428870ff | 3105 | &new_uid, sizeof (new_uid)); |
428870ff BB |
3106 | if (attrzp) { |
3107 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
cae5b340 | 3108 | SA_ZPL_UID(zfsvfs), NULL, &new_uid, |
428870ff | 3109 | sizeof (new_uid)); |
cae5b340 | 3110 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
3111 | } |
3112 | } | |
3113 | ||
3558fd73 | 3114 | if (mask & ATTR_GID) { |
cae5b340 AX |
3115 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
3116 | new_gid = zfs_gid_read(ZTOI(zp)); | |
3117 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), | |
428870ff | 3118 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
3119 | if (attrzp) { |
3120 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
cae5b340 | 3121 | SA_ZPL_GID(zfsvfs), NULL, &new_gid, |
428870ff | 3122 | sizeof (new_gid)); |
cae5b340 | 3123 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
3124 | } |
3125 | } | |
3558fd73 | 3126 | if (!(mask & ATTR_MODE)) { |
cae5b340 | 3127 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), |
428870ff BB |
3128 | NULL, &new_mode, sizeof (new_mode)); |
3129 | new_mode = zp->z_mode; | |
3130 | } | |
3131 | err = zfs_acl_chown_setattr(zp); | |
3132 | ASSERT(err == 0); | |
3133 | if (attrzp) { | |
3134 | err = zfs_acl_chown_setattr(attrzp); | |
3135 | ASSERT(err == 0); | |
3136 | } | |
3137 | } | |
3138 | ||
3558fd73 | 3139 | if (mask & ATTR_MODE) { |
cae5b340 | 3140 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 3141 | &new_mode, sizeof (new_mode)); |
cae5b340 | 3142 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 3143 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 3144 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c06d4368 | 3145 | ASSERT0(err); |
572e2857 BB |
3146 | if (zp->z_acl_cached) |
3147 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
3148 | zp->z_acl_cached = aclp; |
3149 | aclp = NULL; | |
34dc7c2f BB |
3150 | } |
3151 | ||
68d83c55 AX |
3152 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
3153 | zp->z_atime_dirty = 0; | |
3154 | ZFS_TIME_ENCODE(&ip->i_atime, atime); | |
cae5b340 | 3155 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
68d83c55 | 3156 | &atime, sizeof (atime)); |
34dc7c2f BB |
3157 | } |
3158 | ||
41d74433 | 3159 | if (mask & (ATTR_MTIME | ATTR_SIZE)) { |
5484965a | 3160 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
cae5b340 AX |
3161 | ZTOI(zp)->i_mtime = timespec_trunc(vap->va_mtime, |
3162 | ZTOI(zp)->i_sb->s_time_gran); | |
3163 | ||
3164 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, | |
428870ff | 3165 | mtime, sizeof (mtime)); |
34dc7c2f BB |
3166 | } |
3167 | ||
41d74433 | 3168 | if (mask & (ATTR_CTIME | ATTR_SIZE)) { |
cae5b340 AX |
3169 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); |
3170 | ZTOI(zp)->i_ctime = timespec_trunc(vap->va_ctime, | |
3171 | ZTOI(zp)->i_sb->s_time_gran); | |
3172 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, | |
3173 | ctime, sizeof (ctime)); | |
3174 | } | |
3175 | ||
3176 | if (attrzp && mask) { | |
3177 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3178 | SA_ZPL_CTIME(zfsvfs), NULL, &ctime, | |
3179 | sizeof (ctime)); | |
428870ff | 3180 | } |
cae5b340 | 3181 | |
34dc7c2f BB |
3182 | /* |
3183 | * Do this after setting timestamps to prevent timestamp | |
3184 | * update from toggling bit | |
3185 | */ | |
3186 | ||
5484965a BB |
3187 | if (xoap && (mask & ATTR_XVATTR)) { |
3188 | ||
3189 | /* | |
3190 | * restore trimmed off masks | |
3191 | * so that return masks can be set for caller. | |
3192 | */ | |
3193 | ||
f4ea75d4 | 3194 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
3195 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
3196 | } | |
f4ea75d4 | 3197 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
3198 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
3199 | } | |
f4ea75d4 | 3200 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
3201 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
3202 | } | |
f4ea75d4 | 3203 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
3204 | XVA_SET_REQ(xvap, XAT_NODUMP); |
3205 | } | |
f4ea75d4 | 3206 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
3207 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
3208 | } | |
f4ea75d4 | 3209 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
3210 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
3211 | } | |
3212 | ||
3213 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3214 | ASSERT(S_ISREG(ip->i_mode)); | |
3215 | ||
3216 | zfs_xvattr_set(zp, xvap, tx); | |
3217 | } | |
3218 | ||
9babb374 | 3219 | if (fuid_dirtied) |
cae5b340 | 3220 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 3221 | |
34dc7c2f | 3222 | if (mask != 0) |
5484965a | 3223 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3224 | |
34dc7c2f | 3225 | mutex_exit(&zp->z_lock); |
3558fd73 | 3226 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3227 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3228 | |
572e2857 | 3229 | if (attrzp) { |
3558fd73 | 3230 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3231 | mutex_exit(&attrzp->z_acl_lock); |
3232 | mutex_exit(&attrzp->z_lock); | |
3233 | } | |
9babb374 | 3234 | out: |
428870ff BB |
3235 | if (err == 0 && attrzp) { |
3236 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, | |
3237 | xattr_count, tx); | |
3238 | ASSERT(err2 == 0); | |
3239 | } | |
3240 | ||
45d1cae3 | 3241 | if (aclp) |
9babb374 | 3242 | zfs_acl_free(aclp); |
9babb374 BB |
3243 | |
3244 | if (fuidp) { | |
3245 | zfs_fuid_info_free(fuidp); | |
3246 | fuidp = NULL; | |
3247 | } | |
3248 | ||
428870ff | 3249 | if (err) { |
9babb374 | 3250 | dmu_tx_abort(tx); |
22929307 AX |
3251 | if (attrzp) |
3252 | iput(ZTOI(attrzp)); | |
428870ff BB |
3253 | if (err == ERESTART) |
3254 | goto top; | |
3255 | } else { | |
3256 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3257 | dmu_tx_commit(tx); |
22929307 AX |
3258 | if (attrzp) |
3259 | iput(ZTOI(attrzp)); | |
037849f8 | 3260 | zfs_inode_update(zp); |
428870ff BB |
3261 | } |
3262 | ||
428870ff | 3263 | out2: |
cae5b340 | 3264 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3265 | zil_commit(zilog, 0); |
34dc7c2f | 3266 | |
f4ea75d4 | 3267 | out3: |
a08ee875 LG |
3268 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * 7); |
3269 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * 7); | |
3270 | kmem_free(tmpxvattr, sizeof (xvattr_t)); | |
cae5b340 | 3271 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3272 | return (err); |
3273 | } | |
3274 | ||
3275 | typedef struct zfs_zlock { | |
3276 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3277 | znode_t *zl_znode; /* znode we held */ | |
3278 | struct zfs_zlock *zl_next; /* next in list */ | |
3279 | } zfs_zlock_t; | |
3280 | ||
3281 | /* | |
3282 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3283 | */ | |
3284 | static void | |
3285 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3286 | { | |
3287 | zfs_zlock_t *zl; | |
3288 | ||
3289 | while ((zl = *zlpp) != NULL) { | |
3290 | if (zl->zl_znode != NULL) | |
22929307 | 3291 | zfs_iput_async(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3292 | rw_exit(zl->zl_rwlock); |
3293 | *zlpp = zl->zl_next; | |
3294 | kmem_free(zl, sizeof (*zl)); | |
3295 | } | |
3296 | } | |
3297 | ||
3298 | /* | |
3299 | * Search back through the directory tree, using the ".." entries. | |
3300 | * Lock each directory in the chain to prevent concurrent renames. | |
3301 | * Fail any attempt to move a directory into one of its own descendants. | |
3302 | * XXX - z_parent_lock can overlap with map or grow locks | |
3303 | */ | |
3304 | static int | |
3305 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3306 | { | |
3307 | zfs_zlock_t *zl; | |
3308 | znode_t *zp = tdzp; | |
3558fd73 | 3309 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3310 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3311 | krwlock_t *rwlp = &szp->z_parent_lock; |
3312 | krw_t rw = RW_WRITER; | |
3313 | ||
3314 | /* | |
3315 | * First pass write-locks szp and compares to zp->z_id. | |
3316 | * Later passes read-lock zp and compare to zp->z_parent. | |
3317 | */ | |
3318 | do { | |
3319 | if (!rw_tryenter(rwlp, rw)) { | |
3320 | /* | |
3321 | * Another thread is renaming in this path. | |
3322 | * Note that if we are a WRITER, we don't have any | |
3323 | * parent_locks held yet. | |
3324 | */ | |
3325 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3326 | /* | |
3327 | * Drop our locks and restart | |
3328 | */ | |
3329 | zfs_rename_unlock(&zl); | |
3330 | *zlpp = NULL; | |
3331 | zp = tdzp; | |
428870ff | 3332 | oidp = zp->z_id; |
34dc7c2f BB |
3333 | rwlp = &szp->z_parent_lock; |
3334 | rw = RW_WRITER; | |
3335 | continue; | |
3336 | } else { | |
3337 | /* | |
3338 | * Wait for other thread to drop its locks | |
3339 | */ | |
3340 | rw_enter(rwlp, rw); | |
3341 | } | |
3342 | } | |
3343 | ||
3344 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3345 | zl->zl_rwlock = rwlp; | |
3346 | zl->zl_znode = NULL; | |
3347 | zl->zl_next = *zlpp; | |
3348 | *zlpp = zl; | |
3349 | ||
428870ff | 3350 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
a08ee875 | 3351 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3352 | |
428870ff | 3353 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3354 | return (0); |
3355 | ||
3356 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3357 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3358 | if (error) |
3359 | return (error); | |
3360 | zl->zl_znode = zp; | |
3361 | } | |
3558fd73 | 3362 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3363 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3364 | rwlp = &zp->z_parent_lock; |
3365 | rw = RW_READER; | |
3366 | ||
3367 | } while (zp->z_id != sdzp->z_id); | |
3368 | ||
3369 | return (0); | |
3370 | } | |
3371 | ||
3372 | /* | |
3373 | * Move an entry from the provided source directory to the target | |
3374 | * directory. Change the entry name as indicated. | |
3375 | * | |
3558fd73 | 3376 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3377 | * snm - Old entry name. |
3558fd73 | 3378 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3379 | * tnm - New entry name. |
3380 | * cr - credentials of caller. | |
34dc7c2f BB |
3381 | * flags - case flags |
3382 | * | |
a08ee875 | 3383 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3384 | * |
3385 | * Timestamps: | |
3558fd73 | 3386 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3387 | */ |
3388 | /*ARGSUSED*/ | |
e5c39b95 | 3389 | int |
3558fd73 BB |
3390 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3391 | cred_t *cr, int flags) | |
34dc7c2f BB |
3392 | { |
3393 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 | 3394 | znode_t *sdzp = ITOZ(sdip); |
cae5b340 | 3395 | zfsvfs_t *zfsvfs = ITOZSB(sdip); |
34dc7c2f | 3396 | zilog_t *zilog; |
34dc7c2f BB |
3397 | zfs_dirlock_t *sdl, *tdl; |
3398 | dmu_tx_t *tx; | |
3399 | zfs_zlock_t *zl; | |
3400 | int cmp, serr, terr; | |
3401 | int error = 0; | |
3402 | int zflg = 0; | |
a08ee875 | 3403 | boolean_t waited = B_FALSE; |
34dc7c2f | 3404 | |
cae5b340 AX |
3405 | if (snm == NULL || tnm == NULL) |
3406 | return (SET_ERROR(EINVAL)); | |
3407 | ||
3408 | ZFS_ENTER(zfsvfs); | |
34dc7c2f | 3409 | ZFS_VERIFY_ZP(sdzp); |
cae5b340 AX |
3410 | zilog = zfsvfs->z_log; |
3411 | ||
3412 | tdzp = ITOZ(tdip); | |
3413 | ZFS_VERIFY_ZP(tdzp); | |
34dc7c2f | 3414 | |
cae5b340 AX |
3415 | /* |
3416 | * We check i_sb because snapshots and the ctldir must have different | |
3417 | * super blocks. | |
3418 | */ | |
a08ee875 | 3419 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
cae5b340 | 3420 | ZFS_EXIT(zfsvfs); |
a08ee875 | 3421 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3422 | } |
3423 | ||
cae5b340 | 3424 | if (zfsvfs->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3425 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
cae5b340 | 3426 | ZFS_EXIT(zfsvfs); |
a08ee875 | 3427 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3428 | } |
3429 | ||
3430 | if (flags & FIGNORECASE) | |
3431 | zflg |= ZCILOOK; | |
3432 | ||
3433 | top: | |
3434 | szp = NULL; | |
3435 | tzp = NULL; | |
3436 | zl = NULL; | |
3437 | ||
3438 | /* | |
3439 | * This is to prevent the creation of links into attribute space | |
3440 | * by renaming a linked file into/outof an attribute directory. | |
3441 | * See the comment in zfs_link() for why this is considered bad. | |
3442 | */ | |
428870ff | 3443 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
cae5b340 | 3444 | ZFS_EXIT(zfsvfs); |
a08ee875 | 3445 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3446 | } |
3447 | ||
3448 | /* | |
3449 | * Lock source and target directory entries. To prevent deadlock, | |
3450 | * a lock ordering must be defined. We lock the directory with | |
3451 | * the smallest object id first, or if it's a tie, the one with | |
3452 | * the lexically first name. | |
3453 | */ | |
3454 | if (sdzp->z_id < tdzp->z_id) { | |
3455 | cmp = -1; | |
3456 | } else if (sdzp->z_id > tdzp->z_id) { | |
3457 | cmp = 1; | |
3458 | } else { | |
3459 | /* | |
3460 | * First compare the two name arguments without | |
3461 | * considering any case folding. | |
3462 | */ | |
cae5b340 | 3463 | int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3464 | |
3465 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
cae5b340 | 3466 | ASSERT(error == 0 || !zfsvfs->z_utf8); |
34dc7c2f BB |
3467 | if (cmp == 0) { |
3468 | /* | |
3469 | * POSIX: "If the old argument and the new argument | |
3470 | * both refer to links to the same existing file, | |
3471 | * the rename() function shall return successfully | |
3472 | * and perform no other action." | |
3473 | */ | |
cae5b340 | 3474 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3475 | return (0); |
3476 | } | |
3477 | /* | |
3478 | * If the file system is case-folding, then we may | |
3479 | * have some more checking to do. A case-folding file | |
3480 | * system is either supporting mixed case sensitivity | |
3481 | * access or is completely case-insensitive. Note | |
3482 | * that the file system is always case preserving. | |
3483 | * | |
3484 | * In mixed sensitivity mode case sensitive behavior | |
3485 | * is the default. FIGNORECASE must be used to | |
3486 | * explicitly request case insensitive behavior. | |
3487 | * | |
3488 | * If the source and target names provided differ only | |
3489 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3490 | * we will treat this as a special case in the | |
3491 | * case-insensitive mode: as long as the source name | |
3492 | * is an exact match, we will allow this to proceed as | |
3493 | * a name-change request. | |
3494 | */ | |
cae5b340 AX |
3495 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
3496 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3497 | flags & FIGNORECASE)) && |
cae5b340 | 3498 | u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3499 | &error) == 0) { |
3500 | /* | |
3501 | * case preserving rename request, require exact | |
3502 | * name matches | |
3503 | */ | |
3504 | zflg |= ZCIEXACT; | |
3505 | zflg &= ~ZCILOOK; | |
3506 | } | |
3507 | } | |
3508 | ||
428870ff BB |
3509 | /* |
3510 | * If the source and destination directories are the same, we should | |
3511 | * grab the z_name_lock of that directory only once. | |
3512 | */ | |
3513 | if (sdzp == tdzp) { | |
3514 | zflg |= ZHAVELOCK; | |
3515 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3516 | } | |
3517 | ||
34dc7c2f BB |
3518 | if (cmp < 0) { |
3519 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3520 | ZEXISTS | zflg, NULL, NULL); | |
3521 | terr = zfs_dirent_lock(&tdl, | |
3522 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3523 | } else { | |
3524 | terr = zfs_dirent_lock(&tdl, | |
3525 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3526 | serr = zfs_dirent_lock(&sdl, | |
3527 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3528 | NULL, NULL); | |
3529 | } | |
3530 | ||
3531 | if (serr) { | |
3532 | /* | |
3533 | * Source entry invalid or not there. | |
3534 | */ | |
3535 | if (!terr) { | |
3536 | zfs_dirent_unlock(tdl); | |
3537 | if (tzp) | |
3558fd73 | 3538 | iput(ZTOI(tzp)); |
34dc7c2f | 3539 | } |
428870ff BB |
3540 | |
3541 | if (sdzp == tdzp) | |
3542 | rw_exit(&sdzp->z_name_lock); | |
3543 | ||
34dc7c2f BB |
3544 | if (strcmp(snm, "..") == 0) |
3545 | serr = EINVAL; | |
cae5b340 | 3546 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3547 | return (serr); |
3548 | } | |
3549 | if (terr) { | |
3550 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3551 | iput(ZTOI(szp)); |
428870ff BB |
3552 | |
3553 | if (sdzp == tdzp) | |
3554 | rw_exit(&sdzp->z_name_lock); | |
3555 | ||
34dc7c2f BB |
3556 | if (strcmp(tnm, "..") == 0) |
3557 | terr = EINVAL; | |
cae5b340 | 3558 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3559 | return (terr); |
3560 | } | |
3561 | ||
3562 | /* | |
3563 | * Must have write access at the source to remove the old entry | |
3564 | * and write access at the target to create the new entry. | |
3565 | * Note that if target and source are the same, this can be | |
3566 | * done in a single check. | |
3567 | */ | |
3568 | ||
149e873a | 3569 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3570 | goto out; |
3571 | ||
3558fd73 | 3572 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3573 | /* |
3574 | * Check to make sure rename is valid. | |
3575 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3576 | */ | |
149e873a | 3577 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3578 | goto out; |
3579 | } | |
3580 | ||
3581 | /* | |
3582 | * Does target exist? | |
3583 | */ | |
3584 | if (tzp) { | |
3585 | /* | |
3586 | * Source and target must be the same type. | |
3587 | */ | |
3558fd73 BB |
3588 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3589 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
a08ee875 | 3590 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3591 | goto out; |
3592 | } | |
3593 | } else { | |
3558fd73 | 3594 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
a08ee875 | 3595 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3596 | goto out; |
3597 | } | |
3598 | } | |
3599 | /* | |
3600 | * POSIX dictates that when the source and target | |
3601 | * entries refer to the same file object, rename | |
3602 | * must do nothing and exit without error. | |
3603 | */ | |
3604 | if (szp->z_id == tzp->z_id) { | |
3605 | error = 0; | |
3606 | goto out; | |
3607 | } | |
3608 | } | |
3609 | ||
cae5b340 | 3610 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
3611 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3612 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3613 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3614 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3615 | if (sdzp != tdzp) { |
3616 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3617 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3618 | } | |
3619 | if (tzp) { | |
3620 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3621 | zfs_sa_upgrade_txholds(tx, tzp); | |
3622 | } | |
3623 | ||
3624 | zfs_sa_upgrade_txholds(tx, szp); | |
cae5b340 | 3625 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
a08ee875 | 3626 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3627 | if (error) { |
3628 | if (zl != NULL) | |
3629 | zfs_rename_unlock(&zl); | |
3630 | zfs_dirent_unlock(sdl); | |
3631 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3632 | |
3633 | if (sdzp == tdzp) | |
3634 | rw_exit(&sdzp->z_name_lock); | |
3635 | ||
fb5f0bc8 | 3636 | if (error == ERESTART) { |
a08ee875 | 3637 | waited = B_TRUE; |
34dc7c2f BB |
3638 | dmu_tx_wait(tx); |
3639 | dmu_tx_abort(tx); | |
22929307 AX |
3640 | iput(ZTOI(szp)); |
3641 | if (tzp) | |
3642 | iput(ZTOI(tzp)); | |
34dc7c2f BB |
3643 | goto top; |
3644 | } | |
3645 | dmu_tx_abort(tx); | |
22929307 AX |
3646 | iput(ZTOI(szp)); |
3647 | if (tzp) | |
3648 | iput(ZTOI(tzp)); | |
cae5b340 | 3649 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3650 | return (error); |
3651 | } | |
3652 | ||
3653 | if (tzp) /* Attempt to remove the existing target */ | |
3654 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3655 | ||
3656 | if (error == 0) { | |
3657 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3658 | if (error == 0) { | |
428870ff | 3659 | szp->z_pflags |= ZFS_AV_MODIFIED; |
34dc7c2f | 3660 | |
cae5b340 | 3661 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 3662 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c06d4368 | 3663 | ASSERT0(error); |
34dc7c2f | 3664 | |
428870ff BB |
3665 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3666 | if (error == 0) { | |
3667 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3668 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3669 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3670 | } else { |
3671 | /* | |
3672 | * At this point, we have successfully created | |
3673 | * the target name, but have failed to remove | |
3674 | * the source name. Since the create was done | |
3675 | * with the ZRENAMING flag, there are | |
3676 | * complications; for one, the link count is | |
3677 | * wrong. The easiest way to deal with this | |
3678 | * is to remove the newly created target, and | |
3679 | * return the original error. This must | |
3680 | * succeed; fortunately, it is very unlikely to | |
3681 | * fail, since we just created it. | |
3682 | */ | |
3683 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3684 | ZRENAMING, NULL), ==, 0); | |
3685 | } | |
34dc7c2f BB |
3686 | } |
3687 | } | |
3688 | ||
3689 | dmu_tx_commit(tx); | |
3690 | out: | |
3691 | if (zl != NULL) | |
3692 | zfs_rename_unlock(&zl); | |
3693 | ||
3694 | zfs_dirent_unlock(sdl); | |
3695 | zfs_dirent_unlock(tdl); | |
3696 | ||
960e08fe | 3697 | zfs_inode_update(sdzp); |
428870ff BB |
3698 | if (sdzp == tdzp) |
3699 | rw_exit(&sdzp->z_name_lock); | |
3700 | ||
960e08fe BB |
3701 | if (sdzp != tdzp) |
3702 | zfs_inode_update(tdzp); | |
428870ff | 3703 | |
960e08fe | 3704 | zfs_inode_update(szp); |
3558fd73 | 3705 | iput(ZTOI(szp)); |
960e08fe BB |
3706 | if (tzp) { |
3707 | zfs_inode_update(tzp); | |
3558fd73 | 3708 | iput(ZTOI(tzp)); |
960e08fe | 3709 | } |
34dc7c2f | 3710 | |
cae5b340 | 3711 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3712 | zil_commit(zilog, 0); |
428870ff | 3713 | |
cae5b340 | 3714 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3715 | return (error); |
3716 | } | |
3717 | ||
3718 | /* | |
3719 | * Insert the indicated symbolic reference entry into the directory. | |
3720 | * | |
3558fd73 | 3721 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
3722 | * link - Name for new symlink entry. |
3723 | * vap - Attributes of new entry. | |
3724 | * target - Target path of new symlink. | |
3558fd73 | 3725 | * |
34dc7c2f | 3726 | * cr - credentials of caller. |
34dc7c2f BB |
3727 | * flags - case flags |
3728 | * | |
a08ee875 | 3729 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3730 | * |
3731 | * Timestamps: | |
3558fd73 | 3732 | * dip - ctime|mtime updated |
34dc7c2f BB |
3733 | */ |
3734 | /*ARGSUSED*/ | |
e5c39b95 | 3735 | int |
3558fd73 BB |
3736 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
3737 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 3738 | { |
3558fd73 | 3739 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
3740 | zfs_dirlock_t *dl; |
3741 | dmu_tx_t *tx; | |
cae5b340 | 3742 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 3743 | zilog_t *zilog; |
428870ff | 3744 | uint64_t len = strlen(link); |
34dc7c2f BB |
3745 | int error; |
3746 | int zflg = ZNEW; | |
9babb374 BB |
3747 | zfs_acl_ids_t acl_ids; |
3748 | boolean_t fuid_dirtied; | |
428870ff | 3749 | uint64_t txtype = TX_SYMLINK; |
a08ee875 | 3750 | boolean_t waited = B_FALSE; |
34dc7c2f | 3751 | |
3558fd73 | 3752 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 3753 | |
cae5b340 AX |
3754 | if (name == NULL) |
3755 | return (SET_ERROR(EINVAL)); | |
3756 | ||
3757 | ZFS_ENTER(zfsvfs); | |
34dc7c2f | 3758 | ZFS_VERIFY_ZP(dzp); |
cae5b340 | 3759 | zilog = zfsvfs->z_log; |
34dc7c2f | 3760 | |
cae5b340 | 3761 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 3762 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
cae5b340 | 3763 | ZFS_EXIT(zfsvfs); |
a08ee875 | 3764 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3765 | } |
3766 | if (flags & FIGNORECASE) | |
3767 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3768 | |
3769 | if (len > MAXPATHLEN) { | |
cae5b340 | 3770 | ZFS_EXIT(zfsvfs); |
a08ee875 | 3771 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
3772 | } |
3773 | ||
428870ff BB |
3774 | if ((error = zfs_acl_ids_create(dzp, 0, |
3775 | vap, cr, NULL, &acl_ids)) != 0) { | |
cae5b340 | 3776 | ZFS_EXIT(zfsvfs); |
428870ff BB |
3777 | return (error); |
3778 | } | |
3779 | top: | |
3558fd73 BB |
3780 | *ipp = NULL; |
3781 | ||
34dc7c2f BB |
3782 | /* |
3783 | * Attempt to lock directory; fail if entry already exists. | |
3784 | */ | |
3785 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3786 | if (error) { | |
428870ff | 3787 | zfs_acl_ids_free(&acl_ids); |
cae5b340 | 3788 | ZFS_EXIT(zfsvfs); |
428870ff BB |
3789 | return (error); |
3790 | } | |
3791 | ||
149e873a | 3792 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
3793 | zfs_acl_ids_free(&acl_ids); |
3794 | zfs_dirent_unlock(dl); | |
cae5b340 | 3795 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3796 | return (error); |
3797 | } | |
3798 | ||
cae5b340 | 3799 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
9babb374 BB |
3800 | zfs_acl_ids_free(&acl_ids); |
3801 | zfs_dirent_unlock(dl); | |
cae5b340 | 3802 | ZFS_EXIT(zfsvfs); |
a08ee875 | 3803 | return (SET_ERROR(EDQUOT)); |
9babb374 | 3804 | } |
cae5b340 AX |
3805 | tx = dmu_tx_create(zfsvfs->z_os); |
3806 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
34dc7c2f | 3807 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3808 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3809 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3810 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3811 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
cae5b340 | 3812 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
3813 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
3814 | acl_ids.z_aclp->z_acl_bytes); | |
3815 | } | |
9babb374 | 3816 | if (fuid_dirtied) |
cae5b340 | 3817 | zfs_fuid_txhold(zfsvfs, tx); |
a08ee875 | 3818 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3819 | if (error) { |
3820 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3821 | if (error == ERESTART) { |
a08ee875 | 3822 | waited = B_TRUE; |
34dc7c2f BB |
3823 | dmu_tx_wait(tx); |
3824 | dmu_tx_abort(tx); | |
3825 | goto top; | |
3826 | } | |
428870ff | 3827 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 3828 | dmu_tx_abort(tx); |
cae5b340 | 3829 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3830 | return (error); |
3831 | } | |
3832 | ||
34dc7c2f BB |
3833 | /* |
3834 | * Create a new object for the symlink. | |
428870ff | 3835 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 3836 | */ |
428870ff | 3837 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3838 | |
428870ff | 3839 | if (fuid_dirtied) |
cae5b340 | 3840 | zfs_fuid_sync(zfsvfs, tx); |
34dc7c2f | 3841 | |
572e2857 | 3842 | mutex_enter(&zp->z_lock); |
428870ff | 3843 | if (zp->z_is_sa) |
cae5b340 | 3844 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), |
428870ff BB |
3845 | link, len, tx); |
3846 | else | |
3847 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3848 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3849 | |
428870ff | 3850 | zp->z_size = len; |
cae5b340 | 3851 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 3852 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
3853 | /* |
3854 | * Insert the new object into the directory. | |
3855 | */ | |
3856 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
428870ff BB |
3857 | |
3858 | if (flags & FIGNORECASE) | |
3859 | txtype |= TX_CI; | |
3860 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
9babb374 | 3861 | |
960e08fe BB |
3862 | zfs_inode_update(dzp); |
3863 | zfs_inode_update(zp); | |
3864 | ||
9babb374 | 3865 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3866 | |
3867 | dmu_tx_commit(tx); | |
3868 | ||
3869 | zfs_dirent_unlock(dl); | |
3870 | ||
3558fd73 | 3871 | *ipp = ZTOI(zp); |
34dc7c2f | 3872 | |
cae5b340 | 3873 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3874 | zil_commit(zilog, 0); |
428870ff | 3875 | |
cae5b340 | 3876 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3877 | return (error); |
3878 | } | |
3879 | ||
3880 | /* | |
3881 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 3882 | * the symbolic path referred to by ip. |
34dc7c2f | 3883 | * |
8b4f9a2d BB |
3884 | * IN: ip - inode of symbolic link |
3885 | * uio - structure to contain the link path. | |
3886 | * cr - credentials of caller. | |
34dc7c2f BB |
3887 | * |
3888 | * RETURN: 0 if success | |
3889 | * error code if failure | |
3890 | * | |
3891 | * Timestamps: | |
3558fd73 | 3892 | * ip - atime updated |
34dc7c2f BB |
3893 | */ |
3894 | /* ARGSUSED */ | |
e5c39b95 | 3895 | int |
8b4f9a2d | 3896 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 3897 | { |
3558fd73 | 3898 | znode_t *zp = ITOZ(ip); |
cae5b340 | 3899 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
3900 | int error; |
3901 | ||
cae5b340 | 3902 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
3903 | ZFS_VERIFY_ZP(zp); |
3904 | ||
572e2857 | 3905 | mutex_enter(&zp->z_lock); |
428870ff | 3906 | if (zp->z_is_sa) |
8b4f9a2d | 3907 | error = sa_lookup_uio(zp->z_sa_hdl, |
cae5b340 | 3908 | SA_ZPL_SYMLINK(zfsvfs), uio); |
428870ff | 3909 | else |
8b4f9a2d | 3910 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 3911 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3912 | |
cae5b340 | 3913 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3914 | return (error); |
3915 | } | |
3916 | ||
3917 | /* | |
3558fd73 | 3918 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 3919 | * |
3558fd73 BB |
3920 | * IN: tdip - Directory to contain new entry. |
3921 | * sip - inode of new entry. | |
34dc7c2f BB |
3922 | * name - name of new entry. |
3923 | * cr - credentials of caller. | |
34dc7c2f BB |
3924 | * |
3925 | * RETURN: 0 if success | |
3926 | * error code if failure | |
3927 | * | |
3928 | * Timestamps: | |
3558fd73 BB |
3929 | * tdip - ctime|mtime updated |
3930 | * sip - ctime updated | |
34dc7c2f BB |
3931 | */ |
3932 | /* ARGSUSED */ | |
e5c39b95 | 3933 | int |
cae5b340 AX |
3934 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr, |
3935 | int flags) | |
34dc7c2f | 3936 | { |
3558fd73 | 3937 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 3938 | znode_t *tzp, *szp; |
cae5b340 | 3939 | zfsvfs_t *zfsvfs = ITOZSB(tdip); |
34dc7c2f BB |
3940 | zilog_t *zilog; |
3941 | zfs_dirlock_t *dl; | |
3942 | dmu_tx_t *tx; | |
34dc7c2f BB |
3943 | int error; |
3944 | int zf = ZNEW; | |
428870ff | 3945 | uint64_t parent; |
572e2857 | 3946 | uid_t owner; |
a08ee875 | 3947 | boolean_t waited = B_FALSE; |
cae5b340 AX |
3948 | boolean_t is_tmpfile = 0; |
3949 | uint64_t txg; | |
3950 | #ifdef HAVE_TMPFILE | |
3951 | is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); | |
3952 | #endif | |
3558fd73 | 3953 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 3954 | |
cae5b340 AX |
3955 | if (name == NULL) |
3956 | return (SET_ERROR(EINVAL)); | |
3957 | ||
3958 | ZFS_ENTER(zfsvfs); | |
34dc7c2f | 3959 | ZFS_VERIFY_ZP(dzp); |
cae5b340 | 3960 | zilog = zfsvfs->z_log; |
34dc7c2f | 3961 | |
428870ff BB |
3962 | /* |
3963 | * POSIX dictates that we return EPERM here. | |
3964 | * Better choices include ENOTSUP or EISDIR. | |
3965 | */ | |
3558fd73 | 3966 | if (S_ISDIR(sip->i_mode)) { |
cae5b340 | 3967 | ZFS_EXIT(zfsvfs); |
a08ee875 | 3968 | return (SET_ERROR(EPERM)); |
428870ff BB |
3969 | } |
3970 | ||
cae5b340 AX |
3971 | szp = ITOZ(sip); |
3972 | ZFS_VERIFY_ZP(szp); | |
3973 | ||
3974 | /* | |
3975 | * We check i_sb because snapshots and the ctldir must have different | |
3976 | * super blocks. | |
3977 | */ | |
a08ee875 | 3978 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
cae5b340 | 3979 | ZFS_EXIT(zfsvfs); |
a08ee875 | 3980 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 3981 | } |
428870ff | 3982 | |
428870ff BB |
3983 | /* Prevent links to .zfs/shares files */ |
3984 | ||
cae5b340 | 3985 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff | 3986 | &parent, sizeof (uint64_t))) != 0) { |
cae5b340 | 3987 | ZFS_EXIT(zfsvfs); |
428870ff BB |
3988 | return (error); |
3989 | } | |
cae5b340 AX |
3990 | if (parent == zfsvfs->z_shares_dir) { |
3991 | ZFS_EXIT(zfsvfs); | |
a08ee875 | 3992 | return (SET_ERROR(EPERM)); |
428870ff BB |
3993 | } |
3994 | ||
cae5b340 | 3995 | if (zfsvfs->z_utf8 && u8_validate(name, |
34dc7c2f | 3996 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
cae5b340 | 3997 | ZFS_EXIT(zfsvfs); |
a08ee875 | 3998 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3999 | } |
4000 | if (flags & FIGNORECASE) | |
4001 | zf |= ZCILOOK; | |
4002 | ||
34dc7c2f BB |
4003 | /* |
4004 | * We do not support links between attributes and non-attributes | |
4005 | * because of the potential security risk of creating links | |
4006 | * into "normal" file space in order to circumvent restrictions | |
4007 | * imposed in attribute space. | |
4008 | */ | |
428870ff | 4009 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
cae5b340 | 4010 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4011 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4012 | } |
4013 | ||
cae5b340 AX |
4014 | owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid), |
4015 | cr, ZFS_OWNER); | |
572e2857 | 4016 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
cae5b340 | 4017 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4018 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4019 | } |
4020 | ||
149e873a | 4021 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
cae5b340 | 4022 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4023 | return (error); |
4024 | } | |
4025 | ||
428870ff | 4026 | top: |
34dc7c2f BB |
4027 | /* |
4028 | * Attempt to lock directory; fail if entry already exists. | |
4029 | */ | |
4030 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
4031 | if (error) { | |
cae5b340 | 4032 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4033 | return (error); |
4034 | } | |
4035 | ||
cae5b340 | 4036 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 4037 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 4038 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
cae5b340 AX |
4039 | if (is_tmpfile) |
4040 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); | |
4041 | ||
428870ff BB |
4042 | zfs_sa_upgrade_txholds(tx, szp); |
4043 | zfs_sa_upgrade_txholds(tx, dzp); | |
a08ee875 | 4044 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
4045 | if (error) { |
4046 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4047 | if (error == ERESTART) { |
a08ee875 | 4048 | waited = B_TRUE; |
34dc7c2f BB |
4049 | dmu_tx_wait(tx); |
4050 | dmu_tx_abort(tx); | |
4051 | goto top; | |
4052 | } | |
4053 | dmu_tx_abort(tx); | |
cae5b340 | 4054 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4055 | return (error); |
4056 | } | |
cae5b340 AX |
4057 | /* unmark z_unlinked so zfs_link_create will not reject */ |
4058 | if (is_tmpfile) | |
4059 | szp->z_unlinked = 0; | |
34dc7c2f BB |
4060 | error = zfs_link_create(dl, szp, tx, 0); |
4061 | ||
4062 | if (error == 0) { | |
4063 | uint64_t txtype = TX_LINK; | |
cae5b340 AX |
4064 | /* |
4065 | * tmpfile is created to be in z_unlinkedobj, so remove it. | |
4066 | * Also, we don't log in ZIL, be cause all previous file | |
4067 | * operation on the tmpfile are ignored by ZIL. Instead we | |
4068 | * always wait for txg to sync to make sure all previous | |
4069 | * operation are sync safe. | |
4070 | */ | |
4071 | if (is_tmpfile) { | |
4072 | VERIFY(zap_remove_int(zfsvfs->z_os, | |
4073 | zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); | |
4074 | } else { | |
4075 | if (flags & FIGNORECASE) | |
4076 | txtype |= TX_CI; | |
4077 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); | |
4078 | } | |
4079 | } else if (is_tmpfile) { | |
4080 | /* restore z_unlinked since when linking failed */ | |
4081 | szp->z_unlinked = 1; | |
34dc7c2f | 4082 | } |
cae5b340 | 4083 | txg = dmu_tx_get_txg(tx); |
34dc7c2f BB |
4084 | dmu_tx_commit(tx); |
4085 | ||
4086 | zfs_dirent_unlock(dl); | |
4087 | ||
cae5b340 | 4088 | if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4089 | zil_commit(zilog, 0); |
428870ff | 4090 | |
cae5b340 AX |
4091 | if (is_tmpfile) |
4092 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg); | |
4093 | ||
960e08fe BB |
4094 | zfs_inode_update(dzp); |
4095 | zfs_inode_update(szp); | |
cae5b340 | 4096 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4097 | return (error); |
4098 | } | |
4099 | ||
3c0e5c0f | 4100 | static void |
a08ee875 | 4101 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
4102 | { |
4103 | struct page *pp = arg; | |
4104 | ||
a08ee875 | 4105 | ClearPageError(pp); |
3c0e5c0f BB |
4106 | end_page_writeback(pp); |
4107 | } | |
4108 | ||
34dc7c2f | 4109 | /* |
3c0e5c0f BB |
4110 | * Push a page out to disk, once the page is on stable storage the |
4111 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 4112 | * |
3c0e5c0f BB |
4113 | * IN: ip - page mapped for inode. |
4114 | * pp - page to push (page is locked) | |
4115 | * wbc - writeback control data | |
34dc7c2f BB |
4116 | * |
4117 | * RETURN: 0 if success | |
4118 | * error code if failure | |
4119 | * | |
3c0e5c0f BB |
4120 | * Timestamps: |
4121 | * ip - ctime|mtime updated | |
34dc7c2f BB |
4122 | */ |
4123 | /* ARGSUSED */ | |
3c0e5c0f BB |
4124 | int |
4125 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 4126 | { |
3c0e5c0f | 4127 | znode_t *zp = ITOZ(ip); |
cae5b340 | 4128 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
3c0e5c0f BB |
4129 | loff_t offset; |
4130 | loff_t pgoff; | |
4c837f0d BB |
4131 | unsigned int pglen; |
4132 | rl_t *rl; | |
3c0e5c0f BB |
4133 | dmu_tx_t *tx; |
4134 | caddr_t va; | |
4135 | int err = 0; | |
4136 | uint64_t mtime[2], ctime[2]; | |
4137 | sa_bulk_attr_t bulk[3]; | |
4138 | int cnt = 0; | |
e10b0808 | 4139 | struct address_space *mapping; |
3c0e5c0f | 4140 | |
cae5b340 | 4141 | ZFS_ENTER(zfsvfs); |
4c837f0d | 4142 | ZFS_VERIFY_ZP(zp); |
d164b209 | 4143 | |
3c0e5c0f BB |
4144 | ASSERT(PageLocked(pp)); |
4145 | ||
a08ee875 LG |
4146 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
4147 | offset = i_size_read(ip); /* File length in bytes */ | |
5eacc075 AX |
4148 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
4149 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
4150 | |
4151 | /* Page is beyond end of file */ | |
4152 | if (pgoff >= offset) { | |
4153 | unlock_page(pp); | |
cae5b340 | 4154 | ZFS_EXIT(zfsvfs); |
3c0e5c0f BB |
4155 | return (0); |
4156 | } | |
4157 | ||
4158 | /* Truncate page length to end of file */ | |
4159 | if (pgoff + pglen > offset) | |
4160 | pglen = offset - pgoff; | |
4161 | ||
4162 | #if 0 | |
34dc7c2f | 4163 | /* |
3c0e5c0f BB |
4164 | * FIXME: Allow mmap writes past its quota. The correct fix |
4165 | * is to register a page_mkwrite() handler to count the page | |
4166 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 4167 | */ |
cae5b340 AX |
4168 | if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || |
4169 | zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { | |
9babb374 | 4170 | err = EDQUOT; |
9babb374 | 4171 | } |
3c0e5c0f BB |
4172 | #endif |
4173 | ||
ea04106b AX |
4174 | /* |
4175 | * The ordering here is critical and must adhere to the following | |
4176 | * rules in order to avoid deadlocking in either zfs_read() or | |
4177 | * zfs_free_range() due to a lock inversion. | |
4178 | * | |
4179 | * 1) The page must be unlocked prior to acquiring the range lock. | |
4180 | * This is critical because zfs_read() calls find_lock_page() | |
4181 | * which may block on the page lock while holding the range lock. | |
4182 | * | |
4183 | * 2) Before setting or clearing write back on a page the range lock | |
4184 | * must be held in order to prevent a lock inversion with the | |
4185 | * zfs_free_range() function. | |
e10b0808 AX |
4186 | * |
4187 | * This presents a problem because upon entering this function the | |
4188 | * page lock is already held. To safely acquire the range lock the | |
4189 | * page lock must be dropped. This creates a window where another | |
4190 | * process could truncate, invalidate, dirty, or write out the page. | |
4191 | * | |
4192 | * Therefore, after successfully reacquiring the range and page locks | |
4193 | * the current page state is checked. In the common case everything | |
4194 | * will be as is expected and it can be written out. However, if | |
4195 | * the page state has changed it must be handled accordingly. | |
ea04106b | 4196 | */ |
e10b0808 AX |
4197 | mapping = pp->mapping; |
4198 | redirty_page_for_writepage(wbc, pp); | |
3c0e5c0f | 4199 | unlock_page(pp); |
e10b0808 | 4200 | |
87dac73d | 4201 | rl = zfs_range_lock(&zp->z_range_lock, pgoff, pglen, RL_WRITER); |
e10b0808 AX |
4202 | lock_page(pp); |
4203 | ||
4204 | /* Page mapping changed or it was no longer dirty, we're done */ | |
4205 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
4206 | unlock_page(pp); | |
4207 | zfs_range_unlock(rl); | |
cae5b340 | 4208 | ZFS_EXIT(zfsvfs); |
e10b0808 AX |
4209 | return (0); |
4210 | } | |
4211 | ||
4212 | /* Another process started write block if required */ | |
4213 | if (PageWriteback(pp)) { | |
4214 | unlock_page(pp); | |
4215 | zfs_range_unlock(rl); | |
4216 | ||
4217 | if (wbc->sync_mode != WB_SYNC_NONE) | |
4218 | wait_on_page_writeback(pp); | |
4219 | ||
cae5b340 | 4220 | ZFS_EXIT(zfsvfs); |
e10b0808 AX |
4221 | return (0); |
4222 | } | |
4223 | ||
4224 | /* Clear the dirty flag the required locks are held */ | |
4225 | if (!clear_page_dirty_for_io(pp)) { | |
4226 | unlock_page(pp); | |
4227 | zfs_range_unlock(rl); | |
cae5b340 | 4228 | ZFS_EXIT(zfsvfs); |
e10b0808 AX |
4229 | return (0); |
4230 | } | |
4231 | ||
4232 | /* | |
4233 | * Counterpart for redirty_page_for_writepage() above. This page | |
4234 | * was in fact not skipped and should not be counted as if it were. | |
4235 | */ | |
4236 | wbc->pages_skipped--; | |
ea04106b | 4237 | set_page_writeback(pp); |
e10b0808 | 4238 | unlock_page(pp); |
3c0e5c0f | 4239 | |
cae5b340 | 4240 | tx = dmu_tx_create(zfsvfs->z_os); |
3c0e5c0f | 4241 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
4242 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4243 | zfs_sa_upgrade_txholds(tx, zp); | |
ea04106b | 4244 | |
fb5f0bc8 | 4245 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 4246 | if (err != 0) { |
3c0e5c0f | 4247 | if (err == ERESTART) |
34dc7c2f | 4248 | dmu_tx_wait(tx); |
3c0e5c0f | 4249 | |
34dc7c2f | 4250 | dmu_tx_abort(tx); |
a08ee875 LG |
4251 | __set_page_dirty_nobuffers(pp); |
4252 | ClearPageError(pp); | |
4253 | end_page_writeback(pp); | |
4c837f0d | 4254 | zfs_range_unlock(rl); |
cae5b340 | 4255 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4256 | return (err); |
34dc7c2f BB |
4257 | } |
4258 | ||
dde471ef | 4259 | va = kmap(pp); |
5eacc075 | 4260 | ASSERT3U(pglen, <=, PAGE_SIZE); |
cae5b340 | 4261 | dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 4262 | kunmap(pp); |
34dc7c2f | 4263 | |
cae5b340 AX |
4264 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
4265 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
4266 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, | |
4267 | &zp->z_pflags, 8); | |
428870ff | 4268 | |
d3aa3ea9 BB |
4269 | /* Preserve the mtime and ctime provided by the inode */ |
4270 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4271 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
4272 | zp->z_atime_dirty = 0; | |
4273 | zp->z_seq++; | |
4274 | ||
4275 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4276 | ||
cae5b340 | 4277 | zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
a08ee875 | 4278 | zfs_putpage_commit_cb, pp); |
45d1cae3 | 4279 | dmu_tx_commit(tx); |
d3aa3ea9 | 4280 | |
4c837f0d | 4281 | zfs_range_unlock(rl); |
34dc7c2f | 4282 | |
a08ee875 LG |
4283 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4284 | /* | |
4285 | * Note that this is rarely called under writepages(), because | |
4286 | * writepages() normally handles the entire commit for | |
4287 | * performance reasons. | |
4288 | */ | |
cae5b340 | 4289 | zil_commit(zfsvfs->z_log, zp->z_id); |
2b286136 | 4290 | } |
3c0e5c0f | 4291 | |
cae5b340 | 4292 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4293 | return (err); |
34dc7c2f BB |
4294 | } |
4295 | ||
8780c539 BB |
4296 | /* |
4297 | * Update the system attributes when the inode has been dirtied. For the | |
a08ee875 | 4298 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
4299 | */ |
4300 | int | |
4301 | zfs_dirty_inode(struct inode *ip, int flags) | |
4302 | { | |
4303 | znode_t *zp = ITOZ(ip); | |
cae5b340 | 4304 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
8780c539 | 4305 | dmu_tx_t *tx; |
a08ee875 LG |
4306 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
4307 | sa_bulk_attr_t bulk[4]; | |
68d83c55 | 4308 | int error = 0; |
8780c539 BB |
4309 | int cnt = 0; |
4310 | ||
cae5b340 | 4311 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
ea04106b AX |
4312 | return (0); |
4313 | ||
cae5b340 | 4314 | ZFS_ENTER(zfsvfs); |
8780c539 BB |
4315 | ZFS_VERIFY_ZP(zp); |
4316 | ||
68d83c55 AX |
4317 | #ifdef I_DIRTY_TIME |
4318 | /* | |
4319 | * This is the lazytime semantic indroduced in Linux 4.0 | |
4320 | * This flag will only be called from update_time when lazytime is set. | |
4321 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
4322 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
4323 | * only need to dirty atime. | |
4324 | */ | |
4325 | if (flags == I_DIRTY_TIME) { | |
4326 | zp->z_atime_dirty = 1; | |
4327 | goto out; | |
4328 | } | |
4329 | #endif | |
4330 | ||
cae5b340 | 4331 | tx = dmu_tx_create(zfsvfs->z_os); |
8780c539 BB |
4332 | |
4333 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4334 | zfs_sa_upgrade_txholds(tx, zp); | |
4335 | ||
4336 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4337 | if (error) { | |
4338 | dmu_tx_abort(tx); | |
4339 | goto out; | |
4340 | } | |
4341 | ||
4342 | mutex_enter(&zp->z_lock); | |
68d83c55 AX |
4343 | zp->z_atime_dirty = 0; |
4344 | ||
cae5b340 AX |
4345 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
4346 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
4347 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
4348 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
8780c539 | 4349 | |
a08ee875 | 4350 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4351 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4352 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4353 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
a08ee875 LG |
4354 | mode = ip->i_mode; |
4355 | ||
4356 | zp->z_mode = mode; | |
8780c539 BB |
4357 | |
4358 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4359 | mutex_exit(&zp->z_lock); | |
4360 | ||
4361 | dmu_tx_commit(tx); | |
4362 | out: | |
cae5b340 | 4363 | ZFS_EXIT(zfsvfs); |
8780c539 BB |
4364 | return (error); |
4365 | } | |
8780c539 | 4366 | |
34dc7c2f BB |
4367 | /*ARGSUSED*/ |
4368 | void | |
c0d35759 | 4369 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4370 | { |
c0d35759 | 4371 | znode_t *zp = ITOZ(ip); |
cae5b340 | 4372 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
68d83c55 | 4373 | uint64_t atime[2]; |
34dc7c2f | 4374 | int error; |
e10b0808 | 4375 | int need_unlock = 0; |
34dc7c2f | 4376 | |
e10b0808 | 4377 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
cae5b340 | 4378 | if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) { |
e10b0808 | 4379 | need_unlock = 1; |
cae5b340 | 4380 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); |
ebe7e575 | 4381 | } |
c0d35759 | 4382 | if (zp->z_sa_hdl == NULL) { |
e10b0808 | 4383 | if (need_unlock) |
cae5b340 | 4384 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
c0d35759 | 4385 | return; |
34dc7c2f BB |
4386 | } |
4387 | ||
4388 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
cae5b340 | 4389 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 4390 | |
428870ff BB |
4391 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4392 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4393 | error = dmu_tx_assign(tx, TXG_WAIT); |
4394 | if (error) { | |
4395 | dmu_tx_abort(tx); | |
4396 | } else { | |
68d83c55 | 4397 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
34dc7c2f | 4398 | mutex_enter(&zp->z_lock); |
cae5b340 | 4399 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), |
68d83c55 | 4400 | (void *)&atime, sizeof (atime), tx); |
34dc7c2f BB |
4401 | zp->z_atime_dirty = 0; |
4402 | mutex_exit(&zp->z_lock); | |
4403 | dmu_tx_commit(tx); | |
4404 | } | |
4405 | } | |
4406 | ||
4407 | zfs_zinactive(zp); | |
e10b0808 | 4408 | if (need_unlock) |
cae5b340 | 4409 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
34dc7c2f BB |
4410 | } |
4411 | ||
4412 | /* | |
4413 | * Bounds-check the seek operation. | |
4414 | * | |
3558fd73 | 4415 | * IN: ip - inode seeking within |
34dc7c2f BB |
4416 | * ooff - old file offset |
4417 | * noffp - pointer to new file offset | |
4418 | * ct - caller context | |
4419 | * | |
4420 | * RETURN: 0 if success | |
4421 | * EINVAL if new offset invalid | |
4422 | */ | |
4423 | /* ARGSUSED */ | |
3558fd73 | 4424 | int |
9623f736 | 4425 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4426 | { |
3558fd73 | 4427 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4428 | return (0); |
4429 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4430 | } | |
4431 | ||
34dc7c2f | 4432 | /* |
dde471ef | 4433 | * Fill pages with data from the disk. |
34dc7c2f BB |
4434 | */ |
4435 | static int | |
dde471ef | 4436 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4437 | { |
a08ee875 | 4438 | znode_t *zp = ITOZ(ip); |
cae5b340 | 4439 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
a08ee875 | 4440 | objset_t *os; |
dde471ef | 4441 | struct page *cur_pp; |
a08ee875 LG |
4442 | u_offset_t io_off, total; |
4443 | size_t io_len; | |
4444 | loff_t i_size; | |
4445 | unsigned page_idx; | |
4446 | int err; | |
34dc7c2f | 4447 | |
cae5b340 | 4448 | os = zfsvfs->z_os; |
5eacc075 | 4449 | io_len = nr_pages << PAGE_SHIFT; |
dde471ef PJ |
4450 | i_size = i_size_read(ip); |
4451 | io_off = page_offset(pl[0]); | |
4452 | ||
4453 | if (io_off + io_len > i_size) | |
4454 | io_len = i_size - io_off; | |
34dc7c2f BB |
4455 | |
4456 | /* | |
dde471ef | 4457 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4458 | */ |
dde471ef | 4459 | page_idx = 0; |
34dc7c2f | 4460 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4461 | caddr_t va; |
4462 | ||
87dac73d | 4463 | cur_pp = pl[page_idx++]; |
dde471ef | 4464 | va = kmap(cur_pp); |
9babb374 BB |
4465 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4466 | DMU_READ_PREFETCH); | |
dde471ef | 4467 | kunmap(cur_pp); |
34dc7c2f | 4468 | if (err) { |
b128c09f BB |
4469 | /* convert checksum errors into IO errors */ |
4470 | if (err == ECKSUM) | |
a08ee875 | 4471 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4472 | return (err); |
4473 | } | |
34dc7c2f | 4474 | } |
d164b209 | 4475 | |
34dc7c2f BB |
4476 | return (0); |
4477 | } | |
4478 | ||
4479 | /* | |
dde471ef | 4480 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4481 | * |
dde471ef PJ |
4482 | * IN: ip - inode of file to get data from. |
4483 | * pl - list of pages to read | |
4484 | * nr_pages - number of pages to read | |
34dc7c2f | 4485 | * |
a08ee875 | 4486 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4487 | * |
4488 | * Timestamps: | |
4489 | * vp - atime updated | |
4490 | */ | |
4491 | /* ARGSUSED */ | |
dde471ef PJ |
4492 | int |
4493 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4494 | { |
dde471ef | 4495 | znode_t *zp = ITOZ(ip); |
cae5b340 | 4496 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
dde471ef | 4497 | int err; |
d164b209 | 4498 | |
d164b209 BB |
4499 | if (pl == NULL) |
4500 | return (0); | |
34dc7c2f | 4501 | |
cae5b340 | 4502 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4503 | ZFS_VERIFY_ZP(zp); |
4504 | ||
dde471ef | 4505 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4506 | |
cae5b340 | 4507 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4508 | return (err); |
4509 | } | |
4510 | ||
4511 | /* | |
e2e7aa2d | 4512 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4513 | * |
e2e7aa2d BB |
4514 | * IN: ip - inode of the file to mmap |
4515 | * off - file offset | |
4516 | * addrp - start address in memory region | |
4517 | * len - length of memory region | |
4518 | * vm_flags- address flags | |
34dc7c2f | 4519 | * |
e2e7aa2d BB |
4520 | * RETURN: 0 if success |
4521 | * error code if failure | |
34dc7c2f BB |
4522 | */ |
4523 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4524 | int |
4525 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4526 | unsigned long vm_flags) | |
34dc7c2f | 4527 | { |
e2e7aa2d | 4528 | znode_t *zp = ITOZ(ip); |
cae5b340 | 4529 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4530 | |
cae5b340 | 4531 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4532 | ZFS_VERIFY_ZP(zp); |
4533 | ||
e2e7aa2d | 4534 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4535 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
cae5b340 | 4536 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4537 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4538 | } |
4539 | ||
e2e7aa2d | 4540 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4541 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
cae5b340 | 4542 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4543 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4544 | } |
4545 | ||
34dc7c2f | 4546 | if (off < 0 || len > MAXOFFSET_T - off) { |
cae5b340 | 4547 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4548 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4549 | } |
4550 | ||
cae5b340 | 4551 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4552 | return (0); |
4553 | } | |
4554 | ||
3558fd73 BB |
4555 | /* |
4556 | * convoff - converts the given data (start, whence) to the | |
4557 | * given whence. | |
4558 | */ | |
4559 | int | |
4560 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4561 | { | |
5484965a | 4562 | vattr_t vap; |
3558fd73 BB |
4563 | int error; |
4564 | ||
4565 | if ((lckdat->l_whence == 2) || (whence == 2)) { | |
8ec27e97 | 4566 | if ((error = zfs_getattr(ip, &vap, 0, CRED()))) |
3558fd73 BB |
4567 | return (error); |
4568 | } | |
4569 | ||
4570 | switch (lckdat->l_whence) { | |
4571 | case 1: | |
4572 | lckdat->l_start += offset; | |
4573 | break; | |
4574 | case 2: | |
5484965a | 4575 | lckdat->l_start += vap.va_size; |
3558fd73 BB |
4576 | /* FALLTHRU */ |
4577 | case 0: | |
4578 | break; | |
4579 | default: | |
a08ee875 | 4580 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4581 | } |
4582 | ||
4583 | if (lckdat->l_start < 0) | |
a08ee875 | 4584 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4585 | |
4586 | switch (whence) { | |
4587 | case 1: | |
4588 | lckdat->l_start -= offset; | |
4589 | break; | |
4590 | case 2: | |
5484965a | 4591 | lckdat->l_start -= vap.va_size; |
3558fd73 BB |
4592 | /* FALLTHRU */ |
4593 | case 0: | |
4594 | break; | |
4595 | default: | |
a08ee875 | 4596 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4597 | } |
4598 | ||
4599 | lckdat->l_whence = (short)whence; | |
4600 | return (0); | |
4601 | } | |
4602 | ||
34dc7c2f BB |
4603 | /* |
4604 | * Free or allocate space in a file. Currently, this function only | |
4605 | * supports the `F_FREESP' command. However, this command is somewhat | |
4606 | * misnamed, as its functionality includes the ability to allocate as | |
4607 | * well as free space. | |
4608 | * | |
3558fd73 | 4609 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4610 | * cmd - action to take (only F_FREESP supported). |
4611 | * bfp - section of file to free/alloc. | |
4612 | * flag - current file open mode flags. | |
4613 | * offset - current file offset. | |
4614 | * cr - credentials of caller [UNUSED]. | |
34dc7c2f | 4615 | * |
a08ee875 | 4616 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4617 | * |
4618 | * Timestamps: | |
3558fd73 | 4619 | * ip - ctime|mtime updated |
34dc7c2f BB |
4620 | */ |
4621 | /* ARGSUSED */ | |
e5c39b95 | 4622 | int |
3558fd73 BB |
4623 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4624 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4625 | { |
3558fd73 | 4626 | znode_t *zp = ITOZ(ip); |
cae5b340 | 4627 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4628 | uint64_t off, len; |
4629 | int error; | |
4630 | ||
cae5b340 | 4631 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4632 | ZFS_VERIFY_ZP(zp); |
4633 | ||
34dc7c2f | 4634 | if (cmd != F_FREESP) { |
cae5b340 | 4635 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4636 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4637 | } |
4638 | ||
cae5b340 AX |
4639 | /* |
4640 | * Callers might not be able to detect properly that we are read-only, | |
4641 | * so check it explicitly here. | |
4642 | */ | |
4643 | if (zfs_is_readonly(zfsvfs)) { | |
4644 | ZFS_EXIT(zfsvfs); | |
4645 | return (SET_ERROR(EROFS)); | |
4646 | } | |
4647 | ||
3558fd73 | 4648 | if ((error = convoff(ip, bfp, 0, offset))) { |
cae5b340 | 4649 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4650 | return (error); |
4651 | } | |
4652 | ||
4653 | if (bfp->l_len < 0) { | |
cae5b340 | 4654 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4655 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4656 | } |
4657 | ||
aec69371 ED |
4658 | /* |
4659 | * Permissions aren't checked on Solaris because on this OS | |
4660 | * zfs_space() can only be called with an opened file handle. | |
4661 | * On Linux we can get here through truncate_range() which | |
4662 | * operates directly on inodes, so we need to check access rights. | |
4663 | */ | |
4664 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
cae5b340 | 4665 | ZFS_EXIT(zfsvfs); |
aec69371 ED |
4666 | return (error); |
4667 | } | |
4668 | ||
34dc7c2f BB |
4669 | off = bfp->l_start; |
4670 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4671 | ||
b128c09f | 4672 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4673 | |
cae5b340 | 4674 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4675 | return (error); |
4676 | } | |
4677 | ||
4678 | /*ARGSUSED*/ | |
e5c39b95 | 4679 | int |
3558fd73 | 4680 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4681 | { |
3558fd73 | 4682 | znode_t *zp = ITOZ(ip); |
cae5b340 | 4683 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4684 | uint32_t gen; |
428870ff | 4685 | uint64_t gen64; |
34dc7c2f BB |
4686 | uint64_t object = zp->z_id; |
4687 | zfid_short_t *zfid; | |
428870ff | 4688 | int size, i, error; |
34dc7c2f | 4689 | |
cae5b340 | 4690 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4691 | ZFS_VERIFY_ZP(zp); |
428870ff | 4692 | |
cae5b340 | 4693 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), |
428870ff | 4694 | &gen64, sizeof (uint64_t))) != 0) { |
cae5b340 | 4695 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4696 | return (error); |
4697 | } | |
4698 | ||
4699 | gen = (uint32_t)gen64; | |
34dc7c2f | 4700 | |
cae5b340 | 4701 | size = SHORT_FID_LEN; |
34dc7c2f BB |
4702 | |
4703 | zfid = (zfid_short_t *)fidp; | |
4704 | ||
4705 | zfid->zf_len = size; | |
4706 | ||
4707 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4708 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4709 | ||
4710 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4711 | if (gen == 0) | |
4712 | gen = 1; | |
4713 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4714 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4715 | ||
cae5b340 | 4716 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4717 | return (0); |
4718 | } | |
4719 | ||
34dc7c2f | 4720 | /*ARGSUSED*/ |
e5c39b95 | 4721 | int |
3558fd73 | 4722 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4723 | { |
3558fd73 | 4724 | znode_t *zp = ITOZ(ip); |
cae5b340 | 4725 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4726 | int error; |
4727 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
4728 | ||
cae5b340 | 4729 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4730 | ZFS_VERIFY_ZP(zp); |
4731 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
cae5b340 | 4732 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4733 | |
4734 | return (error); | |
4735 | } | |
4736 | ||
4737 | /*ARGSUSED*/ | |
e5c39b95 | 4738 | int |
3558fd73 | 4739 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4740 | { |
3558fd73 | 4741 | znode_t *zp = ITOZ(ip); |
cae5b340 | 4742 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4743 | int error; |
4744 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
cae5b340 | 4745 | zilog_t *zilog = zfsvfs->z_log; |
34dc7c2f | 4746 | |
cae5b340 | 4747 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4748 | ZFS_VERIFY_ZP(zp); |
428870ff | 4749 | |
34dc7c2f | 4750 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 4751 | |
cae5b340 | 4752 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4753 | zil_commit(zilog, 0); |
428870ff | 4754 | |
cae5b340 | 4755 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4756 | return (error); |
4757 | } | |
4758 | ||
3558fd73 | 4759 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
4760 | /* |
4761 | * Tunable, both must be a power of 2. | |
4762 | * | |
4763 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
4764 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 4765 | * an arcbuf for a partial block read |
428870ff BB |
4766 | */ |
4767 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
4768 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
4769 | ||
4770 | /*ARGSUSED*/ | |
4771 | static int | |
3558fd73 | 4772 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 4773 | { |
3558fd73 | 4774 | znode_t *zp = ITOZ(ip); |
cae5b340 AX |
4775 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
4776 | int max_blksz = zfsvfs->z_max_blksz; | |
428870ff BB |
4777 | uio_t *uio = &xuio->xu_uio; |
4778 | ssize_t size = uio->uio_resid; | |
4779 | offset_t offset = uio->uio_loffset; | |
4780 | int blksz; | |
4781 | int fullblk, i; | |
4782 | arc_buf_t *abuf; | |
4783 | ssize_t maxsize; | |
4784 | int preamble, postamble; | |
4785 | ||
4786 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
a08ee875 | 4787 | return (SET_ERROR(EINVAL)); |
428870ff | 4788 | |
cae5b340 | 4789 | ZFS_ENTER(zfsvfs); |
428870ff BB |
4790 | ZFS_VERIFY_ZP(zp); |
4791 | switch (ioflag) { | |
4792 | case UIO_WRITE: | |
4793 | /* | |
4794 | * Loan out an arc_buf for write if write size is bigger than | |
4795 | * max_blksz, and the file's block size is also max_blksz. | |
4796 | */ | |
4797 | blksz = max_blksz; | |
4798 | if (size < blksz || zp->z_blksz != blksz) { | |
cae5b340 | 4799 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4800 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4801 | } |
4802 | /* | |
4803 | * Caller requests buffers for write before knowing where the | |
4804 | * write offset might be (e.g. NFS TCP write). | |
4805 | */ | |
4806 | if (offset == -1) { | |
4807 | preamble = 0; | |
4808 | } else { | |
4809 | preamble = P2PHASE(offset, blksz); | |
4810 | if (preamble) { | |
4811 | preamble = blksz - preamble; | |
4812 | size -= preamble; | |
4813 | } | |
4814 | } | |
4815 | ||
4816 | postamble = P2PHASE(size, blksz); | |
4817 | size -= postamble; | |
4818 | ||
4819 | fullblk = size / blksz; | |
4820 | (void) dmu_xuio_init(xuio, | |
4821 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
4822 | |
4823 | /* | |
4824 | * Have to fix iov base/len for partial buffers. They | |
4825 | * currently represent full arc_buf's. | |
4826 | */ | |
4827 | if (preamble) { | |
4828 | /* data begins in the middle of the arc_buf */ | |
4829 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4830 | blksz); | |
4831 | ASSERT(abuf); | |
4832 | (void) dmu_xuio_add(xuio, abuf, | |
4833 | blksz - preamble, preamble); | |
4834 | } | |
4835 | ||
4836 | for (i = 0; i < fullblk; i++) { | |
4837 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4838 | blksz); | |
4839 | ASSERT(abuf); | |
4840 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
4841 | } | |
4842 | ||
4843 | if (postamble) { | |
4844 | /* data ends in the middle of the arc_buf */ | |
4845 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4846 | blksz); | |
4847 | ASSERT(abuf); | |
4848 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
4849 | } | |
4850 | break; | |
4851 | case UIO_READ: | |
4852 | /* | |
4853 | * Loan out an arc_buf for read if the read size is larger than | |
4854 | * the current file block size. Block alignment is not | |
4855 | * considered. Partial arc_buf will be loaned out for read. | |
4856 | */ | |
4857 | blksz = zp->z_blksz; | |
4858 | if (blksz < zcr_blksz_min) | |
4859 | blksz = zcr_blksz_min; | |
4860 | if (blksz > zcr_blksz_max) | |
4861 | blksz = zcr_blksz_max; | |
4862 | /* avoid potential complexity of dealing with it */ | |
4863 | if (blksz > max_blksz) { | |
cae5b340 | 4864 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4865 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4866 | } |
4867 | ||
4868 | maxsize = zp->z_size - uio->uio_loffset; | |
4869 | if (size > maxsize) | |
4870 | size = maxsize; | |
4871 | ||
3558fd73 | 4872 | if (size < blksz) { |
cae5b340 | 4873 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4874 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4875 | } |
4876 | break; | |
4877 | default: | |
cae5b340 | 4878 | ZFS_EXIT(zfsvfs); |
a08ee875 | 4879 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4880 | } |
4881 | ||
4882 | uio->uio_extflg = UIO_XUIO; | |
4883 | XUIO_XUZC_RW(xuio) = ioflag; | |
cae5b340 | 4884 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4885 | return (0); |
4886 | } | |
4887 | ||
4888 | /*ARGSUSED*/ | |
4889 | static int | |
3558fd73 | 4890 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
4891 | { |
4892 | int i; | |
4893 | arc_buf_t *abuf; | |
4894 | int ioflag = XUIO_XUZC_RW(xuio); | |
4895 | ||
4896 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
4897 | ||
4898 | i = dmu_xuio_cnt(xuio); | |
4899 | while (i-- > 0) { | |
4900 | abuf = dmu_xuio_arcbuf(xuio, i); | |
4901 | /* | |
4902 | * if abuf == NULL, it must be a write buffer | |
4903 | * that has been returned in zfs_write(). | |
4904 | */ | |
4905 | if (abuf) | |
4906 | dmu_return_arcbuf(abuf); | |
4907 | ASSERT(abuf || ioflag == UIO_WRITE); | |
4908 | } | |
4909 | ||
4910 | dmu_xuio_fini(xuio); | |
4911 | return (0); | |
4912 | } | |
3558fd73 | 4913 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 BB |
4914 | |
4915 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
cae5b340 AX |
4916 | EXPORT_SYMBOL(zfs_open); |
4917 | EXPORT_SYMBOL(zfs_close); | |
4918 | EXPORT_SYMBOL(zfs_read); | |
4919 | EXPORT_SYMBOL(zfs_write); | |
4920 | EXPORT_SYMBOL(zfs_access); | |
4921 | EXPORT_SYMBOL(zfs_lookup); | |
4922 | EXPORT_SYMBOL(zfs_create); | |
4923 | EXPORT_SYMBOL(zfs_tmpfile); | |
4924 | EXPORT_SYMBOL(zfs_remove); | |
4925 | EXPORT_SYMBOL(zfs_mkdir); | |
4926 | EXPORT_SYMBOL(zfs_rmdir); | |
4927 | EXPORT_SYMBOL(zfs_readdir); | |
4928 | EXPORT_SYMBOL(zfs_fsync); | |
4929 | EXPORT_SYMBOL(zfs_getattr); | |
4930 | EXPORT_SYMBOL(zfs_getattr_fast); | |
4931 | EXPORT_SYMBOL(zfs_setattr); | |
4932 | EXPORT_SYMBOL(zfs_rename); | |
4933 | EXPORT_SYMBOL(zfs_symlink); | |
4934 | EXPORT_SYMBOL(zfs_readlink); | |
4935 | EXPORT_SYMBOL(zfs_link); | |
4936 | EXPORT_SYMBOL(zfs_inactive); | |
4937 | EXPORT_SYMBOL(zfs_space); | |
4938 | EXPORT_SYMBOL(zfs_fid); | |
4939 | EXPORT_SYMBOL(zfs_getsecattr); | |
4940 | EXPORT_SYMBOL(zfs_setsecattr); | |
4941 | EXPORT_SYMBOL(zfs_getpage); | |
4942 | EXPORT_SYMBOL(zfs_putpage); | |
4943 | EXPORT_SYMBOL(zfs_dirty_inode); | |
4944 | EXPORT_SYMBOL(zfs_map); | |
4945 | ||
4946 | /* CSTYLED */ | |
4947 | module_param(zfs_delete_blocks, ulong, 0644); | |
4948 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
c409e464 BB |
4949 | module_param(zfs_read_chunk_size, long, 0644); |
4950 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
4951 | #endif |