]>
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 BB |
598 | xuio_t *xuio = NULL; |
599 | int i_iov = 0; | |
5475aada | 600 | const iovec_t *iovp = uio->uio_iov; |
428870ff BB |
601 | int write_eof; |
602 | int count = 0; | |
603 | sa_bulk_attr_t bulk[4]; | |
604 | uint64_t mtime[2], ctime[2]; | |
2c6abf15 | 605 | uint32_t uid; |
3558fd73 | 606 | ASSERTV(int iovcnt = uio->uio_iovcnt); |
34dc7c2f | 607 | |
34dc7c2f BB |
608 | /* |
609 | * Fasttrack empty write | |
610 | */ | |
611 | n = start_resid; | |
612 | if (n == 0) | |
613 | return (0); | |
614 | ||
615 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) | |
616 | limit = MAXOFFSET_T; | |
617 | ||
3558fd73 | 618 | ZFS_ENTER(zsb); |
34dc7c2f | 619 | ZFS_VERIFY_ZP(zp); |
b128c09f | 620 | |
3558fd73 BB |
621 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
622 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
623 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); | |
624 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, | |
428870ff BB |
625 | &zp->z_pflags, 8); |
626 | ||
f3c9dca0 MT |
627 | /* |
628 | * Callers might not be able to detect properly that we are read-only, | |
629 | * so check it explicitly here. | |
630 | */ | |
631 | if (zfs_is_readonly(zsb)) { | |
632 | ZFS_EXIT(zsb); | |
633 | return (SET_ERROR(EROFS)); | |
634 | } | |
635 | ||
b128c09f BB |
636 | /* |
637 | * If immutable or not appending then return EPERM | |
638 | */ | |
428870ff BB |
639 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
640 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
641 | (uio->uio_loffset < zp->z_size))) { | |
3558fd73 | 642 | ZFS_EXIT(zsb); |
2e528b49 | 643 | return (SET_ERROR(EPERM)); |
b128c09f BB |
644 | } |
645 | ||
3558fd73 | 646 | zilog = zsb->z_log; |
34dc7c2f | 647 | |
428870ff BB |
648 | /* |
649 | * Validate file offset | |
650 | */ | |
651 | woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; | |
652 | if (woff < 0) { | |
3558fd73 | 653 | ZFS_EXIT(zsb); |
2e528b49 | 654 | return (SET_ERROR(EINVAL)); |
428870ff BB |
655 | } |
656 | ||
34dc7c2f BB |
657 | /* |
658 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
659 | * don't hold up txg. | |
428870ff | 660 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 661 | */ |
9cac042c | 662 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
663 | if ((uio->uio_extflg == UIO_XUIO) && |
664 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
665 | xuio = (xuio_t *)uio; | |
666 | else | |
9cac042c | 667 | #endif |
572e2857 | 668 | uio_prefaultpages(MIN(n, max_blksz), uio); |
34dc7c2f BB |
669 | |
670 | /* | |
671 | * If in append mode, set the io offset pointer to eof. | |
672 | */ | |
673 | if (ioflag & FAPPEND) { | |
674 | /* | |
428870ff BB |
675 | * Obtain an appending range lock to guarantee file append |
676 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f | 677 | */ |
d88895a0 | 678 | rl = zfs_range_lock(&zp->z_range_lock, 0, n, RL_APPEND); |
428870ff | 679 | woff = rl->r_off; |
34dc7c2f | 680 | if (rl->r_len == UINT64_MAX) { |
428870ff BB |
681 | /* |
682 | * We overlocked the file because this write will cause | |
683 | * the file block size to increase. | |
684 | * Note that zp_size cannot change with this lock held. | |
685 | */ | |
686 | woff = zp->z_size; | |
34dc7c2f | 687 | } |
428870ff | 688 | uio->uio_loffset = woff; |
34dc7c2f | 689 | } else { |
34dc7c2f | 690 | /* |
428870ff BB |
691 | * Note that if the file block size will change as a result of |
692 | * this write, then this range lock will lock the entire file | |
693 | * so that we can re-write the block safely. | |
34dc7c2f | 694 | */ |
d88895a0 | 695 | rl = zfs_range_lock(&zp->z_range_lock, woff, n, RL_WRITER); |
34dc7c2f BB |
696 | } |
697 | ||
698 | if (woff >= limit) { | |
699 | zfs_range_unlock(rl); | |
3558fd73 | 700 | ZFS_EXIT(zsb); |
2e528b49 | 701 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
702 | } |
703 | ||
704 | if ((woff + n) > limit || woff > (limit - n)) | |
705 | n = limit - woff; | |
706 | ||
428870ff BB |
707 | /* Will this write extend the file length? */ |
708 | write_eof = (woff + n > zp->z_size); | |
709 | ||
710 | end_size = MAX(zp->z_size, woff + n); | |
34dc7c2f BB |
711 | |
712 | /* | |
713 | * Write the file in reasonable size chunks. Each chunk is written | |
714 | * in a separate transaction; this keeps the intent log records small | |
715 | * and allows us to do more fine-grained space accounting. | |
716 | */ | |
717 | while (n > 0) { | |
9babb374 BB |
718 | abuf = NULL; |
719 | woff = uio->uio_loffset; | |
3558fd73 BB |
720 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
721 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 BB |
722 | if (abuf != NULL) |
723 | dmu_return_arcbuf(abuf); | |
2e528b49 | 724 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
725 | break; |
726 | } | |
727 | ||
428870ff BB |
728 | if (xuio && abuf == NULL) { |
729 | ASSERT(i_iov < iovcnt); | |
5475aada | 730 | ASSERT3U(uio->uio_segflg, !=, UIO_BVEC); |
428870ff BB |
731 | aiov = &iovp[i_iov]; |
732 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
733 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
734 | ASSERT((aiov->iov_base == abuf->b_data) || |
735 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
736 | aiov->iov_len == arc_buf_size(abuf))); | |
737 | i_iov++; | |
738 | } else if (abuf == NULL && n >= max_blksz && | |
739 | woff >= zp->z_size && | |
9babb374 BB |
740 | P2PHASE(woff, max_blksz) == 0 && |
741 | zp->z_blksz == max_blksz) { | |
428870ff BB |
742 | /* |
743 | * This write covers a full block. "Borrow" a buffer | |
744 | * from the dmu so that we can fill it before we enter | |
745 | * a transaction. This avoids the possibility of | |
746 | * holding up the transaction if the data copy hangs | |
747 | * up on a pagefault (e.g., from an NFS server mapping). | |
748 | */ | |
9babb374 BB |
749 | size_t cbytes; |
750 | ||
428870ff BB |
751 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
752 | max_blksz); | |
9babb374 BB |
753 | ASSERT(abuf != NULL); |
754 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
755 | if ((error = uiocopy(abuf->b_data, max_blksz, |
756 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
757 | dmu_return_arcbuf(abuf); |
758 | break; | |
759 | } | |
760 | ASSERT(cbytes == max_blksz); | |
761 | } | |
762 | ||
34dc7c2f BB |
763 | /* |
764 | * Start a transaction. | |
765 | */ | |
3558fd73 | 766 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 767 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 768 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 769 | zfs_sa_upgrade_txholds(tx, zp); |
384f8a09 | 770 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 771 | if (error) { |
34dc7c2f | 772 | dmu_tx_abort(tx); |
9babb374 BB |
773 | if (abuf != NULL) |
774 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
775 | break; |
776 | } | |
777 | ||
778 | /* | |
779 | * If zfs_range_lock() over-locked we grow the blocksize | |
780 | * and then reduce the lock range. This will only happen | |
781 | * on the first iteration since zfs_range_reduce() will | |
782 | * shrink down r_len to the appropriate size. | |
783 | */ | |
784 | if (rl->r_len == UINT64_MAX) { | |
785 | uint64_t new_blksz; | |
786 | ||
787 | if (zp->z_blksz > max_blksz) { | |
f1512ee6 MA |
788 | /* |
789 | * File's blocksize is already larger than the | |
790 | * "recordsize" property. Only let it grow to | |
791 | * the next power of 2. | |
792 | */ | |
34dc7c2f | 793 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 MA |
794 | new_blksz = MIN(end_size, |
795 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
796 | } else { |
797 | new_blksz = MIN(end_size, max_blksz); | |
798 | } | |
799 | zfs_grow_blocksize(zp, new_blksz, tx); | |
800 | zfs_range_reduce(rl, woff, n); | |
801 | } | |
802 | ||
803 | /* | |
804 | * XXX - should we really limit each write to z_max_blksz? | |
805 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
806 | */ | |
807 | nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); | |
34dc7c2f | 808 | |
9babb374 BB |
809 | if (abuf == NULL) { |
810 | tx_bytes = uio->uio_resid; | |
428870ff BB |
811 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
812 | uio, nbytes, tx); | |
9babb374 BB |
813 | tx_bytes -= uio->uio_resid; |
814 | } else { | |
815 | tx_bytes = nbytes; | |
428870ff BB |
816 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
817 | /* | |
818 | * If this is not a full block write, but we are | |
819 | * extending the file past EOF and this data starts | |
820 | * block-aligned, use assign_arcbuf(). Otherwise, | |
821 | * write via dmu_write(). | |
822 | */ | |
823 | if (tx_bytes < max_blksz && (!write_eof || | |
824 | aiov->iov_base != abuf->b_data)) { | |
825 | ASSERT(xuio); | |
3558fd73 | 826 | dmu_write(zsb->z_os, zp->z_id, woff, |
428870ff BB |
827 | aiov->iov_len, aiov->iov_base, tx); |
828 | dmu_return_arcbuf(abuf); | |
829 | xuio_stat_wbuf_copied(); | |
830 | } else { | |
831 | ASSERT(xuio || tx_bytes == max_blksz); | |
832 | dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl), | |
833 | woff, abuf, tx); | |
834 | } | |
9babb374 BB |
835 | ASSERT(tx_bytes <= uio->uio_resid); |
836 | uioskip(uio, tx_bytes); | |
837 | } | |
c0d35759 BB |
838 | |
839 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) | |
840 | update_pages(ip, woff, tx_bytes, zsb->z_os, zp->z_id); | |
34dc7c2f BB |
841 | |
842 | /* | |
843 | * If we made no progress, we're done. If we made even | |
844 | * partial progress, update the znode and ZIL accordingly. | |
845 | */ | |
846 | if (tx_bytes == 0) { | |
3558fd73 | 847 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 848 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
849 | dmu_tx_commit(tx); |
850 | ASSERT(error != 0); | |
851 | break; | |
852 | } | |
853 | ||
854 | /* | |
855 | * Clear Set-UID/Set-GID bits on successful write if not | |
856 | * privileged and at least one of the excute bits is set. | |
857 | * | |
858 | * It would be nice to to this after all writes have | |
859 | * been done, but that would still expose the ISUID/ISGID | |
860 | * to another app after the partial write is committed. | |
861 | * | |
572e2857 BB |
862 | * Note: we don't call zfs_fuid_map_id() here because |
863 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
864 | */ |
865 | mutex_enter(&zp->z_acl_lock); | |
2c6abf15 | 866 | uid = KUID_TO_SUID(ip->i_uid); |
428870ff | 867 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 868 | (S_IXUSR >> 6))) != 0 && |
428870ff | 869 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 870 | secpolicy_vnode_setid_retain(cr, |
2c6abf15 | 871 | ((zp->z_mode & S_ISUID) != 0 && uid == 0)) != 0) { |
428870ff BB |
872 | uint64_t newmode; |
873 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
12fa7f34 | 874 | ip->i_mode = newmode = zp->z_mode; |
3558fd73 | 875 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zsb), |
428870ff | 876 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
877 | } |
878 | mutex_exit(&zp->z_acl_lock); | |
879 | ||
0df9673f | 880 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
34dc7c2f BB |
881 | |
882 | /* | |
883 | * Update the file size (zp_size) if it has changed; | |
884 | * account for possible concurrent updates. | |
885 | */ | |
428870ff BB |
886 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
887 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 888 | uio->uio_loffset); |
428870ff BB |
889 | ASSERT(error == 0); |
890 | } | |
572e2857 BB |
891 | /* |
892 | * If we are replaying and eof is non zero then force | |
893 | * the file size to the specified eof. Note, there's no | |
894 | * concurrency during replay. | |
895 | */ | |
3558fd73 BB |
896 | if (zsb->z_replay && zsb->z_replay_eof != 0) |
897 | zp->z_size = zsb->z_replay_eof; | |
572e2857 | 898 | |
428870ff BB |
899 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
900 | ||
119a394a ED |
901 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
902 | NULL, NULL); | |
34dc7c2f BB |
903 | dmu_tx_commit(tx); |
904 | ||
905 | if (error != 0) | |
906 | break; | |
907 | ASSERT(tx_bytes == nbytes); | |
908 | n -= nbytes; | |
572e2857 BB |
909 | |
910 | if (!xuio && n > 0) | |
911 | uio_prefaultpages(MIN(n, max_blksz), uio); | |
34dc7c2f BB |
912 | } |
913 | ||
2a53e2da | 914 | zfs_inode_update(zp); |
34dc7c2f BB |
915 | zfs_range_unlock(rl); |
916 | ||
917 | /* | |
918 | * If we're in replay mode, or we made no progress, return error. | |
919 | * Otherwise, it's at least a partial write, so it's successful. | |
920 | */ | |
3558fd73 BB |
921 | if (zsb->z_replay || uio->uio_resid == start_resid) { |
922 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
923 | return (error); |
924 | } | |
925 | ||
428870ff | 926 | if (ioflag & (FSYNC | FDSYNC) || |
3558fd73 | 927 | zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 928 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 929 | |
3558fd73 | 930 | ZFS_EXIT(zsb); |
34dc7c2f BB |
931 | return (0); |
932 | } | |
e5c39b95 | 933 | EXPORT_SYMBOL(zfs_write); |
34dc7c2f | 934 | |
7c502b0b | 935 | /* |
936 | * Drop a reference on the passed inode asynchronously. This ensures | |
937 | * that the caller will never drop the last reference on an inode in | |
938 | * the current context. Doing so while holding open a tx could result | |
939 | * in a deadlock if iput_final() re-enters the filesystem code. | |
940 | */ | |
0a50679c BB |
941 | void |
942 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 943 | { |
0a50679c BB |
944 | objset_t *os = ITOZSB(ip)->z_os; |
945 | ||
3558fd73 | 946 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
947 | ASSERT(os != NULL); |
948 | ||
3558fd73 | 949 | if (atomic_read(&ip->i_count) == 1) |
7c502b0b | 950 | VERIFY(taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
951 | (task_func_t *)iput, ip, TQ_SLEEP) != 0); | |
3558fd73 BB |
952 | else |
953 | iput(ip); | |
954 | } | |
955 | ||
34dc7c2f | 956 | void |
428870ff | 957 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 958 | { |
428870ff | 959 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
960 | |
961 | if (zgd->zgd_db) | |
962 | dmu_buf_rele(zgd->zgd_db, zgd); | |
963 | ||
964 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f | 965 | |
9babb374 BB |
966 | /* |
967 | * Release the vnode asynchronously as we currently have the | |
968 | * txg stopped from syncing. | |
969 | */ | |
0a50679c | 970 | zfs_iput_async(ZTOI(zp)); |
428870ff BB |
971 | |
972 | if (error == 0 && zgd->zgd_bp) | |
973 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
974 | ||
34dc7c2f BB |
975 | kmem_free(zgd, sizeof (zgd_t)); |
976 | } | |
977 | ||
45d1cae3 BB |
978 | #ifdef DEBUG |
979 | static int zil_fault_io = 0; | |
980 | #endif | |
981 | ||
34dc7c2f BB |
982 | /* |
983 | * Get data to generate a TX_WRITE intent log record. | |
984 | */ | |
985 | int | |
986 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
987 | { | |
3558fd73 BB |
988 | zfs_sb_t *zsb = arg; |
989 | objset_t *os = zsb->z_os; | |
34dc7c2f | 990 | znode_t *zp; |
428870ff BB |
991 | uint64_t object = lr->lr_foid; |
992 | uint64_t offset = lr->lr_offset; | |
993 | uint64_t size = lr->lr_length; | |
994 | blkptr_t *bp = &lr->lr_blkptr; | |
34dc7c2f | 995 | dmu_buf_t *db; |
34dc7c2f | 996 | zgd_t *zgd; |
34dc7c2f BB |
997 | int error = 0; |
998 | ||
428870ff BB |
999 | ASSERT(zio != NULL); |
1000 | ASSERT(size != 0); | |
34dc7c2f BB |
1001 | |
1002 | /* | |
1003 | * Nothing to do if the file has been removed | |
1004 | */ | |
3558fd73 | 1005 | if (zfs_zget(zsb, object, &zp) != 0) |
2e528b49 | 1006 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 1007 | if (zp->z_unlinked) { |
9babb374 BB |
1008 | /* |
1009 | * Release the vnode asynchronously as we currently have the | |
1010 | * txg stopped from syncing. | |
1011 | */ | |
0a50679c | 1012 | zfs_iput_async(ZTOI(zp)); |
2e528b49 | 1013 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1014 | } |
1015 | ||
79c76d5b | 1016 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
3558fd73 | 1017 | zgd->zgd_zilog = zsb->z_log; |
428870ff BB |
1018 | zgd->zgd_private = zp; |
1019 | ||
34dc7c2f BB |
1020 | /* |
1021 | * Write records come in two flavors: immediate and indirect. | |
1022 | * For small writes it's cheaper to store the data with the | |
1023 | * log record (immediate); for large writes it's cheaper to | |
1024 | * sync the data and get a pointer to it (indirect) so that | |
1025 | * we don't have to write the data twice. | |
1026 | */ | |
1027 | if (buf != NULL) { /* immediate write */ | |
d88895a0 CC |
1028 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, size, |
1029 | RL_READER); | |
34dc7c2f | 1030 | /* test for truncation needs to be done while range locked */ |
428870ff | 1031 | if (offset >= zp->z_size) { |
2e528b49 | 1032 | error = SET_ERROR(ENOENT); |
428870ff BB |
1033 | } else { |
1034 | error = dmu_read(os, object, offset, size, buf, | |
1035 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1036 | } |
428870ff | 1037 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1038 | } else { /* indirect write */ |
34dc7c2f BB |
1039 | /* |
1040 | * Have to lock the whole block to ensure when it's | |
1041 | * written out and it's checksum is being calculated | |
1042 | * that no one can change the data. We need to re-check | |
1043 | * blocksize after we get the lock in case it's changed! | |
1044 | */ | |
1045 | for (;;) { | |
428870ff BB |
1046 | uint64_t blkoff; |
1047 | size = zp->z_blksz; | |
1048 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1049 | offset -= blkoff; | |
d88895a0 CC |
1050 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, |
1051 | size, RL_READER); | |
428870ff | 1052 | if (zp->z_blksz == size) |
34dc7c2f | 1053 | break; |
428870ff BB |
1054 | offset += blkoff; |
1055 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f BB |
1056 | } |
1057 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1058 | if (lr->lr_offset >= zp->z_size) |
2e528b49 | 1059 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1060 | #ifdef DEBUG |
1061 | if (zil_fault_io) { | |
2e528b49 | 1062 | error = SET_ERROR(EIO); |
45d1cae3 | 1063 | zil_fault_io = 0; |
45d1cae3 | 1064 | } |
45d1cae3 | 1065 | #endif |
34dc7c2f | 1066 | if (error == 0) |
428870ff BB |
1067 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1068 | DMU_READ_NO_PREFETCH); | |
1069 | ||
1070 | if (error == 0) { | |
03c6040b GW |
1071 | blkptr_t *obp = dmu_buf_get_blkptr(db); |
1072 | if (obp) { | |
1073 | ASSERT(BP_IS_HOLE(bp)); | |
1074 | *bp = *obp; | |
1075 | } | |
1076 | ||
428870ff BB |
1077 | zgd->zgd_db = db; |
1078 | zgd->zgd_bp = bp; | |
1079 | ||
1080 | ASSERT(db->db_offset == offset); | |
1081 | ASSERT(db->db_size == size); | |
1082 | ||
1083 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1084 | zfs_get_done, zgd); | |
1085 | ASSERT(error || lr->lr_length <= zp->z_blksz); | |
1086 | ||
1087 | /* | |
1088 | * On success, we need to wait for the write I/O | |
1089 | * initiated by dmu_sync() to complete before we can | |
1090 | * release this dbuf. We will finish everything up | |
1091 | * in the zfs_get_done() callback. | |
1092 | */ | |
1093 | if (error == 0) | |
1094 | return (0); | |
1095 | ||
1096 | if (error == EALREADY) { | |
1097 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
1098 | error = 0; | |
1099 | } | |
1100 | } | |
34dc7c2f | 1101 | } |
428870ff BB |
1102 | |
1103 | zfs_get_done(zgd, error); | |
1104 | ||
34dc7c2f BB |
1105 | return (error); |
1106 | } | |
1107 | ||
1108 | /*ARGSUSED*/ | |
3558fd73 BB |
1109 | int |
1110 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1111 | { |
3558fd73 BB |
1112 | znode_t *zp = ITOZ(ip); |
1113 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
1114 | int error; |
1115 | ||
3558fd73 | 1116 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1117 | ZFS_VERIFY_ZP(zp); |
1118 | ||
1119 | if (flag & V_ACE_MASK) | |
1120 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1121 | else | |
1122 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1123 | ||
3558fd73 | 1124 | ZFS_EXIT(zsb); |
45d1cae3 BB |
1125 | return (error); |
1126 | } | |
3558fd73 | 1127 | EXPORT_SYMBOL(zfs_access); |
45d1cae3 | 1128 | |
34dc7c2f BB |
1129 | /* |
1130 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1131 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1132 | * |
3558fd73 | 1133 | * IN: dip - inode of directory to search. |
34dc7c2f | 1134 | * nm - name of entry to lookup. |
34dc7c2f | 1135 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1136 | * cr - credentials of caller. |
34dc7c2f BB |
1137 | * direntflags - directory lookup flags |
1138 | * realpnp - returned pathname. | |
1139 | * | |
3558fd73 | 1140 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1141 | * |
d3cc8b15 | 1142 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1143 | * |
1144 | * Timestamps: | |
1145 | * NA | |
1146 | */ | |
1147 | /* ARGSUSED */ | |
e5c39b95 | 1148 | int |
3558fd73 BB |
1149 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1150 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1151 | { |
3558fd73 BB |
1152 | znode_t *zdp = ITOZ(dip); |
1153 | zfs_sb_t *zsb = ITOZSB(dip); | |
1154 | int error = 0; | |
45d1cae3 BB |
1155 | |
1156 | /* fast path */ | |
1157 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { | |
1158 | ||
3558fd73 | 1159 | if (!S_ISDIR(dip->i_mode)) { |
2e528b49 | 1160 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1161 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 1162 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1163 | } |
1164 | ||
1165 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1166 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1167 | if (!error) { | |
3558fd73 BB |
1168 | *ipp = dip; |
1169 | igrab(*ipp); | |
45d1cae3 BB |
1170 | return (0); |
1171 | } | |
1172 | return (error); | |
3558fd73 | 1173 | #ifdef HAVE_DNLC |
45d1cae3 BB |
1174 | } else { |
1175 | vnode_t *tvp = dnlc_lookup(dvp, nm); | |
1176 | ||
1177 | if (tvp) { | |
1178 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1179 | if (error) { | |
3558fd73 | 1180 | iput(tvp); |
45d1cae3 BB |
1181 | return (error); |
1182 | } | |
1183 | if (tvp == DNLC_NO_VNODE) { | |
3558fd73 | 1184 | iput(tvp); |
2e528b49 | 1185 | return (SET_ERROR(ENOENT)); |
45d1cae3 BB |
1186 | } else { |
1187 | *vpp = tvp; | |
1188 | return (specvp_check(vpp, cr)); | |
1189 | } | |
1190 | } | |
3558fd73 | 1191 | #endif /* HAVE_DNLC */ |
45d1cae3 BB |
1192 | } |
1193 | } | |
1194 | ||
3558fd73 | 1195 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1196 | ZFS_VERIFY_ZP(zdp); |
1197 | ||
3558fd73 | 1198 | *ipp = NULL; |
34dc7c2f BB |
1199 | |
1200 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1201 | /* |
1202 | * We don't allow recursive attributes.. | |
1203 | * Maybe someday we will. | |
1204 | */ | |
428870ff | 1205 | if (zdp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1206 | ZFS_EXIT(zsb); |
2e528b49 | 1207 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1208 | } |
1209 | ||
3558fd73 BB |
1210 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
1211 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1212 | return (error); |
1213 | } | |
1214 | ||
1215 | /* | |
1216 | * Do we have permission to get into attribute directory? | |
1217 | */ | |
1218 | ||
3558fd73 | 1219 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1220 | B_FALSE, cr))) { |
3558fd73 BB |
1221 | iput(*ipp); |
1222 | *ipp = NULL; | |
34dc7c2f BB |
1223 | } |
1224 | ||
3558fd73 | 1225 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1226 | return (error); |
1227 | } | |
1228 | ||
3558fd73 BB |
1229 | if (!S_ISDIR(dip->i_mode)) { |
1230 | ZFS_EXIT(zsb); | |
2e528b49 | 1231 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1232 | } |
1233 | ||
1234 | /* | |
1235 | * Check accessibility of directory. | |
1236 | */ | |
1237 | ||
149e873a | 1238 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
3558fd73 | 1239 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1240 | return (error); |
1241 | } | |
1242 | ||
3558fd73 | 1243 | if (zsb->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1244 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1245 | ZFS_EXIT(zsb); |
2e528b49 | 1246 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1247 | } |
1248 | ||
3558fd73 BB |
1249 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1250 | if ((error == 0) && (*ipp)) | |
1251 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1252 | |
3558fd73 | 1253 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1254 | return (error); |
1255 | } | |
e5c39b95 | 1256 | EXPORT_SYMBOL(zfs_lookup); |
34dc7c2f BB |
1257 | |
1258 | /* | |
1259 | * Attempt to create a new entry in a directory. If the entry | |
1260 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1261 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1262 | * |
3558fd73 | 1263 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1264 | * name - name of new file entry. |
1265 | * vap - attributes of new file. | |
1266 | * excl - flag indicating exclusive or non-exclusive mode. | |
1267 | * mode - mode to open file with. | |
1268 | * cr - credentials of caller. | |
1269 | * flag - large file flag [UNUSED]. | |
3558fd73 | 1270 | * vsecp - ACL to be set |
34dc7c2f | 1271 | * |
3558fd73 | 1272 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1273 | * |
d3cc8b15 | 1274 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1275 | * |
1276 | * Timestamps: | |
3558fd73 BB |
1277 | * dip - ctime|mtime updated if new entry created |
1278 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1279 | */ |
1280 | ||
1281 | /* ARGSUSED */ | |
e5c39b95 | 1282 | int |
3558fd73 BB |
1283 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1284 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1285 | { |
3558fd73 BB |
1286 | znode_t *zp, *dzp = ITOZ(dip); |
1287 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1288 | zilog_t *zilog; |
1289 | objset_t *os; | |
1290 | zfs_dirlock_t *dl; | |
1291 | dmu_tx_t *tx; | |
1292 | int error; | |
b128c09f | 1293 | uid_t uid; |
149e873a | 1294 | gid_t gid; |
428870ff | 1295 | zfs_acl_ids_t acl_ids; |
9babb374 | 1296 | boolean_t fuid_dirtied; |
428870ff | 1297 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 1298 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1299 | |
1300 | /* | |
1301 | * If we have an ephemeral id, ACL, or XVATTR then | |
1302 | * make sure file system is at proper version | |
1303 | */ | |
1304 | ||
149e873a | 1305 | gid = crgetgid(cr); |
3558fd73 | 1306 | uid = crgetuid(cr); |
b128c09f | 1307 | |
3558fd73 BB |
1308 | if (zsb->z_use_fuids == B_FALSE && |
1309 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
2e528b49 | 1310 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1311 | |
3558fd73 | 1312 | ZFS_ENTER(zsb); |
34dc7c2f | 1313 | ZFS_VERIFY_ZP(dzp); |
3558fd73 BB |
1314 | os = zsb->z_os; |
1315 | zilog = zsb->z_log; | |
34dc7c2f | 1316 | |
3558fd73 | 1317 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1318 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1319 | ZFS_EXIT(zsb); |
2e528b49 | 1320 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1321 | } |
1322 | ||
5484965a | 1323 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1324 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1325 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1326 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1327 | return (error); |
1328 | } | |
1329 | } | |
34dc7c2f | 1330 | |
3558fd73 BB |
1331 | top: |
1332 | *ipp = NULL; | |
34dc7c2f BB |
1333 | if (*name == '\0') { |
1334 | /* | |
1335 | * Null component name refers to the directory itself. | |
1336 | */ | |
3558fd73 | 1337 | igrab(dip); |
34dc7c2f BB |
1338 | zp = dzp; |
1339 | dl = NULL; | |
1340 | error = 0; | |
1341 | } else { | |
3558fd73 | 1342 | /* possible igrab(zp) */ |
34dc7c2f BB |
1343 | int zflg = 0; |
1344 | ||
1345 | if (flag & FIGNORECASE) | |
1346 | zflg |= ZCILOOK; | |
1347 | ||
1348 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1349 | NULL, NULL); | |
1350 | if (error) { | |
572e2857 BB |
1351 | if (have_acl) |
1352 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1353 | if (strcmp(name, "..") == 0) |
2e528b49 | 1354 | error = SET_ERROR(EISDIR); |
3558fd73 | 1355 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1356 | return (error); |
1357 | } | |
1358 | } | |
428870ff | 1359 | |
34dc7c2f BB |
1360 | if (zp == NULL) { |
1361 | uint64_t txtype; | |
1362 | ||
1363 | /* | |
1364 | * Create a new file object and update the directory | |
1365 | * to reference it. | |
1366 | */ | |
149e873a | 1367 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1368 | if (have_acl) |
1369 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1370 | goto out; |
1371 | } | |
1372 | ||
1373 | /* | |
1374 | * We only support the creation of regular files in | |
1375 | * extended attribute directories. | |
1376 | */ | |
428870ff | 1377 | |
3558fd73 | 1378 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1379 | if (have_acl) |
1380 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 1381 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1382 | goto out; |
1383 | } | |
1384 | ||
428870ff BB |
1385 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1386 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1387 | goto out; |
428870ff BB |
1388 | have_acl = B_TRUE; |
1389 | ||
3558fd73 | 1390 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1391 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1392 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1393 | goto out; |
1394 | } | |
1395 | ||
34dc7c2f | 1396 | tx = dmu_tx_create(os); |
428870ff BB |
1397 | |
1398 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1399 | ZFS_SA_BASE_ATTR_SIZE); | |
1400 | ||
3558fd73 | 1401 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1402 | if (fuid_dirtied) |
3558fd73 | 1403 | zfs_fuid_txhold(zsb, tx); |
34dc7c2f | 1404 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1405 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
3558fd73 | 1406 | if (!zsb->z_use_sa && |
428870ff | 1407 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1408 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1409 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1410 | } |
e8b96c60 | 1411 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1412 | if (error) { |
1413 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1414 | if (error == ERESTART) { |
e8b96c60 | 1415 | waited = B_TRUE; |
34dc7c2f BB |
1416 | dmu_tx_wait(tx); |
1417 | dmu_tx_abort(tx); | |
1418 | goto top; | |
1419 | } | |
428870ff | 1420 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1421 | dmu_tx_abort(tx); |
3558fd73 | 1422 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1423 | return (error); |
1424 | } | |
428870ff | 1425 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 BB |
1426 | |
1427 | if (fuid_dirtied) | |
3558fd73 | 1428 | zfs_fuid_sync(zsb, tx); |
9babb374 | 1429 | |
34dc7c2f BB |
1430 | (void) zfs_link_create(dl, zp, tx, ZNEW); |
1431 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); | |
1432 | if (flag & FIGNORECASE) | |
1433 | txtype |= TX_CI; | |
1434 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1435 | vsecp, acl_ids.z_fuidp, vap); |
1436 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1437 | dmu_tx_commit(tx); |
1438 | } else { | |
1439 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1440 | ||
572e2857 BB |
1441 | if (have_acl) |
1442 | zfs_acl_ids_free(&acl_ids); | |
1443 | have_acl = B_FALSE; | |
1444 | ||
34dc7c2f BB |
1445 | /* |
1446 | * A directory entry already exists for this name. | |
1447 | */ | |
1448 | /* | |
1449 | * Can't truncate an existing file if in exclusive mode. | |
1450 | */ | |
3558fd73 | 1451 | if (excl) { |
2e528b49 | 1452 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1453 | goto out; |
1454 | } | |
1455 | /* | |
1456 | * Can't open a directory for writing. | |
1457 | */ | |
3558fd73 | 1458 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1459 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1460 | goto out; |
1461 | } | |
1462 | /* | |
1463 | * Verify requested access to file. | |
1464 | */ | |
1465 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1466 | goto out; | |
1467 | } | |
1468 | ||
1469 | mutex_enter(&dzp->z_lock); | |
1470 | dzp->z_seq++; | |
1471 | mutex_exit(&dzp->z_lock); | |
1472 | ||
1473 | /* | |
1474 | * Truncate regular files if requested. | |
1475 | */ | |
3558fd73 BB |
1476 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1477 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 1478 | /* we can't hold any locks when calling zfs_freesp() */ |
609603a5 B |
1479 | if (dl) { |
1480 | zfs_dirent_unlock(dl); | |
1481 | dl = NULL; | |
1482 | } | |
34dc7c2f | 1483 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1484 | } |
1485 | } | |
1486 | out: | |
1487 | ||
1488 | if (dl) | |
1489 | zfs_dirent_unlock(dl); | |
1490 | ||
1491 | if (error) { | |
1492 | if (zp) | |
3558fd73 | 1493 | iput(ZTOI(zp)); |
34dc7c2f | 1494 | } else { |
960e08fe BB |
1495 | zfs_inode_update(dzp); |
1496 | zfs_inode_update(zp); | |
3558fd73 | 1497 | *ipp = ZTOI(zp); |
34dc7c2f | 1498 | } |
34dc7c2f | 1499 | |
3558fd73 | 1500 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1501 | zil_commit(zilog, 0); |
428870ff | 1502 | |
3558fd73 | 1503 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1504 | return (error); |
1505 | } | |
e5c39b95 | 1506 | EXPORT_SYMBOL(zfs_create); |
34dc7c2f BB |
1507 | |
1508 | /* | |
1509 | * Remove an entry from a directory. | |
1510 | * | |
3558fd73 | 1511 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1512 | * name - name of entry to remove. |
1513 | * cr - credentials of caller. | |
34dc7c2f BB |
1514 | * |
1515 | * RETURN: 0 if success | |
1516 | * error code if failure | |
1517 | * | |
1518 | * Timestamps: | |
3558fd73 BB |
1519 | * dip - ctime|mtime |
1520 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1521 | */ |
428870ff BB |
1522 | |
1523 | uint64_t null_xattr = 0; | |
1524 | ||
34dc7c2f | 1525 | /*ARGSUSED*/ |
e5c39b95 | 1526 | int |
da5e151f | 1527 | zfs_remove(struct inode *dip, char *name, cred_t *cr, int flags) |
34dc7c2f | 1528 | { |
3558fd73 | 1529 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1530 | znode_t *xzp; |
3558fd73 BB |
1531 | struct inode *ip; |
1532 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f | 1533 | zilog_t *zilog; |
a966c564 | 1534 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 1535 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 1536 | uint64_t obj = 0; |
dfbc8630 | 1537 | uint64_t links; |
34dc7c2f BB |
1538 | zfs_dirlock_t *dl; |
1539 | dmu_tx_t *tx; | |
a966c564 K |
1540 | boolean_t may_delete_now, delete_now = FALSE; |
1541 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
1542 | uint64_t txtype; |
1543 | pathname_t *realnmp = NULL; | |
1544 | pathname_t realnm; | |
1545 | int error; | |
1546 | int zflg = ZEXISTS; | |
e8b96c60 | 1547 | boolean_t waited = B_FALSE; |
34dc7c2f | 1548 | |
3558fd73 | 1549 | ZFS_ENTER(zsb); |
34dc7c2f | 1550 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1551 | zilog = zsb->z_log; |
34dc7c2f BB |
1552 | |
1553 | if (flags & FIGNORECASE) { | |
1554 | zflg |= ZCILOOK; | |
1555 | pn_alloc(&realnm); | |
1556 | realnmp = &realnm; | |
1557 | } | |
1558 | ||
1559 | top: | |
572e2857 BB |
1560 | xattr_obj = 0; |
1561 | xzp = NULL; | |
34dc7c2f BB |
1562 | /* |
1563 | * Attempt to lock directory; fail if entry doesn't exist. | |
1564 | */ | |
149e873a BB |
1565 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1566 | NULL, realnmp))) { | |
34dc7c2f BB |
1567 | if (realnmp) |
1568 | pn_free(realnmp); | |
3558fd73 | 1569 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1570 | return (error); |
1571 | } | |
1572 | ||
3558fd73 | 1573 | ip = ZTOI(zp); |
34dc7c2f | 1574 | |
149e873a | 1575 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1576 | goto out; |
1577 | } | |
1578 | ||
1579 | /* | |
1580 | * Need to use rmdir for removing directories. | |
1581 | */ | |
3558fd73 | 1582 | if (S_ISDIR(ip->i_mode)) { |
2e528b49 | 1583 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1584 | goto out; |
1585 | } | |
1586 | ||
3558fd73 | 1587 | #ifdef HAVE_DNLC |
34dc7c2f BB |
1588 | if (realnmp) |
1589 | dnlc_remove(dvp, realnmp->pn_buf); | |
1590 | else | |
1591 | dnlc_remove(dvp, name); | |
3558fd73 | 1592 | #endif /* HAVE_DNLC */ |
34dc7c2f | 1593 | |
19d55079 MA |
1594 | mutex_enter(&zp->z_lock); |
1595 | may_delete_now = atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped); | |
1596 | mutex_exit(&zp->z_lock); | |
1597 | ||
34dc7c2f | 1598 | /* |
a966c564 K |
1599 | * We may delete the znode now, or we may put it in the unlinked set; |
1600 | * it depends on whether we're the last link, and on whether there are | |
1601 | * other holds on the inode. So we dmu_tx_hold() the right things to | |
1602 | * allow for either case. | |
34dc7c2f | 1603 | */ |
572e2857 | 1604 | obj = zp->z_id; |
3558fd73 | 1605 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 1606 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1607 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1608 | zfs_sa_upgrade_txholds(tx, zp); | |
1609 | zfs_sa_upgrade_txholds(tx, dzp); | |
a966c564 K |
1610 | if (may_delete_now) { |
1611 | toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; | |
1612 | /* if the file is too big, only hold_free a token amount */ | |
1613 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1614 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1615 | } | |
34dc7c2f BB |
1616 | |
1617 | /* are there any extended attributes? */ | |
3558fd73 | 1618 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1619 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1620 | if (error == 0 && xattr_obj) { |
3558fd73 | 1621 | error = zfs_zget(zsb, xattr_obj, &xzp); |
c99c9001 | 1622 | ASSERT0(error); |
428870ff BB |
1623 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1624 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1625 | } |
1626 | ||
a966c564 K |
1627 | mutex_enter(&zp->z_lock); |
1628 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
1629 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
1630 | mutex_exit(&zp->z_lock); | |
1631 | ||
34dc7c2f | 1632 | /* charge as an update -- would be nice not to charge at all */ |
3558fd73 | 1633 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1634 | |
19d55079 | 1635 | /* |
1a04bab3 | 1636 | * Mark this transaction as typically resulting in a net free of space |
19d55079 | 1637 | */ |
1a04bab3 | 1638 | dmu_tx_mark_netfree(tx); |
19d55079 | 1639 | |
e8b96c60 | 1640 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1641 | if (error) { |
1642 | zfs_dirent_unlock(dl); | |
3558fd73 | 1643 | iput(ip); |
572e2857 | 1644 | if (xzp) |
3558fd73 | 1645 | iput(ZTOI(xzp)); |
fb5f0bc8 | 1646 | if (error == ERESTART) { |
e8b96c60 | 1647 | waited = B_TRUE; |
34dc7c2f BB |
1648 | dmu_tx_wait(tx); |
1649 | dmu_tx_abort(tx); | |
1650 | goto top; | |
1651 | } | |
1652 | if (realnmp) | |
1653 | pn_free(realnmp); | |
1654 | dmu_tx_abort(tx); | |
3558fd73 | 1655 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1656 | return (error); |
1657 | } | |
1658 | ||
1659 | /* | |
1660 | * Remove the directory entry. | |
1661 | */ | |
1662 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1663 | ||
1664 | if (error) { | |
1665 | dmu_tx_commit(tx); | |
1666 | goto out; | |
1667 | } | |
1668 | ||
1669 | if (unlinked) { | |
572e2857 BB |
1670 | /* |
1671 | * Hold z_lock so that we can make sure that the ACL obj | |
1672 | * hasn't changed. Could have been deleted due to | |
1673 | * zfs_sa_upgrade(). | |
1674 | */ | |
1675 | mutex_enter(&zp->z_lock); | |
3558fd73 | 1676 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1677 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
a966c564 K |
1678 | delete_now = may_delete_now && !toobig && |
1679 | atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped) && | |
1680 | xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == | |
1681 | acl_obj; | |
1682 | } | |
1683 | ||
1684 | if (delete_now) { | |
1685 | if (xattr_obj_unlinked) { | |
dfbc8630 | 1686 | ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); |
a966c564 K |
1687 | mutex_enter(&xzp->z_lock); |
1688 | xzp->z_unlinked = 1; | |
dfbc8630 CD |
1689 | clear_nlink(ZTOI(xzp)); |
1690 | links = 0; | |
a966c564 | 1691 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zsb), |
dfbc8630 | 1692 | &links, sizeof (links), tx); |
a966c564 K |
1693 | ASSERT3U(error, ==, 0); |
1694 | mutex_exit(&xzp->z_lock); | |
1695 | zfs_unlinked_add(xzp, tx); | |
1696 | ||
1697 | if (zp->z_is_sa) | |
1698 | error = sa_remove(zp->z_sa_hdl, | |
1699 | SA_ZPL_XATTR(zsb), tx); | |
1700 | else | |
1701 | error = sa_update(zp->z_sa_hdl, | |
1702 | SA_ZPL_XATTR(zsb), &null_xattr, | |
1703 | sizeof (uint64_t), tx); | |
1704 | ASSERT0(error); | |
1705 | } | |
1706 | /* | |
1707 | * Add to the unlinked set because a new reference could be | |
1708 | * taken concurrently resulting in a deferred destruction. | |
1709 | */ | |
1710 | zfs_unlinked_add(zp, tx); | |
1711 | mutex_exit(&zp->z_lock); | |
1712 | zfs_inode_update(zp); | |
1713 | iput(ip); | |
1714 | } else if (unlinked) { | |
572e2857 | 1715 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1716 | zfs_unlinked_add(zp, tx); |
1717 | } | |
1718 | ||
1719 | txtype = TX_REMOVE; | |
1720 | if (flags & FIGNORECASE) | |
1721 | txtype |= TX_CI; | |
572e2857 | 1722 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1723 | |
1724 | dmu_tx_commit(tx); | |
1725 | out: | |
1726 | if (realnmp) | |
1727 | pn_free(realnmp); | |
1728 | ||
1729 | zfs_dirent_unlock(dl); | |
960e08fe | 1730 | zfs_inode_update(dzp); |
34dc7c2f | 1731 | |
a966c564 K |
1732 | if (!delete_now) { |
1733 | zfs_inode_update(zp); | |
1734 | zfs_iput_async(ip); | |
1735 | } | |
1736 | ||
1737 | if (xzp) { | |
1738 | zfs_inode_update(xzp); | |
1739 | zfs_iput_async(ZTOI(xzp)); | |
1740 | } | |
428870ff | 1741 | |
3558fd73 | 1742 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1743 | zil_commit(zilog, 0); |
34dc7c2f | 1744 | |
3558fd73 | 1745 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1746 | return (error); |
1747 | } | |
e5c39b95 | 1748 | EXPORT_SYMBOL(zfs_remove); |
34dc7c2f BB |
1749 | |
1750 | /* | |
3558fd73 | 1751 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1752 | * provided. Return a pointer to the inserted directory. |
1753 | * | |
3558fd73 | 1754 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1755 | * dirname - name of new directory. |
1756 | * vap - attributes of new directory. | |
1757 | * cr - credentials of caller. | |
34dc7c2f BB |
1758 | * vsecp - ACL to be set |
1759 | * | |
3558fd73 | 1760 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1761 | * |
1762 | * RETURN: 0 if success | |
1763 | * error code if failure | |
1764 | * | |
1765 | * Timestamps: | |
3558fd73 BB |
1766 | * dip - ctime|mtime updated |
1767 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1768 | */ |
1769 | /*ARGSUSED*/ | |
e5c39b95 | 1770 | int |
3558fd73 BB |
1771 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1772 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1773 | { |
3558fd73 BB |
1774 | znode_t *zp, *dzp = ITOZ(dip); |
1775 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1776 | zilog_t *zilog; |
1777 | zfs_dirlock_t *dl; | |
1778 | uint64_t txtype; | |
1779 | dmu_tx_t *tx; | |
1780 | int error; | |
34dc7c2f | 1781 | int zf = ZNEW; |
b128c09f BB |
1782 | uid_t uid; |
1783 | gid_t gid = crgetgid(cr); | |
428870ff | 1784 | zfs_acl_ids_t acl_ids; |
9babb374 | 1785 | boolean_t fuid_dirtied; |
e8b96c60 | 1786 | boolean_t waited = B_FALSE; |
34dc7c2f | 1787 | |
3558fd73 | 1788 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1789 | |
1790 | /* | |
1791 | * If we have an ephemeral id, ACL, or XVATTR then | |
1792 | * make sure file system is at proper version | |
1793 | */ | |
1794 | ||
3558fd73 BB |
1795 | uid = crgetuid(cr); |
1796 | if (zsb->z_use_fuids == B_FALSE && | |
1797 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
2e528b49 | 1798 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1799 | |
3558fd73 | 1800 | ZFS_ENTER(zsb); |
34dc7c2f | 1801 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1802 | zilog = zsb->z_log; |
34dc7c2f | 1803 | |
428870ff | 1804 | if (dzp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1805 | ZFS_EXIT(zsb); |
2e528b49 | 1806 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1807 | } |
1808 | ||
3558fd73 | 1809 | if (zsb->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1810 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1811 | ZFS_EXIT(zsb); |
2e528b49 | 1812 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1813 | } |
1814 | if (flags & FIGNORECASE) | |
1815 | zf |= ZCILOOK; | |
1816 | ||
5484965a | 1817 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1818 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1819 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1820 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1821 | return (error); |
1822 | } | |
428870ff | 1823 | } |
34dc7c2f | 1824 | |
428870ff BB |
1825 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1826 | vsecp, &acl_ids)) != 0) { | |
3558fd73 | 1827 | ZFS_EXIT(zsb); |
428870ff BB |
1828 | return (error); |
1829 | } | |
34dc7c2f BB |
1830 | /* |
1831 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1832 | * |
1833 | * Existence is checked first to make sure we don't return | |
1834 | * EACCES instead of EEXIST which can cause some applications | |
1835 | * to fail. | |
34dc7c2f BB |
1836 | */ |
1837 | top: | |
3558fd73 | 1838 | *ipp = NULL; |
34dc7c2f | 1839 | |
149e873a BB |
1840 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1841 | NULL, NULL))) { | |
428870ff | 1842 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 1843 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1844 | return (error); |
1845 | } | |
1846 | ||
149e873a | 1847 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 1848 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1849 | zfs_dirent_unlock(dl); |
3558fd73 | 1850 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1851 | return (error); |
1852 | } | |
1853 | ||
3558fd73 | 1854 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1855 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 1856 | zfs_dirent_unlock(dl); |
3558fd73 | 1857 | ZFS_EXIT(zsb); |
2e528b49 | 1858 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
1859 | } |
1860 | ||
34dc7c2f BB |
1861 | /* |
1862 | * Add a new entry to the directory. | |
1863 | */ | |
3558fd73 | 1864 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f BB |
1865 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
1866 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
3558fd73 | 1867 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1868 | if (fuid_dirtied) |
3558fd73 BB |
1869 | zfs_fuid_txhold(zsb, tx); |
1870 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
1871 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
1872 | acl_ids.z_aclp->z_acl_bytes); | |
1873 | } | |
1874 | ||
1875 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1876 | ZFS_SA_BASE_ATTR_SIZE); | |
1877 | ||
e8b96c60 | 1878 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1879 | if (error) { |
1880 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1881 | if (error == ERESTART) { |
e8b96c60 | 1882 | waited = B_TRUE; |
34dc7c2f BB |
1883 | dmu_tx_wait(tx); |
1884 | dmu_tx_abort(tx); | |
1885 | goto top; | |
1886 | } | |
428870ff | 1887 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1888 | dmu_tx_abort(tx); |
3558fd73 | 1889 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1890 | return (error); |
1891 | } | |
1892 | ||
1893 | /* | |
1894 | * Create new node. | |
1895 | */ | |
428870ff | 1896 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f | 1897 | |
9babb374 | 1898 | if (fuid_dirtied) |
3558fd73 | 1899 | zfs_fuid_sync(zsb, tx); |
428870ff | 1900 | |
34dc7c2f BB |
1901 | /* |
1902 | * Now put new name in parent dir. | |
1903 | */ | |
1904 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
1905 | ||
3558fd73 | 1906 | *ipp = ZTOI(zp); |
34dc7c2f BB |
1907 | |
1908 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
1909 | if (flags & FIGNORECASE) | |
1910 | txtype |= TX_CI; | |
9babb374 BB |
1911 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
1912 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 1913 | |
9babb374 | 1914 | zfs_acl_ids_free(&acl_ids); |
428870ff | 1915 | |
34dc7c2f BB |
1916 | dmu_tx_commit(tx); |
1917 | ||
1918 | zfs_dirent_unlock(dl); | |
1919 | ||
3558fd73 | 1920 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1921 | zil_commit(zilog, 0); |
428870ff | 1922 | |
960e08fe BB |
1923 | zfs_inode_update(dzp); |
1924 | zfs_inode_update(zp); | |
3558fd73 | 1925 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1926 | return (0); |
1927 | } | |
e5c39b95 | 1928 | EXPORT_SYMBOL(zfs_mkdir); |
34dc7c2f BB |
1929 | |
1930 | /* | |
1931 | * Remove a directory subdir entry. If the current working | |
1932 | * directory is the same as the subdir to be removed, the | |
1933 | * remove will fail. | |
1934 | * | |
3558fd73 | 1935 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 1936 | * name - name of directory to be removed. |
3558fd73 | 1937 | * cwd - inode of current working directory. |
34dc7c2f | 1938 | * cr - credentials of caller. |
34dc7c2f BB |
1939 | * flags - case flags |
1940 | * | |
d3cc8b15 | 1941 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1942 | * |
1943 | * Timestamps: | |
3558fd73 | 1944 | * dip - ctime|mtime updated |
34dc7c2f BB |
1945 | */ |
1946 | /*ARGSUSED*/ | |
e5c39b95 | 1947 | int |
3558fd73 BB |
1948 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
1949 | int flags) | |
34dc7c2f | 1950 | { |
3558fd73 | 1951 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 1952 | znode_t *zp; |
3558fd73 BB |
1953 | struct inode *ip; |
1954 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1955 | zilog_t *zilog; |
1956 | zfs_dirlock_t *dl; | |
1957 | dmu_tx_t *tx; | |
1958 | int error; | |
1959 | int zflg = ZEXISTS; | |
e8b96c60 | 1960 | boolean_t waited = B_FALSE; |
34dc7c2f | 1961 | |
3558fd73 | 1962 | ZFS_ENTER(zsb); |
34dc7c2f | 1963 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1964 | zilog = zsb->z_log; |
34dc7c2f BB |
1965 | |
1966 | if (flags & FIGNORECASE) | |
1967 | zflg |= ZCILOOK; | |
1968 | top: | |
1969 | zp = NULL; | |
1970 | ||
1971 | /* | |
1972 | * Attempt to lock directory; fail if entry doesn't exist. | |
1973 | */ | |
149e873a BB |
1974 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1975 | NULL, NULL))) { | |
3558fd73 | 1976 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1977 | return (error); |
1978 | } | |
1979 | ||
3558fd73 | 1980 | ip = ZTOI(zp); |
34dc7c2f | 1981 | |
149e873a | 1982 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1983 | goto out; |
1984 | } | |
1985 | ||
3558fd73 | 1986 | if (!S_ISDIR(ip->i_mode)) { |
2e528b49 | 1987 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
1988 | goto out; |
1989 | } | |
1990 | ||
3558fd73 | 1991 | if (ip == cwd) { |
2e528b49 | 1992 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1993 | goto out; |
1994 | } | |
1995 | ||
34dc7c2f BB |
1996 | /* |
1997 | * Grab a lock on the directory to make sure that noone is | |
1998 | * trying to add (or lookup) entries while we are removing it. | |
1999 | */ | |
2000 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
2001 | ||
2002 | /* | |
2003 | * Grab a lock on the parent pointer to make sure we play well | |
2004 | * with the treewalk and directory rename code. | |
2005 | */ | |
2006 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
2007 | ||
3558fd73 | 2008 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 2009 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 2010 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3558fd73 | 2011 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
2012 | zfs_sa_upgrade_txholds(tx, zp); |
2013 | zfs_sa_upgrade_txholds(tx, dzp); | |
db707ad0 | 2014 | dmu_tx_mark_netfree(tx); |
e8b96c60 | 2015 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
2016 | if (error) { |
2017 | rw_exit(&zp->z_parent_lock); | |
2018 | rw_exit(&zp->z_name_lock); | |
2019 | zfs_dirent_unlock(dl); | |
3558fd73 | 2020 | iput(ip); |
fb5f0bc8 | 2021 | if (error == ERESTART) { |
e8b96c60 | 2022 | waited = B_TRUE; |
34dc7c2f BB |
2023 | dmu_tx_wait(tx); |
2024 | dmu_tx_abort(tx); | |
2025 | goto top; | |
2026 | } | |
2027 | dmu_tx_abort(tx); | |
3558fd73 | 2028 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2029 | return (error); |
2030 | } | |
2031 | ||
2032 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
2033 | ||
2034 | if (error == 0) { | |
2035 | uint64_t txtype = TX_RMDIR; | |
2036 | if (flags & FIGNORECASE) | |
2037 | txtype |= TX_CI; | |
572e2857 | 2038 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
2039 | } |
2040 | ||
2041 | dmu_tx_commit(tx); | |
2042 | ||
2043 | rw_exit(&zp->z_parent_lock); | |
2044 | rw_exit(&zp->z_name_lock); | |
2045 | out: | |
2046 | zfs_dirent_unlock(dl); | |
2047 | ||
59157910 BB |
2048 | zfs_inode_update(dzp); |
2049 | zfs_inode_update(zp); | |
3558fd73 | 2050 | iput(ip); |
34dc7c2f | 2051 | |
3558fd73 | 2052 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2053 | zil_commit(zilog, 0); |
428870ff | 2054 | |
3558fd73 | 2055 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2056 | return (error); |
2057 | } | |
e5c39b95 | 2058 | EXPORT_SYMBOL(zfs_rmdir); |
34dc7c2f BB |
2059 | |
2060 | /* | |
2061 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 2062 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 2063 | * |
3558fd73 BB |
2064 | * IN: ip - inode of directory to read. |
2065 | * dirent - buffer for directory entries. | |
34dc7c2f | 2066 | * |
3558fd73 | 2067 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
2068 | * |
2069 | * RETURN: 0 if success | |
2070 | * error code if failure | |
2071 | * | |
2072 | * Timestamps: | |
3558fd73 | 2073 | * ip - atime updated |
34dc7c2f BB |
2074 | * |
2075 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2076 | * This allows us to use the low range for "special" directory entries: | |
2077 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2078 | * we use the offset 2 for the '.zfs' directory. | |
2079 | */ | |
2080 | /* ARGSUSED */ | |
3558fd73 | 2081 | int |
0f37d0c8 | 2082 | zfs_readdir(struct inode *ip, struct dir_context *ctx, cred_t *cr) |
34dc7c2f | 2083 | { |
3558fd73 BB |
2084 | znode_t *zp = ITOZ(ip); |
2085 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 2086 | objset_t *os; |
34dc7c2f BB |
2087 | zap_cursor_t zc; |
2088 | zap_attribute_t zap; | |
34dc7c2f BB |
2089 | int error; |
2090 | uint8_t prefetch; | |
c12e3a59 | 2091 | uint8_t type; |
3558fd73 BB |
2092 | int done = 0; |
2093 | uint64_t parent; | |
c12e3a59 | 2094 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2095 | |
3558fd73 | 2096 | ZFS_ENTER(zsb); |
34dc7c2f BB |
2097 | ZFS_VERIFY_ZP(zp); |
2098 | ||
3558fd73 BB |
2099 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
2100 | &parent, sizeof (parent))) != 0) | |
2101 | goto out; | |
34dc7c2f BB |
2102 | |
2103 | /* | |
2104 | * Quit if directory has been removed (posix) | |
2105 | */ | |
3558fd73 BB |
2106 | if (zp->z_unlinked) |
2107 | goto out; | |
2108 | ||
c12e3a59 | 2109 | error = 0; |
3558fd73 | 2110 | os = zsb->z_os; |
c12e3a59 | 2111 | offset = ctx->pos; |
34dc7c2f BB |
2112 | prefetch = zp->z_zn_prefetch; |
2113 | ||
2114 | /* | |
2115 | * Initialize the iterator cursor. | |
2116 | */ | |
c12e3a59 | 2117 | if (offset <= 3) { |
34dc7c2f BB |
2118 | /* |
2119 | * Start iteration from the beginning of the directory. | |
2120 | */ | |
2121 | zap_cursor_init(&zc, os, zp->z_id); | |
2122 | } else { | |
2123 | /* | |
2124 | * The offset is a serialized cursor. | |
2125 | */ | |
c12e3a59 | 2126 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2127 | } |
2128 | ||
34dc7c2f BB |
2129 | /* |
2130 | * Transform to file-system independent format | |
2131 | */ | |
3558fd73 BB |
2132 | while (!done) { |
2133 | uint64_t objnum; | |
34dc7c2f BB |
2134 | /* |
2135 | * Special case `.', `..', and `.zfs'. | |
2136 | */ | |
c12e3a59 | 2137 | if (offset == 0) { |
34dc7c2f BB |
2138 | (void) strcpy(zap.za_name, "."); |
2139 | zap.za_normalization_conflict = 0; | |
2140 | objnum = zp->z_id; | |
c12e3a59 RY |
2141 | type = DT_DIR; |
2142 | } else if (offset == 1) { | |
34dc7c2f BB |
2143 | (void) strcpy(zap.za_name, ".."); |
2144 | zap.za_normalization_conflict = 0; | |
428870ff | 2145 | objnum = parent; |
c12e3a59 RY |
2146 | type = DT_DIR; |
2147 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2148 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2149 | zap.za_normalization_conflict = 0; | |
2150 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 2151 | type = DT_DIR; |
34dc7c2f BB |
2152 | } else { |
2153 | /* | |
2154 | * Grab next entry. | |
2155 | */ | |
3558fd73 BB |
2156 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2157 | if (error == ENOENT) | |
34dc7c2f BB |
2158 | break; |
2159 | else | |
2160 | goto update; | |
2161 | } | |
2162 | ||
0c5dde49 BB |
2163 | /* |
2164 | * Allow multiple entries provided the first entry is | |
2165 | * the object id. Non-zpl consumers may safely make | |
2166 | * use of the additional space. | |
2167 | * | |
2168 | * XXX: This should be a feature flag for compatibility | |
2169 | */ | |
34dc7c2f | 2170 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2171 | zap.za_num_integers == 0) { |
34dc7c2f | 2172 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2173 | "entry, obj = %lld, offset = %lld, " |
2174 | "length = %d, num = %lld\n", | |
34dc7c2f | 2175 | (u_longlong_t)zp->z_id, |
c12e3a59 | 2176 | (u_longlong_t)offset, |
0c5dde49 BB |
2177 | zap.za_integer_length, |
2178 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 2179 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2180 | goto update; |
2181 | } | |
2182 | ||
2183 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 2184 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2185 | } |
0f37d0c8 RY |
2186 | |
2187 | done = !dir_emit(ctx, zap.za_name, strlen(zap.za_name), | |
c12e3a59 | 2188 | objnum, type); |
0f37d0c8 | 2189 | if (done) |
34dc7c2f | 2190 | break; |
34dc7c2f BB |
2191 | |
2192 | /* Prefetch znode */ | |
3558fd73 | 2193 | if (prefetch) { |
fcff0f35 PD |
2194 | dmu_prefetch(os, objnum, 0, 0, 0, |
2195 | ZIO_PRIORITY_SYNC_READ); | |
3558fd73 | 2196 | } |
34dc7c2f | 2197 | |
c12e3a59 RY |
2198 | /* |
2199 | * Move to the next entry, fill in the previous offset. | |
2200 | */ | |
2201 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2202 | zap_cursor_advance(&zc); |
c12e3a59 | 2203 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2204 | } else { |
c12e3a59 | 2205 | offset += 1; |
34dc7c2f | 2206 | } |
c12e3a59 | 2207 | ctx->pos = offset; |
34dc7c2f BB |
2208 | } |
2209 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2210 | ||
34dc7c2f BB |
2211 | update: |
2212 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2213 | if (error == ENOENT) |
2214 | error = 0; | |
3558fd73 BB |
2215 | out: |
2216 | ZFS_EXIT(zsb); | |
34dc7c2f | 2217 | |
34dc7c2f BB |
2218 | return (error); |
2219 | } | |
3558fd73 | 2220 | EXPORT_SYMBOL(zfs_readdir); |
34dc7c2f | 2221 | |
d5446cfc BB |
2222 | ulong_t zfs_fsync_sync_cnt = 4; |
2223 | ||
e5c39b95 | 2224 | int |
3558fd73 | 2225 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2226 | { |
3558fd73 BB |
2227 | znode_t *zp = ITOZ(ip); |
2228 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 2229 | |
d5446cfc BB |
2230 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2231 | ||
3558fd73 BB |
2232 | if (zsb->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2233 | ZFS_ENTER(zsb); | |
428870ff | 2234 | ZFS_VERIFY_ZP(zp); |
3558fd73 BB |
2235 | zil_commit(zsb->z_log, zp->z_id); |
2236 | ZFS_EXIT(zsb); | |
428870ff | 2237 | } |
07012da6 CC |
2238 | tsd_set(zfs_fsyncer_key, NULL); |
2239 | ||
34dc7c2f BB |
2240 | return (0); |
2241 | } | |
e5c39b95 | 2242 | EXPORT_SYMBOL(zfs_fsync); |
34dc7c2f BB |
2243 | |
2244 | ||
2245 | /* | |
2246 | * Get the requested file attributes and place them in the provided | |
2247 | * vattr structure. | |
2248 | * | |
3558fd73 | 2249 | * IN: ip - inode of file. |
5484965a BB |
2250 | * vap - va_mask identifies requested attributes. |
2251 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2252 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2253 | * cr - credentials of caller. | |
34dc7c2f | 2254 | * |
5484965a BB |
2255 | * OUT: vap - attribute values. |
2256 | * | |
2257 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2258 | */ |
2259 | /* ARGSUSED */ | |
e5c39b95 | 2260 | int |
5484965a | 2261 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2262 | { |
3558fd73 BB |
2263 | znode_t *zp = ITOZ(ip); |
2264 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2265 | int error = 0; |
2266 | uint64_t links; | |
0df9673f | 2267 | uint64_t atime[2], mtime[2], ctime[2]; |
5484965a BB |
2268 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2269 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2270 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
0df9673f | 2271 | sa_bulk_attr_t bulk[3]; |
428870ff | 2272 | int count = 0; |
34dc7c2f | 2273 | |
3558fd73 | 2274 | ZFS_ENTER(zsb); |
34dc7c2f | 2275 | ZFS_VERIFY_ZP(zp); |
428870ff | 2276 | |
5484965a | 2277 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2278 | |
0df9673f | 2279 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, &atime, 16); |
3558fd73 BB |
2280 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
2281 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
428870ff BB |
2282 | |
2283 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
3558fd73 | 2284 | ZFS_EXIT(zsb); |
428870ff BB |
2285 | return (error); |
2286 | } | |
34dc7c2f | 2287 | |
34dc7c2f BB |
2288 | /* |
2289 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2290 | * Also, if we are the owner don't bother, since owner should | |
2291 | * always be allowed to read basic attributes of file. | |
2292 | */ | |
572e2857 | 2293 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2294 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2295 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2296 | skipaclchk, cr))) { | |
3558fd73 | 2297 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2298 | return (error); |
2299 | } | |
2300 | } | |
2301 | ||
2302 | /* | |
2303 | * Return all attributes. It's cheaper to provide the answer | |
2304 | * than to determine whether we were asked the question. | |
2305 | */ | |
2306 | ||
9babb374 | 2307 | mutex_enter(&zp->z_lock); |
5484965a BB |
2308 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2309 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2310 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2311 | vap->va_nodeid = zp->z_id; |
3558fd73 | 2312 | if ((zp->z_id == zsb->z_root) && zfs_show_ctldir(zp)) |
dfbc8630 | 2313 | links = ZTOI(zp)->i_nlink + 1; |
34dc7c2f | 2314 | else |
dfbc8630 | 2315 | links = ZTOI(zp)->i_nlink; |
5484965a BB |
2316 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2317 | vap->va_size = i_size_read(ip); | |
2318 | vap->va_rdev = ip->i_rdev; | |
2319 | vap->va_seq = ip->i_generation; | |
2320 | ||
2321 | /* | |
2322 | * Add in any requested optional attributes and the create time. | |
2323 | * Also set the corresponding bits in the returned attribute bitmap. | |
2324 | */ | |
2325 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zsb->z_use_fuids) { | |
2326 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
2327 | xoap->xoa_archive = | |
2328 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2329 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2330 | } | |
2331 | ||
2332 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2333 | xoap->xoa_readonly = | |
2334 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2335 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2336 | } | |
2337 | ||
2338 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2339 | xoap->xoa_system = | |
2340 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2341 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2342 | } | |
2343 | ||
2344 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2345 | xoap->xoa_hidden = | |
2346 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2347 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2348 | } | |
2349 | ||
2350 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2351 | xoap->xoa_nounlink = | |
2352 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2353 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2354 | } | |
2355 | ||
2356 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2357 | xoap->xoa_immutable = | |
2358 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2359 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2360 | } | |
2361 | ||
2362 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2363 | xoap->xoa_appendonly = | |
2364 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2365 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2366 | } | |
2367 | ||
2368 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2369 | xoap->xoa_nodump = | |
2370 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2371 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2372 | } | |
2373 | ||
2374 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2375 | xoap->xoa_opaque = | |
2376 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2377 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2378 | } | |
2379 | ||
2380 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2381 | xoap->xoa_av_quarantined = | |
2382 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2383 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2384 | } | |
2385 | ||
2386 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2387 | xoap->xoa_av_modified = | |
2388 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2389 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2390 | } | |
2391 | ||
2392 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2393 | S_ISREG(ip->i_mode)) { | |
2394 | zfs_sa_get_scanstamp(zp, xvap); | |
2395 | } | |
34dc7c2f | 2396 | |
5484965a BB |
2397 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2398 | uint64_t times[2]; | |
2399 | ||
2400 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zsb), | |
2401 | times, sizeof (times)); | |
2402 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2403 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2404 | } | |
2405 | ||
2406 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2407 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2408 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2409 | } | |
2410 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
278f2236 | 2411 | xoap->xoa_generation = ip->i_generation; |
5484965a BB |
2412 | XVA_SET_RTN(xvap, XAT_GEN); |
2413 | } | |
2414 | ||
2415 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2416 | xoap->xoa_offline = | |
2417 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2418 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2419 | } | |
2420 | ||
2421 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2422 | xoap->xoa_sparse = | |
2423 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2424 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2425 | } | |
2426 | } | |
2427 | ||
0df9673f | 2428 | ZFS_TIME_DECODE(&vap->va_atime, atime); |
5484965a BB |
2429 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); |
2430 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2431 | |
2432 | mutex_exit(&zp->z_lock); | |
2433 | ||
5484965a | 2434 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2435 | |
2436 | if (zp->z_blksz == 0) { | |
2437 | /* | |
2438 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2439 | */ | |
5484965a | 2440 | vap->va_blksize = zsb->z_max_blksz; |
34dc7c2f BB |
2441 | } |
2442 | ||
3558fd73 | 2443 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2444 | return (0); |
2445 | } | |
e5c39b95 | 2446 | EXPORT_SYMBOL(zfs_getattr); |
34dc7c2f | 2447 | |
057e8eee BB |
2448 | /* |
2449 | * Get the basic file attributes and place them in the provided kstat | |
2450 | * structure. The inode is assumed to be the authoritative source | |
2451 | * for most of the attributes. However, the znode currently has the | |
2452 | * authoritative atime, blksize, and block count. | |
2453 | * | |
2454 | * IN: ip - inode of file. | |
2455 | * | |
2456 | * OUT: sp - kstat values. | |
2457 | * | |
2458 | * RETURN: 0 (always succeeds) | |
2459 | */ | |
2460 | /* ARGSUSED */ | |
2461 | int | |
2462 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2463 | { | |
2464 | znode_t *zp = ITOZ(ip); | |
2465 | zfs_sb_t *zsb = ITOZSB(ip); | |
b585bc4a BB |
2466 | uint32_t blksize; |
2467 | u_longlong_t nblocks; | |
057e8eee | 2468 | |
a7b125e9 GB |
2469 | ZFS_ENTER(zsb); |
2470 | ZFS_VERIFY_ZP(zp); | |
2471 | ||
057e8eee BB |
2472 | mutex_enter(&zp->z_lock); |
2473 | ||
2474 | generic_fillattr(ip, sp); | |
057e8eee | 2475 | |
b585bc4a BB |
2476 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2477 | sp->blksize = blksize; | |
2478 | sp->blocks = nblocks; | |
2479 | ||
057e8eee BB |
2480 | if (unlikely(zp->z_blksz == 0)) { |
2481 | /* | |
2482 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2483 | */ | |
2484 | sp->blksize = zsb->z_max_blksz; | |
2485 | } | |
2486 | ||
2487 | mutex_exit(&zp->z_lock); | |
2488 | ||
aa9b2708 AV |
2489 | /* |
2490 | * Required to prevent NFS client from detecting different inode | |
2491 | * numbers of snapshot root dentry before and after snapshot mount. | |
2492 | */ | |
2493 | if (zsb->z_issnap) { | |
2494 | if (ip->i_sb->s_root->d_inode == ip) | |
2495 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
2496 | dmu_objset_id(zsb->z_os); | |
2497 | } | |
2498 | ||
a7b125e9 GB |
2499 | ZFS_EXIT(zsb); |
2500 | ||
057e8eee BB |
2501 | return (0); |
2502 | } | |
2503 | EXPORT_SYMBOL(zfs_getattr_fast); | |
2504 | ||
34dc7c2f BB |
2505 | /* |
2506 | * Set the file attributes to the values contained in the | |
2507 | * vattr structure. | |
2508 | * | |
3558fd73 | 2509 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2510 | * vap - new attribute values. |
5484965a | 2511 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2512 | * flags - ATTR_UTIME set if non-default time values provided. |
2513 | * - ATTR_NOACLCHECK (CIFS context only). | |
2514 | * cr - credentials of caller. | |
34dc7c2f BB |
2515 | * |
2516 | * RETURN: 0 if success | |
2517 | * error code if failure | |
2518 | * | |
2519 | * Timestamps: | |
3558fd73 | 2520 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2521 | */ |
2522 | /* ARGSUSED */ | |
e5c39b95 | 2523 | int |
5484965a | 2524 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2525 | { |
3558fd73 BB |
2526 | znode_t *zp = ITOZ(ip); |
2527 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2528 | zilog_t *zilog; |
2529 | dmu_tx_t *tx; | |
2530 | vattr_t oldva; | |
f4ea75d4 | 2531 | xvattr_t *tmpxvattr; |
5484965a | 2532 | uint_t mask = vap->va_mask; |
a117a6d6 | 2533 | uint_t saved_mask = 0; |
34dc7c2f BB |
2534 | int trim_mask = 0; |
2535 | uint64_t new_mode; | |
64aefee1 | 2536 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 2537 | uint64_t xattr_obj; |
0df9673f | 2538 | uint64_t mtime[2], ctime[2], atime[2]; |
34dc7c2f BB |
2539 | znode_t *attrzp; |
2540 | int need_policy = FALSE; | |
428870ff | 2541 | int err, err2; |
34dc7c2f | 2542 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2543 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2544 | xoptattr_t *xoap; | |
2545 | zfs_acl_t *aclp; | |
34dc7c2f | 2546 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2547 | boolean_t fuid_dirtied = B_FALSE; |
17c37660 | 2548 | sa_bulk_attr_t *bulk, *xattr_bulk; |
428870ff | 2549 | int count = 0, xattr_count = 0; |
34dc7c2f BB |
2550 | |
2551 | if (mask == 0) | |
2552 | return (0); | |
2553 | ||
3558fd73 | 2554 | ZFS_ENTER(zsb); |
34dc7c2f BB |
2555 | ZFS_VERIFY_ZP(zp); |
2556 | ||
3558fd73 | 2557 | zilog = zsb->z_log; |
34dc7c2f BB |
2558 | |
2559 | /* | |
2560 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2561 | * that file system is at proper version level | |
2562 | */ | |
5484965a | 2563 | |
3558fd73 | 2564 | if (zsb->z_use_fuids == B_FALSE && |
5484965a BB |
2565 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2566 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2567 | (mask & ATTR_XVATTR))) { | |
3558fd73 | 2568 | ZFS_EXIT(zsb); |
2e528b49 | 2569 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2570 | } |
2571 | ||
3558fd73 BB |
2572 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
2573 | ZFS_EXIT(zsb); | |
2e528b49 | 2574 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2575 | } |
2576 | ||
3558fd73 BB |
2577 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
2578 | ZFS_EXIT(zsb); | |
2e528b49 | 2579 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2580 | } |
2581 | ||
5484965a BB |
2582 | /* |
2583 | * If this is an xvattr_t, then get a pointer to the structure of | |
2584 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2585 | */ | |
2586 | xoap = xva_getxoptattr(xvap); | |
2587 | ||
d1d7e268 | 2588 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2589 | xva_init(tmpxvattr); |
5484965a | 2590 | |
d1d7e268 MK |
2591 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); |
2592 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); | |
17c37660 | 2593 | |
5484965a BB |
2594 | /* |
2595 | * Immutable files can only alter immutable bit and atime | |
2596 | */ | |
2597 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2598 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2599 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
f4ea75d4 BB |
2600 | err = EPERM; |
2601 | goto out3; | |
5484965a BB |
2602 | } |
2603 | ||
3558fd73 | 2604 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
f4ea75d4 BB |
2605 | err = EPERM; |
2606 | goto out3; | |
34dc7c2f BB |
2607 | } |
2608 | ||
5484965a BB |
2609 | /* |
2610 | * Verify timestamps doesn't overflow 32 bits. | |
2611 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2612 | * handle times greater than 2039. This check should be removed | |
2613 | * once large timestamps are fully supported. | |
2614 | */ | |
2615 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
2616 | if (((mask & ATTR_ATIME) && |
2617 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2618 | ((mask & ATTR_MTIME) && | |
2619 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
f4ea75d4 BB |
2620 | err = EOVERFLOW; |
2621 | goto out3; | |
5484965a BB |
2622 | } |
2623 | } | |
2624 | ||
34dc7c2f BB |
2625 | top: |
2626 | attrzp = NULL; | |
572e2857 | 2627 | aclp = NULL; |
34dc7c2f | 2628 | |
45d1cae3 | 2629 | /* Can this be moved to before the top label? */ |
2cf7f52b | 2630 | if (zfs_is_readonly(zsb)) { |
f4ea75d4 BB |
2631 | err = EROFS; |
2632 | goto out3; | |
34dc7c2f BB |
2633 | } |
2634 | ||
2635 | /* | |
2636 | * First validate permissions | |
2637 | */ | |
2638 | ||
3558fd73 | 2639 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2640 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2641 | if (err) |
2642 | goto out3; | |
2643 | ||
34dc7c2f BB |
2644 | /* |
2645 | * XXX - Note, we are not providing any open | |
2646 | * mode flags here (like FNDELAY), so we may | |
2647 | * block if there are locks present... this | |
2648 | * should be addressed in openat(). | |
2649 | */ | |
b128c09f | 2650 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2651 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2652 | if (err) |
2653 | goto out3; | |
428870ff | 2654 | } |
34dc7c2f | 2655 | |
5484965a BB |
2656 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2657 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2658 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2659 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2660 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2661 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2662 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2663 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2664 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2665 | skipaclchk, cr); | |
2666 | } | |
2667 | ||
3558fd73 BB |
2668 | if (mask & (ATTR_UID|ATTR_GID)) { |
2669 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2670 | int take_owner; |
2671 | int take_group; | |
2672 | ||
2673 | /* | |
2674 | * NOTE: even if a new mode is being set, | |
2675 | * we may clear S_ISUID/S_ISGID bits. | |
2676 | */ | |
2677 | ||
3558fd73 | 2678 | if (!(mask & ATTR_MODE)) |
5484965a | 2679 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2680 | |
2681 | /* | |
2682 | * Take ownership or chgrp to group we are a member of | |
2683 | */ | |
2684 | ||
5484965a | 2685 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 2686 | take_group = (mask & ATTR_GID) && |
5484965a | 2687 | zfs_groupmember(zsb, vap->va_gid, cr); |
34dc7c2f BB |
2688 | |
2689 | /* | |
5484965a | 2690 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
2691 | * take_group must both be set in order to allow taking |
2692 | * ownership. | |
2693 | * | |
2694 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2695 | * | |
2696 | */ | |
2697 | ||
3558fd73 BB |
2698 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
2699 | take_owner && take_group) || | |
2700 | ((idmask == ATTR_UID) && take_owner) || | |
2701 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
2702 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
2703 | skipaclchk, cr) == 0) { | |
2704 | /* | |
2705 | * Remove setuid/setgid for non-privileged users | |
2706 | */ | |
5484965a | 2707 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 2708 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
2709 | } else { |
2710 | need_policy = TRUE; | |
2711 | } | |
2712 | } else { | |
2713 | need_policy = TRUE; | |
2714 | } | |
2715 | } | |
2716 | ||
2717 | mutex_enter(&zp->z_lock); | |
428870ff | 2718 | oldva.va_mode = zp->z_mode; |
572e2857 | 2719 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
2720 | if (mask & ATTR_XVATTR) { |
2721 | /* | |
2722 | * Update xvattr mask to include only those attributes | |
2723 | * that are actually changing. | |
2724 | * | |
2725 | * the bits will be restored prior to actually setting | |
2726 | * the attributes so the caller thinks they were set. | |
2727 | */ | |
2728 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2729 | if (xoap->xoa_appendonly != | |
2730 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
2731 | need_policy = TRUE; | |
2732 | } else { | |
2733 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 2734 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
2735 | } |
2736 | } | |
2737 | ||
2738 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2739 | if (xoap->xoa_nounlink != | |
2740 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
2741 | need_policy = TRUE; | |
2742 | } else { | |
2743 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 2744 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
2745 | } |
2746 | } | |
2747 | ||
2748 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2749 | if (xoap->xoa_immutable != | |
2750 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
2751 | need_policy = TRUE; | |
2752 | } else { | |
2753 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 2754 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
2755 | } |
2756 | } | |
2757 | ||
2758 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2759 | if (xoap->xoa_nodump != | |
2760 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
2761 | need_policy = TRUE; | |
2762 | } else { | |
2763 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 2764 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
2765 | } |
2766 | } | |
2767 | ||
2768 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2769 | if (xoap->xoa_av_modified != | |
2770 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
2771 | need_policy = TRUE; | |
2772 | } else { | |
2773 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 2774 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
2775 | } |
2776 | } | |
2777 | ||
2778 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2779 | if ((!S_ISREG(ip->i_mode) && | |
2780 | xoap->xoa_av_quarantined) || | |
2781 | xoap->xoa_av_quarantined != | |
2782 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
2783 | need_policy = TRUE; | |
2784 | } else { | |
2785 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 2786 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
2787 | } |
2788 | } | |
2789 | ||
2790 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2791 | mutex_exit(&zp->z_lock); | |
f4ea75d4 BB |
2792 | err = EPERM; |
2793 | goto out3; | |
5484965a BB |
2794 | } |
2795 | ||
2796 | if (need_policy == FALSE && | |
2797 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2798 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
2799 | need_policy = TRUE; | |
2800 | } | |
2801 | } | |
34dc7c2f BB |
2802 | |
2803 | mutex_exit(&zp->z_lock); | |
2804 | ||
3558fd73 | 2805 | if (mask & ATTR_MODE) { |
34dc7c2f | 2806 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 2807 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 2808 | &oldva, cr); |
f4ea75d4 BB |
2809 | if (err) |
2810 | goto out3; | |
2811 | ||
3558fd73 | 2812 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
2813 | } else { |
2814 | need_policy = TRUE; | |
2815 | } | |
2816 | } | |
2817 | ||
2818 | if (need_policy) { | |
2819 | /* | |
2820 | * If trim_mask is set then take ownership | |
2821 | * has been granted or write_acl is present and user | |
2822 | * has the ability to modify mode. In that case remove | |
2823 | * UID|GID and or MODE from mask so that | |
2824 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2825 | */ | |
2826 | ||
2827 | if (trim_mask) { | |
5484965a BB |
2828 | saved_mask = vap->va_mask; |
2829 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 2830 | } |
5484965a | 2831 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 2832 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
2833 | if (err) |
2834 | goto out3; | |
34dc7c2f BB |
2835 | |
2836 | if (trim_mask) | |
5484965a | 2837 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
2838 | } |
2839 | ||
2840 | /* | |
2841 | * secpolicy_vnode_setattr, or take ownership may have | |
2842 | * changed va_mask | |
2843 | */ | |
5484965a | 2844 | mask = vap->va_mask; |
34dc7c2f | 2845 | |
3558fd73 BB |
2846 | if ((mask & (ATTR_UID | ATTR_GID))) { |
2847 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), | |
572e2857 | 2848 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 2849 | |
572e2857 | 2850 | if (err == 0 && xattr_obj) { |
3558fd73 | 2851 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
2852 | if (err) |
2853 | goto out2; | |
2854 | } | |
3558fd73 | 2855 | if (mask & ATTR_UID) { |
64aefee1 | 2856 | new_kuid = zfs_fuid_create(zsb, |
5484965a | 2857 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
64aefee1 NB |
2858 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
2859 | zfs_fuid_overquota(zsb, B_FALSE, new_kuid)) { | |
572e2857 | 2860 | if (attrzp) |
3558fd73 | 2861 | iput(ZTOI(attrzp)); |
428870ff BB |
2862 | err = EDQUOT; |
2863 | goto out2; | |
2864 | } | |
2865 | } | |
2866 | ||
3558fd73 | 2867 | if (mask & ATTR_GID) { |
64aefee1 | 2868 | new_kgid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid, |
428870ff | 2869 | cr, ZFS_GROUP, &fuidp); |
64aefee1 NB |
2870 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && |
2871 | zfs_fuid_overquota(zsb, B_TRUE, new_kgid)) { | |
572e2857 | 2872 | if (attrzp) |
3558fd73 | 2873 | iput(ZTOI(attrzp)); |
428870ff BB |
2874 | err = EDQUOT; |
2875 | goto out2; | |
2876 | } | |
2877 | } | |
2878 | } | |
3558fd73 | 2879 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 2880 | |
3558fd73 | 2881 | if (mask & ATTR_MODE) { |
428870ff | 2882 | uint64_t pmode = zp->z_mode; |
572e2857 | 2883 | uint64_t acl_obj; |
5484965a | 2884 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 2885 | |
572e2857 | 2886 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 2887 | |
572e2857 BB |
2888 | mutex_enter(&zp->z_lock); |
2889 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
2890 | /* |
2891 | * Are we upgrading ACL from old V0 format | |
2892 | * to V1 format? | |
2893 | */ | |
3558fd73 | 2894 | if (zsb->z_version >= ZPL_VERSION_FUID && |
572e2857 | 2895 | zfs_znode_acl_version(zp) == |
34dc7c2f | 2896 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 2897 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
2898 | DMU_OBJECT_END); |
2899 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
2900 | 0, aclp->z_acl_bytes); | |
2901 | } else { | |
572e2857 | 2902 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
2903 | aclp->z_acl_bytes); |
2904 | } | |
428870ff | 2905 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
2906 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
2907 | 0, aclp->z_acl_bytes); | |
2908 | } | |
572e2857 | 2909 | mutex_exit(&zp->z_lock); |
428870ff BB |
2910 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
2911 | } else { | |
5484965a BB |
2912 | if ((mask & ATTR_XVATTR) && |
2913 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
2914 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
2915 | else | |
2916 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2917 | } |
2918 | ||
428870ff BB |
2919 | if (attrzp) { |
2920 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2921 | } |
2922 | ||
3558fd73 | 2923 | fuid_dirtied = zsb->z_fuid_dirty; |
428870ff | 2924 | if (fuid_dirtied) |
3558fd73 | 2925 | zfs_fuid_txhold(zsb, tx); |
428870ff BB |
2926 | |
2927 | zfs_sa_upgrade_txholds(tx, zp); | |
2928 | ||
384f8a09 MA |
2929 | err = dmu_tx_assign(tx, TXG_WAIT); |
2930 | if (err) | |
9babb374 | 2931 | goto out; |
34dc7c2f | 2932 | |
428870ff | 2933 | count = 0; |
34dc7c2f BB |
2934 | /* |
2935 | * Set each attribute requested. | |
2936 | * We group settings according to the locks they need to acquire. | |
2937 | * | |
2938 | * Note: you cannot set ctime directly, although it will be | |
2939 | * updated as a side-effect of calling this function. | |
2940 | */ | |
2941 | ||
572e2857 | 2942 | |
3558fd73 | 2943 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2944 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
2945 | mutex_enter(&zp->z_lock); |
2946 | ||
3558fd73 | 2947 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff BB |
2948 | &zp->z_pflags, sizeof (zp->z_pflags)); |
2949 | ||
2950 | if (attrzp) { | |
3558fd73 | 2951 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2952 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
2953 | mutex_enter(&attrzp->z_lock); |
2954 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2955 | SA_ZPL_FLAGS(zsb), NULL, &attrzp->z_pflags, |
428870ff BB |
2956 | sizeof (attrzp->z_pflags)); |
2957 | } | |
2958 | ||
3558fd73 | 2959 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 2960 | |
3558fd73 | 2961 | if (mask & ATTR_UID) { |
64aefee1 NB |
2962 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
2963 | new_uid = zfs_uid_read(ZTOI(zp)); | |
3558fd73 | 2964 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, |
428870ff | 2965 | &new_uid, sizeof (new_uid)); |
428870ff BB |
2966 | if (attrzp) { |
2967 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2968 | SA_ZPL_UID(zsb), NULL, &new_uid, |
428870ff | 2969 | sizeof (new_uid)); |
2c6abf15 | 2970 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
2971 | } |
2972 | } | |
2973 | ||
3558fd73 | 2974 | if (mask & ATTR_GID) { |
64aefee1 NB |
2975 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
2976 | new_gid = zfs_gid_read(ZTOI(zp)); | |
3558fd73 | 2977 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), |
428870ff | 2978 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
2979 | if (attrzp) { |
2980 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2981 | SA_ZPL_GID(zsb), NULL, &new_gid, |
428870ff | 2982 | sizeof (new_gid)); |
64aefee1 | 2983 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
2984 | } |
2985 | } | |
3558fd73 BB |
2986 | if (!(mask & ATTR_MODE)) { |
2987 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), | |
428870ff BB |
2988 | NULL, &new_mode, sizeof (new_mode)); |
2989 | new_mode = zp->z_mode; | |
2990 | } | |
2991 | err = zfs_acl_chown_setattr(zp); | |
2992 | ASSERT(err == 0); | |
2993 | if (attrzp) { | |
2994 | err = zfs_acl_chown_setattr(attrzp); | |
2995 | ASSERT(err == 0); | |
2996 | } | |
2997 | } | |
2998 | ||
3558fd73 BB |
2999 | if (mask & ATTR_MODE) { |
3000 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, | |
428870ff | 3001 | &new_mode, sizeof (new_mode)); |
12fa7f34 | 3002 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 3003 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 3004 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 3005 | ASSERT0(err); |
572e2857 BB |
3006 | if (zp->z_acl_cached) |
3007 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
3008 | zp->z_acl_cached = aclp; |
3009 | aclp = NULL; | |
34dc7c2f BB |
3010 | } |
3011 | ||
704cd075 CC |
3012 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
3013 | zp->z_atime_dirty = 0; | |
3014 | ZFS_TIME_ENCODE(&ip->i_atime, atime); | |
3558fd73 | 3015 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, |
0df9673f | 3016 | &atime, sizeof (atime)); |
34dc7c2f BB |
3017 | } |
3018 | ||
3558fd73 | 3019 | if (mask & ATTR_MTIME) { |
5484965a | 3020 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
87f9371a NB |
3021 | ZTOI(zp)->i_mtime = timespec_trunc(vap->va_mtime, |
3022 | ZTOI(zp)->i_sb->s_time_gran); | |
3023 | ||
3558fd73 | 3024 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, |
428870ff | 3025 | mtime, sizeof (mtime)); |
34dc7c2f BB |
3026 | } |
3027 | ||
87f9371a NB |
3028 | if (mask & ATTR_CTIME) { |
3029 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); | |
3030 | ZTOI(zp)->i_ctime = timespec_trunc(vap->va_ctime, | |
3031 | ZTOI(zp)->i_sb->s_time_gran); | |
3558fd73 | 3032 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
87f9371a | 3033 | ctime, sizeof (ctime)); |
428870ff | 3034 | } |
87f9371a NB |
3035 | |
3036 | if (attrzp && mask) { | |
3037 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3038 | SA_ZPL_CTIME(zsb), NULL, &ctime, | |
3039 | sizeof (ctime)); | |
3040 | } | |
3041 | ||
34dc7c2f BB |
3042 | /* |
3043 | * Do this after setting timestamps to prevent timestamp | |
3044 | * update from toggling bit | |
3045 | */ | |
3046 | ||
5484965a BB |
3047 | if (xoap && (mask & ATTR_XVATTR)) { |
3048 | ||
3049 | /* | |
3050 | * restore trimmed off masks | |
3051 | * so that return masks can be set for caller. | |
3052 | */ | |
3053 | ||
f4ea75d4 | 3054 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
3055 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
3056 | } | |
f4ea75d4 | 3057 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
3058 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
3059 | } | |
f4ea75d4 | 3060 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
3061 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
3062 | } | |
f4ea75d4 | 3063 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
3064 | XVA_SET_REQ(xvap, XAT_NODUMP); |
3065 | } | |
f4ea75d4 | 3066 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
3067 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
3068 | } | |
f4ea75d4 | 3069 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
3070 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
3071 | } | |
3072 | ||
3073 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3074 | ASSERT(S_ISREG(ip->i_mode)); | |
3075 | ||
3076 | zfs_xvattr_set(zp, xvap, tx); | |
3077 | } | |
3078 | ||
9babb374 | 3079 | if (fuid_dirtied) |
3558fd73 | 3080 | zfs_fuid_sync(zsb, tx); |
9babb374 | 3081 | |
34dc7c2f | 3082 | if (mask != 0) |
5484965a | 3083 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3084 | |
34dc7c2f | 3085 | mutex_exit(&zp->z_lock); |
3558fd73 | 3086 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3087 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3088 | |
572e2857 | 3089 | if (attrzp) { |
3558fd73 | 3090 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3091 | mutex_exit(&attrzp->z_acl_lock); |
3092 | mutex_exit(&attrzp->z_lock); | |
3093 | } | |
9babb374 | 3094 | out: |
428870ff BB |
3095 | if (err == 0 && attrzp) { |
3096 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, | |
3097 | xattr_count, tx); | |
3098 | ASSERT(err2 == 0); | |
3099 | } | |
3100 | ||
34dc7c2f | 3101 | if (attrzp) |
3558fd73 | 3102 | iput(ZTOI(attrzp)); |
45d1cae3 | 3103 | if (aclp) |
9babb374 | 3104 | zfs_acl_free(aclp); |
9babb374 BB |
3105 | |
3106 | if (fuidp) { | |
3107 | zfs_fuid_info_free(fuidp); | |
3108 | fuidp = NULL; | |
3109 | } | |
3110 | ||
428870ff | 3111 | if (err) { |
9babb374 | 3112 | dmu_tx_abort(tx); |
428870ff BB |
3113 | if (err == ERESTART) |
3114 | goto top; | |
3115 | } else { | |
3116 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3117 | dmu_tx_commit(tx); |
037849f8 | 3118 | zfs_inode_update(zp); |
428870ff BB |
3119 | } |
3120 | ||
428870ff | 3121 | out2: |
3558fd73 | 3122 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3123 | zil_commit(zilog, 0); |
34dc7c2f | 3124 | |
f4ea75d4 | 3125 | out3: |
d1d7e268 MK |
3126 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * 7); |
3127 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * 7); | |
3128 | kmem_free(tmpxvattr, sizeof (xvattr_t)); | |
3558fd73 | 3129 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3130 | return (err); |
3131 | } | |
e5c39b95 | 3132 | EXPORT_SYMBOL(zfs_setattr); |
34dc7c2f BB |
3133 | |
3134 | typedef struct zfs_zlock { | |
3135 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3136 | znode_t *zl_znode; /* znode we held */ | |
3137 | struct zfs_zlock *zl_next; /* next in list */ | |
3138 | } zfs_zlock_t; | |
3139 | ||
3140 | /* | |
3141 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3142 | */ | |
3143 | static void | |
3144 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3145 | { | |
3146 | zfs_zlock_t *zl; | |
3147 | ||
3148 | while ((zl = *zlpp) != NULL) { | |
3149 | if (zl->zl_znode != NULL) | |
3558fd73 | 3150 | iput(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3151 | rw_exit(zl->zl_rwlock); |
3152 | *zlpp = zl->zl_next; | |
3153 | kmem_free(zl, sizeof (*zl)); | |
3154 | } | |
3155 | } | |
3156 | ||
3157 | /* | |
3158 | * Search back through the directory tree, using the ".." entries. | |
3159 | * Lock each directory in the chain to prevent concurrent renames. | |
3160 | * Fail any attempt to move a directory into one of its own descendants. | |
3161 | * XXX - z_parent_lock can overlap with map or grow locks | |
3162 | */ | |
3163 | static int | |
3164 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3165 | { | |
3166 | zfs_zlock_t *zl; | |
3167 | znode_t *zp = tdzp; | |
3558fd73 | 3168 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3169 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3170 | krwlock_t *rwlp = &szp->z_parent_lock; |
3171 | krw_t rw = RW_WRITER; | |
3172 | ||
3173 | /* | |
3174 | * First pass write-locks szp and compares to zp->z_id. | |
3175 | * Later passes read-lock zp and compare to zp->z_parent. | |
3176 | */ | |
3177 | do { | |
3178 | if (!rw_tryenter(rwlp, rw)) { | |
3179 | /* | |
3180 | * Another thread is renaming in this path. | |
3181 | * Note that if we are a WRITER, we don't have any | |
3182 | * parent_locks held yet. | |
3183 | */ | |
3184 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3185 | /* | |
3186 | * Drop our locks and restart | |
3187 | */ | |
3188 | zfs_rename_unlock(&zl); | |
3189 | *zlpp = NULL; | |
3190 | zp = tdzp; | |
428870ff | 3191 | oidp = zp->z_id; |
34dc7c2f BB |
3192 | rwlp = &szp->z_parent_lock; |
3193 | rw = RW_WRITER; | |
3194 | continue; | |
3195 | } else { | |
3196 | /* | |
3197 | * Wait for other thread to drop its locks | |
3198 | */ | |
3199 | rw_enter(rwlp, rw); | |
3200 | } | |
3201 | } | |
3202 | ||
3203 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3204 | zl->zl_rwlock = rwlp; | |
3205 | zl->zl_znode = NULL; | |
3206 | zl->zl_next = *zlpp; | |
3207 | *zlpp = zl; | |
3208 | ||
428870ff | 3209 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 3210 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3211 | |
428870ff | 3212 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3213 | return (0); |
3214 | ||
3215 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3216 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3217 | if (error) |
3218 | return (error); | |
3219 | zl->zl_znode = zp; | |
3220 | } | |
3558fd73 | 3221 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3222 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3223 | rwlp = &zp->z_parent_lock; |
3224 | rw = RW_READER; | |
3225 | ||
3226 | } while (zp->z_id != sdzp->z_id); | |
3227 | ||
3228 | return (0); | |
3229 | } | |
3230 | ||
3231 | /* | |
3232 | * Move an entry from the provided source directory to the target | |
3233 | * directory. Change the entry name as indicated. | |
3234 | * | |
3558fd73 | 3235 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3236 | * snm - Old entry name. |
3558fd73 | 3237 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3238 | * tnm - New entry name. |
3239 | * cr - credentials of caller. | |
34dc7c2f BB |
3240 | * flags - case flags |
3241 | * | |
d3cc8b15 | 3242 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3243 | * |
3244 | * Timestamps: | |
3558fd73 | 3245 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3246 | */ |
3247 | /*ARGSUSED*/ | |
e5c39b95 | 3248 | int |
3558fd73 BB |
3249 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3250 | cred_t *cr, int flags) | |
34dc7c2f BB |
3251 | { |
3252 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 BB |
3253 | znode_t *sdzp = ITOZ(sdip); |
3254 | zfs_sb_t *zsb = ITOZSB(sdip); | |
34dc7c2f | 3255 | zilog_t *zilog; |
34dc7c2f BB |
3256 | zfs_dirlock_t *sdl, *tdl; |
3257 | dmu_tx_t *tx; | |
3258 | zfs_zlock_t *zl; | |
3259 | int cmp, serr, terr; | |
3260 | int error = 0; | |
3261 | int zflg = 0; | |
e8b96c60 | 3262 | boolean_t waited = B_FALSE; |
34dc7c2f | 3263 | |
3558fd73 | 3264 | ZFS_ENTER(zsb); |
34dc7c2f | 3265 | ZFS_VERIFY_ZP(sdzp); |
3558fd73 | 3266 | zilog = zsb->z_log; |
34dc7c2f | 3267 | |
812e91a7 MT |
3268 | tdzp = ITOZ(tdip); |
3269 | ZFS_VERIFY_ZP(tdzp); | |
3270 | ||
3271 | /* | |
3272 | * We check i_sb because snapshots and the ctldir must have different | |
3273 | * super blocks. | |
3274 | */ | |
c0ebc844 | 3275 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
3558fd73 | 3276 | ZFS_EXIT(zsb); |
2e528b49 | 3277 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3278 | } |
3279 | ||
3558fd73 | 3280 | if (zsb->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3281 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3282 | ZFS_EXIT(zsb); |
2e528b49 | 3283 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3284 | } |
3285 | ||
3286 | if (flags & FIGNORECASE) | |
3287 | zflg |= ZCILOOK; | |
3288 | ||
3289 | top: | |
3290 | szp = NULL; | |
3291 | tzp = NULL; | |
3292 | zl = NULL; | |
3293 | ||
3294 | /* | |
3295 | * This is to prevent the creation of links into attribute space | |
3296 | * by renaming a linked file into/outof an attribute directory. | |
3297 | * See the comment in zfs_link() for why this is considered bad. | |
3298 | */ | |
428870ff | 3299 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 3300 | ZFS_EXIT(zsb); |
2e528b49 | 3301 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3302 | } |
3303 | ||
3304 | /* | |
3305 | * Lock source and target directory entries. To prevent deadlock, | |
3306 | * a lock ordering must be defined. We lock the directory with | |
3307 | * the smallest object id first, or if it's a tie, the one with | |
3308 | * the lexically first name. | |
3309 | */ | |
3310 | if (sdzp->z_id < tdzp->z_id) { | |
3311 | cmp = -1; | |
3312 | } else if (sdzp->z_id > tdzp->z_id) { | |
3313 | cmp = 1; | |
3314 | } else { | |
3315 | /* | |
3316 | * First compare the two name arguments without | |
3317 | * considering any case folding. | |
3318 | */ | |
3558fd73 | 3319 | int nofold = (zsb->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3320 | |
3321 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
3558fd73 | 3322 | ASSERT(error == 0 || !zsb->z_utf8); |
34dc7c2f BB |
3323 | if (cmp == 0) { |
3324 | /* | |
3325 | * POSIX: "If the old argument and the new argument | |
3326 | * both refer to links to the same existing file, | |
3327 | * the rename() function shall return successfully | |
3328 | * and perform no other action." | |
3329 | */ | |
3558fd73 | 3330 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3331 | return (0); |
3332 | } | |
3333 | /* | |
3334 | * If the file system is case-folding, then we may | |
3335 | * have some more checking to do. A case-folding file | |
3336 | * system is either supporting mixed case sensitivity | |
3337 | * access or is completely case-insensitive. Note | |
3338 | * that the file system is always case preserving. | |
3339 | * | |
3340 | * In mixed sensitivity mode case sensitive behavior | |
3341 | * is the default. FIGNORECASE must be used to | |
3342 | * explicitly request case insensitive behavior. | |
3343 | * | |
3344 | * If the source and target names provided differ only | |
3345 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3346 | * we will treat this as a special case in the | |
3347 | * case-insensitive mode: as long as the source name | |
3348 | * is an exact match, we will allow this to proceed as | |
3349 | * a name-change request. | |
3350 | */ | |
3558fd73 BB |
3351 | if ((zsb->z_case == ZFS_CASE_INSENSITIVE || |
3352 | (zsb->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3353 | flags & FIGNORECASE)) && |
3558fd73 | 3354 | u8_strcmp(snm, tnm, 0, zsb->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3355 | &error) == 0) { |
3356 | /* | |
3357 | * case preserving rename request, require exact | |
3358 | * name matches | |
3359 | */ | |
3360 | zflg |= ZCIEXACT; | |
3361 | zflg &= ~ZCILOOK; | |
3362 | } | |
3363 | } | |
3364 | ||
428870ff BB |
3365 | /* |
3366 | * If the source and destination directories are the same, we should | |
3367 | * grab the z_name_lock of that directory only once. | |
3368 | */ | |
3369 | if (sdzp == tdzp) { | |
3370 | zflg |= ZHAVELOCK; | |
3371 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3372 | } | |
3373 | ||
34dc7c2f BB |
3374 | if (cmp < 0) { |
3375 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3376 | ZEXISTS | zflg, NULL, NULL); | |
3377 | terr = zfs_dirent_lock(&tdl, | |
3378 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3379 | } else { | |
3380 | terr = zfs_dirent_lock(&tdl, | |
3381 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3382 | serr = zfs_dirent_lock(&sdl, | |
3383 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3384 | NULL, NULL); | |
3385 | } | |
3386 | ||
3387 | if (serr) { | |
3388 | /* | |
3389 | * Source entry invalid or not there. | |
3390 | */ | |
3391 | if (!terr) { | |
3392 | zfs_dirent_unlock(tdl); | |
3393 | if (tzp) | |
3558fd73 | 3394 | iput(ZTOI(tzp)); |
34dc7c2f | 3395 | } |
428870ff BB |
3396 | |
3397 | if (sdzp == tdzp) | |
3398 | rw_exit(&sdzp->z_name_lock); | |
3399 | ||
34dc7c2f BB |
3400 | if (strcmp(snm, "..") == 0) |
3401 | serr = EINVAL; | |
3558fd73 | 3402 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3403 | return (serr); |
3404 | } | |
3405 | if (terr) { | |
3406 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3407 | iput(ZTOI(szp)); |
428870ff BB |
3408 | |
3409 | if (sdzp == tdzp) | |
3410 | rw_exit(&sdzp->z_name_lock); | |
3411 | ||
34dc7c2f BB |
3412 | if (strcmp(tnm, "..") == 0) |
3413 | terr = EINVAL; | |
3558fd73 | 3414 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3415 | return (terr); |
3416 | } | |
3417 | ||
3418 | /* | |
3419 | * Must have write access at the source to remove the old entry | |
3420 | * and write access at the target to create the new entry. | |
3421 | * Note that if target and source are the same, this can be | |
3422 | * done in a single check. | |
3423 | */ | |
3424 | ||
149e873a | 3425 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3426 | goto out; |
3427 | ||
3558fd73 | 3428 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3429 | /* |
3430 | * Check to make sure rename is valid. | |
3431 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3432 | */ | |
149e873a | 3433 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3434 | goto out; |
3435 | } | |
3436 | ||
3437 | /* | |
3438 | * Does target exist? | |
3439 | */ | |
3440 | if (tzp) { | |
3441 | /* | |
3442 | * Source and target must be the same type. | |
3443 | */ | |
3558fd73 BB |
3444 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3445 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
2e528b49 | 3446 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3447 | goto out; |
3448 | } | |
3449 | } else { | |
3558fd73 | 3450 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
2e528b49 | 3451 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3452 | goto out; |
3453 | } | |
3454 | } | |
3455 | /* | |
3456 | * POSIX dictates that when the source and target | |
3457 | * entries refer to the same file object, rename | |
3458 | * must do nothing and exit without error. | |
3459 | */ | |
3460 | if (szp->z_id == tzp->z_id) { | |
3461 | error = 0; | |
3462 | goto out; | |
3463 | } | |
3464 | } | |
3465 | ||
3558fd73 | 3466 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
3467 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3468 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3469 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3470 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3471 | if (sdzp != tdzp) { |
3472 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3473 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3474 | } | |
3475 | if (tzp) { | |
3476 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3477 | zfs_sa_upgrade_txholds(tx, tzp); | |
3478 | } | |
3479 | ||
3480 | zfs_sa_upgrade_txholds(tx, szp); | |
3558fd73 | 3481 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
e8b96c60 | 3482 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3483 | if (error) { |
3484 | if (zl != NULL) | |
3485 | zfs_rename_unlock(&zl); | |
3486 | zfs_dirent_unlock(sdl); | |
3487 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3488 | |
3489 | if (sdzp == tdzp) | |
3490 | rw_exit(&sdzp->z_name_lock); | |
3491 | ||
3558fd73 | 3492 | iput(ZTOI(szp)); |
34dc7c2f | 3493 | if (tzp) |
3558fd73 | 3494 | iput(ZTOI(tzp)); |
fb5f0bc8 | 3495 | if (error == ERESTART) { |
e8b96c60 | 3496 | waited = B_TRUE; |
34dc7c2f BB |
3497 | dmu_tx_wait(tx); |
3498 | dmu_tx_abort(tx); | |
3499 | goto top; | |
3500 | } | |
3501 | dmu_tx_abort(tx); | |
3558fd73 | 3502 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3503 | return (error); |
3504 | } | |
3505 | ||
3506 | if (tzp) /* Attempt to remove the existing target */ | |
3507 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3508 | ||
3509 | if (error == 0) { | |
3510 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3511 | if (error == 0) { | |
428870ff | 3512 | szp->z_pflags |= ZFS_AV_MODIFIED; |
34dc7c2f | 3513 | |
3558fd73 | 3514 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zsb), |
428870ff | 3515 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c99c9001 | 3516 | ASSERT0(error); |
34dc7c2f | 3517 | |
428870ff BB |
3518 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3519 | if (error == 0) { | |
3520 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3521 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3522 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3523 | } else { |
3524 | /* | |
3525 | * At this point, we have successfully created | |
3526 | * the target name, but have failed to remove | |
3527 | * the source name. Since the create was done | |
3528 | * with the ZRENAMING flag, there are | |
3529 | * complications; for one, the link count is | |
3530 | * wrong. The easiest way to deal with this | |
3531 | * is to remove the newly created target, and | |
3532 | * return the original error. This must | |
3533 | * succeed; fortunately, it is very unlikely to | |
3534 | * fail, since we just created it. | |
3535 | */ | |
3536 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3537 | ZRENAMING, NULL), ==, 0); | |
3538 | } | |
34dc7c2f BB |
3539 | } |
3540 | } | |
3541 | ||
3542 | dmu_tx_commit(tx); | |
3543 | out: | |
3544 | if (zl != NULL) | |
3545 | zfs_rename_unlock(&zl); | |
3546 | ||
3547 | zfs_dirent_unlock(sdl); | |
3548 | zfs_dirent_unlock(tdl); | |
3549 | ||
960e08fe | 3550 | zfs_inode_update(sdzp); |
428870ff BB |
3551 | if (sdzp == tdzp) |
3552 | rw_exit(&sdzp->z_name_lock); | |
3553 | ||
960e08fe BB |
3554 | if (sdzp != tdzp) |
3555 | zfs_inode_update(tdzp); | |
428870ff | 3556 | |
960e08fe | 3557 | zfs_inode_update(szp); |
3558fd73 | 3558 | iput(ZTOI(szp)); |
960e08fe BB |
3559 | if (tzp) { |
3560 | zfs_inode_update(tzp); | |
3558fd73 | 3561 | iput(ZTOI(tzp)); |
960e08fe | 3562 | } |
34dc7c2f | 3563 | |
3558fd73 | 3564 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3565 | zil_commit(zilog, 0); |
428870ff | 3566 | |
3558fd73 | 3567 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3568 | return (error); |
3569 | } | |
e5c39b95 | 3570 | EXPORT_SYMBOL(zfs_rename); |
34dc7c2f BB |
3571 | |
3572 | /* | |
3573 | * Insert the indicated symbolic reference entry into the directory. | |
3574 | * | |
3558fd73 | 3575 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
3576 | * link - Name for new symlink entry. |
3577 | * vap - Attributes of new entry. | |
3578 | * target - Target path of new symlink. | |
3558fd73 | 3579 | * |
34dc7c2f | 3580 | * cr - credentials of caller. |
34dc7c2f BB |
3581 | * flags - case flags |
3582 | * | |
d3cc8b15 | 3583 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3584 | * |
3585 | * Timestamps: | |
3558fd73 | 3586 | * dip - ctime|mtime updated |
34dc7c2f BB |
3587 | */ |
3588 | /*ARGSUSED*/ | |
e5c39b95 | 3589 | int |
3558fd73 BB |
3590 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
3591 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 3592 | { |
3558fd73 | 3593 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
3594 | zfs_dirlock_t *dl; |
3595 | dmu_tx_t *tx; | |
3558fd73 | 3596 | zfs_sb_t *zsb = ITOZSB(dip); |
34dc7c2f | 3597 | zilog_t *zilog; |
428870ff | 3598 | uint64_t len = strlen(link); |
34dc7c2f BB |
3599 | int error; |
3600 | int zflg = ZNEW; | |
9babb374 BB |
3601 | zfs_acl_ids_t acl_ids; |
3602 | boolean_t fuid_dirtied; | |
428870ff | 3603 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 3604 | boolean_t waited = B_FALSE; |
34dc7c2f | 3605 | |
3558fd73 | 3606 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 3607 | |
3558fd73 | 3608 | ZFS_ENTER(zsb); |
34dc7c2f | 3609 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3610 | zilog = zsb->z_log; |
34dc7c2f | 3611 | |
3558fd73 | 3612 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 3613 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3614 | ZFS_EXIT(zsb); |
2e528b49 | 3615 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3616 | } |
3617 | if (flags & FIGNORECASE) | |
3618 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3619 | |
3620 | if (len > MAXPATHLEN) { | |
3558fd73 | 3621 | ZFS_EXIT(zsb); |
2e528b49 | 3622 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
3623 | } |
3624 | ||
428870ff BB |
3625 | if ((error = zfs_acl_ids_create(dzp, 0, |
3626 | vap, cr, NULL, &acl_ids)) != 0) { | |
3558fd73 | 3627 | ZFS_EXIT(zsb); |
428870ff BB |
3628 | return (error); |
3629 | } | |
3630 | top: | |
3558fd73 BB |
3631 | *ipp = NULL; |
3632 | ||
34dc7c2f BB |
3633 | /* |
3634 | * Attempt to lock directory; fail if entry already exists. | |
3635 | */ | |
3636 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3637 | if (error) { | |
428870ff | 3638 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 3639 | ZFS_EXIT(zsb); |
428870ff BB |
3640 | return (error); |
3641 | } | |
3642 | ||
149e873a | 3643 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
3644 | zfs_acl_ids_free(&acl_ids); |
3645 | zfs_dirent_unlock(dl); | |
3558fd73 | 3646 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3647 | return (error); |
3648 | } | |
3649 | ||
3558fd73 | 3650 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
9babb374 BB |
3651 | zfs_acl_ids_free(&acl_ids); |
3652 | zfs_dirent_unlock(dl); | |
3558fd73 | 3653 | ZFS_EXIT(zsb); |
2e528b49 | 3654 | return (SET_ERROR(EDQUOT)); |
9babb374 | 3655 | } |
3558fd73 BB |
3656 | tx = dmu_tx_create(zsb->z_os); |
3657 | fuid_dirtied = zsb->z_fuid_dirty; | |
34dc7c2f | 3658 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3659 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3660 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3661 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3662 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
3558fd73 | 3663 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
3664 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
3665 | acl_ids.z_aclp->z_acl_bytes); | |
3666 | } | |
9babb374 | 3667 | if (fuid_dirtied) |
3558fd73 | 3668 | zfs_fuid_txhold(zsb, tx); |
e8b96c60 | 3669 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3670 | if (error) { |
3671 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3672 | if (error == ERESTART) { |
e8b96c60 | 3673 | waited = B_TRUE; |
34dc7c2f BB |
3674 | dmu_tx_wait(tx); |
3675 | dmu_tx_abort(tx); | |
3676 | goto top; | |
3677 | } | |
428870ff | 3678 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 3679 | dmu_tx_abort(tx); |
3558fd73 | 3680 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3681 | return (error); |
3682 | } | |
3683 | ||
34dc7c2f BB |
3684 | /* |
3685 | * Create a new object for the symlink. | |
428870ff | 3686 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 3687 | */ |
428870ff | 3688 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3689 | |
428870ff | 3690 | if (fuid_dirtied) |
3558fd73 | 3691 | zfs_fuid_sync(zsb, tx); |
34dc7c2f | 3692 | |
572e2857 | 3693 | mutex_enter(&zp->z_lock); |
428870ff | 3694 | if (zp->z_is_sa) |
3558fd73 | 3695 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zsb), |
428870ff BB |
3696 | link, len, tx); |
3697 | else | |
3698 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3699 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3700 | |
428870ff | 3701 | zp->z_size = len; |
3558fd73 | 3702 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 3703 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
3704 | /* |
3705 | * Insert the new object into the directory. | |
3706 | */ | |
3707 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
428870ff BB |
3708 | |
3709 | if (flags & FIGNORECASE) | |
3710 | txtype |= TX_CI; | |
3711 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
9babb374 | 3712 | |
960e08fe BB |
3713 | zfs_inode_update(dzp); |
3714 | zfs_inode_update(zp); | |
3715 | ||
9babb374 | 3716 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3717 | |
3718 | dmu_tx_commit(tx); | |
3719 | ||
3720 | zfs_dirent_unlock(dl); | |
3721 | ||
3558fd73 | 3722 | *ipp = ZTOI(zp); |
34dc7c2f | 3723 | |
3558fd73 | 3724 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3725 | zil_commit(zilog, 0); |
428870ff | 3726 | |
3558fd73 | 3727 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3728 | return (error); |
3729 | } | |
e5c39b95 | 3730 | EXPORT_SYMBOL(zfs_symlink); |
34dc7c2f BB |
3731 | |
3732 | /* | |
3733 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 3734 | * the symbolic path referred to by ip. |
34dc7c2f | 3735 | * |
8b4f9a2d BB |
3736 | * IN: ip - inode of symbolic link |
3737 | * uio - structure to contain the link path. | |
3738 | * cr - credentials of caller. | |
34dc7c2f BB |
3739 | * |
3740 | * RETURN: 0 if success | |
3741 | * error code if failure | |
3742 | * | |
3743 | * Timestamps: | |
3558fd73 | 3744 | * ip - atime updated |
34dc7c2f BB |
3745 | */ |
3746 | /* ARGSUSED */ | |
e5c39b95 | 3747 | int |
8b4f9a2d | 3748 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 3749 | { |
3558fd73 BB |
3750 | znode_t *zp = ITOZ(ip); |
3751 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
3752 | int error; |
3753 | ||
3558fd73 | 3754 | ZFS_ENTER(zsb); |
34dc7c2f BB |
3755 | ZFS_VERIFY_ZP(zp); |
3756 | ||
572e2857 | 3757 | mutex_enter(&zp->z_lock); |
428870ff | 3758 | if (zp->z_is_sa) |
8b4f9a2d BB |
3759 | error = sa_lookup_uio(zp->z_sa_hdl, |
3760 | SA_ZPL_SYMLINK(zsb), uio); | |
428870ff | 3761 | else |
8b4f9a2d | 3762 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 3763 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3764 | |
3558fd73 | 3765 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3766 | return (error); |
3767 | } | |
8b4f9a2d | 3768 | EXPORT_SYMBOL(zfs_readlink); |
34dc7c2f BB |
3769 | |
3770 | /* | |
3558fd73 | 3771 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 3772 | * |
3558fd73 BB |
3773 | * IN: tdip - Directory to contain new entry. |
3774 | * sip - inode of new entry. | |
34dc7c2f BB |
3775 | * name - name of new entry. |
3776 | * cr - credentials of caller. | |
34dc7c2f BB |
3777 | * |
3778 | * RETURN: 0 if success | |
3779 | * error code if failure | |
3780 | * | |
3781 | * Timestamps: | |
3558fd73 BB |
3782 | * tdip - ctime|mtime updated |
3783 | * sip - ctime updated | |
34dc7c2f BB |
3784 | */ |
3785 | /* ARGSUSED */ | |
e5c39b95 | 3786 | int |
da5e151f BB |
3787 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr, |
3788 | int flags) | |
34dc7c2f | 3789 | { |
3558fd73 | 3790 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 3791 | znode_t *tzp, *szp; |
3558fd73 | 3792 | zfs_sb_t *zsb = ITOZSB(tdip); |
34dc7c2f BB |
3793 | zilog_t *zilog; |
3794 | zfs_dirlock_t *dl; | |
3795 | dmu_tx_t *tx; | |
34dc7c2f BB |
3796 | int error; |
3797 | int zf = ZNEW; | |
428870ff | 3798 | uint64_t parent; |
572e2857 | 3799 | uid_t owner; |
e8b96c60 | 3800 | boolean_t waited = B_FALSE; |
34dc7c2f | 3801 | |
3558fd73 | 3802 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 3803 | |
3558fd73 | 3804 | ZFS_ENTER(zsb); |
34dc7c2f | 3805 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3806 | zilog = zsb->z_log; |
34dc7c2f | 3807 | |
428870ff BB |
3808 | /* |
3809 | * POSIX dictates that we return EPERM here. | |
3810 | * Better choices include ENOTSUP or EISDIR. | |
3811 | */ | |
3558fd73 BB |
3812 | if (S_ISDIR(sip->i_mode)) { |
3813 | ZFS_EXIT(zsb); | |
2e528b49 | 3814 | return (SET_ERROR(EPERM)); |
428870ff BB |
3815 | } |
3816 | ||
812e91a7 MT |
3817 | szp = ITOZ(sip); |
3818 | ZFS_VERIFY_ZP(szp); | |
3819 | ||
3820 | /* | |
3821 | * We check i_sb because snapshots and the ctldir must have different | |
3822 | * super blocks. | |
3823 | */ | |
c0ebc844 | 3824 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
3558fd73 | 3825 | ZFS_EXIT(zsb); |
2e528b49 | 3826 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 3827 | } |
428870ff | 3828 | |
428870ff BB |
3829 | /* Prevent links to .zfs/shares files */ |
3830 | ||
3558fd73 | 3831 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
428870ff | 3832 | &parent, sizeof (uint64_t))) != 0) { |
3558fd73 | 3833 | ZFS_EXIT(zsb); |
428870ff BB |
3834 | return (error); |
3835 | } | |
3558fd73 BB |
3836 | if (parent == zsb->z_shares_dir) { |
3837 | ZFS_EXIT(zsb); | |
2e528b49 | 3838 | return (SET_ERROR(EPERM)); |
428870ff BB |
3839 | } |
3840 | ||
3558fd73 | 3841 | if (zsb->z_utf8 && u8_validate(name, |
34dc7c2f | 3842 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3843 | ZFS_EXIT(zsb); |
2e528b49 | 3844 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3845 | } |
3846 | if (flags & FIGNORECASE) | |
3847 | zf |= ZCILOOK; | |
3848 | ||
34dc7c2f BB |
3849 | /* |
3850 | * We do not support links between attributes and non-attributes | |
3851 | * because of the potential security risk of creating links | |
3852 | * into "normal" file space in order to circumvent restrictions | |
3853 | * imposed in attribute space. | |
3854 | */ | |
428870ff | 3855 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 3856 | ZFS_EXIT(zsb); |
2e528b49 | 3857 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3858 | } |
3859 | ||
2c6abf15 | 3860 | owner = zfs_fuid_map_id(zsb, KUID_TO_SUID(sip->i_uid), cr, ZFS_OWNER); |
572e2857 | 3861 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
3558fd73 | 3862 | ZFS_EXIT(zsb); |
2e528b49 | 3863 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
3864 | } |
3865 | ||
149e873a | 3866 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
3558fd73 | 3867 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3868 | return (error); |
3869 | } | |
3870 | ||
428870ff | 3871 | top: |
34dc7c2f BB |
3872 | /* |
3873 | * Attempt to lock directory; fail if entry already exists. | |
3874 | */ | |
3875 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
3876 | if (error) { | |
3558fd73 | 3877 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3878 | return (error); |
3879 | } | |
3880 | ||
3558fd73 | 3881 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 3882 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 3883 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3884 | zfs_sa_upgrade_txholds(tx, szp); |
3885 | zfs_sa_upgrade_txholds(tx, dzp); | |
e8b96c60 | 3886 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3887 | if (error) { |
3888 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3889 | if (error == ERESTART) { |
e8b96c60 | 3890 | waited = B_TRUE; |
34dc7c2f BB |
3891 | dmu_tx_wait(tx); |
3892 | dmu_tx_abort(tx); | |
3893 | goto top; | |
3894 | } | |
3895 | dmu_tx_abort(tx); | |
3558fd73 | 3896 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3897 | return (error); |
3898 | } | |
3899 | ||
3900 | error = zfs_link_create(dl, szp, tx, 0); | |
3901 | ||
3902 | if (error == 0) { | |
3903 | uint64_t txtype = TX_LINK; | |
3904 | if (flags & FIGNORECASE) | |
3905 | txtype |= TX_CI; | |
3906 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); | |
3907 | } | |
3908 | ||
3909 | dmu_tx_commit(tx); | |
3910 | ||
3911 | zfs_dirent_unlock(dl); | |
3912 | ||
3558fd73 | 3913 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3914 | zil_commit(zilog, 0); |
428870ff | 3915 | |
960e08fe BB |
3916 | zfs_inode_update(dzp); |
3917 | zfs_inode_update(szp); | |
3558fd73 | 3918 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3919 | return (error); |
3920 | } | |
e5c39b95 | 3921 | EXPORT_SYMBOL(zfs_link); |
34dc7c2f | 3922 | |
3c0e5c0f | 3923 | static void |
119a394a | 3924 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
3925 | { |
3926 | struct page *pp = arg; | |
3927 | ||
119a394a | 3928 | ClearPageError(pp); |
3c0e5c0f BB |
3929 | end_page_writeback(pp); |
3930 | } | |
3931 | ||
34dc7c2f | 3932 | /* |
3c0e5c0f BB |
3933 | * Push a page out to disk, once the page is on stable storage the |
3934 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 3935 | * |
3c0e5c0f BB |
3936 | * IN: ip - page mapped for inode. |
3937 | * pp - page to push (page is locked) | |
3938 | * wbc - writeback control data | |
34dc7c2f BB |
3939 | * |
3940 | * RETURN: 0 if success | |
3941 | * error code if failure | |
3942 | * | |
3c0e5c0f BB |
3943 | * Timestamps: |
3944 | * ip - ctime|mtime updated | |
34dc7c2f BB |
3945 | */ |
3946 | /* ARGSUSED */ | |
3c0e5c0f BB |
3947 | int |
3948 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 3949 | { |
3c0e5c0f BB |
3950 | znode_t *zp = ITOZ(ip); |
3951 | zfs_sb_t *zsb = ITOZSB(ip); | |
3952 | loff_t offset; | |
3953 | loff_t pgoff; | |
4c837f0d BB |
3954 | unsigned int pglen; |
3955 | rl_t *rl; | |
3c0e5c0f BB |
3956 | dmu_tx_t *tx; |
3957 | caddr_t va; | |
3958 | int err = 0; | |
3959 | uint64_t mtime[2], ctime[2]; | |
3960 | sa_bulk_attr_t bulk[3]; | |
3961 | int cnt = 0; | |
21a96fb6 | 3962 | struct address_space *mapping; |
3c0e5c0f | 3963 | |
4c837f0d BB |
3964 | ZFS_ENTER(zsb); |
3965 | ZFS_VERIFY_ZP(zp); | |
d164b209 | 3966 | |
3c0e5c0f BB |
3967 | ASSERT(PageLocked(pp)); |
3968 | ||
d1d7e268 MK |
3969 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
3970 | offset = i_size_read(ip); /* File length in bytes */ | |
8b1899d3 BB |
3971 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
3972 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
3973 | |
3974 | /* Page is beyond end of file */ | |
3975 | if (pgoff >= offset) { | |
3976 | unlock_page(pp); | |
4c837f0d | 3977 | ZFS_EXIT(zsb); |
3c0e5c0f BB |
3978 | return (0); |
3979 | } | |
3980 | ||
3981 | /* Truncate page length to end of file */ | |
3982 | if (pgoff + pglen > offset) | |
3983 | pglen = offset - pgoff; | |
3984 | ||
3985 | #if 0 | |
34dc7c2f | 3986 | /* |
3c0e5c0f BB |
3987 | * FIXME: Allow mmap writes past its quota. The correct fix |
3988 | * is to register a page_mkwrite() handler to count the page | |
3989 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 3990 | */ |
dde471ef PJ |
3991 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
3992 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 | 3993 | err = EDQUOT; |
9babb374 | 3994 | } |
3c0e5c0f BB |
3995 | #endif |
3996 | ||
d958324f BB |
3997 | /* |
3998 | * The ordering here is critical and must adhere to the following | |
3999 | * rules in order to avoid deadlocking in either zfs_read() or | |
4000 | * zfs_free_range() due to a lock inversion. | |
4001 | * | |
4002 | * 1) The page must be unlocked prior to acquiring the range lock. | |
4003 | * This is critical because zfs_read() calls find_lock_page() | |
4004 | * which may block on the page lock while holding the range lock. | |
4005 | * | |
4006 | * 2) Before setting or clearing write back on a page the range lock | |
4007 | * must be held in order to prevent a lock inversion with the | |
4008 | * zfs_free_range() function. | |
21a96fb6 CC |
4009 | * |
4010 | * This presents a problem because upon entering this function the | |
4011 | * page lock is already held. To safely acquire the range lock the | |
4012 | * page lock must be dropped. This creates a window where another | |
4013 | * process could truncate, invalidate, dirty, or write out the page. | |
4014 | * | |
4015 | * Therefore, after successfully reacquiring the range and page locks | |
4016 | * the current page state is checked. In the common case everything | |
4017 | * will be as is expected and it can be written out. However, if | |
4018 | * the page state has changed it must be handled accordingly. | |
d958324f | 4019 | */ |
21a96fb6 CC |
4020 | mapping = pp->mapping; |
4021 | redirty_page_for_writepage(wbc, pp); | |
d958324f | 4022 | unlock_page(pp); |
21a96fb6 | 4023 | |
d88895a0 | 4024 | rl = zfs_range_lock(&zp->z_range_lock, pgoff, pglen, RL_WRITER); |
21a96fb6 CC |
4025 | lock_page(pp); |
4026 | ||
4027 | /* Page mapping changed or it was no longer dirty, we're done */ | |
4028 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
4029 | unlock_page(pp); | |
4030 | zfs_range_unlock(rl); | |
4031 | ZFS_EXIT(zsb); | |
4032 | return (0); | |
4033 | } | |
4034 | ||
4035 | /* Another process started write block if required */ | |
4036 | if (PageWriteback(pp)) { | |
4037 | unlock_page(pp); | |
4038 | zfs_range_unlock(rl); | |
4039 | ||
4040 | if (wbc->sync_mode != WB_SYNC_NONE) | |
4041 | wait_on_page_writeback(pp); | |
4042 | ||
4043 | ZFS_EXIT(zsb); | |
4044 | return (0); | |
4045 | } | |
4046 | ||
4047 | /* Clear the dirty flag the required locks are held */ | |
4048 | if (!clear_page_dirty_for_io(pp)) { | |
4049 | unlock_page(pp); | |
4050 | zfs_range_unlock(rl); | |
4051 | ZFS_EXIT(zsb); | |
4052 | return (0); | |
4053 | } | |
4054 | ||
4055 | /* | |
4056 | * Counterpart for redirty_page_for_writepage() above. This page | |
4057 | * was in fact not skipped and should not be counted as if it were. | |
4058 | */ | |
4059 | wbc->pages_skipped--; | |
3c0e5c0f | 4060 | set_page_writeback(pp); |
21a96fb6 | 4061 | unlock_page(pp); |
3c0e5c0f | 4062 | |
dde471ef | 4063 | tx = dmu_tx_create(zsb->z_os); |
3c0e5c0f | 4064 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
4065 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4066 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 4067 | |
fb5f0bc8 | 4068 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 4069 | if (err != 0) { |
3c0e5c0f | 4070 | if (err == ERESTART) |
34dc7c2f | 4071 | dmu_tx_wait(tx); |
3c0e5c0f | 4072 | |
34dc7c2f | 4073 | dmu_tx_abort(tx); |
119a394a ED |
4074 | __set_page_dirty_nobuffers(pp); |
4075 | ClearPageError(pp); | |
4076 | end_page_writeback(pp); | |
4c837f0d BB |
4077 | zfs_range_unlock(rl); |
4078 | ZFS_EXIT(zsb); | |
3c0e5c0f | 4079 | return (err); |
34dc7c2f BB |
4080 | } |
4081 | ||
dde471ef | 4082 | va = kmap(pp); |
8b1899d3 | 4083 | ASSERT3U(pglen, <=, PAGE_SIZE); |
3c0e5c0f | 4084 | dmu_write(zsb->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 4085 | kunmap(pp); |
34dc7c2f | 4086 | |
3c0e5c0f BB |
4087 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
4088 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
4089 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, 8); | |
428870ff | 4090 | |
d3aa3ea9 BB |
4091 | /* Preserve the mtime and ctime provided by the inode */ |
4092 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4093 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
4094 | zp->z_atime_dirty = 0; | |
4095 | zp->z_seq++; | |
4096 | ||
4097 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4098 | ||
119a394a ED |
4099 | zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
4100 | zfs_putpage_commit_cb, pp); | |
45d1cae3 | 4101 | dmu_tx_commit(tx); |
d3aa3ea9 | 4102 | |
4c837f0d | 4103 | zfs_range_unlock(rl); |
34dc7c2f | 4104 | |
119a394a ED |
4105 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4106 | /* | |
4107 | * Note that this is rarely called under writepages(), because | |
4108 | * writepages() normally handles the entire commit for | |
4109 | * performance reasons. | |
4110 | */ | |
1e8db771 BB |
4111 | if (zsb->z_log != NULL) |
4112 | zil_commit(zsb->z_log, zp->z_id); | |
2b286136 | 4113 | } |
3c0e5c0f | 4114 | |
4c837f0d | 4115 | ZFS_EXIT(zsb); |
3c0e5c0f | 4116 | return (err); |
34dc7c2f BB |
4117 | } |
4118 | ||
8780c539 BB |
4119 | /* |
4120 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 4121 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
4122 | */ |
4123 | int | |
4124 | zfs_dirty_inode(struct inode *ip, int flags) | |
4125 | { | |
4126 | znode_t *zp = ITOZ(ip); | |
4127 | zfs_sb_t *zsb = ITOZSB(ip); | |
4128 | dmu_tx_t *tx; | |
023699cd MM |
4129 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
4130 | sa_bulk_attr_t bulk[4]; | |
704cd075 | 4131 | int error = 0; |
8780c539 BB |
4132 | int cnt = 0; |
4133 | ||
c944be5d BB |
4134 | if (zfs_is_readonly(zsb) || dmu_objset_is_snapshot(zsb->z_os)) |
4135 | return (0); | |
4136 | ||
8780c539 BB |
4137 | ZFS_ENTER(zsb); |
4138 | ZFS_VERIFY_ZP(zp); | |
4139 | ||
704cd075 CC |
4140 | #ifdef I_DIRTY_TIME |
4141 | /* | |
4142 | * This is the lazytime semantic indroduced in Linux 4.0 | |
4143 | * This flag will only be called from update_time when lazytime is set. | |
4144 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
4145 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
4146 | * only need to dirty atime. | |
4147 | */ | |
4148 | if (flags == I_DIRTY_TIME) { | |
4149 | zp->z_atime_dirty = 1; | |
4150 | goto out; | |
4151 | } | |
4152 | #endif | |
4153 | ||
8780c539 BB |
4154 | tx = dmu_tx_create(zsb->z_os); |
4155 | ||
4156 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4157 | zfs_sa_upgrade_txholds(tx, zp); | |
4158 | ||
4159 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4160 | if (error) { | |
4161 | dmu_tx_abort(tx); | |
4162 | goto out; | |
4163 | } | |
4164 | ||
4165 | mutex_enter(&zp->z_lock); | |
704cd075 CC |
4166 | zp->z_atime_dirty = 0; |
4167 | ||
023699cd | 4168 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zsb), NULL, &mode, 8); |
8780c539 BB |
4169 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zsb), NULL, &atime, 16); |
4170 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); | |
4171 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
4172 | ||
023699cd | 4173 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4174 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4175 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4176 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
023699cd MM |
4177 | mode = ip->i_mode; |
4178 | ||
4179 | zp->z_mode = mode; | |
8780c539 BB |
4180 | |
4181 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4182 | mutex_exit(&zp->z_lock); | |
4183 | ||
4184 | dmu_tx_commit(tx); | |
4185 | out: | |
4186 | ZFS_EXIT(zsb); | |
4187 | return (error); | |
4188 | } | |
4189 | EXPORT_SYMBOL(zfs_dirty_inode); | |
4190 | ||
34dc7c2f BB |
4191 | /*ARGSUSED*/ |
4192 | void | |
c0d35759 | 4193 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4194 | { |
c0d35759 BB |
4195 | znode_t *zp = ITOZ(ip); |
4196 | zfs_sb_t *zsb = ITOZSB(ip); | |
0df9673f | 4197 | uint64_t atime[2]; |
34dc7c2f | 4198 | int error; |
cafbd2ac | 4199 | int need_unlock = 0; |
34dc7c2f | 4200 | |
cafbd2ac | 4201 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
4202 | if (!RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)) { | |
4203 | need_unlock = 1; | |
4204 | rw_enter(&zsb->z_teardown_inactive_lock, RW_READER); | |
4205 | } | |
c0d35759 | 4206 | if (zp->z_sa_hdl == NULL) { |
cafbd2ac | 4207 | if (need_unlock) |
4208 | rw_exit(&zsb->z_teardown_inactive_lock); | |
c0d35759 | 4209 | return; |
34dc7c2f BB |
4210 | } |
4211 | ||
4212 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
c0d35759 | 4213 | dmu_tx_t *tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 4214 | |
428870ff BB |
4215 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4216 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4217 | error = dmu_tx_assign(tx, TXG_WAIT); |
4218 | if (error) { | |
4219 | dmu_tx_abort(tx); | |
4220 | } else { | |
0df9673f | 4221 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
34dc7c2f | 4222 | mutex_enter(&zp->z_lock); |
3558fd73 | 4223 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), |
0df9673f | 4224 | (void *)&atime, sizeof (atime), tx); |
34dc7c2f BB |
4225 | zp->z_atime_dirty = 0; |
4226 | mutex_exit(&zp->z_lock); | |
4227 | dmu_tx_commit(tx); | |
4228 | } | |
4229 | } | |
4230 | ||
4231 | zfs_zinactive(zp); | |
cafbd2ac | 4232 | if (need_unlock) |
4233 | rw_exit(&zsb->z_teardown_inactive_lock); | |
34dc7c2f | 4234 | } |
e5c39b95 | 4235 | EXPORT_SYMBOL(zfs_inactive); |
34dc7c2f BB |
4236 | |
4237 | /* | |
4238 | * Bounds-check the seek operation. | |
4239 | * | |
3558fd73 | 4240 | * IN: ip - inode seeking within |
34dc7c2f BB |
4241 | * ooff - old file offset |
4242 | * noffp - pointer to new file offset | |
4243 | * ct - caller context | |
4244 | * | |
4245 | * RETURN: 0 if success | |
4246 | * EINVAL if new offset invalid | |
4247 | */ | |
4248 | /* ARGSUSED */ | |
3558fd73 | 4249 | int |
9623f736 | 4250 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4251 | { |
3558fd73 | 4252 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4253 | return (0); |
4254 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4255 | } | |
3558fd73 | 4256 | EXPORT_SYMBOL(zfs_seek); |
34dc7c2f | 4257 | |
34dc7c2f | 4258 | /* |
dde471ef | 4259 | * Fill pages with data from the disk. |
34dc7c2f BB |
4260 | */ |
4261 | static int | |
dde471ef | 4262 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4263 | { |
d1d7e268 MK |
4264 | znode_t *zp = ITOZ(ip); |
4265 | zfs_sb_t *zsb = ITOZSB(ip); | |
4266 | objset_t *os; | |
dde471ef | 4267 | struct page *cur_pp; |
d1d7e268 MK |
4268 | u_offset_t io_off, total; |
4269 | size_t io_len; | |
4270 | loff_t i_size; | |
4271 | unsigned page_idx; | |
4272 | int err; | |
34dc7c2f | 4273 | |
d1d7e268 | 4274 | os = zsb->z_os; |
8b1899d3 | 4275 | io_len = nr_pages << PAGE_SHIFT; |
dde471ef PJ |
4276 | i_size = i_size_read(ip); |
4277 | io_off = page_offset(pl[0]); | |
4278 | ||
4279 | if (io_off + io_len > i_size) | |
4280 | io_len = i_size - io_off; | |
34dc7c2f BB |
4281 | |
4282 | /* | |
dde471ef | 4283 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4284 | */ |
dde471ef | 4285 | page_idx = 0; |
34dc7c2f | 4286 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4287 | caddr_t va; |
4288 | ||
540c3927 | 4289 | cur_pp = pl[page_idx++]; |
dde471ef | 4290 | va = kmap(cur_pp); |
9babb374 BB |
4291 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4292 | DMU_READ_PREFETCH); | |
dde471ef | 4293 | kunmap(cur_pp); |
34dc7c2f | 4294 | if (err) { |
b128c09f BB |
4295 | /* convert checksum errors into IO errors */ |
4296 | if (err == ECKSUM) | |
2e528b49 | 4297 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4298 | return (err); |
4299 | } | |
34dc7c2f | 4300 | } |
d164b209 | 4301 | |
34dc7c2f BB |
4302 | return (0); |
4303 | } | |
4304 | ||
4305 | /* | |
dde471ef | 4306 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4307 | * |
dde471ef PJ |
4308 | * IN: ip - inode of file to get data from. |
4309 | * pl - list of pages to read | |
4310 | * nr_pages - number of pages to read | |
34dc7c2f | 4311 | * |
d3cc8b15 | 4312 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4313 | * |
4314 | * Timestamps: | |
4315 | * vp - atime updated | |
4316 | */ | |
4317 | /* ARGSUSED */ | |
dde471ef PJ |
4318 | int |
4319 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4320 | { |
dde471ef PJ |
4321 | znode_t *zp = ITOZ(ip); |
4322 | zfs_sb_t *zsb = ITOZSB(ip); | |
4323 | int err; | |
d164b209 | 4324 | |
d164b209 BB |
4325 | if (pl == NULL) |
4326 | return (0); | |
34dc7c2f | 4327 | |
dde471ef | 4328 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4329 | ZFS_VERIFY_ZP(zp); |
4330 | ||
dde471ef | 4331 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4332 | |
dde471ef | 4333 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4334 | return (err); |
4335 | } | |
dde471ef | 4336 | EXPORT_SYMBOL(zfs_getpage); |
34dc7c2f BB |
4337 | |
4338 | /* | |
e2e7aa2d | 4339 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4340 | * |
e2e7aa2d BB |
4341 | * IN: ip - inode of the file to mmap |
4342 | * off - file offset | |
4343 | * addrp - start address in memory region | |
4344 | * len - length of memory region | |
4345 | * vm_flags- address flags | |
34dc7c2f | 4346 | * |
e2e7aa2d BB |
4347 | * RETURN: 0 if success |
4348 | * error code if failure | |
34dc7c2f BB |
4349 | */ |
4350 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4351 | int |
4352 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4353 | unsigned long vm_flags) | |
34dc7c2f | 4354 | { |
e2e7aa2d BB |
4355 | znode_t *zp = ITOZ(ip); |
4356 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4357 | |
e2e7aa2d | 4358 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4359 | ZFS_VERIFY_ZP(zp); |
4360 | ||
e2e7aa2d | 4361 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4362 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
e2e7aa2d | 4363 | ZFS_EXIT(zsb); |
2e528b49 | 4364 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4365 | } |
4366 | ||
e2e7aa2d | 4367 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4368 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
e2e7aa2d | 4369 | ZFS_EXIT(zsb); |
2e528b49 | 4370 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4371 | } |
4372 | ||
34dc7c2f | 4373 | if (off < 0 || len > MAXOFFSET_T - off) { |
e2e7aa2d | 4374 | ZFS_EXIT(zsb); |
2e528b49 | 4375 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4376 | } |
4377 | ||
e2e7aa2d | 4378 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4379 | return (0); |
4380 | } | |
e2e7aa2d | 4381 | EXPORT_SYMBOL(zfs_map); |
34dc7c2f | 4382 | |
3558fd73 BB |
4383 | /* |
4384 | * convoff - converts the given data (start, whence) to the | |
4385 | * given whence. | |
4386 | */ | |
4387 | int | |
4388 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4389 | { | |
5484965a | 4390 | vattr_t vap; |
3558fd73 BB |
4391 | int error; |
4392 | ||
4393 | if ((lckdat->l_whence == 2) || (whence == 2)) { | |
5484965a | 4394 | if ((error = zfs_getattr(ip, &vap, 0, CRED()) != 0)) |
3558fd73 BB |
4395 | return (error); |
4396 | } | |
4397 | ||
4398 | switch (lckdat->l_whence) { | |
4399 | case 1: | |
4400 | lckdat->l_start += offset; | |
4401 | break; | |
4402 | case 2: | |
5484965a | 4403 | lckdat->l_start += vap.va_size; |
3558fd73 BB |
4404 | /* FALLTHRU */ |
4405 | case 0: | |
4406 | break; | |
4407 | default: | |
2e528b49 | 4408 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4409 | } |
4410 | ||
4411 | if (lckdat->l_start < 0) | |
2e528b49 | 4412 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4413 | |
4414 | switch (whence) { | |
4415 | case 1: | |
4416 | lckdat->l_start -= offset; | |
4417 | break; | |
4418 | case 2: | |
5484965a | 4419 | lckdat->l_start -= vap.va_size; |
3558fd73 BB |
4420 | /* FALLTHRU */ |
4421 | case 0: | |
4422 | break; | |
4423 | default: | |
2e528b49 | 4424 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4425 | } |
4426 | ||
4427 | lckdat->l_whence = (short)whence; | |
4428 | return (0); | |
4429 | } | |
4430 | ||
34dc7c2f BB |
4431 | /* |
4432 | * Free or allocate space in a file. Currently, this function only | |
4433 | * supports the `F_FREESP' command. However, this command is somewhat | |
4434 | * misnamed, as its functionality includes the ability to allocate as | |
4435 | * well as free space. | |
4436 | * | |
3558fd73 | 4437 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4438 | * cmd - action to take (only F_FREESP supported). |
4439 | * bfp - section of file to free/alloc. | |
4440 | * flag - current file open mode flags. | |
4441 | * offset - current file offset. | |
4442 | * cr - credentials of caller [UNUSED]. | |
34dc7c2f | 4443 | * |
d3cc8b15 | 4444 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4445 | * |
4446 | * Timestamps: | |
3558fd73 | 4447 | * ip - ctime|mtime updated |
34dc7c2f BB |
4448 | */ |
4449 | /* ARGSUSED */ | |
e5c39b95 | 4450 | int |
3558fd73 BB |
4451 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4452 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4453 | { |
3558fd73 BB |
4454 | znode_t *zp = ITOZ(ip); |
4455 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4456 | uint64_t off, len; |
4457 | int error; | |
4458 | ||
3558fd73 | 4459 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4460 | ZFS_VERIFY_ZP(zp); |
4461 | ||
34dc7c2f | 4462 | if (cmd != F_FREESP) { |
3558fd73 | 4463 | ZFS_EXIT(zsb); |
2e528b49 | 4464 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4465 | } |
4466 | ||
f3c9dca0 MT |
4467 | /* |
4468 | * Callers might not be able to detect properly that we are read-only, | |
4469 | * so check it explicitly here. | |
4470 | */ | |
4471 | if (zfs_is_readonly(zsb)) { | |
4472 | ZFS_EXIT(zsb); | |
4473 | return (SET_ERROR(EROFS)); | |
4474 | } | |
4475 | ||
3558fd73 BB |
4476 | if ((error = convoff(ip, bfp, 0, offset))) { |
4477 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
4478 | return (error); |
4479 | } | |
4480 | ||
4481 | if (bfp->l_len < 0) { | |
3558fd73 | 4482 | ZFS_EXIT(zsb); |
2e528b49 | 4483 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4484 | } |
4485 | ||
aec69371 ED |
4486 | /* |
4487 | * Permissions aren't checked on Solaris because on this OS | |
4488 | * zfs_space() can only be called with an opened file handle. | |
4489 | * On Linux we can get here through truncate_range() which | |
4490 | * operates directly on inodes, so we need to check access rights. | |
4491 | */ | |
4492 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
4493 | ZFS_EXIT(zsb); | |
4494 | return (error); | |
4495 | } | |
4496 | ||
34dc7c2f BB |
4497 | off = bfp->l_start; |
4498 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4499 | ||
b128c09f | 4500 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4501 | |
3558fd73 | 4502 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4503 | return (error); |
4504 | } | |
e5c39b95 | 4505 | EXPORT_SYMBOL(zfs_space); |
34dc7c2f BB |
4506 | |
4507 | /*ARGSUSED*/ | |
e5c39b95 | 4508 | int |
3558fd73 | 4509 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4510 | { |
3558fd73 BB |
4511 | znode_t *zp = ITOZ(ip); |
4512 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4513 | uint32_t gen; |
428870ff | 4514 | uint64_t gen64; |
34dc7c2f BB |
4515 | uint64_t object = zp->z_id; |
4516 | zfid_short_t *zfid; | |
428870ff | 4517 | int size, i, error; |
34dc7c2f | 4518 | |
3558fd73 | 4519 | ZFS_ENTER(zsb); |
34dc7c2f | 4520 | ZFS_VERIFY_ZP(zp); |
428870ff | 4521 | |
3558fd73 | 4522 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), |
428870ff | 4523 | &gen64, sizeof (uint64_t))) != 0) { |
3558fd73 | 4524 | ZFS_EXIT(zsb); |
428870ff BB |
4525 | return (error); |
4526 | } | |
4527 | ||
4528 | gen = (uint32_t)gen64; | |
34dc7c2f | 4529 | |
3558fd73 | 4530 | size = (zsb->z_parent != zsb) ? LONG_FID_LEN : SHORT_FID_LEN; |
34dc7c2f BB |
4531 | if (fidp->fid_len < size) { |
4532 | fidp->fid_len = size; | |
3558fd73 | 4533 | ZFS_EXIT(zsb); |
2e528b49 | 4534 | return (SET_ERROR(ENOSPC)); |
34dc7c2f BB |
4535 | } |
4536 | ||
4537 | zfid = (zfid_short_t *)fidp; | |
4538 | ||
4539 | zfid->zf_len = size; | |
4540 | ||
4541 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4542 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4543 | ||
4544 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4545 | if (gen == 0) | |
4546 | gen = 1; | |
4547 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4548 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4549 | ||
4550 | if (size == LONG_FID_LEN) { | |
3558fd73 | 4551 | uint64_t objsetid = dmu_objset_id(zsb->z_os); |
34dc7c2f BB |
4552 | zfid_long_t *zlfid; |
4553 | ||
4554 | zlfid = (zfid_long_t *)fidp; | |
4555 | ||
4556 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
4557 | zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); | |
4558 | ||
4559 | /* XXX - this should be the generation number for the objset */ | |
4560 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
4561 | zlfid->zf_setgen[i] = 0; | |
4562 | } | |
4563 | ||
3558fd73 | 4564 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4565 | return (0); |
4566 | } | |
e5c39b95 | 4567 | EXPORT_SYMBOL(zfs_fid); |
34dc7c2f | 4568 | |
34dc7c2f | 4569 | /*ARGSUSED*/ |
e5c39b95 | 4570 | int |
3558fd73 | 4571 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4572 | { |
3558fd73 BB |
4573 | znode_t *zp = ITOZ(ip); |
4574 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4575 | int error; |
4576 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
4577 | ||
3558fd73 | 4578 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4579 | ZFS_VERIFY_ZP(zp); |
4580 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
3558fd73 | 4581 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4582 | |
4583 | return (error); | |
4584 | } | |
e5c39b95 | 4585 | EXPORT_SYMBOL(zfs_getsecattr); |
34dc7c2f BB |
4586 | |
4587 | /*ARGSUSED*/ | |
e5c39b95 | 4588 | int |
3558fd73 | 4589 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4590 | { |
3558fd73 BB |
4591 | znode_t *zp = ITOZ(ip); |
4592 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4593 | int error; |
4594 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
3558fd73 | 4595 | zilog_t *zilog = zsb->z_log; |
34dc7c2f | 4596 | |
3558fd73 | 4597 | ZFS_ENTER(zsb); |
34dc7c2f | 4598 | ZFS_VERIFY_ZP(zp); |
428870ff | 4599 | |
34dc7c2f | 4600 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 4601 | |
3558fd73 | 4602 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4603 | zil_commit(zilog, 0); |
428870ff | 4604 | |
3558fd73 | 4605 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4606 | return (error); |
4607 | } | |
e5c39b95 | 4608 | EXPORT_SYMBOL(zfs_setsecattr); |
34dc7c2f | 4609 | |
3558fd73 | 4610 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
4611 | /* |
4612 | * Tunable, both must be a power of 2. | |
4613 | * | |
4614 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
4615 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 4616 | * an arcbuf for a partial block read |
428870ff BB |
4617 | */ |
4618 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
4619 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
4620 | ||
4621 | /*ARGSUSED*/ | |
4622 | static int | |
3558fd73 | 4623 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 4624 | { |
3558fd73 BB |
4625 | znode_t *zp = ITOZ(ip); |
4626 | zfs_sb_t *zsb = ITOZSB(ip); | |
4627 | int max_blksz = zsb->z_max_blksz; | |
428870ff BB |
4628 | uio_t *uio = &xuio->xu_uio; |
4629 | ssize_t size = uio->uio_resid; | |
4630 | offset_t offset = uio->uio_loffset; | |
4631 | int blksz; | |
4632 | int fullblk, i; | |
4633 | arc_buf_t *abuf; | |
4634 | ssize_t maxsize; | |
4635 | int preamble, postamble; | |
4636 | ||
4637 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
2e528b49 | 4638 | return (SET_ERROR(EINVAL)); |
428870ff | 4639 | |
3558fd73 | 4640 | ZFS_ENTER(zsb); |
428870ff BB |
4641 | ZFS_VERIFY_ZP(zp); |
4642 | switch (ioflag) { | |
4643 | case UIO_WRITE: | |
4644 | /* | |
4645 | * Loan out an arc_buf for write if write size is bigger than | |
4646 | * max_blksz, and the file's block size is also max_blksz. | |
4647 | */ | |
4648 | blksz = max_blksz; | |
4649 | if (size < blksz || zp->z_blksz != blksz) { | |
3558fd73 | 4650 | ZFS_EXIT(zsb); |
2e528b49 | 4651 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4652 | } |
4653 | /* | |
4654 | * Caller requests buffers for write before knowing where the | |
4655 | * write offset might be (e.g. NFS TCP write). | |
4656 | */ | |
4657 | if (offset == -1) { | |
4658 | preamble = 0; | |
4659 | } else { | |
4660 | preamble = P2PHASE(offset, blksz); | |
4661 | if (preamble) { | |
4662 | preamble = blksz - preamble; | |
4663 | size -= preamble; | |
4664 | } | |
4665 | } | |
4666 | ||
4667 | postamble = P2PHASE(size, blksz); | |
4668 | size -= postamble; | |
4669 | ||
4670 | fullblk = size / blksz; | |
4671 | (void) dmu_xuio_init(xuio, | |
4672 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
4673 | |
4674 | /* | |
4675 | * Have to fix iov base/len for partial buffers. They | |
4676 | * currently represent full arc_buf's. | |
4677 | */ | |
4678 | if (preamble) { | |
4679 | /* data begins in the middle of the arc_buf */ | |
4680 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4681 | blksz); | |
4682 | ASSERT(abuf); | |
4683 | (void) dmu_xuio_add(xuio, abuf, | |
4684 | blksz - preamble, preamble); | |
4685 | } | |
4686 | ||
4687 | for (i = 0; i < fullblk; i++) { | |
4688 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4689 | blksz); | |
4690 | ASSERT(abuf); | |
4691 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
4692 | } | |
4693 | ||
4694 | if (postamble) { | |
4695 | /* data ends in the middle of the arc_buf */ | |
4696 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4697 | blksz); | |
4698 | ASSERT(abuf); | |
4699 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
4700 | } | |
4701 | break; | |
4702 | case UIO_READ: | |
4703 | /* | |
4704 | * Loan out an arc_buf for read if the read size is larger than | |
4705 | * the current file block size. Block alignment is not | |
4706 | * considered. Partial arc_buf will be loaned out for read. | |
4707 | */ | |
4708 | blksz = zp->z_blksz; | |
4709 | if (blksz < zcr_blksz_min) | |
4710 | blksz = zcr_blksz_min; | |
4711 | if (blksz > zcr_blksz_max) | |
4712 | blksz = zcr_blksz_max; | |
4713 | /* avoid potential complexity of dealing with it */ | |
4714 | if (blksz > max_blksz) { | |
3558fd73 | 4715 | ZFS_EXIT(zsb); |
2e528b49 | 4716 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4717 | } |
4718 | ||
4719 | maxsize = zp->z_size - uio->uio_loffset; | |
4720 | if (size > maxsize) | |
4721 | size = maxsize; | |
4722 | ||
3558fd73 BB |
4723 | if (size < blksz) { |
4724 | ZFS_EXIT(zsb); | |
2e528b49 | 4725 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4726 | } |
4727 | break; | |
4728 | default: | |
3558fd73 | 4729 | ZFS_EXIT(zsb); |
2e528b49 | 4730 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4731 | } |
4732 | ||
4733 | uio->uio_extflg = UIO_XUIO; | |
4734 | XUIO_XUZC_RW(xuio) = ioflag; | |
3558fd73 | 4735 | ZFS_EXIT(zsb); |
428870ff BB |
4736 | return (0); |
4737 | } | |
4738 | ||
4739 | /*ARGSUSED*/ | |
4740 | static int | |
3558fd73 | 4741 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
4742 | { |
4743 | int i; | |
4744 | arc_buf_t *abuf; | |
4745 | int ioflag = XUIO_XUZC_RW(xuio); | |
4746 | ||
4747 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
4748 | ||
4749 | i = dmu_xuio_cnt(xuio); | |
4750 | while (i-- > 0) { | |
4751 | abuf = dmu_xuio_arcbuf(xuio, i); | |
4752 | /* | |
4753 | * if abuf == NULL, it must be a write buffer | |
4754 | * that has been returned in zfs_write(). | |
4755 | */ | |
4756 | if (abuf) | |
4757 | dmu_return_arcbuf(abuf); | |
4758 | ASSERT(abuf || ioflag == UIO_WRITE); | |
4759 | } | |
4760 | ||
4761 | dmu_xuio_fini(xuio); | |
4762 | return (0); | |
4763 | } | |
3558fd73 | 4764 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 BB |
4765 | |
4766 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
a966c564 K |
4767 | module_param(zfs_delete_blocks, ulong, 0644); |
4768 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
c409e464 BB |
4769 | module_param(zfs_read_chunk_size, long, 0644); |
4770 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
4771 | #endif |