]>
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. |
19d55079 | 23 | * Copyright (c) 2012, 2015 by Delphix. All rights reserved. |
f3c9dca0 | 24 | * Copyright 2014 Nexenta Systems, Inc. All rights reserved. |
5475aada | 25 | * Copyright (c) 2015 by Chunwei Chen. All rights reserved. |
34dc7c2f BB |
26 | */ |
27 | ||
28 | /* Portions Copyright 2007 Jeremy Teo */ | |
428870ff | 29 | /* Portions Copyright 2010 Robert Milkowski */ |
34dc7c2f | 30 | |
60101509 | 31 | |
34dc7c2f BB |
32 | #include <sys/types.h> |
33 | #include <sys/param.h> | |
34 | #include <sys/time.h> | |
35 | #include <sys/systm.h> | |
36 | #include <sys/sysmacros.h> | |
37 | #include <sys/resource.h> | |
38 | #include <sys/vfs.h> | |
39 | #include <sys/vfs_opreg.h> | |
34dc7c2f BB |
40 | #include <sys/file.h> |
41 | #include <sys/stat.h> | |
42 | #include <sys/kmem.h> | |
43 | #include <sys/taskq.h> | |
44 | #include <sys/uio.h> | |
45 | #include <sys/vmsystm.h> | |
46 | #include <sys/atomic.h> | |
34dc7c2f | 47 | #include <vm/pvn.h> |
34dc7c2f BB |
48 | #include <sys/pathname.h> |
49 | #include <sys/cmn_err.h> | |
50 | #include <sys/errno.h> | |
51 | #include <sys/unistd.h> | |
52 | #include <sys/zfs_dir.h> | |
53 | #include <sys/zfs_acl.h> | |
54 | #include <sys/zfs_ioctl.h> | |
55 | #include <sys/fs/zfs.h> | |
56 | #include <sys/dmu.h> | |
428870ff | 57 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
58 | #include <sys/spa.h> |
59 | #include <sys/txg.h> | |
60 | #include <sys/dbuf.h> | |
61 | #include <sys/zap.h> | |
428870ff | 62 | #include <sys/sa.h> |
34dc7c2f BB |
63 | #include <sys/dirent.h> |
64 | #include <sys/policy.h> | |
65 | #include <sys/sunddi.h> | |
b128c09f | 66 | #include <sys/sid.h> |
bcf30822 | 67 | #include <sys/mode.h> |
34dc7c2f | 68 | #include "fs/fs_subr.h" |
ebe7e575 | 69 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 70 | #include <sys/zfs_fuid.h> |
428870ff | 71 | #include <sys/zfs_sa.h> |
e5c39b95 | 72 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
73 | #include <sys/dnlc.h> |
74 | #include <sys/zfs_rlock.h> | |
75 | #include <sys/extdirent.h> | |
76 | #include <sys/kidmap.h> | |
428870ff | 77 | #include <sys/cred.h> |
34dc7c2f | 78 | #include <sys/attr.h> |
218b8eaf | 79 | #include <sys/zpl.h> |
34dc7c2f BB |
80 | |
81 | /* | |
82 | * Programming rules. | |
83 | * | |
84 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
85 | * properly lock its in-core state, create a DMU transaction, do the work, | |
86 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
87 | * and wait for the intent log to commit if it is a synchronous operation. | |
88 | * Moreover, the vnode ops must work in both normal and log replay context. | |
89 | * The ordering of events is important to avoid deadlocks and references | |
90 | * to freed memory. The example below illustrates the following Big Rules: | |
91 | * | |
92 | * (1) A check must be made in each zfs thread for a mounted file system. | |
3558fd73 BB |
93 | * This is done avoiding races using ZFS_ENTER(zsb). |
94 | * A ZFS_EXIT(zsb) is needed before all returns. Any znodes | |
34dc7c2f BB |
95 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros |
96 | * can return EIO from the calling function. | |
97 | * | |
3558fd73 | 98 | * (2) iput() should always be the last thing except for zil_commit() |
34dc7c2f BB |
99 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: |
100 | * First, if it's the last reference, the vnode/znode | |
101 | * can be freed, so the zp may point to freed memory. Second, the last | |
102 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
103 | * pushing cached pages (which acquires range locks) and syncing out | |
104 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
105 | * which could deadlock the system if you were already holding one. | |
0a50679c | 106 | * If you must call iput() within a tx then use zfs_iput_async(). |
34dc7c2f BB |
107 | * |
108 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
109 | * as they can span dmu_tx_assign() calls. | |
110 | * | |
384f8a09 MA |
111 | * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to |
112 | * dmu_tx_assign(). This is critical because we don't want to block | |
113 | * while holding locks. | |
114 | * | |
115 | * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This | |
116 | * reduces lock contention and CPU usage when we must wait (note that if | |
117 | * throughput is constrained by the storage, nearly every transaction | |
118 | * must wait). | |
119 | * | |
120 | * Note, in particular, that if a lock is sometimes acquired before | |
121 | * the tx assigns, and sometimes after (e.g. z_lock), then failing | |
122 | * to use a non-blocking assign can deadlock the system. The scenario: | |
34dc7c2f BB |
123 | * |
124 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
125 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
126 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
127 | * forever, because the previous txg can't quiesce until B's tx commits. | |
128 | * | |
3558fd73 | 129 | * If dmu_tx_assign() returns ERESTART and zsb->z_assign is TXG_NOWAIT, |
e8b96c60 MA |
130 | * then drop all locks, call dmu_tx_wait(), and try again. On subsequent |
131 | * calls to dmu_tx_assign(), pass TXG_WAITED rather than TXG_NOWAIT, | |
132 | * to indicate that this operation has already called dmu_tx_wait(). | |
133 | * This will ensure that we don't retry forever, waiting a short bit | |
134 | * each time. | |
34dc7c2f BB |
135 | * |
136 | * (5) If the operation succeeded, generate the intent log entry for it | |
137 | * before dropping locks. This ensures that the ordering of events | |
138 | * in the intent log matches the order in which they actually occurred. | |
d3cc8b15 | 139 | * During ZIL replay the zfs_log_* functions will update the sequence |
fb5f0bc8 | 140 | * number to indicate the zil transaction has replayed. |
34dc7c2f BB |
141 | * |
142 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
143 | * regardless of whether there were any errors. | |
144 | * | |
572e2857 | 145 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
146 | * to ensure that synchronous semantics are provided when necessary. |
147 | * | |
148 | * In general, this is how things should be ordered in each vnode op: | |
149 | * | |
3558fd73 | 150 | * ZFS_ENTER(zsb); // exit if unmounted |
34dc7c2f | 151 | * top: |
3558fd73 | 152 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
153 | * rw_enter(...); // grab any other locks you need |
154 | * tx = dmu_tx_create(...); // get DMU tx | |
155 | * dmu_tx_hold_*(); // hold each object you might modify | |
e8b96c60 | 156 | * error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
157 | * if (error) { |
158 | * rw_exit(...); // drop locks | |
159 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 160 | * iput(...); // release held vnodes |
fb5f0bc8 | 161 | * if (error == ERESTART) { |
e8b96c60 | 162 | * waited = B_TRUE; |
34dc7c2f BB |
163 | * dmu_tx_wait(tx); |
164 | * dmu_tx_abort(tx); | |
165 | * goto top; | |
166 | * } | |
167 | * dmu_tx_abort(tx); // abort DMU tx | |
3558fd73 | 168 | * ZFS_EXIT(zsb); // finished in zfs |
34dc7c2f BB |
169 | * return (error); // really out of space |
170 | * } | |
171 | * error = do_real_work(); // do whatever this VOP does | |
172 | * if (error == 0) | |
173 | * zfs_log_*(...); // on success, make ZIL entry | |
174 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
175 | * rw_exit(...); // drop locks | |
176 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 177 | * iput(...); // release held vnodes |
572e2857 | 178 | * zil_commit(zilog, foid); // synchronous when necessary |
3558fd73 | 179 | * ZFS_EXIT(zsb); // finished in zfs |
34dc7c2f BB |
180 | * return (error); // done, report error |
181 | */ | |
182 | ||
126400a1 BB |
183 | /* |
184 | * Virus scanning is unsupported. It would be possible to add a hook | |
185 | * here to performance the required virus scan. This could be done | |
186 | * entirely in the kernel or potentially as an update to invoke a | |
187 | * scanning utility. | |
188 | */ | |
189 | static int | |
190 | zfs_vscan(struct inode *ip, cred_t *cr, int async) | |
191 | { | |
192 | return (0); | |
193 | } | |
194 | ||
195 | /* ARGSUSED */ | |
196 | int | |
197 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
198 | { | |
199 | znode_t *zp = ITOZ(ip); | |
200 | zfs_sb_t *zsb = ITOZSB(ip); | |
201 | ||
202 | ZFS_ENTER(zsb); | |
203 | ZFS_VERIFY_ZP(zp); | |
204 | ||
205 | /* Honor ZFS_APPENDONLY file attribute */ | |
206 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
207 | ((flag & O_APPEND) == 0)) { | |
208 | ZFS_EXIT(zsb); | |
2e528b49 | 209 | return (SET_ERROR(EPERM)); |
126400a1 BB |
210 | } |
211 | ||
212 | /* Virus scan eligible files on open */ | |
213 | if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && | |
214 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { | |
215 | if (zfs_vscan(ip, cr, 0) != 0) { | |
216 | ZFS_EXIT(zsb); | |
2e528b49 | 217 | return (SET_ERROR(EACCES)); |
126400a1 BB |
218 | } |
219 | } | |
220 | ||
221 | /* Keep a count of the synchronous opens in the znode */ | |
222 | if (flag & O_SYNC) | |
223 | atomic_inc_32(&zp->z_sync_cnt); | |
224 | ||
225 | ZFS_EXIT(zsb); | |
226 | return (0); | |
227 | } | |
228 | EXPORT_SYMBOL(zfs_open); | |
229 | ||
230 | /* ARGSUSED */ | |
231 | int | |
232 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
233 | { | |
234 | znode_t *zp = ITOZ(ip); | |
235 | zfs_sb_t *zsb = ITOZSB(ip); | |
236 | ||
237 | ZFS_ENTER(zsb); | |
238 | ZFS_VERIFY_ZP(zp); | |
239 | ||
7dc71949 | 240 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 241 | if (flag & O_SYNC) |
7dc71949 | 242 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 BB |
243 | |
244 | if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && | |
245 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) | |
246 | VERIFY(zfs_vscan(ip, cr, 1) == 0); | |
247 | ||
248 | ZFS_EXIT(zsb); | |
8780c539 | 249 | return (0); |
126400a1 BB |
250 | } |
251 | EXPORT_SYMBOL(zfs_close); | |
252 | ||
802e7b5f | 253 | #if defined(SEEK_HOLE) && defined(SEEK_DATA) |
cf91b2b6 | 254 | /* |
802e7b5f LD |
255 | * Lseek support for finding holes (cmd == SEEK_HOLE) and |
256 | * data (cmd == SEEK_DATA). "off" is an in/out parameter. | |
cf91b2b6 MA |
257 | */ |
258 | static int | |
802e7b5f | 259 | zfs_holey_common(struct inode *ip, int cmd, loff_t *off) |
cf91b2b6 | 260 | { |
802e7b5f | 261 | znode_t *zp = ITOZ(ip); |
cf91b2b6 MA |
262 | uint64_t noff = (uint64_t)*off; /* new offset */ |
263 | uint64_t file_sz; | |
264 | int error; | |
265 | boolean_t hole; | |
266 | ||
267 | file_sz = zp->z_size; | |
268 | if (noff >= file_sz) { | |
2e528b49 | 269 | return (SET_ERROR(ENXIO)); |
cf91b2b6 MA |
270 | } |
271 | ||
802e7b5f | 272 | if (cmd == SEEK_HOLE) |
cf91b2b6 MA |
273 | hole = B_TRUE; |
274 | else | |
275 | hole = B_FALSE; | |
276 | ||
802e7b5f | 277 | error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff); |
cf91b2b6 | 278 | |
d97aa48f | 279 | if (error == ESRCH) |
2e528b49 | 280 | return (SET_ERROR(ENXIO)); |
d97aa48f MA |
281 | |
282 | /* | |
283 | * We could find a hole that begins after the logical end-of-file, | |
284 | * because dmu_offset_next() only works on whole blocks. If the | |
285 | * EOF falls mid-block, then indicate that the "virtual hole" | |
286 | * at the end of the file begins at the logical EOF, rather than | |
287 | * at the end of the last block. | |
288 | */ | |
289 | if (noff > file_sz) { | |
290 | ASSERT(hole); | |
291 | noff = file_sz; | |
cf91b2b6 MA |
292 | } |
293 | ||
294 | if (noff < *off) | |
295 | return (error); | |
296 | *off = noff; | |
297 | return (error); | |
298 | } | |
802e7b5f LD |
299 | |
300 | int | |
301 | zfs_holey(struct inode *ip, int cmd, loff_t *off) | |
302 | { | |
303 | znode_t *zp = ITOZ(ip); | |
304 | zfs_sb_t *zsb = ITOZSB(ip); | |
305 | int error; | |
306 | ||
307 | ZFS_ENTER(zsb); | |
308 | ZFS_VERIFY_ZP(zp); | |
309 | ||
310 | error = zfs_holey_common(ip, cmd, off); | |
311 | ||
312 | ZFS_EXIT(zsb); | |
313 | return (error); | |
314 | } | |
315 | EXPORT_SYMBOL(zfs_holey); | |
316 | #endif /* SEEK_HOLE && SEEK_DATA */ | |
cf91b2b6 | 317 | |
c0d35759 | 318 | #if defined(_KERNEL) |
34dc7c2f BB |
319 | /* |
320 | * When a file is memory mapped, we must keep the IO data synchronized | |
321 | * between the DMU cache and the memory mapped pages. What this means: | |
322 | * | |
323 | * On Write: If we find a memory mapped page, we write to *both* | |
324 | * the page and the dmu buffer. | |
34dc7c2f | 325 | */ |
d164b209 | 326 | static void |
c0d35759 BB |
327 | update_pages(struct inode *ip, int64_t start, int len, |
328 | objset_t *os, uint64_t oid) | |
34dc7c2f | 329 | { |
c0d35759 BB |
330 | struct address_space *mp = ip->i_mapping; |
331 | struct page *pp; | |
332 | uint64_t nbytes; | |
d164b209 | 333 | int64_t off; |
c0d35759 | 334 | void *pb; |
34dc7c2f | 335 | |
8b1899d3 BB |
336 | off = start & (PAGE_SIZE-1); |
337 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
338 | nbytes = MIN(PAGE_SIZE - off, len); | |
34dc7c2f | 339 | |
8b1899d3 | 340 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
341 | if (pp) { |
342 | if (mapping_writably_mapped(mp)) | |
343 | flush_dcache_page(pp); | |
34dc7c2f | 344 | |
c0d35759 BB |
345 | pb = kmap(pp); |
346 | (void) dmu_read(os, oid, start+off, nbytes, pb+off, | |
9babb374 | 347 | DMU_READ_PREFETCH); |
c0d35759 BB |
348 | kunmap(pp); |
349 | ||
350 | if (mapping_writably_mapped(mp)) | |
351 | flush_dcache_page(pp); | |
352 | ||
353 | mark_page_accessed(pp); | |
354 | SetPageUptodate(pp); | |
355 | ClearPageError(pp); | |
356 | unlock_page(pp); | |
8b1899d3 | 357 | put_page(pp); |
34dc7c2f | 358 | } |
c0d35759 | 359 | |
d164b209 | 360 | len -= nbytes; |
34dc7c2f | 361 | off = 0; |
34dc7c2f | 362 | } |
34dc7c2f BB |
363 | } |
364 | ||
365 | /* | |
366 | * When a file is memory mapped, we must keep the IO data synchronized | |
367 | * between the DMU cache and the memory mapped pages. What this means: | |
368 | * | |
369 | * On Read: We "read" preferentially from memory mapped pages, | |
370 | * else we default from the dmu buffer. | |
371 | * | |
372 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
d3cc8b15 | 373 | * the file is memory mapped. |
34dc7c2f BB |
374 | */ |
375 | static int | |
3558fd73 | 376 | mappedread(struct inode *ip, int nbytes, uio_t *uio) |
34dc7c2f | 377 | { |
c0d35759 BB |
378 | struct address_space *mp = ip->i_mapping; |
379 | struct page *pp; | |
3558fd73 | 380 | znode_t *zp = ITOZ(ip); |
34dc7c2f | 381 | int64_t start, off; |
c0d35759 | 382 | uint64_t bytes; |
34dc7c2f BB |
383 | int len = nbytes; |
384 | int error = 0; | |
c0d35759 | 385 | void *pb; |
34dc7c2f BB |
386 | |
387 | start = uio->uio_loffset; | |
8b1899d3 BB |
388 | off = start & (PAGE_SIZE-1); |
389 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
390 | bytes = MIN(PAGE_SIZE - off, len); | |
c0d35759 | 391 | |
8b1899d3 | 392 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
393 | if (pp) { |
394 | ASSERT(PageUptodate(pp)); | |
395 | ||
396 | pb = kmap(pp); | |
397 | error = uiomove(pb + off, bytes, UIO_READ, uio); | |
398 | kunmap(pp); | |
399 | ||
400 | if (mapping_writably_mapped(mp)) | |
401 | flush_dcache_page(pp); | |
402 | ||
403 | mark_page_accessed(pp); | |
404 | unlock_page(pp); | |
8b1899d3 | 405 | put_page(pp); |
34dc7c2f | 406 | } else { |
804e0504 MA |
407 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
408 | uio, bytes); | |
34dc7c2f | 409 | } |
c0d35759 | 410 | |
34dc7c2f BB |
411 | len -= bytes; |
412 | off = 0; | |
413 | if (error) | |
414 | break; | |
415 | } | |
416 | return (error); | |
417 | } | |
c0d35759 | 418 | #endif /* _KERNEL */ |
34dc7c2f | 419 | |
c409e464 | 420 | unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ |
a966c564 | 421 | unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT; |
34dc7c2f BB |
422 | |
423 | /* | |
424 | * Read bytes from specified file into supplied buffer. | |
425 | * | |
3558fd73 | 426 | * IN: ip - inode of file to be read from. |
34dc7c2f BB |
427 | * uio - structure supplying read location, range info, |
428 | * and return buffer. | |
c0d35759 BB |
429 | * ioflag - FSYNC flags; used to provide FRSYNC semantics. |
430 | * O_DIRECT flag; used to bypass page cache. | |
34dc7c2f | 431 | * cr - credentials of caller. |
34dc7c2f BB |
432 | * |
433 | * OUT: uio - updated offset and range, buffer filled. | |
434 | * | |
d3cc8b15 | 435 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
436 | * |
437 | * Side Effects: | |
3558fd73 | 438 | * inode - atime updated if byte count > 0 |
34dc7c2f BB |
439 | */ |
440 | /* ARGSUSED */ | |
e5c39b95 | 441 | int |
3558fd73 | 442 | zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 443 | { |
3558fd73 BB |
444 | znode_t *zp = ITOZ(ip); |
445 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 446 | ssize_t n, nbytes; |
149e873a | 447 | int error = 0; |
34dc7c2f | 448 | rl_t *rl; |
3558fd73 | 449 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 450 | xuio_t *xuio = NULL; |
3558fd73 | 451 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f | 452 | |
3558fd73 | 453 | ZFS_ENTER(zsb); |
34dc7c2f | 454 | ZFS_VERIFY_ZP(zp); |
34dc7c2f | 455 | |
428870ff | 456 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
3558fd73 | 457 | ZFS_EXIT(zsb); |
2e528b49 | 458 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
459 | } |
460 | ||
461 | /* | |
462 | * Validate file offset | |
463 | */ | |
464 | if (uio->uio_loffset < (offset_t)0) { | |
3558fd73 | 465 | ZFS_EXIT(zsb); |
2e528b49 | 466 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
467 | } |
468 | ||
469 | /* | |
470 | * Fasttrack empty reads | |
471 | */ | |
472 | if (uio->uio_resid == 0) { | |
3558fd73 | 473 | ZFS_EXIT(zsb); |
34dc7c2f BB |
474 | return (0); |
475 | } | |
476 | ||
34dc7c2f BB |
477 | /* |
478 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
479 | */ | |
3558fd73 BB |
480 | if (ioflag & FRSYNC || zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
481 | zil_commit(zsb->z_log, zp->z_id); | |
34dc7c2f BB |
482 | |
483 | /* | |
484 | * Lock the range against changes. | |
485 | */ | |
d88895a0 CC |
486 | rl = zfs_range_lock(&zp->z_range_lock, uio->uio_loffset, uio->uio_resid, |
487 | RL_READER); | |
34dc7c2f BB |
488 | |
489 | /* | |
490 | * If we are reading past end-of-file we can skip | |
491 | * to the end; but we might still need to set atime. | |
492 | */ | |
428870ff | 493 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
494 | error = 0; |
495 | goto out; | |
496 | } | |
497 | ||
428870ff BB |
498 | ASSERT(uio->uio_loffset < zp->z_size); |
499 | n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); | |
500 | ||
3558fd73 | 501 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
502 | if ((uio->uio_extflg == UIO_XUIO) && |
503 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
504 | int nblk; | |
505 | int blksz = zp->z_blksz; | |
506 | uint64_t offset = uio->uio_loffset; | |
507 | ||
508 | xuio = (xuio_t *)uio; | |
509 | if ((ISP2(blksz))) { | |
510 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
511 | blksz)) / blksz; | |
512 | } else { | |
513 | ASSERT(offset + n <= blksz); | |
514 | nblk = 1; | |
515 | } | |
516 | (void) dmu_xuio_init(xuio, nblk); | |
517 | ||
3558fd73 | 518 | if (vn_has_cached_data(ip)) { |
428870ff BB |
519 | /* |
520 | * For simplicity, we always allocate a full buffer | |
521 | * even if we only expect to read a portion of a block. | |
522 | */ | |
523 | while (--nblk >= 0) { | |
524 | (void) dmu_xuio_add(xuio, | |
525 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
526 | blksz), 0, blksz); | |
527 | } | |
528 | } | |
529 | } | |
3558fd73 | 530 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
531 | |
532 | while (n > 0) { | |
533 | nbytes = MIN(n, zfs_read_chunk_size - | |
534 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); | |
535 | ||
804e0504 | 536 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
3558fd73 | 537 | error = mappedread(ip, nbytes, uio); |
804e0504 MA |
538 | } else { |
539 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), | |
540 | uio, nbytes); | |
541 | } | |
c0d35759 | 542 | |
b128c09f BB |
543 | if (error) { |
544 | /* convert checksum errors into IO errors */ | |
545 | if (error == ECKSUM) | |
2e528b49 | 546 | error = SET_ERROR(EIO); |
34dc7c2f | 547 | break; |
b128c09f | 548 | } |
34dc7c2f BB |
549 | |
550 | n -= nbytes; | |
551 | } | |
34dc7c2f BB |
552 | out: |
553 | zfs_range_unlock(rl); | |
554 | ||
3558fd73 | 555 | ZFS_EXIT(zsb); |
34dc7c2f BB |
556 | return (error); |
557 | } | |
e5c39b95 | 558 | EXPORT_SYMBOL(zfs_read); |
34dc7c2f | 559 | |
34dc7c2f BB |
560 | /* |
561 | * Write the bytes to a file. | |
562 | * | |
3558fd73 | 563 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
564 | * uio - structure supplying write location, range info, |
565 | * and data buffer. | |
566 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 567 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 568 | * cr - credentials of caller. |
34dc7c2f BB |
569 | * |
570 | * OUT: uio - updated offset and range. | |
571 | * | |
572 | * RETURN: 0 if success | |
573 | * error code if failure | |
574 | * | |
575 | * Timestamps: | |
3558fd73 | 576 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 577 | */ |
428870ff | 578 | |
34dc7c2f | 579 | /* ARGSUSED */ |
e5c39b95 | 580 | int |
3558fd73 | 581 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 582 | { |
3558fd73 BB |
583 | znode_t *zp = ITOZ(ip); |
584 | rlim64_t limit = uio->uio_limit; | |
34dc7c2f BB |
585 | ssize_t start_resid = uio->uio_resid; |
586 | ssize_t tx_bytes; | |
587 | uint64_t end_size; | |
588 | dmu_tx_t *tx; | |
3558fd73 | 589 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
590 | zilog_t *zilog; |
591 | offset_t woff; | |
592 | ssize_t n, nbytes; | |
593 | rl_t *rl; | |
3558fd73 BB |
594 | int max_blksz = zsb->z_max_blksz; |
595 | int error = 0; | |
9babb374 | 596 | arc_buf_t *abuf; |
5475aada | 597 | const iovec_t *aiov = NULL; |
428870ff | 598 | xuio_t *xuio = NULL; |
428870ff BB |
599 | int write_eof; |
600 | int count = 0; | |
601 | sa_bulk_attr_t bulk[4]; | |
602 | uint64_t mtime[2], ctime[2]; | |
2c6abf15 | 603 | uint32_t uid; |
5a6765cf | 604 | #ifdef HAVE_UIO_ZEROCOPY |
605 | int i_iov = 0; | |
606 | const iovec_t *iovp = uio->uio_iov; | |
3558fd73 | 607 | ASSERTV(int iovcnt = uio->uio_iovcnt); |
5a6765cf | 608 | #endif |
34dc7c2f | 609 | |
34dc7c2f BB |
610 | /* |
611 | * Fasttrack empty write | |
612 | */ | |
613 | n = start_resid; | |
614 | if (n == 0) | |
615 | return (0); | |
616 | ||
617 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) | |
618 | limit = MAXOFFSET_T; | |
619 | ||
3558fd73 | 620 | ZFS_ENTER(zsb); |
34dc7c2f | 621 | ZFS_VERIFY_ZP(zp); |
b128c09f | 622 | |
3558fd73 BB |
623 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
624 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
625 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); | |
626 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, | |
428870ff BB |
627 | &zp->z_pflags, 8); |
628 | ||
f3c9dca0 MT |
629 | /* |
630 | * Callers might not be able to detect properly that we are read-only, | |
631 | * so check it explicitly here. | |
632 | */ | |
633 | if (zfs_is_readonly(zsb)) { | |
634 | ZFS_EXIT(zsb); | |
635 | return (SET_ERROR(EROFS)); | |
636 | } | |
637 | ||
b128c09f BB |
638 | /* |
639 | * If immutable or not appending then return EPERM | |
640 | */ | |
428870ff BB |
641 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
642 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
643 | (uio->uio_loffset < zp->z_size))) { | |
3558fd73 | 644 | ZFS_EXIT(zsb); |
2e528b49 | 645 | return (SET_ERROR(EPERM)); |
b128c09f BB |
646 | } |
647 | ||
3558fd73 | 648 | zilog = zsb->z_log; |
34dc7c2f | 649 | |
428870ff BB |
650 | /* |
651 | * Validate file offset | |
652 | */ | |
653 | woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; | |
654 | if (woff < 0) { | |
3558fd73 | 655 | ZFS_EXIT(zsb); |
2e528b49 | 656 | return (SET_ERROR(EINVAL)); |
428870ff BB |
657 | } |
658 | ||
34dc7c2f BB |
659 | /* |
660 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
661 | * don't hold up txg. | |
428870ff | 662 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 663 | */ |
9cac042c | 664 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
665 | if ((uio->uio_extflg == UIO_XUIO) && |
666 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
667 | xuio = (xuio_t *)uio; | |
668 | else | |
9cac042c | 669 | #endif |
572e2857 | 670 | uio_prefaultpages(MIN(n, max_blksz), uio); |
34dc7c2f BB |
671 | |
672 | /* | |
673 | * If in append mode, set the io offset pointer to eof. | |
674 | */ | |
675 | if (ioflag & FAPPEND) { | |
676 | /* | |
428870ff BB |
677 | * Obtain an appending range lock to guarantee file append |
678 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f | 679 | */ |
d88895a0 | 680 | rl = zfs_range_lock(&zp->z_range_lock, 0, n, RL_APPEND); |
428870ff | 681 | woff = rl->r_off; |
34dc7c2f | 682 | if (rl->r_len == UINT64_MAX) { |
428870ff BB |
683 | /* |
684 | * We overlocked the file because this write will cause | |
685 | * the file block size to increase. | |
686 | * Note that zp_size cannot change with this lock held. | |
687 | */ | |
688 | woff = zp->z_size; | |
34dc7c2f | 689 | } |
428870ff | 690 | uio->uio_loffset = woff; |
34dc7c2f | 691 | } else { |
34dc7c2f | 692 | /* |
428870ff BB |
693 | * Note that if the file block size will change as a result of |
694 | * this write, then this range lock will lock the entire file | |
695 | * so that we can re-write the block safely. | |
34dc7c2f | 696 | */ |
d88895a0 | 697 | rl = zfs_range_lock(&zp->z_range_lock, woff, n, RL_WRITER); |
34dc7c2f BB |
698 | } |
699 | ||
700 | if (woff >= limit) { | |
701 | zfs_range_unlock(rl); | |
3558fd73 | 702 | ZFS_EXIT(zsb); |
2e528b49 | 703 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
704 | } |
705 | ||
706 | if ((woff + n) > limit || woff > (limit - n)) | |
707 | n = limit - woff; | |
708 | ||
428870ff BB |
709 | /* Will this write extend the file length? */ |
710 | write_eof = (woff + n > zp->z_size); | |
711 | ||
712 | end_size = MAX(zp->z_size, woff + n); | |
34dc7c2f BB |
713 | |
714 | /* | |
715 | * Write the file in reasonable size chunks. Each chunk is written | |
716 | * in a separate transaction; this keeps the intent log records small | |
717 | * and allows us to do more fine-grained space accounting. | |
718 | */ | |
719 | while (n > 0) { | |
9babb374 BB |
720 | abuf = NULL; |
721 | woff = uio->uio_loffset; | |
3558fd73 BB |
722 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
723 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 BB |
724 | if (abuf != NULL) |
725 | dmu_return_arcbuf(abuf); | |
2e528b49 | 726 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
727 | break; |
728 | } | |
729 | ||
428870ff | 730 | if (xuio && abuf == NULL) { |
5a6765cf | 731 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 732 | ASSERT(i_iov < iovcnt); |
5475aada | 733 | ASSERT3U(uio->uio_segflg, !=, UIO_BVEC); |
428870ff BB |
734 | aiov = &iovp[i_iov]; |
735 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
736 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
737 | ASSERT((aiov->iov_base == abuf->b_data) || |
738 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
739 | aiov->iov_len == arc_buf_size(abuf))); | |
740 | i_iov++; | |
5a6765cf | 741 | #endif |
428870ff BB |
742 | } else if (abuf == NULL && n >= max_blksz && |
743 | woff >= zp->z_size && | |
9babb374 BB |
744 | P2PHASE(woff, max_blksz) == 0 && |
745 | zp->z_blksz == max_blksz) { | |
428870ff BB |
746 | /* |
747 | * This write covers a full block. "Borrow" a buffer | |
748 | * from the dmu so that we can fill it before we enter | |
749 | * a transaction. This avoids the possibility of | |
750 | * holding up the transaction if the data copy hangs | |
751 | * up on a pagefault (e.g., from an NFS server mapping). | |
752 | */ | |
9babb374 BB |
753 | size_t cbytes; |
754 | ||
428870ff BB |
755 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
756 | max_blksz); | |
9babb374 BB |
757 | ASSERT(abuf != NULL); |
758 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
759 | if ((error = uiocopy(abuf->b_data, max_blksz, |
760 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
761 | dmu_return_arcbuf(abuf); |
762 | break; | |
763 | } | |
764 | ASSERT(cbytes == max_blksz); | |
765 | } | |
766 | ||
34dc7c2f BB |
767 | /* |
768 | * Start a transaction. | |
769 | */ | |
3558fd73 | 770 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 771 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 772 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 773 | zfs_sa_upgrade_txholds(tx, zp); |
384f8a09 | 774 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 775 | if (error) { |
34dc7c2f | 776 | dmu_tx_abort(tx); |
9babb374 BB |
777 | if (abuf != NULL) |
778 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
779 | break; |
780 | } | |
781 | ||
782 | /* | |
783 | * If zfs_range_lock() over-locked we grow the blocksize | |
784 | * and then reduce the lock range. This will only happen | |
785 | * on the first iteration since zfs_range_reduce() will | |
786 | * shrink down r_len to the appropriate size. | |
787 | */ | |
788 | if (rl->r_len == UINT64_MAX) { | |
789 | uint64_t new_blksz; | |
790 | ||
791 | if (zp->z_blksz > max_blksz) { | |
f1512ee6 MA |
792 | /* |
793 | * File's blocksize is already larger than the | |
794 | * "recordsize" property. Only let it grow to | |
795 | * the next power of 2. | |
796 | */ | |
34dc7c2f | 797 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 MA |
798 | new_blksz = MIN(end_size, |
799 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
800 | } else { |
801 | new_blksz = MIN(end_size, max_blksz); | |
802 | } | |
803 | zfs_grow_blocksize(zp, new_blksz, tx); | |
804 | zfs_range_reduce(rl, woff, n); | |
805 | } | |
806 | ||
807 | /* | |
808 | * XXX - should we really limit each write to z_max_blksz? | |
809 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
810 | */ | |
811 | nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); | |
34dc7c2f | 812 | |
9babb374 BB |
813 | if (abuf == NULL) { |
814 | tx_bytes = uio->uio_resid; | |
428870ff BB |
815 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
816 | uio, nbytes, tx); | |
9babb374 BB |
817 | tx_bytes -= uio->uio_resid; |
818 | } else { | |
819 | tx_bytes = nbytes; | |
428870ff BB |
820 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
821 | /* | |
822 | * If this is not a full block write, but we are | |
823 | * extending the file past EOF and this data starts | |
824 | * block-aligned, use assign_arcbuf(). Otherwise, | |
825 | * write via dmu_write(). | |
826 | */ | |
827 | if (tx_bytes < max_blksz && (!write_eof || | |
828 | aiov->iov_base != abuf->b_data)) { | |
829 | ASSERT(xuio); | |
3558fd73 | 830 | dmu_write(zsb->z_os, zp->z_id, woff, |
428870ff BB |
831 | aiov->iov_len, aiov->iov_base, tx); |
832 | dmu_return_arcbuf(abuf); | |
833 | xuio_stat_wbuf_copied(); | |
834 | } else { | |
835 | ASSERT(xuio || tx_bytes == max_blksz); | |
836 | dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl), | |
837 | woff, abuf, tx); | |
838 | } | |
9babb374 BB |
839 | ASSERT(tx_bytes <= uio->uio_resid); |
840 | uioskip(uio, tx_bytes); | |
841 | } | |
c0d35759 BB |
842 | |
843 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) | |
844 | update_pages(ip, woff, tx_bytes, zsb->z_os, zp->z_id); | |
34dc7c2f BB |
845 | |
846 | /* | |
847 | * If we made no progress, we're done. If we made even | |
848 | * partial progress, update the znode and ZIL accordingly. | |
849 | */ | |
850 | if (tx_bytes == 0) { | |
3558fd73 | 851 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 852 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
853 | dmu_tx_commit(tx); |
854 | ASSERT(error != 0); | |
855 | break; | |
856 | } | |
857 | ||
858 | /* | |
859 | * Clear Set-UID/Set-GID bits on successful write if not | |
4e33ba4c | 860 | * privileged and at least one of the execute bits is set. |
34dc7c2f BB |
861 | * |
862 | * It would be nice to to this after all writes have | |
863 | * been done, but that would still expose the ISUID/ISGID | |
864 | * to another app after the partial write is committed. | |
865 | * | |
572e2857 BB |
866 | * Note: we don't call zfs_fuid_map_id() here because |
867 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
868 | */ |
869 | mutex_enter(&zp->z_acl_lock); | |
2c6abf15 | 870 | uid = KUID_TO_SUID(ip->i_uid); |
428870ff | 871 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 872 | (S_IXUSR >> 6))) != 0 && |
428870ff | 873 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 874 | secpolicy_vnode_setid_retain(cr, |
2c6abf15 | 875 | ((zp->z_mode & S_ISUID) != 0 && uid == 0)) != 0) { |
428870ff BB |
876 | uint64_t newmode; |
877 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
12fa7f34 | 878 | ip->i_mode = newmode = zp->z_mode; |
3558fd73 | 879 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zsb), |
428870ff | 880 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
881 | } |
882 | mutex_exit(&zp->z_acl_lock); | |
883 | ||
0df9673f | 884 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
34dc7c2f BB |
885 | |
886 | /* | |
887 | * Update the file size (zp_size) if it has changed; | |
888 | * account for possible concurrent updates. | |
889 | */ | |
428870ff BB |
890 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
891 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 892 | uio->uio_loffset); |
428870ff BB |
893 | ASSERT(error == 0); |
894 | } | |
572e2857 BB |
895 | /* |
896 | * If we are replaying and eof is non zero then force | |
897 | * the file size to the specified eof. Note, there's no | |
898 | * concurrency during replay. | |
899 | */ | |
3558fd73 BB |
900 | if (zsb->z_replay && zsb->z_replay_eof != 0) |
901 | zp->z_size = zsb->z_replay_eof; | |
572e2857 | 902 | |
428870ff BB |
903 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
904 | ||
119a394a ED |
905 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
906 | NULL, NULL); | |
34dc7c2f BB |
907 | dmu_tx_commit(tx); |
908 | ||
909 | if (error != 0) | |
910 | break; | |
911 | ASSERT(tx_bytes == nbytes); | |
912 | n -= nbytes; | |
572e2857 BB |
913 | |
914 | if (!xuio && n > 0) | |
915 | uio_prefaultpages(MIN(n, max_blksz), uio); | |
34dc7c2f BB |
916 | } |
917 | ||
2a53e2da | 918 | zfs_inode_update(zp); |
34dc7c2f BB |
919 | zfs_range_unlock(rl); |
920 | ||
921 | /* | |
922 | * If we're in replay mode, or we made no progress, return error. | |
923 | * Otherwise, it's at least a partial write, so it's successful. | |
924 | */ | |
3558fd73 BB |
925 | if (zsb->z_replay || uio->uio_resid == start_resid) { |
926 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
927 | return (error); |
928 | } | |
929 | ||
428870ff | 930 | if (ioflag & (FSYNC | FDSYNC) || |
3558fd73 | 931 | zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 932 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 933 | |
3558fd73 | 934 | ZFS_EXIT(zsb); |
34dc7c2f BB |
935 | return (0); |
936 | } | |
e5c39b95 | 937 | EXPORT_SYMBOL(zfs_write); |
34dc7c2f | 938 | |
7c502b0b | 939 | /* |
940 | * Drop a reference on the passed inode asynchronously. This ensures | |
941 | * that the caller will never drop the last reference on an inode in | |
942 | * the current context. Doing so while holding open a tx could result | |
943 | * in a deadlock if iput_final() re-enters the filesystem code. | |
944 | */ | |
0a50679c BB |
945 | void |
946 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 947 | { |
0a50679c BB |
948 | objset_t *os = ITOZSB(ip)->z_os; |
949 | ||
3558fd73 | 950 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
951 | ASSERT(os != NULL); |
952 | ||
3558fd73 | 953 | if (atomic_read(&ip->i_count) == 1) |
7c502b0b | 954 | VERIFY(taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
48d3eb40 | 955 | (task_func_t *)iput, ip, TQ_SLEEP) != TASKQID_INVALID); |
3558fd73 BB |
956 | else |
957 | iput(ip); | |
958 | } | |
959 | ||
34dc7c2f | 960 | void |
428870ff | 961 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 962 | { |
428870ff | 963 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
964 | |
965 | if (zgd->zgd_db) | |
966 | dmu_buf_rele(zgd->zgd_db, zgd); | |
967 | ||
968 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f | 969 | |
9babb374 BB |
970 | /* |
971 | * Release the vnode asynchronously as we currently have the | |
972 | * txg stopped from syncing. | |
973 | */ | |
0a50679c | 974 | zfs_iput_async(ZTOI(zp)); |
428870ff BB |
975 | |
976 | if (error == 0 && zgd->zgd_bp) | |
977 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
978 | ||
34dc7c2f BB |
979 | kmem_free(zgd, sizeof (zgd_t)); |
980 | } | |
981 | ||
45d1cae3 BB |
982 | #ifdef DEBUG |
983 | static int zil_fault_io = 0; | |
984 | #endif | |
985 | ||
34dc7c2f BB |
986 | /* |
987 | * Get data to generate a TX_WRITE intent log record. | |
988 | */ | |
989 | int | |
990 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
991 | { | |
3558fd73 BB |
992 | zfs_sb_t *zsb = arg; |
993 | objset_t *os = zsb->z_os; | |
34dc7c2f | 994 | znode_t *zp; |
428870ff BB |
995 | uint64_t object = lr->lr_foid; |
996 | uint64_t offset = lr->lr_offset; | |
997 | uint64_t size = lr->lr_length; | |
998 | blkptr_t *bp = &lr->lr_blkptr; | |
34dc7c2f | 999 | dmu_buf_t *db; |
34dc7c2f | 1000 | zgd_t *zgd; |
34dc7c2f BB |
1001 | int error = 0; |
1002 | ||
428870ff BB |
1003 | ASSERT(zio != NULL); |
1004 | ASSERT(size != 0); | |
34dc7c2f BB |
1005 | |
1006 | /* | |
1007 | * Nothing to do if the file has been removed | |
1008 | */ | |
3558fd73 | 1009 | if (zfs_zget(zsb, object, &zp) != 0) |
2e528b49 | 1010 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 1011 | if (zp->z_unlinked) { |
9babb374 BB |
1012 | /* |
1013 | * Release the vnode asynchronously as we currently have the | |
1014 | * txg stopped from syncing. | |
1015 | */ | |
0a50679c | 1016 | zfs_iput_async(ZTOI(zp)); |
2e528b49 | 1017 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1018 | } |
1019 | ||
79c76d5b | 1020 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
3558fd73 | 1021 | zgd->zgd_zilog = zsb->z_log; |
428870ff BB |
1022 | zgd->zgd_private = zp; |
1023 | ||
34dc7c2f BB |
1024 | /* |
1025 | * Write records come in two flavors: immediate and indirect. | |
1026 | * For small writes it's cheaper to store the data with the | |
1027 | * log record (immediate); for large writes it's cheaper to | |
1028 | * sync the data and get a pointer to it (indirect) so that | |
1029 | * we don't have to write the data twice. | |
1030 | */ | |
1031 | if (buf != NULL) { /* immediate write */ | |
d88895a0 CC |
1032 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, size, |
1033 | RL_READER); | |
34dc7c2f | 1034 | /* test for truncation needs to be done while range locked */ |
428870ff | 1035 | if (offset >= zp->z_size) { |
2e528b49 | 1036 | error = SET_ERROR(ENOENT); |
428870ff BB |
1037 | } else { |
1038 | error = dmu_read(os, object, offset, size, buf, | |
1039 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1040 | } |
428870ff | 1041 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1042 | } else { /* indirect write */ |
34dc7c2f BB |
1043 | /* |
1044 | * Have to lock the whole block to ensure when it's | |
1045 | * written out and it's checksum is being calculated | |
1046 | * that no one can change the data. We need to re-check | |
1047 | * blocksize after we get the lock in case it's changed! | |
1048 | */ | |
1049 | for (;;) { | |
428870ff BB |
1050 | uint64_t blkoff; |
1051 | size = zp->z_blksz; | |
1052 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1053 | offset -= blkoff; | |
d88895a0 CC |
1054 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, |
1055 | size, RL_READER); | |
428870ff | 1056 | if (zp->z_blksz == size) |
34dc7c2f | 1057 | break; |
428870ff BB |
1058 | offset += blkoff; |
1059 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f BB |
1060 | } |
1061 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1062 | if (lr->lr_offset >= zp->z_size) |
2e528b49 | 1063 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1064 | #ifdef DEBUG |
1065 | if (zil_fault_io) { | |
2e528b49 | 1066 | error = SET_ERROR(EIO); |
45d1cae3 | 1067 | zil_fault_io = 0; |
45d1cae3 | 1068 | } |
45d1cae3 | 1069 | #endif |
34dc7c2f | 1070 | if (error == 0) |
428870ff BB |
1071 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1072 | DMU_READ_NO_PREFETCH); | |
1073 | ||
1074 | if (error == 0) { | |
03c6040b GW |
1075 | blkptr_t *obp = dmu_buf_get_blkptr(db); |
1076 | if (obp) { | |
1077 | ASSERT(BP_IS_HOLE(bp)); | |
1078 | *bp = *obp; | |
1079 | } | |
1080 | ||
428870ff BB |
1081 | zgd->zgd_db = db; |
1082 | zgd->zgd_bp = bp; | |
1083 | ||
1084 | ASSERT(db->db_offset == offset); | |
1085 | ASSERT(db->db_size == size); | |
1086 | ||
1087 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1088 | zfs_get_done, zgd); | |
1089 | ASSERT(error || lr->lr_length <= zp->z_blksz); | |
1090 | ||
1091 | /* | |
1092 | * On success, we need to wait for the write I/O | |
1093 | * initiated by dmu_sync() to complete before we can | |
1094 | * release this dbuf. We will finish everything up | |
1095 | * in the zfs_get_done() callback. | |
1096 | */ | |
1097 | if (error == 0) | |
1098 | return (0); | |
1099 | ||
1100 | if (error == EALREADY) { | |
1101 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
1102 | error = 0; | |
1103 | } | |
1104 | } | |
34dc7c2f | 1105 | } |
428870ff BB |
1106 | |
1107 | zfs_get_done(zgd, error); | |
1108 | ||
34dc7c2f BB |
1109 | return (error); |
1110 | } | |
1111 | ||
1112 | /*ARGSUSED*/ | |
3558fd73 BB |
1113 | int |
1114 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1115 | { |
3558fd73 BB |
1116 | znode_t *zp = ITOZ(ip); |
1117 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
1118 | int error; |
1119 | ||
3558fd73 | 1120 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1121 | ZFS_VERIFY_ZP(zp); |
1122 | ||
1123 | if (flag & V_ACE_MASK) | |
1124 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1125 | else | |
1126 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1127 | ||
3558fd73 | 1128 | ZFS_EXIT(zsb); |
45d1cae3 BB |
1129 | return (error); |
1130 | } | |
3558fd73 | 1131 | EXPORT_SYMBOL(zfs_access); |
45d1cae3 | 1132 | |
34dc7c2f BB |
1133 | /* |
1134 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1135 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1136 | * |
3558fd73 | 1137 | * IN: dip - inode of directory to search. |
34dc7c2f | 1138 | * nm - name of entry to lookup. |
34dc7c2f | 1139 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1140 | * cr - credentials of caller. |
34dc7c2f BB |
1141 | * direntflags - directory lookup flags |
1142 | * realpnp - returned pathname. | |
1143 | * | |
3558fd73 | 1144 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1145 | * |
d3cc8b15 | 1146 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1147 | * |
1148 | * Timestamps: | |
1149 | * NA | |
1150 | */ | |
1151 | /* ARGSUSED */ | |
e5c39b95 | 1152 | int |
3558fd73 BB |
1153 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1154 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1155 | { |
3558fd73 BB |
1156 | znode_t *zdp = ITOZ(dip); |
1157 | zfs_sb_t *zsb = ITOZSB(dip); | |
1158 | int error = 0; | |
45d1cae3 BB |
1159 | |
1160 | /* fast path */ | |
1161 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { | |
1162 | ||
3558fd73 | 1163 | if (!S_ISDIR(dip->i_mode)) { |
2e528b49 | 1164 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1165 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 1166 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1167 | } |
1168 | ||
1169 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1170 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1171 | if (!error) { | |
3558fd73 BB |
1172 | *ipp = dip; |
1173 | igrab(*ipp); | |
45d1cae3 BB |
1174 | return (0); |
1175 | } | |
1176 | return (error); | |
3558fd73 | 1177 | #ifdef HAVE_DNLC |
45d1cae3 BB |
1178 | } else { |
1179 | vnode_t *tvp = dnlc_lookup(dvp, nm); | |
1180 | ||
1181 | if (tvp) { | |
1182 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1183 | if (error) { | |
3558fd73 | 1184 | iput(tvp); |
45d1cae3 BB |
1185 | return (error); |
1186 | } | |
1187 | if (tvp == DNLC_NO_VNODE) { | |
3558fd73 | 1188 | iput(tvp); |
2e528b49 | 1189 | return (SET_ERROR(ENOENT)); |
45d1cae3 BB |
1190 | } else { |
1191 | *vpp = tvp; | |
1192 | return (specvp_check(vpp, cr)); | |
1193 | } | |
1194 | } | |
3558fd73 | 1195 | #endif /* HAVE_DNLC */ |
45d1cae3 BB |
1196 | } |
1197 | } | |
1198 | ||
3558fd73 | 1199 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1200 | ZFS_VERIFY_ZP(zdp); |
1201 | ||
3558fd73 | 1202 | *ipp = NULL; |
34dc7c2f BB |
1203 | |
1204 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1205 | /* |
1206 | * We don't allow recursive attributes.. | |
1207 | * Maybe someday we will. | |
1208 | */ | |
428870ff | 1209 | if (zdp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1210 | ZFS_EXIT(zsb); |
2e528b49 | 1211 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1212 | } |
1213 | ||
3558fd73 BB |
1214 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
1215 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1216 | return (error); |
1217 | } | |
1218 | ||
1219 | /* | |
1220 | * Do we have permission to get into attribute directory? | |
1221 | */ | |
1222 | ||
3558fd73 | 1223 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1224 | B_FALSE, cr))) { |
3558fd73 BB |
1225 | iput(*ipp); |
1226 | *ipp = NULL; | |
34dc7c2f BB |
1227 | } |
1228 | ||
3558fd73 | 1229 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1230 | return (error); |
1231 | } | |
1232 | ||
3558fd73 BB |
1233 | if (!S_ISDIR(dip->i_mode)) { |
1234 | ZFS_EXIT(zsb); | |
2e528b49 | 1235 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1236 | } |
1237 | ||
1238 | /* | |
1239 | * Check accessibility of directory. | |
1240 | */ | |
1241 | ||
149e873a | 1242 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
3558fd73 | 1243 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1244 | return (error); |
1245 | } | |
1246 | ||
3558fd73 | 1247 | if (zsb->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1248 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1249 | ZFS_EXIT(zsb); |
2e528b49 | 1250 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1251 | } |
1252 | ||
3558fd73 BB |
1253 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1254 | if ((error == 0) && (*ipp)) | |
1255 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1256 | |
3558fd73 | 1257 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1258 | return (error); |
1259 | } | |
e5c39b95 | 1260 | EXPORT_SYMBOL(zfs_lookup); |
34dc7c2f BB |
1261 | |
1262 | /* | |
1263 | * Attempt to create a new entry in a directory. If the entry | |
1264 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1265 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1266 | * |
3558fd73 | 1267 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1268 | * name - name of new file entry. |
1269 | * vap - attributes of new file. | |
1270 | * excl - flag indicating exclusive or non-exclusive mode. | |
1271 | * mode - mode to open file with. | |
1272 | * cr - credentials of caller. | |
1273 | * flag - large file flag [UNUSED]. | |
3558fd73 | 1274 | * vsecp - ACL to be set |
34dc7c2f | 1275 | * |
3558fd73 | 1276 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1277 | * |
d3cc8b15 | 1278 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1279 | * |
1280 | * Timestamps: | |
3558fd73 BB |
1281 | * dip - ctime|mtime updated if new entry created |
1282 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1283 | */ |
1284 | ||
1285 | /* ARGSUSED */ | |
e5c39b95 | 1286 | int |
3558fd73 BB |
1287 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1288 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1289 | { |
3558fd73 BB |
1290 | znode_t *zp, *dzp = ITOZ(dip); |
1291 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1292 | zilog_t *zilog; |
1293 | objset_t *os; | |
1294 | zfs_dirlock_t *dl; | |
1295 | dmu_tx_t *tx; | |
1296 | int error; | |
b128c09f | 1297 | uid_t uid; |
149e873a | 1298 | gid_t gid; |
428870ff | 1299 | zfs_acl_ids_t acl_ids; |
9babb374 | 1300 | boolean_t fuid_dirtied; |
428870ff | 1301 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 1302 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1303 | |
1304 | /* | |
1305 | * If we have an ephemeral id, ACL, or XVATTR then | |
1306 | * make sure file system is at proper version | |
1307 | */ | |
1308 | ||
149e873a | 1309 | gid = crgetgid(cr); |
3558fd73 | 1310 | uid = crgetuid(cr); |
b128c09f | 1311 | |
3558fd73 BB |
1312 | if (zsb->z_use_fuids == B_FALSE && |
1313 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
2e528b49 | 1314 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1315 | |
32dec7bd | 1316 | if (name == NULL) |
1317 | return (SET_ERROR(EINVAL)); | |
1318 | ||
3558fd73 | 1319 | ZFS_ENTER(zsb); |
34dc7c2f | 1320 | ZFS_VERIFY_ZP(dzp); |
3558fd73 BB |
1321 | os = zsb->z_os; |
1322 | zilog = zsb->z_log; | |
34dc7c2f | 1323 | |
3558fd73 | 1324 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1325 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1326 | ZFS_EXIT(zsb); |
2e528b49 | 1327 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1328 | } |
1329 | ||
5484965a | 1330 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1331 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1332 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1333 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1334 | return (error); |
1335 | } | |
1336 | } | |
34dc7c2f | 1337 | |
3558fd73 BB |
1338 | top: |
1339 | *ipp = NULL; | |
34dc7c2f BB |
1340 | if (*name == '\0') { |
1341 | /* | |
1342 | * Null component name refers to the directory itself. | |
1343 | */ | |
3558fd73 | 1344 | igrab(dip); |
34dc7c2f BB |
1345 | zp = dzp; |
1346 | dl = NULL; | |
1347 | error = 0; | |
1348 | } else { | |
3558fd73 | 1349 | /* possible igrab(zp) */ |
34dc7c2f BB |
1350 | int zflg = 0; |
1351 | ||
1352 | if (flag & FIGNORECASE) | |
1353 | zflg |= ZCILOOK; | |
1354 | ||
1355 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1356 | NULL, NULL); | |
1357 | if (error) { | |
572e2857 BB |
1358 | if (have_acl) |
1359 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1360 | if (strcmp(name, "..") == 0) |
2e528b49 | 1361 | error = SET_ERROR(EISDIR); |
3558fd73 | 1362 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1363 | return (error); |
1364 | } | |
1365 | } | |
428870ff | 1366 | |
34dc7c2f BB |
1367 | if (zp == NULL) { |
1368 | uint64_t txtype; | |
1369 | ||
1370 | /* | |
1371 | * Create a new file object and update the directory | |
1372 | * to reference it. | |
1373 | */ | |
149e873a | 1374 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1375 | if (have_acl) |
1376 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1377 | goto out; |
1378 | } | |
1379 | ||
1380 | /* | |
1381 | * We only support the creation of regular files in | |
1382 | * extended attribute directories. | |
1383 | */ | |
428870ff | 1384 | |
3558fd73 | 1385 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1386 | if (have_acl) |
1387 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 1388 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1389 | goto out; |
1390 | } | |
1391 | ||
428870ff BB |
1392 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1393 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1394 | goto out; |
428870ff BB |
1395 | have_acl = B_TRUE; |
1396 | ||
3558fd73 | 1397 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1398 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1399 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1400 | goto out; |
1401 | } | |
1402 | ||
34dc7c2f | 1403 | tx = dmu_tx_create(os); |
428870ff BB |
1404 | |
1405 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1406 | ZFS_SA_BASE_ATTR_SIZE); | |
1407 | ||
3558fd73 | 1408 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1409 | if (fuid_dirtied) |
3558fd73 | 1410 | zfs_fuid_txhold(zsb, tx); |
34dc7c2f | 1411 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1412 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
3558fd73 | 1413 | if (!zsb->z_use_sa && |
428870ff | 1414 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1415 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1416 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1417 | } |
e8b96c60 | 1418 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1419 | if (error) { |
1420 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1421 | if (error == ERESTART) { |
e8b96c60 | 1422 | waited = B_TRUE; |
34dc7c2f BB |
1423 | dmu_tx_wait(tx); |
1424 | dmu_tx_abort(tx); | |
1425 | goto top; | |
1426 | } | |
428870ff | 1427 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1428 | dmu_tx_abort(tx); |
3558fd73 | 1429 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1430 | return (error); |
1431 | } | |
428870ff | 1432 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 BB |
1433 | |
1434 | if (fuid_dirtied) | |
3558fd73 | 1435 | zfs_fuid_sync(zsb, tx); |
9babb374 | 1436 | |
34dc7c2f BB |
1437 | (void) zfs_link_create(dl, zp, tx, ZNEW); |
1438 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); | |
1439 | if (flag & FIGNORECASE) | |
1440 | txtype |= TX_CI; | |
1441 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1442 | vsecp, acl_ids.z_fuidp, vap); |
1443 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1444 | dmu_tx_commit(tx); |
1445 | } else { | |
1446 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1447 | ||
572e2857 BB |
1448 | if (have_acl) |
1449 | zfs_acl_ids_free(&acl_ids); | |
1450 | have_acl = B_FALSE; | |
1451 | ||
34dc7c2f BB |
1452 | /* |
1453 | * A directory entry already exists for this name. | |
1454 | */ | |
1455 | /* | |
1456 | * Can't truncate an existing file if in exclusive mode. | |
1457 | */ | |
3558fd73 | 1458 | if (excl) { |
2e528b49 | 1459 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1460 | goto out; |
1461 | } | |
1462 | /* | |
1463 | * Can't open a directory for writing. | |
1464 | */ | |
3558fd73 | 1465 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1466 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1467 | goto out; |
1468 | } | |
1469 | /* | |
1470 | * Verify requested access to file. | |
1471 | */ | |
1472 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1473 | goto out; | |
1474 | } | |
1475 | ||
1476 | mutex_enter(&dzp->z_lock); | |
1477 | dzp->z_seq++; | |
1478 | mutex_exit(&dzp->z_lock); | |
1479 | ||
1480 | /* | |
1481 | * Truncate regular files if requested. | |
1482 | */ | |
3558fd73 BB |
1483 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1484 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 1485 | /* we can't hold any locks when calling zfs_freesp() */ |
609603a5 B |
1486 | if (dl) { |
1487 | zfs_dirent_unlock(dl); | |
1488 | dl = NULL; | |
1489 | } | |
34dc7c2f | 1490 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1491 | } |
1492 | } | |
1493 | out: | |
1494 | ||
1495 | if (dl) | |
1496 | zfs_dirent_unlock(dl); | |
1497 | ||
1498 | if (error) { | |
1499 | if (zp) | |
3558fd73 | 1500 | iput(ZTOI(zp)); |
34dc7c2f | 1501 | } else { |
960e08fe BB |
1502 | zfs_inode_update(dzp); |
1503 | zfs_inode_update(zp); | |
3558fd73 | 1504 | *ipp = ZTOI(zp); |
34dc7c2f | 1505 | } |
34dc7c2f | 1506 | |
3558fd73 | 1507 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1508 | zil_commit(zilog, 0); |
428870ff | 1509 | |
3558fd73 | 1510 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1511 | return (error); |
1512 | } | |
e5c39b95 | 1513 | EXPORT_SYMBOL(zfs_create); |
34dc7c2f | 1514 | |
ace1eae8 CC |
1515 | /* ARGSUSED */ |
1516 | int | |
1517 | zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, | |
1518 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
1519 | { | |
1520 | znode_t *zp = NULL, *dzp = ITOZ(dip); | |
1521 | zfs_sb_t *zsb = ITOZSB(dip); | |
1522 | objset_t *os; | |
1523 | dmu_tx_t *tx; | |
1524 | int error; | |
1525 | uid_t uid; | |
1526 | gid_t gid; | |
1527 | zfs_acl_ids_t acl_ids; | |
1528 | boolean_t fuid_dirtied; | |
1529 | boolean_t have_acl = B_FALSE; | |
1530 | boolean_t waited = B_FALSE; | |
1531 | ||
1532 | /* | |
1533 | * If we have an ephemeral id, ACL, or XVATTR then | |
1534 | * make sure file system is at proper version | |
1535 | */ | |
1536 | ||
1537 | gid = crgetgid(cr); | |
1538 | uid = crgetuid(cr); | |
1539 | ||
1540 | if (zsb->z_use_fuids == B_FALSE && | |
1541 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
1542 | return (SET_ERROR(EINVAL)); | |
1543 | ||
1544 | ZFS_ENTER(zsb); | |
1545 | ZFS_VERIFY_ZP(dzp); | |
1546 | os = zsb->z_os; | |
1547 | ||
1548 | if (vap->va_mask & ATTR_XVATTR) { | |
1549 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
1550 | crgetuid(cr), cr, vap->va_mode)) != 0) { | |
1551 | ZFS_EXIT(zsb); | |
1552 | return (error); | |
1553 | } | |
1554 | } | |
1555 | ||
1556 | top: | |
1557 | *ipp = NULL; | |
1558 | ||
1559 | /* | |
1560 | * Create a new file object and update the directory | |
1561 | * to reference it. | |
1562 | */ | |
1563 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { | |
1564 | if (have_acl) | |
1565 | zfs_acl_ids_free(&acl_ids); | |
1566 | goto out; | |
1567 | } | |
1568 | ||
1569 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, | |
1570 | cr, vsecp, &acl_ids)) != 0) | |
1571 | goto out; | |
1572 | have_acl = B_TRUE; | |
1573 | ||
1574 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { | |
1575 | zfs_acl_ids_free(&acl_ids); | |
1576 | error = SET_ERROR(EDQUOT); | |
1577 | goto out; | |
1578 | } | |
1579 | ||
1580 | tx = dmu_tx_create(os); | |
1581 | ||
1582 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1583 | ZFS_SA_BASE_ATTR_SIZE); | |
1584 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); | |
1585 | ||
1586 | fuid_dirtied = zsb->z_fuid_dirty; | |
1587 | if (fuid_dirtied) | |
1588 | zfs_fuid_txhold(zsb, tx); | |
1589 | if (!zsb->z_use_sa && | |
1590 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
1591 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
1592 | 0, acl_ids.z_aclp->z_acl_bytes); | |
1593 | } | |
1594 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); | |
1595 | if (error) { | |
1596 | if (error == ERESTART) { | |
1597 | waited = B_TRUE; | |
1598 | dmu_tx_wait(tx); | |
1599 | dmu_tx_abort(tx); | |
1600 | goto top; | |
1601 | } | |
1602 | zfs_acl_ids_free(&acl_ids); | |
1603 | dmu_tx_abort(tx); | |
1604 | ZFS_EXIT(zsb); | |
1605 | return (error); | |
1606 | } | |
1607 | zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); | |
1608 | ||
1609 | if (fuid_dirtied) | |
1610 | zfs_fuid_sync(zsb, tx); | |
1611 | ||
1612 | /* Add to unlinked set */ | |
1613 | zp->z_unlinked = 1; | |
1614 | zfs_unlinked_add(zp, tx); | |
1615 | zfs_acl_ids_free(&acl_ids); | |
1616 | dmu_tx_commit(tx); | |
1617 | out: | |
1618 | ||
1619 | if (error) { | |
1620 | if (zp) | |
1621 | iput(ZTOI(zp)); | |
1622 | } else { | |
1623 | zfs_inode_update(dzp); | |
1624 | zfs_inode_update(zp); | |
1625 | *ipp = ZTOI(zp); | |
1626 | } | |
1627 | ||
1628 | ZFS_EXIT(zsb); | |
1629 | return (error); | |
1630 | } | |
1631 | ||
34dc7c2f BB |
1632 | /* |
1633 | * Remove an entry from a directory. | |
1634 | * | |
3558fd73 | 1635 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1636 | * name - name of entry to remove. |
1637 | * cr - credentials of caller. | |
34dc7c2f BB |
1638 | * |
1639 | * RETURN: 0 if success | |
1640 | * error code if failure | |
1641 | * | |
1642 | * Timestamps: | |
3558fd73 BB |
1643 | * dip - ctime|mtime |
1644 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1645 | */ |
428870ff BB |
1646 | |
1647 | uint64_t null_xattr = 0; | |
1648 | ||
34dc7c2f | 1649 | /*ARGSUSED*/ |
e5c39b95 | 1650 | int |
da5e151f | 1651 | zfs_remove(struct inode *dip, char *name, cred_t *cr, int flags) |
34dc7c2f | 1652 | { |
3558fd73 | 1653 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1654 | znode_t *xzp; |
3558fd73 BB |
1655 | struct inode *ip; |
1656 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f | 1657 | zilog_t *zilog; |
a966c564 | 1658 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 1659 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 1660 | uint64_t obj = 0; |
dfbc8630 | 1661 | uint64_t links; |
34dc7c2f BB |
1662 | zfs_dirlock_t *dl; |
1663 | dmu_tx_t *tx; | |
a966c564 K |
1664 | boolean_t may_delete_now, delete_now = FALSE; |
1665 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
1666 | uint64_t txtype; |
1667 | pathname_t *realnmp = NULL; | |
1668 | pathname_t realnm; | |
1669 | int error; | |
1670 | int zflg = ZEXISTS; | |
e8b96c60 | 1671 | boolean_t waited = B_FALSE; |
34dc7c2f | 1672 | |
32dec7bd | 1673 | if (name == NULL) |
1674 | return (SET_ERROR(EINVAL)); | |
1675 | ||
3558fd73 | 1676 | ZFS_ENTER(zsb); |
34dc7c2f | 1677 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1678 | zilog = zsb->z_log; |
34dc7c2f BB |
1679 | |
1680 | if (flags & FIGNORECASE) { | |
1681 | zflg |= ZCILOOK; | |
1682 | pn_alloc(&realnm); | |
1683 | realnmp = &realnm; | |
1684 | } | |
1685 | ||
1686 | top: | |
572e2857 BB |
1687 | xattr_obj = 0; |
1688 | xzp = NULL; | |
34dc7c2f BB |
1689 | /* |
1690 | * Attempt to lock directory; fail if entry doesn't exist. | |
1691 | */ | |
149e873a BB |
1692 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1693 | NULL, realnmp))) { | |
34dc7c2f BB |
1694 | if (realnmp) |
1695 | pn_free(realnmp); | |
3558fd73 | 1696 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1697 | return (error); |
1698 | } | |
1699 | ||
3558fd73 | 1700 | ip = ZTOI(zp); |
34dc7c2f | 1701 | |
149e873a | 1702 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1703 | goto out; |
1704 | } | |
1705 | ||
1706 | /* | |
1707 | * Need to use rmdir for removing directories. | |
1708 | */ | |
3558fd73 | 1709 | if (S_ISDIR(ip->i_mode)) { |
2e528b49 | 1710 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1711 | goto out; |
1712 | } | |
1713 | ||
3558fd73 | 1714 | #ifdef HAVE_DNLC |
34dc7c2f BB |
1715 | if (realnmp) |
1716 | dnlc_remove(dvp, realnmp->pn_buf); | |
1717 | else | |
1718 | dnlc_remove(dvp, name); | |
3558fd73 | 1719 | #endif /* HAVE_DNLC */ |
34dc7c2f | 1720 | |
19d55079 MA |
1721 | mutex_enter(&zp->z_lock); |
1722 | may_delete_now = atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped); | |
1723 | mutex_exit(&zp->z_lock); | |
1724 | ||
34dc7c2f | 1725 | /* |
a966c564 K |
1726 | * We may delete the znode now, or we may put it in the unlinked set; |
1727 | * it depends on whether we're the last link, and on whether there are | |
1728 | * other holds on the inode. So we dmu_tx_hold() the right things to | |
1729 | * allow for either case. | |
34dc7c2f | 1730 | */ |
572e2857 | 1731 | obj = zp->z_id; |
3558fd73 | 1732 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 1733 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1734 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1735 | zfs_sa_upgrade_txholds(tx, zp); | |
1736 | zfs_sa_upgrade_txholds(tx, dzp); | |
a966c564 K |
1737 | if (may_delete_now) { |
1738 | toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; | |
1739 | /* if the file is too big, only hold_free a token amount */ | |
1740 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1741 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1742 | } | |
34dc7c2f BB |
1743 | |
1744 | /* are there any extended attributes? */ | |
3558fd73 | 1745 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1746 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1747 | if (error == 0 && xattr_obj) { |
3558fd73 | 1748 | error = zfs_zget(zsb, xattr_obj, &xzp); |
c99c9001 | 1749 | ASSERT0(error); |
428870ff BB |
1750 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1751 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1752 | } |
1753 | ||
a966c564 K |
1754 | mutex_enter(&zp->z_lock); |
1755 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
1756 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
1757 | mutex_exit(&zp->z_lock); | |
1758 | ||
34dc7c2f | 1759 | /* charge as an update -- would be nice not to charge at all */ |
3558fd73 | 1760 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1761 | |
19d55079 | 1762 | /* |
1a04bab3 | 1763 | * Mark this transaction as typically resulting in a net free of space |
19d55079 | 1764 | */ |
1a04bab3 | 1765 | dmu_tx_mark_netfree(tx); |
19d55079 | 1766 | |
e8b96c60 | 1767 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1768 | if (error) { |
1769 | zfs_dirent_unlock(dl); | |
3558fd73 | 1770 | iput(ip); |
572e2857 | 1771 | if (xzp) |
3558fd73 | 1772 | iput(ZTOI(xzp)); |
fb5f0bc8 | 1773 | if (error == ERESTART) { |
e8b96c60 | 1774 | waited = B_TRUE; |
34dc7c2f BB |
1775 | dmu_tx_wait(tx); |
1776 | dmu_tx_abort(tx); | |
1777 | goto top; | |
1778 | } | |
1779 | if (realnmp) | |
1780 | pn_free(realnmp); | |
1781 | dmu_tx_abort(tx); | |
3558fd73 | 1782 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1783 | return (error); |
1784 | } | |
1785 | ||
1786 | /* | |
1787 | * Remove the directory entry. | |
1788 | */ | |
1789 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1790 | ||
1791 | if (error) { | |
1792 | dmu_tx_commit(tx); | |
1793 | goto out; | |
1794 | } | |
1795 | ||
1796 | if (unlinked) { | |
572e2857 BB |
1797 | /* |
1798 | * Hold z_lock so that we can make sure that the ACL obj | |
1799 | * hasn't changed. Could have been deleted due to | |
1800 | * zfs_sa_upgrade(). | |
1801 | */ | |
1802 | mutex_enter(&zp->z_lock); | |
3558fd73 | 1803 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1804 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
a966c564 K |
1805 | delete_now = may_delete_now && !toobig && |
1806 | atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped) && | |
1807 | xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == | |
1808 | acl_obj; | |
1809 | } | |
1810 | ||
1811 | if (delete_now) { | |
1812 | if (xattr_obj_unlinked) { | |
dfbc8630 | 1813 | ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); |
a966c564 K |
1814 | mutex_enter(&xzp->z_lock); |
1815 | xzp->z_unlinked = 1; | |
dfbc8630 CD |
1816 | clear_nlink(ZTOI(xzp)); |
1817 | links = 0; | |
a966c564 | 1818 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zsb), |
dfbc8630 | 1819 | &links, sizeof (links), tx); |
a966c564 K |
1820 | ASSERT3U(error, ==, 0); |
1821 | mutex_exit(&xzp->z_lock); | |
1822 | zfs_unlinked_add(xzp, tx); | |
1823 | ||
1824 | if (zp->z_is_sa) | |
1825 | error = sa_remove(zp->z_sa_hdl, | |
1826 | SA_ZPL_XATTR(zsb), tx); | |
1827 | else | |
1828 | error = sa_update(zp->z_sa_hdl, | |
1829 | SA_ZPL_XATTR(zsb), &null_xattr, | |
1830 | sizeof (uint64_t), tx); | |
1831 | ASSERT0(error); | |
1832 | } | |
1833 | /* | |
1834 | * Add to the unlinked set because a new reference could be | |
1835 | * taken concurrently resulting in a deferred destruction. | |
1836 | */ | |
1837 | zfs_unlinked_add(zp, tx); | |
1838 | mutex_exit(&zp->z_lock); | |
1839 | zfs_inode_update(zp); | |
1840 | iput(ip); | |
1841 | } else if (unlinked) { | |
572e2857 | 1842 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1843 | zfs_unlinked_add(zp, tx); |
1844 | } | |
1845 | ||
1846 | txtype = TX_REMOVE; | |
1847 | if (flags & FIGNORECASE) | |
1848 | txtype |= TX_CI; | |
572e2857 | 1849 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1850 | |
1851 | dmu_tx_commit(tx); | |
1852 | out: | |
1853 | if (realnmp) | |
1854 | pn_free(realnmp); | |
1855 | ||
1856 | zfs_dirent_unlock(dl); | |
960e08fe | 1857 | zfs_inode_update(dzp); |
34dc7c2f | 1858 | |
a966c564 K |
1859 | if (!delete_now) { |
1860 | zfs_inode_update(zp); | |
1861 | zfs_iput_async(ip); | |
1862 | } | |
1863 | ||
1864 | if (xzp) { | |
1865 | zfs_inode_update(xzp); | |
1866 | zfs_iput_async(ZTOI(xzp)); | |
1867 | } | |
428870ff | 1868 | |
3558fd73 | 1869 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1870 | zil_commit(zilog, 0); |
34dc7c2f | 1871 | |
3558fd73 | 1872 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1873 | return (error); |
1874 | } | |
e5c39b95 | 1875 | EXPORT_SYMBOL(zfs_remove); |
34dc7c2f BB |
1876 | |
1877 | /* | |
3558fd73 | 1878 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1879 | * provided. Return a pointer to the inserted directory. |
1880 | * | |
3558fd73 | 1881 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1882 | * dirname - name of new directory. |
1883 | * vap - attributes of new directory. | |
1884 | * cr - credentials of caller. | |
34dc7c2f BB |
1885 | * vsecp - ACL to be set |
1886 | * | |
3558fd73 | 1887 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1888 | * |
1889 | * RETURN: 0 if success | |
1890 | * error code if failure | |
1891 | * | |
1892 | * Timestamps: | |
3558fd73 BB |
1893 | * dip - ctime|mtime updated |
1894 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1895 | */ |
1896 | /*ARGSUSED*/ | |
e5c39b95 | 1897 | int |
3558fd73 BB |
1898 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1899 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1900 | { |
3558fd73 BB |
1901 | znode_t *zp, *dzp = ITOZ(dip); |
1902 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1903 | zilog_t *zilog; |
1904 | zfs_dirlock_t *dl; | |
1905 | uint64_t txtype; | |
1906 | dmu_tx_t *tx; | |
1907 | int error; | |
34dc7c2f | 1908 | int zf = ZNEW; |
b128c09f BB |
1909 | uid_t uid; |
1910 | gid_t gid = crgetgid(cr); | |
428870ff | 1911 | zfs_acl_ids_t acl_ids; |
9babb374 | 1912 | boolean_t fuid_dirtied; |
e8b96c60 | 1913 | boolean_t waited = B_FALSE; |
34dc7c2f | 1914 | |
3558fd73 | 1915 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1916 | |
1917 | /* | |
1918 | * If we have an ephemeral id, ACL, or XVATTR then | |
1919 | * make sure file system is at proper version | |
1920 | */ | |
1921 | ||
3558fd73 BB |
1922 | uid = crgetuid(cr); |
1923 | if (zsb->z_use_fuids == B_FALSE && | |
1924 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
2e528b49 | 1925 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1926 | |
32dec7bd | 1927 | if (dirname == NULL) |
1928 | return (SET_ERROR(EINVAL)); | |
1929 | ||
3558fd73 | 1930 | ZFS_ENTER(zsb); |
34dc7c2f | 1931 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1932 | zilog = zsb->z_log; |
34dc7c2f | 1933 | |
428870ff | 1934 | if (dzp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1935 | ZFS_EXIT(zsb); |
2e528b49 | 1936 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1937 | } |
1938 | ||
3558fd73 | 1939 | if (zsb->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1940 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1941 | ZFS_EXIT(zsb); |
2e528b49 | 1942 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1943 | } |
1944 | if (flags & FIGNORECASE) | |
1945 | zf |= ZCILOOK; | |
1946 | ||
5484965a | 1947 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1948 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1949 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1950 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1951 | return (error); |
1952 | } | |
428870ff | 1953 | } |
34dc7c2f | 1954 | |
428870ff BB |
1955 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1956 | vsecp, &acl_ids)) != 0) { | |
3558fd73 | 1957 | ZFS_EXIT(zsb); |
428870ff BB |
1958 | return (error); |
1959 | } | |
34dc7c2f BB |
1960 | /* |
1961 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1962 | * |
1963 | * Existence is checked first to make sure we don't return | |
1964 | * EACCES instead of EEXIST which can cause some applications | |
1965 | * to fail. | |
34dc7c2f BB |
1966 | */ |
1967 | top: | |
3558fd73 | 1968 | *ipp = NULL; |
34dc7c2f | 1969 | |
149e873a BB |
1970 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1971 | NULL, NULL))) { | |
428870ff | 1972 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 1973 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1974 | return (error); |
1975 | } | |
1976 | ||
149e873a | 1977 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 1978 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1979 | zfs_dirent_unlock(dl); |
3558fd73 | 1980 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1981 | return (error); |
1982 | } | |
1983 | ||
3558fd73 | 1984 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1985 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 1986 | zfs_dirent_unlock(dl); |
3558fd73 | 1987 | ZFS_EXIT(zsb); |
2e528b49 | 1988 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
1989 | } |
1990 | ||
34dc7c2f BB |
1991 | /* |
1992 | * Add a new entry to the directory. | |
1993 | */ | |
3558fd73 | 1994 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f BB |
1995 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
1996 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
3558fd73 | 1997 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1998 | if (fuid_dirtied) |
3558fd73 BB |
1999 | zfs_fuid_txhold(zsb, tx); |
2000 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
2001 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
2002 | acl_ids.z_aclp->z_acl_bytes); | |
2003 | } | |
2004 | ||
2005 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
2006 | ZFS_SA_BASE_ATTR_SIZE); | |
2007 | ||
e8b96c60 | 2008 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
2009 | if (error) { |
2010 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2011 | if (error == ERESTART) { |
e8b96c60 | 2012 | waited = B_TRUE; |
34dc7c2f BB |
2013 | dmu_tx_wait(tx); |
2014 | dmu_tx_abort(tx); | |
2015 | goto top; | |
2016 | } | |
428870ff | 2017 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2018 | dmu_tx_abort(tx); |
3558fd73 | 2019 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2020 | return (error); |
2021 | } | |
2022 | ||
2023 | /* | |
2024 | * Create new node. | |
2025 | */ | |
428870ff | 2026 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f | 2027 | |
9babb374 | 2028 | if (fuid_dirtied) |
3558fd73 | 2029 | zfs_fuid_sync(zsb, tx); |
428870ff | 2030 | |
34dc7c2f BB |
2031 | /* |
2032 | * Now put new name in parent dir. | |
2033 | */ | |
2034 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
2035 | ||
3558fd73 | 2036 | *ipp = ZTOI(zp); |
34dc7c2f BB |
2037 | |
2038 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
2039 | if (flags & FIGNORECASE) | |
2040 | txtype |= TX_CI; | |
9babb374 BB |
2041 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
2042 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 2043 | |
9babb374 | 2044 | zfs_acl_ids_free(&acl_ids); |
428870ff | 2045 | |
34dc7c2f BB |
2046 | dmu_tx_commit(tx); |
2047 | ||
2048 | zfs_dirent_unlock(dl); | |
2049 | ||
3558fd73 | 2050 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2051 | zil_commit(zilog, 0); |
428870ff | 2052 | |
960e08fe BB |
2053 | zfs_inode_update(dzp); |
2054 | zfs_inode_update(zp); | |
3558fd73 | 2055 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2056 | return (0); |
2057 | } | |
e5c39b95 | 2058 | EXPORT_SYMBOL(zfs_mkdir); |
34dc7c2f BB |
2059 | |
2060 | /* | |
2061 | * Remove a directory subdir entry. If the current working | |
2062 | * directory is the same as the subdir to be removed, the | |
2063 | * remove will fail. | |
2064 | * | |
3558fd73 | 2065 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 2066 | * name - name of directory to be removed. |
3558fd73 | 2067 | * cwd - inode of current working directory. |
34dc7c2f | 2068 | * cr - credentials of caller. |
34dc7c2f BB |
2069 | * flags - case flags |
2070 | * | |
d3cc8b15 | 2071 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
2072 | * |
2073 | * Timestamps: | |
3558fd73 | 2074 | * dip - ctime|mtime updated |
34dc7c2f BB |
2075 | */ |
2076 | /*ARGSUSED*/ | |
e5c39b95 | 2077 | int |
3558fd73 BB |
2078 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
2079 | int flags) | |
34dc7c2f | 2080 | { |
3558fd73 | 2081 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 2082 | znode_t *zp; |
3558fd73 BB |
2083 | struct inode *ip; |
2084 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
2085 | zilog_t *zilog; |
2086 | zfs_dirlock_t *dl; | |
2087 | dmu_tx_t *tx; | |
2088 | int error; | |
2089 | int zflg = ZEXISTS; | |
e8b96c60 | 2090 | boolean_t waited = B_FALSE; |
34dc7c2f | 2091 | |
32dec7bd | 2092 | if (name == NULL) |
2093 | return (SET_ERROR(EINVAL)); | |
2094 | ||
3558fd73 | 2095 | ZFS_ENTER(zsb); |
34dc7c2f | 2096 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 2097 | zilog = zsb->z_log; |
34dc7c2f BB |
2098 | |
2099 | if (flags & FIGNORECASE) | |
2100 | zflg |= ZCILOOK; | |
2101 | top: | |
2102 | zp = NULL; | |
2103 | ||
2104 | /* | |
2105 | * Attempt to lock directory; fail if entry doesn't exist. | |
2106 | */ | |
149e873a BB |
2107 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
2108 | NULL, NULL))) { | |
3558fd73 | 2109 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2110 | return (error); |
2111 | } | |
2112 | ||
3558fd73 | 2113 | ip = ZTOI(zp); |
34dc7c2f | 2114 | |
149e873a | 2115 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
2116 | goto out; |
2117 | } | |
2118 | ||
3558fd73 | 2119 | if (!S_ISDIR(ip->i_mode)) { |
2e528b49 | 2120 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
2121 | goto out; |
2122 | } | |
2123 | ||
3558fd73 | 2124 | if (ip == cwd) { |
2e528b49 | 2125 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
2126 | goto out; |
2127 | } | |
2128 | ||
34dc7c2f | 2129 | /* |
4e33ba4c | 2130 | * Grab a lock on the directory to make sure that no one is |
34dc7c2f BB |
2131 | * trying to add (or lookup) entries while we are removing it. |
2132 | */ | |
2133 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
2134 | ||
2135 | /* | |
2136 | * Grab a lock on the parent pointer to make sure we play well | |
2137 | * with the treewalk and directory rename code. | |
2138 | */ | |
2139 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
2140 | ||
3558fd73 | 2141 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 2142 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 2143 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3558fd73 | 2144 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
2145 | zfs_sa_upgrade_txholds(tx, zp); |
2146 | zfs_sa_upgrade_txholds(tx, dzp); | |
db707ad0 | 2147 | dmu_tx_mark_netfree(tx); |
e8b96c60 | 2148 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
2149 | if (error) { |
2150 | rw_exit(&zp->z_parent_lock); | |
2151 | rw_exit(&zp->z_name_lock); | |
2152 | zfs_dirent_unlock(dl); | |
3558fd73 | 2153 | iput(ip); |
fb5f0bc8 | 2154 | if (error == ERESTART) { |
e8b96c60 | 2155 | waited = B_TRUE; |
34dc7c2f BB |
2156 | dmu_tx_wait(tx); |
2157 | dmu_tx_abort(tx); | |
2158 | goto top; | |
2159 | } | |
2160 | dmu_tx_abort(tx); | |
3558fd73 | 2161 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2162 | return (error); |
2163 | } | |
2164 | ||
2165 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
2166 | ||
2167 | if (error == 0) { | |
2168 | uint64_t txtype = TX_RMDIR; | |
2169 | if (flags & FIGNORECASE) | |
2170 | txtype |= TX_CI; | |
572e2857 | 2171 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
2172 | } |
2173 | ||
2174 | dmu_tx_commit(tx); | |
2175 | ||
2176 | rw_exit(&zp->z_parent_lock); | |
2177 | rw_exit(&zp->z_name_lock); | |
2178 | out: | |
2179 | zfs_dirent_unlock(dl); | |
2180 | ||
59157910 BB |
2181 | zfs_inode_update(dzp); |
2182 | zfs_inode_update(zp); | |
3558fd73 | 2183 | iput(ip); |
34dc7c2f | 2184 | |
3558fd73 | 2185 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2186 | zil_commit(zilog, 0); |
428870ff | 2187 | |
3558fd73 | 2188 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2189 | return (error); |
2190 | } | |
e5c39b95 | 2191 | EXPORT_SYMBOL(zfs_rmdir); |
34dc7c2f BB |
2192 | |
2193 | /* | |
2194 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 2195 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 2196 | * |
3558fd73 BB |
2197 | * IN: ip - inode of directory to read. |
2198 | * dirent - buffer for directory entries. | |
34dc7c2f | 2199 | * |
3558fd73 | 2200 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
2201 | * |
2202 | * RETURN: 0 if success | |
2203 | * error code if failure | |
2204 | * | |
2205 | * Timestamps: | |
3558fd73 | 2206 | * ip - atime updated |
34dc7c2f BB |
2207 | * |
2208 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2209 | * This allows us to use the low range for "special" directory entries: | |
2210 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2211 | * we use the offset 2 for the '.zfs' directory. | |
2212 | */ | |
2213 | /* ARGSUSED */ | |
3558fd73 | 2214 | int |
0f37d0c8 | 2215 | zfs_readdir(struct inode *ip, struct dir_context *ctx, cred_t *cr) |
34dc7c2f | 2216 | { |
3558fd73 BB |
2217 | znode_t *zp = ITOZ(ip); |
2218 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 2219 | objset_t *os; |
34dc7c2f BB |
2220 | zap_cursor_t zc; |
2221 | zap_attribute_t zap; | |
34dc7c2f BB |
2222 | int error; |
2223 | uint8_t prefetch; | |
c12e3a59 | 2224 | uint8_t type; |
3558fd73 BB |
2225 | int done = 0; |
2226 | uint64_t parent; | |
c12e3a59 | 2227 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2228 | |
3558fd73 | 2229 | ZFS_ENTER(zsb); |
34dc7c2f BB |
2230 | ZFS_VERIFY_ZP(zp); |
2231 | ||
3558fd73 BB |
2232 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
2233 | &parent, sizeof (parent))) != 0) | |
2234 | goto out; | |
34dc7c2f BB |
2235 | |
2236 | /* | |
2237 | * Quit if directory has been removed (posix) | |
2238 | */ | |
3558fd73 BB |
2239 | if (zp->z_unlinked) |
2240 | goto out; | |
2241 | ||
c12e3a59 | 2242 | error = 0; |
3558fd73 | 2243 | os = zsb->z_os; |
c12e3a59 | 2244 | offset = ctx->pos; |
34dc7c2f BB |
2245 | prefetch = zp->z_zn_prefetch; |
2246 | ||
2247 | /* | |
2248 | * Initialize the iterator cursor. | |
2249 | */ | |
c12e3a59 | 2250 | if (offset <= 3) { |
34dc7c2f BB |
2251 | /* |
2252 | * Start iteration from the beginning of the directory. | |
2253 | */ | |
2254 | zap_cursor_init(&zc, os, zp->z_id); | |
2255 | } else { | |
2256 | /* | |
2257 | * The offset is a serialized cursor. | |
2258 | */ | |
c12e3a59 | 2259 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2260 | } |
2261 | ||
34dc7c2f BB |
2262 | /* |
2263 | * Transform to file-system independent format | |
2264 | */ | |
3558fd73 BB |
2265 | while (!done) { |
2266 | uint64_t objnum; | |
34dc7c2f BB |
2267 | /* |
2268 | * Special case `.', `..', and `.zfs'. | |
2269 | */ | |
c12e3a59 | 2270 | if (offset == 0) { |
34dc7c2f BB |
2271 | (void) strcpy(zap.za_name, "."); |
2272 | zap.za_normalization_conflict = 0; | |
2273 | objnum = zp->z_id; | |
c12e3a59 RY |
2274 | type = DT_DIR; |
2275 | } else if (offset == 1) { | |
34dc7c2f BB |
2276 | (void) strcpy(zap.za_name, ".."); |
2277 | zap.za_normalization_conflict = 0; | |
428870ff | 2278 | objnum = parent; |
c12e3a59 RY |
2279 | type = DT_DIR; |
2280 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2281 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2282 | zap.za_normalization_conflict = 0; | |
2283 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 2284 | type = DT_DIR; |
34dc7c2f BB |
2285 | } else { |
2286 | /* | |
2287 | * Grab next entry. | |
2288 | */ | |
3558fd73 BB |
2289 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2290 | if (error == ENOENT) | |
34dc7c2f BB |
2291 | break; |
2292 | else | |
2293 | goto update; | |
2294 | } | |
2295 | ||
0c5dde49 BB |
2296 | /* |
2297 | * Allow multiple entries provided the first entry is | |
2298 | * the object id. Non-zpl consumers may safely make | |
2299 | * use of the additional space. | |
2300 | * | |
2301 | * XXX: This should be a feature flag for compatibility | |
2302 | */ | |
34dc7c2f | 2303 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2304 | zap.za_num_integers == 0) { |
34dc7c2f | 2305 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2306 | "entry, obj = %lld, offset = %lld, " |
2307 | "length = %d, num = %lld\n", | |
34dc7c2f | 2308 | (u_longlong_t)zp->z_id, |
c12e3a59 | 2309 | (u_longlong_t)offset, |
0c5dde49 BB |
2310 | zap.za_integer_length, |
2311 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 2312 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2313 | goto update; |
2314 | } | |
2315 | ||
2316 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 2317 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2318 | } |
0f37d0c8 RY |
2319 | |
2320 | done = !dir_emit(ctx, zap.za_name, strlen(zap.za_name), | |
c12e3a59 | 2321 | objnum, type); |
0f37d0c8 | 2322 | if (done) |
34dc7c2f | 2323 | break; |
34dc7c2f BB |
2324 | |
2325 | /* Prefetch znode */ | |
3558fd73 | 2326 | if (prefetch) { |
fcff0f35 PD |
2327 | dmu_prefetch(os, objnum, 0, 0, 0, |
2328 | ZIO_PRIORITY_SYNC_READ); | |
3558fd73 | 2329 | } |
34dc7c2f | 2330 | |
c12e3a59 RY |
2331 | /* |
2332 | * Move to the next entry, fill in the previous offset. | |
2333 | */ | |
2334 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2335 | zap_cursor_advance(&zc); |
c12e3a59 | 2336 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2337 | } else { |
c12e3a59 | 2338 | offset += 1; |
34dc7c2f | 2339 | } |
c12e3a59 | 2340 | ctx->pos = offset; |
34dc7c2f BB |
2341 | } |
2342 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2343 | ||
34dc7c2f BB |
2344 | update: |
2345 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2346 | if (error == ENOENT) |
2347 | error = 0; | |
3558fd73 BB |
2348 | out: |
2349 | ZFS_EXIT(zsb); | |
34dc7c2f | 2350 | |
34dc7c2f BB |
2351 | return (error); |
2352 | } | |
3558fd73 | 2353 | EXPORT_SYMBOL(zfs_readdir); |
34dc7c2f | 2354 | |
d5446cfc BB |
2355 | ulong_t zfs_fsync_sync_cnt = 4; |
2356 | ||
e5c39b95 | 2357 | int |
3558fd73 | 2358 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2359 | { |
3558fd73 BB |
2360 | znode_t *zp = ITOZ(ip); |
2361 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 2362 | |
d5446cfc BB |
2363 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2364 | ||
3558fd73 BB |
2365 | if (zsb->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2366 | ZFS_ENTER(zsb); | |
428870ff | 2367 | ZFS_VERIFY_ZP(zp); |
3558fd73 BB |
2368 | zil_commit(zsb->z_log, zp->z_id); |
2369 | ZFS_EXIT(zsb); | |
428870ff | 2370 | } |
07012da6 CC |
2371 | tsd_set(zfs_fsyncer_key, NULL); |
2372 | ||
34dc7c2f BB |
2373 | return (0); |
2374 | } | |
e5c39b95 | 2375 | EXPORT_SYMBOL(zfs_fsync); |
34dc7c2f BB |
2376 | |
2377 | ||
2378 | /* | |
2379 | * Get the requested file attributes and place them in the provided | |
2380 | * vattr structure. | |
2381 | * | |
3558fd73 | 2382 | * IN: ip - inode of file. |
5484965a BB |
2383 | * vap - va_mask identifies requested attributes. |
2384 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2385 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2386 | * cr - credentials of caller. | |
34dc7c2f | 2387 | * |
5484965a BB |
2388 | * OUT: vap - attribute values. |
2389 | * | |
2390 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2391 | */ |
2392 | /* ARGSUSED */ | |
e5c39b95 | 2393 | int |
5484965a | 2394 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2395 | { |
3558fd73 BB |
2396 | znode_t *zp = ITOZ(ip); |
2397 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2398 | int error = 0; |
2399 | uint64_t links; | |
0df9673f | 2400 | uint64_t atime[2], mtime[2], ctime[2]; |
5484965a BB |
2401 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2402 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2403 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
0df9673f | 2404 | sa_bulk_attr_t bulk[3]; |
428870ff | 2405 | int count = 0; |
34dc7c2f | 2406 | |
3558fd73 | 2407 | ZFS_ENTER(zsb); |
34dc7c2f | 2408 | ZFS_VERIFY_ZP(zp); |
428870ff | 2409 | |
5484965a | 2410 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2411 | |
0df9673f | 2412 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, &atime, 16); |
3558fd73 BB |
2413 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
2414 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
428870ff BB |
2415 | |
2416 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
3558fd73 | 2417 | ZFS_EXIT(zsb); |
428870ff BB |
2418 | return (error); |
2419 | } | |
34dc7c2f | 2420 | |
34dc7c2f BB |
2421 | /* |
2422 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2423 | * Also, if we are the owner don't bother, since owner should | |
2424 | * always be allowed to read basic attributes of file. | |
2425 | */ | |
572e2857 | 2426 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2427 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2428 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2429 | skipaclchk, cr))) { | |
3558fd73 | 2430 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2431 | return (error); |
2432 | } | |
2433 | } | |
2434 | ||
2435 | /* | |
2436 | * Return all attributes. It's cheaper to provide the answer | |
2437 | * than to determine whether we were asked the question. | |
2438 | */ | |
2439 | ||
9babb374 | 2440 | mutex_enter(&zp->z_lock); |
5484965a BB |
2441 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2442 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2443 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2444 | vap->va_nodeid = zp->z_id; |
3558fd73 | 2445 | if ((zp->z_id == zsb->z_root) && zfs_show_ctldir(zp)) |
dfbc8630 | 2446 | links = ZTOI(zp)->i_nlink + 1; |
34dc7c2f | 2447 | else |
dfbc8630 | 2448 | links = ZTOI(zp)->i_nlink; |
5484965a BB |
2449 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2450 | vap->va_size = i_size_read(ip); | |
2451 | vap->va_rdev = ip->i_rdev; | |
2452 | vap->va_seq = ip->i_generation; | |
2453 | ||
2454 | /* | |
2455 | * Add in any requested optional attributes and the create time. | |
2456 | * Also set the corresponding bits in the returned attribute bitmap. | |
2457 | */ | |
2458 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zsb->z_use_fuids) { | |
2459 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
2460 | xoap->xoa_archive = | |
2461 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2462 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2463 | } | |
2464 | ||
2465 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2466 | xoap->xoa_readonly = | |
2467 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2468 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2469 | } | |
2470 | ||
2471 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2472 | xoap->xoa_system = | |
2473 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2474 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2475 | } | |
2476 | ||
2477 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2478 | xoap->xoa_hidden = | |
2479 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2480 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2481 | } | |
2482 | ||
2483 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2484 | xoap->xoa_nounlink = | |
2485 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2486 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2487 | } | |
2488 | ||
2489 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2490 | xoap->xoa_immutable = | |
2491 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2492 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2493 | } | |
2494 | ||
2495 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2496 | xoap->xoa_appendonly = | |
2497 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2498 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2499 | } | |
2500 | ||
2501 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2502 | xoap->xoa_nodump = | |
2503 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2504 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2505 | } | |
2506 | ||
2507 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2508 | xoap->xoa_opaque = | |
2509 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2510 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2511 | } | |
2512 | ||
2513 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2514 | xoap->xoa_av_quarantined = | |
2515 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2516 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2517 | } | |
2518 | ||
2519 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2520 | xoap->xoa_av_modified = | |
2521 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2522 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2523 | } | |
2524 | ||
2525 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2526 | S_ISREG(ip->i_mode)) { | |
2527 | zfs_sa_get_scanstamp(zp, xvap); | |
2528 | } | |
34dc7c2f | 2529 | |
5484965a BB |
2530 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2531 | uint64_t times[2]; | |
2532 | ||
2533 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zsb), | |
2534 | times, sizeof (times)); | |
2535 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2536 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2537 | } | |
2538 | ||
2539 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2540 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2541 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2542 | } | |
2543 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
278f2236 | 2544 | xoap->xoa_generation = ip->i_generation; |
5484965a BB |
2545 | XVA_SET_RTN(xvap, XAT_GEN); |
2546 | } | |
2547 | ||
2548 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2549 | xoap->xoa_offline = | |
2550 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2551 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2552 | } | |
2553 | ||
2554 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2555 | xoap->xoa_sparse = | |
2556 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2557 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2558 | } | |
2559 | } | |
2560 | ||
0df9673f | 2561 | ZFS_TIME_DECODE(&vap->va_atime, atime); |
5484965a BB |
2562 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); |
2563 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2564 | |
2565 | mutex_exit(&zp->z_lock); | |
2566 | ||
5484965a | 2567 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2568 | |
2569 | if (zp->z_blksz == 0) { | |
2570 | /* | |
2571 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2572 | */ | |
5484965a | 2573 | vap->va_blksize = zsb->z_max_blksz; |
34dc7c2f BB |
2574 | } |
2575 | ||
3558fd73 | 2576 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2577 | return (0); |
2578 | } | |
e5c39b95 | 2579 | EXPORT_SYMBOL(zfs_getattr); |
34dc7c2f | 2580 | |
057e8eee BB |
2581 | /* |
2582 | * Get the basic file attributes and place them in the provided kstat | |
2583 | * structure. The inode is assumed to be the authoritative source | |
2584 | * for most of the attributes. However, the znode currently has the | |
2585 | * authoritative atime, blksize, and block count. | |
2586 | * | |
2587 | * IN: ip - inode of file. | |
2588 | * | |
2589 | * OUT: sp - kstat values. | |
2590 | * | |
2591 | * RETURN: 0 (always succeeds) | |
2592 | */ | |
2593 | /* ARGSUSED */ | |
2594 | int | |
2595 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2596 | { | |
2597 | znode_t *zp = ITOZ(ip); | |
2598 | zfs_sb_t *zsb = ITOZSB(ip); | |
b585bc4a BB |
2599 | uint32_t blksize; |
2600 | u_longlong_t nblocks; | |
057e8eee | 2601 | |
a7b125e9 GB |
2602 | ZFS_ENTER(zsb); |
2603 | ZFS_VERIFY_ZP(zp); | |
2604 | ||
057e8eee BB |
2605 | mutex_enter(&zp->z_lock); |
2606 | ||
2607 | generic_fillattr(ip, sp); | |
057e8eee | 2608 | |
b585bc4a BB |
2609 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2610 | sp->blksize = blksize; | |
2611 | sp->blocks = nblocks; | |
2612 | ||
057e8eee BB |
2613 | if (unlikely(zp->z_blksz == 0)) { |
2614 | /* | |
2615 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2616 | */ | |
2617 | sp->blksize = zsb->z_max_blksz; | |
2618 | } | |
2619 | ||
2620 | mutex_exit(&zp->z_lock); | |
2621 | ||
aa9b2708 AV |
2622 | /* |
2623 | * Required to prevent NFS client from detecting different inode | |
2624 | * numbers of snapshot root dentry before and after snapshot mount. | |
2625 | */ | |
2626 | if (zsb->z_issnap) { | |
2627 | if (ip->i_sb->s_root->d_inode == ip) | |
2628 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
02730c33 | 2629 | dmu_objset_id(zsb->z_os); |
aa9b2708 AV |
2630 | } |
2631 | ||
a7b125e9 GB |
2632 | ZFS_EXIT(zsb); |
2633 | ||
057e8eee BB |
2634 | return (0); |
2635 | } | |
2636 | EXPORT_SYMBOL(zfs_getattr_fast); | |
2637 | ||
34dc7c2f BB |
2638 | /* |
2639 | * Set the file attributes to the values contained in the | |
2640 | * vattr structure. | |
2641 | * | |
3558fd73 | 2642 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2643 | * vap - new attribute values. |
5484965a | 2644 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2645 | * flags - ATTR_UTIME set if non-default time values provided. |
2646 | * - ATTR_NOACLCHECK (CIFS context only). | |
2647 | * cr - credentials of caller. | |
34dc7c2f BB |
2648 | * |
2649 | * RETURN: 0 if success | |
2650 | * error code if failure | |
2651 | * | |
2652 | * Timestamps: | |
3558fd73 | 2653 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2654 | */ |
2655 | /* ARGSUSED */ | |
e5c39b95 | 2656 | int |
5484965a | 2657 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2658 | { |
3558fd73 BB |
2659 | znode_t *zp = ITOZ(ip); |
2660 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2661 | zilog_t *zilog; |
2662 | dmu_tx_t *tx; | |
2663 | vattr_t oldva; | |
f4ea75d4 | 2664 | xvattr_t *tmpxvattr; |
5484965a | 2665 | uint_t mask = vap->va_mask; |
a117a6d6 | 2666 | uint_t saved_mask = 0; |
34dc7c2f BB |
2667 | int trim_mask = 0; |
2668 | uint64_t new_mode; | |
64aefee1 | 2669 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 2670 | uint64_t xattr_obj; |
0df9673f | 2671 | uint64_t mtime[2], ctime[2], atime[2]; |
34dc7c2f BB |
2672 | znode_t *attrzp; |
2673 | int need_policy = FALSE; | |
428870ff | 2674 | int err, err2; |
34dc7c2f | 2675 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2676 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2677 | xoptattr_t *xoap; | |
2678 | zfs_acl_t *aclp; | |
34dc7c2f | 2679 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2680 | boolean_t fuid_dirtied = B_FALSE; |
17c37660 | 2681 | sa_bulk_attr_t *bulk, *xattr_bulk; |
428870ff | 2682 | int count = 0, xattr_count = 0; |
34dc7c2f BB |
2683 | |
2684 | if (mask == 0) | |
2685 | return (0); | |
2686 | ||
3558fd73 | 2687 | ZFS_ENTER(zsb); |
34dc7c2f BB |
2688 | ZFS_VERIFY_ZP(zp); |
2689 | ||
3558fd73 | 2690 | zilog = zsb->z_log; |
34dc7c2f BB |
2691 | |
2692 | /* | |
2693 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2694 | * that file system is at proper version level | |
2695 | */ | |
5484965a | 2696 | |
3558fd73 | 2697 | if (zsb->z_use_fuids == B_FALSE && |
5484965a BB |
2698 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2699 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2700 | (mask & ATTR_XVATTR))) { | |
3558fd73 | 2701 | ZFS_EXIT(zsb); |
2e528b49 | 2702 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2703 | } |
2704 | ||
3558fd73 BB |
2705 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
2706 | ZFS_EXIT(zsb); | |
2e528b49 | 2707 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2708 | } |
2709 | ||
3558fd73 BB |
2710 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
2711 | ZFS_EXIT(zsb); | |
2e528b49 | 2712 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2713 | } |
2714 | ||
5484965a BB |
2715 | /* |
2716 | * If this is an xvattr_t, then get a pointer to the structure of | |
2717 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2718 | */ | |
2719 | xoap = xva_getxoptattr(xvap); | |
2720 | ||
d1d7e268 | 2721 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2722 | xva_init(tmpxvattr); |
5484965a | 2723 | |
d1d7e268 MK |
2724 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); |
2725 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); | |
17c37660 | 2726 | |
5484965a BB |
2727 | /* |
2728 | * Immutable files can only alter immutable bit and atime | |
2729 | */ | |
2730 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2731 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2732 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
f4ea75d4 BB |
2733 | err = EPERM; |
2734 | goto out3; | |
5484965a BB |
2735 | } |
2736 | ||
3558fd73 | 2737 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
f4ea75d4 BB |
2738 | err = EPERM; |
2739 | goto out3; | |
34dc7c2f BB |
2740 | } |
2741 | ||
5484965a BB |
2742 | /* |
2743 | * Verify timestamps doesn't overflow 32 bits. | |
2744 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2745 | * handle times greater than 2039. This check should be removed | |
2746 | * once large timestamps are fully supported. | |
2747 | */ | |
2748 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
2749 | if (((mask & ATTR_ATIME) && |
2750 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2751 | ((mask & ATTR_MTIME) && | |
2752 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
f4ea75d4 BB |
2753 | err = EOVERFLOW; |
2754 | goto out3; | |
5484965a BB |
2755 | } |
2756 | } | |
2757 | ||
34dc7c2f BB |
2758 | top: |
2759 | attrzp = NULL; | |
572e2857 | 2760 | aclp = NULL; |
34dc7c2f | 2761 | |
45d1cae3 | 2762 | /* Can this be moved to before the top label? */ |
2cf7f52b | 2763 | if (zfs_is_readonly(zsb)) { |
f4ea75d4 BB |
2764 | err = EROFS; |
2765 | goto out3; | |
34dc7c2f BB |
2766 | } |
2767 | ||
2768 | /* | |
2769 | * First validate permissions | |
2770 | */ | |
2771 | ||
3558fd73 | 2772 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2773 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2774 | if (err) |
2775 | goto out3; | |
2776 | ||
34dc7c2f BB |
2777 | /* |
2778 | * XXX - Note, we are not providing any open | |
2779 | * mode flags here (like FNDELAY), so we may | |
2780 | * block if there are locks present... this | |
2781 | * should be addressed in openat(). | |
2782 | */ | |
b128c09f | 2783 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2784 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2785 | if (err) |
2786 | goto out3; | |
428870ff | 2787 | } |
34dc7c2f | 2788 | |
5484965a BB |
2789 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2790 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2791 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2792 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2793 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2794 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2795 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2796 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2797 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2798 | skipaclchk, cr); | |
2799 | } | |
2800 | ||
3558fd73 BB |
2801 | if (mask & (ATTR_UID|ATTR_GID)) { |
2802 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2803 | int take_owner; |
2804 | int take_group; | |
2805 | ||
2806 | /* | |
2807 | * NOTE: even if a new mode is being set, | |
2808 | * we may clear S_ISUID/S_ISGID bits. | |
2809 | */ | |
2810 | ||
3558fd73 | 2811 | if (!(mask & ATTR_MODE)) |
5484965a | 2812 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2813 | |
2814 | /* | |
2815 | * Take ownership or chgrp to group we are a member of | |
2816 | */ | |
2817 | ||
5484965a | 2818 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 2819 | take_group = (mask & ATTR_GID) && |
5484965a | 2820 | zfs_groupmember(zsb, vap->va_gid, cr); |
34dc7c2f BB |
2821 | |
2822 | /* | |
5484965a | 2823 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
2824 | * take_group must both be set in order to allow taking |
2825 | * ownership. | |
2826 | * | |
2827 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2828 | * | |
2829 | */ | |
2830 | ||
3558fd73 BB |
2831 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
2832 | take_owner && take_group) || | |
2833 | ((idmask == ATTR_UID) && take_owner) || | |
2834 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
2835 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
2836 | skipaclchk, cr) == 0) { | |
2837 | /* | |
2838 | * Remove setuid/setgid for non-privileged users | |
2839 | */ | |
5484965a | 2840 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 2841 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
2842 | } else { |
2843 | need_policy = TRUE; | |
2844 | } | |
2845 | } else { | |
2846 | need_policy = TRUE; | |
2847 | } | |
2848 | } | |
2849 | ||
2850 | mutex_enter(&zp->z_lock); | |
428870ff | 2851 | oldva.va_mode = zp->z_mode; |
572e2857 | 2852 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
2853 | if (mask & ATTR_XVATTR) { |
2854 | /* | |
2855 | * Update xvattr mask to include only those attributes | |
2856 | * that are actually changing. | |
2857 | * | |
2858 | * the bits will be restored prior to actually setting | |
2859 | * the attributes so the caller thinks they were set. | |
2860 | */ | |
2861 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2862 | if (xoap->xoa_appendonly != | |
2863 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
2864 | need_policy = TRUE; | |
2865 | } else { | |
2866 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 2867 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
2868 | } |
2869 | } | |
2870 | ||
2871 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2872 | if (xoap->xoa_nounlink != | |
2873 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
2874 | need_policy = TRUE; | |
2875 | } else { | |
2876 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 2877 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
2878 | } |
2879 | } | |
2880 | ||
2881 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2882 | if (xoap->xoa_immutable != | |
2883 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
2884 | need_policy = TRUE; | |
2885 | } else { | |
2886 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 2887 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
2888 | } |
2889 | } | |
2890 | ||
2891 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2892 | if (xoap->xoa_nodump != | |
2893 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
2894 | need_policy = TRUE; | |
2895 | } else { | |
2896 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 2897 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
2898 | } |
2899 | } | |
2900 | ||
2901 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2902 | if (xoap->xoa_av_modified != | |
2903 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
2904 | need_policy = TRUE; | |
2905 | } else { | |
2906 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 2907 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
2908 | } |
2909 | } | |
2910 | ||
2911 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2912 | if ((!S_ISREG(ip->i_mode) && | |
2913 | xoap->xoa_av_quarantined) || | |
2914 | xoap->xoa_av_quarantined != | |
2915 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
2916 | need_policy = TRUE; | |
2917 | } else { | |
2918 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 2919 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
2920 | } |
2921 | } | |
2922 | ||
2923 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2924 | mutex_exit(&zp->z_lock); | |
f4ea75d4 BB |
2925 | err = EPERM; |
2926 | goto out3; | |
5484965a BB |
2927 | } |
2928 | ||
2929 | if (need_policy == FALSE && | |
2930 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2931 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
2932 | need_policy = TRUE; | |
2933 | } | |
2934 | } | |
34dc7c2f BB |
2935 | |
2936 | mutex_exit(&zp->z_lock); | |
2937 | ||
3558fd73 | 2938 | if (mask & ATTR_MODE) { |
34dc7c2f | 2939 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 2940 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 2941 | &oldva, cr); |
f4ea75d4 BB |
2942 | if (err) |
2943 | goto out3; | |
2944 | ||
3558fd73 | 2945 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
2946 | } else { |
2947 | need_policy = TRUE; | |
2948 | } | |
2949 | } | |
2950 | ||
2951 | if (need_policy) { | |
2952 | /* | |
2953 | * If trim_mask is set then take ownership | |
2954 | * has been granted or write_acl is present and user | |
2955 | * has the ability to modify mode. In that case remove | |
2956 | * UID|GID and or MODE from mask so that | |
2957 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2958 | */ | |
2959 | ||
2960 | if (trim_mask) { | |
5484965a BB |
2961 | saved_mask = vap->va_mask; |
2962 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 2963 | } |
5484965a | 2964 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 2965 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
2966 | if (err) |
2967 | goto out3; | |
34dc7c2f BB |
2968 | |
2969 | if (trim_mask) | |
5484965a | 2970 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
2971 | } |
2972 | ||
2973 | /* | |
2974 | * secpolicy_vnode_setattr, or take ownership may have | |
2975 | * changed va_mask | |
2976 | */ | |
5484965a | 2977 | mask = vap->va_mask; |
34dc7c2f | 2978 | |
3558fd73 BB |
2979 | if ((mask & (ATTR_UID | ATTR_GID))) { |
2980 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), | |
572e2857 | 2981 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 2982 | |
572e2857 | 2983 | if (err == 0 && xattr_obj) { |
3558fd73 | 2984 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
2985 | if (err) |
2986 | goto out2; | |
2987 | } | |
3558fd73 | 2988 | if (mask & ATTR_UID) { |
64aefee1 | 2989 | new_kuid = zfs_fuid_create(zsb, |
5484965a | 2990 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
64aefee1 NB |
2991 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
2992 | zfs_fuid_overquota(zsb, B_FALSE, new_kuid)) { | |
572e2857 | 2993 | if (attrzp) |
3558fd73 | 2994 | iput(ZTOI(attrzp)); |
428870ff BB |
2995 | err = EDQUOT; |
2996 | goto out2; | |
2997 | } | |
2998 | } | |
2999 | ||
3558fd73 | 3000 | if (mask & ATTR_GID) { |
64aefee1 | 3001 | new_kgid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid, |
428870ff | 3002 | cr, ZFS_GROUP, &fuidp); |
64aefee1 NB |
3003 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && |
3004 | zfs_fuid_overquota(zsb, B_TRUE, new_kgid)) { | |
572e2857 | 3005 | if (attrzp) |
3558fd73 | 3006 | iput(ZTOI(attrzp)); |
428870ff BB |
3007 | err = EDQUOT; |
3008 | goto out2; | |
3009 | } | |
3010 | } | |
3011 | } | |
3558fd73 | 3012 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 3013 | |
3558fd73 | 3014 | if (mask & ATTR_MODE) { |
428870ff | 3015 | uint64_t pmode = zp->z_mode; |
572e2857 | 3016 | uint64_t acl_obj; |
5484965a | 3017 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 3018 | |
572e2857 | 3019 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 3020 | |
572e2857 BB |
3021 | mutex_enter(&zp->z_lock); |
3022 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
3023 | /* |
3024 | * Are we upgrading ACL from old V0 format | |
3025 | * to V1 format? | |
3026 | */ | |
3558fd73 | 3027 | if (zsb->z_version >= ZPL_VERSION_FUID && |
572e2857 | 3028 | zfs_znode_acl_version(zp) == |
34dc7c2f | 3029 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 3030 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
3031 | DMU_OBJECT_END); |
3032 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
3033 | 0, aclp->z_acl_bytes); | |
3034 | } else { | |
572e2857 | 3035 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
3036 | aclp->z_acl_bytes); |
3037 | } | |
428870ff | 3038 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
3039 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
3040 | 0, aclp->z_acl_bytes); | |
3041 | } | |
572e2857 | 3042 | mutex_exit(&zp->z_lock); |
428870ff BB |
3043 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3044 | } else { | |
5484965a BB |
3045 | if ((mask & ATTR_XVATTR) && |
3046 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3047 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
3048 | else | |
3049 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3050 | } |
3051 | ||
428870ff BB |
3052 | if (attrzp) { |
3053 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3054 | } |
3055 | ||
3558fd73 | 3056 | fuid_dirtied = zsb->z_fuid_dirty; |
428870ff | 3057 | if (fuid_dirtied) |
3558fd73 | 3058 | zfs_fuid_txhold(zsb, tx); |
428870ff BB |
3059 | |
3060 | zfs_sa_upgrade_txholds(tx, zp); | |
3061 | ||
384f8a09 MA |
3062 | err = dmu_tx_assign(tx, TXG_WAIT); |
3063 | if (err) | |
9babb374 | 3064 | goto out; |
34dc7c2f | 3065 | |
428870ff | 3066 | count = 0; |
34dc7c2f BB |
3067 | /* |
3068 | * Set each attribute requested. | |
3069 | * We group settings according to the locks they need to acquire. | |
3070 | * | |
3071 | * Note: you cannot set ctime directly, although it will be | |
3072 | * updated as a side-effect of calling this function. | |
3073 | */ | |
3074 | ||
572e2857 | 3075 | |
3558fd73 | 3076 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3077 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
3078 | mutex_enter(&zp->z_lock); |
3079 | ||
3558fd73 | 3080 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff BB |
3081 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3082 | ||
3083 | if (attrzp) { | |
3558fd73 | 3084 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3085 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
3086 | mutex_enter(&attrzp->z_lock); |
3087 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 3088 | SA_ZPL_FLAGS(zsb), NULL, &attrzp->z_pflags, |
428870ff BB |
3089 | sizeof (attrzp->z_pflags)); |
3090 | } | |
3091 | ||
3558fd73 | 3092 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 3093 | |
3558fd73 | 3094 | if (mask & ATTR_UID) { |
64aefee1 NB |
3095 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
3096 | new_uid = zfs_uid_read(ZTOI(zp)); | |
3558fd73 | 3097 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, |
428870ff | 3098 | &new_uid, sizeof (new_uid)); |
428870ff BB |
3099 | if (attrzp) { |
3100 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 3101 | SA_ZPL_UID(zsb), NULL, &new_uid, |
428870ff | 3102 | sizeof (new_uid)); |
2c6abf15 | 3103 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
3104 | } |
3105 | } | |
3106 | ||
3558fd73 | 3107 | if (mask & ATTR_GID) { |
64aefee1 NB |
3108 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
3109 | new_gid = zfs_gid_read(ZTOI(zp)); | |
3558fd73 | 3110 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), |
428870ff | 3111 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
3112 | if (attrzp) { |
3113 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 3114 | SA_ZPL_GID(zsb), NULL, &new_gid, |
428870ff | 3115 | sizeof (new_gid)); |
64aefee1 | 3116 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
3117 | } |
3118 | } | |
3558fd73 BB |
3119 | if (!(mask & ATTR_MODE)) { |
3120 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), | |
428870ff BB |
3121 | NULL, &new_mode, sizeof (new_mode)); |
3122 | new_mode = zp->z_mode; | |
3123 | } | |
3124 | err = zfs_acl_chown_setattr(zp); | |
3125 | ASSERT(err == 0); | |
3126 | if (attrzp) { | |
3127 | err = zfs_acl_chown_setattr(attrzp); | |
3128 | ASSERT(err == 0); | |
3129 | } | |
3130 | } | |
3131 | ||
3558fd73 BB |
3132 | if (mask & ATTR_MODE) { |
3133 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, | |
428870ff | 3134 | &new_mode, sizeof (new_mode)); |
12fa7f34 | 3135 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 3136 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 3137 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 3138 | ASSERT0(err); |
572e2857 BB |
3139 | if (zp->z_acl_cached) |
3140 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
3141 | zp->z_acl_cached = aclp; |
3142 | aclp = NULL; | |
34dc7c2f BB |
3143 | } |
3144 | ||
704cd075 CC |
3145 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
3146 | zp->z_atime_dirty = 0; | |
3147 | ZFS_TIME_ENCODE(&ip->i_atime, atime); | |
3558fd73 | 3148 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, |
0df9673f | 3149 | &atime, sizeof (atime)); |
34dc7c2f BB |
3150 | } |
3151 | ||
3558fd73 | 3152 | if (mask & ATTR_MTIME) { |
5484965a | 3153 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
87f9371a NB |
3154 | ZTOI(zp)->i_mtime = timespec_trunc(vap->va_mtime, |
3155 | ZTOI(zp)->i_sb->s_time_gran); | |
3156 | ||
3558fd73 | 3157 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, |
428870ff | 3158 | mtime, sizeof (mtime)); |
34dc7c2f BB |
3159 | } |
3160 | ||
87f9371a NB |
3161 | if (mask & ATTR_CTIME) { |
3162 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); | |
3163 | ZTOI(zp)->i_ctime = timespec_trunc(vap->va_ctime, | |
3164 | ZTOI(zp)->i_sb->s_time_gran); | |
3558fd73 | 3165 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
87f9371a | 3166 | ctime, sizeof (ctime)); |
428870ff | 3167 | } |
87f9371a NB |
3168 | |
3169 | if (attrzp && mask) { | |
3170 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3171 | SA_ZPL_CTIME(zsb), NULL, &ctime, | |
3172 | sizeof (ctime)); | |
3173 | } | |
3174 | ||
34dc7c2f BB |
3175 | /* |
3176 | * Do this after setting timestamps to prevent timestamp | |
3177 | * update from toggling bit | |
3178 | */ | |
3179 | ||
5484965a BB |
3180 | if (xoap && (mask & ATTR_XVATTR)) { |
3181 | ||
3182 | /* | |
3183 | * restore trimmed off masks | |
3184 | * so that return masks can be set for caller. | |
3185 | */ | |
3186 | ||
f4ea75d4 | 3187 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
3188 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
3189 | } | |
f4ea75d4 | 3190 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
3191 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
3192 | } | |
f4ea75d4 | 3193 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
3194 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
3195 | } | |
f4ea75d4 | 3196 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
3197 | XVA_SET_REQ(xvap, XAT_NODUMP); |
3198 | } | |
f4ea75d4 | 3199 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
3200 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
3201 | } | |
f4ea75d4 | 3202 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
3203 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
3204 | } | |
3205 | ||
3206 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3207 | ASSERT(S_ISREG(ip->i_mode)); | |
3208 | ||
3209 | zfs_xvattr_set(zp, xvap, tx); | |
3210 | } | |
3211 | ||
9babb374 | 3212 | if (fuid_dirtied) |
3558fd73 | 3213 | zfs_fuid_sync(zsb, tx); |
9babb374 | 3214 | |
34dc7c2f | 3215 | if (mask != 0) |
5484965a | 3216 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3217 | |
34dc7c2f | 3218 | mutex_exit(&zp->z_lock); |
3558fd73 | 3219 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3220 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3221 | |
572e2857 | 3222 | if (attrzp) { |
3558fd73 | 3223 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3224 | mutex_exit(&attrzp->z_acl_lock); |
3225 | mutex_exit(&attrzp->z_lock); | |
3226 | } | |
9babb374 | 3227 | out: |
428870ff BB |
3228 | if (err == 0 && attrzp) { |
3229 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, | |
3230 | xattr_count, tx); | |
3231 | ASSERT(err2 == 0); | |
3232 | } | |
3233 | ||
34dc7c2f | 3234 | if (attrzp) |
3558fd73 | 3235 | iput(ZTOI(attrzp)); |
45d1cae3 | 3236 | if (aclp) |
9babb374 | 3237 | zfs_acl_free(aclp); |
9babb374 BB |
3238 | |
3239 | if (fuidp) { | |
3240 | zfs_fuid_info_free(fuidp); | |
3241 | fuidp = NULL; | |
3242 | } | |
3243 | ||
428870ff | 3244 | if (err) { |
9babb374 | 3245 | dmu_tx_abort(tx); |
428870ff BB |
3246 | if (err == ERESTART) |
3247 | goto top; | |
3248 | } else { | |
3249 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3250 | dmu_tx_commit(tx); |
037849f8 | 3251 | zfs_inode_update(zp); |
428870ff BB |
3252 | } |
3253 | ||
428870ff | 3254 | out2: |
3558fd73 | 3255 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3256 | zil_commit(zilog, 0); |
34dc7c2f | 3257 | |
f4ea75d4 | 3258 | out3: |
d1d7e268 MK |
3259 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * 7); |
3260 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * 7); | |
3261 | kmem_free(tmpxvattr, sizeof (xvattr_t)); | |
3558fd73 | 3262 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3263 | return (err); |
3264 | } | |
e5c39b95 | 3265 | EXPORT_SYMBOL(zfs_setattr); |
34dc7c2f BB |
3266 | |
3267 | typedef struct zfs_zlock { | |
3268 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3269 | znode_t *zl_znode; /* znode we held */ | |
3270 | struct zfs_zlock *zl_next; /* next in list */ | |
3271 | } zfs_zlock_t; | |
3272 | ||
3273 | /* | |
3274 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3275 | */ | |
3276 | static void | |
3277 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3278 | { | |
3279 | zfs_zlock_t *zl; | |
3280 | ||
3281 | while ((zl = *zlpp) != NULL) { | |
3282 | if (zl->zl_znode != NULL) | |
3558fd73 | 3283 | iput(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3284 | rw_exit(zl->zl_rwlock); |
3285 | *zlpp = zl->zl_next; | |
3286 | kmem_free(zl, sizeof (*zl)); | |
3287 | } | |
3288 | } | |
3289 | ||
3290 | /* | |
3291 | * Search back through the directory tree, using the ".." entries. | |
3292 | * Lock each directory in the chain to prevent concurrent renames. | |
3293 | * Fail any attempt to move a directory into one of its own descendants. | |
3294 | * XXX - z_parent_lock can overlap with map or grow locks | |
3295 | */ | |
3296 | static int | |
3297 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3298 | { | |
3299 | zfs_zlock_t *zl; | |
3300 | znode_t *zp = tdzp; | |
3558fd73 | 3301 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3302 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3303 | krwlock_t *rwlp = &szp->z_parent_lock; |
3304 | krw_t rw = RW_WRITER; | |
3305 | ||
3306 | /* | |
3307 | * First pass write-locks szp and compares to zp->z_id. | |
3308 | * Later passes read-lock zp and compare to zp->z_parent. | |
3309 | */ | |
3310 | do { | |
3311 | if (!rw_tryenter(rwlp, rw)) { | |
3312 | /* | |
3313 | * Another thread is renaming in this path. | |
3314 | * Note that if we are a WRITER, we don't have any | |
3315 | * parent_locks held yet. | |
3316 | */ | |
3317 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3318 | /* | |
3319 | * Drop our locks and restart | |
3320 | */ | |
3321 | zfs_rename_unlock(&zl); | |
3322 | *zlpp = NULL; | |
3323 | zp = tdzp; | |
428870ff | 3324 | oidp = zp->z_id; |
34dc7c2f BB |
3325 | rwlp = &szp->z_parent_lock; |
3326 | rw = RW_WRITER; | |
3327 | continue; | |
3328 | } else { | |
3329 | /* | |
3330 | * Wait for other thread to drop its locks | |
3331 | */ | |
3332 | rw_enter(rwlp, rw); | |
3333 | } | |
3334 | } | |
3335 | ||
3336 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3337 | zl->zl_rwlock = rwlp; | |
3338 | zl->zl_znode = NULL; | |
3339 | zl->zl_next = *zlpp; | |
3340 | *zlpp = zl; | |
3341 | ||
428870ff | 3342 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 3343 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3344 | |
428870ff | 3345 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3346 | return (0); |
3347 | ||
3348 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3349 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3350 | if (error) |
3351 | return (error); | |
3352 | zl->zl_znode = zp; | |
3353 | } | |
3558fd73 | 3354 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3355 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3356 | rwlp = &zp->z_parent_lock; |
3357 | rw = RW_READER; | |
3358 | ||
3359 | } while (zp->z_id != sdzp->z_id); | |
3360 | ||
3361 | return (0); | |
3362 | } | |
3363 | ||
3364 | /* | |
3365 | * Move an entry from the provided source directory to the target | |
3366 | * directory. Change the entry name as indicated. | |
3367 | * | |
3558fd73 | 3368 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3369 | * snm - Old entry name. |
3558fd73 | 3370 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3371 | * tnm - New entry name. |
3372 | * cr - credentials of caller. | |
34dc7c2f BB |
3373 | * flags - case flags |
3374 | * | |
d3cc8b15 | 3375 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3376 | * |
3377 | * Timestamps: | |
3558fd73 | 3378 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3379 | */ |
3380 | /*ARGSUSED*/ | |
e5c39b95 | 3381 | int |
3558fd73 BB |
3382 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3383 | cred_t *cr, int flags) | |
34dc7c2f BB |
3384 | { |
3385 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 BB |
3386 | znode_t *sdzp = ITOZ(sdip); |
3387 | zfs_sb_t *zsb = ITOZSB(sdip); | |
34dc7c2f | 3388 | zilog_t *zilog; |
34dc7c2f BB |
3389 | zfs_dirlock_t *sdl, *tdl; |
3390 | dmu_tx_t *tx; | |
3391 | zfs_zlock_t *zl; | |
3392 | int cmp, serr, terr; | |
3393 | int error = 0; | |
3394 | int zflg = 0; | |
e8b96c60 | 3395 | boolean_t waited = B_FALSE; |
34dc7c2f | 3396 | |
32dec7bd | 3397 | if (snm == NULL || tnm == NULL) |
3398 | return (SET_ERROR(EINVAL)); | |
3399 | ||
3558fd73 | 3400 | ZFS_ENTER(zsb); |
34dc7c2f | 3401 | ZFS_VERIFY_ZP(sdzp); |
3558fd73 | 3402 | zilog = zsb->z_log; |
34dc7c2f | 3403 | |
812e91a7 MT |
3404 | tdzp = ITOZ(tdip); |
3405 | ZFS_VERIFY_ZP(tdzp); | |
3406 | ||
3407 | /* | |
3408 | * We check i_sb because snapshots and the ctldir must have different | |
3409 | * super blocks. | |
3410 | */ | |
c0ebc844 | 3411 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
3558fd73 | 3412 | ZFS_EXIT(zsb); |
2e528b49 | 3413 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3414 | } |
3415 | ||
3558fd73 | 3416 | if (zsb->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3417 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3418 | ZFS_EXIT(zsb); |
2e528b49 | 3419 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3420 | } |
3421 | ||
3422 | if (flags & FIGNORECASE) | |
3423 | zflg |= ZCILOOK; | |
3424 | ||
3425 | top: | |
3426 | szp = NULL; | |
3427 | tzp = NULL; | |
3428 | zl = NULL; | |
3429 | ||
3430 | /* | |
3431 | * This is to prevent the creation of links into attribute space | |
3432 | * by renaming a linked file into/outof an attribute directory. | |
3433 | * See the comment in zfs_link() for why this is considered bad. | |
3434 | */ | |
428870ff | 3435 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 3436 | ZFS_EXIT(zsb); |
2e528b49 | 3437 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3438 | } |
3439 | ||
3440 | /* | |
3441 | * Lock source and target directory entries. To prevent deadlock, | |
3442 | * a lock ordering must be defined. We lock the directory with | |
3443 | * the smallest object id first, or if it's a tie, the one with | |
3444 | * the lexically first name. | |
3445 | */ | |
3446 | if (sdzp->z_id < tdzp->z_id) { | |
3447 | cmp = -1; | |
3448 | } else if (sdzp->z_id > tdzp->z_id) { | |
3449 | cmp = 1; | |
3450 | } else { | |
3451 | /* | |
3452 | * First compare the two name arguments without | |
3453 | * considering any case folding. | |
3454 | */ | |
3558fd73 | 3455 | int nofold = (zsb->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3456 | |
3457 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
3558fd73 | 3458 | ASSERT(error == 0 || !zsb->z_utf8); |
34dc7c2f BB |
3459 | if (cmp == 0) { |
3460 | /* | |
3461 | * POSIX: "If the old argument and the new argument | |
3462 | * both refer to links to the same existing file, | |
3463 | * the rename() function shall return successfully | |
3464 | * and perform no other action." | |
3465 | */ | |
3558fd73 | 3466 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3467 | return (0); |
3468 | } | |
3469 | /* | |
3470 | * If the file system is case-folding, then we may | |
3471 | * have some more checking to do. A case-folding file | |
3472 | * system is either supporting mixed case sensitivity | |
3473 | * access or is completely case-insensitive. Note | |
3474 | * that the file system is always case preserving. | |
3475 | * | |
3476 | * In mixed sensitivity mode case sensitive behavior | |
3477 | * is the default. FIGNORECASE must be used to | |
3478 | * explicitly request case insensitive behavior. | |
3479 | * | |
3480 | * If the source and target names provided differ only | |
3481 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3482 | * we will treat this as a special case in the | |
3483 | * case-insensitive mode: as long as the source name | |
3484 | * is an exact match, we will allow this to proceed as | |
3485 | * a name-change request. | |
3486 | */ | |
3558fd73 BB |
3487 | if ((zsb->z_case == ZFS_CASE_INSENSITIVE || |
3488 | (zsb->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3489 | flags & FIGNORECASE)) && |
3558fd73 | 3490 | u8_strcmp(snm, tnm, 0, zsb->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3491 | &error) == 0) { |
3492 | /* | |
3493 | * case preserving rename request, require exact | |
3494 | * name matches | |
3495 | */ | |
3496 | zflg |= ZCIEXACT; | |
3497 | zflg &= ~ZCILOOK; | |
3498 | } | |
3499 | } | |
3500 | ||
428870ff BB |
3501 | /* |
3502 | * If the source and destination directories are the same, we should | |
3503 | * grab the z_name_lock of that directory only once. | |
3504 | */ | |
3505 | if (sdzp == tdzp) { | |
3506 | zflg |= ZHAVELOCK; | |
3507 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3508 | } | |
3509 | ||
34dc7c2f BB |
3510 | if (cmp < 0) { |
3511 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3512 | ZEXISTS | zflg, NULL, NULL); | |
3513 | terr = zfs_dirent_lock(&tdl, | |
3514 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3515 | } else { | |
3516 | terr = zfs_dirent_lock(&tdl, | |
3517 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3518 | serr = zfs_dirent_lock(&sdl, | |
3519 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3520 | NULL, NULL); | |
3521 | } | |
3522 | ||
3523 | if (serr) { | |
3524 | /* | |
3525 | * Source entry invalid or not there. | |
3526 | */ | |
3527 | if (!terr) { | |
3528 | zfs_dirent_unlock(tdl); | |
3529 | if (tzp) | |
3558fd73 | 3530 | iput(ZTOI(tzp)); |
34dc7c2f | 3531 | } |
428870ff BB |
3532 | |
3533 | if (sdzp == tdzp) | |
3534 | rw_exit(&sdzp->z_name_lock); | |
3535 | ||
34dc7c2f BB |
3536 | if (strcmp(snm, "..") == 0) |
3537 | serr = EINVAL; | |
3558fd73 | 3538 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3539 | return (serr); |
3540 | } | |
3541 | if (terr) { | |
3542 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3543 | iput(ZTOI(szp)); |
428870ff BB |
3544 | |
3545 | if (sdzp == tdzp) | |
3546 | rw_exit(&sdzp->z_name_lock); | |
3547 | ||
34dc7c2f BB |
3548 | if (strcmp(tnm, "..") == 0) |
3549 | terr = EINVAL; | |
3558fd73 | 3550 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3551 | return (terr); |
3552 | } | |
3553 | ||
3554 | /* | |
3555 | * Must have write access at the source to remove the old entry | |
3556 | * and write access at the target to create the new entry. | |
3557 | * Note that if target and source are the same, this can be | |
3558 | * done in a single check. | |
3559 | */ | |
3560 | ||
149e873a | 3561 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3562 | goto out; |
3563 | ||
3558fd73 | 3564 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3565 | /* |
3566 | * Check to make sure rename is valid. | |
3567 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3568 | */ | |
149e873a | 3569 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3570 | goto out; |
3571 | } | |
3572 | ||
3573 | /* | |
3574 | * Does target exist? | |
3575 | */ | |
3576 | if (tzp) { | |
3577 | /* | |
3578 | * Source and target must be the same type. | |
3579 | */ | |
3558fd73 BB |
3580 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3581 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
2e528b49 | 3582 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3583 | goto out; |
3584 | } | |
3585 | } else { | |
3558fd73 | 3586 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
2e528b49 | 3587 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3588 | goto out; |
3589 | } | |
3590 | } | |
3591 | /* | |
3592 | * POSIX dictates that when the source and target | |
3593 | * entries refer to the same file object, rename | |
3594 | * must do nothing and exit without error. | |
3595 | */ | |
3596 | if (szp->z_id == tzp->z_id) { | |
3597 | error = 0; | |
3598 | goto out; | |
3599 | } | |
3600 | } | |
3601 | ||
3558fd73 | 3602 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
3603 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3604 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3605 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3606 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3607 | if (sdzp != tdzp) { |
3608 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3609 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3610 | } | |
3611 | if (tzp) { | |
3612 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3613 | zfs_sa_upgrade_txholds(tx, tzp); | |
3614 | } | |
3615 | ||
3616 | zfs_sa_upgrade_txholds(tx, szp); | |
3558fd73 | 3617 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
e8b96c60 | 3618 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3619 | if (error) { |
3620 | if (zl != NULL) | |
3621 | zfs_rename_unlock(&zl); | |
3622 | zfs_dirent_unlock(sdl); | |
3623 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3624 | |
3625 | if (sdzp == tdzp) | |
3626 | rw_exit(&sdzp->z_name_lock); | |
3627 | ||
3558fd73 | 3628 | iput(ZTOI(szp)); |
34dc7c2f | 3629 | if (tzp) |
3558fd73 | 3630 | iput(ZTOI(tzp)); |
fb5f0bc8 | 3631 | if (error == ERESTART) { |
e8b96c60 | 3632 | waited = B_TRUE; |
34dc7c2f BB |
3633 | dmu_tx_wait(tx); |
3634 | dmu_tx_abort(tx); | |
3635 | goto top; | |
3636 | } | |
3637 | dmu_tx_abort(tx); | |
3558fd73 | 3638 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3639 | return (error); |
3640 | } | |
3641 | ||
3642 | if (tzp) /* Attempt to remove the existing target */ | |
3643 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3644 | ||
3645 | if (error == 0) { | |
3646 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3647 | if (error == 0) { | |
428870ff | 3648 | szp->z_pflags |= ZFS_AV_MODIFIED; |
34dc7c2f | 3649 | |
3558fd73 | 3650 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zsb), |
428870ff | 3651 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c99c9001 | 3652 | ASSERT0(error); |
34dc7c2f | 3653 | |
428870ff BB |
3654 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3655 | if (error == 0) { | |
3656 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3657 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3658 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3659 | } else { |
3660 | /* | |
3661 | * At this point, we have successfully created | |
3662 | * the target name, but have failed to remove | |
3663 | * the source name. Since the create was done | |
3664 | * with the ZRENAMING flag, there are | |
3665 | * complications; for one, the link count is | |
3666 | * wrong. The easiest way to deal with this | |
3667 | * is to remove the newly created target, and | |
3668 | * return the original error. This must | |
3669 | * succeed; fortunately, it is very unlikely to | |
3670 | * fail, since we just created it. | |
3671 | */ | |
3672 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3673 | ZRENAMING, NULL), ==, 0); | |
3674 | } | |
34dc7c2f BB |
3675 | } |
3676 | } | |
3677 | ||
3678 | dmu_tx_commit(tx); | |
3679 | out: | |
3680 | if (zl != NULL) | |
3681 | zfs_rename_unlock(&zl); | |
3682 | ||
3683 | zfs_dirent_unlock(sdl); | |
3684 | zfs_dirent_unlock(tdl); | |
3685 | ||
960e08fe | 3686 | zfs_inode_update(sdzp); |
428870ff BB |
3687 | if (sdzp == tdzp) |
3688 | rw_exit(&sdzp->z_name_lock); | |
3689 | ||
960e08fe BB |
3690 | if (sdzp != tdzp) |
3691 | zfs_inode_update(tdzp); | |
428870ff | 3692 | |
960e08fe | 3693 | zfs_inode_update(szp); |
3558fd73 | 3694 | iput(ZTOI(szp)); |
960e08fe BB |
3695 | if (tzp) { |
3696 | zfs_inode_update(tzp); | |
3558fd73 | 3697 | iput(ZTOI(tzp)); |
960e08fe | 3698 | } |
34dc7c2f | 3699 | |
3558fd73 | 3700 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3701 | zil_commit(zilog, 0); |
428870ff | 3702 | |
3558fd73 | 3703 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3704 | return (error); |
3705 | } | |
e5c39b95 | 3706 | EXPORT_SYMBOL(zfs_rename); |
34dc7c2f BB |
3707 | |
3708 | /* | |
3709 | * Insert the indicated symbolic reference entry into the directory. | |
3710 | * | |
3558fd73 | 3711 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
3712 | * link - Name for new symlink entry. |
3713 | * vap - Attributes of new entry. | |
3714 | * target - Target path of new symlink. | |
3558fd73 | 3715 | * |
34dc7c2f | 3716 | * cr - credentials of caller. |
34dc7c2f BB |
3717 | * flags - case flags |
3718 | * | |
d3cc8b15 | 3719 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3720 | * |
3721 | * Timestamps: | |
3558fd73 | 3722 | * dip - ctime|mtime updated |
34dc7c2f BB |
3723 | */ |
3724 | /*ARGSUSED*/ | |
e5c39b95 | 3725 | int |
3558fd73 BB |
3726 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
3727 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 3728 | { |
3558fd73 | 3729 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
3730 | zfs_dirlock_t *dl; |
3731 | dmu_tx_t *tx; | |
3558fd73 | 3732 | zfs_sb_t *zsb = ITOZSB(dip); |
34dc7c2f | 3733 | zilog_t *zilog; |
428870ff | 3734 | uint64_t len = strlen(link); |
34dc7c2f BB |
3735 | int error; |
3736 | int zflg = ZNEW; | |
9babb374 BB |
3737 | zfs_acl_ids_t acl_ids; |
3738 | boolean_t fuid_dirtied; | |
428870ff | 3739 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 3740 | boolean_t waited = B_FALSE; |
34dc7c2f | 3741 | |
3558fd73 | 3742 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 3743 | |
32dec7bd | 3744 | if (name == NULL) |
3745 | return (SET_ERROR(EINVAL)); | |
3746 | ||
3558fd73 | 3747 | ZFS_ENTER(zsb); |
34dc7c2f | 3748 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3749 | zilog = zsb->z_log; |
34dc7c2f | 3750 | |
3558fd73 | 3751 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 3752 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3753 | ZFS_EXIT(zsb); |
2e528b49 | 3754 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3755 | } |
3756 | if (flags & FIGNORECASE) | |
3757 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3758 | |
3759 | if (len > MAXPATHLEN) { | |
3558fd73 | 3760 | ZFS_EXIT(zsb); |
2e528b49 | 3761 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
3762 | } |
3763 | ||
428870ff BB |
3764 | if ((error = zfs_acl_ids_create(dzp, 0, |
3765 | vap, cr, NULL, &acl_ids)) != 0) { | |
3558fd73 | 3766 | ZFS_EXIT(zsb); |
428870ff BB |
3767 | return (error); |
3768 | } | |
3769 | top: | |
3558fd73 BB |
3770 | *ipp = NULL; |
3771 | ||
34dc7c2f BB |
3772 | /* |
3773 | * Attempt to lock directory; fail if entry already exists. | |
3774 | */ | |
3775 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3776 | if (error) { | |
428870ff | 3777 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 3778 | ZFS_EXIT(zsb); |
428870ff BB |
3779 | return (error); |
3780 | } | |
3781 | ||
149e873a | 3782 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
3783 | zfs_acl_ids_free(&acl_ids); |
3784 | zfs_dirent_unlock(dl); | |
3558fd73 | 3785 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3786 | return (error); |
3787 | } | |
3788 | ||
3558fd73 | 3789 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
9babb374 BB |
3790 | zfs_acl_ids_free(&acl_ids); |
3791 | zfs_dirent_unlock(dl); | |
3558fd73 | 3792 | ZFS_EXIT(zsb); |
2e528b49 | 3793 | return (SET_ERROR(EDQUOT)); |
9babb374 | 3794 | } |
3558fd73 BB |
3795 | tx = dmu_tx_create(zsb->z_os); |
3796 | fuid_dirtied = zsb->z_fuid_dirty; | |
34dc7c2f | 3797 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3798 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3799 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3800 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3801 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
3558fd73 | 3802 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
3803 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
3804 | acl_ids.z_aclp->z_acl_bytes); | |
3805 | } | |
9babb374 | 3806 | if (fuid_dirtied) |
3558fd73 | 3807 | zfs_fuid_txhold(zsb, tx); |
e8b96c60 | 3808 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3809 | if (error) { |
3810 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3811 | if (error == ERESTART) { |
e8b96c60 | 3812 | waited = B_TRUE; |
34dc7c2f BB |
3813 | dmu_tx_wait(tx); |
3814 | dmu_tx_abort(tx); | |
3815 | goto top; | |
3816 | } | |
428870ff | 3817 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 3818 | dmu_tx_abort(tx); |
3558fd73 | 3819 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3820 | return (error); |
3821 | } | |
3822 | ||
34dc7c2f BB |
3823 | /* |
3824 | * Create a new object for the symlink. | |
428870ff | 3825 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 3826 | */ |
428870ff | 3827 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3828 | |
428870ff | 3829 | if (fuid_dirtied) |
3558fd73 | 3830 | zfs_fuid_sync(zsb, tx); |
34dc7c2f | 3831 | |
572e2857 | 3832 | mutex_enter(&zp->z_lock); |
428870ff | 3833 | if (zp->z_is_sa) |
3558fd73 | 3834 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zsb), |
428870ff BB |
3835 | link, len, tx); |
3836 | else | |
3837 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3838 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3839 | |
428870ff | 3840 | zp->z_size = len; |
3558fd73 | 3841 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 3842 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
3843 | /* |
3844 | * Insert the new object into the directory. | |
3845 | */ | |
3846 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
428870ff BB |
3847 | |
3848 | if (flags & FIGNORECASE) | |
3849 | txtype |= TX_CI; | |
3850 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
9babb374 | 3851 | |
960e08fe BB |
3852 | zfs_inode_update(dzp); |
3853 | zfs_inode_update(zp); | |
3854 | ||
9babb374 | 3855 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3856 | |
3857 | dmu_tx_commit(tx); | |
3858 | ||
3859 | zfs_dirent_unlock(dl); | |
3860 | ||
3558fd73 | 3861 | *ipp = ZTOI(zp); |
34dc7c2f | 3862 | |
3558fd73 | 3863 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3864 | zil_commit(zilog, 0); |
428870ff | 3865 | |
3558fd73 | 3866 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3867 | return (error); |
3868 | } | |
e5c39b95 | 3869 | EXPORT_SYMBOL(zfs_symlink); |
34dc7c2f BB |
3870 | |
3871 | /* | |
3872 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 3873 | * the symbolic path referred to by ip. |
34dc7c2f | 3874 | * |
8b4f9a2d BB |
3875 | * IN: ip - inode of symbolic link |
3876 | * uio - structure to contain the link path. | |
3877 | * cr - credentials of caller. | |
34dc7c2f BB |
3878 | * |
3879 | * RETURN: 0 if success | |
3880 | * error code if failure | |
3881 | * | |
3882 | * Timestamps: | |
3558fd73 | 3883 | * ip - atime updated |
34dc7c2f BB |
3884 | */ |
3885 | /* ARGSUSED */ | |
e5c39b95 | 3886 | int |
8b4f9a2d | 3887 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 3888 | { |
3558fd73 BB |
3889 | znode_t *zp = ITOZ(ip); |
3890 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
3891 | int error; |
3892 | ||
3558fd73 | 3893 | ZFS_ENTER(zsb); |
34dc7c2f BB |
3894 | ZFS_VERIFY_ZP(zp); |
3895 | ||
572e2857 | 3896 | mutex_enter(&zp->z_lock); |
428870ff | 3897 | if (zp->z_is_sa) |
8b4f9a2d BB |
3898 | error = sa_lookup_uio(zp->z_sa_hdl, |
3899 | SA_ZPL_SYMLINK(zsb), uio); | |
428870ff | 3900 | else |
8b4f9a2d | 3901 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 3902 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3903 | |
3558fd73 | 3904 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3905 | return (error); |
3906 | } | |
8b4f9a2d | 3907 | EXPORT_SYMBOL(zfs_readlink); |
34dc7c2f BB |
3908 | |
3909 | /* | |
3558fd73 | 3910 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 3911 | * |
3558fd73 BB |
3912 | * IN: tdip - Directory to contain new entry. |
3913 | * sip - inode of new entry. | |
34dc7c2f BB |
3914 | * name - name of new entry. |
3915 | * cr - credentials of caller. | |
34dc7c2f BB |
3916 | * |
3917 | * RETURN: 0 if success | |
3918 | * error code if failure | |
3919 | * | |
3920 | * Timestamps: | |
3558fd73 BB |
3921 | * tdip - ctime|mtime updated |
3922 | * sip - ctime updated | |
34dc7c2f BB |
3923 | */ |
3924 | /* ARGSUSED */ | |
e5c39b95 | 3925 | int |
da5e151f BB |
3926 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr, |
3927 | int flags) | |
34dc7c2f | 3928 | { |
3558fd73 | 3929 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 3930 | znode_t *tzp, *szp; |
3558fd73 | 3931 | zfs_sb_t *zsb = ITOZSB(tdip); |
34dc7c2f BB |
3932 | zilog_t *zilog; |
3933 | zfs_dirlock_t *dl; | |
3934 | dmu_tx_t *tx; | |
34dc7c2f BB |
3935 | int error; |
3936 | int zf = ZNEW; | |
428870ff | 3937 | uint64_t parent; |
572e2857 | 3938 | uid_t owner; |
e8b96c60 | 3939 | boolean_t waited = B_FALSE; |
ace1eae8 CC |
3940 | boolean_t is_tmpfile = 0; |
3941 | uint64_t txg; | |
3942 | #ifdef HAVE_TMPFILE | |
3943 | is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); | |
3944 | #endif | |
3558fd73 | 3945 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 3946 | |
32dec7bd | 3947 | if (name == NULL) |
3948 | return (SET_ERROR(EINVAL)); | |
3949 | ||
3558fd73 | 3950 | ZFS_ENTER(zsb); |
34dc7c2f | 3951 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3952 | zilog = zsb->z_log; |
34dc7c2f | 3953 | |
428870ff BB |
3954 | /* |
3955 | * POSIX dictates that we return EPERM here. | |
3956 | * Better choices include ENOTSUP or EISDIR. | |
3957 | */ | |
3558fd73 BB |
3958 | if (S_ISDIR(sip->i_mode)) { |
3959 | ZFS_EXIT(zsb); | |
2e528b49 | 3960 | return (SET_ERROR(EPERM)); |
428870ff BB |
3961 | } |
3962 | ||
812e91a7 MT |
3963 | szp = ITOZ(sip); |
3964 | ZFS_VERIFY_ZP(szp); | |
3965 | ||
3966 | /* | |
3967 | * We check i_sb because snapshots and the ctldir must have different | |
3968 | * super blocks. | |
3969 | */ | |
c0ebc844 | 3970 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
3558fd73 | 3971 | ZFS_EXIT(zsb); |
2e528b49 | 3972 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 3973 | } |
428870ff | 3974 | |
428870ff BB |
3975 | /* Prevent links to .zfs/shares files */ |
3976 | ||
3558fd73 | 3977 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
428870ff | 3978 | &parent, sizeof (uint64_t))) != 0) { |
3558fd73 | 3979 | ZFS_EXIT(zsb); |
428870ff BB |
3980 | return (error); |
3981 | } | |
3558fd73 BB |
3982 | if (parent == zsb->z_shares_dir) { |
3983 | ZFS_EXIT(zsb); | |
2e528b49 | 3984 | return (SET_ERROR(EPERM)); |
428870ff BB |
3985 | } |
3986 | ||
3558fd73 | 3987 | if (zsb->z_utf8 && u8_validate(name, |
34dc7c2f | 3988 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3989 | ZFS_EXIT(zsb); |
2e528b49 | 3990 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3991 | } |
3992 | if (flags & FIGNORECASE) | |
3993 | zf |= ZCILOOK; | |
3994 | ||
34dc7c2f BB |
3995 | /* |
3996 | * We do not support links between attributes and non-attributes | |
3997 | * because of the potential security risk of creating links | |
3998 | * into "normal" file space in order to circumvent restrictions | |
3999 | * imposed in attribute space. | |
4000 | */ | |
428870ff | 4001 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 4002 | ZFS_EXIT(zsb); |
2e528b49 | 4003 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4004 | } |
4005 | ||
2c6abf15 | 4006 | owner = zfs_fuid_map_id(zsb, KUID_TO_SUID(sip->i_uid), cr, ZFS_OWNER); |
572e2857 | 4007 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
3558fd73 | 4008 | ZFS_EXIT(zsb); |
2e528b49 | 4009 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4010 | } |
4011 | ||
149e873a | 4012 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
3558fd73 | 4013 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4014 | return (error); |
4015 | } | |
4016 | ||
428870ff | 4017 | top: |
34dc7c2f BB |
4018 | /* |
4019 | * Attempt to lock directory; fail if entry already exists. | |
4020 | */ | |
4021 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
4022 | if (error) { | |
3558fd73 | 4023 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4024 | return (error); |
4025 | } | |
4026 | ||
3558fd73 | 4027 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 4028 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 4029 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
ace1eae8 CC |
4030 | if (is_tmpfile) |
4031 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); | |
4032 | ||
428870ff BB |
4033 | zfs_sa_upgrade_txholds(tx, szp); |
4034 | zfs_sa_upgrade_txholds(tx, dzp); | |
e8b96c60 | 4035 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
4036 | if (error) { |
4037 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4038 | if (error == ERESTART) { |
e8b96c60 | 4039 | waited = B_TRUE; |
34dc7c2f BB |
4040 | dmu_tx_wait(tx); |
4041 | dmu_tx_abort(tx); | |
4042 | goto top; | |
4043 | } | |
4044 | dmu_tx_abort(tx); | |
3558fd73 | 4045 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4046 | return (error); |
4047 | } | |
ace1eae8 CC |
4048 | /* unmark z_unlinked so zfs_link_create will not reject */ |
4049 | if (is_tmpfile) | |
4050 | szp->z_unlinked = 0; | |
34dc7c2f BB |
4051 | error = zfs_link_create(dl, szp, tx, 0); |
4052 | ||
4053 | if (error == 0) { | |
4054 | uint64_t txtype = TX_LINK; | |
ace1eae8 CC |
4055 | /* |
4056 | * tmpfile is created to be in z_unlinkedobj, so remove it. | |
4057 | * Also, we don't log in ZIL, be cause all previous file | |
4058 | * operation on the tmpfile are ignored by ZIL. Instead we | |
4059 | * always wait for txg to sync to make sure all previous | |
4060 | * operation are sync safe. | |
4061 | */ | |
4062 | if (is_tmpfile) { | |
4063 | VERIFY(zap_remove_int(zsb->z_os, zsb->z_unlinkedobj, | |
4064 | szp->z_id, tx) == 0); | |
4065 | } else { | |
4066 | if (flags & FIGNORECASE) | |
4067 | txtype |= TX_CI; | |
4068 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); | |
4069 | } | |
4070 | } else if (is_tmpfile) { | |
4071 | /* restore z_unlinked since when linking failed */ | |
4072 | szp->z_unlinked = 1; | |
34dc7c2f | 4073 | } |
ace1eae8 | 4074 | txg = dmu_tx_get_txg(tx); |
34dc7c2f BB |
4075 | dmu_tx_commit(tx); |
4076 | ||
4077 | zfs_dirent_unlock(dl); | |
4078 | ||
ace1eae8 | 4079 | if (!is_tmpfile && zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4080 | zil_commit(zilog, 0); |
428870ff | 4081 | |
ace1eae8 CC |
4082 | if (is_tmpfile) |
4083 | txg_wait_synced(dmu_objset_pool(zsb->z_os), txg); | |
4084 | ||
960e08fe BB |
4085 | zfs_inode_update(dzp); |
4086 | zfs_inode_update(szp); | |
3558fd73 | 4087 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4088 | return (error); |
4089 | } | |
e5c39b95 | 4090 | EXPORT_SYMBOL(zfs_link); |
34dc7c2f | 4091 | |
3c0e5c0f | 4092 | static void |
119a394a | 4093 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
4094 | { |
4095 | struct page *pp = arg; | |
4096 | ||
119a394a | 4097 | ClearPageError(pp); |
3c0e5c0f BB |
4098 | end_page_writeback(pp); |
4099 | } | |
4100 | ||
34dc7c2f | 4101 | /* |
3c0e5c0f BB |
4102 | * Push a page out to disk, once the page is on stable storage the |
4103 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 4104 | * |
3c0e5c0f BB |
4105 | * IN: ip - page mapped for inode. |
4106 | * pp - page to push (page is locked) | |
4107 | * wbc - writeback control data | |
34dc7c2f BB |
4108 | * |
4109 | * RETURN: 0 if success | |
4110 | * error code if failure | |
4111 | * | |
3c0e5c0f BB |
4112 | * Timestamps: |
4113 | * ip - ctime|mtime updated | |
34dc7c2f BB |
4114 | */ |
4115 | /* ARGSUSED */ | |
3c0e5c0f BB |
4116 | int |
4117 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 4118 | { |
3c0e5c0f BB |
4119 | znode_t *zp = ITOZ(ip); |
4120 | zfs_sb_t *zsb = ITOZSB(ip); | |
4121 | loff_t offset; | |
4122 | loff_t pgoff; | |
4c837f0d BB |
4123 | unsigned int pglen; |
4124 | rl_t *rl; | |
3c0e5c0f BB |
4125 | dmu_tx_t *tx; |
4126 | caddr_t va; | |
4127 | int err = 0; | |
4128 | uint64_t mtime[2], ctime[2]; | |
4129 | sa_bulk_attr_t bulk[3]; | |
4130 | int cnt = 0; | |
21a96fb6 | 4131 | struct address_space *mapping; |
3c0e5c0f | 4132 | |
4c837f0d BB |
4133 | ZFS_ENTER(zsb); |
4134 | ZFS_VERIFY_ZP(zp); | |
d164b209 | 4135 | |
3c0e5c0f BB |
4136 | ASSERT(PageLocked(pp)); |
4137 | ||
d1d7e268 MK |
4138 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
4139 | offset = i_size_read(ip); /* File length in bytes */ | |
8b1899d3 BB |
4140 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
4141 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
4142 | |
4143 | /* Page is beyond end of file */ | |
4144 | if (pgoff >= offset) { | |
4145 | unlock_page(pp); | |
4c837f0d | 4146 | ZFS_EXIT(zsb); |
3c0e5c0f BB |
4147 | return (0); |
4148 | } | |
4149 | ||
4150 | /* Truncate page length to end of file */ | |
4151 | if (pgoff + pglen > offset) | |
4152 | pglen = offset - pgoff; | |
4153 | ||
4154 | #if 0 | |
34dc7c2f | 4155 | /* |
3c0e5c0f BB |
4156 | * FIXME: Allow mmap writes past its quota. The correct fix |
4157 | * is to register a page_mkwrite() handler to count the page | |
4158 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 4159 | */ |
dde471ef PJ |
4160 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
4161 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 | 4162 | err = EDQUOT; |
9babb374 | 4163 | } |
3c0e5c0f BB |
4164 | #endif |
4165 | ||
d958324f BB |
4166 | /* |
4167 | * The ordering here is critical and must adhere to the following | |
4168 | * rules in order to avoid deadlocking in either zfs_read() or | |
4169 | * zfs_free_range() due to a lock inversion. | |
4170 | * | |
4171 | * 1) The page must be unlocked prior to acquiring the range lock. | |
4172 | * This is critical because zfs_read() calls find_lock_page() | |
4173 | * which may block on the page lock while holding the range lock. | |
4174 | * | |
4175 | * 2) Before setting or clearing write back on a page the range lock | |
4176 | * must be held in order to prevent a lock inversion with the | |
4177 | * zfs_free_range() function. | |
21a96fb6 CC |
4178 | * |
4179 | * This presents a problem because upon entering this function the | |
4180 | * page lock is already held. To safely acquire the range lock the | |
4181 | * page lock must be dropped. This creates a window where another | |
4182 | * process could truncate, invalidate, dirty, or write out the page. | |
4183 | * | |
4184 | * Therefore, after successfully reacquiring the range and page locks | |
4185 | * the current page state is checked. In the common case everything | |
4186 | * will be as is expected and it can be written out. However, if | |
4187 | * the page state has changed it must be handled accordingly. | |
d958324f | 4188 | */ |
21a96fb6 CC |
4189 | mapping = pp->mapping; |
4190 | redirty_page_for_writepage(wbc, pp); | |
d958324f | 4191 | unlock_page(pp); |
21a96fb6 | 4192 | |
d88895a0 | 4193 | rl = zfs_range_lock(&zp->z_range_lock, pgoff, pglen, RL_WRITER); |
21a96fb6 CC |
4194 | lock_page(pp); |
4195 | ||
4196 | /* Page mapping changed or it was no longer dirty, we're done */ | |
4197 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
4198 | unlock_page(pp); | |
4199 | zfs_range_unlock(rl); | |
4200 | ZFS_EXIT(zsb); | |
4201 | return (0); | |
4202 | } | |
4203 | ||
4204 | /* Another process started write block if required */ | |
4205 | if (PageWriteback(pp)) { | |
4206 | unlock_page(pp); | |
4207 | zfs_range_unlock(rl); | |
4208 | ||
4209 | if (wbc->sync_mode != WB_SYNC_NONE) | |
4210 | wait_on_page_writeback(pp); | |
4211 | ||
4212 | ZFS_EXIT(zsb); | |
4213 | return (0); | |
4214 | } | |
4215 | ||
4216 | /* Clear the dirty flag the required locks are held */ | |
4217 | if (!clear_page_dirty_for_io(pp)) { | |
4218 | unlock_page(pp); | |
4219 | zfs_range_unlock(rl); | |
4220 | ZFS_EXIT(zsb); | |
4221 | return (0); | |
4222 | } | |
4223 | ||
4224 | /* | |
4225 | * Counterpart for redirty_page_for_writepage() above. This page | |
4226 | * was in fact not skipped and should not be counted as if it were. | |
4227 | */ | |
4228 | wbc->pages_skipped--; | |
3c0e5c0f | 4229 | set_page_writeback(pp); |
21a96fb6 | 4230 | unlock_page(pp); |
3c0e5c0f | 4231 | |
dde471ef | 4232 | tx = dmu_tx_create(zsb->z_os); |
3c0e5c0f | 4233 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
4234 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4235 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 4236 | |
fb5f0bc8 | 4237 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 4238 | if (err != 0) { |
3c0e5c0f | 4239 | if (err == ERESTART) |
34dc7c2f | 4240 | dmu_tx_wait(tx); |
3c0e5c0f | 4241 | |
34dc7c2f | 4242 | dmu_tx_abort(tx); |
119a394a ED |
4243 | __set_page_dirty_nobuffers(pp); |
4244 | ClearPageError(pp); | |
4245 | end_page_writeback(pp); | |
4c837f0d BB |
4246 | zfs_range_unlock(rl); |
4247 | ZFS_EXIT(zsb); | |
3c0e5c0f | 4248 | return (err); |
34dc7c2f BB |
4249 | } |
4250 | ||
dde471ef | 4251 | va = kmap(pp); |
8b1899d3 | 4252 | ASSERT3U(pglen, <=, PAGE_SIZE); |
3c0e5c0f | 4253 | dmu_write(zsb->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 4254 | kunmap(pp); |
34dc7c2f | 4255 | |
3c0e5c0f BB |
4256 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
4257 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
4258 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, 8); | |
428870ff | 4259 | |
d3aa3ea9 BB |
4260 | /* Preserve the mtime and ctime provided by the inode */ |
4261 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4262 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
4263 | zp->z_atime_dirty = 0; | |
4264 | zp->z_seq++; | |
4265 | ||
4266 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4267 | ||
119a394a ED |
4268 | zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
4269 | zfs_putpage_commit_cb, pp); | |
45d1cae3 | 4270 | dmu_tx_commit(tx); |
d3aa3ea9 | 4271 | |
4c837f0d | 4272 | zfs_range_unlock(rl); |
34dc7c2f | 4273 | |
119a394a ED |
4274 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4275 | /* | |
4276 | * Note that this is rarely called under writepages(), because | |
4277 | * writepages() normally handles the entire commit for | |
4278 | * performance reasons. | |
4279 | */ | |
a36cc8d2 | 4280 | zil_commit(zsb->z_log, zp->z_id); |
2b286136 | 4281 | } |
3c0e5c0f | 4282 | |
4c837f0d | 4283 | ZFS_EXIT(zsb); |
3c0e5c0f | 4284 | return (err); |
34dc7c2f BB |
4285 | } |
4286 | ||
8780c539 BB |
4287 | /* |
4288 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 4289 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
4290 | */ |
4291 | int | |
4292 | zfs_dirty_inode(struct inode *ip, int flags) | |
4293 | { | |
4294 | znode_t *zp = ITOZ(ip); | |
4295 | zfs_sb_t *zsb = ITOZSB(ip); | |
4296 | dmu_tx_t *tx; | |
023699cd MM |
4297 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
4298 | sa_bulk_attr_t bulk[4]; | |
704cd075 | 4299 | int error = 0; |
8780c539 BB |
4300 | int cnt = 0; |
4301 | ||
c944be5d BB |
4302 | if (zfs_is_readonly(zsb) || dmu_objset_is_snapshot(zsb->z_os)) |
4303 | return (0); | |
4304 | ||
8780c539 BB |
4305 | ZFS_ENTER(zsb); |
4306 | ZFS_VERIFY_ZP(zp); | |
4307 | ||
704cd075 CC |
4308 | #ifdef I_DIRTY_TIME |
4309 | /* | |
4310 | * This is the lazytime semantic indroduced in Linux 4.0 | |
4311 | * This flag will only be called from update_time when lazytime is set. | |
4312 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
4313 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
4314 | * only need to dirty atime. | |
4315 | */ | |
4316 | if (flags == I_DIRTY_TIME) { | |
4317 | zp->z_atime_dirty = 1; | |
4318 | goto out; | |
4319 | } | |
4320 | #endif | |
4321 | ||
8780c539 BB |
4322 | tx = dmu_tx_create(zsb->z_os); |
4323 | ||
4324 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4325 | zfs_sa_upgrade_txholds(tx, zp); | |
4326 | ||
4327 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4328 | if (error) { | |
4329 | dmu_tx_abort(tx); | |
4330 | goto out; | |
4331 | } | |
4332 | ||
4333 | mutex_enter(&zp->z_lock); | |
704cd075 CC |
4334 | zp->z_atime_dirty = 0; |
4335 | ||
023699cd | 4336 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zsb), NULL, &mode, 8); |
8780c539 BB |
4337 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zsb), NULL, &atime, 16); |
4338 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); | |
4339 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
4340 | ||
023699cd | 4341 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4342 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4343 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4344 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
023699cd MM |
4345 | mode = ip->i_mode; |
4346 | ||
4347 | zp->z_mode = mode; | |
8780c539 BB |
4348 | |
4349 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4350 | mutex_exit(&zp->z_lock); | |
4351 | ||
4352 | dmu_tx_commit(tx); | |
4353 | out: | |
4354 | ZFS_EXIT(zsb); | |
4355 | return (error); | |
4356 | } | |
4357 | EXPORT_SYMBOL(zfs_dirty_inode); | |
4358 | ||
34dc7c2f BB |
4359 | /*ARGSUSED*/ |
4360 | void | |
c0d35759 | 4361 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4362 | { |
c0d35759 BB |
4363 | znode_t *zp = ITOZ(ip); |
4364 | zfs_sb_t *zsb = ITOZSB(ip); | |
0df9673f | 4365 | uint64_t atime[2]; |
34dc7c2f | 4366 | int error; |
cafbd2ac | 4367 | int need_unlock = 0; |
34dc7c2f | 4368 | |
cafbd2ac | 4369 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
4370 | if (!RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)) { | |
4371 | need_unlock = 1; | |
4372 | rw_enter(&zsb->z_teardown_inactive_lock, RW_READER); | |
4373 | } | |
c0d35759 | 4374 | if (zp->z_sa_hdl == NULL) { |
cafbd2ac | 4375 | if (need_unlock) |
4376 | rw_exit(&zsb->z_teardown_inactive_lock); | |
c0d35759 | 4377 | return; |
34dc7c2f BB |
4378 | } |
4379 | ||
4380 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
c0d35759 | 4381 | dmu_tx_t *tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 4382 | |
428870ff BB |
4383 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4384 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4385 | error = dmu_tx_assign(tx, TXG_WAIT); |
4386 | if (error) { | |
4387 | dmu_tx_abort(tx); | |
4388 | } else { | |
0df9673f | 4389 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
34dc7c2f | 4390 | mutex_enter(&zp->z_lock); |
3558fd73 | 4391 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), |
0df9673f | 4392 | (void *)&atime, sizeof (atime), tx); |
34dc7c2f BB |
4393 | zp->z_atime_dirty = 0; |
4394 | mutex_exit(&zp->z_lock); | |
4395 | dmu_tx_commit(tx); | |
4396 | } | |
4397 | } | |
4398 | ||
4399 | zfs_zinactive(zp); | |
cafbd2ac | 4400 | if (need_unlock) |
4401 | rw_exit(&zsb->z_teardown_inactive_lock); | |
34dc7c2f | 4402 | } |
e5c39b95 | 4403 | EXPORT_SYMBOL(zfs_inactive); |
34dc7c2f BB |
4404 | |
4405 | /* | |
4406 | * Bounds-check the seek operation. | |
4407 | * | |
3558fd73 | 4408 | * IN: ip - inode seeking within |
34dc7c2f BB |
4409 | * ooff - old file offset |
4410 | * noffp - pointer to new file offset | |
4411 | * ct - caller context | |
4412 | * | |
4413 | * RETURN: 0 if success | |
4414 | * EINVAL if new offset invalid | |
4415 | */ | |
4416 | /* ARGSUSED */ | |
3558fd73 | 4417 | int |
9623f736 | 4418 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4419 | { |
3558fd73 | 4420 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4421 | return (0); |
4422 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4423 | } | |
3558fd73 | 4424 | EXPORT_SYMBOL(zfs_seek); |
34dc7c2f | 4425 | |
34dc7c2f | 4426 | /* |
dde471ef | 4427 | * Fill pages with data from the disk. |
34dc7c2f BB |
4428 | */ |
4429 | static int | |
dde471ef | 4430 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4431 | { |
d1d7e268 MK |
4432 | znode_t *zp = ITOZ(ip); |
4433 | zfs_sb_t *zsb = ITOZSB(ip); | |
4434 | objset_t *os; | |
dde471ef | 4435 | struct page *cur_pp; |
d1d7e268 MK |
4436 | u_offset_t io_off, total; |
4437 | size_t io_len; | |
4438 | loff_t i_size; | |
4439 | unsigned page_idx; | |
4440 | int err; | |
34dc7c2f | 4441 | |
d1d7e268 | 4442 | os = zsb->z_os; |
8b1899d3 | 4443 | io_len = nr_pages << PAGE_SHIFT; |
dde471ef PJ |
4444 | i_size = i_size_read(ip); |
4445 | io_off = page_offset(pl[0]); | |
4446 | ||
4447 | if (io_off + io_len > i_size) | |
4448 | io_len = i_size - io_off; | |
34dc7c2f BB |
4449 | |
4450 | /* | |
dde471ef | 4451 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4452 | */ |
dde471ef | 4453 | page_idx = 0; |
34dc7c2f | 4454 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4455 | caddr_t va; |
4456 | ||
540c3927 | 4457 | cur_pp = pl[page_idx++]; |
dde471ef | 4458 | va = kmap(cur_pp); |
9babb374 BB |
4459 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4460 | DMU_READ_PREFETCH); | |
dde471ef | 4461 | kunmap(cur_pp); |
34dc7c2f | 4462 | if (err) { |
b128c09f BB |
4463 | /* convert checksum errors into IO errors */ |
4464 | if (err == ECKSUM) | |
2e528b49 | 4465 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4466 | return (err); |
4467 | } | |
34dc7c2f | 4468 | } |
d164b209 | 4469 | |
34dc7c2f BB |
4470 | return (0); |
4471 | } | |
4472 | ||
4473 | /* | |
dde471ef | 4474 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4475 | * |
dde471ef PJ |
4476 | * IN: ip - inode of file to get data from. |
4477 | * pl - list of pages to read | |
4478 | * nr_pages - number of pages to read | |
34dc7c2f | 4479 | * |
d3cc8b15 | 4480 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4481 | * |
4482 | * Timestamps: | |
4483 | * vp - atime updated | |
4484 | */ | |
4485 | /* ARGSUSED */ | |
dde471ef PJ |
4486 | int |
4487 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4488 | { |
dde471ef PJ |
4489 | znode_t *zp = ITOZ(ip); |
4490 | zfs_sb_t *zsb = ITOZSB(ip); | |
4491 | int err; | |
d164b209 | 4492 | |
d164b209 BB |
4493 | if (pl == NULL) |
4494 | return (0); | |
34dc7c2f | 4495 | |
dde471ef | 4496 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4497 | ZFS_VERIFY_ZP(zp); |
4498 | ||
dde471ef | 4499 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4500 | |
dde471ef | 4501 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4502 | return (err); |
4503 | } | |
dde471ef | 4504 | EXPORT_SYMBOL(zfs_getpage); |
34dc7c2f BB |
4505 | |
4506 | /* | |
e2e7aa2d | 4507 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4508 | * |
e2e7aa2d BB |
4509 | * IN: ip - inode of the file to mmap |
4510 | * off - file offset | |
4511 | * addrp - start address in memory region | |
4512 | * len - length of memory region | |
4513 | * vm_flags- address flags | |
34dc7c2f | 4514 | * |
e2e7aa2d BB |
4515 | * RETURN: 0 if success |
4516 | * error code if failure | |
34dc7c2f BB |
4517 | */ |
4518 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4519 | int |
4520 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4521 | unsigned long vm_flags) | |
34dc7c2f | 4522 | { |
e2e7aa2d BB |
4523 | znode_t *zp = ITOZ(ip); |
4524 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4525 | |
e2e7aa2d | 4526 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4527 | ZFS_VERIFY_ZP(zp); |
4528 | ||
e2e7aa2d | 4529 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4530 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
e2e7aa2d | 4531 | ZFS_EXIT(zsb); |
2e528b49 | 4532 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4533 | } |
4534 | ||
e2e7aa2d | 4535 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4536 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
e2e7aa2d | 4537 | ZFS_EXIT(zsb); |
2e528b49 | 4538 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4539 | } |
4540 | ||
34dc7c2f | 4541 | if (off < 0 || len > MAXOFFSET_T - off) { |
e2e7aa2d | 4542 | ZFS_EXIT(zsb); |
2e528b49 | 4543 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4544 | } |
4545 | ||
e2e7aa2d | 4546 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4547 | return (0); |
4548 | } | |
e2e7aa2d | 4549 | EXPORT_SYMBOL(zfs_map); |
34dc7c2f | 4550 | |
3558fd73 BB |
4551 | /* |
4552 | * convoff - converts the given data (start, whence) to the | |
4553 | * given whence. | |
4554 | */ | |
4555 | int | |
4556 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4557 | { | |
5484965a | 4558 | vattr_t vap; |
3558fd73 BB |
4559 | int error; |
4560 | ||
4561 | if ((lckdat->l_whence == 2) || (whence == 2)) { | |
5484965a | 4562 | if ((error = zfs_getattr(ip, &vap, 0, CRED()) != 0)) |
3558fd73 BB |
4563 | return (error); |
4564 | } | |
4565 | ||
4566 | switch (lckdat->l_whence) { | |
4567 | case 1: | |
4568 | lckdat->l_start += offset; | |
4569 | break; | |
4570 | case 2: | |
5484965a | 4571 | lckdat->l_start += vap.va_size; |
3558fd73 BB |
4572 | /* FALLTHRU */ |
4573 | case 0: | |
4574 | break; | |
4575 | default: | |
2e528b49 | 4576 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4577 | } |
4578 | ||
4579 | if (lckdat->l_start < 0) | |
2e528b49 | 4580 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4581 | |
4582 | switch (whence) { | |
4583 | case 1: | |
4584 | lckdat->l_start -= offset; | |
4585 | break; | |
4586 | case 2: | |
5484965a | 4587 | lckdat->l_start -= vap.va_size; |
3558fd73 BB |
4588 | /* FALLTHRU */ |
4589 | case 0: | |
4590 | break; | |
4591 | default: | |
2e528b49 | 4592 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4593 | } |
4594 | ||
4595 | lckdat->l_whence = (short)whence; | |
4596 | return (0); | |
4597 | } | |
4598 | ||
34dc7c2f BB |
4599 | /* |
4600 | * Free or allocate space in a file. Currently, this function only | |
4601 | * supports the `F_FREESP' command. However, this command is somewhat | |
4602 | * misnamed, as its functionality includes the ability to allocate as | |
4603 | * well as free space. | |
4604 | * | |
3558fd73 | 4605 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4606 | * cmd - action to take (only F_FREESP supported). |
4607 | * bfp - section of file to free/alloc. | |
4608 | * flag - current file open mode flags. | |
4609 | * offset - current file offset. | |
4610 | * cr - credentials of caller [UNUSED]. | |
34dc7c2f | 4611 | * |
d3cc8b15 | 4612 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4613 | * |
4614 | * Timestamps: | |
3558fd73 | 4615 | * ip - ctime|mtime updated |
34dc7c2f BB |
4616 | */ |
4617 | /* ARGSUSED */ | |
e5c39b95 | 4618 | int |
3558fd73 BB |
4619 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4620 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4621 | { |
3558fd73 BB |
4622 | znode_t *zp = ITOZ(ip); |
4623 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4624 | uint64_t off, len; |
4625 | int error; | |
4626 | ||
3558fd73 | 4627 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4628 | ZFS_VERIFY_ZP(zp); |
4629 | ||
34dc7c2f | 4630 | if (cmd != F_FREESP) { |
3558fd73 | 4631 | ZFS_EXIT(zsb); |
2e528b49 | 4632 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4633 | } |
4634 | ||
f3c9dca0 MT |
4635 | /* |
4636 | * Callers might not be able to detect properly that we are read-only, | |
4637 | * so check it explicitly here. | |
4638 | */ | |
4639 | if (zfs_is_readonly(zsb)) { | |
4640 | ZFS_EXIT(zsb); | |
4641 | return (SET_ERROR(EROFS)); | |
4642 | } | |
4643 | ||
3558fd73 BB |
4644 | if ((error = convoff(ip, bfp, 0, offset))) { |
4645 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
4646 | return (error); |
4647 | } | |
4648 | ||
4649 | if (bfp->l_len < 0) { | |
3558fd73 | 4650 | ZFS_EXIT(zsb); |
2e528b49 | 4651 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4652 | } |
4653 | ||
aec69371 ED |
4654 | /* |
4655 | * Permissions aren't checked on Solaris because on this OS | |
4656 | * zfs_space() can only be called with an opened file handle. | |
4657 | * On Linux we can get here through truncate_range() which | |
4658 | * operates directly on inodes, so we need to check access rights. | |
4659 | */ | |
4660 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
4661 | ZFS_EXIT(zsb); | |
4662 | return (error); | |
4663 | } | |
4664 | ||
34dc7c2f BB |
4665 | off = bfp->l_start; |
4666 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4667 | ||
b128c09f | 4668 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4669 | |
3558fd73 | 4670 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4671 | return (error); |
4672 | } | |
e5c39b95 | 4673 | EXPORT_SYMBOL(zfs_space); |
34dc7c2f BB |
4674 | |
4675 | /*ARGSUSED*/ | |
e5c39b95 | 4676 | int |
3558fd73 | 4677 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4678 | { |
3558fd73 BB |
4679 | znode_t *zp = ITOZ(ip); |
4680 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4681 | uint32_t gen; |
428870ff | 4682 | uint64_t gen64; |
34dc7c2f BB |
4683 | uint64_t object = zp->z_id; |
4684 | zfid_short_t *zfid; | |
428870ff | 4685 | int size, i, error; |
34dc7c2f | 4686 | |
3558fd73 | 4687 | ZFS_ENTER(zsb); |
34dc7c2f | 4688 | ZFS_VERIFY_ZP(zp); |
428870ff | 4689 | |
3558fd73 | 4690 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), |
428870ff | 4691 | &gen64, sizeof (uint64_t))) != 0) { |
3558fd73 | 4692 | ZFS_EXIT(zsb); |
428870ff BB |
4693 | return (error); |
4694 | } | |
4695 | ||
4696 | gen = (uint32_t)gen64; | |
34dc7c2f | 4697 | |
3558fd73 | 4698 | size = (zsb->z_parent != zsb) ? LONG_FID_LEN : SHORT_FID_LEN; |
34dc7c2f BB |
4699 | if (fidp->fid_len < size) { |
4700 | fidp->fid_len = size; | |
3558fd73 | 4701 | ZFS_EXIT(zsb); |
2e528b49 | 4702 | return (SET_ERROR(ENOSPC)); |
34dc7c2f BB |
4703 | } |
4704 | ||
4705 | zfid = (zfid_short_t *)fidp; | |
4706 | ||
4707 | zfid->zf_len = size; | |
4708 | ||
4709 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4710 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4711 | ||
4712 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4713 | if (gen == 0) | |
4714 | gen = 1; | |
4715 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4716 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4717 | ||
4718 | if (size == LONG_FID_LEN) { | |
3558fd73 | 4719 | uint64_t objsetid = dmu_objset_id(zsb->z_os); |
34dc7c2f BB |
4720 | zfid_long_t *zlfid; |
4721 | ||
4722 | zlfid = (zfid_long_t *)fidp; | |
4723 | ||
4724 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
4725 | zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); | |
4726 | ||
4727 | /* XXX - this should be the generation number for the objset */ | |
4728 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
4729 | zlfid->zf_setgen[i] = 0; | |
4730 | } | |
4731 | ||
3558fd73 | 4732 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4733 | return (0); |
4734 | } | |
e5c39b95 | 4735 | EXPORT_SYMBOL(zfs_fid); |
34dc7c2f | 4736 | |
34dc7c2f | 4737 | /*ARGSUSED*/ |
e5c39b95 | 4738 | int |
3558fd73 | 4739 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4740 | { |
3558fd73 BB |
4741 | znode_t *zp = ITOZ(ip); |
4742 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4743 | int error; |
4744 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
4745 | ||
3558fd73 | 4746 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4747 | ZFS_VERIFY_ZP(zp); |
4748 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
3558fd73 | 4749 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4750 | |
4751 | return (error); | |
4752 | } | |
e5c39b95 | 4753 | EXPORT_SYMBOL(zfs_getsecattr); |
34dc7c2f BB |
4754 | |
4755 | /*ARGSUSED*/ | |
e5c39b95 | 4756 | int |
3558fd73 | 4757 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4758 | { |
3558fd73 BB |
4759 | znode_t *zp = ITOZ(ip); |
4760 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4761 | int error; |
4762 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
3558fd73 | 4763 | zilog_t *zilog = zsb->z_log; |
34dc7c2f | 4764 | |
3558fd73 | 4765 | ZFS_ENTER(zsb); |
34dc7c2f | 4766 | ZFS_VERIFY_ZP(zp); |
428870ff | 4767 | |
34dc7c2f | 4768 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 4769 | |
3558fd73 | 4770 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4771 | zil_commit(zilog, 0); |
428870ff | 4772 | |
3558fd73 | 4773 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4774 | return (error); |
4775 | } | |
e5c39b95 | 4776 | EXPORT_SYMBOL(zfs_setsecattr); |
34dc7c2f | 4777 | |
3558fd73 | 4778 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
4779 | /* |
4780 | * Tunable, both must be a power of 2. | |
4781 | * | |
4782 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
4783 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 4784 | * an arcbuf for a partial block read |
428870ff BB |
4785 | */ |
4786 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
4787 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
4788 | ||
4789 | /*ARGSUSED*/ | |
4790 | static int | |
3558fd73 | 4791 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 4792 | { |
3558fd73 BB |
4793 | znode_t *zp = ITOZ(ip); |
4794 | zfs_sb_t *zsb = ITOZSB(ip); | |
4795 | int max_blksz = zsb->z_max_blksz; | |
428870ff BB |
4796 | uio_t *uio = &xuio->xu_uio; |
4797 | ssize_t size = uio->uio_resid; | |
4798 | offset_t offset = uio->uio_loffset; | |
4799 | int blksz; | |
4800 | int fullblk, i; | |
4801 | arc_buf_t *abuf; | |
4802 | ssize_t maxsize; | |
4803 | int preamble, postamble; | |
4804 | ||
4805 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
2e528b49 | 4806 | return (SET_ERROR(EINVAL)); |
428870ff | 4807 | |
3558fd73 | 4808 | ZFS_ENTER(zsb); |
428870ff BB |
4809 | ZFS_VERIFY_ZP(zp); |
4810 | switch (ioflag) { | |
4811 | case UIO_WRITE: | |
4812 | /* | |
4813 | * Loan out an arc_buf for write if write size is bigger than | |
4814 | * max_blksz, and the file's block size is also max_blksz. | |
4815 | */ | |
4816 | blksz = max_blksz; | |
4817 | if (size < blksz || zp->z_blksz != blksz) { | |
3558fd73 | 4818 | ZFS_EXIT(zsb); |
2e528b49 | 4819 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4820 | } |
4821 | /* | |
4822 | * Caller requests buffers for write before knowing where the | |
4823 | * write offset might be (e.g. NFS TCP write). | |
4824 | */ | |
4825 | if (offset == -1) { | |
4826 | preamble = 0; | |
4827 | } else { | |
4828 | preamble = P2PHASE(offset, blksz); | |
4829 | if (preamble) { | |
4830 | preamble = blksz - preamble; | |
4831 | size -= preamble; | |
4832 | } | |
4833 | } | |
4834 | ||
4835 | postamble = P2PHASE(size, blksz); | |
4836 | size -= postamble; | |
4837 | ||
4838 | fullblk = size / blksz; | |
4839 | (void) dmu_xuio_init(xuio, | |
4840 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
4841 | |
4842 | /* | |
4843 | * Have to fix iov base/len for partial buffers. They | |
4844 | * currently represent full arc_buf's. | |
4845 | */ | |
4846 | if (preamble) { | |
4847 | /* data begins in the middle of the arc_buf */ | |
4848 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4849 | blksz); | |
4850 | ASSERT(abuf); | |
4851 | (void) dmu_xuio_add(xuio, abuf, | |
4852 | blksz - preamble, preamble); | |
4853 | } | |
4854 | ||
4855 | for (i = 0; i < fullblk; i++) { | |
4856 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4857 | blksz); | |
4858 | ASSERT(abuf); | |
4859 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
4860 | } | |
4861 | ||
4862 | if (postamble) { | |
4863 | /* data ends in the middle of the arc_buf */ | |
4864 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4865 | blksz); | |
4866 | ASSERT(abuf); | |
4867 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
4868 | } | |
4869 | break; | |
4870 | case UIO_READ: | |
4871 | /* | |
4872 | * Loan out an arc_buf for read if the read size is larger than | |
4873 | * the current file block size. Block alignment is not | |
4874 | * considered. Partial arc_buf will be loaned out for read. | |
4875 | */ | |
4876 | blksz = zp->z_blksz; | |
4877 | if (blksz < zcr_blksz_min) | |
4878 | blksz = zcr_blksz_min; | |
4879 | if (blksz > zcr_blksz_max) | |
4880 | blksz = zcr_blksz_max; | |
4881 | /* avoid potential complexity of dealing with it */ | |
4882 | if (blksz > max_blksz) { | |
3558fd73 | 4883 | ZFS_EXIT(zsb); |
2e528b49 | 4884 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4885 | } |
4886 | ||
4887 | maxsize = zp->z_size - uio->uio_loffset; | |
4888 | if (size > maxsize) | |
4889 | size = maxsize; | |
4890 | ||
3558fd73 BB |
4891 | if (size < blksz) { |
4892 | ZFS_EXIT(zsb); | |
2e528b49 | 4893 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4894 | } |
4895 | break; | |
4896 | default: | |
3558fd73 | 4897 | ZFS_EXIT(zsb); |
2e528b49 | 4898 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4899 | } |
4900 | ||
4901 | uio->uio_extflg = UIO_XUIO; | |
4902 | XUIO_XUZC_RW(xuio) = ioflag; | |
3558fd73 | 4903 | ZFS_EXIT(zsb); |
428870ff BB |
4904 | return (0); |
4905 | } | |
4906 | ||
4907 | /*ARGSUSED*/ | |
4908 | static int | |
3558fd73 | 4909 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
4910 | { |
4911 | int i; | |
4912 | arc_buf_t *abuf; | |
4913 | int ioflag = XUIO_XUZC_RW(xuio); | |
4914 | ||
4915 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
4916 | ||
4917 | i = dmu_xuio_cnt(xuio); | |
4918 | while (i-- > 0) { | |
4919 | abuf = dmu_xuio_arcbuf(xuio, i); | |
4920 | /* | |
4921 | * if abuf == NULL, it must be a write buffer | |
4922 | * that has been returned in zfs_write(). | |
4923 | */ | |
4924 | if (abuf) | |
4925 | dmu_return_arcbuf(abuf); | |
4926 | ASSERT(abuf || ioflag == UIO_WRITE); | |
4927 | } | |
4928 | ||
4929 | dmu_xuio_fini(xuio); | |
4930 | return (0); | |
4931 | } | |
3558fd73 | 4932 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 BB |
4933 | |
4934 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
02730c33 | 4935 | /* CSTYLED */ |
a966c564 K |
4936 | module_param(zfs_delete_blocks, ulong, 0644); |
4937 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
c409e464 BB |
4938 | module_param(zfs_read_chunk_size, long, 0644); |
4939 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
4940 | #endif |