]>
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> |
9c5167d1 | 82 | #include <sys/sa_impl.h> |
34dc7c2f BB |
83 | |
84 | /* | |
85 | * Programming rules. | |
86 | * | |
87 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
88 | * properly lock its in-core state, create a DMU transaction, do the work, | |
89 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
90 | * and wait for the intent log to commit if it is a synchronous operation. | |
91 | * Moreover, the vnode ops must work in both normal and log replay context. | |
92 | * The ordering of events is important to avoid deadlocks and references | |
93 | * to freed memory. The example below illustrates the following Big Rules: | |
94 | * | |
95 | * (1) A check must be made in each zfs thread for a mounted file system. | |
0037b49e BB |
96 | * This is done avoiding races using ZFS_ENTER(zfsvfs). |
97 | * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes | |
34dc7c2f BB |
98 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros |
99 | * can return EIO from the calling function. | |
100 | * | |
3558fd73 | 101 | * (2) iput() should always be the last thing except for zil_commit() |
34dc7c2f BB |
102 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: |
103 | * First, if it's the last reference, the vnode/znode | |
104 | * can be freed, so the zp may point to freed memory. Second, the last | |
105 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
106 | * pushing cached pages (which acquires range locks) and syncing out | |
107 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
108 | * which could deadlock the system if you were already holding one. | |
0a50679c | 109 | * If you must call iput() within a tx then use zfs_iput_async(). |
34dc7c2f BB |
110 | * |
111 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
112 | * as they can span dmu_tx_assign() calls. | |
113 | * | |
384f8a09 MA |
114 | * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to |
115 | * dmu_tx_assign(). This is critical because we don't want to block | |
116 | * while holding locks. | |
117 | * | |
118 | * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This | |
119 | * reduces lock contention and CPU usage when we must wait (note that if | |
120 | * throughput is constrained by the storage, nearly every transaction | |
121 | * must wait). | |
122 | * | |
123 | * Note, in particular, that if a lock is sometimes acquired before | |
124 | * the tx assigns, and sometimes after (e.g. z_lock), then failing | |
125 | * to use a non-blocking assign can deadlock the system. The scenario: | |
34dc7c2f BB |
126 | * |
127 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
128 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
129 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
130 | * forever, because the previous txg can't quiesce until B's tx commits. | |
131 | * | |
0037b49e | 132 | * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, |
e8b96c60 | 133 | * then drop all locks, call dmu_tx_wait(), and try again. On subsequent |
0735ecb3 | 134 | * calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT, |
e8b96c60 MA |
135 | * to indicate that this operation has already called dmu_tx_wait(). |
136 | * This will ensure that we don't retry forever, waiting a short bit | |
137 | * each time. | |
34dc7c2f BB |
138 | * |
139 | * (5) If the operation succeeded, generate the intent log entry for it | |
140 | * before dropping locks. This ensures that the ordering of events | |
141 | * in the intent log matches the order in which they actually occurred. | |
d3cc8b15 | 142 | * During ZIL replay the zfs_log_* functions will update the sequence |
fb5f0bc8 | 143 | * number to indicate the zil transaction has replayed. |
34dc7c2f BB |
144 | * |
145 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
146 | * regardless of whether there were any errors. | |
147 | * | |
572e2857 | 148 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
149 | * to ensure that synchronous semantics are provided when necessary. |
150 | * | |
151 | * In general, this is how things should be ordered in each vnode op: | |
152 | * | |
0037b49e | 153 | * ZFS_ENTER(zfsvfs); // exit if unmounted |
34dc7c2f | 154 | * top: |
3558fd73 | 155 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
156 | * rw_enter(...); // grab any other locks you need |
157 | * tx = dmu_tx_create(...); // get DMU tx | |
158 | * dmu_tx_hold_*(); // hold each object you might modify | |
0735ecb3 | 159 | * error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
160 | * if (error) { |
161 | * rw_exit(...); // drop locks | |
162 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 163 | * iput(...); // release held vnodes |
fb5f0bc8 | 164 | * if (error == ERESTART) { |
e8b96c60 | 165 | * waited = B_TRUE; |
34dc7c2f BB |
166 | * dmu_tx_wait(tx); |
167 | * dmu_tx_abort(tx); | |
168 | * goto top; | |
169 | * } | |
170 | * dmu_tx_abort(tx); // abort DMU tx | |
0037b49e | 171 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
172 | * return (error); // really out of space |
173 | * } | |
174 | * error = do_real_work(); // do whatever this VOP does | |
175 | * if (error == 0) | |
176 | * zfs_log_*(...); // on success, make ZIL entry | |
177 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
178 | * rw_exit(...); // drop locks | |
179 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 180 | * iput(...); // release held vnodes |
572e2857 | 181 | * zil_commit(zilog, foid); // synchronous when necessary |
0037b49e | 182 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
183 | * return (error); // done, report error |
184 | */ | |
185 | ||
126400a1 BB |
186 | /* |
187 | * Virus scanning is unsupported. It would be possible to add a hook | |
188 | * here to performance the required virus scan. This could be done | |
189 | * entirely in the kernel or potentially as an update to invoke a | |
190 | * scanning utility. | |
191 | */ | |
192 | static int | |
193 | zfs_vscan(struct inode *ip, cred_t *cr, int async) | |
194 | { | |
195 | return (0); | |
196 | } | |
197 | ||
198 | /* ARGSUSED */ | |
199 | int | |
200 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
201 | { | |
202 | znode_t *zp = ITOZ(ip); | |
0037b49e | 203 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 204 | |
0037b49e | 205 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
206 | ZFS_VERIFY_ZP(zp); |
207 | ||
208 | /* Honor ZFS_APPENDONLY file attribute */ | |
209 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
210 | ((flag & O_APPEND) == 0)) { | |
0037b49e | 211 | ZFS_EXIT(zfsvfs); |
2e528b49 | 212 | return (SET_ERROR(EPERM)); |
126400a1 BB |
213 | } |
214 | ||
215 | /* Virus scan eligible files on open */ | |
0037b49e | 216 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
217 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { |
218 | if (zfs_vscan(ip, cr, 0) != 0) { | |
0037b49e | 219 | ZFS_EXIT(zfsvfs); |
2e528b49 | 220 | return (SET_ERROR(EACCES)); |
126400a1 BB |
221 | } |
222 | } | |
223 | ||
224 | /* Keep a count of the synchronous opens in the znode */ | |
225 | if (flag & O_SYNC) | |
226 | atomic_inc_32(&zp->z_sync_cnt); | |
227 | ||
0037b49e | 228 | ZFS_EXIT(zfsvfs); |
126400a1 BB |
229 | return (0); |
230 | } | |
126400a1 BB |
231 | |
232 | /* ARGSUSED */ | |
233 | int | |
234 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
235 | { | |
236 | znode_t *zp = ITOZ(ip); | |
0037b49e | 237 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 238 | |
0037b49e | 239 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
240 | ZFS_VERIFY_ZP(zp); |
241 | ||
7dc71949 | 242 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 243 | if (flag & O_SYNC) |
7dc71949 | 244 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 | 245 | |
0037b49e | 246 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
247 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) |
248 | VERIFY(zfs_vscan(ip, cr, 1) == 0); | |
249 | ||
0037b49e | 250 | ZFS_EXIT(zfsvfs); |
8780c539 | 251 | return (0); |
126400a1 | 252 | } |
126400a1 | 253 | |
802e7b5f | 254 | #if defined(SEEK_HOLE) && defined(SEEK_DATA) |
cf91b2b6 | 255 | /* |
802e7b5f LD |
256 | * Lseek support for finding holes (cmd == SEEK_HOLE) and |
257 | * data (cmd == SEEK_DATA). "off" is an in/out parameter. | |
cf91b2b6 MA |
258 | */ |
259 | static int | |
802e7b5f | 260 | zfs_holey_common(struct inode *ip, int cmd, loff_t *off) |
cf91b2b6 | 261 | { |
802e7b5f | 262 | znode_t *zp = ITOZ(ip); |
cf91b2b6 MA |
263 | uint64_t noff = (uint64_t)*off; /* new offset */ |
264 | uint64_t file_sz; | |
265 | int error; | |
266 | boolean_t hole; | |
267 | ||
268 | file_sz = zp->z_size; | |
269 | if (noff >= file_sz) { | |
2e528b49 | 270 | return (SET_ERROR(ENXIO)); |
cf91b2b6 MA |
271 | } |
272 | ||
802e7b5f | 273 | if (cmd == SEEK_HOLE) |
cf91b2b6 MA |
274 | hole = B_TRUE; |
275 | else | |
276 | hole = B_FALSE; | |
277 | ||
802e7b5f | 278 | error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff); |
cf91b2b6 | 279 | |
d97aa48f | 280 | if (error == ESRCH) |
2e528b49 | 281 | return (SET_ERROR(ENXIO)); |
d97aa48f | 282 | |
6e03ec4f DB |
283 | /* file was dirty, so fall back to using generic logic */ |
284 | if (error == EBUSY) { | |
285 | if (hole) | |
286 | *off = file_sz; | |
287 | ||
288 | return (0); | |
289 | } | |
66aca247 | 290 | |
d97aa48f MA |
291 | /* |
292 | * We could find a hole that begins after the logical end-of-file, | |
293 | * because dmu_offset_next() only works on whole blocks. If the | |
294 | * EOF falls mid-block, then indicate that the "virtual hole" | |
295 | * at the end of the file begins at the logical EOF, rather than | |
296 | * at the end of the last block. | |
297 | */ | |
298 | if (noff > file_sz) { | |
299 | ASSERT(hole); | |
300 | noff = file_sz; | |
cf91b2b6 MA |
301 | } |
302 | ||
303 | if (noff < *off) | |
304 | return (error); | |
305 | *off = noff; | |
306 | return (error); | |
307 | } | |
802e7b5f LD |
308 | |
309 | int | |
310 | zfs_holey(struct inode *ip, int cmd, loff_t *off) | |
311 | { | |
312 | znode_t *zp = ITOZ(ip); | |
0037b49e | 313 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
802e7b5f LD |
314 | int error; |
315 | ||
0037b49e | 316 | ZFS_ENTER(zfsvfs); |
802e7b5f LD |
317 | ZFS_VERIFY_ZP(zp); |
318 | ||
319 | error = zfs_holey_common(ip, cmd, off); | |
320 | ||
0037b49e | 321 | ZFS_EXIT(zfsvfs); |
802e7b5f LD |
322 | return (error); |
323 | } | |
802e7b5f | 324 | #endif /* SEEK_HOLE && SEEK_DATA */ |
cf91b2b6 | 325 | |
c0d35759 | 326 | #if defined(_KERNEL) |
34dc7c2f BB |
327 | /* |
328 | * When a file is memory mapped, we must keep the IO data synchronized | |
329 | * between the DMU cache and the memory mapped pages. What this means: | |
330 | * | |
331 | * On Write: If we find a memory mapped page, we write to *both* | |
332 | * the page and the dmu buffer. | |
34dc7c2f | 333 | */ |
d164b209 | 334 | static void |
c0d35759 BB |
335 | update_pages(struct inode *ip, int64_t start, int len, |
336 | objset_t *os, uint64_t oid) | |
34dc7c2f | 337 | { |
c0d35759 BB |
338 | struct address_space *mp = ip->i_mapping; |
339 | struct page *pp; | |
340 | uint64_t nbytes; | |
d164b209 | 341 | int64_t off; |
c0d35759 | 342 | void *pb; |
34dc7c2f | 343 | |
8b1899d3 BB |
344 | off = start & (PAGE_SIZE-1); |
345 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
346 | nbytes = MIN(PAGE_SIZE - off, len); | |
34dc7c2f | 347 | |
8b1899d3 | 348 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
349 | if (pp) { |
350 | if (mapping_writably_mapped(mp)) | |
351 | flush_dcache_page(pp); | |
34dc7c2f | 352 | |
c0d35759 BB |
353 | pb = kmap(pp); |
354 | (void) dmu_read(os, oid, start+off, nbytes, pb+off, | |
9babb374 | 355 | DMU_READ_PREFETCH); |
c0d35759 BB |
356 | kunmap(pp); |
357 | ||
358 | if (mapping_writably_mapped(mp)) | |
359 | flush_dcache_page(pp); | |
360 | ||
361 | mark_page_accessed(pp); | |
362 | SetPageUptodate(pp); | |
363 | ClearPageError(pp); | |
364 | unlock_page(pp); | |
8b1899d3 | 365 | put_page(pp); |
34dc7c2f | 366 | } |
c0d35759 | 367 | |
d164b209 | 368 | len -= nbytes; |
34dc7c2f | 369 | off = 0; |
34dc7c2f | 370 | } |
34dc7c2f BB |
371 | } |
372 | ||
373 | /* | |
374 | * When a file is memory mapped, we must keep the IO data synchronized | |
375 | * between the DMU cache and the memory mapped pages. What this means: | |
376 | * | |
377 | * On Read: We "read" preferentially from memory mapped pages, | |
378 | * else we default from the dmu buffer. | |
379 | * | |
380 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
d3cc8b15 | 381 | * the file is memory mapped. |
34dc7c2f BB |
382 | */ |
383 | static int | |
3558fd73 | 384 | mappedread(struct inode *ip, int nbytes, uio_t *uio) |
34dc7c2f | 385 | { |
c0d35759 BB |
386 | struct address_space *mp = ip->i_mapping; |
387 | struct page *pp; | |
3558fd73 | 388 | znode_t *zp = ITOZ(ip); |
34dc7c2f | 389 | int64_t start, off; |
c0d35759 | 390 | uint64_t bytes; |
34dc7c2f BB |
391 | int len = nbytes; |
392 | int error = 0; | |
c0d35759 | 393 | void *pb; |
34dc7c2f BB |
394 | |
395 | start = uio->uio_loffset; | |
8b1899d3 BB |
396 | off = start & (PAGE_SIZE-1); |
397 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
398 | bytes = MIN(PAGE_SIZE - off, len); | |
c0d35759 | 399 | |
8b1899d3 | 400 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
401 | if (pp) { |
402 | ASSERT(PageUptodate(pp)); | |
403 | ||
404 | pb = kmap(pp); | |
405 | error = uiomove(pb + off, bytes, UIO_READ, uio); | |
406 | kunmap(pp); | |
407 | ||
408 | if (mapping_writably_mapped(mp)) | |
409 | flush_dcache_page(pp); | |
410 | ||
411 | mark_page_accessed(pp); | |
412 | unlock_page(pp); | |
8b1899d3 | 413 | put_page(pp); |
34dc7c2f | 414 | } else { |
804e0504 MA |
415 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
416 | uio, bytes); | |
34dc7c2f | 417 | } |
c0d35759 | 418 | |
34dc7c2f BB |
419 | len -= bytes; |
420 | off = 0; | |
421 | if (error) | |
422 | break; | |
423 | } | |
424 | return (error); | |
425 | } | |
c0d35759 | 426 | #endif /* _KERNEL */ |
34dc7c2f | 427 | |
c409e464 | 428 | unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ |
a966c564 | 429 | unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT; |
34dc7c2f BB |
430 | |
431 | /* | |
432 | * Read bytes from specified file into supplied buffer. | |
433 | * | |
3558fd73 | 434 | * IN: ip - inode of file to be read from. |
34dc7c2f BB |
435 | * uio - structure supplying read location, range info, |
436 | * and return buffer. | |
c0d35759 BB |
437 | * ioflag - FSYNC flags; used to provide FRSYNC semantics. |
438 | * O_DIRECT flag; used to bypass page cache. | |
34dc7c2f | 439 | * cr - credentials of caller. |
34dc7c2f BB |
440 | * |
441 | * OUT: uio - updated offset and range, buffer filled. | |
442 | * | |
d3cc8b15 | 443 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
444 | * |
445 | * Side Effects: | |
3558fd73 | 446 | * inode - atime updated if byte count > 0 |
34dc7c2f BB |
447 | */ |
448 | /* ARGSUSED */ | |
e5c39b95 | 449 | int |
3558fd73 | 450 | zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 451 | { |
3558fd73 | 452 | znode_t *zp = ITOZ(ip); |
0037b49e | 453 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 454 | ssize_t n, nbytes; |
149e873a | 455 | int error = 0; |
34dc7c2f | 456 | rl_t *rl; |
3558fd73 | 457 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 458 | xuio_t *xuio = NULL; |
3558fd73 | 459 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f | 460 | |
0037b49e | 461 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 462 | ZFS_VERIFY_ZP(zp); |
34dc7c2f | 463 | |
428870ff | 464 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
0037b49e | 465 | ZFS_EXIT(zfsvfs); |
2e528b49 | 466 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
467 | } |
468 | ||
469 | /* | |
470 | * Validate file offset | |
471 | */ | |
472 | if (uio->uio_loffset < (offset_t)0) { | |
0037b49e | 473 | ZFS_EXIT(zfsvfs); |
2e528b49 | 474 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
475 | } |
476 | ||
477 | /* | |
478 | * Fasttrack empty reads | |
479 | */ | |
480 | if (uio->uio_resid == 0) { | |
0037b49e | 481 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
482 | return (0); |
483 | } | |
484 | ||
34dc7c2f BB |
485 | /* |
486 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
37699482 | 487 | * Only do this for non-snapshots. |
34dc7c2f | 488 | */ |
37699482 CC |
489 | if (zfsvfs->z_log && |
490 | (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)) | |
0037b49e | 491 | zil_commit(zfsvfs->z_log, zp->z_id); |
34dc7c2f BB |
492 | |
493 | /* | |
494 | * Lock the range against changes. | |
495 | */ | |
d88895a0 CC |
496 | rl = zfs_range_lock(&zp->z_range_lock, uio->uio_loffset, uio->uio_resid, |
497 | RL_READER); | |
34dc7c2f BB |
498 | |
499 | /* | |
500 | * If we are reading past end-of-file we can skip | |
501 | * to the end; but we might still need to set atime. | |
502 | */ | |
428870ff | 503 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
504 | error = 0; |
505 | goto out; | |
506 | } | |
507 | ||
428870ff BB |
508 | ASSERT(uio->uio_loffset < zp->z_size); |
509 | n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); | |
510 | ||
3558fd73 | 511 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
512 | if ((uio->uio_extflg == UIO_XUIO) && |
513 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
514 | int nblk; | |
515 | int blksz = zp->z_blksz; | |
516 | uint64_t offset = uio->uio_loffset; | |
517 | ||
518 | xuio = (xuio_t *)uio; | |
519 | if ((ISP2(blksz))) { | |
520 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
521 | blksz)) / blksz; | |
522 | } else { | |
523 | ASSERT(offset + n <= blksz); | |
524 | nblk = 1; | |
525 | } | |
526 | (void) dmu_xuio_init(xuio, nblk); | |
527 | ||
3558fd73 | 528 | if (vn_has_cached_data(ip)) { |
428870ff BB |
529 | /* |
530 | * For simplicity, we always allocate a full buffer | |
531 | * even if we only expect to read a portion of a block. | |
532 | */ | |
533 | while (--nblk >= 0) { | |
534 | (void) dmu_xuio_add(xuio, | |
535 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
536 | blksz), 0, blksz); | |
537 | } | |
538 | } | |
539 | } | |
3558fd73 | 540 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
541 | |
542 | while (n > 0) { | |
543 | nbytes = MIN(n, zfs_read_chunk_size - | |
544 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); | |
545 | ||
804e0504 | 546 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
3558fd73 | 547 | error = mappedread(ip, nbytes, uio); |
804e0504 MA |
548 | } else { |
549 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), | |
550 | uio, nbytes); | |
551 | } | |
c0d35759 | 552 | |
b128c09f BB |
553 | if (error) { |
554 | /* convert checksum errors into IO errors */ | |
555 | if (error == ECKSUM) | |
2e528b49 | 556 | error = SET_ERROR(EIO); |
34dc7c2f | 557 | break; |
b128c09f | 558 | } |
34dc7c2f BB |
559 | |
560 | n -= nbytes; | |
561 | } | |
34dc7c2f BB |
562 | out: |
563 | zfs_range_unlock(rl); | |
564 | ||
0037b49e | 565 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
566 | return (error); |
567 | } | |
568 | ||
34dc7c2f BB |
569 | /* |
570 | * Write the bytes to a file. | |
571 | * | |
3558fd73 | 572 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
573 | * uio - structure supplying write location, range info, |
574 | * and data buffer. | |
575 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 576 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 577 | * cr - credentials of caller. |
34dc7c2f BB |
578 | * |
579 | * OUT: uio - updated offset and range. | |
580 | * | |
581 | * RETURN: 0 if success | |
582 | * error code if failure | |
583 | * | |
584 | * Timestamps: | |
3558fd73 | 585 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 586 | */ |
428870ff | 587 | |
34dc7c2f | 588 | /* ARGSUSED */ |
e5c39b95 | 589 | int |
3558fd73 | 590 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 591 | { |
3558fd73 BB |
592 | znode_t *zp = ITOZ(ip); |
593 | rlim64_t limit = uio->uio_limit; | |
34dc7c2f BB |
594 | ssize_t start_resid = uio->uio_resid; |
595 | ssize_t tx_bytes; | |
596 | uint64_t end_size; | |
597 | dmu_tx_t *tx; | |
0037b49e | 598 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
599 | zilog_t *zilog; |
600 | offset_t woff; | |
601 | ssize_t n, nbytes; | |
602 | rl_t *rl; | |
0037b49e | 603 | int max_blksz = zfsvfs->z_max_blksz; |
3558fd73 | 604 | int error = 0; |
9babb374 | 605 | arc_buf_t *abuf; |
5475aada | 606 | const iovec_t *aiov = NULL; |
428870ff | 607 | xuio_t *xuio = NULL; |
428870ff BB |
608 | int write_eof; |
609 | int count = 0; | |
610 | sa_bulk_attr_t bulk[4]; | |
611 | uint64_t mtime[2], ctime[2]; | |
2c6abf15 | 612 | uint32_t uid; |
5a6765cf | 613 | #ifdef HAVE_UIO_ZEROCOPY |
614 | int i_iov = 0; | |
615 | const iovec_t *iovp = uio->uio_iov; | |
3558fd73 | 616 | ASSERTV(int iovcnt = uio->uio_iovcnt); |
5a6765cf | 617 | #endif |
34dc7c2f | 618 | |
34dc7c2f BB |
619 | /* |
620 | * Fasttrack empty write | |
621 | */ | |
622 | n = start_resid; | |
623 | if (n == 0) | |
624 | return (0); | |
625 | ||
626 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) | |
627 | limit = MAXOFFSET_T; | |
628 | ||
0037b49e | 629 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 630 | ZFS_VERIFY_ZP(zp); |
b128c09f | 631 | |
0037b49e BB |
632 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
633 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
634 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
635 | &zp->z_size, 8); | |
636 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
428870ff BB |
637 | &zp->z_pflags, 8); |
638 | ||
f3c9dca0 MT |
639 | /* |
640 | * Callers might not be able to detect properly that we are read-only, | |
641 | * so check it explicitly here. | |
642 | */ | |
0037b49e BB |
643 | if (zfs_is_readonly(zfsvfs)) { |
644 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
645 | return (SET_ERROR(EROFS)); |
646 | } | |
647 | ||
b128c09f BB |
648 | /* |
649 | * If immutable or not appending then return EPERM | |
650 | */ | |
428870ff BB |
651 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
652 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
653 | (uio->uio_loffset < zp->z_size))) { | |
0037b49e | 654 | ZFS_EXIT(zfsvfs); |
2e528b49 | 655 | return (SET_ERROR(EPERM)); |
b128c09f BB |
656 | } |
657 | ||
0037b49e | 658 | zilog = zfsvfs->z_log; |
34dc7c2f | 659 | |
428870ff BB |
660 | /* |
661 | * Validate file offset | |
662 | */ | |
663 | woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; | |
664 | if (woff < 0) { | |
0037b49e | 665 | ZFS_EXIT(zfsvfs); |
2e528b49 | 666 | return (SET_ERROR(EINVAL)); |
428870ff BB |
667 | } |
668 | ||
34dc7c2f BB |
669 | /* |
670 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
671 | * don't hold up txg. | |
428870ff | 672 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 673 | */ |
9cac042c | 674 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
675 | if ((uio->uio_extflg == UIO_XUIO) && |
676 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
677 | xuio = (xuio_t *)uio; | |
678 | else | |
9cac042c | 679 | #endif |
572e2857 | 680 | uio_prefaultpages(MIN(n, max_blksz), uio); |
34dc7c2f BB |
681 | |
682 | /* | |
683 | * If in append mode, set the io offset pointer to eof. | |
684 | */ | |
685 | if (ioflag & FAPPEND) { | |
686 | /* | |
428870ff BB |
687 | * Obtain an appending range lock to guarantee file append |
688 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f | 689 | */ |
d88895a0 | 690 | rl = zfs_range_lock(&zp->z_range_lock, 0, n, RL_APPEND); |
428870ff | 691 | woff = rl->r_off; |
34dc7c2f | 692 | if (rl->r_len == UINT64_MAX) { |
428870ff BB |
693 | /* |
694 | * We overlocked the file because this write will cause | |
695 | * the file block size to increase. | |
696 | * Note that zp_size cannot change with this lock held. | |
697 | */ | |
698 | woff = zp->z_size; | |
34dc7c2f | 699 | } |
428870ff | 700 | uio->uio_loffset = woff; |
34dc7c2f | 701 | } else { |
34dc7c2f | 702 | /* |
428870ff BB |
703 | * Note that if the file block size will change as a result of |
704 | * this write, then this range lock will lock the entire file | |
705 | * so that we can re-write the block safely. | |
34dc7c2f | 706 | */ |
d88895a0 | 707 | rl = zfs_range_lock(&zp->z_range_lock, woff, n, RL_WRITER); |
34dc7c2f BB |
708 | } |
709 | ||
710 | if (woff >= limit) { | |
711 | zfs_range_unlock(rl); | |
0037b49e | 712 | ZFS_EXIT(zfsvfs); |
2e528b49 | 713 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
714 | } |
715 | ||
716 | if ((woff + n) > limit || woff > (limit - n)) | |
717 | n = limit - woff; | |
718 | ||
428870ff BB |
719 | /* Will this write extend the file length? */ |
720 | write_eof = (woff + n > zp->z_size); | |
721 | ||
722 | end_size = MAX(zp->z_size, woff + n); | |
34dc7c2f BB |
723 | |
724 | /* | |
725 | * Write the file in reasonable size chunks. Each chunk is written | |
726 | * in a separate transaction; this keeps the intent log records small | |
727 | * and allows us to do more fine-grained space accounting. | |
728 | */ | |
729 | while (n > 0) { | |
9babb374 BB |
730 | abuf = NULL; |
731 | woff = uio->uio_loffset; | |
9c5167d1 NF |
732 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
733 | KUID_TO_SUID(ip->i_uid)) || | |
734 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
735 | KGID_TO_SGID(ip->i_gid)) || | |
736 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
737 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
738 | zp->z_projid))) { | |
9babb374 BB |
739 | if (abuf != NULL) |
740 | dmu_return_arcbuf(abuf); | |
2e528b49 | 741 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
742 | break; |
743 | } | |
744 | ||
428870ff | 745 | if (xuio && abuf == NULL) { |
5a6765cf | 746 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 747 | ASSERT(i_iov < iovcnt); |
5475aada | 748 | ASSERT3U(uio->uio_segflg, !=, UIO_BVEC); |
428870ff BB |
749 | aiov = &iovp[i_iov]; |
750 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
751 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
752 | ASSERT((aiov->iov_base == abuf->b_data) || |
753 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
754 | aiov->iov_len == arc_buf_size(abuf))); | |
755 | i_iov++; | |
5a6765cf | 756 | #endif |
428870ff BB |
757 | } else if (abuf == NULL && n >= max_blksz && |
758 | woff >= zp->z_size && | |
9babb374 BB |
759 | P2PHASE(woff, max_blksz) == 0 && |
760 | zp->z_blksz == max_blksz) { | |
428870ff BB |
761 | /* |
762 | * This write covers a full block. "Borrow" a buffer | |
763 | * from the dmu so that we can fill it before we enter | |
764 | * a transaction. This avoids the possibility of | |
765 | * holding up the transaction if the data copy hangs | |
766 | * up on a pagefault (e.g., from an NFS server mapping). | |
767 | */ | |
9babb374 BB |
768 | size_t cbytes; |
769 | ||
428870ff BB |
770 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
771 | max_blksz); | |
9babb374 BB |
772 | ASSERT(abuf != NULL); |
773 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
774 | if ((error = uiocopy(abuf->b_data, max_blksz, |
775 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
776 | dmu_return_arcbuf(abuf); |
777 | break; | |
778 | } | |
779 | ASSERT(cbytes == max_blksz); | |
780 | } | |
781 | ||
34dc7c2f BB |
782 | /* |
783 | * Start a transaction. | |
784 | */ | |
0037b49e | 785 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 786 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 787 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 788 | zfs_sa_upgrade_txholds(tx, zp); |
384f8a09 | 789 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 790 | if (error) { |
34dc7c2f | 791 | dmu_tx_abort(tx); |
9babb374 BB |
792 | if (abuf != NULL) |
793 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
794 | break; |
795 | } | |
796 | ||
797 | /* | |
798 | * If zfs_range_lock() over-locked we grow the blocksize | |
799 | * and then reduce the lock range. This will only happen | |
800 | * on the first iteration since zfs_range_reduce() will | |
801 | * shrink down r_len to the appropriate size. | |
802 | */ | |
803 | if (rl->r_len == UINT64_MAX) { | |
804 | uint64_t new_blksz; | |
805 | ||
806 | if (zp->z_blksz > max_blksz) { | |
f1512ee6 MA |
807 | /* |
808 | * File's blocksize is already larger than the | |
809 | * "recordsize" property. Only let it grow to | |
810 | * the next power of 2. | |
811 | */ | |
34dc7c2f | 812 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 MA |
813 | new_blksz = MIN(end_size, |
814 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
815 | } else { |
816 | new_blksz = MIN(end_size, max_blksz); | |
817 | } | |
818 | zfs_grow_blocksize(zp, new_blksz, tx); | |
819 | zfs_range_reduce(rl, woff, n); | |
820 | } | |
821 | ||
822 | /* | |
823 | * XXX - should we really limit each write to z_max_blksz? | |
824 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
825 | */ | |
826 | nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); | |
34dc7c2f | 827 | |
9babb374 BB |
828 | if (abuf == NULL) { |
829 | tx_bytes = uio->uio_resid; | |
428870ff BB |
830 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
831 | uio, nbytes, tx); | |
9babb374 BB |
832 | tx_bytes -= uio->uio_resid; |
833 | } else { | |
834 | tx_bytes = nbytes; | |
428870ff BB |
835 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
836 | /* | |
837 | * If this is not a full block write, but we are | |
838 | * extending the file past EOF and this data starts | |
839 | * block-aligned, use assign_arcbuf(). Otherwise, | |
840 | * write via dmu_write(). | |
841 | */ | |
842 | if (tx_bytes < max_blksz && (!write_eof || | |
843 | aiov->iov_base != abuf->b_data)) { | |
844 | ASSERT(xuio); | |
0037b49e | 845 | dmu_write(zfsvfs->z_os, zp->z_id, woff, |
94183a9d | 846 | /* cppcheck-suppress nullPointer */ |
428870ff BB |
847 | aiov->iov_len, aiov->iov_base, tx); |
848 | dmu_return_arcbuf(abuf); | |
849 | xuio_stat_wbuf_copied(); | |
850 | } else { | |
851 | ASSERT(xuio || tx_bytes == max_blksz); | |
440a3eb9 TC |
852 | dmu_assign_arcbuf_by_dbuf( |
853 | sa_get_db(zp->z_sa_hdl), woff, abuf, tx); | |
428870ff | 854 | } |
9babb374 BB |
855 | ASSERT(tx_bytes <= uio->uio_resid); |
856 | uioskip(uio, tx_bytes); | |
857 | } | |
0037b49e BB |
858 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
859 | update_pages(ip, woff, | |
860 | tx_bytes, zfsvfs->z_os, zp->z_id); | |
861 | } | |
34dc7c2f BB |
862 | |
863 | /* | |
864 | * If we made no progress, we're done. If we made even | |
865 | * partial progress, update the znode and ZIL accordingly. | |
866 | */ | |
867 | if (tx_bytes == 0) { | |
0037b49e | 868 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 869 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
870 | dmu_tx_commit(tx); |
871 | ASSERT(error != 0); | |
872 | break; | |
873 | } | |
874 | ||
875 | /* | |
876 | * Clear Set-UID/Set-GID bits on successful write if not | |
4e33ba4c | 877 | * privileged and at least one of the execute bits is set. |
34dc7c2f BB |
878 | * |
879 | * It would be nice to to this after all writes have | |
880 | * been done, but that would still expose the ISUID/ISGID | |
881 | * to another app after the partial write is committed. | |
882 | * | |
572e2857 BB |
883 | * Note: we don't call zfs_fuid_map_id() here because |
884 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
885 | */ |
886 | mutex_enter(&zp->z_acl_lock); | |
2c6abf15 | 887 | uid = KUID_TO_SUID(ip->i_uid); |
428870ff | 888 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 889 | (S_IXUSR >> 6))) != 0 && |
428870ff | 890 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 891 | secpolicy_vnode_setid_retain(cr, |
2c6abf15 | 892 | ((zp->z_mode & S_ISUID) != 0 && uid == 0)) != 0) { |
428870ff BB |
893 | uint64_t newmode; |
894 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
12fa7f34 | 895 | ip->i_mode = newmode = zp->z_mode; |
0037b49e | 896 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), |
428870ff | 897 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
898 | } |
899 | mutex_exit(&zp->z_acl_lock); | |
900 | ||
0df9673f | 901 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
34dc7c2f BB |
902 | |
903 | /* | |
904 | * Update the file size (zp_size) if it has changed; | |
905 | * account for possible concurrent updates. | |
906 | */ | |
428870ff BB |
907 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
908 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 909 | uio->uio_loffset); |
428870ff BB |
910 | ASSERT(error == 0); |
911 | } | |
572e2857 BB |
912 | /* |
913 | * If we are replaying and eof is non zero then force | |
914 | * the file size to the specified eof. Note, there's no | |
915 | * concurrency during replay. | |
916 | */ | |
0037b49e BB |
917 | if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) |
918 | zp->z_size = zfsvfs->z_replay_eof; | |
572e2857 | 919 | |
428870ff BB |
920 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
921 | ||
119a394a ED |
922 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
923 | NULL, NULL); | |
34dc7c2f BB |
924 | dmu_tx_commit(tx); |
925 | ||
926 | if (error != 0) | |
927 | break; | |
928 | ASSERT(tx_bytes == nbytes); | |
929 | n -= nbytes; | |
572e2857 BB |
930 | |
931 | if (!xuio && n > 0) | |
932 | uio_prefaultpages(MIN(n, max_blksz), uio); | |
34dc7c2f BB |
933 | } |
934 | ||
2a53e2da | 935 | zfs_inode_update(zp); |
34dc7c2f BB |
936 | zfs_range_unlock(rl); |
937 | ||
938 | /* | |
939 | * If we're in replay mode, or we made no progress, return error. | |
940 | * Otherwise, it's at least a partial write, so it's successful. | |
941 | */ | |
0037b49e BB |
942 | if (zfsvfs->z_replay || uio->uio_resid == start_resid) { |
943 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
944 | return (error); |
945 | } | |
946 | ||
428870ff | 947 | if (ioflag & (FSYNC | FDSYNC) || |
0037b49e | 948 | zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 949 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 950 | |
0037b49e | 951 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
952 | return (0); |
953 | } | |
954 | ||
7c502b0b | 955 | /* |
956 | * Drop a reference on the passed inode asynchronously. This ensures | |
957 | * that the caller will never drop the last reference on an inode in | |
958 | * the current context. Doing so while holding open a tx could result | |
959 | * in a deadlock if iput_final() re-enters the filesystem code. | |
960 | */ | |
0a50679c BB |
961 | void |
962 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 963 | { |
0a50679c BB |
964 | objset_t *os = ITOZSB(ip)->z_os; |
965 | ||
3558fd73 | 966 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
967 | ASSERT(os != NULL); |
968 | ||
3558fd73 | 969 | if (atomic_read(&ip->i_count) == 1) |
7c502b0b | 970 | VERIFY(taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
48d3eb40 | 971 | (task_func_t *)iput, ip, TQ_SLEEP) != TASKQID_INVALID); |
3558fd73 BB |
972 | else |
973 | iput(ip); | |
974 | } | |
975 | ||
34dc7c2f | 976 | void |
428870ff | 977 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 978 | { |
428870ff | 979 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
980 | |
981 | if (zgd->zgd_db) | |
982 | dmu_buf_rele(zgd->zgd_db, zgd); | |
983 | ||
984 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f | 985 | |
9babb374 BB |
986 | /* |
987 | * Release the vnode asynchronously as we currently have the | |
988 | * txg stopped from syncing. | |
989 | */ | |
0a50679c | 990 | zfs_iput_async(ZTOI(zp)); |
428870ff BB |
991 | |
992 | if (error == 0 && zgd->zgd_bp) | |
1ce23dca | 993 | zil_lwb_add_block(zgd->zgd_lwb, zgd->zgd_bp); |
428870ff | 994 | |
34dc7c2f BB |
995 | kmem_free(zgd, sizeof (zgd_t)); |
996 | } | |
997 | ||
45d1cae3 BB |
998 | #ifdef DEBUG |
999 | static int zil_fault_io = 0; | |
1000 | #endif | |
1001 | ||
34dc7c2f BB |
1002 | /* |
1003 | * Get data to generate a TX_WRITE intent log record. | |
1004 | */ | |
1005 | int | |
1ce23dca | 1006 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, zio_t *zio) |
34dc7c2f | 1007 | { |
0037b49e BB |
1008 | zfsvfs_t *zfsvfs = arg; |
1009 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 1010 | znode_t *zp; |
428870ff BB |
1011 | uint64_t object = lr->lr_foid; |
1012 | uint64_t offset = lr->lr_offset; | |
1013 | uint64_t size = lr->lr_length; | |
34dc7c2f | 1014 | dmu_buf_t *db; |
34dc7c2f | 1015 | zgd_t *zgd; |
34dc7c2f BB |
1016 | int error = 0; |
1017 | ||
1ce23dca PS |
1018 | ASSERT3P(lwb, !=, NULL); |
1019 | ASSERT3P(zio, !=, NULL); | |
1020 | ASSERT3U(size, !=, 0); | |
34dc7c2f BB |
1021 | |
1022 | /* | |
1023 | * Nothing to do if the file has been removed | |
1024 | */ | |
0037b49e | 1025 | if (zfs_zget(zfsvfs, object, &zp) != 0) |
2e528b49 | 1026 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 1027 | if (zp->z_unlinked) { |
9babb374 BB |
1028 | /* |
1029 | * Release the vnode asynchronously as we currently have the | |
1030 | * txg stopped from syncing. | |
1031 | */ | |
0a50679c | 1032 | zfs_iput_async(ZTOI(zp)); |
2e528b49 | 1033 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1034 | } |
1035 | ||
79c76d5b | 1036 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
1ce23dca | 1037 | zgd->zgd_lwb = lwb; |
428870ff BB |
1038 | zgd->zgd_private = zp; |
1039 | ||
34dc7c2f BB |
1040 | /* |
1041 | * Write records come in two flavors: immediate and indirect. | |
1042 | * For small writes it's cheaper to store the data with the | |
1043 | * log record (immediate); for large writes it's cheaper to | |
1044 | * sync the data and get a pointer to it (indirect) so that | |
1045 | * we don't have to write the data twice. | |
1046 | */ | |
1047 | if (buf != NULL) { /* immediate write */ | |
d88895a0 CC |
1048 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, size, |
1049 | RL_READER); | |
34dc7c2f | 1050 | /* test for truncation needs to be done while range locked */ |
428870ff | 1051 | if (offset >= zp->z_size) { |
2e528b49 | 1052 | error = SET_ERROR(ENOENT); |
428870ff BB |
1053 | } else { |
1054 | error = dmu_read(os, object, offset, size, buf, | |
1055 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1056 | } |
428870ff | 1057 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1058 | } else { /* indirect write */ |
34dc7c2f BB |
1059 | /* |
1060 | * Have to lock the whole block to ensure when it's | |
f763c3d1 | 1061 | * written out and its checksum is being calculated |
34dc7c2f BB |
1062 | * that no one can change the data. We need to re-check |
1063 | * blocksize after we get the lock in case it's changed! | |
1064 | */ | |
1065 | for (;;) { | |
428870ff BB |
1066 | uint64_t blkoff; |
1067 | size = zp->z_blksz; | |
1068 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1069 | offset -= blkoff; | |
d88895a0 CC |
1070 | zgd->zgd_rl = zfs_range_lock(&zp->z_range_lock, offset, |
1071 | size, RL_READER); | |
428870ff | 1072 | if (zp->z_blksz == size) |
34dc7c2f | 1073 | break; |
428870ff BB |
1074 | offset += blkoff; |
1075 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f BB |
1076 | } |
1077 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1078 | if (lr->lr_offset >= zp->z_size) |
2e528b49 | 1079 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1080 | #ifdef DEBUG |
1081 | if (zil_fault_io) { | |
2e528b49 | 1082 | error = SET_ERROR(EIO); |
45d1cae3 | 1083 | zil_fault_io = 0; |
45d1cae3 | 1084 | } |
45d1cae3 | 1085 | #endif |
34dc7c2f | 1086 | if (error == 0) |
428870ff BB |
1087 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1088 | DMU_READ_NO_PREFETCH); | |
1089 | ||
1090 | if (error == 0) { | |
02dc43bc | 1091 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 1092 | |
428870ff BB |
1093 | zgd->zgd_db = db; |
1094 | zgd->zgd_bp = bp; | |
1095 | ||
1096 | ASSERT(db->db_offset == offset); | |
1097 | ASSERT(db->db_size == size); | |
1098 | ||
1099 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1100 | zfs_get_done, zgd); | |
61ca48ff | 1101 | ASSERT(error || lr->lr_length <= size); |
428870ff BB |
1102 | |
1103 | /* | |
1104 | * On success, we need to wait for the write I/O | |
1105 | * initiated by dmu_sync() to complete before we can | |
1106 | * release this dbuf. We will finish everything up | |
1107 | * in the zfs_get_done() callback. | |
1108 | */ | |
1109 | if (error == 0) | |
1110 | return (0); | |
1111 | ||
1112 | if (error == EALREADY) { | |
1113 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
1114 | error = 0; | |
1115 | } | |
1116 | } | |
34dc7c2f | 1117 | } |
428870ff BB |
1118 | |
1119 | zfs_get_done(zgd, error); | |
1120 | ||
34dc7c2f BB |
1121 | return (error); |
1122 | } | |
1123 | ||
1124 | /*ARGSUSED*/ | |
3558fd73 BB |
1125 | int |
1126 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1127 | { |
3558fd73 | 1128 | znode_t *zp = ITOZ(ip); |
0037b49e | 1129 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
1130 | int error; |
1131 | ||
0037b49e | 1132 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1133 | ZFS_VERIFY_ZP(zp); |
1134 | ||
1135 | if (flag & V_ACE_MASK) | |
1136 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1137 | else | |
1138 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1139 | ||
0037b49e | 1140 | ZFS_EXIT(zfsvfs); |
45d1cae3 BB |
1141 | return (error); |
1142 | } | |
45d1cae3 | 1143 | |
34dc7c2f BB |
1144 | /* |
1145 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1146 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1147 | * |
3558fd73 | 1148 | * IN: dip - inode of directory to search. |
34dc7c2f | 1149 | * nm - name of entry to lookup. |
34dc7c2f | 1150 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1151 | * cr - credentials of caller. |
34dc7c2f BB |
1152 | * direntflags - directory lookup flags |
1153 | * realpnp - returned pathname. | |
1154 | * | |
3558fd73 | 1155 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1156 | * |
d3cc8b15 | 1157 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1158 | * |
1159 | * Timestamps: | |
1160 | * NA | |
1161 | */ | |
1162 | /* ARGSUSED */ | |
e5c39b95 | 1163 | int |
3558fd73 BB |
1164 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1165 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1166 | { |
3558fd73 | 1167 | znode_t *zdp = ITOZ(dip); |
0037b49e | 1168 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
3558fd73 | 1169 | int error = 0; |
45d1cae3 | 1170 | |
9b7b9cd3 GM |
1171 | /* |
1172 | * Fast path lookup, however we must skip DNLC lookup | |
1173 | * for case folding or normalizing lookups because the | |
1174 | * DNLC code only stores the passed in name. This means | |
1175 | * creating 'a' and removing 'A' on a case insensitive | |
1176 | * file system would work, but DNLC still thinks 'a' | |
1177 | * exists and won't let you create it again on the next | |
1178 | * pass through fast path. | |
1179 | */ | |
45d1cae3 BB |
1180 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { |
1181 | ||
3558fd73 | 1182 | if (!S_ISDIR(dip->i_mode)) { |
2e528b49 | 1183 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1184 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 1185 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1186 | } |
1187 | ||
1188 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1189 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1190 | if (!error) { | |
3558fd73 BB |
1191 | *ipp = dip; |
1192 | igrab(*ipp); | |
45d1cae3 BB |
1193 | return (0); |
1194 | } | |
1195 | return (error); | |
3558fd73 | 1196 | #ifdef HAVE_DNLC |
9b7b9cd3 GM |
1197 | } else if (!zdp->z_zfsvfs->z_norm && |
1198 | (zdp->z_zfsvfs->z_case == ZFS_CASE_SENSITIVE)) { | |
1199 | ||
45d1cae3 BB |
1200 | vnode_t *tvp = dnlc_lookup(dvp, nm); |
1201 | ||
1202 | if (tvp) { | |
1203 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1204 | if (error) { | |
3558fd73 | 1205 | iput(tvp); |
45d1cae3 BB |
1206 | return (error); |
1207 | } | |
1208 | if (tvp == DNLC_NO_VNODE) { | |
3558fd73 | 1209 | iput(tvp); |
2e528b49 | 1210 | return (SET_ERROR(ENOENT)); |
45d1cae3 BB |
1211 | } else { |
1212 | *vpp = tvp; | |
1213 | return (specvp_check(vpp, cr)); | |
1214 | } | |
1215 | } | |
3558fd73 | 1216 | #endif /* HAVE_DNLC */ |
45d1cae3 BB |
1217 | } |
1218 | } | |
1219 | ||
0037b49e | 1220 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1221 | ZFS_VERIFY_ZP(zdp); |
1222 | ||
3558fd73 | 1223 | *ipp = NULL; |
34dc7c2f BB |
1224 | |
1225 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1226 | /* |
1227 | * We don't allow recursive attributes.. | |
1228 | * Maybe someday we will. | |
1229 | */ | |
428870ff | 1230 | if (zdp->z_pflags & ZFS_XATTR) { |
0037b49e | 1231 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1232 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1233 | } |
1234 | ||
3558fd73 | 1235 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
0037b49e | 1236 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1237 | return (error); |
1238 | } | |
1239 | ||
1240 | /* | |
1241 | * Do we have permission to get into attribute directory? | |
1242 | */ | |
1243 | ||
3558fd73 | 1244 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1245 | B_FALSE, cr))) { |
3558fd73 BB |
1246 | iput(*ipp); |
1247 | *ipp = NULL; | |
34dc7c2f BB |
1248 | } |
1249 | ||
0037b49e | 1250 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1251 | return (error); |
1252 | } | |
1253 | ||
3558fd73 | 1254 | if (!S_ISDIR(dip->i_mode)) { |
0037b49e | 1255 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1256 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1257 | } |
1258 | ||
1259 | /* | |
1260 | * Check accessibility of directory. | |
1261 | */ | |
1262 | ||
149e873a | 1263 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
0037b49e | 1264 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1265 | return (error); |
1266 | } | |
1267 | ||
0037b49e | 1268 | if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1269 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1270 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1271 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1272 | } |
1273 | ||
3558fd73 BB |
1274 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1275 | if ((error == 0) && (*ipp)) | |
1276 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1277 | |
0037b49e | 1278 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1279 | return (error); |
1280 | } | |
1281 | ||
1282 | /* | |
1283 | * Attempt to create a new entry in a directory. If the entry | |
1284 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1285 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1286 | * |
3558fd73 | 1287 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1288 | * name - name of new file entry. |
1289 | * vap - attributes of new file. | |
1290 | * excl - flag indicating exclusive or non-exclusive mode. | |
1291 | * mode - mode to open file with. | |
1292 | * cr - credentials of caller. | |
1293 | * flag - large file flag [UNUSED]. | |
3558fd73 | 1294 | * vsecp - ACL to be set |
34dc7c2f | 1295 | * |
3558fd73 | 1296 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1297 | * |
d3cc8b15 | 1298 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1299 | * |
1300 | * Timestamps: | |
3558fd73 BB |
1301 | * dip - ctime|mtime updated if new entry created |
1302 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1303 | */ |
1304 | ||
1305 | /* ARGSUSED */ | |
e5c39b95 | 1306 | int |
3558fd73 BB |
1307 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1308 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1309 | { |
3558fd73 | 1310 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1311 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1312 | zilog_t *zilog; |
1313 | objset_t *os; | |
1314 | zfs_dirlock_t *dl; | |
1315 | dmu_tx_t *tx; | |
1316 | int error; | |
b128c09f | 1317 | uid_t uid; |
149e873a | 1318 | gid_t gid; |
428870ff | 1319 | zfs_acl_ids_t acl_ids; |
9babb374 | 1320 | boolean_t fuid_dirtied; |
428870ff | 1321 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 1322 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1323 | |
1324 | /* | |
1325 | * If we have an ephemeral id, ACL, or XVATTR then | |
1326 | * make sure file system is at proper version | |
1327 | */ | |
1328 | ||
149e873a | 1329 | gid = crgetgid(cr); |
3558fd73 | 1330 | uid = crgetuid(cr); |
b128c09f | 1331 | |
0037b49e | 1332 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1333 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1334 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1335 | |
32dec7bd | 1336 | if (name == NULL) |
1337 | return (SET_ERROR(EINVAL)); | |
1338 | ||
0037b49e | 1339 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1340 | ZFS_VERIFY_ZP(dzp); |
0037b49e BB |
1341 | os = zfsvfs->z_os; |
1342 | zilog = zfsvfs->z_log; | |
34dc7c2f | 1343 | |
0037b49e | 1344 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1345 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1346 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1347 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1348 | } |
1349 | ||
5484965a | 1350 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1351 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1352 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1353 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1354 | return (error); |
1355 | } | |
1356 | } | |
34dc7c2f | 1357 | |
3558fd73 BB |
1358 | top: |
1359 | *ipp = NULL; | |
34dc7c2f BB |
1360 | if (*name == '\0') { |
1361 | /* | |
1362 | * Null component name refers to the directory itself. | |
1363 | */ | |
3558fd73 | 1364 | igrab(dip); |
34dc7c2f BB |
1365 | zp = dzp; |
1366 | dl = NULL; | |
1367 | error = 0; | |
1368 | } else { | |
3558fd73 | 1369 | /* possible igrab(zp) */ |
34dc7c2f BB |
1370 | int zflg = 0; |
1371 | ||
1372 | if (flag & FIGNORECASE) | |
1373 | zflg |= ZCILOOK; | |
1374 | ||
1375 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1376 | NULL, NULL); | |
1377 | if (error) { | |
572e2857 BB |
1378 | if (have_acl) |
1379 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1380 | if (strcmp(name, "..") == 0) |
2e528b49 | 1381 | error = SET_ERROR(EISDIR); |
0037b49e | 1382 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1383 | return (error); |
1384 | } | |
1385 | } | |
428870ff | 1386 | |
34dc7c2f BB |
1387 | if (zp == NULL) { |
1388 | uint64_t txtype; | |
9c5167d1 | 1389 | uint64_t projid = ZFS_DEFAULT_PROJID; |
34dc7c2f BB |
1390 | |
1391 | /* | |
1392 | * Create a new file object and update the directory | |
1393 | * to reference it. | |
1394 | */ | |
149e873a | 1395 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1396 | if (have_acl) |
1397 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1398 | goto out; |
1399 | } | |
1400 | ||
1401 | /* | |
1402 | * We only support the creation of regular files in | |
1403 | * extended attribute directories. | |
1404 | */ | |
428870ff | 1405 | |
3558fd73 | 1406 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1407 | if (have_acl) |
1408 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 1409 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1410 | goto out; |
1411 | } | |
1412 | ||
428870ff BB |
1413 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1414 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1415 | goto out; |
428870ff BB |
1416 | have_acl = B_TRUE; |
1417 | ||
9c5167d1 NF |
1418 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
1419 | projid = zfs_inherit_projid(dzp); | |
1420 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
45d1cae3 | 1421 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1422 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1423 | goto out; |
1424 | } | |
1425 | ||
34dc7c2f | 1426 | tx = dmu_tx_create(os); |
428870ff BB |
1427 | |
1428 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1429 | ZFS_SA_BASE_ATTR_SIZE); | |
1430 | ||
0037b49e | 1431 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1432 | if (fuid_dirtied) |
0037b49e | 1433 | zfs_fuid_txhold(zfsvfs, tx); |
34dc7c2f | 1434 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1435 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
0037b49e | 1436 | if (!zfsvfs->z_use_sa && |
428870ff | 1437 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1438 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1439 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1440 | } |
cc63068e | 1441 | |
0735ecb3 PS |
1442 | error = dmu_tx_assign(tx, |
1443 | (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); | |
34dc7c2f BB |
1444 | if (error) { |
1445 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1446 | if (error == ERESTART) { |
e8b96c60 | 1447 | waited = B_TRUE; |
34dc7c2f BB |
1448 | dmu_tx_wait(tx); |
1449 | dmu_tx_abort(tx); | |
1450 | goto top; | |
1451 | } | |
428870ff | 1452 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1453 | dmu_tx_abort(tx); |
0037b49e | 1454 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1455 | return (error); |
1456 | } | |
428870ff | 1457 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 1458 | |
cc63068e SB |
1459 | error = zfs_link_create(dl, zp, tx, ZNEW); |
1460 | if (error != 0) { | |
1461 | /* | |
1462 | * Since, we failed to add the directory entry for it, | |
1463 | * delete the newly created dnode. | |
1464 | */ | |
1465 | zfs_znode_delete(zp, tx); | |
1466 | remove_inode_hash(ZTOI(zp)); | |
1467 | zfs_acl_ids_free(&acl_ids); | |
1468 | dmu_tx_commit(tx); | |
1469 | goto out; | |
1470 | } | |
1471 | ||
9babb374 | 1472 | if (fuid_dirtied) |
0037b49e | 1473 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 1474 | |
34dc7c2f BB |
1475 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); |
1476 | if (flag & FIGNORECASE) | |
1477 | txtype |= TX_CI; | |
1478 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1479 | vsecp, acl_ids.z_fuidp, vap); |
1480 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1481 | dmu_tx_commit(tx); |
1482 | } else { | |
1483 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1484 | ||
572e2857 BB |
1485 | if (have_acl) |
1486 | zfs_acl_ids_free(&acl_ids); | |
1487 | have_acl = B_FALSE; | |
1488 | ||
34dc7c2f BB |
1489 | /* |
1490 | * A directory entry already exists for this name. | |
1491 | */ | |
1492 | /* | |
1493 | * Can't truncate an existing file if in exclusive mode. | |
1494 | */ | |
3558fd73 | 1495 | if (excl) { |
2e528b49 | 1496 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1497 | goto out; |
1498 | } | |
1499 | /* | |
1500 | * Can't open a directory for writing. | |
1501 | */ | |
3558fd73 | 1502 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1503 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1504 | goto out; |
1505 | } | |
1506 | /* | |
1507 | * Verify requested access to file. | |
1508 | */ | |
1509 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1510 | goto out; | |
1511 | } | |
1512 | ||
1513 | mutex_enter(&dzp->z_lock); | |
1514 | dzp->z_seq++; | |
1515 | mutex_exit(&dzp->z_lock); | |
1516 | ||
1517 | /* | |
1518 | * Truncate regular files if requested. | |
1519 | */ | |
3558fd73 BB |
1520 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1521 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 1522 | /* we can't hold any locks when calling zfs_freesp() */ |
609603a5 B |
1523 | if (dl) { |
1524 | zfs_dirent_unlock(dl); | |
1525 | dl = NULL; | |
1526 | } | |
34dc7c2f | 1527 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1528 | } |
1529 | } | |
1530 | out: | |
1531 | ||
1532 | if (dl) | |
1533 | zfs_dirent_unlock(dl); | |
1534 | ||
1535 | if (error) { | |
1536 | if (zp) | |
3558fd73 | 1537 | iput(ZTOI(zp)); |
34dc7c2f | 1538 | } else { |
960e08fe BB |
1539 | zfs_inode_update(dzp); |
1540 | zfs_inode_update(zp); | |
3558fd73 | 1541 | *ipp = ZTOI(zp); |
34dc7c2f | 1542 | } |
34dc7c2f | 1543 | |
0037b49e | 1544 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1545 | zil_commit(zilog, 0); |
428870ff | 1546 | |
0037b49e | 1547 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1548 | return (error); |
1549 | } | |
1550 | ||
ace1eae8 CC |
1551 | /* ARGSUSED */ |
1552 | int | |
1553 | zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, | |
1554 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
1555 | { | |
1556 | znode_t *zp = NULL, *dzp = ITOZ(dip); | |
0037b49e | 1557 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
ace1eae8 CC |
1558 | objset_t *os; |
1559 | dmu_tx_t *tx; | |
1560 | int error; | |
1561 | uid_t uid; | |
1562 | gid_t gid; | |
1563 | zfs_acl_ids_t acl_ids; | |
9c5167d1 | 1564 | uint64_t projid = ZFS_DEFAULT_PROJID; |
ace1eae8 CC |
1565 | boolean_t fuid_dirtied; |
1566 | boolean_t have_acl = B_FALSE; | |
1567 | boolean_t waited = B_FALSE; | |
1568 | ||
1569 | /* | |
1570 | * If we have an ephemeral id, ACL, or XVATTR then | |
1571 | * make sure file system is at proper version | |
1572 | */ | |
1573 | ||
1574 | gid = crgetgid(cr); | |
1575 | uid = crgetuid(cr); | |
1576 | ||
0037b49e | 1577 | if (zfsvfs->z_use_fuids == B_FALSE && |
ace1eae8 CC |
1578 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
1579 | return (SET_ERROR(EINVAL)); | |
1580 | ||
0037b49e | 1581 | ZFS_ENTER(zfsvfs); |
ace1eae8 | 1582 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1583 | os = zfsvfs->z_os; |
ace1eae8 CC |
1584 | |
1585 | if (vap->va_mask & ATTR_XVATTR) { | |
1586 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
1587 | crgetuid(cr), cr, vap->va_mode)) != 0) { | |
0037b49e | 1588 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1589 | return (error); |
1590 | } | |
1591 | } | |
1592 | ||
1593 | top: | |
1594 | *ipp = NULL; | |
1595 | ||
1596 | /* | |
1597 | * Create a new file object and update the directory | |
1598 | * to reference it. | |
1599 | */ | |
1600 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { | |
1601 | if (have_acl) | |
1602 | zfs_acl_ids_free(&acl_ids); | |
1603 | goto out; | |
1604 | } | |
1605 | ||
1606 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, | |
1607 | cr, vsecp, &acl_ids)) != 0) | |
1608 | goto out; | |
1609 | have_acl = B_TRUE; | |
1610 | ||
9c5167d1 NF |
1611 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
1612 | projid = zfs_inherit_projid(dzp); | |
1613 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
ace1eae8 CC |
1614 | zfs_acl_ids_free(&acl_ids); |
1615 | error = SET_ERROR(EDQUOT); | |
1616 | goto out; | |
1617 | } | |
1618 | ||
1619 | tx = dmu_tx_create(os); | |
1620 | ||
1621 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1622 | ZFS_SA_BASE_ATTR_SIZE); | |
0037b49e | 1623 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 1624 | |
0037b49e | 1625 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
ace1eae8 | 1626 | if (fuid_dirtied) |
0037b49e BB |
1627 | zfs_fuid_txhold(zfsvfs, tx); |
1628 | if (!zfsvfs->z_use_sa && | |
ace1eae8 CC |
1629 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
1630 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
1631 | 0, acl_ids.z_aclp->z_acl_bytes); | |
1632 | } | |
0735ecb3 | 1633 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
ace1eae8 CC |
1634 | if (error) { |
1635 | if (error == ERESTART) { | |
1636 | waited = B_TRUE; | |
1637 | dmu_tx_wait(tx); | |
1638 | dmu_tx_abort(tx); | |
1639 | goto top; | |
1640 | } | |
1641 | zfs_acl_ids_free(&acl_ids); | |
1642 | dmu_tx_abort(tx); | |
0037b49e | 1643 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1644 | return (error); |
1645 | } | |
1646 | zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); | |
1647 | ||
1648 | if (fuid_dirtied) | |
0037b49e | 1649 | zfs_fuid_sync(zfsvfs, tx); |
ace1eae8 CC |
1650 | |
1651 | /* Add to unlinked set */ | |
1652 | zp->z_unlinked = 1; | |
1653 | zfs_unlinked_add(zp, tx); | |
1654 | zfs_acl_ids_free(&acl_ids); | |
1655 | dmu_tx_commit(tx); | |
1656 | out: | |
1657 | ||
1658 | if (error) { | |
1659 | if (zp) | |
1660 | iput(ZTOI(zp)); | |
1661 | } else { | |
1662 | zfs_inode_update(dzp); | |
1663 | zfs_inode_update(zp); | |
1664 | *ipp = ZTOI(zp); | |
1665 | } | |
1666 | ||
0037b49e | 1667 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1668 | return (error); |
1669 | } | |
1670 | ||
34dc7c2f BB |
1671 | /* |
1672 | * Remove an entry from a directory. | |
1673 | * | |
3558fd73 | 1674 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1675 | * name - name of entry to remove. |
1676 | * cr - credentials of caller. | |
34dc7c2f BB |
1677 | * |
1678 | * RETURN: 0 if success | |
1679 | * error code if failure | |
1680 | * | |
1681 | * Timestamps: | |
3558fd73 BB |
1682 | * dip - ctime|mtime |
1683 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1684 | */ |
428870ff BB |
1685 | |
1686 | uint64_t null_xattr = 0; | |
1687 | ||
34dc7c2f | 1688 | /*ARGSUSED*/ |
e5c39b95 | 1689 | int |
da5e151f | 1690 | zfs_remove(struct inode *dip, char *name, cred_t *cr, int flags) |
34dc7c2f | 1691 | { |
3558fd73 | 1692 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1693 | znode_t *xzp; |
3558fd73 | 1694 | struct inode *ip; |
0037b49e | 1695 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 1696 | zilog_t *zilog; |
a966c564 | 1697 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 1698 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 1699 | uint64_t obj = 0; |
dfbc8630 | 1700 | uint64_t links; |
34dc7c2f BB |
1701 | zfs_dirlock_t *dl; |
1702 | dmu_tx_t *tx; | |
a966c564 K |
1703 | boolean_t may_delete_now, delete_now = FALSE; |
1704 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
1705 | uint64_t txtype; |
1706 | pathname_t *realnmp = NULL; | |
1707 | pathname_t realnm; | |
1708 | int error; | |
1709 | int zflg = ZEXISTS; | |
e8b96c60 | 1710 | boolean_t waited = B_FALSE; |
34dc7c2f | 1711 | |
32dec7bd | 1712 | if (name == NULL) |
1713 | return (SET_ERROR(EINVAL)); | |
1714 | ||
0037b49e | 1715 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1716 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1717 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1718 | |
1719 | if (flags & FIGNORECASE) { | |
1720 | zflg |= ZCILOOK; | |
1721 | pn_alloc(&realnm); | |
1722 | realnmp = &realnm; | |
1723 | } | |
1724 | ||
1725 | top: | |
572e2857 BB |
1726 | xattr_obj = 0; |
1727 | xzp = NULL; | |
34dc7c2f BB |
1728 | /* |
1729 | * Attempt to lock directory; fail if entry doesn't exist. | |
1730 | */ | |
149e873a BB |
1731 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1732 | NULL, realnmp))) { | |
34dc7c2f BB |
1733 | if (realnmp) |
1734 | pn_free(realnmp); | |
0037b49e | 1735 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1736 | return (error); |
1737 | } | |
1738 | ||
3558fd73 | 1739 | ip = ZTOI(zp); |
34dc7c2f | 1740 | |
149e873a | 1741 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1742 | goto out; |
1743 | } | |
1744 | ||
1745 | /* | |
1746 | * Need to use rmdir for removing directories. | |
1747 | */ | |
3558fd73 | 1748 | if (S_ISDIR(ip->i_mode)) { |
2e528b49 | 1749 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1750 | goto out; |
1751 | } | |
1752 | ||
3558fd73 | 1753 | #ifdef HAVE_DNLC |
34dc7c2f BB |
1754 | if (realnmp) |
1755 | dnlc_remove(dvp, realnmp->pn_buf); | |
1756 | else | |
1757 | dnlc_remove(dvp, name); | |
3558fd73 | 1758 | #endif /* HAVE_DNLC */ |
34dc7c2f | 1759 | |
19d55079 MA |
1760 | mutex_enter(&zp->z_lock); |
1761 | may_delete_now = atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped); | |
1762 | mutex_exit(&zp->z_lock); | |
1763 | ||
34dc7c2f | 1764 | /* |
a966c564 K |
1765 | * We may delete the znode now, or we may put it in the unlinked set; |
1766 | * it depends on whether we're the last link, and on whether there are | |
1767 | * other holds on the inode. So we dmu_tx_hold() the right things to | |
1768 | * allow for either case. | |
34dc7c2f | 1769 | */ |
572e2857 | 1770 | obj = zp->z_id; |
0037b49e | 1771 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 1772 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1773 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1774 | zfs_sa_upgrade_txholds(tx, zp); | |
1775 | zfs_sa_upgrade_txholds(tx, dzp); | |
a966c564 K |
1776 | if (may_delete_now) { |
1777 | toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; | |
1778 | /* if the file is too big, only hold_free a token amount */ | |
1779 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1780 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1781 | } | |
34dc7c2f BB |
1782 | |
1783 | /* are there any extended attributes? */ | |
0037b49e | 1784 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1785 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1786 | if (error == 0 && xattr_obj) { |
0037b49e | 1787 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
c99c9001 | 1788 | ASSERT0(error); |
428870ff BB |
1789 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1790 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1791 | } |
1792 | ||
a966c564 K |
1793 | mutex_enter(&zp->z_lock); |
1794 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
1795 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
1796 | mutex_exit(&zp->z_lock); | |
1797 | ||
34dc7c2f | 1798 | /* charge as an update -- would be nice not to charge at all */ |
0037b49e | 1799 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1800 | |
19d55079 | 1801 | /* |
1a04bab3 | 1802 | * Mark this transaction as typically resulting in a net free of space |
19d55079 | 1803 | */ |
1a04bab3 | 1804 | dmu_tx_mark_netfree(tx); |
19d55079 | 1805 | |
0735ecb3 | 1806 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
1807 | if (error) { |
1808 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1809 | if (error == ERESTART) { |
e8b96c60 | 1810 | waited = B_TRUE; |
34dc7c2f BB |
1811 | dmu_tx_wait(tx); |
1812 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
1813 | iput(ip); |
1814 | if (xzp) | |
1815 | iput(ZTOI(xzp)); | |
34dc7c2f BB |
1816 | goto top; |
1817 | } | |
1818 | if (realnmp) | |
1819 | pn_free(realnmp); | |
1820 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
1821 | iput(ip); |
1822 | if (xzp) | |
1823 | iput(ZTOI(xzp)); | |
0037b49e | 1824 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1825 | return (error); |
1826 | } | |
1827 | ||
1828 | /* | |
1829 | * Remove the directory entry. | |
1830 | */ | |
1831 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1832 | ||
1833 | if (error) { | |
1834 | dmu_tx_commit(tx); | |
1835 | goto out; | |
1836 | } | |
1837 | ||
1838 | if (unlinked) { | |
572e2857 BB |
1839 | /* |
1840 | * Hold z_lock so that we can make sure that the ACL obj | |
1841 | * hasn't changed. Could have been deleted due to | |
1842 | * zfs_sa_upgrade(). | |
1843 | */ | |
1844 | mutex_enter(&zp->z_lock); | |
0037b49e | 1845 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1846 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
a966c564 K |
1847 | delete_now = may_delete_now && !toobig && |
1848 | atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped) && | |
1849 | xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == | |
1850 | acl_obj; | |
1851 | } | |
1852 | ||
1853 | if (delete_now) { | |
1854 | if (xattr_obj_unlinked) { | |
dfbc8630 | 1855 | ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); |
a966c564 K |
1856 | mutex_enter(&xzp->z_lock); |
1857 | xzp->z_unlinked = 1; | |
dfbc8630 CD |
1858 | clear_nlink(ZTOI(xzp)); |
1859 | links = 0; | |
0037b49e | 1860 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 1861 | &links, sizeof (links), tx); |
a966c564 K |
1862 | ASSERT3U(error, ==, 0); |
1863 | mutex_exit(&xzp->z_lock); | |
1864 | zfs_unlinked_add(xzp, tx); | |
1865 | ||
1866 | if (zp->z_is_sa) | |
1867 | error = sa_remove(zp->z_sa_hdl, | |
0037b49e | 1868 | SA_ZPL_XATTR(zfsvfs), tx); |
a966c564 K |
1869 | else |
1870 | error = sa_update(zp->z_sa_hdl, | |
0037b49e | 1871 | SA_ZPL_XATTR(zfsvfs), &null_xattr, |
a966c564 K |
1872 | sizeof (uint64_t), tx); |
1873 | ASSERT0(error); | |
1874 | } | |
1875 | /* | |
1876 | * Add to the unlinked set because a new reference could be | |
1877 | * taken concurrently resulting in a deferred destruction. | |
1878 | */ | |
1879 | zfs_unlinked_add(zp, tx); | |
1880 | mutex_exit(&zp->z_lock); | |
a966c564 | 1881 | } else if (unlinked) { |
572e2857 | 1882 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1883 | zfs_unlinked_add(zp, tx); |
1884 | } | |
1885 | ||
1886 | txtype = TX_REMOVE; | |
1887 | if (flags & FIGNORECASE) | |
1888 | txtype |= TX_CI; | |
572e2857 | 1889 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1890 | |
1891 | dmu_tx_commit(tx); | |
1892 | out: | |
1893 | if (realnmp) | |
1894 | pn_free(realnmp); | |
1895 | ||
1896 | zfs_dirent_unlock(dl); | |
960e08fe | 1897 | zfs_inode_update(dzp); |
ea7e86d8 | 1898 | zfs_inode_update(zp); |
34dc7c2f | 1899 | |
ea7e86d8 BB |
1900 | if (delete_now) |
1901 | iput(ip); | |
1902 | else | |
a966c564 | 1903 | zfs_iput_async(ip); |
a966c564 K |
1904 | |
1905 | if (xzp) { | |
1906 | zfs_inode_update(xzp); | |
1907 | zfs_iput_async(ZTOI(xzp)); | |
1908 | } | |
428870ff | 1909 | |
0037b49e | 1910 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1911 | zil_commit(zilog, 0); |
34dc7c2f | 1912 | |
0037b49e | 1913 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1914 | return (error); |
1915 | } | |
1916 | ||
1917 | /* | |
3558fd73 | 1918 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1919 | * provided. Return a pointer to the inserted directory. |
1920 | * | |
3558fd73 | 1921 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1922 | * dirname - name of new directory. |
1923 | * vap - attributes of new directory. | |
1924 | * cr - credentials of caller. | |
34dc7c2f BB |
1925 | * vsecp - ACL to be set |
1926 | * | |
3558fd73 | 1927 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1928 | * |
1929 | * RETURN: 0 if success | |
1930 | * error code if failure | |
1931 | * | |
1932 | * Timestamps: | |
3558fd73 BB |
1933 | * dip - ctime|mtime updated |
1934 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1935 | */ |
1936 | /*ARGSUSED*/ | |
e5c39b95 | 1937 | int |
3558fd73 BB |
1938 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1939 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1940 | { |
3558fd73 | 1941 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1942 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1943 | zilog_t *zilog; |
1944 | zfs_dirlock_t *dl; | |
1945 | uint64_t txtype; | |
1946 | dmu_tx_t *tx; | |
1947 | int error; | |
34dc7c2f | 1948 | int zf = ZNEW; |
b128c09f BB |
1949 | uid_t uid; |
1950 | gid_t gid = crgetgid(cr); | |
428870ff | 1951 | zfs_acl_ids_t acl_ids; |
9babb374 | 1952 | boolean_t fuid_dirtied; |
e8b96c60 | 1953 | boolean_t waited = B_FALSE; |
34dc7c2f | 1954 | |
3558fd73 | 1955 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1956 | |
1957 | /* | |
1958 | * If we have an ephemeral id, ACL, or XVATTR then | |
1959 | * make sure file system is at proper version | |
1960 | */ | |
1961 | ||
3558fd73 | 1962 | uid = crgetuid(cr); |
0037b49e | 1963 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1964 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1965 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1966 | |
32dec7bd | 1967 | if (dirname == NULL) |
1968 | return (SET_ERROR(EINVAL)); | |
1969 | ||
0037b49e | 1970 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1971 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1972 | zilog = zfsvfs->z_log; |
34dc7c2f | 1973 | |
428870ff | 1974 | if (dzp->z_pflags & ZFS_XATTR) { |
0037b49e | 1975 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1976 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1977 | } |
1978 | ||
0037b49e | 1979 | if (zfsvfs->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1980 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1981 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1982 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1983 | } |
1984 | if (flags & FIGNORECASE) | |
1985 | zf |= ZCILOOK; | |
1986 | ||
5484965a | 1987 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1988 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1989 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1990 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1991 | return (error); |
1992 | } | |
428870ff | 1993 | } |
34dc7c2f | 1994 | |
428870ff BB |
1995 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1996 | vsecp, &acl_ids)) != 0) { | |
0037b49e | 1997 | ZFS_EXIT(zfsvfs); |
428870ff BB |
1998 | return (error); |
1999 | } | |
34dc7c2f BB |
2000 | /* |
2001 | * First make sure the new directory doesn't exist. | |
428870ff BB |
2002 | * |
2003 | * Existence is checked first to make sure we don't return | |
2004 | * EACCES instead of EEXIST which can cause some applications | |
2005 | * to fail. | |
34dc7c2f BB |
2006 | */ |
2007 | top: | |
3558fd73 | 2008 | *ipp = NULL; |
34dc7c2f | 2009 | |
149e873a BB |
2010 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
2011 | NULL, NULL))) { | |
428870ff | 2012 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 2013 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2014 | return (error); |
2015 | } | |
2016 | ||
149e873a | 2017 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 2018 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2019 | zfs_dirent_unlock(dl); |
0037b49e | 2020 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2021 | return (error); |
2022 | } | |
2023 | ||
9c5167d1 | 2024 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) { |
45d1cae3 | 2025 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 2026 | zfs_dirent_unlock(dl); |
0037b49e | 2027 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2028 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
2029 | } |
2030 | ||
34dc7c2f BB |
2031 | /* |
2032 | * Add a new entry to the directory. | |
2033 | */ | |
0037b49e | 2034 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f BB |
2035 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
2036 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
0037b49e | 2037 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 2038 | if (fuid_dirtied) |
0037b49e BB |
2039 | zfs_fuid_txhold(zfsvfs, tx); |
2040 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
2041 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
2042 | acl_ids.z_aclp->z_acl_bytes); | |
2043 | } | |
2044 | ||
2045 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
2046 | ZFS_SA_BASE_ATTR_SIZE); | |
2047 | ||
0735ecb3 | 2048 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
2049 | if (error) { |
2050 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2051 | if (error == ERESTART) { |
e8b96c60 | 2052 | waited = B_TRUE; |
34dc7c2f BB |
2053 | dmu_tx_wait(tx); |
2054 | dmu_tx_abort(tx); | |
2055 | goto top; | |
2056 | } | |
428870ff | 2057 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2058 | dmu_tx_abort(tx); |
0037b49e | 2059 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2060 | return (error); |
2061 | } | |
2062 | ||
2063 | /* | |
2064 | * Create new node. | |
2065 | */ | |
428870ff | 2066 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f BB |
2067 | |
2068 | /* | |
2069 | * Now put new name in parent dir. | |
2070 | */ | |
cc63068e SB |
2071 | error = zfs_link_create(dl, zp, tx, ZNEW); |
2072 | if (error != 0) { | |
2073 | zfs_znode_delete(zp, tx); | |
2074 | remove_inode_hash(ZTOI(zp)); | |
2075 | goto out; | |
2076 | } | |
2077 | ||
2078 | if (fuid_dirtied) | |
2079 | zfs_fuid_sync(zfsvfs, tx); | |
34dc7c2f | 2080 | |
3558fd73 | 2081 | *ipp = ZTOI(zp); |
34dc7c2f BB |
2082 | |
2083 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
2084 | if (flags & FIGNORECASE) | |
2085 | txtype |= TX_CI; | |
9babb374 BB |
2086 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
2087 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 2088 | |
cc63068e | 2089 | out: |
9babb374 | 2090 | zfs_acl_ids_free(&acl_ids); |
428870ff | 2091 | |
34dc7c2f BB |
2092 | dmu_tx_commit(tx); |
2093 | ||
2094 | zfs_dirent_unlock(dl); | |
2095 | ||
0037b49e | 2096 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2097 | zil_commit(zilog, 0); |
428870ff | 2098 | |
cc63068e SB |
2099 | if (error != 0) { |
2100 | iput(ZTOI(zp)); | |
2101 | } else { | |
2102 | zfs_inode_update(dzp); | |
2103 | zfs_inode_update(zp); | |
2104 | } | |
0037b49e | 2105 | ZFS_EXIT(zfsvfs); |
cc63068e | 2106 | return (error); |
34dc7c2f BB |
2107 | } |
2108 | ||
2109 | /* | |
2110 | * Remove a directory subdir entry. If the current working | |
2111 | * directory is the same as the subdir to be removed, the | |
2112 | * remove will fail. | |
2113 | * | |
3558fd73 | 2114 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 2115 | * name - name of directory to be removed. |
3558fd73 | 2116 | * cwd - inode of current working directory. |
34dc7c2f | 2117 | * cr - credentials of caller. |
34dc7c2f BB |
2118 | * flags - case flags |
2119 | * | |
d3cc8b15 | 2120 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
2121 | * |
2122 | * Timestamps: | |
3558fd73 | 2123 | * dip - ctime|mtime updated |
34dc7c2f BB |
2124 | */ |
2125 | /*ARGSUSED*/ | |
e5c39b95 | 2126 | int |
3558fd73 BB |
2127 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
2128 | int flags) | |
34dc7c2f | 2129 | { |
3558fd73 | 2130 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 2131 | znode_t *zp; |
3558fd73 | 2132 | struct inode *ip; |
0037b49e | 2133 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
2134 | zilog_t *zilog; |
2135 | zfs_dirlock_t *dl; | |
2136 | dmu_tx_t *tx; | |
2137 | int error; | |
2138 | int zflg = ZEXISTS; | |
e8b96c60 | 2139 | boolean_t waited = B_FALSE; |
34dc7c2f | 2140 | |
32dec7bd | 2141 | if (name == NULL) |
2142 | return (SET_ERROR(EINVAL)); | |
2143 | ||
0037b49e | 2144 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2145 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 2146 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2147 | |
2148 | if (flags & FIGNORECASE) | |
2149 | zflg |= ZCILOOK; | |
2150 | top: | |
2151 | zp = NULL; | |
2152 | ||
2153 | /* | |
2154 | * Attempt to lock directory; fail if entry doesn't exist. | |
2155 | */ | |
149e873a BB |
2156 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
2157 | NULL, NULL))) { | |
0037b49e | 2158 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2159 | return (error); |
2160 | } | |
2161 | ||
3558fd73 | 2162 | ip = ZTOI(zp); |
34dc7c2f | 2163 | |
149e873a | 2164 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
2165 | goto out; |
2166 | } | |
2167 | ||
3558fd73 | 2168 | if (!S_ISDIR(ip->i_mode)) { |
2e528b49 | 2169 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
2170 | goto out; |
2171 | } | |
2172 | ||
3558fd73 | 2173 | if (ip == cwd) { |
2e528b49 | 2174 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
2175 | goto out; |
2176 | } | |
2177 | ||
34dc7c2f | 2178 | /* |
4e33ba4c | 2179 | * Grab a lock on the directory to make sure that no one is |
34dc7c2f BB |
2180 | * trying to add (or lookup) entries while we are removing it. |
2181 | */ | |
2182 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
2183 | ||
2184 | /* | |
2185 | * Grab a lock on the parent pointer to make sure we play well | |
2186 | * with the treewalk and directory rename code. | |
2187 | */ | |
2188 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
2189 | ||
0037b49e | 2190 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 2191 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 2192 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
0037b49e | 2193 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
2194 | zfs_sa_upgrade_txholds(tx, zp); |
2195 | zfs_sa_upgrade_txholds(tx, dzp); | |
db707ad0 | 2196 | dmu_tx_mark_netfree(tx); |
0735ecb3 | 2197 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
2198 | if (error) { |
2199 | rw_exit(&zp->z_parent_lock); | |
2200 | rw_exit(&zp->z_name_lock); | |
2201 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2202 | if (error == ERESTART) { |
e8b96c60 | 2203 | waited = B_TRUE; |
34dc7c2f BB |
2204 | dmu_tx_wait(tx); |
2205 | dmu_tx_abort(tx); | |
ea7e86d8 | 2206 | iput(ip); |
34dc7c2f BB |
2207 | goto top; |
2208 | } | |
2209 | dmu_tx_abort(tx); | |
ea7e86d8 | 2210 | iput(ip); |
0037b49e | 2211 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2212 | return (error); |
2213 | } | |
2214 | ||
2215 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
2216 | ||
2217 | if (error == 0) { | |
2218 | uint64_t txtype = TX_RMDIR; | |
2219 | if (flags & FIGNORECASE) | |
2220 | txtype |= TX_CI; | |
572e2857 | 2221 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
2222 | } |
2223 | ||
2224 | dmu_tx_commit(tx); | |
2225 | ||
2226 | rw_exit(&zp->z_parent_lock); | |
2227 | rw_exit(&zp->z_name_lock); | |
2228 | out: | |
2229 | zfs_dirent_unlock(dl); | |
2230 | ||
59157910 BB |
2231 | zfs_inode_update(dzp); |
2232 | zfs_inode_update(zp); | |
3558fd73 | 2233 | iput(ip); |
34dc7c2f | 2234 | |
0037b49e | 2235 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2236 | zil_commit(zilog, 0); |
428870ff | 2237 | |
0037b49e | 2238 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2239 | return (error); |
2240 | } | |
2241 | ||
2242 | /* | |
2243 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 2244 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 2245 | * |
3558fd73 BB |
2246 | * IN: ip - inode of directory to read. |
2247 | * dirent - buffer for directory entries. | |
34dc7c2f | 2248 | * |
3558fd73 | 2249 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
2250 | * |
2251 | * RETURN: 0 if success | |
2252 | * error code if failure | |
2253 | * | |
2254 | * Timestamps: | |
3558fd73 | 2255 | * ip - atime updated |
34dc7c2f BB |
2256 | * |
2257 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2258 | * This allows us to use the low range for "special" directory entries: | |
2259 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2260 | * we use the offset 2 for the '.zfs' directory. | |
2261 | */ | |
2262 | /* ARGSUSED */ | |
3558fd73 | 2263 | int |
0f37d0c8 | 2264 | zfs_readdir(struct inode *ip, struct dir_context *ctx, cred_t *cr) |
34dc7c2f | 2265 | { |
3558fd73 | 2266 | znode_t *zp = ITOZ(ip); |
0037b49e | 2267 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2268 | objset_t *os; |
34dc7c2f BB |
2269 | zap_cursor_t zc; |
2270 | zap_attribute_t zap; | |
34dc7c2f BB |
2271 | int error; |
2272 | uint8_t prefetch; | |
c12e3a59 | 2273 | uint8_t type; |
3558fd73 BB |
2274 | int done = 0; |
2275 | uint64_t parent; | |
c12e3a59 | 2276 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2277 | |
0037b49e | 2278 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2279 | ZFS_VERIFY_ZP(zp); |
2280 | ||
0037b49e | 2281 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
3558fd73 BB |
2282 | &parent, sizeof (parent))) != 0) |
2283 | goto out; | |
34dc7c2f BB |
2284 | |
2285 | /* | |
2286 | * Quit if directory has been removed (posix) | |
2287 | */ | |
3558fd73 BB |
2288 | if (zp->z_unlinked) |
2289 | goto out; | |
2290 | ||
c12e3a59 | 2291 | error = 0; |
0037b49e | 2292 | os = zfsvfs->z_os; |
c12e3a59 | 2293 | offset = ctx->pos; |
34dc7c2f BB |
2294 | prefetch = zp->z_zn_prefetch; |
2295 | ||
2296 | /* | |
2297 | * Initialize the iterator cursor. | |
2298 | */ | |
c12e3a59 | 2299 | if (offset <= 3) { |
34dc7c2f BB |
2300 | /* |
2301 | * Start iteration from the beginning of the directory. | |
2302 | */ | |
2303 | zap_cursor_init(&zc, os, zp->z_id); | |
2304 | } else { | |
2305 | /* | |
2306 | * The offset is a serialized cursor. | |
2307 | */ | |
c12e3a59 | 2308 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2309 | } |
2310 | ||
34dc7c2f BB |
2311 | /* |
2312 | * Transform to file-system independent format | |
2313 | */ | |
3558fd73 BB |
2314 | while (!done) { |
2315 | uint64_t objnum; | |
34dc7c2f BB |
2316 | /* |
2317 | * Special case `.', `..', and `.zfs'. | |
2318 | */ | |
c12e3a59 | 2319 | if (offset == 0) { |
34dc7c2f BB |
2320 | (void) strcpy(zap.za_name, "."); |
2321 | zap.za_normalization_conflict = 0; | |
2322 | objnum = zp->z_id; | |
c12e3a59 RY |
2323 | type = DT_DIR; |
2324 | } else if (offset == 1) { | |
34dc7c2f BB |
2325 | (void) strcpy(zap.za_name, ".."); |
2326 | zap.za_normalization_conflict = 0; | |
428870ff | 2327 | objnum = parent; |
c12e3a59 RY |
2328 | type = DT_DIR; |
2329 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2330 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2331 | zap.za_normalization_conflict = 0; | |
2332 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 2333 | type = DT_DIR; |
34dc7c2f BB |
2334 | } else { |
2335 | /* | |
2336 | * Grab next entry. | |
2337 | */ | |
3558fd73 BB |
2338 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2339 | if (error == ENOENT) | |
34dc7c2f BB |
2340 | break; |
2341 | else | |
2342 | goto update; | |
2343 | } | |
2344 | ||
0c5dde49 BB |
2345 | /* |
2346 | * Allow multiple entries provided the first entry is | |
2347 | * the object id. Non-zpl consumers may safely make | |
2348 | * use of the additional space. | |
2349 | * | |
2350 | * XXX: This should be a feature flag for compatibility | |
2351 | */ | |
34dc7c2f | 2352 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2353 | zap.za_num_integers == 0) { |
34dc7c2f | 2354 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2355 | "entry, obj = %lld, offset = %lld, " |
2356 | "length = %d, num = %lld\n", | |
34dc7c2f | 2357 | (u_longlong_t)zp->z_id, |
c12e3a59 | 2358 | (u_longlong_t)offset, |
0c5dde49 BB |
2359 | zap.za_integer_length, |
2360 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 2361 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2362 | goto update; |
2363 | } | |
2364 | ||
2365 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 2366 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2367 | } |
0f37d0c8 RY |
2368 | |
2369 | done = !dir_emit(ctx, zap.za_name, strlen(zap.za_name), | |
c12e3a59 | 2370 | objnum, type); |
0f37d0c8 | 2371 | if (done) |
34dc7c2f | 2372 | break; |
34dc7c2f BB |
2373 | |
2374 | /* Prefetch znode */ | |
3558fd73 | 2375 | if (prefetch) { |
fcff0f35 PD |
2376 | dmu_prefetch(os, objnum, 0, 0, 0, |
2377 | ZIO_PRIORITY_SYNC_READ); | |
3558fd73 | 2378 | } |
34dc7c2f | 2379 | |
c12e3a59 RY |
2380 | /* |
2381 | * Move to the next entry, fill in the previous offset. | |
2382 | */ | |
2383 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2384 | zap_cursor_advance(&zc); |
c12e3a59 | 2385 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2386 | } else { |
c12e3a59 | 2387 | offset += 1; |
34dc7c2f | 2388 | } |
c12e3a59 | 2389 | ctx->pos = offset; |
34dc7c2f BB |
2390 | } |
2391 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2392 | ||
34dc7c2f BB |
2393 | update: |
2394 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2395 | if (error == ENOENT) |
2396 | error = 0; | |
3558fd73 | 2397 | out: |
0037b49e | 2398 | ZFS_EXIT(zfsvfs); |
34dc7c2f | 2399 | |
34dc7c2f BB |
2400 | return (error); |
2401 | } | |
2402 | ||
d5446cfc BB |
2403 | ulong_t zfs_fsync_sync_cnt = 4; |
2404 | ||
e5c39b95 | 2405 | int |
3558fd73 | 2406 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2407 | { |
3558fd73 | 2408 | znode_t *zp = ITOZ(ip); |
0037b49e | 2409 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2410 | |
d5446cfc BB |
2411 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2412 | ||
0037b49e BB |
2413 | if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2414 | ZFS_ENTER(zfsvfs); | |
428870ff | 2415 | ZFS_VERIFY_ZP(zp); |
0037b49e BB |
2416 | zil_commit(zfsvfs->z_log, zp->z_id); |
2417 | ZFS_EXIT(zfsvfs); | |
428870ff | 2418 | } |
07012da6 CC |
2419 | tsd_set(zfs_fsyncer_key, NULL); |
2420 | ||
34dc7c2f BB |
2421 | return (0); |
2422 | } | |
2423 | ||
2424 | ||
2425 | /* | |
2426 | * Get the requested file attributes and place them in the provided | |
2427 | * vattr structure. | |
2428 | * | |
3558fd73 | 2429 | * IN: ip - inode of file. |
5484965a BB |
2430 | * vap - va_mask identifies requested attributes. |
2431 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2432 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2433 | * cr - credentials of caller. | |
34dc7c2f | 2434 | * |
5484965a BB |
2435 | * OUT: vap - attribute values. |
2436 | * | |
2437 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2438 | */ |
2439 | /* ARGSUSED */ | |
e5c39b95 | 2440 | int |
5484965a | 2441 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2442 | { |
3558fd73 | 2443 | znode_t *zp = ITOZ(ip); |
0037b49e | 2444 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
2445 | int error = 0; |
2446 | uint64_t links; | |
0df9673f | 2447 | uint64_t atime[2], mtime[2], ctime[2]; |
5484965a BB |
2448 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2449 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2450 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
0df9673f | 2451 | sa_bulk_attr_t bulk[3]; |
428870ff | 2452 | int count = 0; |
34dc7c2f | 2453 | |
0037b49e | 2454 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2455 | ZFS_VERIFY_ZP(zp); |
428870ff | 2456 | |
5484965a | 2457 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2458 | |
0037b49e BB |
2459 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); |
2460 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
2461 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
428870ff BB |
2462 | |
2463 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
0037b49e | 2464 | ZFS_EXIT(zfsvfs); |
428870ff BB |
2465 | return (error); |
2466 | } | |
34dc7c2f | 2467 | |
34dc7c2f BB |
2468 | /* |
2469 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2470 | * Also, if we are the owner don't bother, since owner should | |
2471 | * always be allowed to read basic attributes of file. | |
2472 | */ | |
572e2857 | 2473 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2474 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2475 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2476 | skipaclchk, cr))) { | |
0037b49e | 2477 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2478 | return (error); |
2479 | } | |
2480 | } | |
2481 | ||
2482 | /* | |
2483 | * Return all attributes. It's cheaper to provide the answer | |
2484 | * than to determine whether we were asked the question. | |
2485 | */ | |
2486 | ||
9babb374 | 2487 | mutex_enter(&zp->z_lock); |
5484965a BB |
2488 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2489 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2490 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2491 | vap->va_nodeid = zp->z_id; |
0037b49e | 2492 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) |
dfbc8630 | 2493 | links = ZTOI(zp)->i_nlink + 1; |
34dc7c2f | 2494 | else |
dfbc8630 | 2495 | links = ZTOI(zp)->i_nlink; |
5484965a BB |
2496 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2497 | vap->va_size = i_size_read(ip); | |
2498 | vap->va_rdev = ip->i_rdev; | |
2499 | vap->va_seq = ip->i_generation; | |
2500 | ||
2501 | /* | |
2502 | * Add in any requested optional attributes and the create time. | |
2503 | * Also set the corresponding bits in the returned attribute bitmap. | |
2504 | */ | |
0037b49e | 2505 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { |
5484965a BB |
2506 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { |
2507 | xoap->xoa_archive = | |
2508 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2509 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2510 | } | |
2511 | ||
2512 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2513 | xoap->xoa_readonly = | |
2514 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2515 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2516 | } | |
2517 | ||
2518 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2519 | xoap->xoa_system = | |
2520 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2521 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2522 | } | |
2523 | ||
2524 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2525 | xoap->xoa_hidden = | |
2526 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2527 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2528 | } | |
2529 | ||
2530 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2531 | xoap->xoa_nounlink = | |
2532 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2533 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2534 | } | |
2535 | ||
2536 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2537 | xoap->xoa_immutable = | |
2538 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2539 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2540 | } | |
2541 | ||
2542 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2543 | xoap->xoa_appendonly = | |
2544 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2545 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2546 | } | |
2547 | ||
2548 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2549 | xoap->xoa_nodump = | |
2550 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2551 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2552 | } | |
2553 | ||
2554 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2555 | xoap->xoa_opaque = | |
2556 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2557 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2558 | } | |
2559 | ||
2560 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2561 | xoap->xoa_av_quarantined = | |
2562 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2563 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2564 | } | |
2565 | ||
2566 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2567 | xoap->xoa_av_modified = | |
2568 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2569 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2570 | } | |
2571 | ||
2572 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2573 | S_ISREG(ip->i_mode)) { | |
2574 | zfs_sa_get_scanstamp(zp, xvap); | |
2575 | } | |
34dc7c2f | 2576 | |
5484965a BB |
2577 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2578 | uint64_t times[2]; | |
2579 | ||
0037b49e | 2580 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), |
5484965a BB |
2581 | times, sizeof (times)); |
2582 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2583 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2584 | } | |
2585 | ||
2586 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2587 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2588 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2589 | } | |
2590 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
278f2236 | 2591 | xoap->xoa_generation = ip->i_generation; |
5484965a BB |
2592 | XVA_SET_RTN(xvap, XAT_GEN); |
2593 | } | |
2594 | ||
2595 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2596 | xoap->xoa_offline = | |
2597 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2598 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2599 | } | |
2600 | ||
2601 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2602 | xoap->xoa_sparse = | |
2603 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2604 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2605 | } | |
9c5167d1 NF |
2606 | |
2607 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { | |
2608 | xoap->xoa_projinherit = | |
2609 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0); | |
2610 | XVA_SET_RTN(xvap, XAT_PROJINHERIT); | |
2611 | } | |
2612 | ||
2613 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
2614 | xoap->xoa_projid = zp->z_projid; | |
2615 | XVA_SET_RTN(xvap, XAT_PROJID); | |
2616 | } | |
5484965a BB |
2617 | } |
2618 | ||
0df9673f | 2619 | ZFS_TIME_DECODE(&vap->va_atime, atime); |
5484965a BB |
2620 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); |
2621 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2622 | |
2623 | mutex_exit(&zp->z_lock); | |
2624 | ||
5484965a | 2625 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2626 | |
2627 | if (zp->z_blksz == 0) { | |
2628 | /* | |
2629 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2630 | */ | |
0037b49e | 2631 | vap->va_blksize = zfsvfs->z_max_blksz; |
34dc7c2f BB |
2632 | } |
2633 | ||
0037b49e | 2634 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2635 | return (0); |
2636 | } | |
2637 | ||
057e8eee BB |
2638 | /* |
2639 | * Get the basic file attributes and place them in the provided kstat | |
2640 | * structure. The inode is assumed to be the authoritative source | |
2641 | * for most of the attributes. However, the znode currently has the | |
2642 | * authoritative atime, blksize, and block count. | |
2643 | * | |
2644 | * IN: ip - inode of file. | |
2645 | * | |
2646 | * OUT: sp - kstat values. | |
2647 | * | |
2648 | * RETURN: 0 (always succeeds) | |
2649 | */ | |
2650 | /* ARGSUSED */ | |
2651 | int | |
2652 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2653 | { | |
2654 | znode_t *zp = ITOZ(ip); | |
0037b49e | 2655 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
b585bc4a BB |
2656 | uint32_t blksize; |
2657 | u_longlong_t nblocks; | |
057e8eee | 2658 | |
0037b49e | 2659 | ZFS_ENTER(zfsvfs); |
a7b125e9 GB |
2660 | ZFS_VERIFY_ZP(zp); |
2661 | ||
057e8eee BB |
2662 | mutex_enter(&zp->z_lock); |
2663 | ||
2664 | generic_fillattr(ip, sp); | |
057e8eee | 2665 | |
b585bc4a BB |
2666 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2667 | sp->blksize = blksize; | |
2668 | sp->blocks = nblocks; | |
2669 | ||
057e8eee BB |
2670 | if (unlikely(zp->z_blksz == 0)) { |
2671 | /* | |
2672 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2673 | */ | |
0037b49e | 2674 | sp->blksize = zfsvfs->z_max_blksz; |
057e8eee BB |
2675 | } |
2676 | ||
2677 | mutex_exit(&zp->z_lock); | |
2678 | ||
aa9b2708 AV |
2679 | /* |
2680 | * Required to prevent NFS client from detecting different inode | |
2681 | * numbers of snapshot root dentry before and after snapshot mount. | |
2682 | */ | |
0037b49e | 2683 | if (zfsvfs->z_issnap) { |
aa9b2708 AV |
2684 | if (ip->i_sb->s_root->d_inode == ip) |
2685 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
0037b49e | 2686 | dmu_objset_id(zfsvfs->z_os); |
aa9b2708 AV |
2687 | } |
2688 | ||
0037b49e | 2689 | ZFS_EXIT(zfsvfs); |
a7b125e9 | 2690 | |
057e8eee BB |
2691 | return (0); |
2692 | } | |
057e8eee | 2693 | |
9c5167d1 NF |
2694 | /* |
2695 | * For the operation of changing file's user/group/project, we need to | |
2696 | * handle not only the main object that is assigned to the file directly, | |
2697 | * but also the ones that are used by the file via hidden xattr directory. | |
2698 | * | |
2699 | * Because the xattr directory may contains many EA entries, as to it may | |
2700 | * be impossible to change all of them via the transaction of changing the | |
2701 | * main object's user/group/project attributes. Then we have to change them | |
2702 | * via other multiple independent transactions one by one. It may be not good | |
2703 | * solution, but we have no better idea yet. | |
2704 | */ | |
2705 | static int | |
2706 | zfs_setattr_dir(znode_t *dzp) | |
2707 | { | |
2708 | struct inode *dxip = ZTOI(dzp); | |
2709 | struct inode *xip = NULL; | |
2710 | zfsvfs_t *zfsvfs = ITOZSB(dxip); | |
2711 | objset_t *os = zfsvfs->z_os; | |
2712 | zap_cursor_t zc; | |
2713 | zap_attribute_t zap; | |
2714 | zfs_dirlock_t *dl; | |
2715 | znode_t *zp; | |
2716 | dmu_tx_t *tx = NULL; | |
2717 | uint64_t uid, gid; | |
2718 | sa_bulk_attr_t bulk[4]; | |
2719 | int count = 0; | |
2720 | int err; | |
2721 | ||
2722 | zap_cursor_init(&zc, os, dzp->z_id); | |
2723 | while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) { | |
2724 | if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { | |
2725 | err = ENXIO; | |
2726 | break; | |
2727 | } | |
2728 | ||
2729 | err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp, | |
2730 | ZEXISTS, NULL, NULL); | |
2731 | if (err == ENOENT) | |
2732 | goto next; | |
2733 | if (err) | |
2734 | break; | |
2735 | ||
2736 | xip = ZTOI(zp); | |
2737 | if (KUID_TO_SUID(xip->i_uid) == KUID_TO_SUID(dxip->i_uid) && | |
2738 | KGID_TO_SGID(xip->i_gid) == KGID_TO_SGID(dxip->i_gid) && | |
2739 | zp->z_projid == dzp->z_projid) | |
2740 | goto next; | |
2741 | ||
2742 | tx = dmu_tx_create(os); | |
2743 | if (!(zp->z_pflags & ZFS_PROJID)) | |
2744 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
2745 | else | |
2746 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
2747 | ||
2748 | err = dmu_tx_assign(tx, TXG_WAIT); | |
2749 | if (err) | |
2750 | break; | |
2751 | ||
2752 | mutex_enter(&dzp->z_lock); | |
2753 | ||
2754 | if (KUID_TO_SUID(xip->i_uid) != KUID_TO_SUID(dxip->i_uid)) { | |
2755 | xip->i_uid = dxip->i_uid; | |
2756 | uid = zfs_uid_read(dxip); | |
2757 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, | |
2758 | &uid, sizeof (uid)); | |
2759 | } | |
2760 | ||
2761 | if (KGID_TO_SGID(xip->i_gid) != KGID_TO_SGID(dxip->i_gid)) { | |
2762 | xip->i_gid = dxip->i_gid; | |
2763 | gid = zfs_gid_read(dxip); | |
2764 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, | |
2765 | &gid, sizeof (gid)); | |
2766 | } | |
2767 | ||
2768 | if (zp->z_projid != dzp->z_projid) { | |
2769 | if (!(zp->z_pflags & ZFS_PROJID)) { | |
2770 | zp->z_pflags |= ZFS_PROJID; | |
2771 | SA_ADD_BULK_ATTR(bulk, count, | |
2772 | SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, | |
2773 | sizeof (zp->z_pflags)); | |
2774 | } | |
2775 | ||
2776 | zp->z_projid = dzp->z_projid; | |
2777 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs), | |
2778 | NULL, &zp->z_projid, sizeof (zp->z_projid)); | |
2779 | } | |
2780 | ||
2781 | mutex_exit(&dzp->z_lock); | |
2782 | ||
2783 | if (likely(count > 0)) { | |
2784 | err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
2785 | dmu_tx_commit(tx); | |
2786 | } else { | |
2787 | dmu_tx_abort(tx); | |
2788 | } | |
2789 | tx = NULL; | |
2790 | if (err != 0 && err != ENOENT) | |
2791 | break; | |
2792 | ||
2793 | next: | |
2794 | if (xip) { | |
2795 | iput(xip); | |
2796 | xip = NULL; | |
2797 | zfs_dirent_unlock(dl); | |
2798 | } | |
2799 | zap_cursor_advance(&zc); | |
2800 | } | |
2801 | ||
2802 | if (tx) | |
2803 | dmu_tx_abort(tx); | |
2804 | if (xip) { | |
2805 | iput(xip); | |
2806 | zfs_dirent_unlock(dl); | |
2807 | } | |
2808 | zap_cursor_fini(&zc); | |
2809 | ||
2810 | return (err == ENOENT ? 0 : err); | |
2811 | } | |
2812 | ||
34dc7c2f BB |
2813 | /* |
2814 | * Set the file attributes to the values contained in the | |
2815 | * vattr structure. | |
2816 | * | |
3558fd73 | 2817 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2818 | * vap - new attribute values. |
5484965a | 2819 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2820 | * flags - ATTR_UTIME set if non-default time values provided. |
2821 | * - ATTR_NOACLCHECK (CIFS context only). | |
2822 | * cr - credentials of caller. | |
34dc7c2f BB |
2823 | * |
2824 | * RETURN: 0 if success | |
2825 | * error code if failure | |
2826 | * | |
2827 | * Timestamps: | |
3558fd73 | 2828 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2829 | */ |
2830 | /* ARGSUSED */ | |
e5c39b95 | 2831 | int |
5484965a | 2832 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2833 | { |
3558fd73 | 2834 | znode_t *zp = ITOZ(ip); |
0037b49e | 2835 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
9c5167d1 | 2836 | objset_t *os = zfsvfs->z_os; |
34dc7c2f BB |
2837 | zilog_t *zilog; |
2838 | dmu_tx_t *tx; | |
2839 | vattr_t oldva; | |
f4ea75d4 | 2840 | xvattr_t *tmpxvattr; |
5484965a | 2841 | uint_t mask = vap->va_mask; |
a117a6d6 | 2842 | uint_t saved_mask = 0; |
34dc7c2f BB |
2843 | int trim_mask = 0; |
2844 | uint64_t new_mode; | |
64aefee1 | 2845 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 2846 | uint64_t xattr_obj; |
0df9673f | 2847 | uint64_t mtime[2], ctime[2], atime[2]; |
9c5167d1 | 2848 | uint64_t projid = ZFS_INVALID_PROJID; |
34dc7c2f BB |
2849 | znode_t *attrzp; |
2850 | int need_policy = FALSE; | |
9c5167d1 | 2851 | int err, err2 = 0; |
34dc7c2f | 2852 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2853 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2854 | xoptattr_t *xoap; | |
2855 | zfs_acl_t *aclp; | |
34dc7c2f | 2856 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2857 | boolean_t fuid_dirtied = B_FALSE; |
9c5167d1 | 2858 | boolean_t handle_eadir = B_FALSE; |
17c37660 | 2859 | sa_bulk_attr_t *bulk, *xattr_bulk; |
9c5167d1 | 2860 | int count = 0, xattr_count = 0, bulks = 8; |
34dc7c2f BB |
2861 | |
2862 | if (mask == 0) | |
2863 | return (0); | |
2864 | ||
0037b49e | 2865 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2866 | ZFS_VERIFY_ZP(zp); |
2867 | ||
9c5167d1 NF |
2868 | /* |
2869 | * If this is a xvattr_t, then get a pointer to the structure of | |
2870 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2871 | */ | |
2872 | xoap = xva_getxoptattr(xvap); | |
2873 | if (xoap != NULL && (mask & ATTR_XVATTR)) { | |
2874 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
2875 | if (!dmu_objset_projectquota_enabled(os) || | |
2876 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode))) { | |
2877 | ZFS_EXIT(zfsvfs); | |
2878 | return (SET_ERROR(ENOTSUP)); | |
2879 | } | |
2880 | ||
2881 | projid = xoap->xoa_projid; | |
2882 | if (unlikely(projid == ZFS_INVALID_PROJID)) { | |
2883 | ZFS_EXIT(zfsvfs); | |
2884 | return (SET_ERROR(EINVAL)); | |
2885 | } | |
2886 | ||
2887 | if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID) | |
2888 | projid = ZFS_INVALID_PROJID; | |
2889 | else | |
2890 | need_policy = TRUE; | |
2891 | } | |
2892 | ||
2893 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) && | |
2894 | (!dmu_objset_projectquota_enabled(os) || | |
2895 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode)))) { | |
2896 | ZFS_EXIT(zfsvfs); | |
2897 | return (SET_ERROR(ENOTSUP)); | |
2898 | } | |
2899 | } | |
2900 | ||
0037b49e | 2901 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2902 | |
2903 | /* | |
2904 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2905 | * that file system is at proper version level | |
2906 | */ | |
5484965a | 2907 | |
0037b49e | 2908 | if (zfsvfs->z_use_fuids == B_FALSE && |
5484965a BB |
2909 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2910 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2911 | (mask & ATTR_XVATTR))) { | |
0037b49e | 2912 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2913 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2914 | } |
2915 | ||
3558fd73 | 2916 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
0037b49e | 2917 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2918 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2919 | } |
2920 | ||
3558fd73 | 2921 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
0037b49e | 2922 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2923 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2924 | } |
2925 | ||
d1d7e268 | 2926 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2927 | xva_init(tmpxvattr); |
5484965a | 2928 | |
9c5167d1 NF |
2929 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); |
2930 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); | |
17c37660 | 2931 | |
5484965a BB |
2932 | /* |
2933 | * Immutable files can only alter immutable bit and atime | |
2934 | */ | |
2935 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2936 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2937 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
ecb2b7dc | 2938 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2939 | goto out3; |
5484965a BB |
2940 | } |
2941 | ||
3558fd73 | 2942 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
ecb2b7dc | 2943 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2944 | goto out3; |
34dc7c2f BB |
2945 | } |
2946 | ||
5484965a BB |
2947 | /* |
2948 | * Verify timestamps doesn't overflow 32 bits. | |
2949 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2950 | * handle times greater than 2039. This check should be removed | |
2951 | * once large timestamps are fully supported. | |
2952 | */ | |
2953 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
2954 | if (((mask & ATTR_ATIME) && |
2955 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2956 | ((mask & ATTR_MTIME) && | |
2957 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
ecb2b7dc | 2958 | err = SET_ERROR(EOVERFLOW); |
f4ea75d4 | 2959 | goto out3; |
5484965a BB |
2960 | } |
2961 | } | |
2962 | ||
34dc7c2f BB |
2963 | top: |
2964 | attrzp = NULL; | |
572e2857 | 2965 | aclp = NULL; |
34dc7c2f | 2966 | |
45d1cae3 | 2967 | /* Can this be moved to before the top label? */ |
0037b49e | 2968 | if (zfs_is_readonly(zfsvfs)) { |
ecb2b7dc | 2969 | err = SET_ERROR(EROFS); |
f4ea75d4 | 2970 | goto out3; |
34dc7c2f BB |
2971 | } |
2972 | ||
2973 | /* | |
2974 | * First validate permissions | |
2975 | */ | |
2976 | ||
3558fd73 | 2977 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2978 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2979 | if (err) |
2980 | goto out3; | |
2981 | ||
34dc7c2f BB |
2982 | /* |
2983 | * XXX - Note, we are not providing any open | |
2984 | * mode flags here (like FNDELAY), so we may | |
2985 | * block if there are locks present... this | |
2986 | * should be addressed in openat(). | |
2987 | */ | |
b128c09f | 2988 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2989 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2990 | if (err) |
2991 | goto out3; | |
428870ff | 2992 | } |
34dc7c2f | 2993 | |
5484965a BB |
2994 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2995 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2996 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2997 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2998 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2999 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
3000 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
3001 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
3002 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
3003 | skipaclchk, cr); | |
3004 | } | |
3005 | ||
3558fd73 BB |
3006 | if (mask & (ATTR_UID|ATTR_GID)) { |
3007 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
3008 | int take_owner; |
3009 | int take_group; | |
3010 | ||
3011 | /* | |
3012 | * NOTE: even if a new mode is being set, | |
3013 | * we may clear S_ISUID/S_ISGID bits. | |
3014 | */ | |
3015 | ||
3558fd73 | 3016 | if (!(mask & ATTR_MODE)) |
5484965a | 3017 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
3018 | |
3019 | /* | |
3020 | * Take ownership or chgrp to group we are a member of | |
3021 | */ | |
3022 | ||
5484965a | 3023 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 3024 | take_group = (mask & ATTR_GID) && |
0037b49e | 3025 | zfs_groupmember(zfsvfs, vap->va_gid, cr); |
34dc7c2f BB |
3026 | |
3027 | /* | |
5484965a | 3028 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
3029 | * take_group must both be set in order to allow taking |
3030 | * ownership. | |
3031 | * | |
3032 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
3033 | * | |
3034 | */ | |
3035 | ||
3558fd73 BB |
3036 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
3037 | take_owner && take_group) || | |
3038 | ((idmask == ATTR_UID) && take_owner) || | |
3039 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
3040 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
3041 | skipaclchk, cr) == 0) { | |
3042 | /* | |
3043 | * Remove setuid/setgid for non-privileged users | |
3044 | */ | |
5484965a | 3045 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 3046 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
3047 | } else { |
3048 | need_policy = TRUE; | |
3049 | } | |
3050 | } else { | |
3051 | need_policy = TRUE; | |
3052 | } | |
3053 | } | |
3054 | ||
3055 | mutex_enter(&zp->z_lock); | |
428870ff | 3056 | oldva.va_mode = zp->z_mode; |
572e2857 | 3057 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
3058 | if (mask & ATTR_XVATTR) { |
3059 | /* | |
3060 | * Update xvattr mask to include only those attributes | |
3061 | * that are actually changing. | |
3062 | * | |
3063 | * the bits will be restored prior to actually setting | |
3064 | * the attributes so the caller thinks they were set. | |
3065 | */ | |
3066 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
3067 | if (xoap->xoa_appendonly != | |
3068 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
3069 | need_policy = TRUE; | |
3070 | } else { | |
3071 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 3072 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
3073 | } |
3074 | } | |
3075 | ||
9c5167d1 NF |
3076 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { |
3077 | if (xoap->xoa_projinherit != | |
3078 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) { | |
3079 | need_policy = TRUE; | |
3080 | } else { | |
3081 | XVA_CLR_REQ(xvap, XAT_PROJINHERIT); | |
3082 | XVA_SET_REQ(tmpxvattr, XAT_PROJINHERIT); | |
3083 | } | |
3084 | } | |
3085 | ||
5484965a BB |
3086 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { |
3087 | if (xoap->xoa_nounlink != | |
3088 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
3089 | need_policy = TRUE; | |
3090 | } else { | |
3091 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 3092 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
3093 | } |
3094 | } | |
3095 | ||
3096 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
3097 | if (xoap->xoa_immutable != | |
3098 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
3099 | need_policy = TRUE; | |
3100 | } else { | |
3101 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 3102 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
3103 | } |
3104 | } | |
3105 | ||
3106 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
3107 | if (xoap->xoa_nodump != | |
3108 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
3109 | need_policy = TRUE; | |
3110 | } else { | |
3111 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 3112 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
3113 | } |
3114 | } | |
3115 | ||
3116 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
3117 | if (xoap->xoa_av_modified != | |
3118 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
3119 | need_policy = TRUE; | |
3120 | } else { | |
3121 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 3122 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
3123 | } |
3124 | } | |
3125 | ||
3126 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
3127 | if ((!S_ISREG(ip->i_mode) && | |
3128 | xoap->xoa_av_quarantined) || | |
3129 | xoap->xoa_av_quarantined != | |
3130 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
3131 | need_policy = TRUE; | |
3132 | } else { | |
3133 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 3134 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
3135 | } |
3136 | } | |
3137 | ||
3138 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
3139 | mutex_exit(&zp->z_lock); | |
ecb2b7dc | 3140 | err = SET_ERROR(EPERM); |
f4ea75d4 | 3141 | goto out3; |
5484965a BB |
3142 | } |
3143 | ||
3144 | if (need_policy == FALSE && | |
3145 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
3146 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
3147 | need_policy = TRUE; | |
3148 | } | |
3149 | } | |
34dc7c2f BB |
3150 | |
3151 | mutex_exit(&zp->z_lock); | |
3152 | ||
3558fd73 | 3153 | if (mask & ATTR_MODE) { |
34dc7c2f | 3154 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 3155 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 3156 | &oldva, cr); |
f4ea75d4 BB |
3157 | if (err) |
3158 | goto out3; | |
3159 | ||
3558fd73 | 3160 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
3161 | } else { |
3162 | need_policy = TRUE; | |
3163 | } | |
3164 | } | |
3165 | ||
3166 | if (need_policy) { | |
3167 | /* | |
3168 | * If trim_mask is set then take ownership | |
3169 | * has been granted or write_acl is present and user | |
3170 | * has the ability to modify mode. In that case remove | |
3171 | * UID|GID and or MODE from mask so that | |
3172 | * secpolicy_vnode_setattr() doesn't revoke it. | |
3173 | */ | |
3174 | ||
3175 | if (trim_mask) { | |
5484965a BB |
3176 | saved_mask = vap->va_mask; |
3177 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 3178 | } |
5484965a | 3179 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 3180 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
3181 | if (err) |
3182 | goto out3; | |
34dc7c2f BB |
3183 | |
3184 | if (trim_mask) | |
5484965a | 3185 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
3186 | } |
3187 | ||
3188 | /* | |
3189 | * secpolicy_vnode_setattr, or take ownership may have | |
3190 | * changed va_mask | |
3191 | */ | |
5484965a | 3192 | mask = vap->va_mask; |
34dc7c2f | 3193 | |
9c5167d1 NF |
3194 | if ((mask & (ATTR_UID | ATTR_GID)) || projid != ZFS_INVALID_PROJID) { |
3195 | handle_eadir = B_TRUE; | |
0037b49e | 3196 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
572e2857 | 3197 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 3198 | |
572e2857 | 3199 | if (err == 0 && xattr_obj) { |
3558fd73 | 3200 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
3201 | if (err) |
3202 | goto out2; | |
3203 | } | |
3558fd73 | 3204 | if (mask & ATTR_UID) { |
0037b49e | 3205 | new_kuid = zfs_fuid_create(zfsvfs, |
5484965a | 3206 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
64aefee1 | 3207 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
9c5167d1 NF |
3208 | zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT, |
3209 | new_kuid)) { | |
572e2857 | 3210 | if (attrzp) |
3558fd73 | 3211 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3212 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3213 | goto out2; |
3214 | } | |
3215 | } | |
3216 | ||
3558fd73 | 3217 | if (mask & ATTR_GID) { |
0037b49e BB |
3218 | new_kgid = zfs_fuid_create(zfsvfs, |
3219 | (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); | |
64aefee1 | 3220 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && |
9c5167d1 NF |
3221 | zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT, |
3222 | new_kgid)) { | |
572e2857 | 3223 | if (attrzp) |
3558fd73 | 3224 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3225 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3226 | goto out2; |
3227 | } | |
3228 | } | |
9c5167d1 NF |
3229 | |
3230 | if (projid != ZFS_INVALID_PROJID && | |
3231 | zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) { | |
3232 | if (attrzp) | |
3233 | iput(ZTOI(attrzp)); | |
3234 | err = EDQUOT; | |
3235 | goto out2; | |
3236 | } | |
428870ff | 3237 | } |
9c5167d1 | 3238 | tx = dmu_tx_create(os); |
34dc7c2f | 3239 | |
3558fd73 | 3240 | if (mask & ATTR_MODE) { |
428870ff | 3241 | uint64_t pmode = zp->z_mode; |
572e2857 | 3242 | uint64_t acl_obj; |
5484965a | 3243 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 3244 | |
572e2857 | 3245 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 3246 | |
572e2857 BB |
3247 | mutex_enter(&zp->z_lock); |
3248 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
3249 | /* |
3250 | * Are we upgrading ACL from old V0 format | |
3251 | * to V1 format? | |
3252 | */ | |
0037b49e | 3253 | if (zfsvfs->z_version >= ZPL_VERSION_FUID && |
572e2857 | 3254 | zfs_znode_acl_version(zp) == |
34dc7c2f | 3255 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 3256 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
3257 | DMU_OBJECT_END); |
3258 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
3259 | 0, aclp->z_acl_bytes); | |
3260 | } else { | |
572e2857 | 3261 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
3262 | aclp->z_acl_bytes); |
3263 | } | |
428870ff | 3264 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
3265 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
3266 | 0, aclp->z_acl_bytes); | |
3267 | } | |
572e2857 | 3268 | mutex_exit(&zp->z_lock); |
428870ff BB |
3269 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3270 | } else { | |
9c5167d1 NF |
3271 | if (((mask & ATTR_XVATTR) && |
3272 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) || | |
3273 | (projid != ZFS_INVALID_PROJID && | |
3274 | !(zp->z_pflags & ZFS_PROJID))) | |
5484965a BB |
3275 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3276 | else | |
3277 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3278 | } |
3279 | ||
428870ff BB |
3280 | if (attrzp) { |
3281 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3282 | } |
3283 | ||
0037b49e | 3284 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
428870ff | 3285 | if (fuid_dirtied) |
0037b49e | 3286 | zfs_fuid_txhold(zfsvfs, tx); |
428870ff BB |
3287 | |
3288 | zfs_sa_upgrade_txholds(tx, zp); | |
3289 | ||
384f8a09 MA |
3290 | err = dmu_tx_assign(tx, TXG_WAIT); |
3291 | if (err) | |
9babb374 | 3292 | goto out; |
34dc7c2f | 3293 | |
428870ff | 3294 | count = 0; |
34dc7c2f BB |
3295 | /* |
3296 | * Set each attribute requested. | |
3297 | * We group settings according to the locks they need to acquire. | |
3298 | * | |
3299 | * Note: you cannot set ctime directly, although it will be | |
3300 | * updated as a side-effect of calling this function. | |
3301 | */ | |
3302 | ||
9c5167d1 NF |
3303 | if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) { |
3304 | /* | |
3305 | * For the existed object that is upgraded from old system, | |
3306 | * its on-disk layout has no slot for the project ID attribute. | |
3307 | * But quota accounting logic needs to access related slots by | |
3308 | * offset directly. So we need to adjust old objects' layout | |
3309 | * to make the project ID to some unified and fixed offset. | |
3310 | */ | |
3311 | if (attrzp) | |
3312 | err = sa_add_projid(attrzp->z_sa_hdl, tx, projid); | |
3313 | if (err == 0) | |
3314 | err = sa_add_projid(zp->z_sa_hdl, tx, projid); | |
3315 | ||
3316 | if (unlikely(err == EEXIST)) | |
3317 | err = 0; | |
3318 | else if (err != 0) | |
3319 | goto out; | |
3320 | else | |
3321 | projid = ZFS_INVALID_PROJID; | |
3322 | } | |
572e2857 | 3323 | |
3558fd73 | 3324 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3325 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
3326 | mutex_enter(&zp->z_lock); |
3327 | ||
0037b49e | 3328 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
3329 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3330 | ||
3331 | if (attrzp) { | |
3558fd73 | 3332 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3333 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
3334 | mutex_enter(&attrzp->z_lock); |
3335 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3336 | SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, |
428870ff | 3337 | sizeof (attrzp->z_pflags)); |
9c5167d1 NF |
3338 | if (projid != ZFS_INVALID_PROJID) { |
3339 | attrzp->z_projid = projid; | |
3340 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3341 | SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid, | |
3342 | sizeof (attrzp->z_projid)); | |
3343 | } | |
428870ff BB |
3344 | } |
3345 | ||
3558fd73 | 3346 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 3347 | |
3558fd73 | 3348 | if (mask & ATTR_UID) { |
64aefee1 NB |
3349 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
3350 | new_uid = zfs_uid_read(ZTOI(zp)); | |
0037b49e | 3351 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 3352 | &new_uid, sizeof (new_uid)); |
428870ff BB |
3353 | if (attrzp) { |
3354 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3355 | SA_ZPL_UID(zfsvfs), NULL, &new_uid, |
428870ff | 3356 | sizeof (new_uid)); |
2c6abf15 | 3357 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
3358 | } |
3359 | } | |
3360 | ||
3558fd73 | 3361 | if (mask & ATTR_GID) { |
64aefee1 NB |
3362 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
3363 | new_gid = zfs_gid_read(ZTOI(zp)); | |
0037b49e | 3364 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), |
428870ff | 3365 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
3366 | if (attrzp) { |
3367 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3368 | SA_ZPL_GID(zfsvfs), NULL, &new_gid, |
428870ff | 3369 | sizeof (new_gid)); |
64aefee1 | 3370 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
3371 | } |
3372 | } | |
3558fd73 | 3373 | if (!(mask & ATTR_MODE)) { |
0037b49e | 3374 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), |
428870ff BB |
3375 | NULL, &new_mode, sizeof (new_mode)); |
3376 | new_mode = zp->z_mode; | |
3377 | } | |
3378 | err = zfs_acl_chown_setattr(zp); | |
3379 | ASSERT(err == 0); | |
3380 | if (attrzp) { | |
3381 | err = zfs_acl_chown_setattr(attrzp); | |
3382 | ASSERT(err == 0); | |
3383 | } | |
3384 | } | |
3385 | ||
3558fd73 | 3386 | if (mask & ATTR_MODE) { |
0037b49e | 3387 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 3388 | &new_mode, sizeof (new_mode)); |
12fa7f34 | 3389 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 3390 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 3391 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 3392 | ASSERT0(err); |
572e2857 BB |
3393 | if (zp->z_acl_cached) |
3394 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
3395 | zp->z_acl_cached = aclp; |
3396 | aclp = NULL; | |
34dc7c2f BB |
3397 | } |
3398 | ||
704cd075 CC |
3399 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
3400 | zp->z_atime_dirty = 0; | |
3401 | ZFS_TIME_ENCODE(&ip->i_atime, atime); | |
0037b49e | 3402 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
0df9673f | 3403 | &atime, sizeof (atime)); |
34dc7c2f BB |
3404 | } |
3405 | ||
99834d19 | 3406 | if (mask & (ATTR_MTIME | ATTR_SIZE)) { |
5484965a | 3407 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
87f9371a NB |
3408 | ZTOI(zp)->i_mtime = timespec_trunc(vap->va_mtime, |
3409 | ZTOI(zp)->i_sb->s_time_gran); | |
3410 | ||
0037b49e | 3411 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 3412 | mtime, sizeof (mtime)); |
34dc7c2f BB |
3413 | } |
3414 | ||
99834d19 | 3415 | if (mask & (ATTR_CTIME | ATTR_SIZE)) { |
87f9371a NB |
3416 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); |
3417 | ZTOI(zp)->i_ctime = timespec_trunc(vap->va_ctime, | |
3418 | ZTOI(zp)->i_sb->s_time_gran); | |
0037b49e | 3419 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
87f9371a | 3420 | ctime, sizeof (ctime)); |
428870ff | 3421 | } |
87f9371a | 3422 | |
9c5167d1 NF |
3423 | if (projid != ZFS_INVALID_PROJID) { |
3424 | zp->z_projid = projid; | |
3425 | SA_ADD_BULK_ATTR(bulk, count, | |
3426 | SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid, | |
3427 | sizeof (zp->z_projid)); | |
3428 | } | |
3429 | ||
87f9371a NB |
3430 | if (attrzp && mask) { |
3431 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3432 | SA_ZPL_CTIME(zfsvfs), NULL, &ctime, |
87f9371a NB |
3433 | sizeof (ctime)); |
3434 | } | |
3435 | ||
34dc7c2f BB |
3436 | /* |
3437 | * Do this after setting timestamps to prevent timestamp | |
3438 | * update from toggling bit | |
3439 | */ | |
3440 | ||
5484965a BB |
3441 | if (xoap && (mask & ATTR_XVATTR)) { |
3442 | ||
3443 | /* | |
3444 | * restore trimmed off masks | |
3445 | * so that return masks can be set for caller. | |
3446 | */ | |
3447 | ||
f4ea75d4 | 3448 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
3449 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
3450 | } | |
f4ea75d4 | 3451 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
3452 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
3453 | } | |
f4ea75d4 | 3454 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
3455 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
3456 | } | |
f4ea75d4 | 3457 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
3458 | XVA_SET_REQ(xvap, XAT_NODUMP); |
3459 | } | |
f4ea75d4 | 3460 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
3461 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
3462 | } | |
f4ea75d4 | 3463 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
3464 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
3465 | } | |
9c5167d1 NF |
3466 | if (XVA_ISSET_REQ(tmpxvattr, XAT_PROJINHERIT)) { |
3467 | XVA_SET_REQ(xvap, XAT_PROJINHERIT); | |
3468 | } | |
5484965a BB |
3469 | |
3470 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3471 | ASSERT(S_ISREG(ip->i_mode)); | |
3472 | ||
3473 | zfs_xvattr_set(zp, xvap, tx); | |
3474 | } | |
3475 | ||
9babb374 | 3476 | if (fuid_dirtied) |
0037b49e | 3477 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 3478 | |
34dc7c2f | 3479 | if (mask != 0) |
5484965a | 3480 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3481 | |
34dc7c2f | 3482 | mutex_exit(&zp->z_lock); |
3558fd73 | 3483 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3484 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3485 | |
572e2857 | 3486 | if (attrzp) { |
3558fd73 | 3487 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3488 | mutex_exit(&attrzp->z_acl_lock); |
3489 | mutex_exit(&attrzp->z_lock); | |
3490 | } | |
9babb374 | 3491 | out: |
9c5167d1 | 3492 | if (err == 0 && xattr_count > 0) { |
428870ff BB |
3493 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, |
3494 | xattr_count, tx); | |
3495 | ASSERT(err2 == 0); | |
3496 | } | |
3497 | ||
45d1cae3 | 3498 | if (aclp) |
9babb374 | 3499 | zfs_acl_free(aclp); |
9babb374 BB |
3500 | |
3501 | if (fuidp) { | |
3502 | zfs_fuid_info_free(fuidp); | |
3503 | fuidp = NULL; | |
3504 | } | |
3505 | ||
428870ff | 3506 | if (err) { |
9babb374 | 3507 | dmu_tx_abort(tx); |
ea7e86d8 BB |
3508 | if (attrzp) |
3509 | iput(ZTOI(attrzp)); | |
428870ff BB |
3510 | if (err == ERESTART) |
3511 | goto top; | |
3512 | } else { | |
9c5167d1 NF |
3513 | if (count > 0) |
3514 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3515 | dmu_tx_commit(tx); |
9c5167d1 NF |
3516 | if (attrzp) { |
3517 | if (err2 == 0 && handle_eadir) | |
3518 | err2 = zfs_setattr_dir(attrzp); | |
ea7e86d8 | 3519 | iput(ZTOI(attrzp)); |
9c5167d1 | 3520 | } |
037849f8 | 3521 | zfs_inode_update(zp); |
428870ff BB |
3522 | } |
3523 | ||
428870ff | 3524 | out2: |
9c5167d1 | 3525 | if (os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3526 | zil_commit(zilog, 0); |
34dc7c2f | 3527 | |
f4ea75d4 | 3528 | out3: |
9c5167d1 NF |
3529 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks); |
3530 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * bulks); | |
d1d7e268 | 3531 | kmem_free(tmpxvattr, sizeof (xvattr_t)); |
0037b49e | 3532 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3533 | return (err); |
3534 | } | |
3535 | ||
3536 | typedef struct zfs_zlock { | |
3537 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3538 | znode_t *zl_znode; /* znode we held */ | |
3539 | struct zfs_zlock *zl_next; /* next in list */ | |
3540 | } zfs_zlock_t; | |
3541 | ||
3542 | /* | |
3543 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3544 | */ | |
3545 | static void | |
3546 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3547 | { | |
3548 | zfs_zlock_t *zl; | |
3549 | ||
3550 | while ((zl = *zlpp) != NULL) { | |
3551 | if (zl->zl_znode != NULL) | |
ea7e86d8 | 3552 | zfs_iput_async(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3553 | rw_exit(zl->zl_rwlock); |
3554 | *zlpp = zl->zl_next; | |
3555 | kmem_free(zl, sizeof (*zl)); | |
3556 | } | |
3557 | } | |
3558 | ||
3559 | /* | |
3560 | * Search back through the directory tree, using the ".." entries. | |
3561 | * Lock each directory in the chain to prevent concurrent renames. | |
3562 | * Fail any attempt to move a directory into one of its own descendants. | |
3563 | * XXX - z_parent_lock can overlap with map or grow locks | |
3564 | */ | |
3565 | static int | |
3566 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3567 | { | |
3568 | zfs_zlock_t *zl; | |
3569 | znode_t *zp = tdzp; | |
3558fd73 | 3570 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3571 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3572 | krwlock_t *rwlp = &szp->z_parent_lock; |
3573 | krw_t rw = RW_WRITER; | |
3574 | ||
3575 | /* | |
3576 | * First pass write-locks szp and compares to zp->z_id. | |
3577 | * Later passes read-lock zp and compare to zp->z_parent. | |
3578 | */ | |
3579 | do { | |
3580 | if (!rw_tryenter(rwlp, rw)) { | |
3581 | /* | |
3582 | * Another thread is renaming in this path. | |
3583 | * Note that if we are a WRITER, we don't have any | |
3584 | * parent_locks held yet. | |
3585 | */ | |
3586 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3587 | /* | |
3588 | * Drop our locks and restart | |
3589 | */ | |
3590 | zfs_rename_unlock(&zl); | |
3591 | *zlpp = NULL; | |
3592 | zp = tdzp; | |
428870ff | 3593 | oidp = zp->z_id; |
34dc7c2f BB |
3594 | rwlp = &szp->z_parent_lock; |
3595 | rw = RW_WRITER; | |
3596 | continue; | |
3597 | } else { | |
3598 | /* | |
3599 | * Wait for other thread to drop its locks | |
3600 | */ | |
3601 | rw_enter(rwlp, rw); | |
3602 | } | |
3603 | } | |
3604 | ||
3605 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3606 | zl->zl_rwlock = rwlp; | |
3607 | zl->zl_znode = NULL; | |
3608 | zl->zl_next = *zlpp; | |
3609 | *zlpp = zl; | |
3610 | ||
428870ff | 3611 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 3612 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3613 | |
428870ff | 3614 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3615 | return (0); |
3616 | ||
3617 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3618 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3619 | if (error) |
3620 | return (error); | |
3621 | zl->zl_znode = zp; | |
3622 | } | |
3558fd73 | 3623 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3624 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3625 | rwlp = &zp->z_parent_lock; |
3626 | rw = RW_READER; | |
3627 | ||
3628 | } while (zp->z_id != sdzp->z_id); | |
3629 | ||
3630 | return (0); | |
3631 | } | |
3632 | ||
3633 | /* | |
3634 | * Move an entry from the provided source directory to the target | |
3635 | * directory. Change the entry name as indicated. | |
3636 | * | |
3558fd73 | 3637 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3638 | * snm - Old entry name. |
3558fd73 | 3639 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3640 | * tnm - New entry name. |
3641 | * cr - credentials of caller. | |
34dc7c2f BB |
3642 | * flags - case flags |
3643 | * | |
d3cc8b15 | 3644 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3645 | * |
3646 | * Timestamps: | |
3558fd73 | 3647 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3648 | */ |
3649 | /*ARGSUSED*/ | |
e5c39b95 | 3650 | int |
3558fd73 BB |
3651 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3652 | cred_t *cr, int flags) | |
34dc7c2f BB |
3653 | { |
3654 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 | 3655 | znode_t *sdzp = ITOZ(sdip); |
0037b49e | 3656 | zfsvfs_t *zfsvfs = ITOZSB(sdip); |
34dc7c2f | 3657 | zilog_t *zilog; |
34dc7c2f BB |
3658 | zfs_dirlock_t *sdl, *tdl; |
3659 | dmu_tx_t *tx; | |
3660 | zfs_zlock_t *zl; | |
3661 | int cmp, serr, terr; | |
3662 | int error = 0; | |
3663 | int zflg = 0; | |
e8b96c60 | 3664 | boolean_t waited = B_FALSE; |
34dc7c2f | 3665 | |
32dec7bd | 3666 | if (snm == NULL || tnm == NULL) |
3667 | return (SET_ERROR(EINVAL)); | |
3668 | ||
0037b49e | 3669 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 3670 | ZFS_VERIFY_ZP(sdzp); |
0037b49e | 3671 | zilog = zfsvfs->z_log; |
34dc7c2f | 3672 | |
812e91a7 MT |
3673 | tdzp = ITOZ(tdip); |
3674 | ZFS_VERIFY_ZP(tdzp); | |
3675 | ||
3676 | /* | |
3677 | * We check i_sb because snapshots and the ctldir must have different | |
3678 | * super blocks. | |
3679 | */ | |
c0ebc844 | 3680 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
0037b49e | 3681 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3682 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3683 | } |
3684 | ||
0037b49e | 3685 | if (zfsvfs->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3686 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 3687 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3688 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3689 | } |
3690 | ||
3691 | if (flags & FIGNORECASE) | |
3692 | zflg |= ZCILOOK; | |
3693 | ||
3694 | top: | |
3695 | szp = NULL; | |
3696 | tzp = NULL; | |
3697 | zl = NULL; | |
3698 | ||
3699 | /* | |
3700 | * This is to prevent the creation of links into attribute space | |
3701 | * by renaming a linked file into/outof an attribute directory. | |
3702 | * See the comment in zfs_link() for why this is considered bad. | |
3703 | */ | |
428870ff | 3704 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 3705 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3706 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3707 | } |
3708 | ||
3709 | /* | |
3710 | * Lock source and target directory entries. To prevent deadlock, | |
3711 | * a lock ordering must be defined. We lock the directory with | |
3712 | * the smallest object id first, or if it's a tie, the one with | |
3713 | * the lexically first name. | |
3714 | */ | |
3715 | if (sdzp->z_id < tdzp->z_id) { | |
3716 | cmp = -1; | |
3717 | } else if (sdzp->z_id > tdzp->z_id) { | |
3718 | cmp = 1; | |
3719 | } else { | |
3720 | /* | |
3721 | * First compare the two name arguments without | |
3722 | * considering any case folding. | |
3723 | */ | |
0037b49e | 3724 | int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3725 | |
3726 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
0037b49e | 3727 | ASSERT(error == 0 || !zfsvfs->z_utf8); |
34dc7c2f BB |
3728 | if (cmp == 0) { |
3729 | /* | |
3730 | * POSIX: "If the old argument and the new argument | |
3731 | * both refer to links to the same existing file, | |
3732 | * the rename() function shall return successfully | |
3733 | * and perform no other action." | |
3734 | */ | |
0037b49e | 3735 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3736 | return (0); |
3737 | } | |
3738 | /* | |
3739 | * If the file system is case-folding, then we may | |
3740 | * have some more checking to do. A case-folding file | |
3741 | * system is either supporting mixed case sensitivity | |
3742 | * access or is completely case-insensitive. Note | |
3743 | * that the file system is always case preserving. | |
3744 | * | |
3745 | * In mixed sensitivity mode case sensitive behavior | |
3746 | * is the default. FIGNORECASE must be used to | |
3747 | * explicitly request case insensitive behavior. | |
3748 | * | |
3749 | * If the source and target names provided differ only | |
3750 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3751 | * we will treat this as a special case in the | |
3752 | * case-insensitive mode: as long as the source name | |
3753 | * is an exact match, we will allow this to proceed as | |
3754 | * a name-change request. | |
3755 | */ | |
0037b49e BB |
3756 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
3757 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3758 | flags & FIGNORECASE)) && |
0037b49e | 3759 | u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3760 | &error) == 0) { |
3761 | /* | |
3762 | * case preserving rename request, require exact | |
3763 | * name matches | |
3764 | */ | |
3765 | zflg |= ZCIEXACT; | |
3766 | zflg &= ~ZCILOOK; | |
3767 | } | |
3768 | } | |
3769 | ||
428870ff BB |
3770 | /* |
3771 | * If the source and destination directories are the same, we should | |
3772 | * grab the z_name_lock of that directory only once. | |
3773 | */ | |
3774 | if (sdzp == tdzp) { | |
3775 | zflg |= ZHAVELOCK; | |
3776 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3777 | } | |
3778 | ||
34dc7c2f BB |
3779 | if (cmp < 0) { |
3780 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3781 | ZEXISTS | zflg, NULL, NULL); | |
3782 | terr = zfs_dirent_lock(&tdl, | |
3783 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3784 | } else { | |
3785 | terr = zfs_dirent_lock(&tdl, | |
3786 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3787 | serr = zfs_dirent_lock(&sdl, | |
3788 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3789 | NULL, NULL); | |
3790 | } | |
3791 | ||
3792 | if (serr) { | |
3793 | /* | |
3794 | * Source entry invalid or not there. | |
3795 | */ | |
3796 | if (!terr) { | |
3797 | zfs_dirent_unlock(tdl); | |
3798 | if (tzp) | |
3558fd73 | 3799 | iput(ZTOI(tzp)); |
34dc7c2f | 3800 | } |
428870ff BB |
3801 | |
3802 | if (sdzp == tdzp) | |
3803 | rw_exit(&sdzp->z_name_lock); | |
3804 | ||
34dc7c2f BB |
3805 | if (strcmp(snm, "..") == 0) |
3806 | serr = EINVAL; | |
0037b49e | 3807 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3808 | return (serr); |
3809 | } | |
3810 | if (terr) { | |
3811 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3812 | iput(ZTOI(szp)); |
428870ff BB |
3813 | |
3814 | if (sdzp == tdzp) | |
3815 | rw_exit(&sdzp->z_name_lock); | |
3816 | ||
34dc7c2f BB |
3817 | if (strcmp(tnm, "..") == 0) |
3818 | terr = EINVAL; | |
0037b49e | 3819 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3820 | return (terr); |
3821 | } | |
3822 | ||
9c5167d1 NF |
3823 | /* |
3824 | * If we are using project inheritance, means if the directory has | |
3825 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
3826 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
3827 | * such case, we only allow renames into our tree when the project | |
3828 | * IDs are the same. | |
3829 | */ | |
3830 | if (tdzp->z_pflags & ZFS_PROJINHERIT && | |
3831 | tdzp->z_projid != szp->z_projid) { | |
3832 | error = SET_ERROR(EXDEV); | |
3833 | goto out; | |
3834 | } | |
3835 | ||
34dc7c2f BB |
3836 | /* |
3837 | * Must have write access at the source to remove the old entry | |
3838 | * and write access at the target to create the new entry. | |
3839 | * Note that if target and source are the same, this can be | |
3840 | * done in a single check. | |
3841 | */ | |
3842 | ||
149e873a | 3843 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3844 | goto out; |
3845 | ||
3558fd73 | 3846 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3847 | /* |
3848 | * Check to make sure rename is valid. | |
3849 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3850 | */ | |
149e873a | 3851 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3852 | goto out; |
3853 | } | |
3854 | ||
3855 | /* | |
3856 | * Does target exist? | |
3857 | */ | |
3858 | if (tzp) { | |
3859 | /* | |
3860 | * Source and target must be the same type. | |
3861 | */ | |
3558fd73 BB |
3862 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3863 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
2e528b49 | 3864 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3865 | goto out; |
3866 | } | |
3867 | } else { | |
3558fd73 | 3868 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
2e528b49 | 3869 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3870 | goto out; |
3871 | } | |
3872 | } | |
3873 | /* | |
3874 | * POSIX dictates that when the source and target | |
3875 | * entries refer to the same file object, rename | |
3876 | * must do nothing and exit without error. | |
3877 | */ | |
3878 | if (szp->z_id == tzp->z_id) { | |
3879 | error = 0; | |
3880 | goto out; | |
3881 | } | |
3882 | } | |
3883 | ||
0037b49e | 3884 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
3885 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3886 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3887 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3888 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3889 | if (sdzp != tdzp) { |
3890 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3891 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3892 | } | |
3893 | if (tzp) { | |
3894 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3895 | zfs_sa_upgrade_txholds(tx, tzp); | |
3896 | } | |
3897 | ||
3898 | zfs_sa_upgrade_txholds(tx, szp); | |
0037b49e | 3899 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
0735ecb3 | 3900 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
3901 | if (error) { |
3902 | if (zl != NULL) | |
3903 | zfs_rename_unlock(&zl); | |
3904 | zfs_dirent_unlock(sdl); | |
3905 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3906 | |
3907 | if (sdzp == tdzp) | |
3908 | rw_exit(&sdzp->z_name_lock); | |
3909 | ||
fb5f0bc8 | 3910 | if (error == ERESTART) { |
e8b96c60 | 3911 | waited = B_TRUE; |
34dc7c2f BB |
3912 | dmu_tx_wait(tx); |
3913 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3914 | iput(ZTOI(szp)); |
3915 | if (tzp) | |
3916 | iput(ZTOI(tzp)); | |
34dc7c2f BB |
3917 | goto top; |
3918 | } | |
3919 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3920 | iput(ZTOI(szp)); |
3921 | if (tzp) | |
3922 | iput(ZTOI(tzp)); | |
0037b49e | 3923 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3924 | return (error); |
3925 | } | |
3926 | ||
3927 | if (tzp) /* Attempt to remove the existing target */ | |
3928 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3929 | ||
3930 | if (error == 0) { | |
3931 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3932 | if (error == 0) { | |
428870ff | 3933 | szp->z_pflags |= ZFS_AV_MODIFIED; |
9c5167d1 NF |
3934 | if (tdzp->z_pflags & ZFS_PROJINHERIT) |
3935 | szp->z_pflags |= ZFS_PROJINHERIT; | |
34dc7c2f | 3936 | |
0037b49e | 3937 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 3938 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c99c9001 | 3939 | ASSERT0(error); |
34dc7c2f | 3940 | |
428870ff BB |
3941 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3942 | if (error == 0) { | |
3943 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3944 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3945 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3946 | } else { |
3947 | /* | |
3948 | * At this point, we have successfully created | |
3949 | * the target name, but have failed to remove | |
3950 | * the source name. Since the create was done | |
3951 | * with the ZRENAMING flag, there are | |
3952 | * complications; for one, the link count is | |
3953 | * wrong. The easiest way to deal with this | |
3954 | * is to remove the newly created target, and | |
3955 | * return the original error. This must | |
3956 | * succeed; fortunately, it is very unlikely to | |
3957 | * fail, since we just created it. | |
3958 | */ | |
3959 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3960 | ZRENAMING, NULL), ==, 0); | |
3961 | } | |
cc63068e SB |
3962 | } else { |
3963 | /* | |
3964 | * If we had removed the existing target, subsequent | |
3965 | * call to zfs_link_create() to add back the same entry | |
3966 | * but, the new dnode (szp) should not fail. | |
3967 | */ | |
3968 | ASSERT(tzp == NULL); | |
34dc7c2f BB |
3969 | } |
3970 | } | |
3971 | ||
3972 | dmu_tx_commit(tx); | |
3973 | out: | |
3974 | if (zl != NULL) | |
3975 | zfs_rename_unlock(&zl); | |
3976 | ||
3977 | zfs_dirent_unlock(sdl); | |
3978 | zfs_dirent_unlock(tdl); | |
3979 | ||
960e08fe | 3980 | zfs_inode_update(sdzp); |
428870ff BB |
3981 | if (sdzp == tdzp) |
3982 | rw_exit(&sdzp->z_name_lock); | |
3983 | ||
960e08fe BB |
3984 | if (sdzp != tdzp) |
3985 | zfs_inode_update(tdzp); | |
428870ff | 3986 | |
960e08fe | 3987 | zfs_inode_update(szp); |
3558fd73 | 3988 | iput(ZTOI(szp)); |
960e08fe BB |
3989 | if (tzp) { |
3990 | zfs_inode_update(tzp); | |
3558fd73 | 3991 | iput(ZTOI(tzp)); |
960e08fe | 3992 | } |
34dc7c2f | 3993 | |
0037b49e | 3994 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3995 | zil_commit(zilog, 0); |
428870ff | 3996 | |
0037b49e | 3997 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3998 | return (error); |
3999 | } | |
4000 | ||
4001 | /* | |
4002 | * Insert the indicated symbolic reference entry into the directory. | |
4003 | * | |
3558fd73 | 4004 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
4005 | * link - Name for new symlink entry. |
4006 | * vap - Attributes of new entry. | |
4007 | * target - Target path of new symlink. | |
3558fd73 | 4008 | * |
34dc7c2f | 4009 | * cr - credentials of caller. |
34dc7c2f BB |
4010 | * flags - case flags |
4011 | * | |
d3cc8b15 | 4012 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4013 | * |
4014 | * Timestamps: | |
3558fd73 | 4015 | * dip - ctime|mtime updated |
34dc7c2f BB |
4016 | */ |
4017 | /*ARGSUSED*/ | |
e5c39b95 | 4018 | int |
3558fd73 BB |
4019 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
4020 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 4021 | { |
3558fd73 | 4022 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
4023 | zfs_dirlock_t *dl; |
4024 | dmu_tx_t *tx; | |
0037b49e | 4025 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 4026 | zilog_t *zilog; |
428870ff | 4027 | uint64_t len = strlen(link); |
34dc7c2f BB |
4028 | int error; |
4029 | int zflg = ZNEW; | |
9babb374 BB |
4030 | zfs_acl_ids_t acl_ids; |
4031 | boolean_t fuid_dirtied; | |
428870ff | 4032 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 4033 | boolean_t waited = B_FALSE; |
34dc7c2f | 4034 | |
3558fd73 | 4035 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 4036 | |
32dec7bd | 4037 | if (name == NULL) |
4038 | return (SET_ERROR(EINVAL)); | |
4039 | ||
0037b49e | 4040 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4041 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 4042 | zilog = zfsvfs->z_log; |
34dc7c2f | 4043 | |
0037b49e | 4044 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 4045 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 4046 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4047 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
4048 | } |
4049 | if (flags & FIGNORECASE) | |
4050 | zflg |= ZCILOOK; | |
34dc7c2f BB |
4051 | |
4052 | if (len > MAXPATHLEN) { | |
0037b49e | 4053 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4054 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
4055 | } |
4056 | ||
428870ff BB |
4057 | if ((error = zfs_acl_ids_create(dzp, 0, |
4058 | vap, cr, NULL, &acl_ids)) != 0) { | |
0037b49e | 4059 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4060 | return (error); |
4061 | } | |
4062 | top: | |
3558fd73 BB |
4063 | *ipp = NULL; |
4064 | ||
34dc7c2f BB |
4065 | /* |
4066 | * Attempt to lock directory; fail if entry already exists. | |
4067 | */ | |
4068 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
4069 | if (error) { | |
428870ff | 4070 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 4071 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4072 | return (error); |
4073 | } | |
4074 | ||
149e873a | 4075 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
4076 | zfs_acl_ids_free(&acl_ids); |
4077 | zfs_dirent_unlock(dl); | |
0037b49e | 4078 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4079 | return (error); |
4080 | } | |
4081 | ||
9c5167d1 | 4082 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, ZFS_DEFAULT_PROJID)) { |
9babb374 BB |
4083 | zfs_acl_ids_free(&acl_ids); |
4084 | zfs_dirent_unlock(dl); | |
0037b49e | 4085 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4086 | return (SET_ERROR(EDQUOT)); |
9babb374 | 4087 | } |
0037b49e BB |
4088 | tx = dmu_tx_create(zfsvfs->z_os); |
4089 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
34dc7c2f | 4090 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 4091 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
4092 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
4093 | ZFS_SA_BASE_ATTR_SIZE + len); | |
4094 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
0037b49e | 4095 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
4096 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
4097 | acl_ids.z_aclp->z_acl_bytes); | |
4098 | } | |
9babb374 | 4099 | if (fuid_dirtied) |
0037b49e | 4100 | zfs_fuid_txhold(zfsvfs, tx); |
0735ecb3 | 4101 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
4102 | if (error) { |
4103 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4104 | if (error == ERESTART) { |
e8b96c60 | 4105 | waited = B_TRUE; |
34dc7c2f BB |
4106 | dmu_tx_wait(tx); |
4107 | dmu_tx_abort(tx); | |
4108 | goto top; | |
4109 | } | |
428870ff | 4110 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 4111 | dmu_tx_abort(tx); |
0037b49e | 4112 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4113 | return (error); |
4114 | } | |
4115 | ||
34dc7c2f BB |
4116 | /* |
4117 | * Create a new object for the symlink. | |
428870ff | 4118 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 4119 | */ |
428870ff | 4120 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 4121 | |
428870ff | 4122 | if (fuid_dirtied) |
0037b49e | 4123 | zfs_fuid_sync(zfsvfs, tx); |
34dc7c2f | 4124 | |
572e2857 | 4125 | mutex_enter(&zp->z_lock); |
428870ff | 4126 | if (zp->z_is_sa) |
0037b49e | 4127 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), |
428870ff BB |
4128 | link, len, tx); |
4129 | else | |
4130 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 4131 | mutex_exit(&zp->z_lock); |
34dc7c2f | 4132 | |
428870ff | 4133 | zp->z_size = len; |
0037b49e | 4134 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 4135 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
4136 | /* |
4137 | * Insert the new object into the directory. | |
4138 | */ | |
cc63068e SB |
4139 | error = zfs_link_create(dl, zp, tx, ZNEW); |
4140 | if (error != 0) { | |
4141 | zfs_znode_delete(zp, tx); | |
4142 | remove_inode_hash(ZTOI(zp)); | |
4143 | } else { | |
4144 | if (flags & FIGNORECASE) | |
4145 | txtype |= TX_CI; | |
4146 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
9babb374 | 4147 | |
cc63068e SB |
4148 | zfs_inode_update(dzp); |
4149 | zfs_inode_update(zp); | |
4150 | } | |
960e08fe | 4151 | |
9babb374 | 4152 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
4153 | |
4154 | dmu_tx_commit(tx); | |
4155 | ||
4156 | zfs_dirent_unlock(dl); | |
4157 | ||
cc63068e SB |
4158 | if (error == 0) { |
4159 | *ipp = ZTOI(zp); | |
34dc7c2f | 4160 | |
cc63068e SB |
4161 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
4162 | zil_commit(zilog, 0); | |
4163 | } else { | |
4164 | iput(ZTOI(zp)); | |
4165 | } | |
428870ff | 4166 | |
0037b49e | 4167 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4168 | return (error); |
4169 | } | |
4170 | ||
4171 | /* | |
4172 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 4173 | * the symbolic path referred to by ip. |
34dc7c2f | 4174 | * |
8b4f9a2d BB |
4175 | * IN: ip - inode of symbolic link |
4176 | * uio - structure to contain the link path. | |
4177 | * cr - credentials of caller. | |
34dc7c2f BB |
4178 | * |
4179 | * RETURN: 0 if success | |
4180 | * error code if failure | |
4181 | * | |
4182 | * Timestamps: | |
3558fd73 | 4183 | * ip - atime updated |
34dc7c2f BB |
4184 | */ |
4185 | /* ARGSUSED */ | |
e5c39b95 | 4186 | int |
8b4f9a2d | 4187 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 4188 | { |
3558fd73 | 4189 | znode_t *zp = ITOZ(ip); |
0037b49e | 4190 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4191 | int error; |
4192 | ||
0037b49e | 4193 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4194 | ZFS_VERIFY_ZP(zp); |
4195 | ||
572e2857 | 4196 | mutex_enter(&zp->z_lock); |
428870ff | 4197 | if (zp->z_is_sa) |
8b4f9a2d | 4198 | error = sa_lookup_uio(zp->z_sa_hdl, |
0037b49e | 4199 | SA_ZPL_SYMLINK(zfsvfs), uio); |
428870ff | 4200 | else |
8b4f9a2d | 4201 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 4202 | mutex_exit(&zp->z_lock); |
34dc7c2f | 4203 | |
0037b49e | 4204 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4205 | return (error); |
4206 | } | |
4207 | ||
4208 | /* | |
3558fd73 | 4209 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 4210 | * |
3558fd73 BB |
4211 | * IN: tdip - Directory to contain new entry. |
4212 | * sip - inode of new entry. | |
34dc7c2f BB |
4213 | * name - name of new entry. |
4214 | * cr - credentials of caller. | |
34dc7c2f BB |
4215 | * |
4216 | * RETURN: 0 if success | |
4217 | * error code if failure | |
4218 | * | |
4219 | * Timestamps: | |
3558fd73 BB |
4220 | * tdip - ctime|mtime updated |
4221 | * sip - ctime updated | |
34dc7c2f BB |
4222 | */ |
4223 | /* ARGSUSED */ | |
e5c39b95 | 4224 | int |
da5e151f BB |
4225 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr, |
4226 | int flags) | |
34dc7c2f | 4227 | { |
3558fd73 | 4228 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 4229 | znode_t *tzp, *szp; |
0037b49e | 4230 | zfsvfs_t *zfsvfs = ITOZSB(tdip); |
34dc7c2f BB |
4231 | zilog_t *zilog; |
4232 | zfs_dirlock_t *dl; | |
4233 | dmu_tx_t *tx; | |
34dc7c2f BB |
4234 | int error; |
4235 | int zf = ZNEW; | |
428870ff | 4236 | uint64_t parent; |
572e2857 | 4237 | uid_t owner; |
e8b96c60 | 4238 | boolean_t waited = B_FALSE; |
ace1eae8 CC |
4239 | boolean_t is_tmpfile = 0; |
4240 | uint64_t txg; | |
4241 | #ifdef HAVE_TMPFILE | |
4242 | is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); | |
4243 | #endif | |
3558fd73 | 4244 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 4245 | |
32dec7bd | 4246 | if (name == NULL) |
4247 | return (SET_ERROR(EINVAL)); | |
4248 | ||
0037b49e | 4249 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4250 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 4251 | zilog = zfsvfs->z_log; |
34dc7c2f | 4252 | |
428870ff BB |
4253 | /* |
4254 | * POSIX dictates that we return EPERM here. | |
4255 | * Better choices include ENOTSUP or EISDIR. | |
4256 | */ | |
3558fd73 | 4257 | if (S_ISDIR(sip->i_mode)) { |
0037b49e | 4258 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4259 | return (SET_ERROR(EPERM)); |
428870ff BB |
4260 | } |
4261 | ||
812e91a7 MT |
4262 | szp = ITOZ(sip); |
4263 | ZFS_VERIFY_ZP(szp); | |
4264 | ||
9c5167d1 NF |
4265 | /* |
4266 | * If we are using project inheritance, means if the directory has | |
4267 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
4268 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
4269 | * such case, we only allow hard link creation in our tree when the | |
4270 | * project IDs are the same. | |
4271 | */ | |
4272 | if (dzp->z_pflags & ZFS_PROJINHERIT && dzp->z_projid != szp->z_projid) { | |
4273 | ZFS_EXIT(zfsvfs); | |
4274 | return (SET_ERROR(EXDEV)); | |
4275 | } | |
4276 | ||
812e91a7 MT |
4277 | /* |
4278 | * We check i_sb because snapshots and the ctldir must have different | |
4279 | * super blocks. | |
4280 | */ | |
c0ebc844 | 4281 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
0037b49e | 4282 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4283 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 4284 | } |
428870ff | 4285 | |
428870ff BB |
4286 | /* Prevent links to .zfs/shares files */ |
4287 | ||
0037b49e | 4288 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff | 4289 | &parent, sizeof (uint64_t))) != 0) { |
0037b49e | 4290 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4291 | return (error); |
4292 | } | |
0037b49e BB |
4293 | if (parent == zfsvfs->z_shares_dir) { |
4294 | ZFS_EXIT(zfsvfs); | |
2e528b49 | 4295 | return (SET_ERROR(EPERM)); |
428870ff BB |
4296 | } |
4297 | ||
0037b49e | 4298 | if (zfsvfs->z_utf8 && u8_validate(name, |
34dc7c2f | 4299 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 4300 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4301 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
4302 | } |
4303 | if (flags & FIGNORECASE) | |
4304 | zf |= ZCILOOK; | |
4305 | ||
34dc7c2f BB |
4306 | /* |
4307 | * We do not support links between attributes and non-attributes | |
4308 | * because of the potential security risk of creating links | |
4309 | * into "normal" file space in order to circumvent restrictions | |
4310 | * imposed in attribute space. | |
4311 | */ | |
428870ff | 4312 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 4313 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4314 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4315 | } |
4316 | ||
0037b49e BB |
4317 | owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid), |
4318 | cr, ZFS_OWNER); | |
572e2857 | 4319 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
0037b49e | 4320 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4321 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4322 | } |
4323 | ||
149e873a | 4324 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
0037b49e | 4325 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4326 | return (error); |
4327 | } | |
4328 | ||
428870ff | 4329 | top: |
34dc7c2f BB |
4330 | /* |
4331 | * Attempt to lock directory; fail if entry already exists. | |
4332 | */ | |
4333 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
4334 | if (error) { | |
0037b49e | 4335 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4336 | return (error); |
4337 | } | |
4338 | ||
0037b49e | 4339 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 4340 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 4341 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
ace1eae8 | 4342 | if (is_tmpfile) |
0037b49e | 4343 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 4344 | |
428870ff BB |
4345 | zfs_sa_upgrade_txholds(tx, szp); |
4346 | zfs_sa_upgrade_txholds(tx, dzp); | |
0735ecb3 | 4347 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
4348 | if (error) { |
4349 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4350 | if (error == ERESTART) { |
e8b96c60 | 4351 | waited = B_TRUE; |
34dc7c2f BB |
4352 | dmu_tx_wait(tx); |
4353 | dmu_tx_abort(tx); | |
4354 | goto top; | |
4355 | } | |
4356 | dmu_tx_abort(tx); | |
0037b49e | 4357 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4358 | return (error); |
4359 | } | |
ace1eae8 CC |
4360 | /* unmark z_unlinked so zfs_link_create will not reject */ |
4361 | if (is_tmpfile) | |
4362 | szp->z_unlinked = 0; | |
34dc7c2f BB |
4363 | error = zfs_link_create(dl, szp, tx, 0); |
4364 | ||
4365 | if (error == 0) { | |
4366 | uint64_t txtype = TX_LINK; | |
ace1eae8 CC |
4367 | /* |
4368 | * tmpfile is created to be in z_unlinkedobj, so remove it. | |
4369 | * Also, we don't log in ZIL, be cause all previous file | |
4370 | * operation on the tmpfile are ignored by ZIL. Instead we | |
4371 | * always wait for txg to sync to make sure all previous | |
4372 | * operation are sync safe. | |
4373 | */ | |
4374 | if (is_tmpfile) { | |
0037b49e BB |
4375 | VERIFY(zap_remove_int(zfsvfs->z_os, |
4376 | zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); | |
ace1eae8 CC |
4377 | } else { |
4378 | if (flags & FIGNORECASE) | |
4379 | txtype |= TX_CI; | |
4380 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); | |
4381 | } | |
4382 | } else if (is_tmpfile) { | |
4383 | /* restore z_unlinked since when linking failed */ | |
4384 | szp->z_unlinked = 1; | |
34dc7c2f | 4385 | } |
ace1eae8 | 4386 | txg = dmu_tx_get_txg(tx); |
34dc7c2f BB |
4387 | dmu_tx_commit(tx); |
4388 | ||
4389 | zfs_dirent_unlock(dl); | |
4390 | ||
0037b49e | 4391 | if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4392 | zil_commit(zilog, 0); |
428870ff | 4393 | |
ace1eae8 | 4394 | if (is_tmpfile) |
0037b49e | 4395 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg); |
ace1eae8 | 4396 | |
960e08fe BB |
4397 | zfs_inode_update(dzp); |
4398 | zfs_inode_update(szp); | |
0037b49e | 4399 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4400 | return (error); |
4401 | } | |
4402 | ||
3c0e5c0f | 4403 | static void |
119a394a | 4404 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
4405 | { |
4406 | struct page *pp = arg; | |
4407 | ||
119a394a | 4408 | ClearPageError(pp); |
3c0e5c0f BB |
4409 | end_page_writeback(pp); |
4410 | } | |
4411 | ||
34dc7c2f | 4412 | /* |
3c0e5c0f BB |
4413 | * Push a page out to disk, once the page is on stable storage the |
4414 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 4415 | * |
3c0e5c0f BB |
4416 | * IN: ip - page mapped for inode. |
4417 | * pp - page to push (page is locked) | |
4418 | * wbc - writeback control data | |
34dc7c2f BB |
4419 | * |
4420 | * RETURN: 0 if success | |
4421 | * error code if failure | |
4422 | * | |
3c0e5c0f BB |
4423 | * Timestamps: |
4424 | * ip - ctime|mtime updated | |
34dc7c2f BB |
4425 | */ |
4426 | /* ARGSUSED */ | |
3c0e5c0f BB |
4427 | int |
4428 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 4429 | { |
3c0e5c0f | 4430 | znode_t *zp = ITOZ(ip); |
0037b49e | 4431 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
3c0e5c0f BB |
4432 | loff_t offset; |
4433 | loff_t pgoff; | |
4c837f0d BB |
4434 | unsigned int pglen; |
4435 | rl_t *rl; | |
3c0e5c0f BB |
4436 | dmu_tx_t *tx; |
4437 | caddr_t va; | |
4438 | int err = 0; | |
4439 | uint64_t mtime[2], ctime[2]; | |
4440 | sa_bulk_attr_t bulk[3]; | |
4441 | int cnt = 0; | |
21a96fb6 | 4442 | struct address_space *mapping; |
3c0e5c0f | 4443 | |
0037b49e | 4444 | ZFS_ENTER(zfsvfs); |
4c837f0d | 4445 | ZFS_VERIFY_ZP(zp); |
d164b209 | 4446 | |
3c0e5c0f BB |
4447 | ASSERT(PageLocked(pp)); |
4448 | ||
d1d7e268 MK |
4449 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
4450 | offset = i_size_read(ip); /* File length in bytes */ | |
8b1899d3 BB |
4451 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
4452 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
4453 | |
4454 | /* Page is beyond end of file */ | |
4455 | if (pgoff >= offset) { | |
4456 | unlock_page(pp); | |
0037b49e | 4457 | ZFS_EXIT(zfsvfs); |
3c0e5c0f BB |
4458 | return (0); |
4459 | } | |
4460 | ||
4461 | /* Truncate page length to end of file */ | |
4462 | if (pgoff + pglen > offset) | |
4463 | pglen = offset - pgoff; | |
4464 | ||
4465 | #if 0 | |
34dc7c2f | 4466 | /* |
3c0e5c0f BB |
4467 | * FIXME: Allow mmap writes past its quota. The correct fix |
4468 | * is to register a page_mkwrite() handler to count the page | |
4469 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 4470 | */ |
9c5167d1 NF |
4471 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
4472 | KUID_TO_SUID(ip->i_uid)) || | |
4473 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
4474 | KGID_TO_SGID(ip->i_gid)) || | |
4475 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
4476 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
4477 | zp->z_projid))) { | |
9babb374 | 4478 | err = EDQUOT; |
9babb374 | 4479 | } |
3c0e5c0f BB |
4480 | #endif |
4481 | ||
d958324f BB |
4482 | /* |
4483 | * The ordering here is critical and must adhere to the following | |
4484 | * rules in order to avoid deadlocking in either zfs_read() or | |
4485 | * zfs_free_range() due to a lock inversion. | |
4486 | * | |
4487 | * 1) The page must be unlocked prior to acquiring the range lock. | |
4488 | * This is critical because zfs_read() calls find_lock_page() | |
4489 | * which may block on the page lock while holding the range lock. | |
4490 | * | |
4491 | * 2) Before setting or clearing write back on a page the range lock | |
4492 | * must be held in order to prevent a lock inversion with the | |
4493 | * zfs_free_range() function. | |
21a96fb6 CC |
4494 | * |
4495 | * This presents a problem because upon entering this function the | |
4496 | * page lock is already held. To safely acquire the range lock the | |
4497 | * page lock must be dropped. This creates a window where another | |
4498 | * process could truncate, invalidate, dirty, or write out the page. | |
4499 | * | |
4500 | * Therefore, after successfully reacquiring the range and page locks | |
4501 | * the current page state is checked. In the common case everything | |
4502 | * will be as is expected and it can be written out. However, if | |
4503 | * the page state has changed it must be handled accordingly. | |
d958324f | 4504 | */ |
21a96fb6 CC |
4505 | mapping = pp->mapping; |
4506 | redirty_page_for_writepage(wbc, pp); | |
d958324f | 4507 | unlock_page(pp); |
21a96fb6 | 4508 | |
d88895a0 | 4509 | rl = zfs_range_lock(&zp->z_range_lock, pgoff, pglen, RL_WRITER); |
21a96fb6 CC |
4510 | lock_page(pp); |
4511 | ||
4512 | /* Page mapping changed or it was no longer dirty, we're done */ | |
4513 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
4514 | unlock_page(pp); | |
4515 | zfs_range_unlock(rl); | |
0037b49e | 4516 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4517 | return (0); |
4518 | } | |
4519 | ||
4520 | /* Another process started write block if required */ | |
4521 | if (PageWriteback(pp)) { | |
4522 | unlock_page(pp); | |
4523 | zfs_range_unlock(rl); | |
4524 | ||
4525 | if (wbc->sync_mode != WB_SYNC_NONE) | |
4526 | wait_on_page_writeback(pp); | |
4527 | ||
0037b49e | 4528 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4529 | return (0); |
4530 | } | |
4531 | ||
4532 | /* Clear the dirty flag the required locks are held */ | |
4533 | if (!clear_page_dirty_for_io(pp)) { | |
4534 | unlock_page(pp); | |
4535 | zfs_range_unlock(rl); | |
0037b49e | 4536 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4537 | return (0); |
4538 | } | |
4539 | ||
4540 | /* | |
4541 | * Counterpart for redirty_page_for_writepage() above. This page | |
4542 | * was in fact not skipped and should not be counted as if it were. | |
4543 | */ | |
4544 | wbc->pages_skipped--; | |
3c0e5c0f | 4545 | set_page_writeback(pp); |
21a96fb6 | 4546 | unlock_page(pp); |
3c0e5c0f | 4547 | |
0037b49e | 4548 | tx = dmu_tx_create(zfsvfs->z_os); |
3c0e5c0f | 4549 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
4550 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4551 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 4552 | |
fb5f0bc8 | 4553 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 4554 | if (err != 0) { |
3c0e5c0f | 4555 | if (err == ERESTART) |
34dc7c2f | 4556 | dmu_tx_wait(tx); |
3c0e5c0f | 4557 | |
34dc7c2f | 4558 | dmu_tx_abort(tx); |
119a394a ED |
4559 | __set_page_dirty_nobuffers(pp); |
4560 | ClearPageError(pp); | |
4561 | end_page_writeback(pp); | |
4c837f0d | 4562 | zfs_range_unlock(rl); |
0037b49e | 4563 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4564 | return (err); |
34dc7c2f BB |
4565 | } |
4566 | ||
dde471ef | 4567 | va = kmap(pp); |
8b1899d3 | 4568 | ASSERT3U(pglen, <=, PAGE_SIZE); |
0037b49e | 4569 | dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 4570 | kunmap(pp); |
34dc7c2f | 4571 | |
0037b49e BB |
4572 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
4573 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
4574 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, | |
4575 | &zp->z_pflags, 8); | |
428870ff | 4576 | |
d3aa3ea9 BB |
4577 | /* Preserve the mtime and ctime provided by the inode */ |
4578 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4579 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
4580 | zp->z_atime_dirty = 0; | |
4581 | zp->z_seq++; | |
4582 | ||
4583 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4584 | ||
0037b49e | 4585 | zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
119a394a | 4586 | zfs_putpage_commit_cb, pp); |
45d1cae3 | 4587 | dmu_tx_commit(tx); |
d3aa3ea9 | 4588 | |
4c837f0d | 4589 | zfs_range_unlock(rl); |
34dc7c2f | 4590 | |
119a394a ED |
4591 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4592 | /* | |
4593 | * Note that this is rarely called under writepages(), because | |
4594 | * writepages() normally handles the entire commit for | |
4595 | * performance reasons. | |
4596 | */ | |
0037b49e | 4597 | zil_commit(zfsvfs->z_log, zp->z_id); |
2b286136 | 4598 | } |
3c0e5c0f | 4599 | |
0037b49e | 4600 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4601 | return (err); |
34dc7c2f BB |
4602 | } |
4603 | ||
8780c539 BB |
4604 | /* |
4605 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 4606 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
4607 | */ |
4608 | int | |
4609 | zfs_dirty_inode(struct inode *ip, int flags) | |
4610 | { | |
4611 | znode_t *zp = ITOZ(ip); | |
0037b49e | 4612 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
8780c539 | 4613 | dmu_tx_t *tx; |
023699cd MM |
4614 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
4615 | sa_bulk_attr_t bulk[4]; | |
704cd075 | 4616 | int error = 0; |
8780c539 BB |
4617 | int cnt = 0; |
4618 | ||
0037b49e | 4619 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
c944be5d BB |
4620 | return (0); |
4621 | ||
0037b49e | 4622 | ZFS_ENTER(zfsvfs); |
8780c539 BB |
4623 | ZFS_VERIFY_ZP(zp); |
4624 | ||
704cd075 CC |
4625 | #ifdef I_DIRTY_TIME |
4626 | /* | |
4627 | * This is the lazytime semantic indroduced in Linux 4.0 | |
4628 | * This flag will only be called from update_time when lazytime is set. | |
4629 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
4630 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
4631 | * only need to dirty atime. | |
4632 | */ | |
4633 | if (flags == I_DIRTY_TIME) { | |
4634 | zp->z_atime_dirty = 1; | |
4635 | goto out; | |
4636 | } | |
4637 | #endif | |
4638 | ||
0037b49e | 4639 | tx = dmu_tx_create(zfsvfs->z_os); |
8780c539 BB |
4640 | |
4641 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4642 | zfs_sa_upgrade_txholds(tx, zp); | |
4643 | ||
4644 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4645 | if (error) { | |
4646 | dmu_tx_abort(tx); | |
4647 | goto out; | |
4648 | } | |
4649 | ||
4650 | mutex_enter(&zp->z_lock); | |
704cd075 CC |
4651 | zp->z_atime_dirty = 0; |
4652 | ||
0037b49e BB |
4653 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
4654 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
4655 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
4656 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
8780c539 | 4657 | |
023699cd | 4658 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4659 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4660 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4661 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
023699cd MM |
4662 | mode = ip->i_mode; |
4663 | ||
4664 | zp->z_mode = mode; | |
8780c539 BB |
4665 | |
4666 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4667 | mutex_exit(&zp->z_lock); | |
4668 | ||
4669 | dmu_tx_commit(tx); | |
4670 | out: | |
0037b49e | 4671 | ZFS_EXIT(zfsvfs); |
8780c539 BB |
4672 | return (error); |
4673 | } | |
8780c539 | 4674 | |
34dc7c2f BB |
4675 | /*ARGSUSED*/ |
4676 | void | |
c0d35759 | 4677 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4678 | { |
c0d35759 | 4679 | znode_t *zp = ITOZ(ip); |
0037b49e | 4680 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
0df9673f | 4681 | uint64_t atime[2]; |
34dc7c2f | 4682 | int error; |
cafbd2ac | 4683 | int need_unlock = 0; |
34dc7c2f | 4684 | |
cafbd2ac | 4685 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
0037b49e | 4686 | if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) { |
cafbd2ac | 4687 | need_unlock = 1; |
0037b49e | 4688 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); |
cafbd2ac | 4689 | } |
c0d35759 | 4690 | if (zp->z_sa_hdl == NULL) { |
cafbd2ac | 4691 | if (need_unlock) |
0037b49e | 4692 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
c0d35759 | 4693 | return; |
34dc7c2f BB |
4694 | } |
4695 | ||
4696 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
0037b49e | 4697 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 4698 | |
428870ff BB |
4699 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4700 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4701 | error = dmu_tx_assign(tx, TXG_WAIT); |
4702 | if (error) { | |
4703 | dmu_tx_abort(tx); | |
4704 | } else { | |
0df9673f | 4705 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
34dc7c2f | 4706 | mutex_enter(&zp->z_lock); |
0037b49e | 4707 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), |
0df9673f | 4708 | (void *)&atime, sizeof (atime), tx); |
34dc7c2f BB |
4709 | zp->z_atime_dirty = 0; |
4710 | mutex_exit(&zp->z_lock); | |
4711 | dmu_tx_commit(tx); | |
4712 | } | |
4713 | } | |
4714 | ||
4715 | zfs_zinactive(zp); | |
cafbd2ac | 4716 | if (need_unlock) |
0037b49e | 4717 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
34dc7c2f BB |
4718 | } |
4719 | ||
4720 | /* | |
4721 | * Bounds-check the seek operation. | |
4722 | * | |
3558fd73 | 4723 | * IN: ip - inode seeking within |
34dc7c2f BB |
4724 | * ooff - old file offset |
4725 | * noffp - pointer to new file offset | |
4726 | * ct - caller context | |
4727 | * | |
4728 | * RETURN: 0 if success | |
4729 | * EINVAL if new offset invalid | |
4730 | */ | |
4731 | /* ARGSUSED */ | |
3558fd73 | 4732 | int |
9623f736 | 4733 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4734 | { |
3558fd73 | 4735 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4736 | return (0); |
4737 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4738 | } | |
4739 | ||
34dc7c2f | 4740 | /* |
dde471ef | 4741 | * Fill pages with data from the disk. |
34dc7c2f BB |
4742 | */ |
4743 | static int | |
dde471ef | 4744 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4745 | { |
d1d7e268 | 4746 | znode_t *zp = ITOZ(ip); |
0037b49e | 4747 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
d1d7e268 | 4748 | objset_t *os; |
dde471ef | 4749 | struct page *cur_pp; |
d1d7e268 MK |
4750 | u_offset_t io_off, total; |
4751 | size_t io_len; | |
4752 | loff_t i_size; | |
4753 | unsigned page_idx; | |
4754 | int err; | |
34dc7c2f | 4755 | |
0037b49e | 4756 | os = zfsvfs->z_os; |
8b1899d3 | 4757 | io_len = nr_pages << PAGE_SHIFT; |
dde471ef PJ |
4758 | i_size = i_size_read(ip); |
4759 | io_off = page_offset(pl[0]); | |
4760 | ||
4761 | if (io_off + io_len > i_size) | |
4762 | io_len = i_size - io_off; | |
34dc7c2f BB |
4763 | |
4764 | /* | |
dde471ef | 4765 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4766 | */ |
dde471ef | 4767 | page_idx = 0; |
34dc7c2f | 4768 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4769 | caddr_t va; |
4770 | ||
540c3927 | 4771 | cur_pp = pl[page_idx++]; |
dde471ef | 4772 | va = kmap(cur_pp); |
9babb374 BB |
4773 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4774 | DMU_READ_PREFETCH); | |
dde471ef | 4775 | kunmap(cur_pp); |
34dc7c2f | 4776 | if (err) { |
b128c09f BB |
4777 | /* convert checksum errors into IO errors */ |
4778 | if (err == ECKSUM) | |
2e528b49 | 4779 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4780 | return (err); |
4781 | } | |
34dc7c2f | 4782 | } |
d164b209 | 4783 | |
34dc7c2f BB |
4784 | return (0); |
4785 | } | |
4786 | ||
4787 | /* | |
dde471ef | 4788 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4789 | * |
dde471ef PJ |
4790 | * IN: ip - inode of file to get data from. |
4791 | * pl - list of pages to read | |
4792 | * nr_pages - number of pages to read | |
34dc7c2f | 4793 | * |
d3cc8b15 | 4794 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4795 | * |
4796 | * Timestamps: | |
4797 | * vp - atime updated | |
4798 | */ | |
4799 | /* ARGSUSED */ | |
dde471ef PJ |
4800 | int |
4801 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4802 | { |
dde471ef | 4803 | znode_t *zp = ITOZ(ip); |
0037b49e | 4804 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
dde471ef | 4805 | int err; |
d164b209 | 4806 | |
d164b209 BB |
4807 | if (pl == NULL) |
4808 | return (0); | |
34dc7c2f | 4809 | |
0037b49e | 4810 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4811 | ZFS_VERIFY_ZP(zp); |
4812 | ||
dde471ef | 4813 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4814 | |
0037b49e | 4815 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4816 | return (err); |
4817 | } | |
4818 | ||
4819 | /* | |
e2e7aa2d | 4820 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4821 | * |
e2e7aa2d BB |
4822 | * IN: ip - inode of the file to mmap |
4823 | * off - file offset | |
4824 | * addrp - start address in memory region | |
4825 | * len - length of memory region | |
4826 | * vm_flags- address flags | |
34dc7c2f | 4827 | * |
e2e7aa2d BB |
4828 | * RETURN: 0 if success |
4829 | * error code if failure | |
34dc7c2f BB |
4830 | */ |
4831 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4832 | int |
4833 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4834 | unsigned long vm_flags) | |
34dc7c2f | 4835 | { |
e2e7aa2d | 4836 | znode_t *zp = ITOZ(ip); |
0037b49e | 4837 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4838 | |
0037b49e | 4839 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4840 | ZFS_VERIFY_ZP(zp); |
4841 | ||
e2e7aa2d | 4842 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4843 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
0037b49e | 4844 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4845 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4846 | } |
4847 | ||
e2e7aa2d | 4848 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4849 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
0037b49e | 4850 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4851 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4852 | } |
4853 | ||
34dc7c2f | 4854 | if (off < 0 || len > MAXOFFSET_T - off) { |
0037b49e | 4855 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4856 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4857 | } |
4858 | ||
0037b49e | 4859 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4860 | return (0); |
4861 | } | |
4862 | ||
3558fd73 BB |
4863 | /* |
4864 | * convoff - converts the given data (start, whence) to the | |
4865 | * given whence. | |
4866 | */ | |
4867 | int | |
4868 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4869 | { | |
5484965a | 4870 | vattr_t vap; |
3558fd73 BB |
4871 | int error; |
4872 | ||
4873 | if ((lckdat->l_whence == 2) || (whence == 2)) { | |
d95a5980 | 4874 | if ((error = zfs_getattr(ip, &vap, 0, CRED()))) |
3558fd73 BB |
4875 | return (error); |
4876 | } | |
4877 | ||
4878 | switch (lckdat->l_whence) { | |
4879 | case 1: | |
4880 | lckdat->l_start += offset; | |
4881 | break; | |
4882 | case 2: | |
5484965a | 4883 | lckdat->l_start += vap.va_size; |
3558fd73 BB |
4884 | /* FALLTHRU */ |
4885 | case 0: | |
4886 | break; | |
4887 | default: | |
2e528b49 | 4888 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4889 | } |
4890 | ||
4891 | if (lckdat->l_start < 0) | |
2e528b49 | 4892 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4893 | |
4894 | switch (whence) { | |
4895 | case 1: | |
4896 | lckdat->l_start -= offset; | |
4897 | break; | |
4898 | case 2: | |
5484965a | 4899 | lckdat->l_start -= vap.va_size; |
3558fd73 BB |
4900 | /* FALLTHRU */ |
4901 | case 0: | |
4902 | break; | |
4903 | default: | |
2e528b49 | 4904 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4905 | } |
4906 | ||
4907 | lckdat->l_whence = (short)whence; | |
4908 | return (0); | |
4909 | } | |
4910 | ||
34dc7c2f BB |
4911 | /* |
4912 | * Free or allocate space in a file. Currently, this function only | |
4913 | * supports the `F_FREESP' command. However, this command is somewhat | |
4914 | * misnamed, as its functionality includes the ability to allocate as | |
4915 | * well as free space. | |
4916 | * | |
3558fd73 | 4917 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4918 | * cmd - action to take (only F_FREESP supported). |
4919 | * bfp - section of file to free/alloc. | |
4920 | * flag - current file open mode flags. | |
4921 | * offset - current file offset. | |
4922 | * cr - credentials of caller [UNUSED]. | |
34dc7c2f | 4923 | * |
d3cc8b15 | 4924 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4925 | * |
4926 | * Timestamps: | |
3558fd73 | 4927 | * ip - ctime|mtime updated |
34dc7c2f BB |
4928 | */ |
4929 | /* ARGSUSED */ | |
e5c39b95 | 4930 | int |
3558fd73 BB |
4931 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4932 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4933 | { |
3558fd73 | 4934 | znode_t *zp = ITOZ(ip); |
0037b49e | 4935 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4936 | uint64_t off, len; |
4937 | int error; | |
4938 | ||
0037b49e | 4939 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4940 | ZFS_VERIFY_ZP(zp); |
4941 | ||
34dc7c2f | 4942 | if (cmd != F_FREESP) { |
0037b49e | 4943 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4944 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4945 | } |
4946 | ||
f3c9dca0 MT |
4947 | /* |
4948 | * Callers might not be able to detect properly that we are read-only, | |
4949 | * so check it explicitly here. | |
4950 | */ | |
0037b49e BB |
4951 | if (zfs_is_readonly(zfsvfs)) { |
4952 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
4953 | return (SET_ERROR(EROFS)); |
4954 | } | |
4955 | ||
3558fd73 | 4956 | if ((error = convoff(ip, bfp, 0, offset))) { |
0037b49e | 4957 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4958 | return (error); |
4959 | } | |
4960 | ||
4961 | if (bfp->l_len < 0) { | |
0037b49e | 4962 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4963 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4964 | } |
4965 | ||
aec69371 ED |
4966 | /* |
4967 | * Permissions aren't checked on Solaris because on this OS | |
4968 | * zfs_space() can only be called with an opened file handle. | |
4969 | * On Linux we can get here through truncate_range() which | |
4970 | * operates directly on inodes, so we need to check access rights. | |
4971 | */ | |
4972 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
0037b49e | 4973 | ZFS_EXIT(zfsvfs); |
aec69371 ED |
4974 | return (error); |
4975 | } | |
4976 | ||
34dc7c2f BB |
4977 | off = bfp->l_start; |
4978 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4979 | ||
b128c09f | 4980 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4981 | |
0037b49e | 4982 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4983 | return (error); |
4984 | } | |
4985 | ||
4986 | /*ARGSUSED*/ | |
e5c39b95 | 4987 | int |
3558fd73 | 4988 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4989 | { |
3558fd73 | 4990 | znode_t *zp = ITOZ(ip); |
0037b49e | 4991 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4992 | uint32_t gen; |
428870ff | 4993 | uint64_t gen64; |
34dc7c2f BB |
4994 | uint64_t object = zp->z_id; |
4995 | zfid_short_t *zfid; | |
428870ff | 4996 | int size, i, error; |
34dc7c2f | 4997 | |
0037b49e | 4998 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4999 | ZFS_VERIFY_ZP(zp); |
428870ff | 5000 | |
0037b49e | 5001 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), |
428870ff | 5002 | &gen64, sizeof (uint64_t))) != 0) { |
0037b49e | 5003 | ZFS_EXIT(zfsvfs); |
428870ff BB |
5004 | return (error); |
5005 | } | |
5006 | ||
5007 | gen = (uint32_t)gen64; | |
34dc7c2f | 5008 | |
9b77d1c9 | 5009 | size = SHORT_FID_LEN; |
34dc7c2f BB |
5010 | |
5011 | zfid = (zfid_short_t *)fidp; | |
5012 | ||
5013 | zfid->zf_len = size; | |
5014 | ||
5015 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
5016 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
5017 | ||
5018 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
5019 | if (gen == 0) | |
5020 | gen = 1; | |
5021 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
5022 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
5023 | ||
0037b49e | 5024 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5025 | return (0); |
5026 | } | |
5027 | ||
34dc7c2f | 5028 | /*ARGSUSED*/ |
e5c39b95 | 5029 | int |
3558fd73 | 5030 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 5031 | { |
3558fd73 | 5032 | znode_t *zp = ITOZ(ip); |
0037b49e | 5033 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
5034 | int error; |
5035 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
5036 | ||
0037b49e | 5037 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
5038 | ZFS_VERIFY_ZP(zp); |
5039 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
0037b49e | 5040 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5041 | |
5042 | return (error); | |
5043 | } | |
5044 | ||
5045 | /*ARGSUSED*/ | |
e5c39b95 | 5046 | int |
3558fd73 | 5047 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 5048 | { |
3558fd73 | 5049 | znode_t *zp = ITOZ(ip); |
0037b49e | 5050 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
5051 | int error; |
5052 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
0037b49e | 5053 | zilog_t *zilog = zfsvfs->z_log; |
34dc7c2f | 5054 | |
0037b49e | 5055 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 5056 | ZFS_VERIFY_ZP(zp); |
428870ff | 5057 | |
34dc7c2f | 5058 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 5059 | |
0037b49e | 5060 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 5061 | zil_commit(zilog, 0); |
428870ff | 5062 | |
0037b49e | 5063 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5064 | return (error); |
5065 | } | |
5066 | ||
3558fd73 | 5067 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
5068 | /* |
5069 | * Tunable, both must be a power of 2. | |
5070 | * | |
5071 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
5072 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 5073 | * an arcbuf for a partial block read |
428870ff BB |
5074 | */ |
5075 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
5076 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
5077 | ||
5078 | /*ARGSUSED*/ | |
5079 | static int | |
3558fd73 | 5080 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 5081 | { |
3558fd73 | 5082 | znode_t *zp = ITOZ(ip); |
0037b49e BB |
5083 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
5084 | int max_blksz = zfsvfs->z_max_blksz; | |
428870ff BB |
5085 | uio_t *uio = &xuio->xu_uio; |
5086 | ssize_t size = uio->uio_resid; | |
5087 | offset_t offset = uio->uio_loffset; | |
5088 | int blksz; | |
5089 | int fullblk, i; | |
5090 | arc_buf_t *abuf; | |
5091 | ssize_t maxsize; | |
5092 | int preamble, postamble; | |
5093 | ||
5094 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
2e528b49 | 5095 | return (SET_ERROR(EINVAL)); |
428870ff | 5096 | |
0037b49e | 5097 | ZFS_ENTER(zfsvfs); |
428870ff BB |
5098 | ZFS_VERIFY_ZP(zp); |
5099 | switch (ioflag) { | |
5100 | case UIO_WRITE: | |
5101 | /* | |
5102 | * Loan out an arc_buf for write if write size is bigger than | |
5103 | * max_blksz, and the file's block size is also max_blksz. | |
5104 | */ | |
5105 | blksz = max_blksz; | |
5106 | if (size < blksz || zp->z_blksz != blksz) { | |
0037b49e | 5107 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5108 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5109 | } |
5110 | /* | |
5111 | * Caller requests buffers for write before knowing where the | |
5112 | * write offset might be (e.g. NFS TCP write). | |
5113 | */ | |
5114 | if (offset == -1) { | |
5115 | preamble = 0; | |
5116 | } else { | |
5117 | preamble = P2PHASE(offset, blksz); | |
5118 | if (preamble) { | |
5119 | preamble = blksz - preamble; | |
5120 | size -= preamble; | |
5121 | } | |
5122 | } | |
5123 | ||
5124 | postamble = P2PHASE(size, blksz); | |
5125 | size -= postamble; | |
5126 | ||
5127 | fullblk = size / blksz; | |
5128 | (void) dmu_xuio_init(xuio, | |
5129 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
5130 | |
5131 | /* | |
5132 | * Have to fix iov base/len for partial buffers. They | |
5133 | * currently represent full arc_buf's. | |
5134 | */ | |
5135 | if (preamble) { | |
5136 | /* data begins in the middle of the arc_buf */ | |
5137 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5138 | blksz); | |
5139 | ASSERT(abuf); | |
5140 | (void) dmu_xuio_add(xuio, abuf, | |
5141 | blksz - preamble, preamble); | |
5142 | } | |
5143 | ||
5144 | for (i = 0; i < fullblk; i++) { | |
5145 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5146 | blksz); | |
5147 | ASSERT(abuf); | |
5148 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
5149 | } | |
5150 | ||
5151 | if (postamble) { | |
5152 | /* data ends in the middle of the arc_buf */ | |
5153 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5154 | blksz); | |
5155 | ASSERT(abuf); | |
5156 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
5157 | } | |
5158 | break; | |
5159 | case UIO_READ: | |
5160 | /* | |
5161 | * Loan out an arc_buf for read if the read size is larger than | |
5162 | * the current file block size. Block alignment is not | |
5163 | * considered. Partial arc_buf will be loaned out for read. | |
5164 | */ | |
5165 | blksz = zp->z_blksz; | |
5166 | if (blksz < zcr_blksz_min) | |
5167 | blksz = zcr_blksz_min; | |
5168 | if (blksz > zcr_blksz_max) | |
5169 | blksz = zcr_blksz_max; | |
5170 | /* avoid potential complexity of dealing with it */ | |
5171 | if (blksz > max_blksz) { | |
0037b49e | 5172 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5173 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5174 | } |
5175 | ||
5176 | maxsize = zp->z_size - uio->uio_loffset; | |
5177 | if (size > maxsize) | |
5178 | size = maxsize; | |
5179 | ||
3558fd73 | 5180 | if (size < blksz) { |
0037b49e | 5181 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5182 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5183 | } |
5184 | break; | |
5185 | default: | |
0037b49e | 5186 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5187 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5188 | } |
5189 | ||
5190 | uio->uio_extflg = UIO_XUIO; | |
5191 | XUIO_XUZC_RW(xuio) = ioflag; | |
0037b49e | 5192 | ZFS_EXIT(zfsvfs); |
428870ff BB |
5193 | return (0); |
5194 | } | |
5195 | ||
5196 | /*ARGSUSED*/ | |
5197 | static int | |
3558fd73 | 5198 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
5199 | { |
5200 | int i; | |
5201 | arc_buf_t *abuf; | |
5202 | int ioflag = XUIO_XUZC_RW(xuio); | |
5203 | ||
5204 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
5205 | ||
5206 | i = dmu_xuio_cnt(xuio); | |
5207 | while (i-- > 0) { | |
5208 | abuf = dmu_xuio_arcbuf(xuio, i); | |
5209 | /* | |
5210 | * if abuf == NULL, it must be a write buffer | |
5211 | * that has been returned in zfs_write(). | |
5212 | */ | |
5213 | if (abuf) | |
5214 | dmu_return_arcbuf(abuf); | |
5215 | ASSERT(abuf || ioflag == UIO_WRITE); | |
5216 | } | |
5217 | ||
5218 | dmu_xuio_fini(xuio); | |
5219 | return (0); | |
5220 | } | |
3558fd73 | 5221 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 BB |
5222 | |
5223 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
f298b24d BB |
5224 | EXPORT_SYMBOL(zfs_open); |
5225 | EXPORT_SYMBOL(zfs_close); | |
5226 | EXPORT_SYMBOL(zfs_read); | |
5227 | EXPORT_SYMBOL(zfs_write); | |
5228 | EXPORT_SYMBOL(zfs_access); | |
5229 | EXPORT_SYMBOL(zfs_lookup); | |
5230 | EXPORT_SYMBOL(zfs_create); | |
5231 | EXPORT_SYMBOL(zfs_tmpfile); | |
5232 | EXPORT_SYMBOL(zfs_remove); | |
5233 | EXPORT_SYMBOL(zfs_mkdir); | |
5234 | EXPORT_SYMBOL(zfs_rmdir); | |
5235 | EXPORT_SYMBOL(zfs_readdir); | |
5236 | EXPORT_SYMBOL(zfs_fsync); | |
5237 | EXPORT_SYMBOL(zfs_getattr); | |
5238 | EXPORT_SYMBOL(zfs_getattr_fast); | |
5239 | EXPORT_SYMBOL(zfs_setattr); | |
5240 | EXPORT_SYMBOL(zfs_rename); | |
5241 | EXPORT_SYMBOL(zfs_symlink); | |
5242 | EXPORT_SYMBOL(zfs_readlink); | |
5243 | EXPORT_SYMBOL(zfs_link); | |
5244 | EXPORT_SYMBOL(zfs_inactive); | |
5245 | EXPORT_SYMBOL(zfs_space); | |
5246 | EXPORT_SYMBOL(zfs_fid); | |
5247 | EXPORT_SYMBOL(zfs_getsecattr); | |
5248 | EXPORT_SYMBOL(zfs_setsecattr); | |
5249 | EXPORT_SYMBOL(zfs_getpage); | |
5250 | EXPORT_SYMBOL(zfs_putpage); | |
5251 | EXPORT_SYMBOL(zfs_dirty_inode); | |
5252 | EXPORT_SYMBOL(zfs_map); | |
5253 | ||
02730c33 | 5254 | /* CSTYLED */ |
a966c564 K |
5255 | module_param(zfs_delete_blocks, ulong, 0644); |
5256 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
c409e464 BB |
5257 | module_param(zfs_read_chunk_size, long, 0644); |
5258 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
5259 | #endif |