]>
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 | */ | |
9b7b9cd3 | 21 | |
34dc7c2f | 22 | /* |
428870ff | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
19d55079 | 24 | * Copyright (c) 2012, 2015 by Delphix. All rights reserved. |
5475aada | 25 | * Copyright (c) 2015 by Chunwei Chen. All rights reserved. |
9b7b9cd3 | 26 | * Copyright 2017 Nexenta Systems, Inc. |
34dc7c2f BB |
27 | */ |
28 | ||
29 | /* Portions Copyright 2007 Jeremy Teo */ | |
428870ff | 30 | /* Portions Copyright 2010 Robert Milkowski */ |
34dc7c2f | 31 | |
60101509 | 32 | |
34dc7c2f BB |
33 | #include <sys/types.h> |
34 | #include <sys/param.h> | |
35 | #include <sys/time.h> | |
36 | #include <sys/systm.h> | |
37 | #include <sys/sysmacros.h> | |
38 | #include <sys/resource.h> | |
39 | #include <sys/vfs.h> | |
40 | #include <sys/vfs_opreg.h> | |
34dc7c2f BB |
41 | #include <sys/file.h> |
42 | #include <sys/stat.h> | |
43 | #include <sys/kmem.h> | |
44 | #include <sys/taskq.h> | |
45 | #include <sys/uio.h> | |
46 | #include <sys/vmsystm.h> | |
47 | #include <sys/atomic.h> | |
34dc7c2f | 48 | #include <vm/pvn.h> |
34dc7c2f BB |
49 | #include <sys/pathname.h> |
50 | #include <sys/cmn_err.h> | |
51 | #include <sys/errno.h> | |
52 | #include <sys/unistd.h> | |
53 | #include <sys/zfs_dir.h> | |
54 | #include <sys/zfs_acl.h> | |
55 | #include <sys/zfs_ioctl.h> | |
56 | #include <sys/fs/zfs.h> | |
57 | #include <sys/dmu.h> | |
428870ff | 58 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
59 | #include <sys/spa.h> |
60 | #include <sys/txg.h> | |
61 | #include <sys/dbuf.h> | |
62 | #include <sys/zap.h> | |
428870ff | 63 | #include <sys/sa.h> |
34dc7c2f BB |
64 | #include <sys/dirent.h> |
65 | #include <sys/policy.h> | |
66 | #include <sys/sunddi.h> | |
b128c09f | 67 | #include <sys/sid.h> |
bcf30822 | 68 | #include <sys/mode.h> |
34dc7c2f | 69 | #include "fs/fs_subr.h" |
ebe7e575 | 70 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 71 | #include <sys/zfs_fuid.h> |
428870ff | 72 | #include <sys/zfs_sa.h> |
e5c39b95 | 73 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
74 | #include <sys/dnlc.h> |
75 | #include <sys/zfs_rlock.h> | |
76 | #include <sys/extdirent.h> | |
77 | #include <sys/kidmap.h> | |
428870ff | 78 | #include <sys/cred.h> |
34dc7c2f | 79 | #include <sys/attr.h> |
218b8eaf | 80 | #include <sys/zpl.h> |
1ce23dca | 81 | #include <sys/zil.h> |
34dc7c2f BB |
82 | |
83 | /* | |
84 | * Programming rules. | |
85 | * | |
86 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
87 | * properly lock its in-core state, create a DMU transaction, do the work, | |
88 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
89 | * and wait for the intent log to commit if it is a synchronous operation. | |
90 | * Moreover, the vnode ops must work in both normal and log replay context. | |
91 | * The ordering of events is important to avoid deadlocks and references | |
92 | * to freed memory. The example below illustrates the following Big Rules: | |
93 | * | |
94 | * (1) A check must be made in each zfs thread for a mounted file system. | |
0037b49e BB |
95 | * This is done avoiding races using ZFS_ENTER(zfsvfs). |
96 | * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes | |
34dc7c2f BB |
97 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros |
98 | * can return EIO from the calling function. | |
99 | * | |
3558fd73 | 100 | * (2) iput() should always be the last thing except for zil_commit() |
34dc7c2f BB |
101 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: |
102 | * First, if it's the last reference, the vnode/znode | |
103 | * can be freed, so the zp may point to freed memory. Second, the last | |
104 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
105 | * pushing cached pages (which acquires range locks) and syncing out | |
106 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
107 | * which could deadlock the system if you were already holding one. | |
0a50679c | 108 | * If you must call iput() within a tx then use zfs_iput_async(). |
34dc7c2f BB |
109 | * |
110 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
111 | * as they can span dmu_tx_assign() calls. | |
112 | * | |
384f8a09 MA |
113 | * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to |
114 | * dmu_tx_assign(). This is critical because we don't want to block | |
115 | * while holding locks. | |
116 | * | |
117 | * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This | |
118 | * reduces lock contention and CPU usage when we must wait (note that if | |
119 | * throughput is constrained by the storage, nearly every transaction | |
120 | * must wait). | |
121 | * | |
122 | * Note, in particular, that if a lock is sometimes acquired before | |
123 | * the tx assigns, and sometimes after (e.g. z_lock), then failing | |
124 | * to use a non-blocking assign can deadlock the system. The scenario: | |
34dc7c2f BB |
125 | * |
126 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
127 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
128 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
129 | * forever, because the previous txg can't quiesce until B's tx commits. | |
130 | * | |
0037b49e | 131 | * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, |
e8b96c60 MA |
132 | * then drop all locks, call dmu_tx_wait(), and try again. On subsequent |
133 | * calls to dmu_tx_assign(), pass TXG_WAITED rather than TXG_NOWAIT, | |
134 | * to indicate that this operation has already called dmu_tx_wait(). | |
135 | * This will ensure that we don't retry forever, waiting a short bit | |
136 | * each time. | |
34dc7c2f BB |
137 | * |
138 | * (5) If the operation succeeded, generate the intent log entry for it | |
139 | * before dropping locks. This ensures that the ordering of events | |
140 | * in the intent log matches the order in which they actually occurred. | |
d3cc8b15 | 141 | * During ZIL replay the zfs_log_* functions will update the sequence |
fb5f0bc8 | 142 | * number to indicate the zil transaction has replayed. |
34dc7c2f BB |
143 | * |
144 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
145 | * regardless of whether there were any errors. | |
146 | * | |
572e2857 | 147 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
148 | * to ensure that synchronous semantics are provided when necessary. |
149 | * | |
150 | * In general, this is how things should be ordered in each vnode op: | |
151 | * | |
0037b49e | 152 | * ZFS_ENTER(zfsvfs); // exit if unmounted |
34dc7c2f | 153 | * top: |
3558fd73 | 154 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
155 | * rw_enter(...); // grab any other locks you need |
156 | * tx = dmu_tx_create(...); // get DMU tx | |
157 | * dmu_tx_hold_*(); // hold each object you might modify | |
e8b96c60 | 158 | * error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
159 | * if (error) { |
160 | * rw_exit(...); // drop locks | |
161 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 162 | * iput(...); // release held vnodes |
fb5f0bc8 | 163 | * if (error == ERESTART) { |
e8b96c60 | 164 | * waited = B_TRUE; |
34dc7c2f BB |
165 | * dmu_tx_wait(tx); |
166 | * dmu_tx_abort(tx); | |
167 | * goto top; | |
168 | * } | |
169 | * dmu_tx_abort(tx); // abort DMU tx | |
0037b49e | 170 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
171 | * return (error); // really out of space |
172 | * } | |
173 | * error = do_real_work(); // do whatever this VOP does | |
174 | * if (error == 0) | |
175 | * zfs_log_*(...); // on success, make ZIL entry | |
176 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
177 | * rw_exit(...); // drop locks | |
178 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 179 | * iput(...); // release held vnodes |
572e2857 | 180 | * zil_commit(zilog, foid); // synchronous when necessary |
0037b49e | 181 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
182 | * return (error); // done, report error |
183 | */ | |
184 | ||
126400a1 BB |
185 | /* |
186 | * Virus scanning is unsupported. It would be possible to add a hook | |
187 | * here to performance the required virus scan. This could be done | |
188 | * entirely in the kernel or potentially as an update to invoke a | |
189 | * scanning utility. | |
190 | */ | |
191 | static int | |
192 | zfs_vscan(struct inode *ip, cred_t *cr, int async) | |
193 | { | |
194 | return (0); | |
195 | } | |
196 | ||
197 | /* ARGSUSED */ | |
198 | int | |
199 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
200 | { | |
201 | znode_t *zp = ITOZ(ip); | |
0037b49e | 202 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 203 | |
0037b49e | 204 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
205 | ZFS_VERIFY_ZP(zp); |
206 | ||
207 | /* Honor ZFS_APPENDONLY file attribute */ | |
208 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
209 | ((flag & O_APPEND) == 0)) { | |
0037b49e | 210 | ZFS_EXIT(zfsvfs); |
2e528b49 | 211 | return (SET_ERROR(EPERM)); |
126400a1 BB |
212 | } |
213 | ||
214 | /* Virus scan eligible files on open */ | |
0037b49e | 215 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
216 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { |
217 | if (zfs_vscan(ip, cr, 0) != 0) { | |
0037b49e | 218 | ZFS_EXIT(zfsvfs); |
2e528b49 | 219 | return (SET_ERROR(EACCES)); |
126400a1 BB |
220 | } |
221 | } | |
222 | ||
223 | /* Keep a count of the synchronous opens in the znode */ | |
224 | if (flag & O_SYNC) | |
225 | atomic_inc_32(&zp->z_sync_cnt); | |
226 | ||
0037b49e | 227 | ZFS_EXIT(zfsvfs); |
126400a1 BB |
228 | return (0); |
229 | } | |
126400a1 BB |
230 | |
231 | /* ARGSUSED */ | |
232 | int | |
233 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
234 | { | |
235 | znode_t *zp = ITOZ(ip); | |
0037b49e | 236 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 237 | |
0037b49e | 238 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
239 | ZFS_VERIFY_ZP(zp); |
240 | ||
7dc71949 | 241 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 242 | if (flag & O_SYNC) |
7dc71949 | 243 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 | 244 | |
0037b49e | 245 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
246 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) |
247 | VERIFY(zfs_vscan(ip, cr, 1) == 0); | |
248 | ||
0037b49e | 249 | ZFS_EXIT(zfsvfs); |
8780c539 | 250 | return (0); |
126400a1 | 251 | } |
126400a1 | 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 | 281 | |
6e03ec4f DB |
282 | /* file was dirty, so fall back to using generic logic */ |
283 | if (error == EBUSY) { | |
284 | if (hole) | |
285 | *off = file_sz; | |
286 | ||
287 | return (0); | |
288 | } | |
66aca247 | 289 | |
d97aa48f MA |
290 | /* |
291 | * We could find a hole that begins after the logical end-of-file, | |
292 | * because dmu_offset_next() only works on whole blocks. If the | |
293 | * EOF falls mid-block, then indicate that the "virtual hole" | |
294 | * at the end of the file begins at the logical EOF, rather than | |
295 | * at the end of the last block. | |
296 | */ | |
297 | if (noff > file_sz) { | |
298 | ASSERT(hole); | |
299 | noff = file_sz; | |
cf91b2b6 MA |
300 | } |
301 | ||
302 | if (noff < *off) | |
303 | return (error); | |
304 | *off = noff; | |
305 | return (error); | |
306 | } | |
802e7b5f LD |
307 | |
308 | int | |
309 | zfs_holey(struct inode *ip, int cmd, loff_t *off) | |
310 | { | |
311 | znode_t *zp = ITOZ(ip); | |
0037b49e | 312 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
802e7b5f LD |
313 | int error; |
314 | ||
0037b49e | 315 | ZFS_ENTER(zfsvfs); |
802e7b5f LD |
316 | ZFS_VERIFY_ZP(zp); |
317 | ||
318 | error = zfs_holey_common(ip, cmd, off); | |
319 | ||
0037b49e | 320 | ZFS_EXIT(zfsvfs); |
802e7b5f LD |
321 | return (error); |
322 | } | |
802e7b5f | 323 | #endif /* SEEK_HOLE && SEEK_DATA */ |
cf91b2b6 | 324 | |
c0d35759 | 325 | #if defined(_KERNEL) |
34dc7c2f BB |
326 | /* |
327 | * When a file is memory mapped, we must keep the IO data synchronized | |
328 | * between the DMU cache and the memory mapped pages. What this means: | |
329 | * | |
330 | * On Write: If we find a memory mapped page, we write to *both* | |
331 | * the page and the dmu buffer. | |
34dc7c2f | 332 | */ |
d164b209 | 333 | static void |
c0d35759 BB |
334 | update_pages(struct inode *ip, int64_t start, int len, |
335 | objset_t *os, uint64_t oid) | |
34dc7c2f | 336 | { |
c0d35759 BB |
337 | struct address_space *mp = ip->i_mapping; |
338 | struct page *pp; | |
339 | uint64_t nbytes; | |
d164b209 | 340 | int64_t off; |
c0d35759 | 341 | void *pb; |
34dc7c2f | 342 | |
8b1899d3 BB |
343 | off = start & (PAGE_SIZE-1); |
344 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
345 | nbytes = MIN(PAGE_SIZE - off, len); | |
34dc7c2f | 346 | |
8b1899d3 | 347 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
348 | if (pp) { |
349 | if (mapping_writably_mapped(mp)) | |
350 | flush_dcache_page(pp); | |
34dc7c2f | 351 | |
c0d35759 BB |
352 | pb = kmap(pp); |
353 | (void) dmu_read(os, oid, start+off, nbytes, pb+off, | |
9babb374 | 354 | DMU_READ_PREFETCH); |
c0d35759 BB |
355 | kunmap(pp); |
356 | ||
357 | if (mapping_writably_mapped(mp)) | |
358 | flush_dcache_page(pp); | |
359 | ||
360 | mark_page_accessed(pp); | |
361 | SetPageUptodate(pp); | |
362 | ClearPageError(pp); | |
363 | unlock_page(pp); | |
8b1899d3 | 364 | put_page(pp); |
34dc7c2f | 365 | } |
c0d35759 | 366 | |
d164b209 | 367 | len -= nbytes; |
34dc7c2f | 368 | off = 0; |
34dc7c2f | 369 | } |
34dc7c2f BB |
370 | } |
371 | ||
372 | /* | |
373 | * When a file is memory mapped, we must keep the IO data synchronized | |
374 | * between the DMU cache and the memory mapped pages. What this means: | |
375 | * | |
376 | * On Read: We "read" preferentially from memory mapped pages, | |
377 | * else we default from the dmu buffer. | |
378 | * | |
379 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
d3cc8b15 | 380 | * the file is memory mapped. |
34dc7c2f BB |
381 | */ |
382 | static int | |
3558fd73 | 383 | mappedread(struct inode *ip, int nbytes, uio_t *uio) |
34dc7c2f | 384 | { |
c0d35759 BB |
385 | struct address_space *mp = ip->i_mapping; |
386 | struct page *pp; | |
3558fd73 | 387 | znode_t *zp = ITOZ(ip); |
34dc7c2f | 388 | int64_t start, off; |
c0d35759 | 389 | uint64_t bytes; |
34dc7c2f BB |
390 | int len = nbytes; |
391 | int error = 0; | |
c0d35759 | 392 | void *pb; |
34dc7c2f BB |
393 | |
394 | start = uio->uio_loffset; | |
8b1899d3 BB |
395 | off = start & (PAGE_SIZE-1); |
396 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
397 | bytes = MIN(PAGE_SIZE - off, len); | |
c0d35759 | 398 | |
8b1899d3 | 399 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
400 | if (pp) { |
401 | ASSERT(PageUptodate(pp)); | |
402 | ||
403 | pb = kmap(pp); | |
404 | error = uiomove(pb + off, bytes, UIO_READ, uio); | |
405 | kunmap(pp); | |
406 | ||
407 | if (mapping_writably_mapped(mp)) | |
408 | flush_dcache_page(pp); | |
409 | ||
410 | mark_page_accessed(pp); | |
411 | unlock_page(pp); | |
8b1899d3 | 412 | put_page(pp); |
34dc7c2f | 413 | } else { |
804e0504 MA |
414 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
415 | uio, bytes); | |
34dc7c2f | 416 | } |
c0d35759 | 417 | |
34dc7c2f BB |
418 | len -= bytes; |
419 | off = 0; | |
420 | if (error) | |
421 | break; | |
422 | } | |
423 | return (error); | |
424 | } | |
c0d35759 | 425 | #endif /* _KERNEL */ |
34dc7c2f | 426 | |
c409e464 | 427 | unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ |
a966c564 | 428 | unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT; |
34dc7c2f BB |
429 | |
430 | /* | |
431 | * Read bytes from specified file into supplied buffer. | |
432 | * | |
3558fd73 | 433 | * IN: ip - inode of file to be read from. |
34dc7c2f BB |
434 | * uio - structure supplying read location, range info, |
435 | * and return buffer. | |
c0d35759 BB |
436 | * ioflag - FSYNC flags; used to provide FRSYNC semantics. |
437 | * O_DIRECT flag; used to bypass page cache. | |
34dc7c2f | 438 | * cr - credentials of caller. |
34dc7c2f BB |
439 | * |
440 | * OUT: uio - updated offset and range, buffer filled. | |
441 | * | |
d3cc8b15 | 442 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
443 | * |
444 | * Side Effects: | |
3558fd73 | 445 | * inode - atime updated if byte count > 0 |
34dc7c2f BB |
446 | */ |
447 | /* ARGSUSED */ | |
e5c39b95 | 448 | int |
3558fd73 | 449 | zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 450 | { |
3558fd73 | 451 | znode_t *zp = ITOZ(ip); |
0037b49e | 452 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 453 | ssize_t n, nbytes; |
149e873a | 454 | int error = 0; |
34dc7c2f | 455 | rl_t *rl; |
3558fd73 | 456 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 457 | xuio_t *xuio = NULL; |
3558fd73 | 458 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f | 459 | |
0037b49e | 460 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 461 | ZFS_VERIFY_ZP(zp); |
34dc7c2f | 462 | |
428870ff | 463 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
0037b49e | 464 | ZFS_EXIT(zfsvfs); |
2e528b49 | 465 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
466 | } |
467 | ||
468 | /* | |
469 | * Validate file offset | |
470 | */ | |
471 | if (uio->uio_loffset < (offset_t)0) { | |
0037b49e | 472 | ZFS_EXIT(zfsvfs); |
2e528b49 | 473 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
474 | } |
475 | ||
476 | /* | |
477 | * Fasttrack empty reads | |
478 | */ | |
479 | if (uio->uio_resid == 0) { | |
0037b49e | 480 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
481 | return (0); |
482 | } | |
483 | ||
34dc7c2f BB |
484 | /* |
485 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
37699482 | 486 | * Only do this for non-snapshots. |
34dc7c2f | 487 | */ |
37699482 CC |
488 | if (zfsvfs->z_log && |
489 | (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)) | |
0037b49e | 490 | zil_commit(zfsvfs->z_log, zp->z_id); |
34dc7c2f BB |
491 | |
492 | /* | |
493 | * Lock the range against changes. | |
494 | */ | |
d88895a0 CC |
495 | rl = zfs_range_lock(&zp->z_range_lock, uio->uio_loffset, uio->uio_resid, |
496 | RL_READER); | |
34dc7c2f BB |
497 | |
498 | /* | |
499 | * If we are reading past end-of-file we can skip | |
500 | * to the end; but we might still need to set atime. | |
501 | */ | |
428870ff | 502 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
503 | error = 0; |
504 | goto out; | |
505 | } | |
506 | ||
428870ff BB |
507 | ASSERT(uio->uio_loffset < zp->z_size); |
508 | n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); | |
509 | ||
3558fd73 | 510 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
511 | if ((uio->uio_extflg == UIO_XUIO) && |
512 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
513 | int nblk; | |
514 | int blksz = zp->z_blksz; | |
515 | uint64_t offset = uio->uio_loffset; | |
516 | ||
517 | xuio = (xuio_t *)uio; | |
518 | if ((ISP2(blksz))) { | |
519 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
520 | blksz)) / blksz; | |
521 | } else { | |
522 | ASSERT(offset + n <= blksz); | |
523 | nblk = 1; | |
524 | } | |
525 | (void) dmu_xuio_init(xuio, nblk); | |
526 | ||
3558fd73 | 527 | if (vn_has_cached_data(ip)) { |
428870ff BB |
528 | /* |
529 | * For simplicity, we always allocate a full buffer | |
530 | * even if we only expect to read a portion of a block. | |
531 | */ | |
532 | while (--nblk >= 0) { | |
533 | (void) dmu_xuio_add(xuio, | |
534 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
535 | blksz), 0, blksz); | |
536 | } | |
537 | } | |
538 | } | |
3558fd73 | 539 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
540 | |
541 | while (n > 0) { | |
542 | nbytes = MIN(n, zfs_read_chunk_size - | |
543 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); | |
544 | ||
804e0504 | 545 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
3558fd73 | 546 | error = mappedread(ip, nbytes, uio); |
804e0504 MA |
547 | } else { |
548 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), | |
549 | uio, nbytes); | |
550 | } | |
c0d35759 | 551 | |
b128c09f BB |
552 | if (error) { |
553 | /* convert checksum errors into IO errors */ | |
554 | if (error == ECKSUM) | |
2e528b49 | 555 | error = SET_ERROR(EIO); |
34dc7c2f | 556 | break; |
b128c09f | 557 | } |
34dc7c2f BB |
558 | |
559 | n -= nbytes; | |
560 | } | |
34dc7c2f BB |
561 | out: |
562 | zfs_range_unlock(rl); | |
563 | ||
0037b49e | 564 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
565 | return (error); |
566 | } | |
567 | ||
34dc7c2f BB |
568 | /* |
569 | * Write the bytes to a file. | |
570 | * | |
3558fd73 | 571 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
572 | * uio - structure supplying write location, range info, |
573 | * and data buffer. | |
574 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 575 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 576 | * cr - credentials of caller. |
34dc7c2f BB |
577 | * |
578 | * OUT: uio - updated offset and range. | |
579 | * | |
580 | * RETURN: 0 if success | |
581 | * error code if failure | |
582 | * | |
583 | * Timestamps: | |
3558fd73 | 584 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 585 | */ |
428870ff | 586 | |
34dc7c2f | 587 | /* ARGSUSED */ |
e5c39b95 | 588 | int |
3558fd73 | 589 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 590 | { |
3558fd73 BB |
591 | znode_t *zp = ITOZ(ip); |
592 | rlim64_t limit = uio->uio_limit; | |
34dc7c2f BB |
593 | ssize_t start_resid = uio->uio_resid; |
594 | ssize_t tx_bytes; | |
595 | uint64_t end_size; | |
596 | dmu_tx_t *tx; | |
0037b49e | 597 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
598 | zilog_t *zilog; |
599 | offset_t woff; | |
600 | ssize_t n, nbytes; | |
601 | rl_t *rl; | |
0037b49e | 602 | int max_blksz = zfsvfs->z_max_blksz; |
3558fd73 | 603 | int error = 0; |
9babb374 | 604 | arc_buf_t *abuf; |
5475aada | 605 | const iovec_t *aiov = NULL; |
428870ff | 606 | xuio_t *xuio = NULL; |
428870ff BB |
607 | int write_eof; |
608 | int count = 0; | |
609 | sa_bulk_attr_t bulk[4]; | |
610 | uint64_t mtime[2], ctime[2]; | |
2c6abf15 | 611 | uint32_t uid; |
5a6765cf | 612 | #ifdef HAVE_UIO_ZEROCOPY |
613 | int i_iov = 0; | |
614 | const iovec_t *iovp = uio->uio_iov; | |
3558fd73 | 615 | ASSERTV(int iovcnt = uio->uio_iovcnt); |
5a6765cf | 616 | #endif |
34dc7c2f | 617 | |
34dc7c2f BB |
618 | /* |
619 | * Fasttrack empty write | |
620 | */ | |
621 | n = start_resid; | |
622 | if (n == 0) | |
623 | return (0); | |
624 | ||
625 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) | |
626 | limit = MAXOFFSET_T; | |
627 | ||
0037b49e | 628 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 629 | ZFS_VERIFY_ZP(zp); |
b128c09f | 630 | |
0037b49e BB |
631 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
632 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
633 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
634 | &zp->z_size, 8); | |
635 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
428870ff BB |
636 | &zp->z_pflags, 8); |
637 | ||
f3c9dca0 MT |
638 | /* |
639 | * Callers might not be able to detect properly that we are read-only, | |
640 | * so check it explicitly here. | |
641 | */ | |
0037b49e BB |
642 | if (zfs_is_readonly(zfsvfs)) { |
643 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
644 | return (SET_ERROR(EROFS)); |
645 | } | |
646 | ||
b128c09f BB |
647 | /* |
648 | * If immutable or not appending then return EPERM | |
649 | */ | |
428870ff BB |
650 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
651 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
652 | (uio->uio_loffset < zp->z_size))) { | |
0037b49e | 653 | ZFS_EXIT(zfsvfs); |
2e528b49 | 654 | return (SET_ERROR(EPERM)); |
b128c09f BB |
655 | } |
656 | ||
0037b49e | 657 | zilog = zfsvfs->z_log; |
34dc7c2f | 658 | |
428870ff BB |
659 | /* |
660 | * Validate file offset | |
661 | */ | |
662 | woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; | |
663 | if (woff < 0) { | |
0037b49e | 664 | ZFS_EXIT(zfsvfs); |
2e528b49 | 665 | return (SET_ERROR(EINVAL)); |
428870ff BB |
666 | } |
667 | ||
34dc7c2f BB |
668 | /* |
669 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
670 | * don't hold up txg. | |
428870ff | 671 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 672 | */ |
9cac042c | 673 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
674 | if ((uio->uio_extflg == UIO_XUIO) && |
675 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
676 | xuio = (xuio_t *)uio; | |
677 | else | |
9cac042c | 678 | #endif |
572e2857 | 679 | uio_prefaultpages(MIN(n, max_blksz), uio); |
34dc7c2f BB |
680 | |
681 | /* | |
682 | * If in append mode, set the io offset pointer to eof. | |
683 | */ | |
684 | if (ioflag & FAPPEND) { | |
685 | /* | |
428870ff BB |
686 | * Obtain an appending range lock to guarantee file append |
687 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f | 688 | */ |
d88895a0 | 689 | rl = zfs_range_lock(&zp->z_range_lock, 0, n, RL_APPEND); |
428870ff | 690 | woff = rl->r_off; |
34dc7c2f | 691 | if (rl->r_len == UINT64_MAX) { |
428870ff BB |
692 | /* |
693 | * We overlocked the file because this write will cause | |
694 | * the file block size to increase. | |
695 | * Note that zp_size cannot change with this lock held. | |
696 | */ | |
697 | woff = zp->z_size; | |
34dc7c2f | 698 | } |
428870ff | 699 | uio->uio_loffset = woff; |
34dc7c2f | 700 | } else { |
34dc7c2f | 701 | /* |
428870ff BB |
702 | * Note that if the file block size will change as a result of |
703 | * this write, then this range lock will lock the entire file | |
704 | * so that we can re-write the block safely. | |
34dc7c2f | 705 | */ |
d88895a0 | 706 | rl = zfs_range_lock(&zp->z_range_lock, woff, n, RL_WRITER); |
34dc7c2f BB |
707 | } |
708 | ||
709 | if (woff >= limit) { | |
710 | zfs_range_unlock(rl); | |
0037b49e | 711 | ZFS_EXIT(zfsvfs); |
2e528b49 | 712 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
713 | } |
714 | ||
715 | if ((woff + n) > limit || woff > (limit - n)) | |
716 | n = limit - woff; | |
717 | ||
428870ff BB |
718 | /* Will this write extend the file length? */ |
719 | write_eof = (woff + n > zp->z_size); | |
720 | ||
721 | end_size = MAX(zp->z_size, woff + n); | |
34dc7c2f BB |
722 | |
723 | /* | |
724 | * Write the file in reasonable size chunks. Each chunk is written | |
725 | * in a separate transaction; this keeps the intent log records small | |
726 | * and allows us to do more fine-grained space accounting. | |
727 | */ | |
728 | while (n > 0) { | |
9babb374 BB |
729 | abuf = NULL; |
730 | woff = uio->uio_loffset; | |
0037b49e BB |
731 | if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || |
732 | zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { | |
9babb374 BB |
733 | if (abuf != NULL) |
734 | dmu_return_arcbuf(abuf); | |
2e528b49 | 735 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
736 | break; |
737 | } | |
738 | ||
428870ff | 739 | if (xuio && abuf == NULL) { |
5a6765cf | 740 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 741 | ASSERT(i_iov < iovcnt); |
5475aada | 742 | ASSERT3U(uio->uio_segflg, !=, UIO_BVEC); |
428870ff BB |
743 | aiov = &iovp[i_iov]; |
744 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
745 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
746 | ASSERT((aiov->iov_base == abuf->b_data) || |
747 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
748 | aiov->iov_len == arc_buf_size(abuf))); | |
749 | i_iov++; | |
5a6765cf | 750 | #endif |
428870ff BB |
751 | } else if (abuf == NULL && n >= max_blksz && |
752 | woff >= zp->z_size && | |
9babb374 BB |
753 | P2PHASE(woff, max_blksz) == 0 && |
754 | zp->z_blksz == max_blksz) { | |
428870ff BB |
755 | /* |
756 | * This write covers a full block. "Borrow" a buffer | |
757 | * from the dmu so that we can fill it before we enter | |
758 | * a transaction. This avoids the possibility of | |
759 | * holding up the transaction if the data copy hangs | |
760 | * up on a pagefault (e.g., from an NFS server mapping). | |
761 | */ | |
9babb374 BB |
762 | size_t cbytes; |
763 | ||
428870ff BB |
764 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
765 | max_blksz); | |
9babb374 BB |
766 | ASSERT(abuf != NULL); |
767 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
768 | if ((error = uiocopy(abuf->b_data, max_blksz, |
769 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
770 | dmu_return_arcbuf(abuf); |
771 | break; | |
772 | } | |
773 | ASSERT(cbytes == max_blksz); | |
774 | } | |
775 | ||
34dc7c2f BB |
776 | /* |
777 | * Start a transaction. | |
778 | */ | |
0037b49e | 779 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 780 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 781 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 782 | zfs_sa_upgrade_txholds(tx, zp); |
384f8a09 | 783 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 784 | if (error) { |
34dc7c2f | 785 | dmu_tx_abort(tx); |
9babb374 BB |
786 | if (abuf != NULL) |
787 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
788 | break; |
789 | } | |
790 | ||
791 | /* | |
792 | * If zfs_range_lock() over-locked we grow the blocksize | |
793 | * and then reduce the lock range. This will only happen | |
794 | * on the first iteration since zfs_range_reduce() will | |
795 | * shrink down r_len to the appropriate size. | |
796 | */ | |
797 | if (rl->r_len == UINT64_MAX) { | |
798 | uint64_t new_blksz; | |
799 | ||
800 | if (zp->z_blksz > max_blksz) { | |
f1512ee6 MA |
801 | /* |
802 | * File's blocksize is already larger than the | |
803 | * "recordsize" property. Only let it grow to | |
804 | * the next power of 2. | |
805 | */ | |
34dc7c2f | 806 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 MA |
807 | new_blksz = MIN(end_size, |
808 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
809 | } else { |
810 | new_blksz = MIN(end_size, max_blksz); | |
811 | } | |
812 | zfs_grow_blocksize(zp, new_blksz, tx); | |
813 | zfs_range_reduce(rl, woff, n); | |
814 | } | |
815 | ||
816 | /* | |
817 | * XXX - should we really limit each write to z_max_blksz? | |
818 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
819 | */ | |
820 | nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); | |
34dc7c2f | 821 | |
9babb374 BB |
822 | if (abuf == NULL) { |
823 | tx_bytes = uio->uio_resid; | |
428870ff BB |
824 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
825 | uio, nbytes, tx); | |
9babb374 BB |
826 | tx_bytes -= uio->uio_resid; |
827 | } else { | |
828 | tx_bytes = nbytes; | |
428870ff BB |
829 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
830 | /* | |
831 | * If this is not a full block write, but we are | |
832 | * extending the file past EOF and this data starts | |
833 | * block-aligned, use assign_arcbuf(). Otherwise, | |
834 | * write via dmu_write(). | |
835 | */ | |
836 | if (tx_bytes < max_blksz && (!write_eof || | |
837 | aiov->iov_base != abuf->b_data)) { | |
838 | ASSERT(xuio); | |
0037b49e | 839 | dmu_write(zfsvfs->z_os, zp->z_id, woff, |
94183a9d | 840 | /* cppcheck-suppress nullPointer */ |
428870ff BB |
841 | aiov->iov_len, aiov->iov_base, tx); |
842 | dmu_return_arcbuf(abuf); | |
843 | xuio_stat_wbuf_copied(); | |
844 | } else { | |
845 | ASSERT(xuio || tx_bytes == max_blksz); | |
440a3eb9 TC |
846 | dmu_assign_arcbuf_by_dbuf( |
847 | sa_get_db(zp->z_sa_hdl), woff, abuf, tx); | |
428870ff | 848 | } |
9babb374 BB |
849 | ASSERT(tx_bytes <= uio->uio_resid); |
850 | uioskip(uio, tx_bytes); | |
851 | } | |
0037b49e BB |
852 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
853 | update_pages(ip, woff, | |
854 | tx_bytes, zfsvfs->z_os, zp->z_id); | |
855 | } | |
34dc7c2f BB |
856 | |
857 | /* | |
858 | * If we made no progress, we're done. If we made even | |
859 | * partial progress, update the znode and ZIL accordingly. | |
860 | */ | |
861 | if (tx_bytes == 0) { | |
0037b49e | 862 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 863 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
864 | dmu_tx_commit(tx); |
865 | ASSERT(error != 0); | |
866 | break; | |
867 | } | |
868 | ||
869 | /* | |
870 | * Clear Set-UID/Set-GID bits on successful write if not | |
4e33ba4c | 871 | * privileged and at least one of the execute bits is set. |
34dc7c2f BB |
872 | * |
873 | * It would be nice to to this after all writes have | |
874 | * been done, but that would still expose the ISUID/ISGID | |
875 | * to another app after the partial write is committed. | |
876 | * | |
572e2857 BB |
877 | * Note: we don't call zfs_fuid_map_id() here because |
878 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
879 | */ |
880 | mutex_enter(&zp->z_acl_lock); | |
2c6abf15 | 881 | uid = KUID_TO_SUID(ip->i_uid); |
428870ff | 882 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 883 | (S_IXUSR >> 6))) != 0 && |
428870ff | 884 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 885 | secpolicy_vnode_setid_retain(cr, |
2c6abf15 | 886 | ((zp->z_mode & S_ISUID) != 0 && uid == 0)) != 0) { |
428870ff BB |
887 | uint64_t newmode; |
888 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
12fa7f34 | 889 | ip->i_mode = newmode = zp->z_mode; |
0037b49e | 890 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), |
428870ff | 891 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
892 | } |
893 | mutex_exit(&zp->z_acl_lock); | |
894 | ||
0df9673f | 895 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
34dc7c2f BB |
896 | |
897 | /* | |
898 | * Update the file size (zp_size) if it has changed; | |
899 | * account for possible concurrent updates. | |
900 | */ | |
428870ff BB |
901 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
902 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 903 | uio->uio_loffset); |
428870ff BB |
904 | ASSERT(error == 0); |
905 | } | |
572e2857 BB |
906 | /* |
907 | * If we are replaying and eof is non zero then force | |
908 | * the file size to the specified eof. Note, there's no | |
909 | * concurrency during replay. | |
910 | */ | |
0037b49e BB |
911 | if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) |
912 | zp->z_size = zfsvfs->z_replay_eof; | |
572e2857 | 913 | |
428870ff BB |
914 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
915 | ||
119a394a ED |
916 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
917 | NULL, NULL); | |
34dc7c2f BB |
918 | dmu_tx_commit(tx); |
919 | ||
920 | if (error != 0) | |
921 | break; | |
922 | ASSERT(tx_bytes == nbytes); | |
923 | n -= nbytes; | |
572e2857 BB |
924 | |
925 | if (!xuio && n > 0) | |
926 | uio_prefaultpages(MIN(n, max_blksz), uio); | |
34dc7c2f BB |
927 | } |
928 | ||
2a53e2da | 929 | zfs_inode_update(zp); |
34dc7c2f BB |
930 | zfs_range_unlock(rl); |
931 | ||
932 | /* | |
933 | * If we're in replay mode, or we made no progress, return error. | |
934 | * Otherwise, it's at least a partial write, so it's successful. | |
935 | */ | |
0037b49e BB |
936 | if (zfsvfs->z_replay || uio->uio_resid == start_resid) { |
937 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
938 | return (error); |
939 | } | |
940 | ||
428870ff | 941 | if (ioflag & (FSYNC | FDSYNC) || |
0037b49e | 942 | zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 943 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 944 | |
0037b49e | 945 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
946 | return (0); |
947 | } | |
948 | ||
7c502b0b | 949 | /* |
950 | * Drop a reference on the passed inode asynchronously. This ensures | |
951 | * that the caller will never drop the last reference on an inode in | |
952 | * the current context. Doing so while holding open a tx could result | |
953 | * in a deadlock if iput_final() re-enters the filesystem code. | |
954 | */ | |
0a50679c BB |
955 | void |
956 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 957 | { |
0a50679c BB |
958 | objset_t *os = ITOZSB(ip)->z_os; |
959 | ||
3558fd73 | 960 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
961 | ASSERT(os != NULL); |
962 | ||
3558fd73 | 963 | if (atomic_read(&ip->i_count) == 1) |
7c502b0b | 964 | VERIFY(taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
48d3eb40 | 965 | (task_func_t *)iput, ip, TQ_SLEEP) != TASKQID_INVALID); |
3558fd73 BB |
966 | else |
967 | iput(ip); | |
968 | } | |
969 | ||
34dc7c2f | 970 | void |
428870ff | 971 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 972 | { |
428870ff | 973 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
974 | |
975 | if (zgd->zgd_db) | |
976 | dmu_buf_rele(zgd->zgd_db, zgd); | |
977 | ||
978 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f | 979 | |
9babb374 BB |
980 | /* |
981 | * Release the vnode asynchronously as we currently have the | |
982 | * txg stopped from syncing. | |
983 | */ | |
0a50679c | 984 | zfs_iput_async(ZTOI(zp)); |
428870ff BB |
985 | |
986 | if (error == 0 && zgd->zgd_bp) | |
1ce23dca | 987 | zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp); |
428870ff | 988 | |
34dc7c2f BB |
989 | kmem_free(zgd, sizeof (zgd_t)); |
990 | } | |
991 | ||
45d1cae3 BB |
992 | #ifdef DEBUG |
993 | static int zil_fault_io = 0; | |
994 | #endif | |
995 | ||
34dc7c2f BB |
996 | /* |
997 | * Get data to generate a TX_WRITE intent log record. | |
998 | */ | |
999 | int | |
1ce23dca | 1000 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, zio_t *zio) |
34dc7c2f | 1001 | { |
0037b49e BB |
1002 | zfsvfs_t *zfsvfs = arg; |
1003 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 1004 | znode_t *zp; |
428870ff BB |
1005 | uint64_t object = lr->lr_foid; |
1006 | uint64_t offset = lr->lr_offset; | |
1007 | uint64_t size = lr->lr_length; | |
34dc7c2f | 1008 | dmu_buf_t *db; |
34dc7c2f | 1009 | zgd_t *zgd; |
34dc7c2f BB |
1010 | int error = 0; |
1011 | ||
1ce23dca PS |
1012 | ASSERT3P(lwb, !=, NULL); |
1013 | ASSERT3P(zio, !=, NULL); | |
1014 | ASSERT3U(size, !=, 0); | |
34dc7c2f BB |
1015 | |
1016 | /* | |
1017 | * Nothing to do if the file has been removed | |
1018 | */ | |
0037b49e | 1019 | if (zfs_zget(zfsvfs, object, &zp) != 0) |
2e528b49 | 1020 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 1021 | if (zp->z_unlinked) { |
9babb374 BB |
1022 | /* |
1023 | * Release the vnode asynchronously as we currently have the | |
1024 | * txg stopped from syncing. | |
1025 | */ | |
0a50679c | 1026 | zfs_iput_async(ZTOI(zp)); |
2e528b49 | 1027 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1028 | } |
1029 | ||
79c76d5b | 1030 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
1ce23dca | 1031 | zgd->zgd_lwb = lwb; |
428870ff BB |
1032 | zgd->zgd_private = zp; |
1033 | ||
34dc7c2f BB |
1034 | /* |
1035 | * Write records come in two flavors: immediate and indirect. | |
1036 | * For small writes it's cheaper to store the data with the | |
1037 | * log record (immediate); for large writes it's cheaper to | |
1038 | * sync the data and get a pointer to it (indirect) so that | |
1039 | * we don't have to write the data twice. | |
1040 | */ | |
1041 | if (buf != NULL) { /* immediate write */ | |
d88895a0 CC |
1042 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, size, |
1043 | RL_READER); | |
34dc7c2f | 1044 | /* test for truncation needs to be done while range locked */ |
428870ff | 1045 | if (offset >= zp->z_size) { |
2e528b49 | 1046 | error = SET_ERROR(ENOENT); |
428870ff BB |
1047 | } else { |
1048 | error = dmu_read(os, object, offset, size, buf, | |
1049 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1050 | } |
428870ff | 1051 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1052 | } else { /* indirect write */ |
34dc7c2f BB |
1053 | /* |
1054 | * Have to lock the whole block to ensure when it's | |
f763c3d1 | 1055 | * written out and its checksum is being calculated |
34dc7c2f BB |
1056 | * that no one can change the data. We need to re-check |
1057 | * blocksize after we get the lock in case it's changed! | |
1058 | */ | |
1059 | for (;;) { | |
428870ff BB |
1060 | uint64_t blkoff; |
1061 | size = zp->z_blksz; | |
1062 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1063 | offset -= blkoff; | |
d88895a0 CC |
1064 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, |
1065 | size, RL_READER); | |
428870ff | 1066 | if (zp->z_blksz == size) |
34dc7c2f | 1067 | break; |
428870ff BB |
1068 | offset += blkoff; |
1069 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f BB |
1070 | } |
1071 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1072 | if (lr->lr_offset >= zp->z_size) |
2e528b49 | 1073 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1074 | #ifdef DEBUG |
1075 | if (zil_fault_io) { | |
2e528b49 | 1076 | error = SET_ERROR(EIO); |
45d1cae3 | 1077 | zil_fault_io = 0; |
45d1cae3 | 1078 | } |
45d1cae3 | 1079 | #endif |
34dc7c2f | 1080 | if (error == 0) |
428870ff BB |
1081 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1082 | DMU_READ_NO_PREFETCH); | |
1083 | ||
1084 | if (error == 0) { | |
02dc43bc | 1085 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 1086 | |
428870ff BB |
1087 | zgd->zgd_db = db; |
1088 | zgd->zgd_bp = bp; | |
1089 | ||
1090 | ASSERT(db->db_offset == offset); | |
1091 | ASSERT(db->db_size == size); | |
1092 | ||
1093 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1094 | zfs_get_done, zgd); | |
61ca48ff | 1095 | ASSERT(error || lr->lr_length <= size); |
428870ff BB |
1096 | |
1097 | /* | |
1098 | * On success, we need to wait for the write I/O | |
1099 | * initiated by dmu_sync() to complete before we can | |
1100 | * release this dbuf. We will finish everything up | |
1101 | * in the zfs_get_done() callback. | |
1102 | */ | |
1103 | if (error == 0) | |
1104 | return (0); | |
1105 | ||
1106 | if (error == EALREADY) { | |
1107 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
1108 | error = 0; | |
1109 | } | |
1110 | } | |
34dc7c2f | 1111 | } |
428870ff BB |
1112 | |
1113 | zfs_get_done(zgd, error); | |
1114 | ||
34dc7c2f BB |
1115 | return (error); |
1116 | } | |
1117 | ||
1118 | /*ARGSUSED*/ | |
3558fd73 BB |
1119 | int |
1120 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1121 | { |
3558fd73 | 1122 | znode_t *zp = ITOZ(ip); |
0037b49e | 1123 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
1124 | int error; |
1125 | ||
0037b49e | 1126 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1127 | ZFS_VERIFY_ZP(zp); |
1128 | ||
1129 | if (flag & V_ACE_MASK) | |
1130 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1131 | else | |
1132 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1133 | ||
0037b49e | 1134 | ZFS_EXIT(zfsvfs); |
45d1cae3 BB |
1135 | return (error); |
1136 | } | |
45d1cae3 | 1137 | |
34dc7c2f BB |
1138 | /* |
1139 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1140 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1141 | * |
3558fd73 | 1142 | * IN: dip - inode of directory to search. |
34dc7c2f | 1143 | * nm - name of entry to lookup. |
34dc7c2f | 1144 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1145 | * cr - credentials of caller. |
34dc7c2f BB |
1146 | * direntflags - directory lookup flags |
1147 | * realpnp - returned pathname. | |
1148 | * | |
3558fd73 | 1149 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1150 | * |
d3cc8b15 | 1151 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1152 | * |
1153 | * Timestamps: | |
1154 | * NA | |
1155 | */ | |
1156 | /* ARGSUSED */ | |
e5c39b95 | 1157 | int |
3558fd73 BB |
1158 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1159 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1160 | { |
3558fd73 | 1161 | znode_t *zdp = ITOZ(dip); |
0037b49e | 1162 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
3558fd73 | 1163 | int error = 0; |
45d1cae3 | 1164 | |
9b7b9cd3 GM |
1165 | /* |
1166 | * Fast path lookup, however we must skip DNLC lookup | |
1167 | * for case folding or normalizing lookups because the | |
1168 | * DNLC code only stores the passed in name. This means | |
1169 | * creating 'a' and removing 'A' on a case insensitive | |
1170 | * file system would work, but DNLC still thinks 'a' | |
1171 | * exists and won't let you create it again on the next | |
1172 | * pass through fast path. | |
1173 | */ | |
45d1cae3 BB |
1174 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { |
1175 | ||
3558fd73 | 1176 | if (!S_ISDIR(dip->i_mode)) { |
2e528b49 | 1177 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1178 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 1179 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1180 | } |
1181 | ||
1182 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1183 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1184 | if (!error) { | |
3558fd73 BB |
1185 | *ipp = dip; |
1186 | igrab(*ipp); | |
45d1cae3 BB |
1187 | return (0); |
1188 | } | |
1189 | return (error); | |
3558fd73 | 1190 | #ifdef HAVE_DNLC |
9b7b9cd3 GM |
1191 | } else if (!zdp->z_zfsvfs->z_norm && |
1192 | (zdp->z_zfsvfs->z_case == ZFS_CASE_SENSITIVE)) { | |
1193 | ||
45d1cae3 BB |
1194 | vnode_t *tvp = dnlc_lookup(dvp, nm); |
1195 | ||
1196 | if (tvp) { | |
1197 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1198 | if (error) { | |
3558fd73 | 1199 | iput(tvp); |
45d1cae3 BB |
1200 | return (error); |
1201 | } | |
1202 | if (tvp == DNLC_NO_VNODE) { | |
3558fd73 | 1203 | iput(tvp); |
2e528b49 | 1204 | return (SET_ERROR(ENOENT)); |
45d1cae3 BB |
1205 | } else { |
1206 | *vpp = tvp; | |
1207 | return (specvp_check(vpp, cr)); | |
1208 | } | |
1209 | } | |
3558fd73 | 1210 | #endif /* HAVE_DNLC */ |
45d1cae3 BB |
1211 | } |
1212 | } | |
1213 | ||
0037b49e | 1214 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1215 | ZFS_VERIFY_ZP(zdp); |
1216 | ||
3558fd73 | 1217 | *ipp = NULL; |
34dc7c2f BB |
1218 | |
1219 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1220 | /* |
1221 | * We don't allow recursive attributes.. | |
1222 | * Maybe someday we will. | |
1223 | */ | |
428870ff | 1224 | if (zdp->z_pflags & ZFS_XATTR) { |
0037b49e | 1225 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1226 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1227 | } |
1228 | ||
3558fd73 | 1229 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
0037b49e | 1230 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1231 | return (error); |
1232 | } | |
1233 | ||
1234 | /* | |
1235 | * Do we have permission to get into attribute directory? | |
1236 | */ | |
1237 | ||
3558fd73 | 1238 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1239 | B_FALSE, cr))) { |
3558fd73 BB |
1240 | iput(*ipp); |
1241 | *ipp = NULL; | |
34dc7c2f BB |
1242 | } |
1243 | ||
0037b49e | 1244 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1245 | return (error); |
1246 | } | |
1247 | ||
3558fd73 | 1248 | if (!S_ISDIR(dip->i_mode)) { |
0037b49e | 1249 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1250 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1251 | } |
1252 | ||
1253 | /* | |
1254 | * Check accessibility of directory. | |
1255 | */ | |
1256 | ||
149e873a | 1257 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
0037b49e | 1258 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1259 | return (error); |
1260 | } | |
1261 | ||
0037b49e | 1262 | if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1263 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1264 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1265 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1266 | } |
1267 | ||
3558fd73 BB |
1268 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1269 | if ((error == 0) && (*ipp)) | |
1270 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1271 | |
0037b49e | 1272 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1273 | return (error); |
1274 | } | |
1275 | ||
1276 | /* | |
1277 | * Attempt to create a new entry in a directory. If the entry | |
1278 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1279 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1280 | * |
3558fd73 | 1281 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1282 | * name - name of new file entry. |
1283 | * vap - attributes of new file. | |
1284 | * excl - flag indicating exclusive or non-exclusive mode. | |
1285 | * mode - mode to open file with. | |
1286 | * cr - credentials of caller. | |
1287 | * flag - large file flag [UNUSED]. | |
3558fd73 | 1288 | * vsecp - ACL to be set |
34dc7c2f | 1289 | * |
3558fd73 | 1290 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1291 | * |
d3cc8b15 | 1292 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1293 | * |
1294 | * Timestamps: | |
3558fd73 BB |
1295 | * dip - ctime|mtime updated if new entry created |
1296 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1297 | */ |
1298 | ||
1299 | /* ARGSUSED */ | |
e5c39b95 | 1300 | int |
3558fd73 BB |
1301 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1302 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1303 | { |
3558fd73 | 1304 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1305 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1306 | zilog_t *zilog; |
1307 | objset_t *os; | |
1308 | zfs_dirlock_t *dl; | |
1309 | dmu_tx_t *tx; | |
1310 | int error; | |
b128c09f | 1311 | uid_t uid; |
149e873a | 1312 | gid_t gid; |
428870ff | 1313 | zfs_acl_ids_t acl_ids; |
9babb374 | 1314 | boolean_t fuid_dirtied; |
428870ff | 1315 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 1316 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1317 | |
1318 | /* | |
1319 | * If we have an ephemeral id, ACL, or XVATTR then | |
1320 | * make sure file system is at proper version | |
1321 | */ | |
1322 | ||
149e873a | 1323 | gid = crgetgid(cr); |
3558fd73 | 1324 | uid = crgetuid(cr); |
b128c09f | 1325 | |
0037b49e | 1326 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1327 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1328 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1329 | |
32dec7bd | 1330 | if (name == NULL) |
1331 | return (SET_ERROR(EINVAL)); | |
1332 | ||
0037b49e | 1333 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1334 | ZFS_VERIFY_ZP(dzp); |
0037b49e BB |
1335 | os = zfsvfs->z_os; |
1336 | zilog = zfsvfs->z_log; | |
34dc7c2f | 1337 | |
0037b49e | 1338 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1339 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1340 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1341 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1342 | } |
1343 | ||
5484965a | 1344 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1345 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1346 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1347 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1348 | return (error); |
1349 | } | |
1350 | } | |
34dc7c2f | 1351 | |
3558fd73 BB |
1352 | top: |
1353 | *ipp = NULL; | |
34dc7c2f BB |
1354 | if (*name == '\0') { |
1355 | /* | |
1356 | * Null component name refers to the directory itself. | |
1357 | */ | |
3558fd73 | 1358 | igrab(dip); |
34dc7c2f BB |
1359 | zp = dzp; |
1360 | dl = NULL; | |
1361 | error = 0; | |
1362 | } else { | |
3558fd73 | 1363 | /* possible igrab(zp) */ |
34dc7c2f BB |
1364 | int zflg = 0; |
1365 | ||
1366 | if (flag & FIGNORECASE) | |
1367 | zflg |= ZCILOOK; | |
1368 | ||
1369 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1370 | NULL, NULL); | |
1371 | if (error) { | |
572e2857 BB |
1372 | if (have_acl) |
1373 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1374 | if (strcmp(name, "..") == 0) |
2e528b49 | 1375 | error = SET_ERROR(EISDIR); |
0037b49e | 1376 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1377 | return (error); |
1378 | } | |
1379 | } | |
428870ff | 1380 | |
34dc7c2f BB |
1381 | if (zp == NULL) { |
1382 | uint64_t txtype; | |
1383 | ||
1384 | /* | |
1385 | * Create a new file object and update the directory | |
1386 | * to reference it. | |
1387 | */ | |
149e873a | 1388 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1389 | if (have_acl) |
1390 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1391 | goto out; |
1392 | } | |
1393 | ||
1394 | /* | |
1395 | * We only support the creation of regular files in | |
1396 | * extended attribute directories. | |
1397 | */ | |
428870ff | 1398 | |
3558fd73 | 1399 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1400 | if (have_acl) |
1401 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 1402 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1403 | goto out; |
1404 | } | |
1405 | ||
428870ff BB |
1406 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1407 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1408 | goto out; |
428870ff BB |
1409 | have_acl = B_TRUE; |
1410 | ||
0037b49e | 1411 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
45d1cae3 | 1412 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1413 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1414 | goto out; |
1415 | } | |
1416 | ||
34dc7c2f | 1417 | tx = dmu_tx_create(os); |
428870ff BB |
1418 | |
1419 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1420 | ZFS_SA_BASE_ATTR_SIZE); | |
1421 | ||
0037b49e | 1422 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1423 | if (fuid_dirtied) |
0037b49e | 1424 | zfs_fuid_txhold(zfsvfs, tx); |
34dc7c2f | 1425 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1426 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
0037b49e | 1427 | if (!zfsvfs->z_use_sa && |
428870ff | 1428 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1429 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1430 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1431 | } |
e8b96c60 | 1432 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1433 | if (error) { |
1434 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1435 | if (error == ERESTART) { |
e8b96c60 | 1436 | waited = B_TRUE; |
34dc7c2f BB |
1437 | dmu_tx_wait(tx); |
1438 | dmu_tx_abort(tx); | |
1439 | goto top; | |
1440 | } | |
428870ff | 1441 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1442 | dmu_tx_abort(tx); |
0037b49e | 1443 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1444 | return (error); |
1445 | } | |
428870ff | 1446 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 BB |
1447 | |
1448 | if (fuid_dirtied) | |
0037b49e | 1449 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 1450 | |
34dc7c2f BB |
1451 | (void) zfs_link_create(dl, zp, tx, ZNEW); |
1452 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); | |
1453 | if (flag & FIGNORECASE) | |
1454 | txtype |= TX_CI; | |
1455 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1456 | vsecp, acl_ids.z_fuidp, vap); |
1457 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1458 | dmu_tx_commit(tx); |
1459 | } else { | |
1460 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1461 | ||
572e2857 BB |
1462 | if (have_acl) |
1463 | zfs_acl_ids_free(&acl_ids); | |
1464 | have_acl = B_FALSE; | |
1465 | ||
34dc7c2f BB |
1466 | /* |
1467 | * A directory entry already exists for this name. | |
1468 | */ | |
1469 | /* | |
1470 | * Can't truncate an existing file if in exclusive mode. | |
1471 | */ | |
3558fd73 | 1472 | if (excl) { |
2e528b49 | 1473 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1474 | goto out; |
1475 | } | |
1476 | /* | |
1477 | * Can't open a directory for writing. | |
1478 | */ | |
3558fd73 | 1479 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1480 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1481 | goto out; |
1482 | } | |
1483 | /* | |
1484 | * Verify requested access to file. | |
1485 | */ | |
1486 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1487 | goto out; | |
1488 | } | |
1489 | ||
1490 | mutex_enter(&dzp->z_lock); | |
1491 | dzp->z_seq++; | |
1492 | mutex_exit(&dzp->z_lock); | |
1493 | ||
1494 | /* | |
1495 | * Truncate regular files if requested. | |
1496 | */ | |
3558fd73 BB |
1497 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1498 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 1499 | /* we can't hold any locks when calling zfs_freesp() */ |
609603a5 B |
1500 | if (dl) { |
1501 | zfs_dirent_unlock(dl); | |
1502 | dl = NULL; | |
1503 | } | |
34dc7c2f | 1504 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1505 | } |
1506 | } | |
1507 | out: | |
1508 | ||
1509 | if (dl) | |
1510 | zfs_dirent_unlock(dl); | |
1511 | ||
1512 | if (error) { | |
1513 | if (zp) | |
3558fd73 | 1514 | iput(ZTOI(zp)); |
34dc7c2f | 1515 | } else { |
960e08fe BB |
1516 | zfs_inode_update(dzp); |
1517 | zfs_inode_update(zp); | |
3558fd73 | 1518 | *ipp = ZTOI(zp); |
34dc7c2f | 1519 | } |
34dc7c2f | 1520 | |
0037b49e | 1521 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1522 | zil_commit(zilog, 0); |
428870ff | 1523 | |
0037b49e | 1524 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1525 | return (error); |
1526 | } | |
1527 | ||
ace1eae8 CC |
1528 | /* ARGSUSED */ |
1529 | int | |
1530 | zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, | |
1531 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
1532 | { | |
1533 | znode_t *zp = NULL, *dzp = ITOZ(dip); | |
0037b49e | 1534 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
ace1eae8 CC |
1535 | objset_t *os; |
1536 | dmu_tx_t *tx; | |
1537 | int error; | |
1538 | uid_t uid; | |
1539 | gid_t gid; | |
1540 | zfs_acl_ids_t acl_ids; | |
1541 | boolean_t fuid_dirtied; | |
1542 | boolean_t have_acl = B_FALSE; | |
1543 | boolean_t waited = B_FALSE; | |
1544 | ||
1545 | /* | |
1546 | * If we have an ephemeral id, ACL, or XVATTR then | |
1547 | * make sure file system is at proper version | |
1548 | */ | |
1549 | ||
1550 | gid = crgetgid(cr); | |
1551 | uid = crgetuid(cr); | |
1552 | ||
0037b49e | 1553 | if (zfsvfs->z_use_fuids == B_FALSE && |
ace1eae8 CC |
1554 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
1555 | return (SET_ERROR(EINVAL)); | |
1556 | ||
0037b49e | 1557 | ZFS_ENTER(zfsvfs); |
ace1eae8 | 1558 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1559 | os = zfsvfs->z_os; |
ace1eae8 CC |
1560 | |
1561 | if (vap->va_mask & ATTR_XVATTR) { | |
1562 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
1563 | crgetuid(cr), cr, vap->va_mode)) != 0) { | |
0037b49e | 1564 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1565 | return (error); |
1566 | } | |
1567 | } | |
1568 | ||
1569 | top: | |
1570 | *ipp = NULL; | |
1571 | ||
1572 | /* | |
1573 | * Create a new file object and update the directory | |
1574 | * to reference it. | |
1575 | */ | |
1576 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { | |
1577 | if (have_acl) | |
1578 | zfs_acl_ids_free(&acl_ids); | |
1579 | goto out; | |
1580 | } | |
1581 | ||
1582 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, | |
1583 | cr, vsecp, &acl_ids)) != 0) | |
1584 | goto out; | |
1585 | have_acl = B_TRUE; | |
1586 | ||
0037b49e | 1587 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
ace1eae8 CC |
1588 | zfs_acl_ids_free(&acl_ids); |
1589 | error = SET_ERROR(EDQUOT); | |
1590 | goto out; | |
1591 | } | |
1592 | ||
1593 | tx = dmu_tx_create(os); | |
1594 | ||
1595 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1596 | ZFS_SA_BASE_ATTR_SIZE); | |
0037b49e | 1597 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 1598 | |
0037b49e | 1599 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
ace1eae8 | 1600 | if (fuid_dirtied) |
0037b49e BB |
1601 | zfs_fuid_txhold(zfsvfs, tx); |
1602 | if (!zfsvfs->z_use_sa && | |
ace1eae8 CC |
1603 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
1604 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
1605 | 0, acl_ids.z_aclp->z_acl_bytes); | |
1606 | } | |
1607 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); | |
1608 | if (error) { | |
1609 | if (error == ERESTART) { | |
1610 | waited = B_TRUE; | |
1611 | dmu_tx_wait(tx); | |
1612 | dmu_tx_abort(tx); | |
1613 | goto top; | |
1614 | } | |
1615 | zfs_acl_ids_free(&acl_ids); | |
1616 | dmu_tx_abort(tx); | |
0037b49e | 1617 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1618 | return (error); |
1619 | } | |
1620 | zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); | |
1621 | ||
1622 | if (fuid_dirtied) | |
0037b49e | 1623 | zfs_fuid_sync(zfsvfs, tx); |
ace1eae8 CC |
1624 | |
1625 | /* Add to unlinked set */ | |
1626 | zp->z_unlinked = 1; | |
1627 | zfs_unlinked_add(zp, tx); | |
1628 | zfs_acl_ids_free(&acl_ids); | |
1629 | dmu_tx_commit(tx); | |
1630 | out: | |
1631 | ||
1632 | if (error) { | |
1633 | if (zp) | |
1634 | iput(ZTOI(zp)); | |
1635 | } else { | |
1636 | zfs_inode_update(dzp); | |
1637 | zfs_inode_update(zp); | |
1638 | *ipp = ZTOI(zp); | |
1639 | } | |
1640 | ||
0037b49e | 1641 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1642 | return (error); |
1643 | } | |
1644 | ||
34dc7c2f BB |
1645 | /* |
1646 | * Remove an entry from a directory. | |
1647 | * | |
3558fd73 | 1648 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1649 | * name - name of entry to remove. |
1650 | * cr - credentials of caller. | |
34dc7c2f BB |
1651 | * |
1652 | * RETURN: 0 if success | |
1653 | * error code if failure | |
1654 | * | |
1655 | * Timestamps: | |
3558fd73 BB |
1656 | * dip - ctime|mtime |
1657 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1658 | */ |
428870ff BB |
1659 | |
1660 | uint64_t null_xattr = 0; | |
1661 | ||
34dc7c2f | 1662 | /*ARGSUSED*/ |
e5c39b95 | 1663 | int |
da5e151f | 1664 | zfs_remove(struct inode *dip, char *name, cred_t *cr, int flags) |
34dc7c2f | 1665 | { |
3558fd73 | 1666 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1667 | znode_t *xzp; |
3558fd73 | 1668 | struct inode *ip; |
0037b49e | 1669 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 1670 | zilog_t *zilog; |
a966c564 | 1671 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 1672 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 1673 | uint64_t obj = 0; |
dfbc8630 | 1674 | uint64_t links; |
34dc7c2f BB |
1675 | zfs_dirlock_t *dl; |
1676 | dmu_tx_t *tx; | |
a966c564 K |
1677 | boolean_t may_delete_now, delete_now = FALSE; |
1678 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
1679 | uint64_t txtype; |
1680 | pathname_t *realnmp = NULL; | |
1681 | pathname_t realnm; | |
1682 | int error; | |
1683 | int zflg = ZEXISTS; | |
e8b96c60 | 1684 | boolean_t waited = B_FALSE; |
34dc7c2f | 1685 | |
32dec7bd | 1686 | if (name == NULL) |
1687 | return (SET_ERROR(EINVAL)); | |
1688 | ||
0037b49e | 1689 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1690 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1691 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1692 | |
1693 | if (flags & FIGNORECASE) { | |
1694 | zflg |= ZCILOOK; | |
1695 | pn_alloc(&realnm); | |
1696 | realnmp = &realnm; | |
1697 | } | |
1698 | ||
1699 | top: | |
572e2857 BB |
1700 | xattr_obj = 0; |
1701 | xzp = NULL; | |
34dc7c2f BB |
1702 | /* |
1703 | * Attempt to lock directory; fail if entry doesn't exist. | |
1704 | */ | |
149e873a BB |
1705 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1706 | NULL, realnmp))) { | |
34dc7c2f BB |
1707 | if (realnmp) |
1708 | pn_free(realnmp); | |
0037b49e | 1709 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1710 | return (error); |
1711 | } | |
1712 | ||
3558fd73 | 1713 | ip = ZTOI(zp); |
34dc7c2f | 1714 | |
149e873a | 1715 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1716 | goto out; |
1717 | } | |
1718 | ||
1719 | /* | |
1720 | * Need to use rmdir for removing directories. | |
1721 | */ | |
3558fd73 | 1722 | if (S_ISDIR(ip->i_mode)) { |
2e528b49 | 1723 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1724 | goto out; |
1725 | } | |
1726 | ||
3558fd73 | 1727 | #ifdef HAVE_DNLC |
34dc7c2f BB |
1728 | if (realnmp) |
1729 | dnlc_remove(dvp, realnmp->pn_buf); | |
1730 | else | |
1731 | dnlc_remove(dvp, name); | |
3558fd73 | 1732 | #endif /* HAVE_DNLC */ |
34dc7c2f | 1733 | |
19d55079 MA |
1734 | mutex_enter(&zp->z_lock); |
1735 | may_delete_now = atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped); | |
1736 | mutex_exit(&zp->z_lock); | |
1737 | ||
34dc7c2f | 1738 | /* |
a966c564 K |
1739 | * We may delete the znode now, or we may put it in the unlinked set; |
1740 | * it depends on whether we're the last link, and on whether there are | |
1741 | * other holds on the inode. So we dmu_tx_hold() the right things to | |
1742 | * allow for either case. | |
34dc7c2f | 1743 | */ |
572e2857 | 1744 | obj = zp->z_id; |
0037b49e | 1745 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 1746 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1747 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1748 | zfs_sa_upgrade_txholds(tx, zp); | |
1749 | zfs_sa_upgrade_txholds(tx, dzp); | |
a966c564 K |
1750 | if (may_delete_now) { |
1751 | toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; | |
1752 | /* if the file is too big, only hold_free a token amount */ | |
1753 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1754 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1755 | } | |
34dc7c2f BB |
1756 | |
1757 | /* are there any extended attributes? */ | |
0037b49e | 1758 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1759 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1760 | if (error == 0 && xattr_obj) { |
0037b49e | 1761 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
c99c9001 | 1762 | ASSERT0(error); |
428870ff BB |
1763 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1764 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1765 | } |
1766 | ||
a966c564 K |
1767 | mutex_enter(&zp->z_lock); |
1768 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
1769 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
1770 | mutex_exit(&zp->z_lock); | |
1771 | ||
34dc7c2f | 1772 | /* charge as an update -- would be nice not to charge at all */ |
0037b49e | 1773 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1774 | |
19d55079 | 1775 | /* |
1a04bab3 | 1776 | * Mark this transaction as typically resulting in a net free of space |
19d55079 | 1777 | */ |
1a04bab3 | 1778 | dmu_tx_mark_netfree(tx); |
19d55079 | 1779 | |
e8b96c60 | 1780 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1781 | if (error) { |
1782 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1783 | if (error == ERESTART) { |
e8b96c60 | 1784 | waited = B_TRUE; |
34dc7c2f BB |
1785 | dmu_tx_wait(tx); |
1786 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
1787 | iput(ip); |
1788 | if (xzp) | |
1789 | iput(ZTOI(xzp)); | |
34dc7c2f BB |
1790 | goto top; |
1791 | } | |
1792 | if (realnmp) | |
1793 | pn_free(realnmp); | |
1794 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
1795 | iput(ip); |
1796 | if (xzp) | |
1797 | iput(ZTOI(xzp)); | |
0037b49e | 1798 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1799 | return (error); |
1800 | } | |
1801 | ||
1802 | /* | |
1803 | * Remove the directory entry. | |
1804 | */ | |
1805 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1806 | ||
1807 | if (error) { | |
1808 | dmu_tx_commit(tx); | |
1809 | goto out; | |
1810 | } | |
1811 | ||
1812 | if (unlinked) { | |
572e2857 BB |
1813 | /* |
1814 | * Hold z_lock so that we can make sure that the ACL obj | |
1815 | * hasn't changed. Could have been deleted due to | |
1816 | * zfs_sa_upgrade(). | |
1817 | */ | |
1818 | mutex_enter(&zp->z_lock); | |
0037b49e | 1819 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1820 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
a966c564 K |
1821 | delete_now = may_delete_now && !toobig && |
1822 | atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped) && | |
1823 | xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == | |
1824 | acl_obj; | |
1825 | } | |
1826 | ||
1827 | if (delete_now) { | |
1828 | if (xattr_obj_unlinked) { | |
dfbc8630 | 1829 | ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); |
a966c564 K |
1830 | mutex_enter(&xzp->z_lock); |
1831 | xzp->z_unlinked = 1; | |
dfbc8630 CD |
1832 | clear_nlink(ZTOI(xzp)); |
1833 | links = 0; | |
0037b49e | 1834 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 1835 | &links, sizeof (links), tx); |
a966c564 K |
1836 | ASSERT3U(error, ==, 0); |
1837 | mutex_exit(&xzp->z_lock); | |
1838 | zfs_unlinked_add(xzp, tx); | |
1839 | ||
1840 | if (zp->z_is_sa) | |
1841 | error = sa_remove(zp->z_sa_hdl, | |
0037b49e | 1842 | SA_ZPL_XATTR(zfsvfs), tx); |
a966c564 K |
1843 | else |
1844 | error = sa_update(zp->z_sa_hdl, | |
0037b49e | 1845 | SA_ZPL_XATTR(zfsvfs), &null_xattr, |
a966c564 K |
1846 | sizeof (uint64_t), tx); |
1847 | ASSERT0(error); | |
1848 | } | |
1849 | /* | |
1850 | * Add to the unlinked set because a new reference could be | |
1851 | * taken concurrently resulting in a deferred destruction. | |
1852 | */ | |
1853 | zfs_unlinked_add(zp, tx); | |
1854 | mutex_exit(&zp->z_lock); | |
a966c564 | 1855 | } else if (unlinked) { |
572e2857 | 1856 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1857 | zfs_unlinked_add(zp, tx); |
1858 | } | |
1859 | ||
1860 | txtype = TX_REMOVE; | |
1861 | if (flags & FIGNORECASE) | |
1862 | txtype |= TX_CI; | |
572e2857 | 1863 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1864 | |
1865 | dmu_tx_commit(tx); | |
1866 | out: | |
1867 | if (realnmp) | |
1868 | pn_free(realnmp); | |
1869 | ||
1870 | zfs_dirent_unlock(dl); | |
960e08fe | 1871 | zfs_inode_update(dzp); |
ea7e86d8 | 1872 | zfs_inode_update(zp); |
34dc7c2f | 1873 | |
ea7e86d8 BB |
1874 | if (delete_now) |
1875 | iput(ip); | |
1876 | else | |
a966c564 | 1877 | zfs_iput_async(ip); |
a966c564 K |
1878 | |
1879 | if (xzp) { | |
1880 | zfs_inode_update(xzp); | |
1881 | zfs_iput_async(ZTOI(xzp)); | |
1882 | } | |
428870ff | 1883 | |
0037b49e | 1884 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1885 | zil_commit(zilog, 0); |
34dc7c2f | 1886 | |
0037b49e | 1887 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1888 | return (error); |
1889 | } | |
1890 | ||
1891 | /* | |
3558fd73 | 1892 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1893 | * provided. Return a pointer to the inserted directory. |
1894 | * | |
3558fd73 | 1895 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1896 | * dirname - name of new directory. |
1897 | * vap - attributes of new directory. | |
1898 | * cr - credentials of caller. | |
34dc7c2f BB |
1899 | * vsecp - ACL to be set |
1900 | * | |
3558fd73 | 1901 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1902 | * |
1903 | * RETURN: 0 if success | |
1904 | * error code if failure | |
1905 | * | |
1906 | * Timestamps: | |
3558fd73 BB |
1907 | * dip - ctime|mtime updated |
1908 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1909 | */ |
1910 | /*ARGSUSED*/ | |
e5c39b95 | 1911 | int |
3558fd73 BB |
1912 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1913 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1914 | { |
3558fd73 | 1915 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1916 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1917 | zilog_t *zilog; |
1918 | zfs_dirlock_t *dl; | |
1919 | uint64_t txtype; | |
1920 | dmu_tx_t *tx; | |
1921 | int error; | |
34dc7c2f | 1922 | int zf = ZNEW; |
b128c09f BB |
1923 | uid_t uid; |
1924 | gid_t gid = crgetgid(cr); | |
428870ff | 1925 | zfs_acl_ids_t acl_ids; |
9babb374 | 1926 | boolean_t fuid_dirtied; |
e8b96c60 | 1927 | boolean_t waited = B_FALSE; |
34dc7c2f | 1928 | |
3558fd73 | 1929 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1930 | |
1931 | /* | |
1932 | * If we have an ephemeral id, ACL, or XVATTR then | |
1933 | * make sure file system is at proper version | |
1934 | */ | |
1935 | ||
3558fd73 | 1936 | uid = crgetuid(cr); |
0037b49e | 1937 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1938 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1939 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1940 | |
32dec7bd | 1941 | if (dirname == NULL) |
1942 | return (SET_ERROR(EINVAL)); | |
1943 | ||
0037b49e | 1944 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1945 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1946 | zilog = zfsvfs->z_log; |
34dc7c2f | 1947 | |
428870ff | 1948 | if (dzp->z_pflags & ZFS_XATTR) { |
0037b49e | 1949 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1950 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1951 | } |
1952 | ||
0037b49e | 1953 | if (zfsvfs->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1954 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1955 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1956 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1957 | } |
1958 | if (flags & FIGNORECASE) | |
1959 | zf |= ZCILOOK; | |
1960 | ||
5484965a | 1961 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1962 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1963 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1964 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1965 | return (error); |
1966 | } | |
428870ff | 1967 | } |
34dc7c2f | 1968 | |
428870ff BB |
1969 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1970 | vsecp, &acl_ids)) != 0) { | |
0037b49e | 1971 | ZFS_EXIT(zfsvfs); |
428870ff BB |
1972 | return (error); |
1973 | } | |
34dc7c2f BB |
1974 | /* |
1975 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1976 | * |
1977 | * Existence is checked first to make sure we don't return | |
1978 | * EACCES instead of EEXIST which can cause some applications | |
1979 | * to fail. | |
34dc7c2f BB |
1980 | */ |
1981 | top: | |
3558fd73 | 1982 | *ipp = NULL; |
34dc7c2f | 1983 | |
149e873a BB |
1984 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1985 | NULL, NULL))) { | |
428870ff | 1986 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 1987 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1988 | return (error); |
1989 | } | |
1990 | ||
149e873a | 1991 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 1992 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1993 | zfs_dirent_unlock(dl); |
0037b49e | 1994 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1995 | return (error); |
1996 | } | |
1997 | ||
0037b49e | 1998 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
45d1cae3 | 1999 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 2000 | zfs_dirent_unlock(dl); |
0037b49e | 2001 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2002 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
2003 | } |
2004 | ||
34dc7c2f BB |
2005 | /* |
2006 | * Add a new entry to the directory. | |
2007 | */ | |
0037b49e | 2008 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f BB |
2009 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
2010 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
0037b49e | 2011 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 2012 | if (fuid_dirtied) |
0037b49e BB |
2013 | zfs_fuid_txhold(zfsvfs, tx); |
2014 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
2015 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
2016 | acl_ids.z_aclp->z_acl_bytes); | |
2017 | } | |
2018 | ||
2019 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
2020 | ZFS_SA_BASE_ATTR_SIZE); | |
2021 | ||
e8b96c60 | 2022 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
2023 | if (error) { |
2024 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2025 | if (error == ERESTART) { |
e8b96c60 | 2026 | waited = B_TRUE; |
34dc7c2f BB |
2027 | dmu_tx_wait(tx); |
2028 | dmu_tx_abort(tx); | |
2029 | goto top; | |
2030 | } | |
428870ff | 2031 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2032 | dmu_tx_abort(tx); |
0037b49e | 2033 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2034 | return (error); |
2035 | } | |
2036 | ||
2037 | /* | |
2038 | * Create new node. | |
2039 | */ | |
428870ff | 2040 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f | 2041 | |
9babb374 | 2042 | if (fuid_dirtied) |
0037b49e | 2043 | zfs_fuid_sync(zfsvfs, tx); |
428870ff | 2044 | |
34dc7c2f BB |
2045 | /* |
2046 | * Now put new name in parent dir. | |
2047 | */ | |
2048 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
2049 | ||
3558fd73 | 2050 | *ipp = ZTOI(zp); |
34dc7c2f BB |
2051 | |
2052 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
2053 | if (flags & FIGNORECASE) | |
2054 | txtype |= TX_CI; | |
9babb374 BB |
2055 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
2056 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 2057 | |
9babb374 | 2058 | zfs_acl_ids_free(&acl_ids); |
428870ff | 2059 | |
34dc7c2f BB |
2060 | dmu_tx_commit(tx); |
2061 | ||
2062 | zfs_dirent_unlock(dl); | |
2063 | ||
0037b49e | 2064 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2065 | zil_commit(zilog, 0); |
428870ff | 2066 | |
960e08fe BB |
2067 | zfs_inode_update(dzp); |
2068 | zfs_inode_update(zp); | |
0037b49e | 2069 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2070 | return (0); |
2071 | } | |
2072 | ||
2073 | /* | |
2074 | * Remove a directory subdir entry. If the current working | |
2075 | * directory is the same as the subdir to be removed, the | |
2076 | * remove will fail. | |
2077 | * | |
3558fd73 | 2078 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 2079 | * name - name of directory to be removed. |
3558fd73 | 2080 | * cwd - inode of current working directory. |
34dc7c2f | 2081 | * cr - credentials of caller. |
34dc7c2f BB |
2082 | * flags - case flags |
2083 | * | |
d3cc8b15 | 2084 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
2085 | * |
2086 | * Timestamps: | |
3558fd73 | 2087 | * dip - ctime|mtime updated |
34dc7c2f BB |
2088 | */ |
2089 | /*ARGSUSED*/ | |
e5c39b95 | 2090 | int |
3558fd73 BB |
2091 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
2092 | int flags) | |
34dc7c2f | 2093 | { |
3558fd73 | 2094 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 2095 | znode_t *zp; |
3558fd73 | 2096 | struct inode *ip; |
0037b49e | 2097 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
2098 | zilog_t *zilog; |
2099 | zfs_dirlock_t *dl; | |
2100 | dmu_tx_t *tx; | |
2101 | int error; | |
2102 | int zflg = ZEXISTS; | |
e8b96c60 | 2103 | boolean_t waited = B_FALSE; |
34dc7c2f | 2104 | |
32dec7bd | 2105 | if (name == NULL) |
2106 | return (SET_ERROR(EINVAL)); | |
2107 | ||
0037b49e | 2108 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2109 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 2110 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2111 | |
2112 | if (flags & FIGNORECASE) | |
2113 | zflg |= ZCILOOK; | |
2114 | top: | |
2115 | zp = NULL; | |
2116 | ||
2117 | /* | |
2118 | * Attempt to lock directory; fail if entry doesn't exist. | |
2119 | */ | |
149e873a BB |
2120 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
2121 | NULL, NULL))) { | |
0037b49e | 2122 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2123 | return (error); |
2124 | } | |
2125 | ||
3558fd73 | 2126 | ip = ZTOI(zp); |
34dc7c2f | 2127 | |
149e873a | 2128 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
2129 | goto out; |
2130 | } | |
2131 | ||
3558fd73 | 2132 | if (!S_ISDIR(ip->i_mode)) { |
2e528b49 | 2133 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
2134 | goto out; |
2135 | } | |
2136 | ||
3558fd73 | 2137 | if (ip == cwd) { |
2e528b49 | 2138 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
2139 | goto out; |
2140 | } | |
2141 | ||
34dc7c2f | 2142 | /* |
4e33ba4c | 2143 | * Grab a lock on the directory to make sure that no one is |
34dc7c2f BB |
2144 | * trying to add (or lookup) entries while we are removing it. |
2145 | */ | |
2146 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
2147 | ||
2148 | /* | |
2149 | * Grab a lock on the parent pointer to make sure we play well | |
2150 | * with the treewalk and directory rename code. | |
2151 | */ | |
2152 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
2153 | ||
0037b49e | 2154 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 2155 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 2156 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
0037b49e | 2157 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
2158 | zfs_sa_upgrade_txholds(tx, zp); |
2159 | zfs_sa_upgrade_txholds(tx, dzp); | |
db707ad0 | 2160 | dmu_tx_mark_netfree(tx); |
e8b96c60 | 2161 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
2162 | if (error) { |
2163 | rw_exit(&zp->z_parent_lock); | |
2164 | rw_exit(&zp->z_name_lock); | |
2165 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2166 | if (error == ERESTART) { |
e8b96c60 | 2167 | waited = B_TRUE; |
34dc7c2f BB |
2168 | dmu_tx_wait(tx); |
2169 | dmu_tx_abort(tx); | |
ea7e86d8 | 2170 | iput(ip); |
34dc7c2f BB |
2171 | goto top; |
2172 | } | |
2173 | dmu_tx_abort(tx); | |
ea7e86d8 | 2174 | iput(ip); |
0037b49e | 2175 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2176 | return (error); |
2177 | } | |
2178 | ||
2179 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
2180 | ||
2181 | if (error == 0) { | |
2182 | uint64_t txtype = TX_RMDIR; | |
2183 | if (flags & FIGNORECASE) | |
2184 | txtype |= TX_CI; | |
572e2857 | 2185 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
2186 | } |
2187 | ||
2188 | dmu_tx_commit(tx); | |
2189 | ||
2190 | rw_exit(&zp->z_parent_lock); | |
2191 | rw_exit(&zp->z_name_lock); | |
2192 | out: | |
2193 | zfs_dirent_unlock(dl); | |
2194 | ||
59157910 BB |
2195 | zfs_inode_update(dzp); |
2196 | zfs_inode_update(zp); | |
3558fd73 | 2197 | iput(ip); |
34dc7c2f | 2198 | |
0037b49e | 2199 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2200 | zil_commit(zilog, 0); |
428870ff | 2201 | |
0037b49e | 2202 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2203 | return (error); |
2204 | } | |
2205 | ||
2206 | /* | |
2207 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 2208 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 2209 | * |
3558fd73 BB |
2210 | * IN: ip - inode of directory to read. |
2211 | * dirent - buffer for directory entries. | |
34dc7c2f | 2212 | * |
3558fd73 | 2213 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
2214 | * |
2215 | * RETURN: 0 if success | |
2216 | * error code if failure | |
2217 | * | |
2218 | * Timestamps: | |
3558fd73 | 2219 | * ip - atime updated |
34dc7c2f BB |
2220 | * |
2221 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2222 | * This allows us to use the low range for "special" directory entries: | |
2223 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2224 | * we use the offset 2 for the '.zfs' directory. | |
2225 | */ | |
2226 | /* ARGSUSED */ | |
3558fd73 | 2227 | int |
0f37d0c8 | 2228 | zfs_readdir(struct inode *ip, struct dir_context *ctx, cred_t *cr) |
34dc7c2f | 2229 | { |
3558fd73 | 2230 | znode_t *zp = ITOZ(ip); |
0037b49e | 2231 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2232 | objset_t *os; |
34dc7c2f BB |
2233 | zap_cursor_t zc; |
2234 | zap_attribute_t zap; | |
34dc7c2f BB |
2235 | int error; |
2236 | uint8_t prefetch; | |
c12e3a59 | 2237 | uint8_t type; |
3558fd73 BB |
2238 | int done = 0; |
2239 | uint64_t parent; | |
c12e3a59 | 2240 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2241 | |
0037b49e | 2242 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2243 | ZFS_VERIFY_ZP(zp); |
2244 | ||
0037b49e | 2245 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
3558fd73 BB |
2246 | &parent, sizeof (parent))) != 0) |
2247 | goto out; | |
34dc7c2f BB |
2248 | |
2249 | /* | |
2250 | * Quit if directory has been removed (posix) | |
2251 | */ | |
3558fd73 BB |
2252 | if (zp->z_unlinked) |
2253 | goto out; | |
2254 | ||
c12e3a59 | 2255 | error = 0; |
0037b49e | 2256 | os = zfsvfs->z_os; |
c12e3a59 | 2257 | offset = ctx->pos; |
34dc7c2f BB |
2258 | prefetch = zp->z_zn_prefetch; |
2259 | ||
2260 | /* | |
2261 | * Initialize the iterator cursor. | |
2262 | */ | |
c12e3a59 | 2263 | if (offset <= 3) { |
34dc7c2f BB |
2264 | /* |
2265 | * Start iteration from the beginning of the directory. | |
2266 | */ | |
2267 | zap_cursor_init(&zc, os, zp->z_id); | |
2268 | } else { | |
2269 | /* | |
2270 | * The offset is a serialized cursor. | |
2271 | */ | |
c12e3a59 | 2272 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2273 | } |
2274 | ||
34dc7c2f BB |
2275 | /* |
2276 | * Transform to file-system independent format | |
2277 | */ | |
3558fd73 BB |
2278 | while (!done) { |
2279 | uint64_t objnum; | |
34dc7c2f BB |
2280 | /* |
2281 | * Special case `.', `..', and `.zfs'. | |
2282 | */ | |
c12e3a59 | 2283 | if (offset == 0) { |
34dc7c2f BB |
2284 | (void) strcpy(zap.za_name, "."); |
2285 | zap.za_normalization_conflict = 0; | |
2286 | objnum = zp->z_id; | |
c12e3a59 RY |
2287 | type = DT_DIR; |
2288 | } else if (offset == 1) { | |
34dc7c2f BB |
2289 | (void) strcpy(zap.za_name, ".."); |
2290 | zap.za_normalization_conflict = 0; | |
428870ff | 2291 | objnum = parent; |
c12e3a59 RY |
2292 | type = DT_DIR; |
2293 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2294 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2295 | zap.za_normalization_conflict = 0; | |
2296 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 2297 | type = DT_DIR; |
34dc7c2f BB |
2298 | } else { |
2299 | /* | |
2300 | * Grab next entry. | |
2301 | */ | |
3558fd73 BB |
2302 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2303 | if (error == ENOENT) | |
34dc7c2f BB |
2304 | break; |
2305 | else | |
2306 | goto update; | |
2307 | } | |
2308 | ||
0c5dde49 BB |
2309 | /* |
2310 | * Allow multiple entries provided the first entry is | |
2311 | * the object id. Non-zpl consumers may safely make | |
2312 | * use of the additional space. | |
2313 | * | |
2314 | * XXX: This should be a feature flag for compatibility | |
2315 | */ | |
34dc7c2f | 2316 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2317 | zap.za_num_integers == 0) { |
34dc7c2f | 2318 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2319 | "entry, obj = %lld, offset = %lld, " |
2320 | "length = %d, num = %lld\n", | |
34dc7c2f | 2321 | (u_longlong_t)zp->z_id, |
c12e3a59 | 2322 | (u_longlong_t)offset, |
0c5dde49 BB |
2323 | zap.za_integer_length, |
2324 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 2325 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2326 | goto update; |
2327 | } | |
2328 | ||
2329 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 2330 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2331 | } |
0f37d0c8 RY |
2332 | |
2333 | done = !dir_emit(ctx, zap.za_name, strlen(zap.za_name), | |
c12e3a59 | 2334 | objnum, type); |
0f37d0c8 | 2335 | if (done) |
34dc7c2f | 2336 | break; |
34dc7c2f BB |
2337 | |
2338 | /* Prefetch znode */ | |
3558fd73 | 2339 | if (prefetch) { |
fcff0f35 PD |
2340 | dmu_prefetch(os, objnum, 0, 0, 0, |
2341 | ZIO_PRIORITY_SYNC_READ); | |
3558fd73 | 2342 | } |
34dc7c2f | 2343 | |
c12e3a59 RY |
2344 | /* |
2345 | * Move to the next entry, fill in the previous offset. | |
2346 | */ | |
2347 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2348 | zap_cursor_advance(&zc); |
c12e3a59 | 2349 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2350 | } else { |
c12e3a59 | 2351 | offset += 1; |
34dc7c2f | 2352 | } |
c12e3a59 | 2353 | ctx->pos = offset; |
34dc7c2f BB |
2354 | } |
2355 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2356 | ||
34dc7c2f BB |
2357 | update: |
2358 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2359 | if (error == ENOENT) |
2360 | error = 0; | |
3558fd73 | 2361 | out: |
0037b49e | 2362 | ZFS_EXIT(zfsvfs); |
34dc7c2f | 2363 | |
34dc7c2f BB |
2364 | return (error); |
2365 | } | |
2366 | ||
d5446cfc BB |
2367 | ulong_t zfs_fsync_sync_cnt = 4; |
2368 | ||
e5c39b95 | 2369 | int |
3558fd73 | 2370 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2371 | { |
3558fd73 | 2372 | znode_t *zp = ITOZ(ip); |
0037b49e | 2373 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2374 | |
d5446cfc BB |
2375 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2376 | ||
0037b49e BB |
2377 | if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2378 | ZFS_ENTER(zfsvfs); | |
428870ff | 2379 | ZFS_VERIFY_ZP(zp); |
0037b49e BB |
2380 | zil_commit(zfsvfs->z_log, zp->z_id); |
2381 | ZFS_EXIT(zfsvfs); | |
428870ff | 2382 | } |
07012da6 CC |
2383 | tsd_set(zfs_fsyncer_key, NULL); |
2384 | ||
34dc7c2f BB |
2385 | return (0); |
2386 | } | |
2387 | ||
2388 | ||
2389 | /* | |
2390 | * Get the requested file attributes and place them in the provided | |
2391 | * vattr structure. | |
2392 | * | |
3558fd73 | 2393 | * IN: ip - inode of file. |
5484965a BB |
2394 | * vap - va_mask identifies requested attributes. |
2395 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2396 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2397 | * cr - credentials of caller. | |
34dc7c2f | 2398 | * |
5484965a BB |
2399 | * OUT: vap - attribute values. |
2400 | * | |
2401 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2402 | */ |
2403 | /* ARGSUSED */ | |
e5c39b95 | 2404 | int |
5484965a | 2405 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2406 | { |
3558fd73 | 2407 | znode_t *zp = ITOZ(ip); |
0037b49e | 2408 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
2409 | int error = 0; |
2410 | uint64_t links; | |
0df9673f | 2411 | uint64_t atime[2], mtime[2], ctime[2]; |
5484965a BB |
2412 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2413 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2414 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
0df9673f | 2415 | sa_bulk_attr_t bulk[3]; |
428870ff | 2416 | int count = 0; |
34dc7c2f | 2417 | |
0037b49e | 2418 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2419 | ZFS_VERIFY_ZP(zp); |
428870ff | 2420 | |
5484965a | 2421 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2422 | |
0037b49e BB |
2423 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); |
2424 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
2425 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
428870ff BB |
2426 | |
2427 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
0037b49e | 2428 | ZFS_EXIT(zfsvfs); |
428870ff BB |
2429 | return (error); |
2430 | } | |
34dc7c2f | 2431 | |
34dc7c2f BB |
2432 | /* |
2433 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2434 | * Also, if we are the owner don't bother, since owner should | |
2435 | * always be allowed to read basic attributes of file. | |
2436 | */ | |
572e2857 | 2437 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2438 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2439 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2440 | skipaclchk, cr))) { | |
0037b49e | 2441 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2442 | return (error); |
2443 | } | |
2444 | } | |
2445 | ||
2446 | /* | |
2447 | * Return all attributes. It's cheaper to provide the answer | |
2448 | * than to determine whether we were asked the question. | |
2449 | */ | |
2450 | ||
9babb374 | 2451 | mutex_enter(&zp->z_lock); |
5484965a BB |
2452 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2453 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2454 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2455 | vap->va_nodeid = zp->z_id; |
0037b49e | 2456 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) |
dfbc8630 | 2457 | links = ZTOI(zp)->i_nlink + 1; |
34dc7c2f | 2458 | else |
dfbc8630 | 2459 | links = ZTOI(zp)->i_nlink; |
5484965a BB |
2460 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2461 | vap->va_size = i_size_read(ip); | |
2462 | vap->va_rdev = ip->i_rdev; | |
2463 | vap->va_seq = ip->i_generation; | |
2464 | ||
2465 | /* | |
2466 | * Add in any requested optional attributes and the create time. | |
2467 | * Also set the corresponding bits in the returned attribute bitmap. | |
2468 | */ | |
0037b49e | 2469 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { |
5484965a BB |
2470 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { |
2471 | xoap->xoa_archive = | |
2472 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2473 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2474 | } | |
2475 | ||
2476 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2477 | xoap->xoa_readonly = | |
2478 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2479 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2480 | } | |
2481 | ||
2482 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2483 | xoap->xoa_system = | |
2484 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2485 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2486 | } | |
2487 | ||
2488 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2489 | xoap->xoa_hidden = | |
2490 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2491 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2492 | } | |
2493 | ||
2494 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2495 | xoap->xoa_nounlink = | |
2496 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2497 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2498 | } | |
2499 | ||
2500 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2501 | xoap->xoa_immutable = | |
2502 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2503 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2504 | } | |
2505 | ||
2506 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2507 | xoap->xoa_appendonly = | |
2508 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2509 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2510 | } | |
2511 | ||
2512 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2513 | xoap->xoa_nodump = | |
2514 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2515 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2516 | } | |
2517 | ||
2518 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2519 | xoap->xoa_opaque = | |
2520 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2521 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2522 | } | |
2523 | ||
2524 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2525 | xoap->xoa_av_quarantined = | |
2526 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2527 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2528 | } | |
2529 | ||
2530 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2531 | xoap->xoa_av_modified = | |
2532 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2533 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2534 | } | |
2535 | ||
2536 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2537 | S_ISREG(ip->i_mode)) { | |
2538 | zfs_sa_get_scanstamp(zp, xvap); | |
2539 | } | |
34dc7c2f | 2540 | |
5484965a BB |
2541 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2542 | uint64_t times[2]; | |
2543 | ||
0037b49e | 2544 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), |
5484965a BB |
2545 | times, sizeof (times)); |
2546 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2547 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2548 | } | |
2549 | ||
2550 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2551 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2552 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2553 | } | |
2554 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
278f2236 | 2555 | xoap->xoa_generation = ip->i_generation; |
5484965a BB |
2556 | XVA_SET_RTN(xvap, XAT_GEN); |
2557 | } | |
2558 | ||
2559 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2560 | xoap->xoa_offline = | |
2561 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2562 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2563 | } | |
2564 | ||
2565 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2566 | xoap->xoa_sparse = | |
2567 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2568 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2569 | } | |
2570 | } | |
2571 | ||
0df9673f | 2572 | ZFS_TIME_DECODE(&vap->va_atime, atime); |
5484965a BB |
2573 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); |
2574 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2575 | |
2576 | mutex_exit(&zp->z_lock); | |
2577 | ||
5484965a | 2578 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2579 | |
2580 | if (zp->z_blksz == 0) { | |
2581 | /* | |
2582 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2583 | */ | |
0037b49e | 2584 | vap->va_blksize = zfsvfs->z_max_blksz; |
34dc7c2f BB |
2585 | } |
2586 | ||
0037b49e | 2587 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2588 | return (0); |
2589 | } | |
2590 | ||
057e8eee BB |
2591 | /* |
2592 | * Get the basic file attributes and place them in the provided kstat | |
2593 | * structure. The inode is assumed to be the authoritative source | |
2594 | * for most of the attributes. However, the znode currently has the | |
2595 | * authoritative atime, blksize, and block count. | |
2596 | * | |
2597 | * IN: ip - inode of file. | |
2598 | * | |
2599 | * OUT: sp - kstat values. | |
2600 | * | |
2601 | * RETURN: 0 (always succeeds) | |
2602 | */ | |
2603 | /* ARGSUSED */ | |
2604 | int | |
2605 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2606 | { | |
2607 | znode_t *zp = ITOZ(ip); | |
0037b49e | 2608 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
b585bc4a BB |
2609 | uint32_t blksize; |
2610 | u_longlong_t nblocks; | |
057e8eee | 2611 | |
0037b49e | 2612 | ZFS_ENTER(zfsvfs); |
a7b125e9 GB |
2613 | ZFS_VERIFY_ZP(zp); |
2614 | ||
057e8eee BB |
2615 | mutex_enter(&zp->z_lock); |
2616 | ||
2617 | generic_fillattr(ip, sp); | |
057e8eee | 2618 | |
b585bc4a BB |
2619 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2620 | sp->blksize = blksize; | |
2621 | sp->blocks = nblocks; | |
2622 | ||
057e8eee BB |
2623 | if (unlikely(zp->z_blksz == 0)) { |
2624 | /* | |
2625 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2626 | */ | |
0037b49e | 2627 | sp->blksize = zfsvfs->z_max_blksz; |
057e8eee BB |
2628 | } |
2629 | ||
2630 | mutex_exit(&zp->z_lock); | |
2631 | ||
aa9b2708 AV |
2632 | /* |
2633 | * Required to prevent NFS client from detecting different inode | |
2634 | * numbers of snapshot root dentry before and after snapshot mount. | |
2635 | */ | |
0037b49e | 2636 | if (zfsvfs->z_issnap) { |
aa9b2708 AV |
2637 | if (ip->i_sb->s_root->d_inode == ip) |
2638 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
0037b49e | 2639 | dmu_objset_id(zfsvfs->z_os); |
aa9b2708 AV |
2640 | } |
2641 | ||
0037b49e | 2642 | ZFS_EXIT(zfsvfs); |
a7b125e9 | 2643 | |
057e8eee BB |
2644 | return (0); |
2645 | } | |
057e8eee | 2646 | |
34dc7c2f BB |
2647 | /* |
2648 | * Set the file attributes to the values contained in the | |
2649 | * vattr structure. | |
2650 | * | |
3558fd73 | 2651 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2652 | * vap - new attribute values. |
5484965a | 2653 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2654 | * flags - ATTR_UTIME set if non-default time values provided. |
2655 | * - ATTR_NOACLCHECK (CIFS context only). | |
2656 | * cr - credentials of caller. | |
34dc7c2f BB |
2657 | * |
2658 | * RETURN: 0 if success | |
2659 | * error code if failure | |
2660 | * | |
2661 | * Timestamps: | |
3558fd73 | 2662 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2663 | */ |
2664 | /* ARGSUSED */ | |
e5c39b95 | 2665 | int |
5484965a | 2666 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2667 | { |
3558fd73 | 2668 | znode_t *zp = ITOZ(ip); |
0037b49e | 2669 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
2670 | zilog_t *zilog; |
2671 | dmu_tx_t *tx; | |
2672 | vattr_t oldva; | |
f4ea75d4 | 2673 | xvattr_t *tmpxvattr; |
5484965a | 2674 | uint_t mask = vap->va_mask; |
a117a6d6 | 2675 | uint_t saved_mask = 0; |
34dc7c2f BB |
2676 | int trim_mask = 0; |
2677 | uint64_t new_mode; | |
64aefee1 | 2678 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 2679 | uint64_t xattr_obj; |
0df9673f | 2680 | uint64_t mtime[2], ctime[2], atime[2]; |
34dc7c2f BB |
2681 | znode_t *attrzp; |
2682 | int need_policy = FALSE; | |
428870ff | 2683 | int err, err2; |
34dc7c2f | 2684 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2685 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2686 | xoptattr_t *xoap; | |
2687 | zfs_acl_t *aclp; | |
34dc7c2f | 2688 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2689 | boolean_t fuid_dirtied = B_FALSE; |
17c37660 | 2690 | sa_bulk_attr_t *bulk, *xattr_bulk; |
428870ff | 2691 | int count = 0, xattr_count = 0; |
34dc7c2f BB |
2692 | |
2693 | if (mask == 0) | |
2694 | return (0); | |
2695 | ||
0037b49e | 2696 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2697 | ZFS_VERIFY_ZP(zp); |
2698 | ||
0037b49e | 2699 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2700 | |
2701 | /* | |
2702 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2703 | * that file system is at proper version level | |
2704 | */ | |
5484965a | 2705 | |
0037b49e | 2706 | if (zfsvfs->z_use_fuids == B_FALSE && |
5484965a BB |
2707 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2708 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2709 | (mask & ATTR_XVATTR))) { | |
0037b49e | 2710 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2711 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2712 | } |
2713 | ||
3558fd73 | 2714 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
0037b49e | 2715 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2716 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2717 | } |
2718 | ||
3558fd73 | 2719 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
0037b49e | 2720 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2721 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2722 | } |
2723 | ||
5484965a BB |
2724 | /* |
2725 | * If this is an xvattr_t, then get a pointer to the structure of | |
2726 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2727 | */ | |
2728 | xoap = xva_getxoptattr(xvap); | |
2729 | ||
d1d7e268 | 2730 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2731 | xva_init(tmpxvattr); |
5484965a | 2732 | |
d1d7e268 MK |
2733 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); |
2734 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); | |
17c37660 | 2735 | |
5484965a BB |
2736 | /* |
2737 | * Immutable files can only alter immutable bit and atime | |
2738 | */ | |
2739 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2740 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2741 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
ecb2b7dc | 2742 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2743 | goto out3; |
5484965a BB |
2744 | } |
2745 | ||
3558fd73 | 2746 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
ecb2b7dc | 2747 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2748 | goto out3; |
34dc7c2f BB |
2749 | } |
2750 | ||
5484965a BB |
2751 | /* |
2752 | * Verify timestamps doesn't overflow 32 bits. | |
2753 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2754 | * handle times greater than 2039. This check should be removed | |
2755 | * once large timestamps are fully supported. | |
2756 | */ | |
2757 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
2758 | if (((mask & ATTR_ATIME) && |
2759 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2760 | ((mask & ATTR_MTIME) && | |
2761 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
ecb2b7dc | 2762 | err = SET_ERROR(EOVERFLOW); |
f4ea75d4 | 2763 | goto out3; |
5484965a BB |
2764 | } |
2765 | } | |
2766 | ||
34dc7c2f BB |
2767 | top: |
2768 | attrzp = NULL; | |
572e2857 | 2769 | aclp = NULL; |
34dc7c2f | 2770 | |
45d1cae3 | 2771 | /* Can this be moved to before the top label? */ |
0037b49e | 2772 | if (zfs_is_readonly(zfsvfs)) { |
ecb2b7dc | 2773 | err = SET_ERROR(EROFS); |
f4ea75d4 | 2774 | goto out3; |
34dc7c2f BB |
2775 | } |
2776 | ||
2777 | /* | |
2778 | * First validate permissions | |
2779 | */ | |
2780 | ||
3558fd73 | 2781 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2782 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2783 | if (err) |
2784 | goto out3; | |
2785 | ||
34dc7c2f BB |
2786 | /* |
2787 | * XXX - Note, we are not providing any open | |
2788 | * mode flags here (like FNDELAY), so we may | |
2789 | * block if there are locks present... this | |
2790 | * should be addressed in openat(). | |
2791 | */ | |
b128c09f | 2792 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2793 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2794 | if (err) |
2795 | goto out3; | |
428870ff | 2796 | } |
34dc7c2f | 2797 | |
5484965a BB |
2798 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2799 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2800 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2801 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2802 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2803 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2804 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2805 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2806 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2807 | skipaclchk, cr); | |
2808 | } | |
2809 | ||
3558fd73 BB |
2810 | if (mask & (ATTR_UID|ATTR_GID)) { |
2811 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2812 | int take_owner; |
2813 | int take_group; | |
2814 | ||
2815 | /* | |
2816 | * NOTE: even if a new mode is being set, | |
2817 | * we may clear S_ISUID/S_ISGID bits. | |
2818 | */ | |
2819 | ||
3558fd73 | 2820 | if (!(mask & ATTR_MODE)) |
5484965a | 2821 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2822 | |
2823 | /* | |
2824 | * Take ownership or chgrp to group we are a member of | |
2825 | */ | |
2826 | ||
5484965a | 2827 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 2828 | take_group = (mask & ATTR_GID) && |
0037b49e | 2829 | zfs_groupmember(zfsvfs, vap->va_gid, cr); |
34dc7c2f BB |
2830 | |
2831 | /* | |
5484965a | 2832 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
2833 | * take_group must both be set in order to allow taking |
2834 | * ownership. | |
2835 | * | |
2836 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2837 | * | |
2838 | */ | |
2839 | ||
3558fd73 BB |
2840 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
2841 | take_owner && take_group) || | |
2842 | ((idmask == ATTR_UID) && take_owner) || | |
2843 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
2844 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
2845 | skipaclchk, cr) == 0) { | |
2846 | /* | |
2847 | * Remove setuid/setgid for non-privileged users | |
2848 | */ | |
5484965a | 2849 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 2850 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
2851 | } else { |
2852 | need_policy = TRUE; | |
2853 | } | |
2854 | } else { | |
2855 | need_policy = TRUE; | |
2856 | } | |
2857 | } | |
2858 | ||
2859 | mutex_enter(&zp->z_lock); | |
428870ff | 2860 | oldva.va_mode = zp->z_mode; |
572e2857 | 2861 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
2862 | if (mask & ATTR_XVATTR) { |
2863 | /* | |
2864 | * Update xvattr mask to include only those attributes | |
2865 | * that are actually changing. | |
2866 | * | |
2867 | * the bits will be restored prior to actually setting | |
2868 | * the attributes so the caller thinks they were set. | |
2869 | */ | |
2870 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2871 | if (xoap->xoa_appendonly != | |
2872 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
2873 | need_policy = TRUE; | |
2874 | } else { | |
2875 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 2876 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
2877 | } |
2878 | } | |
2879 | ||
2880 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2881 | if (xoap->xoa_nounlink != | |
2882 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
2883 | need_policy = TRUE; | |
2884 | } else { | |
2885 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 2886 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
2887 | } |
2888 | } | |
2889 | ||
2890 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2891 | if (xoap->xoa_immutable != | |
2892 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
2893 | need_policy = TRUE; | |
2894 | } else { | |
2895 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 2896 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
2897 | } |
2898 | } | |
2899 | ||
2900 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2901 | if (xoap->xoa_nodump != | |
2902 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
2903 | need_policy = TRUE; | |
2904 | } else { | |
2905 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 2906 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
2907 | } |
2908 | } | |
2909 | ||
2910 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2911 | if (xoap->xoa_av_modified != | |
2912 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
2913 | need_policy = TRUE; | |
2914 | } else { | |
2915 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 2916 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
2917 | } |
2918 | } | |
2919 | ||
2920 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2921 | if ((!S_ISREG(ip->i_mode) && | |
2922 | xoap->xoa_av_quarantined) || | |
2923 | xoap->xoa_av_quarantined != | |
2924 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
2925 | need_policy = TRUE; | |
2926 | } else { | |
2927 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 2928 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
2929 | } |
2930 | } | |
2931 | ||
2932 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2933 | mutex_exit(&zp->z_lock); | |
ecb2b7dc | 2934 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2935 | goto out3; |
5484965a BB |
2936 | } |
2937 | ||
2938 | if (need_policy == FALSE && | |
2939 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2940 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
2941 | need_policy = TRUE; | |
2942 | } | |
2943 | } | |
34dc7c2f BB |
2944 | |
2945 | mutex_exit(&zp->z_lock); | |
2946 | ||
3558fd73 | 2947 | if (mask & ATTR_MODE) { |
34dc7c2f | 2948 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 2949 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 2950 | &oldva, cr); |
f4ea75d4 BB |
2951 | if (err) |
2952 | goto out3; | |
2953 | ||
3558fd73 | 2954 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
2955 | } else { |
2956 | need_policy = TRUE; | |
2957 | } | |
2958 | } | |
2959 | ||
2960 | if (need_policy) { | |
2961 | /* | |
2962 | * If trim_mask is set then take ownership | |
2963 | * has been granted or write_acl is present and user | |
2964 | * has the ability to modify mode. In that case remove | |
2965 | * UID|GID and or MODE from mask so that | |
2966 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2967 | */ | |
2968 | ||
2969 | if (trim_mask) { | |
5484965a BB |
2970 | saved_mask = vap->va_mask; |
2971 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 2972 | } |
5484965a | 2973 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 2974 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
2975 | if (err) |
2976 | goto out3; | |
34dc7c2f BB |
2977 | |
2978 | if (trim_mask) | |
5484965a | 2979 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
2980 | } |
2981 | ||
2982 | /* | |
2983 | * secpolicy_vnode_setattr, or take ownership may have | |
2984 | * changed va_mask | |
2985 | */ | |
5484965a | 2986 | mask = vap->va_mask; |
34dc7c2f | 2987 | |
3558fd73 | 2988 | if ((mask & (ATTR_UID | ATTR_GID))) { |
0037b49e | 2989 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
572e2857 | 2990 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 2991 | |
572e2857 | 2992 | if (err == 0 && xattr_obj) { |
3558fd73 | 2993 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
2994 | if (err) |
2995 | goto out2; | |
2996 | } | |
3558fd73 | 2997 | if (mask & ATTR_UID) { |
0037b49e | 2998 | new_kuid = zfs_fuid_create(zfsvfs, |
5484965a | 2999 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
64aefee1 | 3000 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
0037b49e | 3001 | zfs_fuid_overquota(zfsvfs, B_FALSE, new_kuid)) { |
572e2857 | 3002 | if (attrzp) |
3558fd73 | 3003 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3004 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3005 | goto out2; |
3006 | } | |
3007 | } | |
3008 | ||
3558fd73 | 3009 | if (mask & ATTR_GID) { |
0037b49e BB |
3010 | new_kgid = zfs_fuid_create(zfsvfs, |
3011 | (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); | |
64aefee1 | 3012 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && |
0037b49e | 3013 | zfs_fuid_overquota(zfsvfs, B_TRUE, new_kgid)) { |
572e2857 | 3014 | if (attrzp) |
3558fd73 | 3015 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3016 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3017 | goto out2; |
3018 | } | |
3019 | } | |
3020 | } | |
0037b49e | 3021 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 3022 | |
3558fd73 | 3023 | if (mask & ATTR_MODE) { |
428870ff | 3024 | uint64_t pmode = zp->z_mode; |
572e2857 | 3025 | uint64_t acl_obj; |
5484965a | 3026 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 3027 | |
572e2857 | 3028 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 3029 | |
572e2857 BB |
3030 | mutex_enter(&zp->z_lock); |
3031 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
3032 | /* |
3033 | * Are we upgrading ACL from old V0 format | |
3034 | * to V1 format? | |
3035 | */ | |
0037b49e | 3036 | if (zfsvfs->z_version >= ZPL_VERSION_FUID && |
572e2857 | 3037 | zfs_znode_acl_version(zp) == |
34dc7c2f | 3038 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 3039 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
3040 | DMU_OBJECT_END); |
3041 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
3042 | 0, aclp->z_acl_bytes); | |
3043 | } else { | |
572e2857 | 3044 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
3045 | aclp->z_acl_bytes); |
3046 | } | |
428870ff | 3047 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
3048 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
3049 | 0, aclp->z_acl_bytes); | |
3050 | } | |
572e2857 | 3051 | mutex_exit(&zp->z_lock); |
428870ff BB |
3052 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3053 | } else { | |
5484965a BB |
3054 | if ((mask & ATTR_XVATTR) && |
3055 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3056 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
3057 | else | |
3058 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3059 | } |
3060 | ||
428870ff BB |
3061 | if (attrzp) { |
3062 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3063 | } |
3064 | ||
0037b49e | 3065 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
428870ff | 3066 | if (fuid_dirtied) |
0037b49e | 3067 | zfs_fuid_txhold(zfsvfs, tx); |
428870ff BB |
3068 | |
3069 | zfs_sa_upgrade_txholds(tx, zp); | |
3070 | ||
384f8a09 MA |
3071 | err = dmu_tx_assign(tx, TXG_WAIT); |
3072 | if (err) | |
9babb374 | 3073 | goto out; |
34dc7c2f | 3074 | |
428870ff | 3075 | count = 0; |
34dc7c2f BB |
3076 | /* |
3077 | * Set each attribute requested. | |
3078 | * We group settings according to the locks they need to acquire. | |
3079 | * | |
3080 | * Note: you cannot set ctime directly, although it will be | |
3081 | * updated as a side-effect of calling this function. | |
3082 | */ | |
3083 | ||
572e2857 | 3084 | |
3558fd73 | 3085 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3086 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
3087 | mutex_enter(&zp->z_lock); |
3088 | ||
0037b49e | 3089 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
3090 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3091 | ||
3092 | if (attrzp) { | |
3558fd73 | 3093 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3094 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
3095 | mutex_enter(&attrzp->z_lock); |
3096 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3097 | SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, |
428870ff BB |
3098 | sizeof (attrzp->z_pflags)); |
3099 | } | |
3100 | ||
3558fd73 | 3101 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 3102 | |
3558fd73 | 3103 | if (mask & ATTR_UID) { |
64aefee1 NB |
3104 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
3105 | new_uid = zfs_uid_read(ZTOI(zp)); | |
0037b49e | 3106 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 3107 | &new_uid, sizeof (new_uid)); |
428870ff BB |
3108 | if (attrzp) { |
3109 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3110 | SA_ZPL_UID(zfsvfs), NULL, &new_uid, |
428870ff | 3111 | sizeof (new_uid)); |
2c6abf15 | 3112 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
3113 | } |
3114 | } | |
3115 | ||
3558fd73 | 3116 | if (mask & ATTR_GID) { |
64aefee1 NB |
3117 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
3118 | new_gid = zfs_gid_read(ZTOI(zp)); | |
0037b49e | 3119 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), |
428870ff | 3120 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
3121 | if (attrzp) { |
3122 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3123 | SA_ZPL_GID(zfsvfs), NULL, &new_gid, |
428870ff | 3124 | sizeof (new_gid)); |
64aefee1 | 3125 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
3126 | } |
3127 | } | |
3558fd73 | 3128 | if (!(mask & ATTR_MODE)) { |
0037b49e | 3129 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), |
428870ff BB |
3130 | NULL, &new_mode, sizeof (new_mode)); |
3131 | new_mode = zp->z_mode; | |
3132 | } | |
3133 | err = zfs_acl_chown_setattr(zp); | |
3134 | ASSERT(err == 0); | |
3135 | if (attrzp) { | |
3136 | err = zfs_acl_chown_setattr(attrzp); | |
3137 | ASSERT(err == 0); | |
3138 | } | |
3139 | } | |
3140 | ||
3558fd73 | 3141 | if (mask & ATTR_MODE) { |
0037b49e | 3142 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 3143 | &new_mode, sizeof (new_mode)); |
12fa7f34 | 3144 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 3145 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 3146 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 3147 | ASSERT0(err); |
572e2857 BB |
3148 | if (zp->z_acl_cached) |
3149 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
3150 | zp->z_acl_cached = aclp; |
3151 | aclp = NULL; | |
34dc7c2f BB |
3152 | } |
3153 | ||
704cd075 CC |
3154 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
3155 | zp->z_atime_dirty = 0; | |
3156 | ZFS_TIME_ENCODE(&ip->i_atime, atime); | |
0037b49e | 3157 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
0df9673f | 3158 | &atime, sizeof (atime)); |
34dc7c2f BB |
3159 | } |
3160 | ||
99834d19 | 3161 | if (mask & (ATTR_MTIME | ATTR_SIZE)) { |
5484965a | 3162 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
87f9371a NB |
3163 | ZTOI(zp)->i_mtime = timespec_trunc(vap->va_mtime, |
3164 | ZTOI(zp)->i_sb->s_time_gran); | |
3165 | ||
0037b49e | 3166 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 3167 | mtime, sizeof (mtime)); |
34dc7c2f BB |
3168 | } |
3169 | ||
99834d19 | 3170 | if (mask & (ATTR_CTIME | ATTR_SIZE)) { |
87f9371a NB |
3171 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); |
3172 | ZTOI(zp)->i_ctime = timespec_trunc(vap->va_ctime, | |
3173 | ZTOI(zp)->i_sb->s_time_gran); | |
0037b49e | 3174 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
87f9371a | 3175 | ctime, sizeof (ctime)); |
428870ff | 3176 | } |
87f9371a NB |
3177 | |
3178 | if (attrzp && mask) { | |
3179 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3180 | SA_ZPL_CTIME(zfsvfs), NULL, &ctime, |
87f9371a NB |
3181 | sizeof (ctime)); |
3182 | } | |
3183 | ||
34dc7c2f BB |
3184 | /* |
3185 | * Do this after setting timestamps to prevent timestamp | |
3186 | * update from toggling bit | |
3187 | */ | |
3188 | ||
5484965a BB |
3189 | if (xoap && (mask & ATTR_XVATTR)) { |
3190 | ||
3191 | /* | |
3192 | * restore trimmed off masks | |
3193 | * so that return masks can be set for caller. | |
3194 | */ | |
3195 | ||
f4ea75d4 | 3196 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
3197 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
3198 | } | |
f4ea75d4 | 3199 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
3200 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
3201 | } | |
f4ea75d4 | 3202 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
3203 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
3204 | } | |
f4ea75d4 | 3205 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
3206 | XVA_SET_REQ(xvap, XAT_NODUMP); |
3207 | } | |
f4ea75d4 | 3208 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
3209 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
3210 | } | |
f4ea75d4 | 3211 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
3212 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
3213 | } | |
3214 | ||
3215 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3216 | ASSERT(S_ISREG(ip->i_mode)); | |
3217 | ||
3218 | zfs_xvattr_set(zp, xvap, tx); | |
3219 | } | |
3220 | ||
9babb374 | 3221 | if (fuid_dirtied) |
0037b49e | 3222 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 3223 | |
34dc7c2f | 3224 | if (mask != 0) |
5484965a | 3225 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3226 | |
34dc7c2f | 3227 | mutex_exit(&zp->z_lock); |
3558fd73 | 3228 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3229 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3230 | |
572e2857 | 3231 | if (attrzp) { |
3558fd73 | 3232 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3233 | mutex_exit(&attrzp->z_acl_lock); |
3234 | mutex_exit(&attrzp->z_lock); | |
3235 | } | |
9babb374 | 3236 | out: |
428870ff BB |
3237 | if (err == 0 && attrzp) { |
3238 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, | |
3239 | xattr_count, tx); | |
3240 | ASSERT(err2 == 0); | |
3241 | } | |
3242 | ||
45d1cae3 | 3243 | if (aclp) |
9babb374 | 3244 | zfs_acl_free(aclp); |
9babb374 BB |
3245 | |
3246 | if (fuidp) { | |
3247 | zfs_fuid_info_free(fuidp); | |
3248 | fuidp = NULL; | |
3249 | } | |
3250 | ||
428870ff | 3251 | if (err) { |
9babb374 | 3252 | dmu_tx_abort(tx); |
ea7e86d8 BB |
3253 | if (attrzp) |
3254 | iput(ZTOI(attrzp)); | |
428870ff BB |
3255 | if (err == ERESTART) |
3256 | goto top; | |
3257 | } else { | |
3258 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3259 | dmu_tx_commit(tx); |
ea7e86d8 BB |
3260 | if (attrzp) |
3261 | iput(ZTOI(attrzp)); | |
037849f8 | 3262 | zfs_inode_update(zp); |
428870ff BB |
3263 | } |
3264 | ||
428870ff | 3265 | out2: |
0037b49e | 3266 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3267 | zil_commit(zilog, 0); |
34dc7c2f | 3268 | |
f4ea75d4 | 3269 | out3: |
d1d7e268 MK |
3270 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * 7); |
3271 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * 7); | |
3272 | kmem_free(tmpxvattr, sizeof (xvattr_t)); | |
0037b49e | 3273 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3274 | return (err); |
3275 | } | |
3276 | ||
3277 | typedef struct zfs_zlock { | |
3278 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3279 | znode_t *zl_znode; /* znode we held */ | |
3280 | struct zfs_zlock *zl_next; /* next in list */ | |
3281 | } zfs_zlock_t; | |
3282 | ||
3283 | /* | |
3284 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3285 | */ | |
3286 | static void | |
3287 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3288 | { | |
3289 | zfs_zlock_t *zl; | |
3290 | ||
3291 | while ((zl = *zlpp) != NULL) { | |
3292 | if (zl->zl_znode != NULL) | |
ea7e86d8 | 3293 | zfs_iput_async(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3294 | rw_exit(zl->zl_rwlock); |
3295 | *zlpp = zl->zl_next; | |
3296 | kmem_free(zl, sizeof (*zl)); | |
3297 | } | |
3298 | } | |
3299 | ||
3300 | /* | |
3301 | * Search back through the directory tree, using the ".." entries. | |
3302 | * Lock each directory in the chain to prevent concurrent renames. | |
3303 | * Fail any attempt to move a directory into one of its own descendants. | |
3304 | * XXX - z_parent_lock can overlap with map or grow locks | |
3305 | */ | |
3306 | static int | |
3307 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3308 | { | |
3309 | zfs_zlock_t *zl; | |
3310 | znode_t *zp = tdzp; | |
3558fd73 | 3311 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3312 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3313 | krwlock_t *rwlp = &szp->z_parent_lock; |
3314 | krw_t rw = RW_WRITER; | |
3315 | ||
3316 | /* | |
3317 | * First pass write-locks szp and compares to zp->z_id. | |
3318 | * Later passes read-lock zp and compare to zp->z_parent. | |
3319 | */ | |
3320 | do { | |
3321 | if (!rw_tryenter(rwlp, rw)) { | |
3322 | /* | |
3323 | * Another thread is renaming in this path. | |
3324 | * Note that if we are a WRITER, we don't have any | |
3325 | * parent_locks held yet. | |
3326 | */ | |
3327 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3328 | /* | |
3329 | * Drop our locks and restart | |
3330 | */ | |
3331 | zfs_rename_unlock(&zl); | |
3332 | *zlpp = NULL; | |
3333 | zp = tdzp; | |
428870ff | 3334 | oidp = zp->z_id; |
34dc7c2f BB |
3335 | rwlp = &szp->z_parent_lock; |
3336 | rw = RW_WRITER; | |
3337 | continue; | |
3338 | } else { | |
3339 | /* | |
3340 | * Wait for other thread to drop its locks | |
3341 | */ | |
3342 | rw_enter(rwlp, rw); | |
3343 | } | |
3344 | } | |
3345 | ||
3346 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3347 | zl->zl_rwlock = rwlp; | |
3348 | zl->zl_znode = NULL; | |
3349 | zl->zl_next = *zlpp; | |
3350 | *zlpp = zl; | |
3351 | ||
428870ff | 3352 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 3353 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3354 | |
428870ff | 3355 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3356 | return (0); |
3357 | ||
3358 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3359 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3360 | if (error) |
3361 | return (error); | |
3362 | zl->zl_znode = zp; | |
3363 | } | |
3558fd73 | 3364 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3365 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3366 | rwlp = &zp->z_parent_lock; |
3367 | rw = RW_READER; | |
3368 | ||
3369 | } while (zp->z_id != sdzp->z_id); | |
3370 | ||
3371 | return (0); | |
3372 | } | |
3373 | ||
3374 | /* | |
3375 | * Move an entry from the provided source directory to the target | |
3376 | * directory. Change the entry name as indicated. | |
3377 | * | |
3558fd73 | 3378 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3379 | * snm - Old entry name. |
3558fd73 | 3380 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3381 | * tnm - New entry name. |
3382 | * cr - credentials of caller. | |
34dc7c2f BB |
3383 | * flags - case flags |
3384 | * | |
d3cc8b15 | 3385 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3386 | * |
3387 | * Timestamps: | |
3558fd73 | 3388 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3389 | */ |
3390 | /*ARGSUSED*/ | |
e5c39b95 | 3391 | int |
3558fd73 BB |
3392 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3393 | cred_t *cr, int flags) | |
34dc7c2f BB |
3394 | { |
3395 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 | 3396 | znode_t *sdzp = ITOZ(sdip); |
0037b49e | 3397 | zfsvfs_t *zfsvfs = ITOZSB(sdip); |
34dc7c2f | 3398 | zilog_t *zilog; |
34dc7c2f BB |
3399 | zfs_dirlock_t *sdl, *tdl; |
3400 | dmu_tx_t *tx; | |
3401 | zfs_zlock_t *zl; | |
3402 | int cmp, serr, terr; | |
3403 | int error = 0; | |
3404 | int zflg = 0; | |
e8b96c60 | 3405 | boolean_t waited = B_FALSE; |
34dc7c2f | 3406 | |
32dec7bd | 3407 | if (snm == NULL || tnm == NULL) |
3408 | return (SET_ERROR(EINVAL)); | |
3409 | ||
0037b49e | 3410 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 3411 | ZFS_VERIFY_ZP(sdzp); |
0037b49e | 3412 | zilog = zfsvfs->z_log; |
34dc7c2f | 3413 | |
812e91a7 MT |
3414 | tdzp = ITOZ(tdip); |
3415 | ZFS_VERIFY_ZP(tdzp); | |
3416 | ||
3417 | /* | |
3418 | * We check i_sb because snapshots and the ctldir must have different | |
3419 | * super blocks. | |
3420 | */ | |
c0ebc844 | 3421 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
0037b49e | 3422 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3423 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3424 | } |
3425 | ||
0037b49e | 3426 | if (zfsvfs->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3427 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 3428 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3429 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3430 | } |
3431 | ||
3432 | if (flags & FIGNORECASE) | |
3433 | zflg |= ZCILOOK; | |
3434 | ||
3435 | top: | |
3436 | szp = NULL; | |
3437 | tzp = NULL; | |
3438 | zl = NULL; | |
3439 | ||
3440 | /* | |
3441 | * This is to prevent the creation of links into attribute space | |
3442 | * by renaming a linked file into/outof an attribute directory. | |
3443 | * See the comment in zfs_link() for why this is considered bad. | |
3444 | */ | |
428870ff | 3445 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 3446 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3447 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3448 | } |
3449 | ||
3450 | /* | |
3451 | * Lock source and target directory entries. To prevent deadlock, | |
3452 | * a lock ordering must be defined. We lock the directory with | |
3453 | * the smallest object id first, or if it's a tie, the one with | |
3454 | * the lexically first name. | |
3455 | */ | |
3456 | if (sdzp->z_id < tdzp->z_id) { | |
3457 | cmp = -1; | |
3458 | } else if (sdzp->z_id > tdzp->z_id) { | |
3459 | cmp = 1; | |
3460 | } else { | |
3461 | /* | |
3462 | * First compare the two name arguments without | |
3463 | * considering any case folding. | |
3464 | */ | |
0037b49e | 3465 | int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3466 | |
3467 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
0037b49e | 3468 | ASSERT(error == 0 || !zfsvfs->z_utf8); |
34dc7c2f BB |
3469 | if (cmp == 0) { |
3470 | /* | |
3471 | * POSIX: "If the old argument and the new argument | |
3472 | * both refer to links to the same existing file, | |
3473 | * the rename() function shall return successfully | |
3474 | * and perform no other action." | |
3475 | */ | |
0037b49e | 3476 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3477 | return (0); |
3478 | } | |
3479 | /* | |
3480 | * If the file system is case-folding, then we may | |
3481 | * have some more checking to do. A case-folding file | |
3482 | * system is either supporting mixed case sensitivity | |
3483 | * access or is completely case-insensitive. Note | |
3484 | * that the file system is always case preserving. | |
3485 | * | |
3486 | * In mixed sensitivity mode case sensitive behavior | |
3487 | * is the default. FIGNORECASE must be used to | |
3488 | * explicitly request case insensitive behavior. | |
3489 | * | |
3490 | * If the source and target names provided differ only | |
3491 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3492 | * we will treat this as a special case in the | |
3493 | * case-insensitive mode: as long as the source name | |
3494 | * is an exact match, we will allow this to proceed as | |
3495 | * a name-change request. | |
3496 | */ | |
0037b49e BB |
3497 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
3498 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3499 | flags & FIGNORECASE)) && |
0037b49e | 3500 | u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3501 | &error) == 0) { |
3502 | /* | |
3503 | * case preserving rename request, require exact | |
3504 | * name matches | |
3505 | */ | |
3506 | zflg |= ZCIEXACT; | |
3507 | zflg &= ~ZCILOOK; | |
3508 | } | |
3509 | } | |
3510 | ||
428870ff BB |
3511 | /* |
3512 | * If the source and destination directories are the same, we should | |
3513 | * grab the z_name_lock of that directory only once. | |
3514 | */ | |
3515 | if (sdzp == tdzp) { | |
3516 | zflg |= ZHAVELOCK; | |
3517 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3518 | } | |
3519 | ||
34dc7c2f BB |
3520 | if (cmp < 0) { |
3521 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3522 | ZEXISTS | zflg, NULL, NULL); | |
3523 | terr = zfs_dirent_lock(&tdl, | |
3524 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3525 | } else { | |
3526 | terr = zfs_dirent_lock(&tdl, | |
3527 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3528 | serr = zfs_dirent_lock(&sdl, | |
3529 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3530 | NULL, NULL); | |
3531 | } | |
3532 | ||
3533 | if (serr) { | |
3534 | /* | |
3535 | * Source entry invalid or not there. | |
3536 | */ | |
3537 | if (!terr) { | |
3538 | zfs_dirent_unlock(tdl); | |
3539 | if (tzp) | |
3558fd73 | 3540 | iput(ZTOI(tzp)); |
34dc7c2f | 3541 | } |
428870ff BB |
3542 | |
3543 | if (sdzp == tdzp) | |
3544 | rw_exit(&sdzp->z_name_lock); | |
3545 | ||
34dc7c2f BB |
3546 | if (strcmp(snm, "..") == 0) |
3547 | serr = EINVAL; | |
0037b49e | 3548 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3549 | return (serr); |
3550 | } | |
3551 | if (terr) { | |
3552 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3553 | iput(ZTOI(szp)); |
428870ff BB |
3554 | |
3555 | if (sdzp == tdzp) | |
3556 | rw_exit(&sdzp->z_name_lock); | |
3557 | ||
34dc7c2f BB |
3558 | if (strcmp(tnm, "..") == 0) |
3559 | terr = EINVAL; | |
0037b49e | 3560 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3561 | return (terr); |
3562 | } | |
3563 | ||
3564 | /* | |
3565 | * Must have write access at the source to remove the old entry | |
3566 | * and write access at the target to create the new entry. | |
3567 | * Note that if target and source are the same, this can be | |
3568 | * done in a single check. | |
3569 | */ | |
3570 | ||
149e873a | 3571 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3572 | goto out; |
3573 | ||
3558fd73 | 3574 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3575 | /* |
3576 | * Check to make sure rename is valid. | |
3577 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3578 | */ | |
149e873a | 3579 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3580 | goto out; |
3581 | } | |
3582 | ||
3583 | /* | |
3584 | * Does target exist? | |
3585 | */ | |
3586 | if (tzp) { | |
3587 | /* | |
3588 | * Source and target must be the same type. | |
3589 | */ | |
3558fd73 BB |
3590 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3591 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
2e528b49 | 3592 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3593 | goto out; |
3594 | } | |
3595 | } else { | |
3558fd73 | 3596 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
2e528b49 | 3597 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3598 | goto out; |
3599 | } | |
3600 | } | |
3601 | /* | |
3602 | * POSIX dictates that when the source and target | |
3603 | * entries refer to the same file object, rename | |
3604 | * must do nothing and exit without error. | |
3605 | */ | |
3606 | if (szp->z_id == tzp->z_id) { | |
3607 | error = 0; | |
3608 | goto out; | |
3609 | } | |
3610 | } | |
3611 | ||
0037b49e | 3612 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
3613 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3614 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3615 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3616 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3617 | if (sdzp != tdzp) { |
3618 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3619 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3620 | } | |
3621 | if (tzp) { | |
3622 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3623 | zfs_sa_upgrade_txholds(tx, tzp); | |
3624 | } | |
3625 | ||
3626 | zfs_sa_upgrade_txholds(tx, szp); | |
0037b49e | 3627 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
e8b96c60 | 3628 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3629 | if (error) { |
3630 | if (zl != NULL) | |
3631 | zfs_rename_unlock(&zl); | |
3632 | zfs_dirent_unlock(sdl); | |
3633 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3634 | |
3635 | if (sdzp == tdzp) | |
3636 | rw_exit(&sdzp->z_name_lock); | |
3637 | ||
fb5f0bc8 | 3638 | if (error == ERESTART) { |
e8b96c60 | 3639 | waited = B_TRUE; |
34dc7c2f BB |
3640 | dmu_tx_wait(tx); |
3641 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3642 | iput(ZTOI(szp)); |
3643 | if (tzp) | |
3644 | iput(ZTOI(tzp)); | |
34dc7c2f BB |
3645 | goto top; |
3646 | } | |
3647 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3648 | iput(ZTOI(szp)); |
3649 | if (tzp) | |
3650 | iput(ZTOI(tzp)); | |
0037b49e | 3651 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3652 | return (error); |
3653 | } | |
3654 | ||
3655 | if (tzp) /* Attempt to remove the existing target */ | |
3656 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3657 | ||
3658 | if (error == 0) { | |
3659 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3660 | if (error == 0) { | |
428870ff | 3661 | szp->z_pflags |= ZFS_AV_MODIFIED; |
34dc7c2f | 3662 | |
0037b49e | 3663 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 3664 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c99c9001 | 3665 | ASSERT0(error); |
34dc7c2f | 3666 | |
428870ff BB |
3667 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3668 | if (error == 0) { | |
3669 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3670 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3671 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3672 | } else { |
3673 | /* | |
3674 | * At this point, we have successfully created | |
3675 | * the target name, but have failed to remove | |
3676 | * the source name. Since the create was done | |
3677 | * with the ZRENAMING flag, there are | |
3678 | * complications; for one, the link count is | |
3679 | * wrong. The easiest way to deal with this | |
3680 | * is to remove the newly created target, and | |
3681 | * return the original error. This must | |
3682 | * succeed; fortunately, it is very unlikely to | |
3683 | * fail, since we just created it. | |
3684 | */ | |
3685 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3686 | ZRENAMING, NULL), ==, 0); | |
3687 | } | |
34dc7c2f BB |
3688 | } |
3689 | } | |
3690 | ||
3691 | dmu_tx_commit(tx); | |
3692 | out: | |
3693 | if (zl != NULL) | |
3694 | zfs_rename_unlock(&zl); | |
3695 | ||
3696 | zfs_dirent_unlock(sdl); | |
3697 | zfs_dirent_unlock(tdl); | |
3698 | ||
960e08fe | 3699 | zfs_inode_update(sdzp); |
428870ff BB |
3700 | if (sdzp == tdzp) |
3701 | rw_exit(&sdzp->z_name_lock); | |
3702 | ||
960e08fe BB |
3703 | if (sdzp != tdzp) |
3704 | zfs_inode_update(tdzp); | |
428870ff | 3705 | |
960e08fe | 3706 | zfs_inode_update(szp); |
3558fd73 | 3707 | iput(ZTOI(szp)); |
960e08fe BB |
3708 | if (tzp) { |
3709 | zfs_inode_update(tzp); | |
3558fd73 | 3710 | iput(ZTOI(tzp)); |
960e08fe | 3711 | } |
34dc7c2f | 3712 | |
0037b49e | 3713 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3714 | zil_commit(zilog, 0); |
428870ff | 3715 | |
0037b49e | 3716 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3717 | return (error); |
3718 | } | |
3719 | ||
3720 | /* | |
3721 | * Insert the indicated symbolic reference entry into the directory. | |
3722 | * | |
3558fd73 | 3723 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
3724 | * link - Name for new symlink entry. |
3725 | * vap - Attributes of new entry. | |
3726 | * target - Target path of new symlink. | |
3558fd73 | 3727 | * |
34dc7c2f | 3728 | * cr - credentials of caller. |
34dc7c2f BB |
3729 | * flags - case flags |
3730 | * | |
d3cc8b15 | 3731 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3732 | * |
3733 | * Timestamps: | |
3558fd73 | 3734 | * dip - ctime|mtime updated |
34dc7c2f BB |
3735 | */ |
3736 | /*ARGSUSED*/ | |
e5c39b95 | 3737 | int |
3558fd73 BB |
3738 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
3739 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 3740 | { |
3558fd73 | 3741 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
3742 | zfs_dirlock_t *dl; |
3743 | dmu_tx_t *tx; | |
0037b49e | 3744 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 3745 | zilog_t *zilog; |
428870ff | 3746 | uint64_t len = strlen(link); |
34dc7c2f BB |
3747 | int error; |
3748 | int zflg = ZNEW; | |
9babb374 BB |
3749 | zfs_acl_ids_t acl_ids; |
3750 | boolean_t fuid_dirtied; | |
428870ff | 3751 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 3752 | boolean_t waited = B_FALSE; |
34dc7c2f | 3753 | |
3558fd73 | 3754 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 3755 | |
32dec7bd | 3756 | if (name == NULL) |
3757 | return (SET_ERROR(EINVAL)); | |
3758 | ||
0037b49e | 3759 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 3760 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 3761 | zilog = zfsvfs->z_log; |
34dc7c2f | 3762 | |
0037b49e | 3763 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 3764 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 3765 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3766 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3767 | } |
3768 | if (flags & FIGNORECASE) | |
3769 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3770 | |
3771 | if (len > MAXPATHLEN) { | |
0037b49e | 3772 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3773 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
3774 | } |
3775 | ||
428870ff BB |
3776 | if ((error = zfs_acl_ids_create(dzp, 0, |
3777 | vap, cr, NULL, &acl_ids)) != 0) { | |
0037b49e | 3778 | ZFS_EXIT(zfsvfs); |
428870ff BB |
3779 | return (error); |
3780 | } | |
3781 | top: | |
3558fd73 BB |
3782 | *ipp = NULL; |
3783 | ||
34dc7c2f BB |
3784 | /* |
3785 | * Attempt to lock directory; fail if entry already exists. | |
3786 | */ | |
3787 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3788 | if (error) { | |
428870ff | 3789 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 3790 | ZFS_EXIT(zfsvfs); |
428870ff BB |
3791 | return (error); |
3792 | } | |
3793 | ||
149e873a | 3794 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
3795 | zfs_acl_ids_free(&acl_ids); |
3796 | zfs_dirent_unlock(dl); | |
0037b49e | 3797 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3798 | return (error); |
3799 | } | |
3800 | ||
0037b49e | 3801 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
9babb374 BB |
3802 | zfs_acl_ids_free(&acl_ids); |
3803 | zfs_dirent_unlock(dl); | |
0037b49e | 3804 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3805 | return (SET_ERROR(EDQUOT)); |
9babb374 | 3806 | } |
0037b49e BB |
3807 | tx = dmu_tx_create(zfsvfs->z_os); |
3808 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
34dc7c2f | 3809 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3810 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3811 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3812 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3813 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
0037b49e | 3814 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
3815 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
3816 | acl_ids.z_aclp->z_acl_bytes); | |
3817 | } | |
9babb374 | 3818 | if (fuid_dirtied) |
0037b49e | 3819 | zfs_fuid_txhold(zfsvfs, tx); |
e8b96c60 | 3820 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3821 | if (error) { |
3822 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3823 | if (error == ERESTART) { |
e8b96c60 | 3824 | waited = B_TRUE; |
34dc7c2f BB |
3825 | dmu_tx_wait(tx); |
3826 | dmu_tx_abort(tx); | |
3827 | goto top; | |
3828 | } | |
428870ff | 3829 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 3830 | dmu_tx_abort(tx); |
0037b49e | 3831 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3832 | return (error); |
3833 | } | |
3834 | ||
34dc7c2f BB |
3835 | /* |
3836 | * Create a new object for the symlink. | |
428870ff | 3837 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 3838 | */ |
428870ff | 3839 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3840 | |
428870ff | 3841 | if (fuid_dirtied) |
0037b49e | 3842 | zfs_fuid_sync(zfsvfs, tx); |
34dc7c2f | 3843 | |
572e2857 | 3844 | mutex_enter(&zp->z_lock); |
428870ff | 3845 | if (zp->z_is_sa) |
0037b49e | 3846 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), |
428870ff BB |
3847 | link, len, tx); |
3848 | else | |
3849 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3850 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3851 | |
428870ff | 3852 | zp->z_size = len; |
0037b49e | 3853 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 3854 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
3855 | /* |
3856 | * Insert the new object into the directory. | |
3857 | */ | |
3858 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
428870ff BB |
3859 | |
3860 | if (flags & FIGNORECASE) | |
3861 | txtype |= TX_CI; | |
3862 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
9babb374 | 3863 | |
960e08fe BB |
3864 | zfs_inode_update(dzp); |
3865 | zfs_inode_update(zp); | |
3866 | ||
9babb374 | 3867 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3868 | |
3869 | dmu_tx_commit(tx); | |
3870 | ||
3871 | zfs_dirent_unlock(dl); | |
3872 | ||
3558fd73 | 3873 | *ipp = ZTOI(zp); |
34dc7c2f | 3874 | |
0037b49e | 3875 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3876 | zil_commit(zilog, 0); |
428870ff | 3877 | |
0037b49e | 3878 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3879 | return (error); |
3880 | } | |
3881 | ||
3882 | /* | |
3883 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 3884 | * the symbolic path referred to by ip. |
34dc7c2f | 3885 | * |
8b4f9a2d BB |
3886 | * IN: ip - inode of symbolic link |
3887 | * uio - structure to contain the link path. | |
3888 | * cr - credentials of caller. | |
34dc7c2f BB |
3889 | * |
3890 | * RETURN: 0 if success | |
3891 | * error code if failure | |
3892 | * | |
3893 | * Timestamps: | |
3558fd73 | 3894 | * ip - atime updated |
34dc7c2f BB |
3895 | */ |
3896 | /* ARGSUSED */ | |
e5c39b95 | 3897 | int |
8b4f9a2d | 3898 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 3899 | { |
3558fd73 | 3900 | znode_t *zp = ITOZ(ip); |
0037b49e | 3901 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
3902 | int error; |
3903 | ||
0037b49e | 3904 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
3905 | ZFS_VERIFY_ZP(zp); |
3906 | ||
572e2857 | 3907 | mutex_enter(&zp->z_lock); |
428870ff | 3908 | if (zp->z_is_sa) |
8b4f9a2d | 3909 | error = sa_lookup_uio(zp->z_sa_hdl, |
0037b49e | 3910 | SA_ZPL_SYMLINK(zfsvfs), uio); |
428870ff | 3911 | else |
8b4f9a2d | 3912 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 3913 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3914 | |
0037b49e | 3915 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3916 | return (error); |
3917 | } | |
3918 | ||
3919 | /* | |
3558fd73 | 3920 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 3921 | * |
3558fd73 BB |
3922 | * IN: tdip - Directory to contain new entry. |
3923 | * sip - inode of new entry. | |
34dc7c2f BB |
3924 | * name - name of new entry. |
3925 | * cr - credentials of caller. | |
34dc7c2f BB |
3926 | * |
3927 | * RETURN: 0 if success | |
3928 | * error code if failure | |
3929 | * | |
3930 | * Timestamps: | |
3558fd73 BB |
3931 | * tdip - ctime|mtime updated |
3932 | * sip - ctime updated | |
34dc7c2f BB |
3933 | */ |
3934 | /* ARGSUSED */ | |
e5c39b95 | 3935 | int |
da5e151f BB |
3936 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr, |
3937 | int flags) | |
34dc7c2f | 3938 | { |
3558fd73 | 3939 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 3940 | znode_t *tzp, *szp; |
0037b49e | 3941 | zfsvfs_t *zfsvfs = ITOZSB(tdip); |
34dc7c2f BB |
3942 | zilog_t *zilog; |
3943 | zfs_dirlock_t *dl; | |
3944 | dmu_tx_t *tx; | |
34dc7c2f BB |
3945 | int error; |
3946 | int zf = ZNEW; | |
428870ff | 3947 | uint64_t parent; |
572e2857 | 3948 | uid_t owner; |
e8b96c60 | 3949 | boolean_t waited = B_FALSE; |
ace1eae8 CC |
3950 | boolean_t is_tmpfile = 0; |
3951 | uint64_t txg; | |
3952 | #ifdef HAVE_TMPFILE | |
3953 | is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); | |
3954 | #endif | |
3558fd73 | 3955 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 3956 | |
32dec7bd | 3957 | if (name == NULL) |
3958 | return (SET_ERROR(EINVAL)); | |
3959 | ||
0037b49e | 3960 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 3961 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 3962 | zilog = zfsvfs->z_log; |
34dc7c2f | 3963 | |
428870ff BB |
3964 | /* |
3965 | * POSIX dictates that we return EPERM here. | |
3966 | * Better choices include ENOTSUP or EISDIR. | |
3967 | */ | |
3558fd73 | 3968 | if (S_ISDIR(sip->i_mode)) { |
0037b49e | 3969 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3970 | return (SET_ERROR(EPERM)); |
428870ff BB |
3971 | } |
3972 | ||
812e91a7 MT |
3973 | szp = ITOZ(sip); |
3974 | ZFS_VERIFY_ZP(szp); | |
3975 | ||
3976 | /* | |
3977 | * We check i_sb because snapshots and the ctldir must have different | |
3978 | * super blocks. | |
3979 | */ | |
c0ebc844 | 3980 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
0037b49e | 3981 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3982 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 3983 | } |
428870ff | 3984 | |
428870ff BB |
3985 | /* Prevent links to .zfs/shares files */ |
3986 | ||
0037b49e | 3987 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff | 3988 | &parent, sizeof (uint64_t))) != 0) { |
0037b49e | 3989 | ZFS_EXIT(zfsvfs); |
428870ff BB |
3990 | return (error); |
3991 | } | |
0037b49e BB |
3992 | if (parent == zfsvfs->z_shares_dir) { |
3993 | ZFS_EXIT(zfsvfs); | |
2e528b49 | 3994 | return (SET_ERROR(EPERM)); |
428870ff BB |
3995 | } |
3996 | ||
0037b49e | 3997 | if (zfsvfs->z_utf8 && u8_validate(name, |
34dc7c2f | 3998 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 3999 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4000 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
4001 | } |
4002 | if (flags & FIGNORECASE) | |
4003 | zf |= ZCILOOK; | |
4004 | ||
34dc7c2f BB |
4005 | /* |
4006 | * We do not support links between attributes and non-attributes | |
4007 | * because of the potential security risk of creating links | |
4008 | * into "normal" file space in order to circumvent restrictions | |
4009 | * imposed in attribute space. | |
4010 | */ | |
428870ff | 4011 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 4012 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4013 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4014 | } |
4015 | ||
0037b49e BB |
4016 | owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid), |
4017 | cr, ZFS_OWNER); | |
572e2857 | 4018 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
0037b49e | 4019 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4020 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4021 | } |
4022 | ||
149e873a | 4023 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
0037b49e | 4024 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4025 | return (error); |
4026 | } | |
4027 | ||
428870ff | 4028 | top: |
34dc7c2f BB |
4029 | /* |
4030 | * Attempt to lock directory; fail if entry already exists. | |
4031 | */ | |
4032 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
4033 | if (error) { | |
0037b49e | 4034 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4035 | return (error); |
4036 | } | |
4037 | ||
0037b49e | 4038 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 4039 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 4040 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
ace1eae8 | 4041 | if (is_tmpfile) |
0037b49e | 4042 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 4043 | |
428870ff BB |
4044 | zfs_sa_upgrade_txholds(tx, szp); |
4045 | zfs_sa_upgrade_txholds(tx, dzp); | |
e8b96c60 | 4046 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
4047 | if (error) { |
4048 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4049 | if (error == ERESTART) { |
e8b96c60 | 4050 | waited = B_TRUE; |
34dc7c2f BB |
4051 | dmu_tx_wait(tx); |
4052 | dmu_tx_abort(tx); | |
4053 | goto top; | |
4054 | } | |
4055 | dmu_tx_abort(tx); | |
0037b49e | 4056 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4057 | return (error); |
4058 | } | |
ace1eae8 CC |
4059 | /* unmark z_unlinked so zfs_link_create will not reject */ |
4060 | if (is_tmpfile) | |
4061 | szp->z_unlinked = 0; | |
34dc7c2f BB |
4062 | error = zfs_link_create(dl, szp, tx, 0); |
4063 | ||
4064 | if (error == 0) { | |
4065 | uint64_t txtype = TX_LINK; | |
ace1eae8 CC |
4066 | /* |
4067 | * tmpfile is created to be in z_unlinkedobj, so remove it. | |
4068 | * Also, we don't log in ZIL, be cause all previous file | |
4069 | * operation on the tmpfile are ignored by ZIL. Instead we | |
4070 | * always wait for txg to sync to make sure all previous | |
4071 | * operation are sync safe. | |
4072 | */ | |
4073 | if (is_tmpfile) { | |
0037b49e BB |
4074 | VERIFY(zap_remove_int(zfsvfs->z_os, |
4075 | zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); | |
ace1eae8 CC |
4076 | } else { |
4077 | if (flags & FIGNORECASE) | |
4078 | txtype |= TX_CI; | |
4079 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); | |
4080 | } | |
4081 | } else if (is_tmpfile) { | |
4082 | /* restore z_unlinked since when linking failed */ | |
4083 | szp->z_unlinked = 1; | |
34dc7c2f | 4084 | } |
ace1eae8 | 4085 | txg = dmu_tx_get_txg(tx); |
34dc7c2f BB |
4086 | dmu_tx_commit(tx); |
4087 | ||
4088 | zfs_dirent_unlock(dl); | |
4089 | ||
0037b49e | 4090 | if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4091 | zil_commit(zilog, 0); |
428870ff | 4092 | |
ace1eae8 | 4093 | if (is_tmpfile) |
0037b49e | 4094 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg); |
ace1eae8 | 4095 | |
960e08fe BB |
4096 | zfs_inode_update(dzp); |
4097 | zfs_inode_update(szp); | |
0037b49e | 4098 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4099 | return (error); |
4100 | } | |
4101 | ||
3c0e5c0f | 4102 | static void |
119a394a | 4103 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
4104 | { |
4105 | struct page *pp = arg; | |
4106 | ||
119a394a | 4107 | ClearPageError(pp); |
3c0e5c0f BB |
4108 | end_page_writeback(pp); |
4109 | } | |
4110 | ||
34dc7c2f | 4111 | /* |
3c0e5c0f BB |
4112 | * Push a page out to disk, once the page is on stable storage the |
4113 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 4114 | * |
3c0e5c0f BB |
4115 | * IN: ip - page mapped for inode. |
4116 | * pp - page to push (page is locked) | |
4117 | * wbc - writeback control data | |
34dc7c2f BB |
4118 | * |
4119 | * RETURN: 0 if success | |
4120 | * error code if failure | |
4121 | * | |
3c0e5c0f BB |
4122 | * Timestamps: |
4123 | * ip - ctime|mtime updated | |
34dc7c2f BB |
4124 | */ |
4125 | /* ARGSUSED */ | |
3c0e5c0f BB |
4126 | int |
4127 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 4128 | { |
3c0e5c0f | 4129 | znode_t *zp = ITOZ(ip); |
0037b49e | 4130 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
3c0e5c0f BB |
4131 | loff_t offset; |
4132 | loff_t pgoff; | |
4c837f0d BB |
4133 | unsigned int pglen; |
4134 | rl_t *rl; | |
3c0e5c0f BB |
4135 | dmu_tx_t *tx; |
4136 | caddr_t va; | |
4137 | int err = 0; | |
4138 | uint64_t mtime[2], ctime[2]; | |
4139 | sa_bulk_attr_t bulk[3]; | |
4140 | int cnt = 0; | |
21a96fb6 | 4141 | struct address_space *mapping; |
3c0e5c0f | 4142 | |
0037b49e | 4143 | ZFS_ENTER(zfsvfs); |
4c837f0d | 4144 | ZFS_VERIFY_ZP(zp); |
d164b209 | 4145 | |
3c0e5c0f BB |
4146 | ASSERT(PageLocked(pp)); |
4147 | ||
d1d7e268 MK |
4148 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
4149 | offset = i_size_read(ip); /* File length in bytes */ | |
8b1899d3 BB |
4150 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
4151 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
4152 | |
4153 | /* Page is beyond end of file */ | |
4154 | if (pgoff >= offset) { | |
4155 | unlock_page(pp); | |
0037b49e | 4156 | ZFS_EXIT(zfsvfs); |
3c0e5c0f BB |
4157 | return (0); |
4158 | } | |
4159 | ||
4160 | /* Truncate page length to end of file */ | |
4161 | if (pgoff + pglen > offset) | |
4162 | pglen = offset - pgoff; | |
4163 | ||
4164 | #if 0 | |
34dc7c2f | 4165 | /* |
3c0e5c0f BB |
4166 | * FIXME: Allow mmap writes past its quota. The correct fix |
4167 | * is to register a page_mkwrite() handler to count the page | |
4168 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 4169 | */ |
0037b49e BB |
4170 | if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || |
4171 | zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { | |
9babb374 | 4172 | err = EDQUOT; |
9babb374 | 4173 | } |
3c0e5c0f BB |
4174 | #endif |
4175 | ||
d958324f BB |
4176 | /* |
4177 | * The ordering here is critical and must adhere to the following | |
4178 | * rules in order to avoid deadlocking in either zfs_read() or | |
4179 | * zfs_free_range() due to a lock inversion. | |
4180 | * | |
4181 | * 1) The page must be unlocked prior to acquiring the range lock. | |
4182 | * This is critical because zfs_read() calls find_lock_page() | |
4183 | * which may block on the page lock while holding the range lock. | |
4184 | * | |
4185 | * 2) Before setting or clearing write back on a page the range lock | |
4186 | * must be held in order to prevent a lock inversion with the | |
4187 | * zfs_free_range() function. | |
21a96fb6 CC |
4188 | * |
4189 | * This presents a problem because upon entering this function the | |
4190 | * page lock is already held. To safely acquire the range lock the | |
4191 | * page lock must be dropped. This creates a window where another | |
4192 | * process could truncate, invalidate, dirty, or write out the page. | |
4193 | * | |
4194 | * Therefore, after successfully reacquiring the range and page locks | |
4195 | * the current page state is checked. In the common case everything | |
4196 | * will be as is expected and it can be written out. However, if | |
4197 | * the page state has changed it must be handled accordingly. | |
d958324f | 4198 | */ |
21a96fb6 CC |
4199 | mapping = pp->mapping; |
4200 | redirty_page_for_writepage(wbc, pp); | |
d958324f | 4201 | unlock_page(pp); |
21a96fb6 | 4202 | |
d88895a0 | 4203 | rl = zfs_range_lock(&zp->z_range_lock, pgoff, pglen, RL_WRITER); |
21a96fb6 CC |
4204 | lock_page(pp); |
4205 | ||
4206 | /* Page mapping changed or it was no longer dirty, we're done */ | |
4207 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
4208 | unlock_page(pp); | |
4209 | zfs_range_unlock(rl); | |
0037b49e | 4210 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4211 | return (0); |
4212 | } | |
4213 | ||
4214 | /* Another process started write block if required */ | |
4215 | if (PageWriteback(pp)) { | |
4216 | unlock_page(pp); | |
4217 | zfs_range_unlock(rl); | |
4218 | ||
4219 | if (wbc->sync_mode != WB_SYNC_NONE) | |
4220 | wait_on_page_writeback(pp); | |
4221 | ||
0037b49e | 4222 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4223 | return (0); |
4224 | } | |
4225 | ||
4226 | /* Clear the dirty flag the required locks are held */ | |
4227 | if (!clear_page_dirty_for_io(pp)) { | |
4228 | unlock_page(pp); | |
4229 | zfs_range_unlock(rl); | |
0037b49e | 4230 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4231 | return (0); |
4232 | } | |
4233 | ||
4234 | /* | |
4235 | * Counterpart for redirty_page_for_writepage() above. This page | |
4236 | * was in fact not skipped and should not be counted as if it were. | |
4237 | */ | |
4238 | wbc->pages_skipped--; | |
3c0e5c0f | 4239 | set_page_writeback(pp); |
21a96fb6 | 4240 | unlock_page(pp); |
3c0e5c0f | 4241 | |
0037b49e | 4242 | tx = dmu_tx_create(zfsvfs->z_os); |
3c0e5c0f | 4243 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
4244 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4245 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 4246 | |
fb5f0bc8 | 4247 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 4248 | if (err != 0) { |
3c0e5c0f | 4249 | if (err == ERESTART) |
34dc7c2f | 4250 | dmu_tx_wait(tx); |
3c0e5c0f | 4251 | |
34dc7c2f | 4252 | dmu_tx_abort(tx); |
119a394a ED |
4253 | __set_page_dirty_nobuffers(pp); |
4254 | ClearPageError(pp); | |
4255 | end_page_writeback(pp); | |
4c837f0d | 4256 | zfs_range_unlock(rl); |
0037b49e | 4257 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4258 | return (err); |
34dc7c2f BB |
4259 | } |
4260 | ||
dde471ef | 4261 | va = kmap(pp); |
8b1899d3 | 4262 | ASSERT3U(pglen, <=, PAGE_SIZE); |
0037b49e | 4263 | dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 4264 | kunmap(pp); |
34dc7c2f | 4265 | |
0037b49e BB |
4266 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
4267 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
4268 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, | |
4269 | &zp->z_pflags, 8); | |
428870ff | 4270 | |
d3aa3ea9 BB |
4271 | /* Preserve the mtime and ctime provided by the inode */ |
4272 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4273 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
4274 | zp->z_atime_dirty = 0; | |
4275 | zp->z_seq++; | |
4276 | ||
4277 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4278 | ||
0037b49e | 4279 | zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
119a394a | 4280 | zfs_putpage_commit_cb, pp); |
45d1cae3 | 4281 | dmu_tx_commit(tx); |
d3aa3ea9 | 4282 | |
4c837f0d | 4283 | zfs_range_unlock(rl); |
34dc7c2f | 4284 | |
119a394a ED |
4285 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4286 | /* | |
4287 | * Note that this is rarely called under writepages(), because | |
4288 | * writepages() normally handles the entire commit for | |
4289 | * performance reasons. | |
4290 | */ | |
0037b49e | 4291 | zil_commit(zfsvfs->z_log, zp->z_id); |
2b286136 | 4292 | } |
3c0e5c0f | 4293 | |
0037b49e | 4294 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4295 | return (err); |
34dc7c2f BB |
4296 | } |
4297 | ||
8780c539 BB |
4298 | /* |
4299 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 4300 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
4301 | */ |
4302 | int | |
4303 | zfs_dirty_inode(struct inode *ip, int flags) | |
4304 | { | |
4305 | znode_t *zp = ITOZ(ip); | |
0037b49e | 4306 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
8780c539 | 4307 | dmu_tx_t *tx; |
023699cd MM |
4308 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
4309 | sa_bulk_attr_t bulk[4]; | |
704cd075 | 4310 | int error = 0; |
8780c539 BB |
4311 | int cnt = 0; |
4312 | ||
0037b49e | 4313 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
c944be5d BB |
4314 | return (0); |
4315 | ||
0037b49e | 4316 | ZFS_ENTER(zfsvfs); |
8780c539 BB |
4317 | ZFS_VERIFY_ZP(zp); |
4318 | ||
704cd075 CC |
4319 | #ifdef I_DIRTY_TIME |
4320 | /* | |
4321 | * This is the lazytime semantic indroduced in Linux 4.0 | |
4322 | * This flag will only be called from update_time when lazytime is set. | |
4323 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
4324 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
4325 | * only need to dirty atime. | |
4326 | */ | |
4327 | if (flags == I_DIRTY_TIME) { | |
4328 | zp->z_atime_dirty = 1; | |
4329 | goto out; | |
4330 | } | |
4331 | #endif | |
4332 | ||
0037b49e | 4333 | tx = dmu_tx_create(zfsvfs->z_os); |
8780c539 BB |
4334 | |
4335 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4336 | zfs_sa_upgrade_txholds(tx, zp); | |
4337 | ||
4338 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4339 | if (error) { | |
4340 | dmu_tx_abort(tx); | |
4341 | goto out; | |
4342 | } | |
4343 | ||
4344 | mutex_enter(&zp->z_lock); | |
704cd075 CC |
4345 | zp->z_atime_dirty = 0; |
4346 | ||
0037b49e BB |
4347 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
4348 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
4349 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
4350 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
8780c539 | 4351 | |
023699cd | 4352 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4353 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4354 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4355 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
023699cd MM |
4356 | mode = ip->i_mode; |
4357 | ||
4358 | zp->z_mode = mode; | |
8780c539 BB |
4359 | |
4360 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4361 | mutex_exit(&zp->z_lock); | |
4362 | ||
4363 | dmu_tx_commit(tx); | |
4364 | out: | |
0037b49e | 4365 | ZFS_EXIT(zfsvfs); |
8780c539 BB |
4366 | return (error); |
4367 | } | |
8780c539 | 4368 | |
34dc7c2f BB |
4369 | /*ARGSUSED*/ |
4370 | void | |
c0d35759 | 4371 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4372 | { |
c0d35759 | 4373 | znode_t *zp = ITOZ(ip); |
0037b49e | 4374 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
0df9673f | 4375 | uint64_t atime[2]; |
34dc7c2f | 4376 | int error; |
cafbd2ac | 4377 | int need_unlock = 0; |
34dc7c2f | 4378 | |
cafbd2ac | 4379 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
0037b49e | 4380 | if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) { |
cafbd2ac | 4381 | need_unlock = 1; |
0037b49e | 4382 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); |
cafbd2ac | 4383 | } |
c0d35759 | 4384 | if (zp->z_sa_hdl == NULL) { |
cafbd2ac | 4385 | if (need_unlock) |
0037b49e | 4386 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
c0d35759 | 4387 | return; |
34dc7c2f BB |
4388 | } |
4389 | ||
4390 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
0037b49e | 4391 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 4392 | |
428870ff BB |
4393 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4394 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4395 | error = dmu_tx_assign(tx, TXG_WAIT); |
4396 | if (error) { | |
4397 | dmu_tx_abort(tx); | |
4398 | } else { | |
0df9673f | 4399 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
34dc7c2f | 4400 | mutex_enter(&zp->z_lock); |
0037b49e | 4401 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), |
0df9673f | 4402 | (void *)&atime, sizeof (atime), tx); |
34dc7c2f BB |
4403 | zp->z_atime_dirty = 0; |
4404 | mutex_exit(&zp->z_lock); | |
4405 | dmu_tx_commit(tx); | |
4406 | } | |
4407 | } | |
4408 | ||
4409 | zfs_zinactive(zp); | |
cafbd2ac | 4410 | if (need_unlock) |
0037b49e | 4411 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
34dc7c2f BB |
4412 | } |
4413 | ||
4414 | /* | |
4415 | * Bounds-check the seek operation. | |
4416 | * | |
3558fd73 | 4417 | * IN: ip - inode seeking within |
34dc7c2f BB |
4418 | * ooff - old file offset |
4419 | * noffp - pointer to new file offset | |
4420 | * ct - caller context | |
4421 | * | |
4422 | * RETURN: 0 if success | |
4423 | * EINVAL if new offset invalid | |
4424 | */ | |
4425 | /* ARGSUSED */ | |
3558fd73 | 4426 | int |
9623f736 | 4427 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4428 | { |
3558fd73 | 4429 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4430 | return (0); |
4431 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4432 | } | |
4433 | ||
34dc7c2f | 4434 | /* |
dde471ef | 4435 | * Fill pages with data from the disk. |
34dc7c2f BB |
4436 | */ |
4437 | static int | |
dde471ef | 4438 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4439 | { |
d1d7e268 | 4440 | znode_t *zp = ITOZ(ip); |
0037b49e | 4441 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
d1d7e268 | 4442 | objset_t *os; |
dde471ef | 4443 | struct page *cur_pp; |
d1d7e268 MK |
4444 | u_offset_t io_off, total; |
4445 | size_t io_len; | |
4446 | loff_t i_size; | |
4447 | unsigned page_idx; | |
4448 | int err; | |
34dc7c2f | 4449 | |
0037b49e | 4450 | os = zfsvfs->z_os; |
8b1899d3 | 4451 | io_len = nr_pages << PAGE_SHIFT; |
dde471ef PJ |
4452 | i_size = i_size_read(ip); |
4453 | io_off = page_offset(pl[0]); | |
4454 | ||
4455 | if (io_off + io_len > i_size) | |
4456 | io_len = i_size - io_off; | |
34dc7c2f BB |
4457 | |
4458 | /* | |
dde471ef | 4459 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4460 | */ |
dde471ef | 4461 | page_idx = 0; |
34dc7c2f | 4462 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4463 | caddr_t va; |
4464 | ||
540c3927 | 4465 | cur_pp = pl[page_idx++]; |
dde471ef | 4466 | va = kmap(cur_pp); |
9babb374 BB |
4467 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4468 | DMU_READ_PREFETCH); | |
dde471ef | 4469 | kunmap(cur_pp); |
34dc7c2f | 4470 | if (err) { |
b128c09f BB |
4471 | /* convert checksum errors into IO errors */ |
4472 | if (err == ECKSUM) | |
2e528b49 | 4473 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4474 | return (err); |
4475 | } | |
34dc7c2f | 4476 | } |
d164b209 | 4477 | |
34dc7c2f BB |
4478 | return (0); |
4479 | } | |
4480 | ||
4481 | /* | |
dde471ef | 4482 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4483 | * |
dde471ef PJ |
4484 | * IN: ip - inode of file to get data from. |
4485 | * pl - list of pages to read | |
4486 | * nr_pages - number of pages to read | |
34dc7c2f | 4487 | * |
d3cc8b15 | 4488 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4489 | * |
4490 | * Timestamps: | |
4491 | * vp - atime updated | |
4492 | */ | |
4493 | /* ARGSUSED */ | |
dde471ef PJ |
4494 | int |
4495 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4496 | { |
dde471ef | 4497 | znode_t *zp = ITOZ(ip); |
0037b49e | 4498 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
dde471ef | 4499 | int err; |
d164b209 | 4500 | |
d164b209 BB |
4501 | if (pl == NULL) |
4502 | return (0); | |
34dc7c2f | 4503 | |
0037b49e | 4504 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4505 | ZFS_VERIFY_ZP(zp); |
4506 | ||
dde471ef | 4507 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4508 | |
0037b49e | 4509 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4510 | return (err); |
4511 | } | |
4512 | ||
4513 | /* | |
e2e7aa2d | 4514 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4515 | * |
e2e7aa2d BB |
4516 | * IN: ip - inode of the file to mmap |
4517 | * off - file offset | |
4518 | * addrp - start address in memory region | |
4519 | * len - length of memory region | |
4520 | * vm_flags- address flags | |
34dc7c2f | 4521 | * |
e2e7aa2d BB |
4522 | * RETURN: 0 if success |
4523 | * error code if failure | |
34dc7c2f BB |
4524 | */ |
4525 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4526 | int |
4527 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4528 | unsigned long vm_flags) | |
34dc7c2f | 4529 | { |
e2e7aa2d | 4530 | znode_t *zp = ITOZ(ip); |
0037b49e | 4531 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4532 | |
0037b49e | 4533 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4534 | ZFS_VERIFY_ZP(zp); |
4535 | ||
e2e7aa2d | 4536 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4537 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
0037b49e | 4538 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4539 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4540 | } |
4541 | ||
e2e7aa2d | 4542 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4543 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
0037b49e | 4544 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4545 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4546 | } |
4547 | ||
34dc7c2f | 4548 | if (off < 0 || len > MAXOFFSET_T - off) { |
0037b49e | 4549 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4550 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4551 | } |
4552 | ||
0037b49e | 4553 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4554 | return (0); |
4555 | } | |
4556 | ||
3558fd73 BB |
4557 | /* |
4558 | * convoff - converts the given data (start, whence) to the | |
4559 | * given whence. | |
4560 | */ | |
4561 | int | |
4562 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4563 | { | |
5484965a | 4564 | vattr_t vap; |
3558fd73 BB |
4565 | int error; |
4566 | ||
4567 | if ((lckdat->l_whence == 2) || (whence == 2)) { | |
d95a5980 | 4568 | if ((error = zfs_getattr(ip, &vap, 0, CRED()))) |
3558fd73 BB |
4569 | return (error); |
4570 | } | |
4571 | ||
4572 | switch (lckdat->l_whence) { | |
4573 | case 1: | |
4574 | lckdat->l_start += offset; | |
4575 | break; | |
4576 | case 2: | |
5484965a | 4577 | lckdat->l_start += vap.va_size; |
3558fd73 BB |
4578 | /* FALLTHRU */ |
4579 | case 0: | |
4580 | break; | |
4581 | default: | |
2e528b49 | 4582 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4583 | } |
4584 | ||
4585 | if (lckdat->l_start < 0) | |
2e528b49 | 4586 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4587 | |
4588 | switch (whence) { | |
4589 | case 1: | |
4590 | lckdat->l_start -= offset; | |
4591 | break; | |
4592 | case 2: | |
5484965a | 4593 | lckdat->l_start -= vap.va_size; |
3558fd73 BB |
4594 | /* FALLTHRU */ |
4595 | case 0: | |
4596 | break; | |
4597 | default: | |
2e528b49 | 4598 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4599 | } |
4600 | ||
4601 | lckdat->l_whence = (short)whence; | |
4602 | return (0); | |
4603 | } | |
4604 | ||
34dc7c2f BB |
4605 | /* |
4606 | * Free or allocate space in a file. Currently, this function only | |
4607 | * supports the `F_FREESP' command. However, this command is somewhat | |
4608 | * misnamed, as its functionality includes the ability to allocate as | |
4609 | * well as free space. | |
4610 | * | |
3558fd73 | 4611 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4612 | * cmd - action to take (only F_FREESP supported). |
4613 | * bfp - section of file to free/alloc. | |
4614 | * flag - current file open mode flags. | |
4615 | * offset - current file offset. | |
4616 | * cr - credentials of caller [UNUSED]. | |
34dc7c2f | 4617 | * |
d3cc8b15 | 4618 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4619 | * |
4620 | * Timestamps: | |
3558fd73 | 4621 | * ip - ctime|mtime updated |
34dc7c2f BB |
4622 | */ |
4623 | /* ARGSUSED */ | |
e5c39b95 | 4624 | int |
3558fd73 BB |
4625 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4626 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4627 | { |
3558fd73 | 4628 | znode_t *zp = ITOZ(ip); |
0037b49e | 4629 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4630 | uint64_t off, len; |
4631 | int error; | |
4632 | ||
0037b49e | 4633 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4634 | ZFS_VERIFY_ZP(zp); |
4635 | ||
34dc7c2f | 4636 | if (cmd != F_FREESP) { |
0037b49e | 4637 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4638 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4639 | } |
4640 | ||
f3c9dca0 MT |
4641 | /* |
4642 | * Callers might not be able to detect properly that we are read-only, | |
4643 | * so check it explicitly here. | |
4644 | */ | |
0037b49e BB |
4645 | if (zfs_is_readonly(zfsvfs)) { |
4646 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
4647 | return (SET_ERROR(EROFS)); |
4648 | } | |
4649 | ||
3558fd73 | 4650 | if ((error = convoff(ip, bfp, 0, offset))) { |
0037b49e | 4651 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4652 | return (error); |
4653 | } | |
4654 | ||
4655 | if (bfp->l_len < 0) { | |
0037b49e | 4656 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4657 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4658 | } |
4659 | ||
aec69371 ED |
4660 | /* |
4661 | * Permissions aren't checked on Solaris because on this OS | |
4662 | * zfs_space() can only be called with an opened file handle. | |
4663 | * On Linux we can get here through truncate_range() which | |
4664 | * operates directly on inodes, so we need to check access rights. | |
4665 | */ | |
4666 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
0037b49e | 4667 | ZFS_EXIT(zfsvfs); |
aec69371 ED |
4668 | return (error); |
4669 | } | |
4670 | ||
34dc7c2f BB |
4671 | off = bfp->l_start; |
4672 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4673 | ||
b128c09f | 4674 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4675 | |
0037b49e | 4676 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4677 | return (error); |
4678 | } | |
4679 | ||
4680 | /*ARGSUSED*/ | |
e5c39b95 | 4681 | int |
3558fd73 | 4682 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4683 | { |
3558fd73 | 4684 | znode_t *zp = ITOZ(ip); |
0037b49e | 4685 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4686 | uint32_t gen; |
428870ff | 4687 | uint64_t gen64; |
34dc7c2f BB |
4688 | uint64_t object = zp->z_id; |
4689 | zfid_short_t *zfid; | |
428870ff | 4690 | int size, i, error; |
34dc7c2f | 4691 | |
0037b49e | 4692 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4693 | ZFS_VERIFY_ZP(zp); |
428870ff | 4694 | |
0037b49e | 4695 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), |
428870ff | 4696 | &gen64, sizeof (uint64_t))) != 0) { |
0037b49e | 4697 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4698 | return (error); |
4699 | } | |
4700 | ||
4701 | gen = (uint32_t)gen64; | |
34dc7c2f | 4702 | |
9b77d1c9 | 4703 | size = SHORT_FID_LEN; |
34dc7c2f BB |
4704 | |
4705 | zfid = (zfid_short_t *)fidp; | |
4706 | ||
4707 | zfid->zf_len = size; | |
4708 | ||
4709 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4710 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4711 | ||
4712 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4713 | if (gen == 0) | |
4714 | gen = 1; | |
4715 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4716 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4717 | ||
0037b49e | 4718 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4719 | return (0); |
4720 | } | |
4721 | ||
34dc7c2f | 4722 | /*ARGSUSED*/ |
e5c39b95 | 4723 | int |
3558fd73 | 4724 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4725 | { |
3558fd73 | 4726 | znode_t *zp = ITOZ(ip); |
0037b49e | 4727 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4728 | int error; |
4729 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
4730 | ||
0037b49e | 4731 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4732 | ZFS_VERIFY_ZP(zp); |
4733 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
0037b49e | 4734 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4735 | |
4736 | return (error); | |
4737 | } | |
4738 | ||
4739 | /*ARGSUSED*/ | |
e5c39b95 | 4740 | int |
3558fd73 | 4741 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4742 | { |
3558fd73 | 4743 | znode_t *zp = ITOZ(ip); |
0037b49e | 4744 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4745 | int error; |
4746 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
0037b49e | 4747 | zilog_t *zilog = zfsvfs->z_log; |
34dc7c2f | 4748 | |
0037b49e | 4749 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4750 | ZFS_VERIFY_ZP(zp); |
428870ff | 4751 | |
34dc7c2f | 4752 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 4753 | |
0037b49e | 4754 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4755 | zil_commit(zilog, 0); |
428870ff | 4756 | |
0037b49e | 4757 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4758 | return (error); |
4759 | } | |
4760 | ||
3558fd73 | 4761 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
4762 | /* |
4763 | * Tunable, both must be a power of 2. | |
4764 | * | |
4765 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
4766 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 4767 | * an arcbuf for a partial block read |
428870ff BB |
4768 | */ |
4769 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
4770 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
4771 | ||
4772 | /*ARGSUSED*/ | |
4773 | static int | |
3558fd73 | 4774 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 4775 | { |
3558fd73 | 4776 | znode_t *zp = ITOZ(ip); |
0037b49e BB |
4777 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
4778 | int max_blksz = zfsvfs->z_max_blksz; | |
428870ff BB |
4779 | uio_t *uio = &xuio->xu_uio; |
4780 | ssize_t size = uio->uio_resid; | |
4781 | offset_t offset = uio->uio_loffset; | |
4782 | int blksz; | |
4783 | int fullblk, i; | |
4784 | arc_buf_t *abuf; | |
4785 | ssize_t maxsize; | |
4786 | int preamble, postamble; | |
4787 | ||
4788 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
2e528b49 | 4789 | return (SET_ERROR(EINVAL)); |
428870ff | 4790 | |
0037b49e | 4791 | ZFS_ENTER(zfsvfs); |
428870ff BB |
4792 | ZFS_VERIFY_ZP(zp); |
4793 | switch (ioflag) { | |
4794 | case UIO_WRITE: | |
4795 | /* | |
4796 | * Loan out an arc_buf for write if write size is bigger than | |
4797 | * max_blksz, and the file's block size is also max_blksz. | |
4798 | */ | |
4799 | blksz = max_blksz; | |
4800 | if (size < blksz || zp->z_blksz != blksz) { | |
0037b49e | 4801 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4802 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4803 | } |
4804 | /* | |
4805 | * Caller requests buffers for write before knowing where the | |
4806 | * write offset might be (e.g. NFS TCP write). | |
4807 | */ | |
4808 | if (offset == -1) { | |
4809 | preamble = 0; | |
4810 | } else { | |
4811 | preamble = P2PHASE(offset, blksz); | |
4812 | if (preamble) { | |
4813 | preamble = blksz - preamble; | |
4814 | size -= preamble; | |
4815 | } | |
4816 | } | |
4817 | ||
4818 | postamble = P2PHASE(size, blksz); | |
4819 | size -= postamble; | |
4820 | ||
4821 | fullblk = size / blksz; | |
4822 | (void) dmu_xuio_init(xuio, | |
4823 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
4824 | |
4825 | /* | |
4826 | * Have to fix iov base/len for partial buffers. They | |
4827 | * currently represent full arc_buf's. | |
4828 | */ | |
4829 | if (preamble) { | |
4830 | /* data begins in the middle of the arc_buf */ | |
4831 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4832 | blksz); | |
4833 | ASSERT(abuf); | |
4834 | (void) dmu_xuio_add(xuio, abuf, | |
4835 | blksz - preamble, preamble); | |
4836 | } | |
4837 | ||
4838 | for (i = 0; i < fullblk; i++) { | |
4839 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4840 | blksz); | |
4841 | ASSERT(abuf); | |
4842 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
4843 | } | |
4844 | ||
4845 | if (postamble) { | |
4846 | /* data ends in the middle of the arc_buf */ | |
4847 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4848 | blksz); | |
4849 | ASSERT(abuf); | |
4850 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
4851 | } | |
4852 | break; | |
4853 | case UIO_READ: | |
4854 | /* | |
4855 | * Loan out an arc_buf for read if the read size is larger than | |
4856 | * the current file block size. Block alignment is not | |
4857 | * considered. Partial arc_buf will be loaned out for read. | |
4858 | */ | |
4859 | blksz = zp->z_blksz; | |
4860 | if (blksz < zcr_blksz_min) | |
4861 | blksz = zcr_blksz_min; | |
4862 | if (blksz > zcr_blksz_max) | |
4863 | blksz = zcr_blksz_max; | |
4864 | /* avoid potential complexity of dealing with it */ | |
4865 | if (blksz > max_blksz) { | |
0037b49e | 4866 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4867 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4868 | } |
4869 | ||
4870 | maxsize = zp->z_size - uio->uio_loffset; | |
4871 | if (size > maxsize) | |
4872 | size = maxsize; | |
4873 | ||
3558fd73 | 4874 | if (size < blksz) { |
0037b49e | 4875 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4876 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4877 | } |
4878 | break; | |
4879 | default: | |
0037b49e | 4880 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4881 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4882 | } |
4883 | ||
4884 | uio->uio_extflg = UIO_XUIO; | |
4885 | XUIO_XUZC_RW(xuio) = ioflag; | |
0037b49e | 4886 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4887 | return (0); |
4888 | } | |
4889 | ||
4890 | /*ARGSUSED*/ | |
4891 | static int | |
3558fd73 | 4892 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
4893 | { |
4894 | int i; | |
4895 | arc_buf_t *abuf; | |
4896 | int ioflag = XUIO_XUZC_RW(xuio); | |
4897 | ||
4898 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
4899 | ||
4900 | i = dmu_xuio_cnt(xuio); | |
4901 | while (i-- > 0) { | |
4902 | abuf = dmu_xuio_arcbuf(xuio, i); | |
4903 | /* | |
4904 | * if abuf == NULL, it must be a write buffer | |
4905 | * that has been returned in zfs_write(). | |
4906 | */ | |
4907 | if (abuf) | |
4908 | dmu_return_arcbuf(abuf); | |
4909 | ASSERT(abuf || ioflag == UIO_WRITE); | |
4910 | } | |
4911 | ||
4912 | dmu_xuio_fini(xuio); | |
4913 | return (0); | |
4914 | } | |
3558fd73 | 4915 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 BB |
4916 | |
4917 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
f298b24d BB |
4918 | EXPORT_SYMBOL(zfs_open); |
4919 | EXPORT_SYMBOL(zfs_close); | |
4920 | EXPORT_SYMBOL(zfs_read); | |
4921 | EXPORT_SYMBOL(zfs_write); | |
4922 | EXPORT_SYMBOL(zfs_access); | |
4923 | EXPORT_SYMBOL(zfs_lookup); | |
4924 | EXPORT_SYMBOL(zfs_create); | |
4925 | EXPORT_SYMBOL(zfs_tmpfile); | |
4926 | EXPORT_SYMBOL(zfs_remove); | |
4927 | EXPORT_SYMBOL(zfs_mkdir); | |
4928 | EXPORT_SYMBOL(zfs_rmdir); | |
4929 | EXPORT_SYMBOL(zfs_readdir); | |
4930 | EXPORT_SYMBOL(zfs_fsync); | |
4931 | EXPORT_SYMBOL(zfs_getattr); | |
4932 | EXPORT_SYMBOL(zfs_getattr_fast); | |
4933 | EXPORT_SYMBOL(zfs_setattr); | |
4934 | EXPORT_SYMBOL(zfs_rename); | |
4935 | EXPORT_SYMBOL(zfs_symlink); | |
4936 | EXPORT_SYMBOL(zfs_readlink); | |
4937 | EXPORT_SYMBOL(zfs_link); | |
4938 | EXPORT_SYMBOL(zfs_inactive); | |
4939 | EXPORT_SYMBOL(zfs_space); | |
4940 | EXPORT_SYMBOL(zfs_fid); | |
4941 | EXPORT_SYMBOL(zfs_getsecattr); | |
4942 | EXPORT_SYMBOL(zfs_setsecattr); | |
4943 | EXPORT_SYMBOL(zfs_getpage); | |
4944 | EXPORT_SYMBOL(zfs_putpage); | |
4945 | EXPORT_SYMBOL(zfs_dirty_inode); | |
4946 | EXPORT_SYMBOL(zfs_map); | |
4947 | ||
02730c33 | 4948 | /* CSTYLED */ |
a966c564 K |
4949 | module_param(zfs_delete_blocks, ulong, 0644); |
4950 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
c409e464 BB |
4951 | module_param(zfs_read_chunk_size, long, 0644); |
4952 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
4953 | #endif |