]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
21 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
2e528b49 | 23 | * Copyright (c) 2013 by Delphix. All rights reserved. |
34dc7c2f BB |
24 | */ |
25 | ||
26 | /* Portions Copyright 2007 Jeremy Teo */ | |
428870ff | 27 | /* Portions Copyright 2010 Robert Milkowski */ |
34dc7c2f | 28 | |
60101509 | 29 | |
34dc7c2f BB |
30 | #include <sys/types.h> |
31 | #include <sys/param.h> | |
32 | #include <sys/time.h> | |
33 | #include <sys/systm.h> | |
34 | #include <sys/sysmacros.h> | |
35 | #include <sys/resource.h> | |
36 | #include <sys/vfs.h> | |
37 | #include <sys/vfs_opreg.h> | |
34dc7c2f BB |
38 | #include <sys/file.h> |
39 | #include <sys/stat.h> | |
40 | #include <sys/kmem.h> | |
41 | #include <sys/taskq.h> | |
42 | #include <sys/uio.h> | |
43 | #include <sys/vmsystm.h> | |
44 | #include <sys/atomic.h> | |
34dc7c2f | 45 | #include <vm/pvn.h> |
34dc7c2f BB |
46 | #include <sys/pathname.h> |
47 | #include <sys/cmn_err.h> | |
48 | #include <sys/errno.h> | |
49 | #include <sys/unistd.h> | |
50 | #include <sys/zfs_dir.h> | |
51 | #include <sys/zfs_acl.h> | |
52 | #include <sys/zfs_ioctl.h> | |
53 | #include <sys/fs/zfs.h> | |
54 | #include <sys/dmu.h> | |
428870ff | 55 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
56 | #include <sys/spa.h> |
57 | #include <sys/txg.h> | |
58 | #include <sys/dbuf.h> | |
59 | #include <sys/zap.h> | |
428870ff | 60 | #include <sys/sa.h> |
34dc7c2f BB |
61 | #include <sys/dirent.h> |
62 | #include <sys/policy.h> | |
63 | #include <sys/sunddi.h> | |
b128c09f | 64 | #include <sys/sid.h> |
bcf30822 | 65 | #include <sys/mode.h> |
34dc7c2f | 66 | #include "fs/fs_subr.h" |
ebe7e575 | 67 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 68 | #include <sys/zfs_fuid.h> |
428870ff | 69 | #include <sys/zfs_sa.h> |
e5c39b95 | 70 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
71 | #include <sys/dnlc.h> |
72 | #include <sys/zfs_rlock.h> | |
73 | #include <sys/extdirent.h> | |
74 | #include <sys/kidmap.h> | |
428870ff | 75 | #include <sys/cred.h> |
34dc7c2f | 76 | #include <sys/attr.h> |
218b8eaf | 77 | #include <sys/zpl.h> |
34dc7c2f BB |
78 | |
79 | /* | |
80 | * Programming rules. | |
81 | * | |
82 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
83 | * properly lock its in-core state, create a DMU transaction, do the work, | |
84 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
85 | * and wait for the intent log to commit if it is a synchronous operation. | |
86 | * Moreover, the vnode ops must work in both normal and log replay context. | |
87 | * The ordering of events is important to avoid deadlocks and references | |
88 | * to freed memory. The example below illustrates the following Big Rules: | |
89 | * | |
90 | * (1) A check must be made in each zfs thread for a mounted file system. | |
3558fd73 BB |
91 | * This is done avoiding races using ZFS_ENTER(zsb). |
92 | * A ZFS_EXIT(zsb) is needed before all returns. Any znodes | |
34dc7c2f BB |
93 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros |
94 | * can return EIO from the calling function. | |
95 | * | |
3558fd73 | 96 | * (2) iput() should always be the last thing except for zil_commit() |
34dc7c2f BB |
97 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: |
98 | * First, if it's the last reference, the vnode/znode | |
99 | * can be freed, so the zp may point to freed memory. Second, the last | |
100 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
101 | * pushing cached pages (which acquires range locks) and syncing out | |
102 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
103 | * which could deadlock the system if you were already holding one. | |
0a50679c | 104 | * If you must call iput() within a tx then use zfs_iput_async(). |
34dc7c2f BB |
105 | * |
106 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
107 | * as they can span dmu_tx_assign() calls. | |
108 | * | |
384f8a09 MA |
109 | * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to |
110 | * dmu_tx_assign(). This is critical because we don't want to block | |
111 | * while holding locks. | |
112 | * | |
113 | * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This | |
114 | * reduces lock contention and CPU usage when we must wait (note that if | |
115 | * throughput is constrained by the storage, nearly every transaction | |
116 | * must wait). | |
117 | * | |
118 | * Note, in particular, that if a lock is sometimes acquired before | |
119 | * the tx assigns, and sometimes after (e.g. z_lock), then failing | |
120 | * to use a non-blocking assign can deadlock the system. The scenario: | |
34dc7c2f BB |
121 | * |
122 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
123 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
124 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
125 | * forever, because the previous txg can't quiesce until B's tx commits. | |
126 | * | |
3558fd73 | 127 | * If dmu_tx_assign() returns ERESTART and zsb->z_assign is TXG_NOWAIT, |
e8b96c60 MA |
128 | * then drop all locks, call dmu_tx_wait(), and try again. On subsequent |
129 | * calls to dmu_tx_assign(), pass TXG_WAITED rather than TXG_NOWAIT, | |
130 | * to indicate that this operation has already called dmu_tx_wait(). | |
131 | * This will ensure that we don't retry forever, waiting a short bit | |
132 | * each time. | |
34dc7c2f BB |
133 | * |
134 | * (5) If the operation succeeded, generate the intent log entry for it | |
135 | * before dropping locks. This ensures that the ordering of events | |
136 | * in the intent log matches the order in which they actually occurred. | |
d3cc8b15 | 137 | * During ZIL replay the zfs_log_* functions will update the sequence |
fb5f0bc8 | 138 | * number to indicate the zil transaction has replayed. |
34dc7c2f BB |
139 | * |
140 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
141 | * regardless of whether there were any errors. | |
142 | * | |
572e2857 | 143 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
144 | * to ensure that synchronous semantics are provided when necessary. |
145 | * | |
146 | * In general, this is how things should be ordered in each vnode op: | |
147 | * | |
3558fd73 | 148 | * ZFS_ENTER(zsb); // exit if unmounted |
34dc7c2f | 149 | * top: |
3558fd73 | 150 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
151 | * rw_enter(...); // grab any other locks you need |
152 | * tx = dmu_tx_create(...); // get DMU tx | |
153 | * dmu_tx_hold_*(); // hold each object you might modify | |
e8b96c60 | 154 | * error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
155 | * if (error) { |
156 | * rw_exit(...); // drop locks | |
157 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 158 | * iput(...); // release held vnodes |
fb5f0bc8 | 159 | * if (error == ERESTART) { |
e8b96c60 | 160 | * waited = B_TRUE; |
34dc7c2f BB |
161 | * dmu_tx_wait(tx); |
162 | * dmu_tx_abort(tx); | |
163 | * goto top; | |
164 | * } | |
165 | * dmu_tx_abort(tx); // abort DMU tx | |
3558fd73 | 166 | * ZFS_EXIT(zsb); // finished in zfs |
34dc7c2f BB |
167 | * return (error); // really out of space |
168 | * } | |
169 | * error = do_real_work(); // do whatever this VOP does | |
170 | * if (error == 0) | |
171 | * zfs_log_*(...); // on success, make ZIL entry | |
172 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
173 | * rw_exit(...); // drop locks | |
174 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 175 | * iput(...); // release held vnodes |
572e2857 | 176 | * zil_commit(zilog, foid); // synchronous when necessary |
3558fd73 | 177 | * ZFS_EXIT(zsb); // finished in zfs |
34dc7c2f BB |
178 | * return (error); // done, report error |
179 | */ | |
180 | ||
126400a1 BB |
181 | /* |
182 | * Virus scanning is unsupported. It would be possible to add a hook | |
183 | * here to performance the required virus scan. This could be done | |
184 | * entirely in the kernel or potentially as an update to invoke a | |
185 | * scanning utility. | |
186 | */ | |
187 | static int | |
188 | zfs_vscan(struct inode *ip, cred_t *cr, int async) | |
189 | { | |
190 | return (0); | |
191 | } | |
192 | ||
193 | /* ARGSUSED */ | |
194 | int | |
195 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
196 | { | |
197 | znode_t *zp = ITOZ(ip); | |
198 | zfs_sb_t *zsb = ITOZSB(ip); | |
199 | ||
200 | ZFS_ENTER(zsb); | |
201 | ZFS_VERIFY_ZP(zp); | |
202 | ||
203 | /* Honor ZFS_APPENDONLY file attribute */ | |
204 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
205 | ((flag & O_APPEND) == 0)) { | |
206 | ZFS_EXIT(zsb); | |
2e528b49 | 207 | return (SET_ERROR(EPERM)); |
126400a1 BB |
208 | } |
209 | ||
210 | /* Virus scan eligible files on open */ | |
211 | if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && | |
212 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { | |
213 | if (zfs_vscan(ip, cr, 0) != 0) { | |
214 | ZFS_EXIT(zsb); | |
2e528b49 | 215 | return (SET_ERROR(EACCES)); |
126400a1 BB |
216 | } |
217 | } | |
218 | ||
219 | /* Keep a count of the synchronous opens in the znode */ | |
220 | if (flag & O_SYNC) | |
221 | atomic_inc_32(&zp->z_sync_cnt); | |
222 | ||
223 | ZFS_EXIT(zsb); | |
224 | return (0); | |
225 | } | |
226 | EXPORT_SYMBOL(zfs_open); | |
227 | ||
228 | /* ARGSUSED */ | |
229 | int | |
230 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
231 | { | |
232 | znode_t *zp = ITOZ(ip); | |
233 | zfs_sb_t *zsb = ITOZSB(ip); | |
234 | ||
235 | ZFS_ENTER(zsb); | |
236 | ZFS_VERIFY_ZP(zp); | |
237 | ||
7dc71949 | 238 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 239 | if (flag & O_SYNC) |
7dc71949 | 240 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 BB |
241 | |
242 | if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && | |
243 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) | |
244 | VERIFY(zfs_vscan(ip, cr, 1) == 0); | |
245 | ||
246 | ZFS_EXIT(zsb); | |
8780c539 | 247 | return (0); |
126400a1 BB |
248 | } |
249 | EXPORT_SYMBOL(zfs_close); | |
250 | ||
802e7b5f | 251 | #if defined(SEEK_HOLE) && defined(SEEK_DATA) |
cf91b2b6 | 252 | /* |
802e7b5f LD |
253 | * Lseek support for finding holes (cmd == SEEK_HOLE) and |
254 | * data (cmd == SEEK_DATA). "off" is an in/out parameter. | |
cf91b2b6 MA |
255 | */ |
256 | static int | |
802e7b5f | 257 | zfs_holey_common(struct inode *ip, int cmd, loff_t *off) |
cf91b2b6 | 258 | { |
802e7b5f | 259 | znode_t *zp = ITOZ(ip); |
cf91b2b6 MA |
260 | uint64_t noff = (uint64_t)*off; /* new offset */ |
261 | uint64_t file_sz; | |
262 | int error; | |
263 | boolean_t hole; | |
264 | ||
265 | file_sz = zp->z_size; | |
266 | if (noff >= file_sz) { | |
2e528b49 | 267 | return (SET_ERROR(ENXIO)); |
cf91b2b6 MA |
268 | } |
269 | ||
802e7b5f | 270 | if (cmd == SEEK_HOLE) |
cf91b2b6 MA |
271 | hole = B_TRUE; |
272 | else | |
273 | hole = B_FALSE; | |
274 | ||
802e7b5f | 275 | error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff); |
cf91b2b6 | 276 | |
d97aa48f | 277 | if (error == ESRCH) |
2e528b49 | 278 | return (SET_ERROR(ENXIO)); |
d97aa48f MA |
279 | |
280 | /* | |
281 | * We could find a hole that begins after the logical end-of-file, | |
282 | * because dmu_offset_next() only works on whole blocks. If the | |
283 | * EOF falls mid-block, then indicate that the "virtual hole" | |
284 | * at the end of the file begins at the logical EOF, rather than | |
285 | * at the end of the last block. | |
286 | */ | |
287 | if (noff > file_sz) { | |
288 | ASSERT(hole); | |
289 | noff = file_sz; | |
cf91b2b6 MA |
290 | } |
291 | ||
292 | if (noff < *off) | |
293 | return (error); | |
294 | *off = noff; | |
295 | return (error); | |
296 | } | |
802e7b5f LD |
297 | |
298 | int | |
299 | zfs_holey(struct inode *ip, int cmd, loff_t *off) | |
300 | { | |
301 | znode_t *zp = ITOZ(ip); | |
302 | zfs_sb_t *zsb = ITOZSB(ip); | |
303 | int error; | |
304 | ||
305 | ZFS_ENTER(zsb); | |
306 | ZFS_VERIFY_ZP(zp); | |
307 | ||
308 | error = zfs_holey_common(ip, cmd, off); | |
309 | ||
310 | ZFS_EXIT(zsb); | |
311 | return (error); | |
312 | } | |
313 | EXPORT_SYMBOL(zfs_holey); | |
314 | #endif /* SEEK_HOLE && SEEK_DATA */ | |
cf91b2b6 | 315 | |
c0d35759 | 316 | #if defined(_KERNEL) |
34dc7c2f BB |
317 | /* |
318 | * When a file is memory mapped, we must keep the IO data synchronized | |
319 | * between the DMU cache and the memory mapped pages. What this means: | |
320 | * | |
321 | * On Write: If we find a memory mapped page, we write to *both* | |
322 | * the page and the dmu buffer. | |
34dc7c2f | 323 | */ |
d164b209 | 324 | static void |
c0d35759 BB |
325 | update_pages(struct inode *ip, int64_t start, int len, |
326 | objset_t *os, uint64_t oid) | |
34dc7c2f | 327 | { |
c0d35759 BB |
328 | struct address_space *mp = ip->i_mapping; |
329 | struct page *pp; | |
330 | uint64_t nbytes; | |
d164b209 | 331 | int64_t off; |
c0d35759 | 332 | void *pb; |
34dc7c2f | 333 | |
c0d35759 BB |
334 | off = start & (PAGE_CACHE_SIZE-1); |
335 | for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) { | |
336 | nbytes = MIN(PAGE_CACHE_SIZE - off, len); | |
34dc7c2f | 337 | |
c0d35759 BB |
338 | pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT); |
339 | if (pp) { | |
340 | if (mapping_writably_mapped(mp)) | |
341 | flush_dcache_page(pp); | |
34dc7c2f | 342 | |
c0d35759 BB |
343 | pb = kmap(pp); |
344 | (void) dmu_read(os, oid, start+off, nbytes, pb+off, | |
9babb374 | 345 | DMU_READ_PREFETCH); |
c0d35759 BB |
346 | kunmap(pp); |
347 | ||
348 | if (mapping_writably_mapped(mp)) | |
349 | flush_dcache_page(pp); | |
350 | ||
351 | mark_page_accessed(pp); | |
352 | SetPageUptodate(pp); | |
353 | ClearPageError(pp); | |
354 | unlock_page(pp); | |
355 | page_cache_release(pp); | |
34dc7c2f | 356 | } |
c0d35759 | 357 | |
d164b209 | 358 | len -= nbytes; |
34dc7c2f | 359 | off = 0; |
34dc7c2f | 360 | } |
34dc7c2f BB |
361 | } |
362 | ||
363 | /* | |
364 | * When a file is memory mapped, we must keep the IO data synchronized | |
365 | * between the DMU cache and the memory mapped pages. What this means: | |
366 | * | |
367 | * On Read: We "read" preferentially from memory mapped pages, | |
368 | * else we default from the dmu buffer. | |
369 | * | |
370 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
d3cc8b15 | 371 | * the file is memory mapped. |
34dc7c2f BB |
372 | */ |
373 | static int | |
3558fd73 | 374 | mappedread(struct inode *ip, int nbytes, uio_t *uio) |
34dc7c2f | 375 | { |
c0d35759 BB |
376 | struct address_space *mp = ip->i_mapping; |
377 | struct page *pp; | |
3558fd73 BB |
378 | znode_t *zp = ITOZ(ip); |
379 | objset_t *os = ITOZSB(ip)->z_os; | |
34dc7c2f | 380 | int64_t start, off; |
c0d35759 | 381 | uint64_t bytes; |
34dc7c2f BB |
382 | int len = nbytes; |
383 | int error = 0; | |
c0d35759 | 384 | void *pb; |
34dc7c2f BB |
385 | |
386 | start = uio->uio_loffset; | |
c0d35759 BB |
387 | off = start & (PAGE_CACHE_SIZE-1); |
388 | for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) { | |
389 | bytes = MIN(PAGE_CACHE_SIZE - off, len); | |
390 | ||
391 | pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT); | |
392 | if (pp) { | |
393 | ASSERT(PageUptodate(pp)); | |
394 | ||
395 | pb = kmap(pp); | |
396 | error = uiomove(pb + off, bytes, UIO_READ, uio); | |
397 | kunmap(pp); | |
398 | ||
399 | if (mapping_writably_mapped(mp)) | |
400 | flush_dcache_page(pp); | |
401 | ||
402 | mark_page_accessed(pp); | |
403 | unlock_page(pp); | |
404 | page_cache_release(pp); | |
34dc7c2f BB |
405 | } else { |
406 | error = dmu_read_uio(os, zp->z_id, uio, bytes); | |
407 | } | |
c0d35759 | 408 | |
34dc7c2f BB |
409 | len -= bytes; |
410 | off = 0; | |
411 | if (error) | |
412 | break; | |
413 | } | |
414 | return (error); | |
415 | } | |
c0d35759 | 416 | #endif /* _KERNEL */ |
34dc7c2f | 417 | |
c409e464 | 418 | unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ |
34dc7c2f BB |
419 | |
420 | /* | |
421 | * Read bytes from specified file into supplied buffer. | |
422 | * | |
3558fd73 | 423 | * IN: ip - inode of file to be read from. |
34dc7c2f BB |
424 | * uio - structure supplying read location, range info, |
425 | * and return buffer. | |
c0d35759 BB |
426 | * ioflag - FSYNC flags; used to provide FRSYNC semantics. |
427 | * O_DIRECT flag; used to bypass page cache. | |
34dc7c2f | 428 | * cr - credentials of caller. |
34dc7c2f BB |
429 | * |
430 | * OUT: uio - updated offset and range, buffer filled. | |
431 | * | |
d3cc8b15 | 432 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
433 | * |
434 | * Side Effects: | |
3558fd73 | 435 | * inode - atime updated if byte count > 0 |
34dc7c2f BB |
436 | */ |
437 | /* ARGSUSED */ | |
e5c39b95 | 438 | int |
3558fd73 | 439 | zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 440 | { |
3558fd73 BB |
441 | znode_t *zp = ITOZ(ip); |
442 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
443 | objset_t *os; |
444 | ssize_t n, nbytes; | |
149e873a | 445 | int error = 0; |
34dc7c2f | 446 | rl_t *rl; |
3558fd73 | 447 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 448 | xuio_t *xuio = NULL; |
3558fd73 | 449 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f | 450 | |
3558fd73 | 451 | ZFS_ENTER(zsb); |
34dc7c2f | 452 | ZFS_VERIFY_ZP(zp); |
3558fd73 | 453 | os = zsb->z_os; |
34dc7c2f | 454 | |
428870ff | 455 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
3558fd73 | 456 | ZFS_EXIT(zsb); |
2e528b49 | 457 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
458 | } |
459 | ||
460 | /* | |
461 | * Validate file offset | |
462 | */ | |
463 | if (uio->uio_loffset < (offset_t)0) { | |
3558fd73 | 464 | ZFS_EXIT(zsb); |
2e528b49 | 465 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
466 | } |
467 | ||
468 | /* | |
469 | * Fasttrack empty reads | |
470 | */ | |
471 | if (uio->uio_resid == 0) { | |
3558fd73 | 472 | ZFS_EXIT(zsb); |
34dc7c2f BB |
473 | return (0); |
474 | } | |
475 | ||
34dc7c2f BB |
476 | /* |
477 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
478 | */ | |
3558fd73 BB |
479 | if (ioflag & FRSYNC || zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
480 | zil_commit(zsb->z_log, zp->z_id); | |
34dc7c2f BB |
481 | |
482 | /* | |
483 | * Lock the range against changes. | |
484 | */ | |
485 | rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER); | |
486 | ||
487 | /* | |
488 | * If we are reading past end-of-file we can skip | |
489 | * to the end; but we might still need to set atime. | |
490 | */ | |
428870ff | 491 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
492 | error = 0; |
493 | goto out; | |
494 | } | |
495 | ||
428870ff BB |
496 | ASSERT(uio->uio_loffset < zp->z_size); |
497 | n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); | |
498 | ||
3558fd73 | 499 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
500 | if ((uio->uio_extflg == UIO_XUIO) && |
501 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
502 | int nblk; | |
503 | int blksz = zp->z_blksz; | |
504 | uint64_t offset = uio->uio_loffset; | |
505 | ||
506 | xuio = (xuio_t *)uio; | |
507 | if ((ISP2(blksz))) { | |
508 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
509 | blksz)) / blksz; | |
510 | } else { | |
511 | ASSERT(offset + n <= blksz); | |
512 | nblk = 1; | |
513 | } | |
514 | (void) dmu_xuio_init(xuio, nblk); | |
515 | ||
3558fd73 | 516 | if (vn_has_cached_data(ip)) { |
428870ff BB |
517 | /* |
518 | * For simplicity, we always allocate a full buffer | |
519 | * even if we only expect to read a portion of a block. | |
520 | */ | |
521 | while (--nblk >= 0) { | |
522 | (void) dmu_xuio_add(xuio, | |
523 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
524 | blksz), 0, blksz); | |
525 | } | |
526 | } | |
527 | } | |
3558fd73 | 528 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
529 | |
530 | while (n > 0) { | |
531 | nbytes = MIN(n, zfs_read_chunk_size - | |
532 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); | |
533 | ||
c0d35759 | 534 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) |
3558fd73 | 535 | error = mappedread(ip, nbytes, uio); |
34dc7c2f BB |
536 | else |
537 | error = dmu_read_uio(os, zp->z_id, uio, nbytes); | |
c0d35759 | 538 | |
b128c09f BB |
539 | if (error) { |
540 | /* convert checksum errors into IO errors */ | |
541 | if (error == ECKSUM) | |
2e528b49 | 542 | error = SET_ERROR(EIO); |
34dc7c2f | 543 | break; |
b128c09f | 544 | } |
34dc7c2f BB |
545 | |
546 | n -= nbytes; | |
547 | } | |
34dc7c2f BB |
548 | out: |
549 | zfs_range_unlock(rl); | |
550 | ||
3558fd73 | 551 | ZFS_ACCESSTIME_STAMP(zsb, zp); |
3558fd73 | 552 | ZFS_EXIT(zsb); |
34dc7c2f BB |
553 | return (error); |
554 | } | |
e5c39b95 | 555 | EXPORT_SYMBOL(zfs_read); |
34dc7c2f | 556 | |
34dc7c2f BB |
557 | /* |
558 | * Write the bytes to a file. | |
559 | * | |
3558fd73 | 560 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
561 | * uio - structure supplying write location, range info, |
562 | * and data buffer. | |
563 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 564 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 565 | * cr - credentials of caller. |
34dc7c2f BB |
566 | * |
567 | * OUT: uio - updated offset and range. | |
568 | * | |
569 | * RETURN: 0 if success | |
570 | * error code if failure | |
571 | * | |
572 | * Timestamps: | |
3558fd73 | 573 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 574 | */ |
428870ff | 575 | |
34dc7c2f | 576 | /* ARGSUSED */ |
e5c39b95 | 577 | int |
3558fd73 | 578 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 579 | { |
3558fd73 BB |
580 | znode_t *zp = ITOZ(ip); |
581 | rlim64_t limit = uio->uio_limit; | |
34dc7c2f BB |
582 | ssize_t start_resid = uio->uio_resid; |
583 | ssize_t tx_bytes; | |
584 | uint64_t end_size; | |
585 | dmu_tx_t *tx; | |
3558fd73 | 586 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
587 | zilog_t *zilog; |
588 | offset_t woff; | |
589 | ssize_t n, nbytes; | |
590 | rl_t *rl; | |
3558fd73 BB |
591 | int max_blksz = zsb->z_max_blksz; |
592 | int error = 0; | |
9babb374 | 593 | arc_buf_t *abuf; |
3558fd73 | 594 | iovec_t *aiov = NULL; |
428870ff BB |
595 | xuio_t *xuio = NULL; |
596 | int i_iov = 0; | |
428870ff BB |
597 | iovec_t *iovp = uio->uio_iov; |
598 | int write_eof; | |
599 | int count = 0; | |
600 | sa_bulk_attr_t bulk[4]; | |
601 | uint64_t mtime[2], ctime[2]; | |
3558fd73 | 602 | ASSERTV(int iovcnt = uio->uio_iovcnt); |
34dc7c2f | 603 | |
34dc7c2f BB |
604 | /* |
605 | * Fasttrack empty write | |
606 | */ | |
607 | n = start_resid; | |
608 | if (n == 0) | |
609 | return (0); | |
610 | ||
611 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) | |
612 | limit = MAXOFFSET_T; | |
613 | ||
3558fd73 | 614 | ZFS_ENTER(zsb); |
34dc7c2f | 615 | ZFS_VERIFY_ZP(zp); |
b128c09f | 616 | |
3558fd73 BB |
617 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
618 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
619 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); | |
620 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, | |
428870ff BB |
621 | &zp->z_pflags, 8); |
622 | ||
b128c09f BB |
623 | /* |
624 | * If immutable or not appending then return EPERM | |
625 | */ | |
428870ff BB |
626 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
627 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
628 | (uio->uio_loffset < zp->z_size))) { | |
3558fd73 | 629 | ZFS_EXIT(zsb); |
2e528b49 | 630 | return (SET_ERROR(EPERM)); |
b128c09f BB |
631 | } |
632 | ||
3558fd73 | 633 | zilog = zsb->z_log; |
34dc7c2f | 634 | |
428870ff BB |
635 | /* |
636 | * Validate file offset | |
637 | */ | |
638 | woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; | |
639 | if (woff < 0) { | |
3558fd73 | 640 | ZFS_EXIT(zsb); |
2e528b49 | 641 | return (SET_ERROR(EINVAL)); |
428870ff BB |
642 | } |
643 | ||
34dc7c2f BB |
644 | /* |
645 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
646 | * don't hold up txg. | |
428870ff | 647 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 648 | */ |
9cac042c | 649 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
650 | if ((uio->uio_extflg == UIO_XUIO) && |
651 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
652 | xuio = (xuio_t *)uio; | |
653 | else | |
9cac042c | 654 | #endif |
572e2857 | 655 | uio_prefaultpages(MIN(n, max_blksz), uio); |
34dc7c2f BB |
656 | |
657 | /* | |
658 | * If in append mode, set the io offset pointer to eof. | |
659 | */ | |
660 | if (ioflag & FAPPEND) { | |
661 | /* | |
428870ff BB |
662 | * Obtain an appending range lock to guarantee file append |
663 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f BB |
664 | */ |
665 | rl = zfs_range_lock(zp, 0, n, RL_APPEND); | |
428870ff | 666 | woff = rl->r_off; |
34dc7c2f | 667 | if (rl->r_len == UINT64_MAX) { |
428870ff BB |
668 | /* |
669 | * We overlocked the file because this write will cause | |
670 | * the file block size to increase. | |
671 | * Note that zp_size cannot change with this lock held. | |
672 | */ | |
673 | woff = zp->z_size; | |
34dc7c2f | 674 | } |
428870ff | 675 | uio->uio_loffset = woff; |
34dc7c2f | 676 | } else { |
34dc7c2f | 677 | /* |
428870ff BB |
678 | * Note that if the file block size will change as a result of |
679 | * this write, then this range lock will lock the entire file | |
680 | * so that we can re-write the block safely. | |
34dc7c2f BB |
681 | */ |
682 | rl = zfs_range_lock(zp, woff, n, RL_WRITER); | |
683 | } | |
684 | ||
685 | if (woff >= limit) { | |
686 | zfs_range_unlock(rl); | |
3558fd73 | 687 | ZFS_EXIT(zsb); |
2e528b49 | 688 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
689 | } |
690 | ||
691 | if ((woff + n) > limit || woff > (limit - n)) | |
692 | n = limit - woff; | |
693 | ||
428870ff BB |
694 | /* Will this write extend the file length? */ |
695 | write_eof = (woff + n > zp->z_size); | |
696 | ||
697 | end_size = MAX(zp->z_size, woff + n); | |
34dc7c2f BB |
698 | |
699 | /* | |
700 | * Write the file in reasonable size chunks. Each chunk is written | |
701 | * in a separate transaction; this keeps the intent log records small | |
702 | * and allows us to do more fine-grained space accounting. | |
703 | */ | |
704 | while (n > 0) { | |
9babb374 BB |
705 | abuf = NULL; |
706 | woff = uio->uio_loffset; | |
3558fd73 BB |
707 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
708 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 BB |
709 | if (abuf != NULL) |
710 | dmu_return_arcbuf(abuf); | |
2e528b49 | 711 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
712 | break; |
713 | } | |
714 | ||
428870ff BB |
715 | if (xuio && abuf == NULL) { |
716 | ASSERT(i_iov < iovcnt); | |
717 | aiov = &iovp[i_iov]; | |
718 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
719 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
720 | ASSERT((aiov->iov_base == abuf->b_data) || |
721 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
722 | aiov->iov_len == arc_buf_size(abuf))); | |
723 | i_iov++; | |
724 | } else if (abuf == NULL && n >= max_blksz && | |
725 | woff >= zp->z_size && | |
9babb374 BB |
726 | P2PHASE(woff, max_blksz) == 0 && |
727 | zp->z_blksz == max_blksz) { | |
428870ff BB |
728 | /* |
729 | * This write covers a full block. "Borrow" a buffer | |
730 | * from the dmu so that we can fill it before we enter | |
731 | * a transaction. This avoids the possibility of | |
732 | * holding up the transaction if the data copy hangs | |
733 | * up on a pagefault (e.g., from an NFS server mapping). | |
734 | */ | |
9babb374 BB |
735 | size_t cbytes; |
736 | ||
428870ff BB |
737 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
738 | max_blksz); | |
9babb374 BB |
739 | ASSERT(abuf != NULL); |
740 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
741 | if ((error = uiocopy(abuf->b_data, max_blksz, |
742 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
743 | dmu_return_arcbuf(abuf); |
744 | break; | |
745 | } | |
746 | ASSERT(cbytes == max_blksz); | |
747 | } | |
748 | ||
34dc7c2f BB |
749 | /* |
750 | * Start a transaction. | |
751 | */ | |
3558fd73 | 752 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 753 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 754 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 755 | zfs_sa_upgrade_txholds(tx, zp); |
384f8a09 | 756 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 757 | if (error) { |
34dc7c2f | 758 | dmu_tx_abort(tx); |
9babb374 BB |
759 | if (abuf != NULL) |
760 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
761 | break; |
762 | } | |
763 | ||
764 | /* | |
765 | * If zfs_range_lock() over-locked we grow the blocksize | |
766 | * and then reduce the lock range. This will only happen | |
767 | * on the first iteration since zfs_range_reduce() will | |
768 | * shrink down r_len to the appropriate size. | |
769 | */ | |
770 | if (rl->r_len == UINT64_MAX) { | |
771 | uint64_t new_blksz; | |
772 | ||
773 | if (zp->z_blksz > max_blksz) { | |
f1512ee6 MA |
774 | /* |
775 | * File's blocksize is already larger than the | |
776 | * "recordsize" property. Only let it grow to | |
777 | * the next power of 2. | |
778 | */ | |
34dc7c2f | 779 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 MA |
780 | new_blksz = MIN(end_size, |
781 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
782 | } else { |
783 | new_blksz = MIN(end_size, max_blksz); | |
784 | } | |
785 | zfs_grow_blocksize(zp, new_blksz, tx); | |
786 | zfs_range_reduce(rl, woff, n); | |
787 | } | |
788 | ||
789 | /* | |
790 | * XXX - should we really limit each write to z_max_blksz? | |
791 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
792 | */ | |
793 | nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); | |
34dc7c2f | 794 | |
9babb374 BB |
795 | if (abuf == NULL) { |
796 | tx_bytes = uio->uio_resid; | |
428870ff BB |
797 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
798 | uio, nbytes, tx); | |
9babb374 BB |
799 | tx_bytes -= uio->uio_resid; |
800 | } else { | |
801 | tx_bytes = nbytes; | |
428870ff BB |
802 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
803 | /* | |
804 | * If this is not a full block write, but we are | |
805 | * extending the file past EOF and this data starts | |
806 | * block-aligned, use assign_arcbuf(). Otherwise, | |
807 | * write via dmu_write(). | |
808 | */ | |
809 | if (tx_bytes < max_blksz && (!write_eof || | |
810 | aiov->iov_base != abuf->b_data)) { | |
811 | ASSERT(xuio); | |
3558fd73 | 812 | dmu_write(zsb->z_os, zp->z_id, woff, |
428870ff BB |
813 | aiov->iov_len, aiov->iov_base, tx); |
814 | dmu_return_arcbuf(abuf); | |
815 | xuio_stat_wbuf_copied(); | |
816 | } else { | |
817 | ASSERT(xuio || tx_bytes == max_blksz); | |
818 | dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl), | |
819 | woff, abuf, tx); | |
820 | } | |
9babb374 BB |
821 | ASSERT(tx_bytes <= uio->uio_resid); |
822 | uioskip(uio, tx_bytes); | |
823 | } | |
c0d35759 BB |
824 | |
825 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) | |
826 | update_pages(ip, woff, tx_bytes, zsb->z_os, zp->z_id); | |
34dc7c2f BB |
827 | |
828 | /* | |
829 | * If we made no progress, we're done. If we made even | |
830 | * partial progress, update the znode and ZIL accordingly. | |
831 | */ | |
832 | if (tx_bytes == 0) { | |
3558fd73 | 833 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 834 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
835 | dmu_tx_commit(tx); |
836 | ASSERT(error != 0); | |
837 | break; | |
838 | } | |
839 | ||
840 | /* | |
841 | * Clear Set-UID/Set-GID bits on successful write if not | |
842 | * privileged and at least one of the excute bits is set. | |
843 | * | |
844 | * It would be nice to to this after all writes have | |
845 | * been done, but that would still expose the ISUID/ISGID | |
846 | * to another app after the partial write is committed. | |
847 | * | |
572e2857 BB |
848 | * Note: we don't call zfs_fuid_map_id() here because |
849 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
850 | */ |
851 | mutex_enter(&zp->z_acl_lock); | |
428870ff | 852 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 853 | (S_IXUSR >> 6))) != 0 && |
428870ff | 854 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 855 | secpolicy_vnode_setid_retain(cr, |
428870ff BB |
856 | (zp->z_mode & S_ISUID) != 0 && zp->z_uid == 0) != 0) { |
857 | uint64_t newmode; | |
858 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
859 | newmode = zp->z_mode; | |
3558fd73 | 860 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zsb), |
428870ff | 861 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
862 | } |
863 | mutex_exit(&zp->z_acl_lock); | |
864 | ||
428870ff BB |
865 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, |
866 | B_TRUE); | |
34dc7c2f BB |
867 | |
868 | /* | |
869 | * Update the file size (zp_size) if it has changed; | |
870 | * account for possible concurrent updates. | |
871 | */ | |
428870ff BB |
872 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
873 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 874 | uio->uio_loffset); |
428870ff BB |
875 | ASSERT(error == 0); |
876 | } | |
572e2857 BB |
877 | /* |
878 | * If we are replaying and eof is non zero then force | |
879 | * the file size to the specified eof. Note, there's no | |
880 | * concurrency during replay. | |
881 | */ | |
3558fd73 BB |
882 | if (zsb->z_replay && zsb->z_replay_eof != 0) |
883 | zp->z_size = zsb->z_replay_eof; | |
572e2857 | 884 | |
428870ff BB |
885 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
886 | ||
119a394a ED |
887 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
888 | NULL, NULL); | |
34dc7c2f BB |
889 | dmu_tx_commit(tx); |
890 | ||
891 | if (error != 0) | |
892 | break; | |
893 | ASSERT(tx_bytes == nbytes); | |
894 | n -= nbytes; | |
572e2857 BB |
895 | |
896 | if (!xuio && n > 0) | |
897 | uio_prefaultpages(MIN(n, max_blksz), uio); | |
34dc7c2f BB |
898 | } |
899 | ||
900 | zfs_range_unlock(rl); | |
901 | ||
902 | /* | |
903 | * If we're in replay mode, or we made no progress, return error. | |
904 | * Otherwise, it's at least a partial write, so it's successful. | |
905 | */ | |
3558fd73 BB |
906 | if (zsb->z_replay || uio->uio_resid == start_resid) { |
907 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
908 | return (error); |
909 | } | |
910 | ||
428870ff | 911 | if (ioflag & (FSYNC | FDSYNC) || |
3558fd73 | 912 | zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 913 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 914 | |
960e08fe | 915 | zfs_inode_update(zp); |
3558fd73 | 916 | ZFS_EXIT(zsb); |
34dc7c2f BB |
917 | return (0); |
918 | } | |
e5c39b95 | 919 | EXPORT_SYMBOL(zfs_write); |
34dc7c2f | 920 | |
0a50679c BB |
921 | void |
922 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 923 | { |
0a50679c BB |
924 | objset_t *os = ITOZSB(ip)->z_os; |
925 | ||
3558fd73 | 926 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
927 | ASSERT(os != NULL); |
928 | ||
3558fd73 | 929 | if (atomic_read(&ip->i_count) == 1) |
0a50679c | 930 | taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
79c76d5b | 931 | (task_func_t *)iput, ip, TQ_SLEEP); |
3558fd73 BB |
932 | else |
933 | iput(ip); | |
934 | } | |
935 | ||
34dc7c2f | 936 | void |
428870ff | 937 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 938 | { |
428870ff | 939 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
940 | |
941 | if (zgd->zgd_db) | |
942 | dmu_buf_rele(zgd->zgd_db, zgd); | |
943 | ||
944 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f | 945 | |
9babb374 BB |
946 | /* |
947 | * Release the vnode asynchronously as we currently have the | |
948 | * txg stopped from syncing. | |
949 | */ | |
0a50679c | 950 | zfs_iput_async(ZTOI(zp)); |
428870ff BB |
951 | |
952 | if (error == 0 && zgd->zgd_bp) | |
953 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
954 | ||
34dc7c2f BB |
955 | kmem_free(zgd, sizeof (zgd_t)); |
956 | } | |
957 | ||
45d1cae3 BB |
958 | #ifdef DEBUG |
959 | static int zil_fault_io = 0; | |
960 | #endif | |
961 | ||
34dc7c2f BB |
962 | /* |
963 | * Get data to generate a TX_WRITE intent log record. | |
964 | */ | |
965 | int | |
966 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
967 | { | |
3558fd73 BB |
968 | zfs_sb_t *zsb = arg; |
969 | objset_t *os = zsb->z_os; | |
34dc7c2f | 970 | znode_t *zp; |
428870ff BB |
971 | uint64_t object = lr->lr_foid; |
972 | uint64_t offset = lr->lr_offset; | |
973 | uint64_t size = lr->lr_length; | |
974 | blkptr_t *bp = &lr->lr_blkptr; | |
34dc7c2f | 975 | dmu_buf_t *db; |
34dc7c2f | 976 | zgd_t *zgd; |
34dc7c2f BB |
977 | int error = 0; |
978 | ||
428870ff BB |
979 | ASSERT(zio != NULL); |
980 | ASSERT(size != 0); | |
34dc7c2f BB |
981 | |
982 | /* | |
983 | * Nothing to do if the file has been removed | |
984 | */ | |
3558fd73 | 985 | if (zfs_zget(zsb, object, &zp) != 0) |
2e528b49 | 986 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 987 | if (zp->z_unlinked) { |
9babb374 BB |
988 | /* |
989 | * Release the vnode asynchronously as we currently have the | |
990 | * txg stopped from syncing. | |
991 | */ | |
0a50679c | 992 | zfs_iput_async(ZTOI(zp)); |
2e528b49 | 993 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
994 | } |
995 | ||
79c76d5b | 996 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
3558fd73 | 997 | zgd->zgd_zilog = zsb->z_log; |
428870ff BB |
998 | zgd->zgd_private = zp; |
999 | ||
34dc7c2f BB |
1000 | /* |
1001 | * Write records come in two flavors: immediate and indirect. | |
1002 | * For small writes it's cheaper to store the data with the | |
1003 | * log record (immediate); for large writes it's cheaper to | |
1004 | * sync the data and get a pointer to it (indirect) so that | |
1005 | * we don't have to write the data twice. | |
1006 | */ | |
1007 | if (buf != NULL) { /* immediate write */ | |
428870ff | 1008 | zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER); |
34dc7c2f | 1009 | /* test for truncation needs to be done while range locked */ |
428870ff | 1010 | if (offset >= zp->z_size) { |
2e528b49 | 1011 | error = SET_ERROR(ENOENT); |
428870ff BB |
1012 | } else { |
1013 | error = dmu_read(os, object, offset, size, buf, | |
1014 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1015 | } |
428870ff | 1016 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1017 | } else { /* indirect write */ |
34dc7c2f BB |
1018 | /* |
1019 | * Have to lock the whole block to ensure when it's | |
1020 | * written out and it's checksum is being calculated | |
1021 | * that no one can change the data. We need to re-check | |
1022 | * blocksize after we get the lock in case it's changed! | |
1023 | */ | |
1024 | for (;;) { | |
428870ff BB |
1025 | uint64_t blkoff; |
1026 | size = zp->z_blksz; | |
1027 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1028 | offset -= blkoff; | |
1029 | zgd->zgd_rl = zfs_range_lock(zp, offset, size, | |
1030 | RL_READER); | |
1031 | if (zp->z_blksz == size) | |
34dc7c2f | 1032 | break; |
428870ff BB |
1033 | offset += blkoff; |
1034 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f BB |
1035 | } |
1036 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1037 | if (lr->lr_offset >= zp->z_size) |
2e528b49 | 1038 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1039 | #ifdef DEBUG |
1040 | if (zil_fault_io) { | |
2e528b49 | 1041 | error = SET_ERROR(EIO); |
45d1cae3 | 1042 | zil_fault_io = 0; |
45d1cae3 | 1043 | } |
45d1cae3 | 1044 | #endif |
34dc7c2f | 1045 | if (error == 0) |
428870ff BB |
1046 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1047 | DMU_READ_NO_PREFETCH); | |
1048 | ||
1049 | if (error == 0) { | |
03c6040b GW |
1050 | blkptr_t *obp = dmu_buf_get_blkptr(db); |
1051 | if (obp) { | |
1052 | ASSERT(BP_IS_HOLE(bp)); | |
1053 | *bp = *obp; | |
1054 | } | |
1055 | ||
428870ff BB |
1056 | zgd->zgd_db = db; |
1057 | zgd->zgd_bp = bp; | |
1058 | ||
1059 | ASSERT(db->db_offset == offset); | |
1060 | ASSERT(db->db_size == size); | |
1061 | ||
1062 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1063 | zfs_get_done, zgd); | |
1064 | ASSERT(error || lr->lr_length <= zp->z_blksz); | |
1065 | ||
1066 | /* | |
1067 | * On success, we need to wait for the write I/O | |
1068 | * initiated by dmu_sync() to complete before we can | |
1069 | * release this dbuf. We will finish everything up | |
1070 | * in the zfs_get_done() callback. | |
1071 | */ | |
1072 | if (error == 0) | |
1073 | return (0); | |
1074 | ||
1075 | if (error == EALREADY) { | |
1076 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
1077 | error = 0; | |
1078 | } | |
1079 | } | |
34dc7c2f | 1080 | } |
428870ff BB |
1081 | |
1082 | zfs_get_done(zgd, error); | |
1083 | ||
34dc7c2f BB |
1084 | return (error); |
1085 | } | |
1086 | ||
1087 | /*ARGSUSED*/ | |
3558fd73 BB |
1088 | int |
1089 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1090 | { |
3558fd73 BB |
1091 | znode_t *zp = ITOZ(ip); |
1092 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
1093 | int error; |
1094 | ||
3558fd73 | 1095 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1096 | ZFS_VERIFY_ZP(zp); |
1097 | ||
1098 | if (flag & V_ACE_MASK) | |
1099 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1100 | else | |
1101 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1102 | ||
3558fd73 | 1103 | ZFS_EXIT(zsb); |
45d1cae3 BB |
1104 | return (error); |
1105 | } | |
3558fd73 | 1106 | EXPORT_SYMBOL(zfs_access); |
45d1cae3 | 1107 | |
34dc7c2f BB |
1108 | /* |
1109 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1110 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1111 | * |
3558fd73 | 1112 | * IN: dip - inode of directory to search. |
34dc7c2f | 1113 | * nm - name of entry to lookup. |
34dc7c2f | 1114 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1115 | * cr - credentials of caller. |
34dc7c2f BB |
1116 | * direntflags - directory lookup flags |
1117 | * realpnp - returned pathname. | |
1118 | * | |
3558fd73 | 1119 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1120 | * |
d3cc8b15 | 1121 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1122 | * |
1123 | * Timestamps: | |
1124 | * NA | |
1125 | */ | |
1126 | /* ARGSUSED */ | |
e5c39b95 | 1127 | int |
3558fd73 BB |
1128 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1129 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1130 | { |
3558fd73 BB |
1131 | znode_t *zdp = ITOZ(dip); |
1132 | zfs_sb_t *zsb = ITOZSB(dip); | |
1133 | int error = 0; | |
45d1cae3 BB |
1134 | |
1135 | /* fast path */ | |
1136 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { | |
1137 | ||
3558fd73 | 1138 | if (!S_ISDIR(dip->i_mode)) { |
2e528b49 | 1139 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1140 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 1141 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1142 | } |
1143 | ||
1144 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1145 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1146 | if (!error) { | |
3558fd73 BB |
1147 | *ipp = dip; |
1148 | igrab(*ipp); | |
45d1cae3 BB |
1149 | return (0); |
1150 | } | |
1151 | return (error); | |
3558fd73 | 1152 | #ifdef HAVE_DNLC |
45d1cae3 BB |
1153 | } else { |
1154 | vnode_t *tvp = dnlc_lookup(dvp, nm); | |
1155 | ||
1156 | if (tvp) { | |
1157 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1158 | if (error) { | |
3558fd73 | 1159 | iput(tvp); |
45d1cae3 BB |
1160 | return (error); |
1161 | } | |
1162 | if (tvp == DNLC_NO_VNODE) { | |
3558fd73 | 1163 | iput(tvp); |
2e528b49 | 1164 | return (SET_ERROR(ENOENT)); |
45d1cae3 BB |
1165 | } else { |
1166 | *vpp = tvp; | |
1167 | return (specvp_check(vpp, cr)); | |
1168 | } | |
1169 | } | |
3558fd73 | 1170 | #endif /* HAVE_DNLC */ |
45d1cae3 BB |
1171 | } |
1172 | } | |
1173 | ||
3558fd73 | 1174 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1175 | ZFS_VERIFY_ZP(zdp); |
1176 | ||
3558fd73 | 1177 | *ipp = NULL; |
34dc7c2f BB |
1178 | |
1179 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1180 | /* |
1181 | * We don't allow recursive attributes.. | |
1182 | * Maybe someday we will. | |
1183 | */ | |
428870ff | 1184 | if (zdp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1185 | ZFS_EXIT(zsb); |
2e528b49 | 1186 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1187 | } |
1188 | ||
3558fd73 BB |
1189 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
1190 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1191 | return (error); |
1192 | } | |
1193 | ||
1194 | /* | |
1195 | * Do we have permission to get into attribute directory? | |
1196 | */ | |
1197 | ||
3558fd73 | 1198 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1199 | B_FALSE, cr))) { |
3558fd73 BB |
1200 | iput(*ipp); |
1201 | *ipp = NULL; | |
34dc7c2f BB |
1202 | } |
1203 | ||
3558fd73 | 1204 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1205 | return (error); |
1206 | } | |
1207 | ||
3558fd73 BB |
1208 | if (!S_ISDIR(dip->i_mode)) { |
1209 | ZFS_EXIT(zsb); | |
2e528b49 | 1210 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1211 | } |
1212 | ||
1213 | /* | |
1214 | * Check accessibility of directory. | |
1215 | */ | |
1216 | ||
149e873a | 1217 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
3558fd73 | 1218 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1219 | return (error); |
1220 | } | |
1221 | ||
3558fd73 | 1222 | if (zsb->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1223 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1224 | ZFS_EXIT(zsb); |
2e528b49 | 1225 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1226 | } |
1227 | ||
3558fd73 BB |
1228 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1229 | if ((error == 0) && (*ipp)) | |
1230 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1231 | |
3558fd73 | 1232 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1233 | return (error); |
1234 | } | |
e5c39b95 | 1235 | EXPORT_SYMBOL(zfs_lookup); |
34dc7c2f BB |
1236 | |
1237 | /* | |
1238 | * Attempt to create a new entry in a directory. If the entry | |
1239 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1240 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1241 | * |
3558fd73 | 1242 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1243 | * name - name of new file entry. |
1244 | * vap - attributes of new file. | |
1245 | * excl - flag indicating exclusive or non-exclusive mode. | |
1246 | * mode - mode to open file with. | |
1247 | * cr - credentials of caller. | |
1248 | * flag - large file flag [UNUSED]. | |
3558fd73 | 1249 | * vsecp - ACL to be set |
34dc7c2f | 1250 | * |
3558fd73 | 1251 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1252 | * |
d3cc8b15 | 1253 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1254 | * |
1255 | * Timestamps: | |
3558fd73 BB |
1256 | * dip - ctime|mtime updated if new entry created |
1257 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1258 | */ |
1259 | ||
1260 | /* ARGSUSED */ | |
e5c39b95 | 1261 | int |
3558fd73 BB |
1262 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1263 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1264 | { |
3558fd73 BB |
1265 | znode_t *zp, *dzp = ITOZ(dip); |
1266 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1267 | zilog_t *zilog; |
1268 | objset_t *os; | |
1269 | zfs_dirlock_t *dl; | |
1270 | dmu_tx_t *tx; | |
1271 | int error; | |
b128c09f | 1272 | uid_t uid; |
149e873a | 1273 | gid_t gid; |
428870ff | 1274 | zfs_acl_ids_t acl_ids; |
9babb374 | 1275 | boolean_t fuid_dirtied; |
428870ff | 1276 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 1277 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1278 | |
1279 | /* | |
1280 | * If we have an ephemeral id, ACL, or XVATTR then | |
1281 | * make sure file system is at proper version | |
1282 | */ | |
1283 | ||
149e873a | 1284 | gid = crgetgid(cr); |
3558fd73 | 1285 | uid = crgetuid(cr); |
b128c09f | 1286 | |
3558fd73 BB |
1287 | if (zsb->z_use_fuids == B_FALSE && |
1288 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
2e528b49 | 1289 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1290 | |
3558fd73 | 1291 | ZFS_ENTER(zsb); |
34dc7c2f | 1292 | ZFS_VERIFY_ZP(dzp); |
3558fd73 BB |
1293 | os = zsb->z_os; |
1294 | zilog = zsb->z_log; | |
34dc7c2f | 1295 | |
3558fd73 | 1296 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1297 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1298 | ZFS_EXIT(zsb); |
2e528b49 | 1299 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1300 | } |
1301 | ||
5484965a | 1302 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1303 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1304 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1305 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1306 | return (error); |
1307 | } | |
1308 | } | |
34dc7c2f | 1309 | |
3558fd73 BB |
1310 | top: |
1311 | *ipp = NULL; | |
34dc7c2f BB |
1312 | if (*name == '\0') { |
1313 | /* | |
1314 | * Null component name refers to the directory itself. | |
1315 | */ | |
3558fd73 | 1316 | igrab(dip); |
34dc7c2f BB |
1317 | zp = dzp; |
1318 | dl = NULL; | |
1319 | error = 0; | |
1320 | } else { | |
3558fd73 | 1321 | /* possible igrab(zp) */ |
34dc7c2f BB |
1322 | int zflg = 0; |
1323 | ||
1324 | if (flag & FIGNORECASE) | |
1325 | zflg |= ZCILOOK; | |
1326 | ||
1327 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1328 | NULL, NULL); | |
1329 | if (error) { | |
572e2857 BB |
1330 | if (have_acl) |
1331 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1332 | if (strcmp(name, "..") == 0) |
2e528b49 | 1333 | error = SET_ERROR(EISDIR); |
3558fd73 | 1334 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1335 | return (error); |
1336 | } | |
1337 | } | |
428870ff | 1338 | |
34dc7c2f BB |
1339 | if (zp == NULL) { |
1340 | uint64_t txtype; | |
1341 | ||
1342 | /* | |
1343 | * Create a new file object and update the directory | |
1344 | * to reference it. | |
1345 | */ | |
149e873a | 1346 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1347 | if (have_acl) |
1348 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1349 | goto out; |
1350 | } | |
1351 | ||
1352 | /* | |
1353 | * We only support the creation of regular files in | |
1354 | * extended attribute directories. | |
1355 | */ | |
428870ff | 1356 | |
3558fd73 | 1357 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1358 | if (have_acl) |
1359 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 1360 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1361 | goto out; |
1362 | } | |
1363 | ||
428870ff BB |
1364 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1365 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1366 | goto out; |
428870ff BB |
1367 | have_acl = B_TRUE; |
1368 | ||
3558fd73 | 1369 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1370 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1371 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1372 | goto out; |
1373 | } | |
1374 | ||
34dc7c2f | 1375 | tx = dmu_tx_create(os); |
428870ff BB |
1376 | |
1377 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1378 | ZFS_SA_BASE_ATTR_SIZE); | |
1379 | ||
3558fd73 | 1380 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1381 | if (fuid_dirtied) |
3558fd73 | 1382 | zfs_fuid_txhold(zsb, tx); |
34dc7c2f | 1383 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1384 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
3558fd73 | 1385 | if (!zsb->z_use_sa && |
428870ff | 1386 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1387 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1388 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1389 | } |
e8b96c60 | 1390 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1391 | if (error) { |
1392 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1393 | if (error == ERESTART) { |
e8b96c60 | 1394 | waited = B_TRUE; |
34dc7c2f BB |
1395 | dmu_tx_wait(tx); |
1396 | dmu_tx_abort(tx); | |
1397 | goto top; | |
1398 | } | |
428870ff | 1399 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1400 | dmu_tx_abort(tx); |
3558fd73 | 1401 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1402 | return (error); |
1403 | } | |
428870ff | 1404 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 BB |
1405 | |
1406 | if (fuid_dirtied) | |
3558fd73 | 1407 | zfs_fuid_sync(zsb, tx); |
9babb374 | 1408 | |
34dc7c2f BB |
1409 | (void) zfs_link_create(dl, zp, tx, ZNEW); |
1410 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); | |
1411 | if (flag & FIGNORECASE) | |
1412 | txtype |= TX_CI; | |
1413 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1414 | vsecp, acl_ids.z_fuidp, vap); |
1415 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1416 | dmu_tx_commit(tx); |
1417 | } else { | |
1418 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1419 | ||
572e2857 BB |
1420 | if (have_acl) |
1421 | zfs_acl_ids_free(&acl_ids); | |
1422 | have_acl = B_FALSE; | |
1423 | ||
34dc7c2f BB |
1424 | /* |
1425 | * A directory entry already exists for this name. | |
1426 | */ | |
1427 | /* | |
1428 | * Can't truncate an existing file if in exclusive mode. | |
1429 | */ | |
3558fd73 | 1430 | if (excl) { |
2e528b49 | 1431 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1432 | goto out; |
1433 | } | |
1434 | /* | |
1435 | * Can't open a directory for writing. | |
1436 | */ | |
3558fd73 | 1437 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1438 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1439 | goto out; |
1440 | } | |
1441 | /* | |
1442 | * Verify requested access to file. | |
1443 | */ | |
1444 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1445 | goto out; | |
1446 | } | |
1447 | ||
1448 | mutex_enter(&dzp->z_lock); | |
1449 | dzp->z_seq++; | |
1450 | mutex_exit(&dzp->z_lock); | |
1451 | ||
1452 | /* | |
1453 | * Truncate regular files if requested. | |
1454 | */ | |
3558fd73 BB |
1455 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1456 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f BB |
1457 | /* we can't hold any locks when calling zfs_freesp() */ |
1458 | zfs_dirent_unlock(dl); | |
1459 | dl = NULL; | |
34dc7c2f | 1460 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1461 | } |
1462 | } | |
1463 | out: | |
1464 | ||
1465 | if (dl) | |
1466 | zfs_dirent_unlock(dl); | |
1467 | ||
1468 | if (error) { | |
1469 | if (zp) | |
3558fd73 | 1470 | iput(ZTOI(zp)); |
34dc7c2f | 1471 | } else { |
960e08fe BB |
1472 | zfs_inode_update(dzp); |
1473 | zfs_inode_update(zp); | |
3558fd73 | 1474 | *ipp = ZTOI(zp); |
34dc7c2f | 1475 | } |
34dc7c2f | 1476 | |
3558fd73 | 1477 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1478 | zil_commit(zilog, 0); |
428870ff | 1479 | |
3558fd73 | 1480 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1481 | return (error); |
1482 | } | |
e5c39b95 | 1483 | EXPORT_SYMBOL(zfs_create); |
34dc7c2f BB |
1484 | |
1485 | /* | |
1486 | * Remove an entry from a directory. | |
1487 | * | |
3558fd73 | 1488 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1489 | * name - name of entry to remove. |
1490 | * cr - credentials of caller. | |
34dc7c2f BB |
1491 | * |
1492 | * RETURN: 0 if success | |
1493 | * error code if failure | |
1494 | * | |
1495 | * Timestamps: | |
3558fd73 BB |
1496 | * dip - ctime|mtime |
1497 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1498 | */ |
428870ff BB |
1499 | |
1500 | uint64_t null_xattr = 0; | |
1501 | ||
34dc7c2f | 1502 | /*ARGSUSED*/ |
e5c39b95 | 1503 | int |
3558fd73 | 1504 | zfs_remove(struct inode *dip, char *name, cred_t *cr) |
34dc7c2f | 1505 | { |
3558fd73 | 1506 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1507 | znode_t *xzp; |
3558fd73 BB |
1508 | struct inode *ip; |
1509 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f | 1510 | zilog_t *zilog; |
3558fd73 BB |
1511 | uint64_t xattr_obj; |
1512 | uint64_t xattr_obj_unlinked = 0; | |
572e2857 | 1513 | uint64_t obj = 0; |
34dc7c2f BB |
1514 | zfs_dirlock_t *dl; |
1515 | dmu_tx_t *tx; | |
3558fd73 | 1516 | boolean_t unlinked; |
34dc7c2f BB |
1517 | uint64_t txtype; |
1518 | pathname_t *realnmp = NULL; | |
3558fd73 | 1519 | #ifdef HAVE_PN_UTILS |
34dc7c2f | 1520 | pathname_t realnm; |
3558fd73 | 1521 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
1522 | int error; |
1523 | int zflg = ZEXISTS; | |
e8b96c60 | 1524 | boolean_t waited = B_FALSE; |
34dc7c2f | 1525 | |
3558fd73 | 1526 | ZFS_ENTER(zsb); |
34dc7c2f | 1527 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1528 | zilog = zsb->z_log; |
34dc7c2f | 1529 | |
3558fd73 | 1530 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1531 | if (flags & FIGNORECASE) { |
1532 | zflg |= ZCILOOK; | |
1533 | pn_alloc(&realnm); | |
1534 | realnmp = &realnm; | |
1535 | } | |
3558fd73 | 1536 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
1537 | |
1538 | top: | |
572e2857 BB |
1539 | xattr_obj = 0; |
1540 | xzp = NULL; | |
34dc7c2f BB |
1541 | /* |
1542 | * Attempt to lock directory; fail if entry doesn't exist. | |
1543 | */ | |
149e873a BB |
1544 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1545 | NULL, realnmp))) { | |
3558fd73 | 1546 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1547 | if (realnmp) |
1548 | pn_free(realnmp); | |
3558fd73 BB |
1549 | #endif /* HAVE_PN_UTILS */ |
1550 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1551 | return (error); |
1552 | } | |
1553 | ||
3558fd73 | 1554 | ip = ZTOI(zp); |
34dc7c2f | 1555 | |
149e873a | 1556 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1557 | goto out; |
1558 | } | |
1559 | ||
1560 | /* | |
1561 | * Need to use rmdir for removing directories. | |
1562 | */ | |
3558fd73 | 1563 | if (S_ISDIR(ip->i_mode)) { |
2e528b49 | 1564 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1565 | goto out; |
1566 | } | |
1567 | ||
3558fd73 | 1568 | #ifdef HAVE_DNLC |
34dc7c2f BB |
1569 | if (realnmp) |
1570 | dnlc_remove(dvp, realnmp->pn_buf); | |
1571 | else | |
1572 | dnlc_remove(dvp, name); | |
3558fd73 | 1573 | #endif /* HAVE_DNLC */ |
34dc7c2f BB |
1574 | |
1575 | /* | |
3558fd73 BB |
1576 | * We never delete the znode and always place it in the unlinked |
1577 | * set. The dentry cache will always hold the last reference and | |
1578 | * is responsible for safely freeing the znode. | |
34dc7c2f | 1579 | */ |
572e2857 | 1580 | obj = zp->z_id; |
3558fd73 | 1581 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 1582 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1583 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1584 | zfs_sa_upgrade_txholds(tx, zp); | |
1585 | zfs_sa_upgrade_txholds(tx, dzp); | |
34dc7c2f BB |
1586 | |
1587 | /* are there any extended attributes? */ | |
3558fd73 | 1588 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1589 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1590 | if (error == 0 && xattr_obj) { |
3558fd73 | 1591 | error = zfs_zget(zsb, xattr_obj, &xzp); |
c99c9001 | 1592 | ASSERT0(error); |
428870ff BB |
1593 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1594 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1595 | } |
1596 | ||
34dc7c2f | 1597 | /* charge as an update -- would be nice not to charge at all */ |
3558fd73 | 1598 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1599 | |
e8b96c60 | 1600 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1601 | if (error) { |
1602 | zfs_dirent_unlock(dl); | |
3558fd73 | 1603 | iput(ip); |
572e2857 | 1604 | if (xzp) |
3558fd73 | 1605 | iput(ZTOI(xzp)); |
fb5f0bc8 | 1606 | if (error == ERESTART) { |
e8b96c60 | 1607 | waited = B_TRUE; |
34dc7c2f BB |
1608 | dmu_tx_wait(tx); |
1609 | dmu_tx_abort(tx); | |
1610 | goto top; | |
1611 | } | |
3558fd73 | 1612 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1613 | if (realnmp) |
1614 | pn_free(realnmp); | |
3558fd73 | 1615 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f | 1616 | dmu_tx_abort(tx); |
3558fd73 | 1617 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1618 | return (error); |
1619 | } | |
1620 | ||
1621 | /* | |
1622 | * Remove the directory entry. | |
1623 | */ | |
1624 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1625 | ||
1626 | if (error) { | |
1627 | dmu_tx_commit(tx); | |
1628 | goto out; | |
1629 | } | |
1630 | ||
1631 | if (unlinked) { | |
572e2857 BB |
1632 | /* |
1633 | * Hold z_lock so that we can make sure that the ACL obj | |
1634 | * hasn't changed. Could have been deleted due to | |
1635 | * zfs_sa_upgrade(). | |
1636 | */ | |
1637 | mutex_enter(&zp->z_lock); | |
3558fd73 | 1638 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1639 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
572e2857 | 1640 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1641 | zfs_unlinked_add(zp, tx); |
1642 | } | |
1643 | ||
1644 | txtype = TX_REMOVE; | |
3558fd73 | 1645 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1646 | if (flags & FIGNORECASE) |
1647 | txtype |= TX_CI; | |
3558fd73 | 1648 | #endif /* HAVE_PN_UTILS */ |
572e2857 | 1649 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1650 | |
1651 | dmu_tx_commit(tx); | |
1652 | out: | |
3558fd73 | 1653 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1654 | if (realnmp) |
1655 | pn_free(realnmp); | |
3558fd73 | 1656 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
1657 | |
1658 | zfs_dirent_unlock(dl); | |
960e08fe BB |
1659 | zfs_inode_update(dzp); |
1660 | zfs_inode_update(zp); | |
1661 | if (xzp) | |
1662 | zfs_inode_update(xzp); | |
34dc7c2f | 1663 | |
3558fd73 | 1664 | iput(ip); |
428870ff | 1665 | if (xzp) |
3558fd73 | 1666 | iput(ZTOI(xzp)); |
428870ff | 1667 | |
3558fd73 | 1668 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1669 | zil_commit(zilog, 0); |
34dc7c2f | 1670 | |
3558fd73 | 1671 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1672 | return (error); |
1673 | } | |
e5c39b95 | 1674 | EXPORT_SYMBOL(zfs_remove); |
34dc7c2f BB |
1675 | |
1676 | /* | |
3558fd73 | 1677 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1678 | * provided. Return a pointer to the inserted directory. |
1679 | * | |
3558fd73 | 1680 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1681 | * dirname - name of new directory. |
1682 | * vap - attributes of new directory. | |
1683 | * cr - credentials of caller. | |
34dc7c2f BB |
1684 | * vsecp - ACL to be set |
1685 | * | |
3558fd73 | 1686 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1687 | * |
1688 | * RETURN: 0 if success | |
1689 | * error code if failure | |
1690 | * | |
1691 | * Timestamps: | |
3558fd73 BB |
1692 | * dip - ctime|mtime updated |
1693 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1694 | */ |
1695 | /*ARGSUSED*/ | |
e5c39b95 | 1696 | int |
3558fd73 BB |
1697 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1698 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1699 | { |
3558fd73 BB |
1700 | znode_t *zp, *dzp = ITOZ(dip); |
1701 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1702 | zilog_t *zilog; |
1703 | zfs_dirlock_t *dl; | |
1704 | uint64_t txtype; | |
1705 | dmu_tx_t *tx; | |
1706 | int error; | |
34dc7c2f | 1707 | int zf = ZNEW; |
b128c09f BB |
1708 | uid_t uid; |
1709 | gid_t gid = crgetgid(cr); | |
428870ff | 1710 | zfs_acl_ids_t acl_ids; |
9babb374 | 1711 | boolean_t fuid_dirtied; |
e8b96c60 | 1712 | boolean_t waited = B_FALSE; |
34dc7c2f | 1713 | |
3558fd73 | 1714 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1715 | |
1716 | /* | |
1717 | * If we have an ephemeral id, ACL, or XVATTR then | |
1718 | * make sure file system is at proper version | |
1719 | */ | |
1720 | ||
3558fd73 BB |
1721 | uid = crgetuid(cr); |
1722 | if (zsb->z_use_fuids == B_FALSE && | |
1723 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
2e528b49 | 1724 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1725 | |
3558fd73 | 1726 | ZFS_ENTER(zsb); |
34dc7c2f | 1727 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1728 | zilog = zsb->z_log; |
34dc7c2f | 1729 | |
428870ff | 1730 | if (dzp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1731 | ZFS_EXIT(zsb); |
2e528b49 | 1732 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1733 | } |
1734 | ||
3558fd73 | 1735 | if (zsb->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1736 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1737 | ZFS_EXIT(zsb); |
2e528b49 | 1738 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1739 | } |
1740 | if (flags & FIGNORECASE) | |
1741 | zf |= ZCILOOK; | |
1742 | ||
5484965a | 1743 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1744 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1745 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1746 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1747 | return (error); |
1748 | } | |
428870ff | 1749 | } |
34dc7c2f | 1750 | |
428870ff BB |
1751 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1752 | vsecp, &acl_ids)) != 0) { | |
3558fd73 | 1753 | ZFS_EXIT(zsb); |
428870ff BB |
1754 | return (error); |
1755 | } | |
34dc7c2f BB |
1756 | /* |
1757 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1758 | * |
1759 | * Existence is checked first to make sure we don't return | |
1760 | * EACCES instead of EEXIST which can cause some applications | |
1761 | * to fail. | |
34dc7c2f BB |
1762 | */ |
1763 | top: | |
3558fd73 | 1764 | *ipp = NULL; |
34dc7c2f | 1765 | |
149e873a BB |
1766 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1767 | NULL, NULL))) { | |
428870ff | 1768 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 1769 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1770 | return (error); |
1771 | } | |
1772 | ||
149e873a | 1773 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 1774 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1775 | zfs_dirent_unlock(dl); |
3558fd73 | 1776 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1777 | return (error); |
1778 | } | |
1779 | ||
3558fd73 | 1780 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1781 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 1782 | zfs_dirent_unlock(dl); |
3558fd73 | 1783 | ZFS_EXIT(zsb); |
2e528b49 | 1784 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
1785 | } |
1786 | ||
34dc7c2f BB |
1787 | /* |
1788 | * Add a new entry to the directory. | |
1789 | */ | |
3558fd73 | 1790 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f BB |
1791 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
1792 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
3558fd73 | 1793 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1794 | if (fuid_dirtied) |
3558fd73 BB |
1795 | zfs_fuid_txhold(zsb, tx); |
1796 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
1797 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
1798 | acl_ids.z_aclp->z_acl_bytes); | |
1799 | } | |
1800 | ||
1801 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1802 | ZFS_SA_BASE_ATTR_SIZE); | |
1803 | ||
e8b96c60 | 1804 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1805 | if (error) { |
1806 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1807 | if (error == ERESTART) { |
e8b96c60 | 1808 | waited = B_TRUE; |
34dc7c2f BB |
1809 | dmu_tx_wait(tx); |
1810 | dmu_tx_abort(tx); | |
1811 | goto top; | |
1812 | } | |
428870ff | 1813 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1814 | dmu_tx_abort(tx); |
3558fd73 | 1815 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1816 | return (error); |
1817 | } | |
1818 | ||
1819 | /* | |
1820 | * Create new node. | |
1821 | */ | |
428870ff | 1822 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f | 1823 | |
9babb374 | 1824 | if (fuid_dirtied) |
3558fd73 | 1825 | zfs_fuid_sync(zsb, tx); |
428870ff | 1826 | |
34dc7c2f BB |
1827 | /* |
1828 | * Now put new name in parent dir. | |
1829 | */ | |
1830 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
1831 | ||
3558fd73 | 1832 | *ipp = ZTOI(zp); |
34dc7c2f BB |
1833 | |
1834 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
1835 | if (flags & FIGNORECASE) | |
1836 | txtype |= TX_CI; | |
9babb374 BB |
1837 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
1838 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 1839 | |
9babb374 | 1840 | zfs_acl_ids_free(&acl_ids); |
428870ff | 1841 | |
34dc7c2f BB |
1842 | dmu_tx_commit(tx); |
1843 | ||
1844 | zfs_dirent_unlock(dl); | |
1845 | ||
3558fd73 | 1846 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1847 | zil_commit(zilog, 0); |
428870ff | 1848 | |
960e08fe BB |
1849 | zfs_inode_update(dzp); |
1850 | zfs_inode_update(zp); | |
3558fd73 | 1851 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1852 | return (0); |
1853 | } | |
e5c39b95 | 1854 | EXPORT_SYMBOL(zfs_mkdir); |
34dc7c2f BB |
1855 | |
1856 | /* | |
1857 | * Remove a directory subdir entry. If the current working | |
1858 | * directory is the same as the subdir to be removed, the | |
1859 | * remove will fail. | |
1860 | * | |
3558fd73 | 1861 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 1862 | * name - name of directory to be removed. |
3558fd73 | 1863 | * cwd - inode of current working directory. |
34dc7c2f | 1864 | * cr - credentials of caller. |
34dc7c2f BB |
1865 | * flags - case flags |
1866 | * | |
d3cc8b15 | 1867 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1868 | * |
1869 | * Timestamps: | |
3558fd73 | 1870 | * dip - ctime|mtime updated |
34dc7c2f BB |
1871 | */ |
1872 | /*ARGSUSED*/ | |
e5c39b95 | 1873 | int |
3558fd73 BB |
1874 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
1875 | int flags) | |
34dc7c2f | 1876 | { |
3558fd73 | 1877 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 1878 | znode_t *zp; |
3558fd73 BB |
1879 | struct inode *ip; |
1880 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1881 | zilog_t *zilog; |
1882 | zfs_dirlock_t *dl; | |
1883 | dmu_tx_t *tx; | |
1884 | int error; | |
1885 | int zflg = ZEXISTS; | |
e8b96c60 | 1886 | boolean_t waited = B_FALSE; |
34dc7c2f | 1887 | |
3558fd73 | 1888 | ZFS_ENTER(zsb); |
34dc7c2f | 1889 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1890 | zilog = zsb->z_log; |
34dc7c2f BB |
1891 | |
1892 | if (flags & FIGNORECASE) | |
1893 | zflg |= ZCILOOK; | |
1894 | top: | |
1895 | zp = NULL; | |
1896 | ||
1897 | /* | |
1898 | * Attempt to lock directory; fail if entry doesn't exist. | |
1899 | */ | |
149e873a BB |
1900 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1901 | NULL, NULL))) { | |
3558fd73 | 1902 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1903 | return (error); |
1904 | } | |
1905 | ||
3558fd73 | 1906 | ip = ZTOI(zp); |
34dc7c2f | 1907 | |
149e873a | 1908 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1909 | goto out; |
1910 | } | |
1911 | ||
3558fd73 | 1912 | if (!S_ISDIR(ip->i_mode)) { |
2e528b49 | 1913 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
1914 | goto out; |
1915 | } | |
1916 | ||
3558fd73 | 1917 | if (ip == cwd) { |
2e528b49 | 1918 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1919 | goto out; |
1920 | } | |
1921 | ||
34dc7c2f BB |
1922 | /* |
1923 | * Grab a lock on the directory to make sure that noone is | |
1924 | * trying to add (or lookup) entries while we are removing it. | |
1925 | */ | |
1926 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
1927 | ||
1928 | /* | |
1929 | * Grab a lock on the parent pointer to make sure we play well | |
1930 | * with the treewalk and directory rename code. | |
1931 | */ | |
1932 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
1933 | ||
3558fd73 | 1934 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 1935 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 1936 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3558fd73 | 1937 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
1938 | zfs_sa_upgrade_txholds(tx, zp); |
1939 | zfs_sa_upgrade_txholds(tx, dzp); | |
e8b96c60 | 1940 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1941 | if (error) { |
1942 | rw_exit(&zp->z_parent_lock); | |
1943 | rw_exit(&zp->z_name_lock); | |
1944 | zfs_dirent_unlock(dl); | |
3558fd73 | 1945 | iput(ip); |
fb5f0bc8 | 1946 | if (error == ERESTART) { |
e8b96c60 | 1947 | waited = B_TRUE; |
34dc7c2f BB |
1948 | dmu_tx_wait(tx); |
1949 | dmu_tx_abort(tx); | |
1950 | goto top; | |
1951 | } | |
1952 | dmu_tx_abort(tx); | |
3558fd73 | 1953 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1954 | return (error); |
1955 | } | |
1956 | ||
1957 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
1958 | ||
1959 | if (error == 0) { | |
1960 | uint64_t txtype = TX_RMDIR; | |
1961 | if (flags & FIGNORECASE) | |
1962 | txtype |= TX_CI; | |
572e2857 | 1963 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
1964 | } |
1965 | ||
1966 | dmu_tx_commit(tx); | |
1967 | ||
1968 | rw_exit(&zp->z_parent_lock); | |
1969 | rw_exit(&zp->z_name_lock); | |
1970 | out: | |
1971 | zfs_dirent_unlock(dl); | |
1972 | ||
59157910 BB |
1973 | zfs_inode_update(dzp); |
1974 | zfs_inode_update(zp); | |
3558fd73 | 1975 | iput(ip); |
34dc7c2f | 1976 | |
3558fd73 | 1977 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1978 | zil_commit(zilog, 0); |
428870ff | 1979 | |
3558fd73 | 1980 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1981 | return (error); |
1982 | } | |
e5c39b95 | 1983 | EXPORT_SYMBOL(zfs_rmdir); |
34dc7c2f BB |
1984 | |
1985 | /* | |
1986 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 1987 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 1988 | * |
3558fd73 BB |
1989 | * IN: ip - inode of directory to read. |
1990 | * dirent - buffer for directory entries. | |
34dc7c2f | 1991 | * |
3558fd73 | 1992 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
1993 | * |
1994 | * RETURN: 0 if success | |
1995 | * error code if failure | |
1996 | * | |
1997 | * Timestamps: | |
3558fd73 | 1998 | * ip - atime updated |
34dc7c2f BB |
1999 | * |
2000 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2001 | * This allows us to use the low range for "special" directory entries: | |
2002 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2003 | * we use the offset 2 for the '.zfs' directory. | |
2004 | */ | |
2005 | /* ARGSUSED */ | |
3558fd73 | 2006 | int |
0f37d0c8 | 2007 | zfs_readdir(struct inode *ip, struct dir_context *ctx, cred_t *cr) |
34dc7c2f | 2008 | { |
3558fd73 BB |
2009 | znode_t *zp = ITOZ(ip); |
2010 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 2011 | objset_t *os; |
34dc7c2f BB |
2012 | zap_cursor_t zc; |
2013 | zap_attribute_t zap; | |
34dc7c2f BB |
2014 | int error; |
2015 | uint8_t prefetch; | |
c12e3a59 | 2016 | uint8_t type; |
3558fd73 BB |
2017 | int done = 0; |
2018 | uint64_t parent; | |
c12e3a59 | 2019 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2020 | |
3558fd73 | 2021 | ZFS_ENTER(zsb); |
34dc7c2f BB |
2022 | ZFS_VERIFY_ZP(zp); |
2023 | ||
3558fd73 BB |
2024 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
2025 | &parent, sizeof (parent))) != 0) | |
2026 | goto out; | |
34dc7c2f BB |
2027 | |
2028 | /* | |
2029 | * Quit if directory has been removed (posix) | |
2030 | */ | |
3558fd73 BB |
2031 | if (zp->z_unlinked) |
2032 | goto out; | |
2033 | ||
c12e3a59 | 2034 | error = 0; |
3558fd73 | 2035 | os = zsb->z_os; |
c12e3a59 | 2036 | offset = ctx->pos; |
34dc7c2f BB |
2037 | prefetch = zp->z_zn_prefetch; |
2038 | ||
2039 | /* | |
2040 | * Initialize the iterator cursor. | |
2041 | */ | |
c12e3a59 | 2042 | if (offset <= 3) { |
34dc7c2f BB |
2043 | /* |
2044 | * Start iteration from the beginning of the directory. | |
2045 | */ | |
2046 | zap_cursor_init(&zc, os, zp->z_id); | |
2047 | } else { | |
2048 | /* | |
2049 | * The offset is a serialized cursor. | |
2050 | */ | |
c12e3a59 | 2051 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2052 | } |
2053 | ||
34dc7c2f BB |
2054 | /* |
2055 | * Transform to file-system independent format | |
2056 | */ | |
3558fd73 BB |
2057 | while (!done) { |
2058 | uint64_t objnum; | |
34dc7c2f BB |
2059 | /* |
2060 | * Special case `.', `..', and `.zfs'. | |
2061 | */ | |
c12e3a59 | 2062 | if (offset == 0) { |
34dc7c2f BB |
2063 | (void) strcpy(zap.za_name, "."); |
2064 | zap.za_normalization_conflict = 0; | |
2065 | objnum = zp->z_id; | |
c12e3a59 RY |
2066 | type = DT_DIR; |
2067 | } else if (offset == 1) { | |
34dc7c2f BB |
2068 | (void) strcpy(zap.za_name, ".."); |
2069 | zap.za_normalization_conflict = 0; | |
428870ff | 2070 | objnum = parent; |
c12e3a59 RY |
2071 | type = DT_DIR; |
2072 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2073 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2074 | zap.za_normalization_conflict = 0; | |
2075 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 2076 | type = DT_DIR; |
34dc7c2f BB |
2077 | } else { |
2078 | /* | |
2079 | * Grab next entry. | |
2080 | */ | |
3558fd73 BB |
2081 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2082 | if (error == ENOENT) | |
34dc7c2f BB |
2083 | break; |
2084 | else | |
2085 | goto update; | |
2086 | } | |
2087 | ||
0c5dde49 BB |
2088 | /* |
2089 | * Allow multiple entries provided the first entry is | |
2090 | * the object id. Non-zpl consumers may safely make | |
2091 | * use of the additional space. | |
2092 | * | |
2093 | * XXX: This should be a feature flag for compatibility | |
2094 | */ | |
34dc7c2f | 2095 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2096 | zap.za_num_integers == 0) { |
34dc7c2f | 2097 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2098 | "entry, obj = %lld, offset = %lld, " |
2099 | "length = %d, num = %lld\n", | |
34dc7c2f | 2100 | (u_longlong_t)zp->z_id, |
c12e3a59 | 2101 | (u_longlong_t)offset, |
0c5dde49 BB |
2102 | zap.za_integer_length, |
2103 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 2104 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2105 | goto update; |
2106 | } | |
2107 | ||
2108 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 2109 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2110 | } |
0f37d0c8 RY |
2111 | |
2112 | done = !dir_emit(ctx, zap.za_name, strlen(zap.za_name), | |
c12e3a59 | 2113 | objnum, type); |
0f37d0c8 | 2114 | if (done) |
34dc7c2f | 2115 | break; |
34dc7c2f BB |
2116 | |
2117 | /* Prefetch znode */ | |
3558fd73 | 2118 | if (prefetch) { |
34dc7c2f | 2119 | dmu_prefetch(os, objnum, 0, 0); |
3558fd73 | 2120 | } |
34dc7c2f | 2121 | |
c12e3a59 RY |
2122 | /* |
2123 | * Move to the next entry, fill in the previous offset. | |
2124 | */ | |
2125 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2126 | zap_cursor_advance(&zc); |
c12e3a59 | 2127 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2128 | } else { |
c12e3a59 | 2129 | offset += 1; |
34dc7c2f | 2130 | } |
c12e3a59 | 2131 | ctx->pos = offset; |
34dc7c2f BB |
2132 | } |
2133 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2134 | ||
34dc7c2f BB |
2135 | update: |
2136 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2137 | if (error == ENOENT) |
2138 | error = 0; | |
2139 | ||
3558fd73 | 2140 | ZFS_ACCESSTIME_STAMP(zsb, zp); |
3558fd73 BB |
2141 | |
2142 | out: | |
2143 | ZFS_EXIT(zsb); | |
34dc7c2f | 2144 | |
34dc7c2f BB |
2145 | return (error); |
2146 | } | |
3558fd73 | 2147 | EXPORT_SYMBOL(zfs_readdir); |
34dc7c2f | 2148 | |
d5446cfc BB |
2149 | ulong_t zfs_fsync_sync_cnt = 4; |
2150 | ||
e5c39b95 | 2151 | int |
3558fd73 | 2152 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2153 | { |
3558fd73 BB |
2154 | znode_t *zp = ITOZ(ip); |
2155 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 2156 | |
d5446cfc BB |
2157 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2158 | ||
3558fd73 BB |
2159 | if (zsb->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2160 | ZFS_ENTER(zsb); | |
428870ff | 2161 | ZFS_VERIFY_ZP(zp); |
3558fd73 BB |
2162 | zil_commit(zsb->z_log, zp->z_id); |
2163 | ZFS_EXIT(zsb); | |
428870ff | 2164 | } |
07012da6 CC |
2165 | tsd_set(zfs_fsyncer_key, NULL); |
2166 | ||
34dc7c2f BB |
2167 | return (0); |
2168 | } | |
e5c39b95 | 2169 | EXPORT_SYMBOL(zfs_fsync); |
34dc7c2f BB |
2170 | |
2171 | ||
2172 | /* | |
2173 | * Get the requested file attributes and place them in the provided | |
2174 | * vattr structure. | |
2175 | * | |
3558fd73 | 2176 | * IN: ip - inode of file. |
5484965a BB |
2177 | * vap - va_mask identifies requested attributes. |
2178 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2179 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2180 | * cr - credentials of caller. | |
34dc7c2f | 2181 | * |
5484965a BB |
2182 | * OUT: vap - attribute values. |
2183 | * | |
2184 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2185 | */ |
2186 | /* ARGSUSED */ | |
e5c39b95 | 2187 | int |
5484965a | 2188 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2189 | { |
3558fd73 BB |
2190 | znode_t *zp = ITOZ(ip); |
2191 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2192 | int error = 0; |
2193 | uint64_t links; | |
428870ff | 2194 | uint64_t mtime[2], ctime[2]; |
5484965a BB |
2195 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2196 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2197 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff BB |
2198 | sa_bulk_attr_t bulk[2]; |
2199 | int count = 0; | |
34dc7c2f | 2200 | |
3558fd73 | 2201 | ZFS_ENTER(zsb); |
34dc7c2f | 2202 | ZFS_VERIFY_ZP(zp); |
428870ff | 2203 | |
5484965a | 2204 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2205 | |
3558fd73 BB |
2206 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
2207 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
428870ff BB |
2208 | |
2209 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
3558fd73 | 2210 | ZFS_EXIT(zsb); |
428870ff BB |
2211 | return (error); |
2212 | } | |
34dc7c2f | 2213 | |
34dc7c2f BB |
2214 | /* |
2215 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2216 | * Also, if we are the owner don't bother, since owner should | |
2217 | * always be allowed to read basic attributes of file. | |
2218 | */ | |
572e2857 | 2219 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2220 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2221 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2222 | skipaclchk, cr))) { | |
3558fd73 | 2223 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2224 | return (error); |
2225 | } | |
2226 | } | |
2227 | ||
2228 | /* | |
2229 | * Return all attributes. It's cheaper to provide the answer | |
2230 | * than to determine whether we were asked the question. | |
2231 | */ | |
2232 | ||
9babb374 | 2233 | mutex_enter(&zp->z_lock); |
5484965a BB |
2234 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2235 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2236 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2237 | vap->va_nodeid = zp->z_id; |
3558fd73 | 2238 | if ((zp->z_id == zsb->z_root) && zfs_show_ctldir(zp)) |
428870ff | 2239 | links = zp->z_links + 1; |
34dc7c2f | 2240 | else |
428870ff | 2241 | links = zp->z_links; |
5484965a BB |
2242 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2243 | vap->va_size = i_size_read(ip); | |
2244 | vap->va_rdev = ip->i_rdev; | |
2245 | vap->va_seq = ip->i_generation; | |
2246 | ||
2247 | /* | |
2248 | * Add in any requested optional attributes and the create time. | |
2249 | * Also set the corresponding bits in the returned attribute bitmap. | |
2250 | */ | |
2251 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zsb->z_use_fuids) { | |
2252 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
2253 | xoap->xoa_archive = | |
2254 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2255 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2256 | } | |
2257 | ||
2258 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2259 | xoap->xoa_readonly = | |
2260 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2261 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2262 | } | |
2263 | ||
2264 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2265 | xoap->xoa_system = | |
2266 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2267 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2268 | } | |
2269 | ||
2270 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2271 | xoap->xoa_hidden = | |
2272 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2273 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2274 | } | |
2275 | ||
2276 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2277 | xoap->xoa_nounlink = | |
2278 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2279 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2280 | } | |
2281 | ||
2282 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2283 | xoap->xoa_immutable = | |
2284 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2285 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2286 | } | |
2287 | ||
2288 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2289 | xoap->xoa_appendonly = | |
2290 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2291 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2292 | } | |
2293 | ||
2294 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2295 | xoap->xoa_nodump = | |
2296 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2297 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2298 | } | |
2299 | ||
2300 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2301 | xoap->xoa_opaque = | |
2302 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2303 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2304 | } | |
2305 | ||
2306 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2307 | xoap->xoa_av_quarantined = | |
2308 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2309 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2310 | } | |
2311 | ||
2312 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2313 | xoap->xoa_av_modified = | |
2314 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2315 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2316 | } | |
2317 | ||
2318 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2319 | S_ISREG(ip->i_mode)) { | |
2320 | zfs_sa_get_scanstamp(zp, xvap); | |
2321 | } | |
34dc7c2f | 2322 | |
5484965a BB |
2323 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2324 | uint64_t times[2]; | |
2325 | ||
2326 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zsb), | |
2327 | times, sizeof (times)); | |
2328 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2329 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2330 | } | |
2331 | ||
2332 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2333 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2334 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2335 | } | |
2336 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
2337 | xoap->xoa_generation = zp->z_gen; | |
2338 | XVA_SET_RTN(xvap, XAT_GEN); | |
2339 | } | |
2340 | ||
2341 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2342 | xoap->xoa_offline = | |
2343 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2344 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2345 | } | |
2346 | ||
2347 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2348 | xoap->xoa_sparse = | |
2349 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2350 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2351 | } | |
2352 | } | |
2353 | ||
2354 | ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime); | |
2355 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); | |
2356 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2357 | |
2358 | mutex_exit(&zp->z_lock); | |
2359 | ||
5484965a | 2360 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2361 | |
2362 | if (zp->z_blksz == 0) { | |
2363 | /* | |
2364 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2365 | */ | |
5484965a | 2366 | vap->va_blksize = zsb->z_max_blksz; |
34dc7c2f BB |
2367 | } |
2368 | ||
3558fd73 | 2369 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2370 | return (0); |
2371 | } | |
e5c39b95 | 2372 | EXPORT_SYMBOL(zfs_getattr); |
34dc7c2f | 2373 | |
057e8eee BB |
2374 | /* |
2375 | * Get the basic file attributes and place them in the provided kstat | |
2376 | * structure. The inode is assumed to be the authoritative source | |
2377 | * for most of the attributes. However, the znode currently has the | |
2378 | * authoritative atime, blksize, and block count. | |
2379 | * | |
2380 | * IN: ip - inode of file. | |
2381 | * | |
2382 | * OUT: sp - kstat values. | |
2383 | * | |
2384 | * RETURN: 0 (always succeeds) | |
2385 | */ | |
2386 | /* ARGSUSED */ | |
2387 | int | |
2388 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2389 | { | |
2390 | znode_t *zp = ITOZ(ip); | |
2391 | zfs_sb_t *zsb = ITOZSB(ip); | |
b585bc4a BB |
2392 | uint32_t blksize; |
2393 | u_longlong_t nblocks; | |
057e8eee | 2394 | |
a7b125e9 GB |
2395 | ZFS_ENTER(zsb); |
2396 | ZFS_VERIFY_ZP(zp); | |
2397 | ||
057e8eee BB |
2398 | mutex_enter(&zp->z_lock); |
2399 | ||
2400 | generic_fillattr(ip, sp); | |
2401 | ZFS_TIME_DECODE(&sp->atime, zp->z_atime); | |
2402 | ||
b585bc4a BB |
2403 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2404 | sp->blksize = blksize; | |
2405 | sp->blocks = nblocks; | |
2406 | ||
057e8eee BB |
2407 | if (unlikely(zp->z_blksz == 0)) { |
2408 | /* | |
2409 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2410 | */ | |
2411 | sp->blksize = zsb->z_max_blksz; | |
2412 | } | |
2413 | ||
2414 | mutex_exit(&zp->z_lock); | |
2415 | ||
a7b125e9 GB |
2416 | ZFS_EXIT(zsb); |
2417 | ||
057e8eee BB |
2418 | return (0); |
2419 | } | |
2420 | EXPORT_SYMBOL(zfs_getattr_fast); | |
2421 | ||
34dc7c2f BB |
2422 | /* |
2423 | * Set the file attributes to the values contained in the | |
2424 | * vattr structure. | |
2425 | * | |
3558fd73 | 2426 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2427 | * vap - new attribute values. |
5484965a | 2428 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2429 | * flags - ATTR_UTIME set if non-default time values provided. |
2430 | * - ATTR_NOACLCHECK (CIFS context only). | |
2431 | * cr - credentials of caller. | |
34dc7c2f BB |
2432 | * |
2433 | * RETURN: 0 if success | |
2434 | * error code if failure | |
2435 | * | |
2436 | * Timestamps: | |
3558fd73 | 2437 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2438 | */ |
2439 | /* ARGSUSED */ | |
e5c39b95 | 2440 | int |
5484965a | 2441 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2442 | { |
3558fd73 BB |
2443 | znode_t *zp = ITOZ(ip); |
2444 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2445 | zilog_t *zilog; |
2446 | dmu_tx_t *tx; | |
2447 | vattr_t oldva; | |
f4ea75d4 | 2448 | xvattr_t *tmpxvattr; |
5484965a | 2449 | uint_t mask = vap->va_mask; |
a117a6d6 | 2450 | uint_t saved_mask = 0; |
34dc7c2f BB |
2451 | int trim_mask = 0; |
2452 | uint64_t new_mode; | |
9babb374 | 2453 | uint64_t new_uid, new_gid; |
572e2857 | 2454 | uint64_t xattr_obj; |
428870ff | 2455 | uint64_t mtime[2], ctime[2]; |
34dc7c2f BB |
2456 | znode_t *attrzp; |
2457 | int need_policy = FALSE; | |
428870ff | 2458 | int err, err2; |
34dc7c2f | 2459 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2460 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2461 | xoptattr_t *xoap; | |
2462 | zfs_acl_t *aclp; | |
34dc7c2f | 2463 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2464 | boolean_t fuid_dirtied = B_FALSE; |
17c37660 | 2465 | sa_bulk_attr_t *bulk, *xattr_bulk; |
428870ff | 2466 | int count = 0, xattr_count = 0; |
34dc7c2f BB |
2467 | |
2468 | if (mask == 0) | |
2469 | return (0); | |
2470 | ||
3558fd73 | 2471 | ZFS_ENTER(zsb); |
34dc7c2f BB |
2472 | ZFS_VERIFY_ZP(zp); |
2473 | ||
3558fd73 | 2474 | zilog = zsb->z_log; |
34dc7c2f BB |
2475 | |
2476 | /* | |
2477 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2478 | * that file system is at proper version level | |
2479 | */ | |
5484965a | 2480 | |
3558fd73 | 2481 | if (zsb->z_use_fuids == B_FALSE && |
5484965a BB |
2482 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2483 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2484 | (mask & ATTR_XVATTR))) { | |
3558fd73 | 2485 | ZFS_EXIT(zsb); |
2e528b49 | 2486 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2487 | } |
2488 | ||
3558fd73 BB |
2489 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
2490 | ZFS_EXIT(zsb); | |
2e528b49 | 2491 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2492 | } |
2493 | ||
3558fd73 BB |
2494 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
2495 | ZFS_EXIT(zsb); | |
2e528b49 | 2496 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2497 | } |
2498 | ||
5484965a BB |
2499 | /* |
2500 | * If this is an xvattr_t, then get a pointer to the structure of | |
2501 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2502 | */ | |
2503 | xoap = xva_getxoptattr(xvap); | |
2504 | ||
d1d7e268 | 2505 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2506 | xva_init(tmpxvattr); |
5484965a | 2507 | |
d1d7e268 MK |
2508 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); |
2509 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); | |
17c37660 | 2510 | |
5484965a BB |
2511 | /* |
2512 | * Immutable files can only alter immutable bit and atime | |
2513 | */ | |
2514 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2515 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2516 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
f4ea75d4 BB |
2517 | err = EPERM; |
2518 | goto out3; | |
5484965a BB |
2519 | } |
2520 | ||
3558fd73 | 2521 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
f4ea75d4 BB |
2522 | err = EPERM; |
2523 | goto out3; | |
34dc7c2f BB |
2524 | } |
2525 | ||
5484965a BB |
2526 | /* |
2527 | * Verify timestamps doesn't overflow 32 bits. | |
2528 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2529 | * handle times greater than 2039. This check should be removed | |
2530 | * once large timestamps are fully supported. | |
2531 | */ | |
2532 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
2533 | if (((mask & ATTR_ATIME) && |
2534 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2535 | ((mask & ATTR_MTIME) && | |
2536 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
f4ea75d4 BB |
2537 | err = EOVERFLOW; |
2538 | goto out3; | |
5484965a BB |
2539 | } |
2540 | } | |
2541 | ||
34dc7c2f BB |
2542 | top: |
2543 | attrzp = NULL; | |
572e2857 | 2544 | aclp = NULL; |
34dc7c2f | 2545 | |
45d1cae3 | 2546 | /* Can this be moved to before the top label? */ |
2cf7f52b | 2547 | if (zfs_is_readonly(zsb)) { |
f4ea75d4 BB |
2548 | err = EROFS; |
2549 | goto out3; | |
34dc7c2f BB |
2550 | } |
2551 | ||
2552 | /* | |
2553 | * First validate permissions | |
2554 | */ | |
2555 | ||
3558fd73 | 2556 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2557 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2558 | if (err) |
2559 | goto out3; | |
2560 | ||
34dc7c2f BB |
2561 | /* |
2562 | * XXX - Note, we are not providing any open | |
2563 | * mode flags here (like FNDELAY), so we may | |
2564 | * block if there are locks present... this | |
2565 | * should be addressed in openat(). | |
2566 | */ | |
b128c09f | 2567 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2568 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2569 | if (err) |
2570 | goto out3; | |
428870ff | 2571 | } |
34dc7c2f | 2572 | |
5484965a BB |
2573 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2574 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2575 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2576 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2577 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2578 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2579 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2580 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2581 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2582 | skipaclchk, cr); | |
2583 | } | |
2584 | ||
3558fd73 BB |
2585 | if (mask & (ATTR_UID|ATTR_GID)) { |
2586 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2587 | int take_owner; |
2588 | int take_group; | |
2589 | ||
2590 | /* | |
2591 | * NOTE: even if a new mode is being set, | |
2592 | * we may clear S_ISUID/S_ISGID bits. | |
2593 | */ | |
2594 | ||
3558fd73 | 2595 | if (!(mask & ATTR_MODE)) |
5484965a | 2596 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2597 | |
2598 | /* | |
2599 | * Take ownership or chgrp to group we are a member of | |
2600 | */ | |
2601 | ||
5484965a | 2602 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 2603 | take_group = (mask & ATTR_GID) && |
5484965a | 2604 | zfs_groupmember(zsb, vap->va_gid, cr); |
34dc7c2f BB |
2605 | |
2606 | /* | |
5484965a | 2607 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
2608 | * take_group must both be set in order to allow taking |
2609 | * ownership. | |
2610 | * | |
2611 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2612 | * | |
2613 | */ | |
2614 | ||
3558fd73 BB |
2615 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
2616 | take_owner && take_group) || | |
2617 | ((idmask == ATTR_UID) && take_owner) || | |
2618 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
2619 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
2620 | skipaclchk, cr) == 0) { | |
2621 | /* | |
2622 | * Remove setuid/setgid for non-privileged users | |
2623 | */ | |
5484965a | 2624 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 2625 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
2626 | } else { |
2627 | need_policy = TRUE; | |
2628 | } | |
2629 | } else { | |
2630 | need_policy = TRUE; | |
2631 | } | |
2632 | } | |
2633 | ||
2634 | mutex_enter(&zp->z_lock); | |
428870ff | 2635 | oldva.va_mode = zp->z_mode; |
572e2857 | 2636 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
2637 | if (mask & ATTR_XVATTR) { |
2638 | /* | |
2639 | * Update xvattr mask to include only those attributes | |
2640 | * that are actually changing. | |
2641 | * | |
2642 | * the bits will be restored prior to actually setting | |
2643 | * the attributes so the caller thinks they were set. | |
2644 | */ | |
2645 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2646 | if (xoap->xoa_appendonly != | |
2647 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
2648 | need_policy = TRUE; | |
2649 | } else { | |
2650 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 2651 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
2652 | } |
2653 | } | |
2654 | ||
2655 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2656 | if (xoap->xoa_nounlink != | |
2657 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
2658 | need_policy = TRUE; | |
2659 | } else { | |
2660 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 2661 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
2662 | } |
2663 | } | |
2664 | ||
2665 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2666 | if (xoap->xoa_immutable != | |
2667 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
2668 | need_policy = TRUE; | |
2669 | } else { | |
2670 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 2671 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
2672 | } |
2673 | } | |
2674 | ||
2675 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2676 | if (xoap->xoa_nodump != | |
2677 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
2678 | need_policy = TRUE; | |
2679 | } else { | |
2680 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 2681 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
2682 | } |
2683 | } | |
2684 | ||
2685 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2686 | if (xoap->xoa_av_modified != | |
2687 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
2688 | need_policy = TRUE; | |
2689 | } else { | |
2690 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 2691 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
2692 | } |
2693 | } | |
2694 | ||
2695 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2696 | if ((!S_ISREG(ip->i_mode) && | |
2697 | xoap->xoa_av_quarantined) || | |
2698 | xoap->xoa_av_quarantined != | |
2699 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
2700 | need_policy = TRUE; | |
2701 | } else { | |
2702 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 2703 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
2704 | } |
2705 | } | |
2706 | ||
2707 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2708 | mutex_exit(&zp->z_lock); | |
f4ea75d4 BB |
2709 | err = EPERM; |
2710 | goto out3; | |
5484965a BB |
2711 | } |
2712 | ||
2713 | if (need_policy == FALSE && | |
2714 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2715 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
2716 | need_policy = TRUE; | |
2717 | } | |
2718 | } | |
34dc7c2f BB |
2719 | |
2720 | mutex_exit(&zp->z_lock); | |
2721 | ||
3558fd73 | 2722 | if (mask & ATTR_MODE) { |
34dc7c2f | 2723 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 2724 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 2725 | &oldva, cr); |
f4ea75d4 BB |
2726 | if (err) |
2727 | goto out3; | |
2728 | ||
3558fd73 | 2729 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
2730 | } else { |
2731 | need_policy = TRUE; | |
2732 | } | |
2733 | } | |
2734 | ||
2735 | if (need_policy) { | |
2736 | /* | |
2737 | * If trim_mask is set then take ownership | |
2738 | * has been granted or write_acl is present and user | |
2739 | * has the ability to modify mode. In that case remove | |
2740 | * UID|GID and or MODE from mask so that | |
2741 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2742 | */ | |
2743 | ||
2744 | if (trim_mask) { | |
5484965a BB |
2745 | saved_mask = vap->va_mask; |
2746 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 2747 | } |
5484965a | 2748 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 2749 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
2750 | if (err) |
2751 | goto out3; | |
34dc7c2f BB |
2752 | |
2753 | if (trim_mask) | |
5484965a | 2754 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
2755 | } |
2756 | ||
2757 | /* | |
2758 | * secpolicy_vnode_setattr, or take ownership may have | |
2759 | * changed va_mask | |
2760 | */ | |
5484965a | 2761 | mask = vap->va_mask; |
34dc7c2f | 2762 | |
3558fd73 BB |
2763 | if ((mask & (ATTR_UID | ATTR_GID))) { |
2764 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), | |
572e2857 | 2765 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 2766 | |
572e2857 | 2767 | if (err == 0 && xattr_obj) { |
3558fd73 | 2768 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
2769 | if (err) |
2770 | goto out2; | |
2771 | } | |
3558fd73 BB |
2772 | if (mask & ATTR_UID) { |
2773 | new_uid = zfs_fuid_create(zsb, | |
5484965a | 2774 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
572e2857 | 2775 | if (new_uid != zp->z_uid && |
3558fd73 | 2776 | zfs_fuid_overquota(zsb, B_FALSE, new_uid)) { |
572e2857 | 2777 | if (attrzp) |
3558fd73 | 2778 | iput(ZTOI(attrzp)); |
428870ff BB |
2779 | err = EDQUOT; |
2780 | goto out2; | |
2781 | } | |
2782 | } | |
2783 | ||
3558fd73 | 2784 | if (mask & ATTR_GID) { |
5484965a | 2785 | new_gid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid, |
428870ff BB |
2786 | cr, ZFS_GROUP, &fuidp); |
2787 | if (new_gid != zp->z_gid && | |
3558fd73 | 2788 | zfs_fuid_overquota(zsb, B_TRUE, new_gid)) { |
572e2857 | 2789 | if (attrzp) |
3558fd73 | 2790 | iput(ZTOI(attrzp)); |
428870ff BB |
2791 | err = EDQUOT; |
2792 | goto out2; | |
2793 | } | |
2794 | } | |
2795 | } | |
3558fd73 | 2796 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 2797 | |
3558fd73 | 2798 | if (mask & ATTR_MODE) { |
428870ff | 2799 | uint64_t pmode = zp->z_mode; |
572e2857 | 2800 | uint64_t acl_obj; |
5484965a | 2801 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 2802 | |
572e2857 | 2803 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 2804 | |
572e2857 BB |
2805 | mutex_enter(&zp->z_lock); |
2806 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
2807 | /* |
2808 | * Are we upgrading ACL from old V0 format | |
2809 | * to V1 format? | |
2810 | */ | |
3558fd73 | 2811 | if (zsb->z_version >= ZPL_VERSION_FUID && |
572e2857 | 2812 | zfs_znode_acl_version(zp) == |
34dc7c2f | 2813 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 2814 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
2815 | DMU_OBJECT_END); |
2816 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
2817 | 0, aclp->z_acl_bytes); | |
2818 | } else { | |
572e2857 | 2819 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
2820 | aclp->z_acl_bytes); |
2821 | } | |
428870ff | 2822 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
2823 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
2824 | 0, aclp->z_acl_bytes); | |
2825 | } | |
572e2857 | 2826 | mutex_exit(&zp->z_lock); |
428870ff BB |
2827 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
2828 | } else { | |
5484965a BB |
2829 | if ((mask & ATTR_XVATTR) && |
2830 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
2831 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
2832 | else | |
2833 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2834 | } |
2835 | ||
428870ff BB |
2836 | if (attrzp) { |
2837 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2838 | } |
2839 | ||
3558fd73 | 2840 | fuid_dirtied = zsb->z_fuid_dirty; |
428870ff | 2841 | if (fuid_dirtied) |
3558fd73 | 2842 | zfs_fuid_txhold(zsb, tx); |
428870ff BB |
2843 | |
2844 | zfs_sa_upgrade_txholds(tx, zp); | |
2845 | ||
384f8a09 MA |
2846 | err = dmu_tx_assign(tx, TXG_WAIT); |
2847 | if (err) | |
9babb374 | 2848 | goto out; |
34dc7c2f | 2849 | |
428870ff | 2850 | count = 0; |
34dc7c2f BB |
2851 | /* |
2852 | * Set each attribute requested. | |
2853 | * We group settings according to the locks they need to acquire. | |
2854 | * | |
2855 | * Note: you cannot set ctime directly, although it will be | |
2856 | * updated as a side-effect of calling this function. | |
2857 | */ | |
2858 | ||
572e2857 | 2859 | |
3558fd73 | 2860 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2861 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
2862 | mutex_enter(&zp->z_lock); |
2863 | ||
3558fd73 | 2864 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff BB |
2865 | &zp->z_pflags, sizeof (zp->z_pflags)); |
2866 | ||
2867 | if (attrzp) { | |
3558fd73 | 2868 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2869 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
2870 | mutex_enter(&attrzp->z_lock); |
2871 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2872 | SA_ZPL_FLAGS(zsb), NULL, &attrzp->z_pflags, |
428870ff BB |
2873 | sizeof (attrzp->z_pflags)); |
2874 | } | |
2875 | ||
3558fd73 | 2876 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 2877 | |
3558fd73 BB |
2878 | if (mask & ATTR_UID) { |
2879 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, | |
428870ff | 2880 | &new_uid, sizeof (new_uid)); |
572e2857 | 2881 | zp->z_uid = new_uid; |
428870ff BB |
2882 | if (attrzp) { |
2883 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2884 | SA_ZPL_UID(zsb), NULL, &new_uid, |
428870ff | 2885 | sizeof (new_uid)); |
572e2857 | 2886 | attrzp->z_uid = new_uid; |
428870ff BB |
2887 | } |
2888 | } | |
2889 | ||
3558fd73 BB |
2890 | if (mask & ATTR_GID) { |
2891 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), | |
428870ff | 2892 | NULL, &new_gid, sizeof (new_gid)); |
572e2857 | 2893 | zp->z_gid = new_gid; |
428870ff BB |
2894 | if (attrzp) { |
2895 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2896 | SA_ZPL_GID(zsb), NULL, &new_gid, |
428870ff | 2897 | sizeof (new_gid)); |
572e2857 | 2898 | attrzp->z_gid = new_gid; |
428870ff BB |
2899 | } |
2900 | } | |
3558fd73 BB |
2901 | if (!(mask & ATTR_MODE)) { |
2902 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), | |
428870ff BB |
2903 | NULL, &new_mode, sizeof (new_mode)); |
2904 | new_mode = zp->z_mode; | |
2905 | } | |
2906 | err = zfs_acl_chown_setattr(zp); | |
2907 | ASSERT(err == 0); | |
2908 | if (attrzp) { | |
2909 | err = zfs_acl_chown_setattr(attrzp); | |
2910 | ASSERT(err == 0); | |
2911 | } | |
2912 | } | |
2913 | ||
3558fd73 BB |
2914 | if (mask & ATTR_MODE) { |
2915 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, | |
428870ff BB |
2916 | &new_mode, sizeof (new_mode)); |
2917 | zp->z_mode = new_mode; | |
99c564bc | 2918 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 2919 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 2920 | ASSERT0(err); |
572e2857 BB |
2921 | if (zp->z_acl_cached) |
2922 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
2923 | zp->z_acl_cached = aclp; |
2924 | aclp = NULL; | |
34dc7c2f BB |
2925 | } |
2926 | ||
34dc7c2f | 2927 | |
3558fd73 | 2928 | if (mask & ATTR_ATIME) { |
5484965a | 2929 | ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime); |
3558fd73 | 2930 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, |
428870ff | 2931 | &zp->z_atime, sizeof (zp->z_atime)); |
34dc7c2f BB |
2932 | } |
2933 | ||
3558fd73 | 2934 | if (mask & ATTR_MTIME) { |
5484965a | 2935 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
3558fd73 | 2936 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, |
428870ff | 2937 | mtime, sizeof (mtime)); |
34dc7c2f BB |
2938 | } |
2939 | ||
b128c09f | 2940 | /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */ |
3558fd73 BB |
2941 | if (mask & ATTR_SIZE && !(mask & ATTR_MTIME)) { |
2942 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), | |
428870ff | 2943 | NULL, mtime, sizeof (mtime)); |
3558fd73 | 2944 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
2945 | &ctime, sizeof (ctime)); |
2946 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, | |
2947 | B_TRUE); | |
2948 | } else if (mask != 0) { | |
3558fd73 | 2949 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
2950 | &ctime, sizeof (ctime)); |
2951 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, | |
2952 | B_TRUE); | |
2953 | if (attrzp) { | |
2954 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2955 | SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
2956 | &ctime, sizeof (ctime)); |
2957 | zfs_tstamp_update_setup(attrzp, STATE_CHANGED, | |
2958 | mtime, ctime, B_TRUE); | |
2959 | } | |
2960 | } | |
34dc7c2f BB |
2961 | /* |
2962 | * Do this after setting timestamps to prevent timestamp | |
2963 | * update from toggling bit | |
2964 | */ | |
2965 | ||
5484965a BB |
2966 | if (xoap && (mask & ATTR_XVATTR)) { |
2967 | ||
2968 | /* | |
2969 | * restore trimmed off masks | |
2970 | * so that return masks can be set for caller. | |
2971 | */ | |
2972 | ||
f4ea75d4 | 2973 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
2974 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
2975 | } | |
f4ea75d4 | 2976 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
2977 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
2978 | } | |
f4ea75d4 | 2979 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
2980 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
2981 | } | |
f4ea75d4 | 2982 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
2983 | XVA_SET_REQ(xvap, XAT_NODUMP); |
2984 | } | |
f4ea75d4 | 2985 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
2986 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
2987 | } | |
f4ea75d4 | 2988 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
2989 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
2990 | } | |
2991 | ||
2992 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
2993 | ASSERT(S_ISREG(ip->i_mode)); | |
2994 | ||
2995 | zfs_xvattr_set(zp, xvap, tx); | |
2996 | } | |
2997 | ||
9babb374 | 2998 | if (fuid_dirtied) |
3558fd73 | 2999 | zfs_fuid_sync(zsb, tx); |
9babb374 | 3000 | |
34dc7c2f | 3001 | if (mask != 0) |
5484965a | 3002 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3003 | |
34dc7c2f | 3004 | mutex_exit(&zp->z_lock); |
3558fd73 | 3005 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3006 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3007 | |
572e2857 | 3008 | if (attrzp) { |
3558fd73 | 3009 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3010 | mutex_exit(&attrzp->z_acl_lock); |
3011 | mutex_exit(&attrzp->z_lock); | |
3012 | } | |
9babb374 | 3013 | out: |
428870ff BB |
3014 | if (err == 0 && attrzp) { |
3015 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, | |
3016 | xattr_count, tx); | |
3017 | ASSERT(err2 == 0); | |
3018 | } | |
3019 | ||
34dc7c2f | 3020 | if (attrzp) |
3558fd73 | 3021 | iput(ZTOI(attrzp)); |
45d1cae3 | 3022 | if (aclp) |
9babb374 | 3023 | zfs_acl_free(aclp); |
9babb374 BB |
3024 | |
3025 | if (fuidp) { | |
3026 | zfs_fuid_info_free(fuidp); | |
3027 | fuidp = NULL; | |
3028 | } | |
3029 | ||
428870ff | 3030 | if (err) { |
9babb374 | 3031 | dmu_tx_abort(tx); |
428870ff BB |
3032 | if (err == ERESTART) |
3033 | goto top; | |
3034 | } else { | |
3035 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3036 | dmu_tx_commit(tx); |
037849f8 | 3037 | zfs_inode_update(zp); |
428870ff BB |
3038 | } |
3039 | ||
428870ff | 3040 | out2: |
3558fd73 | 3041 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3042 | zil_commit(zilog, 0); |
34dc7c2f | 3043 | |
f4ea75d4 | 3044 | out3: |
d1d7e268 MK |
3045 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * 7); |
3046 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * 7); | |
3047 | kmem_free(tmpxvattr, sizeof (xvattr_t)); | |
3558fd73 | 3048 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3049 | return (err); |
3050 | } | |
e5c39b95 | 3051 | EXPORT_SYMBOL(zfs_setattr); |
34dc7c2f BB |
3052 | |
3053 | typedef struct zfs_zlock { | |
3054 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3055 | znode_t *zl_znode; /* znode we held */ | |
3056 | struct zfs_zlock *zl_next; /* next in list */ | |
3057 | } zfs_zlock_t; | |
3058 | ||
3059 | /* | |
3060 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3061 | */ | |
3062 | static void | |
3063 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3064 | { | |
3065 | zfs_zlock_t *zl; | |
3066 | ||
3067 | while ((zl = *zlpp) != NULL) { | |
3068 | if (zl->zl_znode != NULL) | |
3558fd73 | 3069 | iput(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3070 | rw_exit(zl->zl_rwlock); |
3071 | *zlpp = zl->zl_next; | |
3072 | kmem_free(zl, sizeof (*zl)); | |
3073 | } | |
3074 | } | |
3075 | ||
3076 | /* | |
3077 | * Search back through the directory tree, using the ".." entries. | |
3078 | * Lock each directory in the chain to prevent concurrent renames. | |
3079 | * Fail any attempt to move a directory into one of its own descendants. | |
3080 | * XXX - z_parent_lock can overlap with map or grow locks | |
3081 | */ | |
3082 | static int | |
3083 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3084 | { | |
3085 | zfs_zlock_t *zl; | |
3086 | znode_t *zp = tdzp; | |
3558fd73 | 3087 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3088 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3089 | krwlock_t *rwlp = &szp->z_parent_lock; |
3090 | krw_t rw = RW_WRITER; | |
3091 | ||
3092 | /* | |
3093 | * First pass write-locks szp and compares to zp->z_id. | |
3094 | * Later passes read-lock zp and compare to zp->z_parent. | |
3095 | */ | |
3096 | do { | |
3097 | if (!rw_tryenter(rwlp, rw)) { | |
3098 | /* | |
3099 | * Another thread is renaming in this path. | |
3100 | * Note that if we are a WRITER, we don't have any | |
3101 | * parent_locks held yet. | |
3102 | */ | |
3103 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3104 | /* | |
3105 | * Drop our locks and restart | |
3106 | */ | |
3107 | zfs_rename_unlock(&zl); | |
3108 | *zlpp = NULL; | |
3109 | zp = tdzp; | |
428870ff | 3110 | oidp = zp->z_id; |
34dc7c2f BB |
3111 | rwlp = &szp->z_parent_lock; |
3112 | rw = RW_WRITER; | |
3113 | continue; | |
3114 | } else { | |
3115 | /* | |
3116 | * Wait for other thread to drop its locks | |
3117 | */ | |
3118 | rw_enter(rwlp, rw); | |
3119 | } | |
3120 | } | |
3121 | ||
3122 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3123 | zl->zl_rwlock = rwlp; | |
3124 | zl->zl_znode = NULL; | |
3125 | zl->zl_next = *zlpp; | |
3126 | *zlpp = zl; | |
3127 | ||
428870ff | 3128 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 3129 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3130 | |
428870ff | 3131 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3132 | return (0); |
3133 | ||
3134 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3135 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3136 | if (error) |
3137 | return (error); | |
3138 | zl->zl_znode = zp; | |
3139 | } | |
3558fd73 | 3140 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3141 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3142 | rwlp = &zp->z_parent_lock; |
3143 | rw = RW_READER; | |
3144 | ||
3145 | } while (zp->z_id != sdzp->z_id); | |
3146 | ||
3147 | return (0); | |
3148 | } | |
3149 | ||
3150 | /* | |
3151 | * Move an entry from the provided source directory to the target | |
3152 | * directory. Change the entry name as indicated. | |
3153 | * | |
3558fd73 | 3154 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3155 | * snm - Old entry name. |
3558fd73 | 3156 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3157 | * tnm - New entry name. |
3158 | * cr - credentials of caller. | |
34dc7c2f BB |
3159 | * flags - case flags |
3160 | * | |
d3cc8b15 | 3161 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3162 | * |
3163 | * Timestamps: | |
3558fd73 | 3164 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3165 | */ |
3166 | /*ARGSUSED*/ | |
e5c39b95 | 3167 | int |
3558fd73 BB |
3168 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3169 | cred_t *cr, int flags) | |
34dc7c2f BB |
3170 | { |
3171 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 BB |
3172 | znode_t *sdzp = ITOZ(sdip); |
3173 | zfs_sb_t *zsb = ITOZSB(sdip); | |
34dc7c2f | 3174 | zilog_t *zilog; |
34dc7c2f BB |
3175 | zfs_dirlock_t *sdl, *tdl; |
3176 | dmu_tx_t *tx; | |
3177 | zfs_zlock_t *zl; | |
3178 | int cmp, serr, terr; | |
3179 | int error = 0; | |
3180 | int zflg = 0; | |
e8b96c60 | 3181 | boolean_t waited = B_FALSE; |
34dc7c2f | 3182 | |
3558fd73 | 3183 | ZFS_ENTER(zsb); |
34dc7c2f | 3184 | ZFS_VERIFY_ZP(sdzp); |
3558fd73 | 3185 | zilog = zsb->z_log; |
34dc7c2f | 3186 | |
c0ebc844 | 3187 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
3558fd73 | 3188 | ZFS_EXIT(zsb); |
2e528b49 | 3189 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3190 | } |
3191 | ||
3558fd73 | 3192 | tdzp = ITOZ(tdip); |
34dc7c2f | 3193 | ZFS_VERIFY_ZP(tdzp); |
3558fd73 | 3194 | if (zsb->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3195 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3196 | ZFS_EXIT(zsb); |
2e528b49 | 3197 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3198 | } |
3199 | ||
3200 | if (flags & FIGNORECASE) | |
3201 | zflg |= ZCILOOK; | |
3202 | ||
3203 | top: | |
3204 | szp = NULL; | |
3205 | tzp = NULL; | |
3206 | zl = NULL; | |
3207 | ||
3208 | /* | |
3209 | * This is to prevent the creation of links into attribute space | |
3210 | * by renaming a linked file into/outof an attribute directory. | |
3211 | * See the comment in zfs_link() for why this is considered bad. | |
3212 | */ | |
428870ff | 3213 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 3214 | ZFS_EXIT(zsb); |
2e528b49 | 3215 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3216 | } |
3217 | ||
3218 | /* | |
3219 | * Lock source and target directory entries. To prevent deadlock, | |
3220 | * a lock ordering must be defined. We lock the directory with | |
3221 | * the smallest object id first, or if it's a tie, the one with | |
3222 | * the lexically first name. | |
3223 | */ | |
3224 | if (sdzp->z_id < tdzp->z_id) { | |
3225 | cmp = -1; | |
3226 | } else if (sdzp->z_id > tdzp->z_id) { | |
3227 | cmp = 1; | |
3228 | } else { | |
3229 | /* | |
3230 | * First compare the two name arguments without | |
3231 | * considering any case folding. | |
3232 | */ | |
3558fd73 | 3233 | int nofold = (zsb->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3234 | |
3235 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
3558fd73 | 3236 | ASSERT(error == 0 || !zsb->z_utf8); |
34dc7c2f BB |
3237 | if (cmp == 0) { |
3238 | /* | |
3239 | * POSIX: "If the old argument and the new argument | |
3240 | * both refer to links to the same existing file, | |
3241 | * the rename() function shall return successfully | |
3242 | * and perform no other action." | |
3243 | */ | |
3558fd73 | 3244 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3245 | return (0); |
3246 | } | |
3247 | /* | |
3248 | * If the file system is case-folding, then we may | |
3249 | * have some more checking to do. A case-folding file | |
3250 | * system is either supporting mixed case sensitivity | |
3251 | * access or is completely case-insensitive. Note | |
3252 | * that the file system is always case preserving. | |
3253 | * | |
3254 | * In mixed sensitivity mode case sensitive behavior | |
3255 | * is the default. FIGNORECASE must be used to | |
3256 | * explicitly request case insensitive behavior. | |
3257 | * | |
3258 | * If the source and target names provided differ only | |
3259 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3260 | * we will treat this as a special case in the | |
3261 | * case-insensitive mode: as long as the source name | |
3262 | * is an exact match, we will allow this to proceed as | |
3263 | * a name-change request. | |
3264 | */ | |
3558fd73 BB |
3265 | if ((zsb->z_case == ZFS_CASE_INSENSITIVE || |
3266 | (zsb->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3267 | flags & FIGNORECASE)) && |
3558fd73 | 3268 | u8_strcmp(snm, tnm, 0, zsb->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3269 | &error) == 0) { |
3270 | /* | |
3271 | * case preserving rename request, require exact | |
3272 | * name matches | |
3273 | */ | |
3274 | zflg |= ZCIEXACT; | |
3275 | zflg &= ~ZCILOOK; | |
3276 | } | |
3277 | } | |
3278 | ||
428870ff BB |
3279 | /* |
3280 | * If the source and destination directories are the same, we should | |
3281 | * grab the z_name_lock of that directory only once. | |
3282 | */ | |
3283 | if (sdzp == tdzp) { | |
3284 | zflg |= ZHAVELOCK; | |
3285 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3286 | } | |
3287 | ||
34dc7c2f BB |
3288 | if (cmp < 0) { |
3289 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3290 | ZEXISTS | zflg, NULL, NULL); | |
3291 | terr = zfs_dirent_lock(&tdl, | |
3292 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3293 | } else { | |
3294 | terr = zfs_dirent_lock(&tdl, | |
3295 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3296 | serr = zfs_dirent_lock(&sdl, | |
3297 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3298 | NULL, NULL); | |
3299 | } | |
3300 | ||
3301 | if (serr) { | |
3302 | /* | |
3303 | * Source entry invalid or not there. | |
3304 | */ | |
3305 | if (!terr) { | |
3306 | zfs_dirent_unlock(tdl); | |
3307 | if (tzp) | |
3558fd73 | 3308 | iput(ZTOI(tzp)); |
34dc7c2f | 3309 | } |
428870ff BB |
3310 | |
3311 | if (sdzp == tdzp) | |
3312 | rw_exit(&sdzp->z_name_lock); | |
3313 | ||
34dc7c2f BB |
3314 | if (strcmp(snm, "..") == 0) |
3315 | serr = EINVAL; | |
3558fd73 | 3316 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3317 | return (serr); |
3318 | } | |
3319 | if (terr) { | |
3320 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3321 | iput(ZTOI(szp)); |
428870ff BB |
3322 | |
3323 | if (sdzp == tdzp) | |
3324 | rw_exit(&sdzp->z_name_lock); | |
3325 | ||
34dc7c2f BB |
3326 | if (strcmp(tnm, "..") == 0) |
3327 | terr = EINVAL; | |
3558fd73 | 3328 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3329 | return (terr); |
3330 | } | |
3331 | ||
3332 | /* | |
3333 | * Must have write access at the source to remove the old entry | |
3334 | * and write access at the target to create the new entry. | |
3335 | * Note that if target and source are the same, this can be | |
3336 | * done in a single check. | |
3337 | */ | |
3338 | ||
149e873a | 3339 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3340 | goto out; |
3341 | ||
3558fd73 | 3342 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3343 | /* |
3344 | * Check to make sure rename is valid. | |
3345 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3346 | */ | |
149e873a | 3347 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3348 | goto out; |
3349 | } | |
3350 | ||
3351 | /* | |
3352 | * Does target exist? | |
3353 | */ | |
3354 | if (tzp) { | |
3355 | /* | |
3356 | * Source and target must be the same type. | |
3357 | */ | |
3558fd73 BB |
3358 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3359 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
2e528b49 | 3360 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3361 | goto out; |
3362 | } | |
3363 | } else { | |
3558fd73 | 3364 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
2e528b49 | 3365 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3366 | goto out; |
3367 | } | |
3368 | } | |
3369 | /* | |
3370 | * POSIX dictates that when the source and target | |
3371 | * entries refer to the same file object, rename | |
3372 | * must do nothing and exit without error. | |
3373 | */ | |
3374 | if (szp->z_id == tzp->z_id) { | |
3375 | error = 0; | |
3376 | goto out; | |
3377 | } | |
3378 | } | |
3379 | ||
3558fd73 | 3380 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
3381 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3382 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3383 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3384 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3385 | if (sdzp != tdzp) { |
3386 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3387 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3388 | } | |
3389 | if (tzp) { | |
3390 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3391 | zfs_sa_upgrade_txholds(tx, tzp); | |
3392 | } | |
3393 | ||
3394 | zfs_sa_upgrade_txholds(tx, szp); | |
3558fd73 | 3395 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
e8b96c60 | 3396 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3397 | if (error) { |
3398 | if (zl != NULL) | |
3399 | zfs_rename_unlock(&zl); | |
3400 | zfs_dirent_unlock(sdl); | |
3401 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3402 | |
3403 | if (sdzp == tdzp) | |
3404 | rw_exit(&sdzp->z_name_lock); | |
3405 | ||
3558fd73 | 3406 | iput(ZTOI(szp)); |
34dc7c2f | 3407 | if (tzp) |
3558fd73 | 3408 | iput(ZTOI(tzp)); |
fb5f0bc8 | 3409 | if (error == ERESTART) { |
e8b96c60 | 3410 | waited = B_TRUE; |
34dc7c2f BB |
3411 | dmu_tx_wait(tx); |
3412 | dmu_tx_abort(tx); | |
3413 | goto top; | |
3414 | } | |
3415 | dmu_tx_abort(tx); | |
3558fd73 | 3416 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3417 | return (error); |
3418 | } | |
3419 | ||
3420 | if (tzp) /* Attempt to remove the existing target */ | |
3421 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3422 | ||
3423 | if (error == 0) { | |
3424 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3425 | if (error == 0) { | |
428870ff | 3426 | szp->z_pflags |= ZFS_AV_MODIFIED; |
34dc7c2f | 3427 | |
3558fd73 | 3428 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zsb), |
428870ff | 3429 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c99c9001 | 3430 | ASSERT0(error); |
34dc7c2f | 3431 | |
428870ff BB |
3432 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3433 | if (error == 0) { | |
3434 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3435 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3436 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3437 | } else { |
3438 | /* | |
3439 | * At this point, we have successfully created | |
3440 | * the target name, but have failed to remove | |
3441 | * the source name. Since the create was done | |
3442 | * with the ZRENAMING flag, there are | |
3443 | * complications; for one, the link count is | |
3444 | * wrong. The easiest way to deal with this | |
3445 | * is to remove the newly created target, and | |
3446 | * return the original error. This must | |
3447 | * succeed; fortunately, it is very unlikely to | |
3448 | * fail, since we just created it. | |
3449 | */ | |
3450 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3451 | ZRENAMING, NULL), ==, 0); | |
3452 | } | |
34dc7c2f BB |
3453 | } |
3454 | } | |
3455 | ||
3456 | dmu_tx_commit(tx); | |
3457 | out: | |
3458 | if (zl != NULL) | |
3459 | zfs_rename_unlock(&zl); | |
3460 | ||
3461 | zfs_dirent_unlock(sdl); | |
3462 | zfs_dirent_unlock(tdl); | |
3463 | ||
960e08fe | 3464 | zfs_inode_update(sdzp); |
428870ff BB |
3465 | if (sdzp == tdzp) |
3466 | rw_exit(&sdzp->z_name_lock); | |
3467 | ||
960e08fe BB |
3468 | if (sdzp != tdzp) |
3469 | zfs_inode_update(tdzp); | |
428870ff | 3470 | |
960e08fe | 3471 | zfs_inode_update(szp); |
3558fd73 | 3472 | iput(ZTOI(szp)); |
960e08fe BB |
3473 | if (tzp) { |
3474 | zfs_inode_update(tzp); | |
3558fd73 | 3475 | iput(ZTOI(tzp)); |
960e08fe | 3476 | } |
34dc7c2f | 3477 | |
3558fd73 | 3478 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3479 | zil_commit(zilog, 0); |
428870ff | 3480 | |
3558fd73 | 3481 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3482 | return (error); |
3483 | } | |
e5c39b95 | 3484 | EXPORT_SYMBOL(zfs_rename); |
34dc7c2f BB |
3485 | |
3486 | /* | |
3487 | * Insert the indicated symbolic reference entry into the directory. | |
3488 | * | |
3558fd73 | 3489 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
3490 | * link - Name for new symlink entry. |
3491 | * vap - Attributes of new entry. | |
3492 | * target - Target path of new symlink. | |
3558fd73 | 3493 | * |
34dc7c2f | 3494 | * cr - credentials of caller. |
34dc7c2f BB |
3495 | * flags - case flags |
3496 | * | |
d3cc8b15 | 3497 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3498 | * |
3499 | * Timestamps: | |
3558fd73 | 3500 | * dip - ctime|mtime updated |
34dc7c2f BB |
3501 | */ |
3502 | /*ARGSUSED*/ | |
e5c39b95 | 3503 | int |
3558fd73 BB |
3504 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
3505 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 3506 | { |
3558fd73 | 3507 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
3508 | zfs_dirlock_t *dl; |
3509 | dmu_tx_t *tx; | |
3558fd73 | 3510 | zfs_sb_t *zsb = ITOZSB(dip); |
34dc7c2f | 3511 | zilog_t *zilog; |
428870ff | 3512 | uint64_t len = strlen(link); |
34dc7c2f BB |
3513 | int error; |
3514 | int zflg = ZNEW; | |
9babb374 BB |
3515 | zfs_acl_ids_t acl_ids; |
3516 | boolean_t fuid_dirtied; | |
428870ff | 3517 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 3518 | boolean_t waited = B_FALSE; |
34dc7c2f | 3519 | |
3558fd73 | 3520 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 3521 | |
3558fd73 | 3522 | ZFS_ENTER(zsb); |
34dc7c2f | 3523 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3524 | zilog = zsb->z_log; |
34dc7c2f | 3525 | |
3558fd73 | 3526 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 3527 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3528 | ZFS_EXIT(zsb); |
2e528b49 | 3529 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3530 | } |
3531 | if (flags & FIGNORECASE) | |
3532 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3533 | |
3534 | if (len > MAXPATHLEN) { | |
3558fd73 | 3535 | ZFS_EXIT(zsb); |
2e528b49 | 3536 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
3537 | } |
3538 | ||
428870ff BB |
3539 | if ((error = zfs_acl_ids_create(dzp, 0, |
3540 | vap, cr, NULL, &acl_ids)) != 0) { | |
3558fd73 | 3541 | ZFS_EXIT(zsb); |
428870ff BB |
3542 | return (error); |
3543 | } | |
3544 | top: | |
3558fd73 BB |
3545 | *ipp = NULL; |
3546 | ||
34dc7c2f BB |
3547 | /* |
3548 | * Attempt to lock directory; fail if entry already exists. | |
3549 | */ | |
3550 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3551 | if (error) { | |
428870ff | 3552 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 3553 | ZFS_EXIT(zsb); |
428870ff BB |
3554 | return (error); |
3555 | } | |
3556 | ||
149e873a | 3557 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
3558 | zfs_acl_ids_free(&acl_ids); |
3559 | zfs_dirent_unlock(dl); | |
3558fd73 | 3560 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3561 | return (error); |
3562 | } | |
3563 | ||
3558fd73 | 3564 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
9babb374 BB |
3565 | zfs_acl_ids_free(&acl_ids); |
3566 | zfs_dirent_unlock(dl); | |
3558fd73 | 3567 | ZFS_EXIT(zsb); |
2e528b49 | 3568 | return (SET_ERROR(EDQUOT)); |
9babb374 | 3569 | } |
3558fd73 BB |
3570 | tx = dmu_tx_create(zsb->z_os); |
3571 | fuid_dirtied = zsb->z_fuid_dirty; | |
34dc7c2f | 3572 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3573 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3574 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3575 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3576 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
3558fd73 | 3577 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
3578 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
3579 | acl_ids.z_aclp->z_acl_bytes); | |
3580 | } | |
9babb374 | 3581 | if (fuid_dirtied) |
3558fd73 | 3582 | zfs_fuid_txhold(zsb, tx); |
e8b96c60 | 3583 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3584 | if (error) { |
3585 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3586 | if (error == ERESTART) { |
e8b96c60 | 3587 | waited = B_TRUE; |
34dc7c2f BB |
3588 | dmu_tx_wait(tx); |
3589 | dmu_tx_abort(tx); | |
3590 | goto top; | |
3591 | } | |
428870ff | 3592 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 3593 | dmu_tx_abort(tx); |
3558fd73 | 3594 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3595 | return (error); |
3596 | } | |
3597 | ||
34dc7c2f BB |
3598 | /* |
3599 | * Create a new object for the symlink. | |
428870ff | 3600 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 3601 | */ |
428870ff | 3602 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3603 | |
428870ff | 3604 | if (fuid_dirtied) |
3558fd73 | 3605 | zfs_fuid_sync(zsb, tx); |
34dc7c2f | 3606 | |
572e2857 | 3607 | mutex_enter(&zp->z_lock); |
428870ff | 3608 | if (zp->z_is_sa) |
3558fd73 | 3609 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zsb), |
428870ff BB |
3610 | link, len, tx); |
3611 | else | |
3612 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3613 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3614 | |
428870ff | 3615 | zp->z_size = len; |
3558fd73 | 3616 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 3617 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
3618 | /* |
3619 | * Insert the new object into the directory. | |
3620 | */ | |
3621 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
428870ff BB |
3622 | |
3623 | if (flags & FIGNORECASE) | |
3624 | txtype |= TX_CI; | |
3625 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
9babb374 | 3626 | |
960e08fe BB |
3627 | zfs_inode_update(dzp); |
3628 | zfs_inode_update(zp); | |
3629 | ||
9babb374 | 3630 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3631 | |
3632 | dmu_tx_commit(tx); | |
3633 | ||
3634 | zfs_dirent_unlock(dl); | |
3635 | ||
3558fd73 | 3636 | *ipp = ZTOI(zp); |
34dc7c2f | 3637 | |
3558fd73 | 3638 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3639 | zil_commit(zilog, 0); |
428870ff | 3640 | |
3558fd73 | 3641 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3642 | return (error); |
3643 | } | |
e5c39b95 | 3644 | EXPORT_SYMBOL(zfs_symlink); |
34dc7c2f BB |
3645 | |
3646 | /* | |
3647 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 3648 | * the symbolic path referred to by ip. |
34dc7c2f | 3649 | * |
8b4f9a2d BB |
3650 | * IN: ip - inode of symbolic link |
3651 | * uio - structure to contain the link path. | |
3652 | * cr - credentials of caller. | |
34dc7c2f BB |
3653 | * |
3654 | * RETURN: 0 if success | |
3655 | * error code if failure | |
3656 | * | |
3657 | * Timestamps: | |
3558fd73 | 3658 | * ip - atime updated |
34dc7c2f BB |
3659 | */ |
3660 | /* ARGSUSED */ | |
e5c39b95 | 3661 | int |
8b4f9a2d | 3662 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 3663 | { |
3558fd73 BB |
3664 | znode_t *zp = ITOZ(ip); |
3665 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
3666 | int error; |
3667 | ||
3558fd73 | 3668 | ZFS_ENTER(zsb); |
34dc7c2f BB |
3669 | ZFS_VERIFY_ZP(zp); |
3670 | ||
572e2857 | 3671 | mutex_enter(&zp->z_lock); |
428870ff | 3672 | if (zp->z_is_sa) |
8b4f9a2d BB |
3673 | error = sa_lookup_uio(zp->z_sa_hdl, |
3674 | SA_ZPL_SYMLINK(zsb), uio); | |
428870ff | 3675 | else |
8b4f9a2d | 3676 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 3677 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3678 | |
3558fd73 | 3679 | ZFS_ACCESSTIME_STAMP(zsb, zp); |
3558fd73 | 3680 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3681 | return (error); |
3682 | } | |
8b4f9a2d | 3683 | EXPORT_SYMBOL(zfs_readlink); |
34dc7c2f BB |
3684 | |
3685 | /* | |
3558fd73 | 3686 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 3687 | * |
3558fd73 BB |
3688 | * IN: tdip - Directory to contain new entry. |
3689 | * sip - inode of new entry. | |
34dc7c2f BB |
3690 | * name - name of new entry. |
3691 | * cr - credentials of caller. | |
34dc7c2f BB |
3692 | * |
3693 | * RETURN: 0 if success | |
3694 | * error code if failure | |
3695 | * | |
3696 | * Timestamps: | |
3558fd73 BB |
3697 | * tdip - ctime|mtime updated |
3698 | * sip - ctime updated | |
34dc7c2f BB |
3699 | */ |
3700 | /* ARGSUSED */ | |
e5c39b95 | 3701 | int |
3558fd73 | 3702 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr) |
34dc7c2f | 3703 | { |
3558fd73 | 3704 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 3705 | znode_t *tzp, *szp; |
3558fd73 | 3706 | zfs_sb_t *zsb = ITOZSB(tdip); |
34dc7c2f BB |
3707 | zilog_t *zilog; |
3708 | zfs_dirlock_t *dl; | |
3709 | dmu_tx_t *tx; | |
34dc7c2f BB |
3710 | int error; |
3711 | int zf = ZNEW; | |
428870ff | 3712 | uint64_t parent; |
572e2857 | 3713 | uid_t owner; |
e8b96c60 | 3714 | boolean_t waited = B_FALSE; |
34dc7c2f | 3715 | |
3558fd73 | 3716 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 3717 | |
3558fd73 | 3718 | ZFS_ENTER(zsb); |
34dc7c2f | 3719 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3720 | zilog = zsb->z_log; |
34dc7c2f | 3721 | |
428870ff BB |
3722 | /* |
3723 | * POSIX dictates that we return EPERM here. | |
3724 | * Better choices include ENOTSUP or EISDIR. | |
3725 | */ | |
3558fd73 BB |
3726 | if (S_ISDIR(sip->i_mode)) { |
3727 | ZFS_EXIT(zsb); | |
2e528b49 | 3728 | return (SET_ERROR(EPERM)); |
428870ff BB |
3729 | } |
3730 | ||
c0ebc844 | 3731 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
3558fd73 | 3732 | ZFS_EXIT(zsb); |
2e528b49 | 3733 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 3734 | } |
428870ff | 3735 | |
3558fd73 | 3736 | szp = ITOZ(sip); |
34dc7c2f BB |
3737 | ZFS_VERIFY_ZP(szp); |
3738 | ||
428870ff BB |
3739 | /* Prevent links to .zfs/shares files */ |
3740 | ||
3558fd73 | 3741 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
428870ff | 3742 | &parent, sizeof (uint64_t))) != 0) { |
3558fd73 | 3743 | ZFS_EXIT(zsb); |
428870ff BB |
3744 | return (error); |
3745 | } | |
3558fd73 BB |
3746 | if (parent == zsb->z_shares_dir) { |
3747 | ZFS_EXIT(zsb); | |
2e528b49 | 3748 | return (SET_ERROR(EPERM)); |
428870ff BB |
3749 | } |
3750 | ||
3558fd73 | 3751 | if (zsb->z_utf8 && u8_validate(name, |
34dc7c2f | 3752 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3753 | ZFS_EXIT(zsb); |
2e528b49 | 3754 | return (SET_ERROR(EILSEQ)); |
34dc7c2f | 3755 | } |
3558fd73 | 3756 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
3757 | if (flags & FIGNORECASE) |
3758 | zf |= ZCILOOK; | |
3558fd73 | 3759 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f | 3760 | |
34dc7c2f BB |
3761 | /* |
3762 | * We do not support links between attributes and non-attributes | |
3763 | * because of the potential security risk of creating links | |
3764 | * into "normal" file space in order to circumvent restrictions | |
3765 | * imposed in attribute space. | |
3766 | */ | |
428870ff | 3767 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 3768 | ZFS_EXIT(zsb); |
2e528b49 | 3769 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3770 | } |
3771 | ||
3558fd73 | 3772 | owner = zfs_fuid_map_id(zsb, szp->z_uid, cr, ZFS_OWNER); |
572e2857 | 3773 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
3558fd73 | 3774 | ZFS_EXIT(zsb); |
2e528b49 | 3775 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
3776 | } |
3777 | ||
149e873a | 3778 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
3558fd73 | 3779 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3780 | return (error); |
3781 | } | |
3782 | ||
428870ff | 3783 | top: |
34dc7c2f BB |
3784 | /* |
3785 | * Attempt to lock directory; fail if entry already exists. | |
3786 | */ | |
3787 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
3788 | if (error) { | |
3558fd73 | 3789 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3790 | return (error); |
3791 | } | |
3792 | ||
3558fd73 | 3793 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 3794 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 3795 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3796 | zfs_sa_upgrade_txholds(tx, szp); |
3797 | zfs_sa_upgrade_txholds(tx, dzp); | |
e8b96c60 | 3798 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3799 | if (error) { |
3800 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3801 | if (error == ERESTART) { |
e8b96c60 | 3802 | waited = B_TRUE; |
34dc7c2f BB |
3803 | dmu_tx_wait(tx); |
3804 | dmu_tx_abort(tx); | |
3805 | goto top; | |
3806 | } | |
3807 | dmu_tx_abort(tx); | |
3558fd73 | 3808 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3809 | return (error); |
3810 | } | |
3811 | ||
3812 | error = zfs_link_create(dl, szp, tx, 0); | |
3813 | ||
3814 | if (error == 0) { | |
3815 | uint64_t txtype = TX_LINK; | |
3558fd73 | 3816 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
3817 | if (flags & FIGNORECASE) |
3818 | txtype |= TX_CI; | |
3558fd73 | 3819 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
3820 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); |
3821 | } | |
3822 | ||
3823 | dmu_tx_commit(tx); | |
3824 | ||
3825 | zfs_dirent_unlock(dl); | |
3826 | ||
3558fd73 | 3827 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3828 | zil_commit(zilog, 0); |
428870ff | 3829 | |
960e08fe BB |
3830 | zfs_inode_update(dzp); |
3831 | zfs_inode_update(szp); | |
3558fd73 | 3832 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3833 | return (error); |
3834 | } | |
e5c39b95 | 3835 | EXPORT_SYMBOL(zfs_link); |
34dc7c2f | 3836 | |
3c0e5c0f | 3837 | static void |
119a394a | 3838 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
3839 | { |
3840 | struct page *pp = arg; | |
3841 | ||
119a394a | 3842 | ClearPageError(pp); |
3c0e5c0f BB |
3843 | end_page_writeback(pp); |
3844 | } | |
3845 | ||
34dc7c2f | 3846 | /* |
3c0e5c0f BB |
3847 | * Push a page out to disk, once the page is on stable storage the |
3848 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 3849 | * |
3c0e5c0f BB |
3850 | * IN: ip - page mapped for inode. |
3851 | * pp - page to push (page is locked) | |
3852 | * wbc - writeback control data | |
34dc7c2f BB |
3853 | * |
3854 | * RETURN: 0 if success | |
3855 | * error code if failure | |
3856 | * | |
3c0e5c0f BB |
3857 | * Timestamps: |
3858 | * ip - ctime|mtime updated | |
34dc7c2f BB |
3859 | */ |
3860 | /* ARGSUSED */ | |
3c0e5c0f BB |
3861 | int |
3862 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 3863 | { |
3c0e5c0f BB |
3864 | znode_t *zp = ITOZ(ip); |
3865 | zfs_sb_t *zsb = ITOZSB(ip); | |
3866 | loff_t offset; | |
3867 | loff_t pgoff; | |
4c837f0d BB |
3868 | unsigned int pglen; |
3869 | rl_t *rl; | |
3c0e5c0f BB |
3870 | dmu_tx_t *tx; |
3871 | caddr_t va; | |
3872 | int err = 0; | |
3873 | uint64_t mtime[2], ctime[2]; | |
3874 | sa_bulk_attr_t bulk[3]; | |
3875 | int cnt = 0; | |
3876 | ||
4c837f0d BB |
3877 | ZFS_ENTER(zsb); |
3878 | ZFS_VERIFY_ZP(zp); | |
d164b209 | 3879 | |
3c0e5c0f BB |
3880 | ASSERT(PageLocked(pp)); |
3881 | ||
d1d7e268 MK |
3882 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
3883 | offset = i_size_read(ip); /* File length in bytes */ | |
3884 | pglen = MIN(PAGE_CACHE_SIZE, /* Page length in bytes */ | |
3c0e5c0f BB |
3885 | P2ROUNDUP(offset, PAGE_CACHE_SIZE)-pgoff); |
3886 | ||
3887 | /* Page is beyond end of file */ | |
3888 | if (pgoff >= offset) { | |
3889 | unlock_page(pp); | |
4c837f0d | 3890 | ZFS_EXIT(zsb); |
3c0e5c0f BB |
3891 | return (0); |
3892 | } | |
3893 | ||
3894 | /* Truncate page length to end of file */ | |
3895 | if (pgoff + pglen > offset) | |
3896 | pglen = offset - pgoff; | |
3897 | ||
3898 | #if 0 | |
34dc7c2f | 3899 | /* |
3c0e5c0f BB |
3900 | * FIXME: Allow mmap writes past its quota. The correct fix |
3901 | * is to register a page_mkwrite() handler to count the page | |
3902 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 3903 | */ |
dde471ef PJ |
3904 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
3905 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 | 3906 | err = EDQUOT; |
9babb374 | 3907 | } |
3c0e5c0f BB |
3908 | #endif |
3909 | ||
d958324f BB |
3910 | /* |
3911 | * The ordering here is critical and must adhere to the following | |
3912 | * rules in order to avoid deadlocking in either zfs_read() or | |
3913 | * zfs_free_range() due to a lock inversion. | |
3914 | * | |
3915 | * 1) The page must be unlocked prior to acquiring the range lock. | |
3916 | * This is critical because zfs_read() calls find_lock_page() | |
3917 | * which may block on the page lock while holding the range lock. | |
3918 | * | |
3919 | * 2) Before setting or clearing write back on a page the range lock | |
3920 | * must be held in order to prevent a lock inversion with the | |
3921 | * zfs_free_range() function. | |
3922 | */ | |
3923 | unlock_page(pp); | |
74328ee1 | 3924 | rl = zfs_range_lock(zp, pgoff, pglen, RL_WRITER); |
3c0e5c0f | 3925 | set_page_writeback(pp); |
3c0e5c0f | 3926 | |
dde471ef | 3927 | tx = dmu_tx_create(zsb->z_os); |
3c0e5c0f | 3928 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
3929 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3930 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 3931 | |
fb5f0bc8 | 3932 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 3933 | if (err != 0) { |
3c0e5c0f | 3934 | if (err == ERESTART) |
34dc7c2f | 3935 | dmu_tx_wait(tx); |
3c0e5c0f | 3936 | |
34dc7c2f | 3937 | dmu_tx_abort(tx); |
119a394a ED |
3938 | __set_page_dirty_nobuffers(pp); |
3939 | ClearPageError(pp); | |
3940 | end_page_writeback(pp); | |
4c837f0d BB |
3941 | zfs_range_unlock(rl); |
3942 | ZFS_EXIT(zsb); | |
3c0e5c0f | 3943 | return (err); |
34dc7c2f BB |
3944 | } |
3945 | ||
dde471ef | 3946 | va = kmap(pp); |
3c0e5c0f BB |
3947 | ASSERT3U(pglen, <=, PAGE_CACHE_SIZE); |
3948 | dmu_write(zsb->z_os, zp->z_id, pgoff, pglen, va, tx); | |
dde471ef | 3949 | kunmap(pp); |
34dc7c2f | 3950 | |
3c0e5c0f BB |
3951 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
3952 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
3953 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, 8); | |
428870ff | 3954 | |
d3aa3ea9 BB |
3955 | /* Preserve the mtime and ctime provided by the inode */ |
3956 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
3957 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
3958 | zp->z_atime_dirty = 0; | |
3959 | zp->z_seq++; | |
3960 | ||
3961 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
3962 | ||
119a394a ED |
3963 | zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
3964 | zfs_putpage_commit_cb, pp); | |
45d1cae3 | 3965 | dmu_tx_commit(tx); |
d3aa3ea9 | 3966 | |
4c837f0d | 3967 | zfs_range_unlock(rl); |
34dc7c2f | 3968 | |
119a394a ED |
3969 | if (wbc->sync_mode != WB_SYNC_NONE) { |
3970 | /* | |
3971 | * Note that this is rarely called under writepages(), because | |
3972 | * writepages() normally handles the entire commit for | |
3973 | * performance reasons. | |
3974 | */ | |
1e8db771 BB |
3975 | if (zsb->z_log != NULL) |
3976 | zil_commit(zsb->z_log, zp->z_id); | |
2b286136 | 3977 | } |
3c0e5c0f | 3978 | |
4c837f0d | 3979 | ZFS_EXIT(zsb); |
3c0e5c0f | 3980 | return (err); |
34dc7c2f BB |
3981 | } |
3982 | ||
8780c539 BB |
3983 | /* |
3984 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 3985 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
3986 | */ |
3987 | int | |
3988 | zfs_dirty_inode(struct inode *ip, int flags) | |
3989 | { | |
3990 | znode_t *zp = ITOZ(ip); | |
3991 | zfs_sb_t *zsb = ITOZSB(ip); | |
3992 | dmu_tx_t *tx; | |
023699cd MM |
3993 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
3994 | sa_bulk_attr_t bulk[4]; | |
8780c539 BB |
3995 | int error; |
3996 | int cnt = 0; | |
3997 | ||
c944be5d BB |
3998 | if (zfs_is_readonly(zsb) || dmu_objset_is_snapshot(zsb->z_os)) |
3999 | return (0); | |
4000 | ||
8780c539 BB |
4001 | ZFS_ENTER(zsb); |
4002 | ZFS_VERIFY_ZP(zp); | |
4003 | ||
4004 | tx = dmu_tx_create(zsb->z_os); | |
4005 | ||
4006 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4007 | zfs_sa_upgrade_txholds(tx, zp); | |
4008 | ||
4009 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4010 | if (error) { | |
4011 | dmu_tx_abort(tx); | |
4012 | goto out; | |
4013 | } | |
4014 | ||
4015 | mutex_enter(&zp->z_lock); | |
023699cd | 4016 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zsb), NULL, &mode, 8); |
8780c539 BB |
4017 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zsb), NULL, &atime, 16); |
4018 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); | |
4019 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
4020 | ||
023699cd | 4021 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4022 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4023 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4024 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
023699cd MM |
4025 | mode = ip->i_mode; |
4026 | ||
4027 | zp->z_mode = mode; | |
8780c539 BB |
4028 | zp->z_atime_dirty = 0; |
4029 | ||
4030 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4031 | mutex_exit(&zp->z_lock); | |
4032 | ||
4033 | dmu_tx_commit(tx); | |
4034 | out: | |
4035 | ZFS_EXIT(zsb); | |
4036 | return (error); | |
4037 | } | |
4038 | EXPORT_SYMBOL(zfs_dirty_inode); | |
4039 | ||
34dc7c2f BB |
4040 | /*ARGSUSED*/ |
4041 | void | |
c0d35759 | 4042 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4043 | { |
c0d35759 BB |
4044 | znode_t *zp = ITOZ(ip); |
4045 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4046 | int error; |
4047 | ||
ebe7e575 BB |
4048 | if (zfsctl_is_node(ip)) { |
4049 | zfsctl_inode_inactive(ip); | |
4050 | return; | |
4051 | } | |
34dc7c2f | 4052 | |
c0d35759 BB |
4053 | rw_enter(&zsb->z_teardown_inactive_lock, RW_READER); |
4054 | if (zp->z_sa_hdl == NULL) { | |
4055 | rw_exit(&zsb->z_teardown_inactive_lock); | |
4056 | return; | |
34dc7c2f BB |
4057 | } |
4058 | ||
4059 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
c0d35759 | 4060 | dmu_tx_t *tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 4061 | |
428870ff BB |
4062 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4063 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4064 | error = dmu_tx_assign(tx, TXG_WAIT); |
4065 | if (error) { | |
4066 | dmu_tx_abort(tx); | |
4067 | } else { | |
34dc7c2f | 4068 | mutex_enter(&zp->z_lock); |
3558fd73 | 4069 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), |
428870ff | 4070 | (void *)&zp->z_atime, sizeof (zp->z_atime), tx); |
34dc7c2f BB |
4071 | zp->z_atime_dirty = 0; |
4072 | mutex_exit(&zp->z_lock); | |
4073 | dmu_tx_commit(tx); | |
4074 | } | |
4075 | } | |
4076 | ||
4077 | zfs_zinactive(zp); | |
3558fd73 | 4078 | rw_exit(&zsb->z_teardown_inactive_lock); |
34dc7c2f | 4079 | } |
e5c39b95 | 4080 | EXPORT_SYMBOL(zfs_inactive); |
34dc7c2f BB |
4081 | |
4082 | /* | |
4083 | * Bounds-check the seek operation. | |
4084 | * | |
3558fd73 | 4085 | * IN: ip - inode seeking within |
34dc7c2f BB |
4086 | * ooff - old file offset |
4087 | * noffp - pointer to new file offset | |
4088 | * ct - caller context | |
4089 | * | |
4090 | * RETURN: 0 if success | |
4091 | * EINVAL if new offset invalid | |
4092 | */ | |
4093 | /* ARGSUSED */ | |
3558fd73 | 4094 | int |
9623f736 | 4095 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4096 | { |
3558fd73 | 4097 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4098 | return (0); |
4099 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4100 | } | |
3558fd73 | 4101 | EXPORT_SYMBOL(zfs_seek); |
34dc7c2f | 4102 | |
34dc7c2f | 4103 | /* |
dde471ef | 4104 | * Fill pages with data from the disk. |
34dc7c2f BB |
4105 | */ |
4106 | static int | |
dde471ef | 4107 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4108 | { |
d1d7e268 MK |
4109 | znode_t *zp = ITOZ(ip); |
4110 | zfs_sb_t *zsb = ITOZSB(ip); | |
4111 | objset_t *os; | |
dde471ef | 4112 | struct page *cur_pp; |
d1d7e268 MK |
4113 | u_offset_t io_off, total; |
4114 | size_t io_len; | |
4115 | loff_t i_size; | |
4116 | unsigned page_idx; | |
4117 | int err; | |
34dc7c2f | 4118 | |
d1d7e268 | 4119 | os = zsb->z_os; |
dde471ef PJ |
4120 | io_len = nr_pages << PAGE_CACHE_SHIFT; |
4121 | i_size = i_size_read(ip); | |
4122 | io_off = page_offset(pl[0]); | |
4123 | ||
4124 | if (io_off + io_len > i_size) | |
4125 | io_len = i_size - io_off; | |
34dc7c2f BB |
4126 | |
4127 | /* | |
dde471ef | 4128 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4129 | */ |
dde471ef PJ |
4130 | page_idx = 0; |
4131 | cur_pp = pl[0]; | |
34dc7c2f | 4132 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4133 | caddr_t va; |
4134 | ||
dde471ef | 4135 | va = kmap(cur_pp); |
9babb374 BB |
4136 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4137 | DMU_READ_PREFETCH); | |
dde471ef | 4138 | kunmap(cur_pp); |
34dc7c2f | 4139 | if (err) { |
b128c09f BB |
4140 | /* convert checksum errors into IO errors */ |
4141 | if (err == ECKSUM) | |
2e528b49 | 4142 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4143 | return (err); |
4144 | } | |
dde471ef | 4145 | cur_pp = pl[++page_idx]; |
34dc7c2f | 4146 | } |
d164b209 | 4147 | |
34dc7c2f BB |
4148 | return (0); |
4149 | } | |
4150 | ||
4151 | /* | |
dde471ef | 4152 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4153 | * |
dde471ef PJ |
4154 | * IN: ip - inode of file to get data from. |
4155 | * pl - list of pages to read | |
4156 | * nr_pages - number of pages to read | |
34dc7c2f | 4157 | * |
d3cc8b15 | 4158 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4159 | * |
4160 | * Timestamps: | |
4161 | * vp - atime updated | |
4162 | */ | |
4163 | /* ARGSUSED */ | |
dde471ef PJ |
4164 | int |
4165 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4166 | { |
dde471ef PJ |
4167 | znode_t *zp = ITOZ(ip); |
4168 | zfs_sb_t *zsb = ITOZSB(ip); | |
4169 | int err; | |
d164b209 | 4170 | |
d164b209 BB |
4171 | if (pl == NULL) |
4172 | return (0); | |
34dc7c2f | 4173 | |
dde471ef | 4174 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4175 | ZFS_VERIFY_ZP(zp); |
4176 | ||
dde471ef | 4177 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4178 | |
dde471ef PJ |
4179 | if (!err) |
4180 | ZFS_ACCESSTIME_STAMP(zsb, zp); | |
34dc7c2f | 4181 | |
dde471ef | 4182 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4183 | return (err); |
4184 | } | |
dde471ef | 4185 | EXPORT_SYMBOL(zfs_getpage); |
34dc7c2f BB |
4186 | |
4187 | /* | |
e2e7aa2d | 4188 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4189 | * |
e2e7aa2d BB |
4190 | * IN: ip - inode of the file to mmap |
4191 | * off - file offset | |
4192 | * addrp - start address in memory region | |
4193 | * len - length of memory region | |
4194 | * vm_flags- address flags | |
34dc7c2f | 4195 | * |
e2e7aa2d BB |
4196 | * RETURN: 0 if success |
4197 | * error code if failure | |
34dc7c2f BB |
4198 | */ |
4199 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4200 | int |
4201 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4202 | unsigned long vm_flags) | |
34dc7c2f | 4203 | { |
e2e7aa2d BB |
4204 | znode_t *zp = ITOZ(ip); |
4205 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4206 | |
e2e7aa2d | 4207 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4208 | ZFS_VERIFY_ZP(zp); |
4209 | ||
e2e7aa2d | 4210 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4211 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
e2e7aa2d | 4212 | ZFS_EXIT(zsb); |
2e528b49 | 4213 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4214 | } |
4215 | ||
e2e7aa2d | 4216 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4217 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
e2e7aa2d | 4218 | ZFS_EXIT(zsb); |
2e528b49 | 4219 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4220 | } |
4221 | ||
34dc7c2f | 4222 | if (off < 0 || len > MAXOFFSET_T - off) { |
e2e7aa2d | 4223 | ZFS_EXIT(zsb); |
2e528b49 | 4224 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4225 | } |
4226 | ||
e2e7aa2d | 4227 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4228 | return (0); |
4229 | } | |
e2e7aa2d | 4230 | EXPORT_SYMBOL(zfs_map); |
34dc7c2f | 4231 | |
3558fd73 BB |
4232 | /* |
4233 | * convoff - converts the given data (start, whence) to the | |
4234 | * given whence. | |
4235 | */ | |
4236 | int | |
4237 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4238 | { | |
5484965a | 4239 | vattr_t vap; |
3558fd73 BB |
4240 | int error; |
4241 | ||
4242 | if ((lckdat->l_whence == 2) || (whence == 2)) { | |
5484965a | 4243 | if ((error = zfs_getattr(ip, &vap, 0, CRED()) != 0)) |
3558fd73 BB |
4244 | return (error); |
4245 | } | |
4246 | ||
4247 | switch (lckdat->l_whence) { | |
4248 | case 1: | |
4249 | lckdat->l_start += offset; | |
4250 | break; | |
4251 | case 2: | |
5484965a | 4252 | lckdat->l_start += vap.va_size; |
3558fd73 BB |
4253 | /* FALLTHRU */ |
4254 | case 0: | |
4255 | break; | |
4256 | default: | |
2e528b49 | 4257 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4258 | } |
4259 | ||
4260 | if (lckdat->l_start < 0) | |
2e528b49 | 4261 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4262 | |
4263 | switch (whence) { | |
4264 | case 1: | |
4265 | lckdat->l_start -= offset; | |
4266 | break; | |
4267 | case 2: | |
5484965a | 4268 | lckdat->l_start -= vap.va_size; |
3558fd73 BB |
4269 | /* FALLTHRU */ |
4270 | case 0: | |
4271 | break; | |
4272 | default: | |
2e528b49 | 4273 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4274 | } |
4275 | ||
4276 | lckdat->l_whence = (short)whence; | |
4277 | return (0); | |
4278 | } | |
4279 | ||
34dc7c2f BB |
4280 | /* |
4281 | * Free or allocate space in a file. Currently, this function only | |
4282 | * supports the `F_FREESP' command. However, this command is somewhat | |
4283 | * misnamed, as its functionality includes the ability to allocate as | |
4284 | * well as free space. | |
4285 | * | |
3558fd73 | 4286 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4287 | * cmd - action to take (only F_FREESP supported). |
4288 | * bfp - section of file to free/alloc. | |
4289 | * flag - current file open mode flags. | |
4290 | * offset - current file offset. | |
4291 | * cr - credentials of caller [UNUSED]. | |
34dc7c2f | 4292 | * |
d3cc8b15 | 4293 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4294 | * |
4295 | * Timestamps: | |
3558fd73 | 4296 | * ip - ctime|mtime updated |
34dc7c2f BB |
4297 | */ |
4298 | /* ARGSUSED */ | |
e5c39b95 | 4299 | int |
3558fd73 BB |
4300 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4301 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4302 | { |
3558fd73 BB |
4303 | znode_t *zp = ITOZ(ip); |
4304 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4305 | uint64_t off, len; |
4306 | int error; | |
4307 | ||
3558fd73 | 4308 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4309 | ZFS_VERIFY_ZP(zp); |
4310 | ||
34dc7c2f | 4311 | if (cmd != F_FREESP) { |
3558fd73 | 4312 | ZFS_EXIT(zsb); |
2e528b49 | 4313 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4314 | } |
4315 | ||
3558fd73 BB |
4316 | if ((error = convoff(ip, bfp, 0, offset))) { |
4317 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
4318 | return (error); |
4319 | } | |
4320 | ||
4321 | if (bfp->l_len < 0) { | |
3558fd73 | 4322 | ZFS_EXIT(zsb); |
2e528b49 | 4323 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4324 | } |
4325 | ||
aec69371 ED |
4326 | /* |
4327 | * Permissions aren't checked on Solaris because on this OS | |
4328 | * zfs_space() can only be called with an opened file handle. | |
4329 | * On Linux we can get here through truncate_range() which | |
4330 | * operates directly on inodes, so we need to check access rights. | |
4331 | */ | |
4332 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
4333 | ZFS_EXIT(zsb); | |
4334 | return (error); | |
4335 | } | |
4336 | ||
34dc7c2f BB |
4337 | off = bfp->l_start; |
4338 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4339 | ||
b128c09f | 4340 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4341 | |
3558fd73 | 4342 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4343 | return (error); |
4344 | } | |
e5c39b95 | 4345 | EXPORT_SYMBOL(zfs_space); |
34dc7c2f BB |
4346 | |
4347 | /*ARGSUSED*/ | |
e5c39b95 | 4348 | int |
3558fd73 | 4349 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4350 | { |
3558fd73 BB |
4351 | znode_t *zp = ITOZ(ip); |
4352 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4353 | uint32_t gen; |
428870ff | 4354 | uint64_t gen64; |
34dc7c2f BB |
4355 | uint64_t object = zp->z_id; |
4356 | zfid_short_t *zfid; | |
428870ff | 4357 | int size, i, error; |
34dc7c2f | 4358 | |
3558fd73 | 4359 | ZFS_ENTER(zsb); |
34dc7c2f | 4360 | ZFS_VERIFY_ZP(zp); |
428870ff | 4361 | |
3558fd73 | 4362 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), |
428870ff | 4363 | &gen64, sizeof (uint64_t))) != 0) { |
3558fd73 | 4364 | ZFS_EXIT(zsb); |
428870ff BB |
4365 | return (error); |
4366 | } | |
4367 | ||
4368 | gen = (uint32_t)gen64; | |
34dc7c2f | 4369 | |
3558fd73 | 4370 | size = (zsb->z_parent != zsb) ? LONG_FID_LEN : SHORT_FID_LEN; |
34dc7c2f BB |
4371 | if (fidp->fid_len < size) { |
4372 | fidp->fid_len = size; | |
3558fd73 | 4373 | ZFS_EXIT(zsb); |
2e528b49 | 4374 | return (SET_ERROR(ENOSPC)); |
34dc7c2f BB |
4375 | } |
4376 | ||
4377 | zfid = (zfid_short_t *)fidp; | |
4378 | ||
4379 | zfid->zf_len = size; | |
4380 | ||
4381 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4382 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4383 | ||
4384 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4385 | if (gen == 0) | |
4386 | gen = 1; | |
4387 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4388 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4389 | ||
4390 | if (size == LONG_FID_LEN) { | |
3558fd73 | 4391 | uint64_t objsetid = dmu_objset_id(zsb->z_os); |
34dc7c2f BB |
4392 | zfid_long_t *zlfid; |
4393 | ||
4394 | zlfid = (zfid_long_t *)fidp; | |
4395 | ||
4396 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
4397 | zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); | |
4398 | ||
4399 | /* XXX - this should be the generation number for the objset */ | |
4400 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
4401 | zlfid->zf_setgen[i] = 0; | |
4402 | } | |
4403 | ||
3558fd73 | 4404 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4405 | return (0); |
4406 | } | |
e5c39b95 | 4407 | EXPORT_SYMBOL(zfs_fid); |
34dc7c2f | 4408 | |
34dc7c2f | 4409 | /*ARGSUSED*/ |
e5c39b95 | 4410 | int |
3558fd73 | 4411 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4412 | { |
3558fd73 BB |
4413 | znode_t *zp = ITOZ(ip); |
4414 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4415 | int error; |
4416 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
4417 | ||
3558fd73 | 4418 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4419 | ZFS_VERIFY_ZP(zp); |
4420 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
3558fd73 | 4421 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4422 | |
4423 | return (error); | |
4424 | } | |
e5c39b95 | 4425 | EXPORT_SYMBOL(zfs_getsecattr); |
34dc7c2f BB |
4426 | |
4427 | /*ARGSUSED*/ | |
e5c39b95 | 4428 | int |
3558fd73 | 4429 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4430 | { |
3558fd73 BB |
4431 | znode_t *zp = ITOZ(ip); |
4432 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4433 | int error; |
4434 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
3558fd73 | 4435 | zilog_t *zilog = zsb->z_log; |
34dc7c2f | 4436 | |
3558fd73 | 4437 | ZFS_ENTER(zsb); |
34dc7c2f | 4438 | ZFS_VERIFY_ZP(zp); |
428870ff | 4439 | |
34dc7c2f | 4440 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 4441 | |
3558fd73 | 4442 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4443 | zil_commit(zilog, 0); |
428870ff | 4444 | |
3558fd73 | 4445 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4446 | return (error); |
4447 | } | |
e5c39b95 | 4448 | EXPORT_SYMBOL(zfs_setsecattr); |
34dc7c2f | 4449 | |
3558fd73 | 4450 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
4451 | /* |
4452 | * Tunable, both must be a power of 2. | |
4453 | * | |
4454 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
4455 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 4456 | * an arcbuf for a partial block read |
428870ff BB |
4457 | */ |
4458 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
4459 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
4460 | ||
4461 | /*ARGSUSED*/ | |
4462 | static int | |
3558fd73 | 4463 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 4464 | { |
3558fd73 BB |
4465 | znode_t *zp = ITOZ(ip); |
4466 | zfs_sb_t *zsb = ITOZSB(ip); | |
4467 | int max_blksz = zsb->z_max_blksz; | |
428870ff BB |
4468 | uio_t *uio = &xuio->xu_uio; |
4469 | ssize_t size = uio->uio_resid; | |
4470 | offset_t offset = uio->uio_loffset; | |
4471 | int blksz; | |
4472 | int fullblk, i; | |
4473 | arc_buf_t *abuf; | |
4474 | ssize_t maxsize; | |
4475 | int preamble, postamble; | |
4476 | ||
4477 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
2e528b49 | 4478 | return (SET_ERROR(EINVAL)); |
428870ff | 4479 | |
3558fd73 | 4480 | ZFS_ENTER(zsb); |
428870ff BB |
4481 | ZFS_VERIFY_ZP(zp); |
4482 | switch (ioflag) { | |
4483 | case UIO_WRITE: | |
4484 | /* | |
4485 | * Loan out an arc_buf for write if write size is bigger than | |
4486 | * max_blksz, and the file's block size is also max_blksz. | |
4487 | */ | |
4488 | blksz = max_blksz; | |
4489 | if (size < blksz || zp->z_blksz != blksz) { | |
3558fd73 | 4490 | ZFS_EXIT(zsb); |
2e528b49 | 4491 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4492 | } |
4493 | /* | |
4494 | * Caller requests buffers for write before knowing where the | |
4495 | * write offset might be (e.g. NFS TCP write). | |
4496 | */ | |
4497 | if (offset == -1) { | |
4498 | preamble = 0; | |
4499 | } else { | |
4500 | preamble = P2PHASE(offset, blksz); | |
4501 | if (preamble) { | |
4502 | preamble = blksz - preamble; | |
4503 | size -= preamble; | |
4504 | } | |
4505 | } | |
4506 | ||
4507 | postamble = P2PHASE(size, blksz); | |
4508 | size -= postamble; | |
4509 | ||
4510 | fullblk = size / blksz; | |
4511 | (void) dmu_xuio_init(xuio, | |
4512 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
4513 | |
4514 | /* | |
4515 | * Have to fix iov base/len for partial buffers. They | |
4516 | * currently represent full arc_buf's. | |
4517 | */ | |
4518 | if (preamble) { | |
4519 | /* data begins in the middle of the arc_buf */ | |
4520 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4521 | blksz); | |
4522 | ASSERT(abuf); | |
4523 | (void) dmu_xuio_add(xuio, abuf, | |
4524 | blksz - preamble, preamble); | |
4525 | } | |
4526 | ||
4527 | for (i = 0; i < fullblk; i++) { | |
4528 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4529 | blksz); | |
4530 | ASSERT(abuf); | |
4531 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
4532 | } | |
4533 | ||
4534 | if (postamble) { | |
4535 | /* data ends in the middle of the arc_buf */ | |
4536 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4537 | blksz); | |
4538 | ASSERT(abuf); | |
4539 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
4540 | } | |
4541 | break; | |
4542 | case UIO_READ: | |
4543 | /* | |
4544 | * Loan out an arc_buf for read if the read size is larger than | |
4545 | * the current file block size. Block alignment is not | |
4546 | * considered. Partial arc_buf will be loaned out for read. | |
4547 | */ | |
4548 | blksz = zp->z_blksz; | |
4549 | if (blksz < zcr_blksz_min) | |
4550 | blksz = zcr_blksz_min; | |
4551 | if (blksz > zcr_blksz_max) | |
4552 | blksz = zcr_blksz_max; | |
4553 | /* avoid potential complexity of dealing with it */ | |
4554 | if (blksz > max_blksz) { | |
3558fd73 | 4555 | ZFS_EXIT(zsb); |
2e528b49 | 4556 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4557 | } |
4558 | ||
4559 | maxsize = zp->z_size - uio->uio_loffset; | |
4560 | if (size > maxsize) | |
4561 | size = maxsize; | |
4562 | ||
3558fd73 BB |
4563 | if (size < blksz) { |
4564 | ZFS_EXIT(zsb); | |
2e528b49 | 4565 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4566 | } |
4567 | break; | |
4568 | default: | |
3558fd73 | 4569 | ZFS_EXIT(zsb); |
2e528b49 | 4570 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4571 | } |
4572 | ||
4573 | uio->uio_extflg = UIO_XUIO; | |
4574 | XUIO_XUZC_RW(xuio) = ioflag; | |
3558fd73 | 4575 | ZFS_EXIT(zsb); |
428870ff BB |
4576 | return (0); |
4577 | } | |
4578 | ||
4579 | /*ARGSUSED*/ | |
4580 | static int | |
3558fd73 | 4581 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
4582 | { |
4583 | int i; | |
4584 | arc_buf_t *abuf; | |
4585 | int ioflag = XUIO_XUZC_RW(xuio); | |
4586 | ||
4587 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
4588 | ||
4589 | i = dmu_xuio_cnt(xuio); | |
4590 | while (i-- > 0) { | |
4591 | abuf = dmu_xuio_arcbuf(xuio, i); | |
4592 | /* | |
4593 | * if abuf == NULL, it must be a write buffer | |
4594 | * that has been returned in zfs_write(). | |
4595 | */ | |
4596 | if (abuf) | |
4597 | dmu_return_arcbuf(abuf); | |
4598 | ASSERT(abuf || ioflag == UIO_WRITE); | |
4599 | } | |
4600 | ||
4601 | dmu_xuio_fini(xuio); | |
4602 | return (0); | |
4603 | } | |
3558fd73 | 4604 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 BB |
4605 | |
4606 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
4607 | module_param(zfs_read_chunk_size, long, 0644); | |
4608 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
4609 | #endif |