]>
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. |
a448a255 | 24 | * Copyright (c) 2012, 2018 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> | |
34dc7c2f | 36 | #include <sys/sysmacros.h> |
34dc7c2f | 37 | #include <sys/vfs.h> |
34dc7c2f BB |
38 | #include <sys/file.h> |
39 | #include <sys/stat.h> | |
40 | #include <sys/kmem.h> | |
41 | #include <sys/taskq.h> | |
42 | #include <sys/uio.h> | |
43 | #include <sys/vmsystm.h> | |
44 | #include <sys/atomic.h> | |
34dc7c2f BB |
45 | #include <sys/pathname.h> |
46 | #include <sys/cmn_err.h> | |
47 | #include <sys/errno.h> | |
34dc7c2f BB |
48 | #include <sys/zfs_dir.h> |
49 | #include <sys/zfs_acl.h> | |
50 | #include <sys/zfs_ioctl.h> | |
51 | #include <sys/fs/zfs.h> | |
52 | #include <sys/dmu.h> | |
428870ff | 53 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
54 | #include <sys/spa.h> |
55 | #include <sys/txg.h> | |
56 | #include <sys/dbuf.h> | |
57 | #include <sys/zap.h> | |
428870ff | 58 | #include <sys/sa.h> |
34dc7c2f BB |
59 | #include <sys/policy.h> |
60 | #include <sys/sunddi.h> | |
b128c09f | 61 | #include <sys/sid.h> |
bcf30822 | 62 | #include <sys/mode.h> |
ebe7e575 | 63 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 64 | #include <sys/zfs_fuid.h> |
428870ff | 65 | #include <sys/zfs_sa.h> |
e5c39b95 | 66 | #include <sys/zfs_vnops.h> |
34dc7c2f | 67 | #include <sys/zfs_rlock.h> |
428870ff | 68 | #include <sys/cred.h> |
218b8eaf | 69 | #include <sys/zpl.h> |
1ce23dca | 70 | #include <sys/zil.h> |
9c5167d1 | 71 | #include <sys/sa_impl.h> |
34dc7c2f BB |
72 | |
73 | /* | |
74 | * Programming rules. | |
75 | * | |
76 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
77 | * properly lock its in-core state, create a DMU transaction, do the work, | |
78 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
79 | * and wait for the intent log to commit if it is a synchronous operation. | |
80 | * Moreover, the vnode ops must work in both normal and log replay context. | |
81 | * The ordering of events is important to avoid deadlocks and references | |
82 | * to freed memory. The example below illustrates the following Big Rules: | |
83 | * | |
84 | * (1) A check must be made in each zfs thread for a mounted file system. | |
0037b49e BB |
85 | * This is done avoiding races using ZFS_ENTER(zfsvfs). |
86 | * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes | |
34dc7c2f BB |
87 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros |
88 | * can return EIO from the calling function. | |
89 | * | |
3558fd73 | 90 | * (2) iput() should always be the last thing except for zil_commit() |
34dc7c2f BB |
91 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: |
92 | * First, if it's the last reference, the vnode/znode | |
93 | * can be freed, so the zp may point to freed memory. Second, the last | |
94 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
95 | * pushing cached pages (which acquires range locks) and syncing out | |
96 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
97 | * which could deadlock the system if you were already holding one. | |
0a50679c | 98 | * If you must call iput() within a tx then use zfs_iput_async(). |
34dc7c2f BB |
99 | * |
100 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
101 | * as they can span dmu_tx_assign() calls. | |
102 | * | |
384f8a09 MA |
103 | * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to |
104 | * dmu_tx_assign(). This is critical because we don't want to block | |
105 | * while holding locks. | |
106 | * | |
107 | * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This | |
108 | * reduces lock contention and CPU usage when we must wait (note that if | |
109 | * throughput is constrained by the storage, nearly every transaction | |
110 | * must wait). | |
111 | * | |
112 | * Note, in particular, that if a lock is sometimes acquired before | |
113 | * the tx assigns, and sometimes after (e.g. z_lock), then failing | |
114 | * to use a non-blocking assign can deadlock the system. The scenario: | |
34dc7c2f BB |
115 | * |
116 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
117 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
118 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
119 | * forever, because the previous txg can't quiesce until B's tx commits. | |
120 | * | |
0037b49e | 121 | * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, |
e8b96c60 | 122 | * then drop all locks, call dmu_tx_wait(), and try again. On subsequent |
0735ecb3 | 123 | * calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT, |
e8b96c60 MA |
124 | * to indicate that this operation has already called dmu_tx_wait(). |
125 | * This will ensure that we don't retry forever, waiting a short bit | |
126 | * each time. | |
34dc7c2f BB |
127 | * |
128 | * (5) If the operation succeeded, generate the intent log entry for it | |
129 | * before dropping locks. This ensures that the ordering of events | |
130 | * in the intent log matches the order in which they actually occurred. | |
d3cc8b15 | 131 | * During ZIL replay the zfs_log_* functions will update the sequence |
fb5f0bc8 | 132 | * number to indicate the zil transaction has replayed. |
34dc7c2f BB |
133 | * |
134 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
135 | * regardless of whether there were any errors. | |
136 | * | |
572e2857 | 137 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
138 | * to ensure that synchronous semantics are provided when necessary. |
139 | * | |
140 | * In general, this is how things should be ordered in each vnode op: | |
141 | * | |
0037b49e | 142 | * ZFS_ENTER(zfsvfs); // exit if unmounted |
34dc7c2f | 143 | * top: |
3558fd73 | 144 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
145 | * rw_enter(...); // grab any other locks you need |
146 | * tx = dmu_tx_create(...); // get DMU tx | |
147 | * dmu_tx_hold_*(); // hold each object you might modify | |
0735ecb3 | 148 | * error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
149 | * if (error) { |
150 | * rw_exit(...); // drop locks | |
151 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 152 | * iput(...); // release held vnodes |
fb5f0bc8 | 153 | * if (error == ERESTART) { |
e8b96c60 | 154 | * waited = B_TRUE; |
34dc7c2f BB |
155 | * dmu_tx_wait(tx); |
156 | * dmu_tx_abort(tx); | |
157 | * goto top; | |
158 | * } | |
159 | * dmu_tx_abort(tx); // abort DMU tx | |
0037b49e | 160 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
161 | * return (error); // really out of space |
162 | * } | |
163 | * error = do_real_work(); // do whatever this VOP does | |
164 | * if (error == 0) | |
165 | * zfs_log_*(...); // on success, make ZIL entry | |
166 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
167 | * rw_exit(...); // drop locks | |
168 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 169 | * iput(...); // release held vnodes |
572e2857 | 170 | * zil_commit(zilog, foid); // synchronous when necessary |
0037b49e | 171 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
172 | * return (error); // done, report error |
173 | */ | |
174 | ||
126400a1 BB |
175 | /* |
176 | * Virus scanning is unsupported. It would be possible to add a hook | |
177 | * here to performance the required virus scan. This could be done | |
178 | * entirely in the kernel or potentially as an update to invoke a | |
179 | * scanning utility. | |
180 | */ | |
181 | static int | |
182 | zfs_vscan(struct inode *ip, cred_t *cr, int async) | |
183 | { | |
184 | return (0); | |
185 | } | |
186 | ||
187 | /* ARGSUSED */ | |
188 | int | |
189 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
190 | { | |
191 | znode_t *zp = ITOZ(ip); | |
0037b49e | 192 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 193 | |
0037b49e | 194 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
195 | ZFS_VERIFY_ZP(zp); |
196 | ||
197 | /* Honor ZFS_APPENDONLY file attribute */ | |
198 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
199 | ((flag & O_APPEND) == 0)) { | |
0037b49e | 200 | ZFS_EXIT(zfsvfs); |
2e528b49 | 201 | return (SET_ERROR(EPERM)); |
126400a1 BB |
202 | } |
203 | ||
204 | /* Virus scan eligible files on open */ | |
0037b49e | 205 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
206 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { |
207 | if (zfs_vscan(ip, cr, 0) != 0) { | |
0037b49e | 208 | ZFS_EXIT(zfsvfs); |
2e528b49 | 209 | return (SET_ERROR(EACCES)); |
126400a1 BB |
210 | } |
211 | } | |
212 | ||
213 | /* Keep a count of the synchronous opens in the znode */ | |
214 | if (flag & O_SYNC) | |
215 | atomic_inc_32(&zp->z_sync_cnt); | |
216 | ||
0037b49e | 217 | ZFS_EXIT(zfsvfs); |
126400a1 BB |
218 | return (0); |
219 | } | |
126400a1 BB |
220 | |
221 | /* ARGSUSED */ | |
222 | int | |
223 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
224 | { | |
225 | znode_t *zp = ITOZ(ip); | |
0037b49e | 226 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 227 | |
0037b49e | 228 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
229 | ZFS_VERIFY_ZP(zp); |
230 | ||
7dc71949 | 231 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 232 | if (flag & O_SYNC) |
7dc71949 | 233 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 | 234 | |
0037b49e | 235 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
236 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) |
237 | VERIFY(zfs_vscan(ip, cr, 1) == 0); | |
238 | ||
0037b49e | 239 | ZFS_EXIT(zfsvfs); |
8780c539 | 240 | return (0); |
126400a1 | 241 | } |
126400a1 | 242 | |
802e7b5f | 243 | #if defined(SEEK_HOLE) && defined(SEEK_DATA) |
cf91b2b6 | 244 | /* |
802e7b5f LD |
245 | * Lseek support for finding holes (cmd == SEEK_HOLE) and |
246 | * data (cmd == SEEK_DATA). "off" is an in/out parameter. | |
cf91b2b6 MA |
247 | */ |
248 | static int | |
802e7b5f | 249 | zfs_holey_common(struct inode *ip, int cmd, loff_t *off) |
cf91b2b6 | 250 | { |
802e7b5f | 251 | znode_t *zp = ITOZ(ip); |
cf91b2b6 MA |
252 | uint64_t noff = (uint64_t)*off; /* new offset */ |
253 | uint64_t file_sz; | |
254 | int error; | |
255 | boolean_t hole; | |
256 | ||
257 | file_sz = zp->z_size; | |
258 | if (noff >= file_sz) { | |
2e528b49 | 259 | return (SET_ERROR(ENXIO)); |
cf91b2b6 MA |
260 | } |
261 | ||
802e7b5f | 262 | if (cmd == SEEK_HOLE) |
cf91b2b6 MA |
263 | hole = B_TRUE; |
264 | else | |
265 | hole = B_FALSE; | |
266 | ||
802e7b5f | 267 | error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff); |
cf91b2b6 | 268 | |
d97aa48f | 269 | if (error == ESRCH) |
2e528b49 | 270 | return (SET_ERROR(ENXIO)); |
d97aa48f | 271 | |
6e03ec4f DB |
272 | /* file was dirty, so fall back to using generic logic */ |
273 | if (error == EBUSY) { | |
274 | if (hole) | |
275 | *off = file_sz; | |
276 | ||
277 | return (0); | |
278 | } | |
66aca247 | 279 | |
d97aa48f MA |
280 | /* |
281 | * We could find a hole that begins after the logical end-of-file, | |
282 | * because dmu_offset_next() only works on whole blocks. If the | |
283 | * EOF falls mid-block, then indicate that the "virtual hole" | |
284 | * at the end of the file begins at the logical EOF, rather than | |
285 | * at the end of the last block. | |
286 | */ | |
287 | if (noff > file_sz) { | |
288 | ASSERT(hole); | |
289 | noff = file_sz; | |
cf91b2b6 MA |
290 | } |
291 | ||
292 | if (noff < *off) | |
293 | return (error); | |
294 | *off = noff; | |
295 | return (error); | |
296 | } | |
802e7b5f LD |
297 | |
298 | int | |
299 | zfs_holey(struct inode *ip, int cmd, loff_t *off) | |
300 | { | |
301 | znode_t *zp = ITOZ(ip); | |
0037b49e | 302 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
802e7b5f LD |
303 | int error; |
304 | ||
0037b49e | 305 | ZFS_ENTER(zfsvfs); |
802e7b5f LD |
306 | ZFS_VERIFY_ZP(zp); |
307 | ||
308 | error = zfs_holey_common(ip, cmd, off); | |
309 | ||
0037b49e | 310 | ZFS_EXIT(zfsvfs); |
802e7b5f LD |
311 | return (error); |
312 | } | |
802e7b5f | 313 | #endif /* SEEK_HOLE && SEEK_DATA */ |
cf91b2b6 | 314 | |
c0d35759 | 315 | #if defined(_KERNEL) |
34dc7c2f BB |
316 | /* |
317 | * When a file is memory mapped, we must keep the IO data synchronized | |
318 | * between the DMU cache and the memory mapped pages. What this means: | |
319 | * | |
320 | * On Write: If we find a memory mapped page, we write to *both* | |
321 | * the page and the dmu buffer. | |
34dc7c2f | 322 | */ |
d164b209 | 323 | static void |
c0d35759 BB |
324 | update_pages(struct inode *ip, int64_t start, int len, |
325 | objset_t *os, uint64_t oid) | |
34dc7c2f | 326 | { |
c0d35759 BB |
327 | struct address_space *mp = ip->i_mapping; |
328 | struct page *pp; | |
329 | uint64_t nbytes; | |
d164b209 | 330 | int64_t off; |
c0d35759 | 331 | void *pb; |
34dc7c2f | 332 | |
8b1899d3 BB |
333 | off = start & (PAGE_SIZE-1); |
334 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
335 | nbytes = MIN(PAGE_SIZE - off, len); | |
34dc7c2f | 336 | |
8b1899d3 | 337 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
338 | if (pp) { |
339 | if (mapping_writably_mapped(mp)) | |
340 | flush_dcache_page(pp); | |
34dc7c2f | 341 | |
c0d35759 BB |
342 | pb = kmap(pp); |
343 | (void) dmu_read(os, oid, start+off, nbytes, pb+off, | |
9babb374 | 344 | DMU_READ_PREFETCH); |
c0d35759 BB |
345 | kunmap(pp); |
346 | ||
347 | if (mapping_writably_mapped(mp)) | |
348 | flush_dcache_page(pp); | |
349 | ||
350 | mark_page_accessed(pp); | |
351 | SetPageUptodate(pp); | |
352 | ClearPageError(pp); | |
353 | unlock_page(pp); | |
8b1899d3 | 354 | put_page(pp); |
34dc7c2f | 355 | } |
c0d35759 | 356 | |
d164b209 | 357 | len -= nbytes; |
34dc7c2f | 358 | off = 0; |
34dc7c2f | 359 | } |
34dc7c2f BB |
360 | } |
361 | ||
362 | /* | |
363 | * When a file is memory mapped, we must keep the IO data synchronized | |
364 | * between the DMU cache and the memory mapped pages. What this means: | |
365 | * | |
366 | * On Read: We "read" preferentially from memory mapped pages, | |
367 | * else we default from the dmu buffer. | |
368 | * | |
369 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
d3cc8b15 | 370 | * the file is memory mapped. |
34dc7c2f BB |
371 | */ |
372 | static int | |
3558fd73 | 373 | mappedread(struct inode *ip, int nbytes, uio_t *uio) |
34dc7c2f | 374 | { |
c0d35759 BB |
375 | struct address_space *mp = ip->i_mapping; |
376 | struct page *pp; | |
3558fd73 | 377 | znode_t *zp = ITOZ(ip); |
34dc7c2f | 378 | int64_t start, off; |
c0d35759 | 379 | uint64_t bytes; |
34dc7c2f BB |
380 | int len = nbytes; |
381 | int error = 0; | |
c0d35759 | 382 | void *pb; |
34dc7c2f BB |
383 | |
384 | start = uio->uio_loffset; | |
8b1899d3 BB |
385 | off = start & (PAGE_SIZE-1); |
386 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
387 | bytes = MIN(PAGE_SIZE - off, len); | |
c0d35759 | 388 | |
8b1899d3 | 389 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
390 | if (pp) { |
391 | ASSERT(PageUptodate(pp)); | |
b2ab468d | 392 | unlock_page(pp); |
c0d35759 BB |
393 | |
394 | pb = kmap(pp); | |
395 | error = uiomove(pb + off, bytes, UIO_READ, uio); | |
396 | kunmap(pp); | |
397 | ||
398 | if (mapping_writably_mapped(mp)) | |
399 | flush_dcache_page(pp); | |
400 | ||
401 | mark_page_accessed(pp); | |
8b1899d3 | 402 | put_page(pp); |
34dc7c2f | 403 | } else { |
804e0504 MA |
404 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
405 | uio, bytes); | |
34dc7c2f | 406 | } |
c0d35759 | 407 | |
34dc7c2f BB |
408 | len -= bytes; |
409 | off = 0; | |
410 | if (error) | |
411 | break; | |
412 | } | |
413 | return (error); | |
414 | } | |
c0d35759 | 415 | #endif /* _KERNEL */ |
34dc7c2f | 416 | |
c409e464 | 417 | unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ |
a966c564 | 418 | unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT; |
34dc7c2f BB |
419 | |
420 | /* | |
421 | * Read bytes from specified file into supplied buffer. | |
422 | * | |
3558fd73 | 423 | * IN: ip - inode of file to be read from. |
34dc7c2f BB |
424 | * uio - structure supplying read location, range info, |
425 | * and return buffer. | |
c0d35759 BB |
426 | * ioflag - FSYNC flags; used to provide FRSYNC semantics. |
427 | * O_DIRECT flag; used to bypass page cache. | |
34dc7c2f | 428 | * cr - credentials of caller. |
34dc7c2f BB |
429 | * |
430 | * OUT: uio - updated offset and range, buffer filled. | |
431 | * | |
d3cc8b15 | 432 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
433 | * |
434 | * Side Effects: | |
3558fd73 | 435 | * inode - atime updated if byte count > 0 |
34dc7c2f BB |
436 | */ |
437 | /* ARGSUSED */ | |
e5c39b95 | 438 | int |
3558fd73 | 439 | zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 440 | { |
a448a255 | 441 | int error = 0; |
2efea7c8 | 442 | boolean_t frsync = B_FALSE; |
34dc7c2f | 443 | |
a448a255 SD |
444 | znode_t *zp = ITOZ(ip); |
445 | zfsvfs_t *zfsvfs = ITOZSB(ip); | |
0037b49e | 446 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 447 | ZFS_VERIFY_ZP(zp); |
34dc7c2f | 448 | |
428870ff | 449 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
0037b49e | 450 | ZFS_EXIT(zfsvfs); |
2e528b49 | 451 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
452 | } |
453 | ||
454 | /* | |
455 | * Validate file offset | |
456 | */ | |
457 | if (uio->uio_loffset < (offset_t)0) { | |
0037b49e | 458 | ZFS_EXIT(zfsvfs); |
2e528b49 | 459 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
460 | } |
461 | ||
462 | /* | |
463 | * Fasttrack empty reads | |
464 | */ | |
465 | if (uio->uio_resid == 0) { | |
0037b49e | 466 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
467 | return (0); |
468 | } | |
469 | ||
2efea7c8 | 470 | #ifdef FRSYNC |
34dc7c2f BB |
471 | /* |
472 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
37699482 | 473 | * Only do this for non-snapshots. |
2efea7c8 GW |
474 | * |
475 | * Some platforms do not support FRSYNC and instead map it | |
476 | * to FSYNC, which results in unnecessary calls to zil_commit. We | |
477 | * only honor FRSYNC requests on platforms which support it. | |
34dc7c2f | 478 | */ |
2efea7c8 GW |
479 | frsync = !!(ioflag & FRSYNC); |
480 | #endif | |
37699482 | 481 | if (zfsvfs->z_log && |
2efea7c8 | 482 | (frsync || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)) |
0037b49e | 483 | zil_commit(zfsvfs->z_log, zp->z_id); |
34dc7c2f BB |
484 | |
485 | /* | |
486 | * Lock the range against changes. | |
487 | */ | |
5d43cc9a | 488 | locked_range_t *lr = rangelock_enter(&zp->z_rangelock, |
a448a255 | 489 | uio->uio_loffset, uio->uio_resid, RL_READER); |
34dc7c2f BB |
490 | |
491 | /* | |
492 | * If we are reading past end-of-file we can skip | |
493 | * to the end; but we might still need to set atime. | |
494 | */ | |
428870ff | 495 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
496 | error = 0; |
497 | goto out; | |
498 | } | |
499 | ||
428870ff | 500 | ASSERT(uio->uio_loffset < zp->z_size); |
a448a255 SD |
501 | ssize_t n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); |
502 | ssize_t start_resid = n; | |
428870ff | 503 | |
3558fd73 | 504 | #ifdef HAVE_UIO_ZEROCOPY |
a448a255 | 505 | xuio_t *xuio = NULL; |
428870ff BB |
506 | if ((uio->uio_extflg == UIO_XUIO) && |
507 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
508 | int nblk; | |
509 | int blksz = zp->z_blksz; | |
510 | uint64_t offset = uio->uio_loffset; | |
511 | ||
512 | xuio = (xuio_t *)uio; | |
513 | if ((ISP2(blksz))) { | |
514 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
515 | blksz)) / blksz; | |
516 | } else { | |
517 | ASSERT(offset + n <= blksz); | |
518 | nblk = 1; | |
519 | } | |
520 | (void) dmu_xuio_init(xuio, nblk); | |
521 | ||
3558fd73 | 522 | if (vn_has_cached_data(ip)) { |
428870ff BB |
523 | /* |
524 | * For simplicity, we always allocate a full buffer | |
525 | * even if we only expect to read a portion of a block. | |
526 | */ | |
527 | while (--nblk >= 0) { | |
528 | (void) dmu_xuio_add(xuio, | |
529 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
530 | blksz), 0, blksz); | |
531 | } | |
532 | } | |
533 | } | |
3558fd73 | 534 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
535 | |
536 | while (n > 0) { | |
a448a255 | 537 | ssize_t nbytes = MIN(n, zfs_read_chunk_size - |
34dc7c2f BB |
538 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); |
539 | ||
804e0504 | 540 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
3558fd73 | 541 | error = mappedread(ip, nbytes, uio); |
804e0504 MA |
542 | } else { |
543 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), | |
544 | uio, nbytes); | |
545 | } | |
c0d35759 | 546 | |
b128c09f BB |
547 | if (error) { |
548 | /* convert checksum errors into IO errors */ | |
549 | if (error == ECKSUM) | |
2e528b49 | 550 | error = SET_ERROR(EIO); |
34dc7c2f | 551 | break; |
b128c09f | 552 | } |
34dc7c2f BB |
553 | |
554 | n -= nbytes; | |
555 | } | |
a448a255 SD |
556 | |
557 | int64_t nread = start_resid - n; | |
558 | dataset_kstats_update_read_kstats(&zfsvfs->z_kstat, nread); | |
559 | task_io_account_read(nread); | |
34dc7c2f | 560 | out: |
5d43cc9a | 561 | rangelock_exit(lr); |
34dc7c2f | 562 | |
0037b49e | 563 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
564 | return (error); |
565 | } | |
566 | ||
34dc7c2f BB |
567 | /* |
568 | * Write the bytes to a file. | |
569 | * | |
3558fd73 | 570 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
571 | * uio - structure supplying write location, range info, |
572 | * and data buffer. | |
573 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 574 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 575 | * cr - credentials of caller. |
34dc7c2f BB |
576 | * |
577 | * OUT: uio - updated offset and range. | |
578 | * | |
579 | * RETURN: 0 if success | |
580 | * error code if failure | |
581 | * | |
582 | * Timestamps: | |
3558fd73 | 583 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 584 | */ |
428870ff | 585 | |
34dc7c2f | 586 | /* ARGSUSED */ |
e5c39b95 | 587 | int |
3558fd73 | 588 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 589 | { |
a448a255 SD |
590 | int error = 0; |
591 | ssize_t start_resid = uio->uio_resid; | |
34dc7c2f | 592 | |
34dc7c2f BB |
593 | /* |
594 | * Fasttrack empty write | |
595 | */ | |
a448a255 | 596 | ssize_t n = start_resid; |
34dc7c2f BB |
597 | if (n == 0) |
598 | return (0); | |
599 | ||
a448a255 | 600 | rlim64_t limit = uio->uio_limit; |
34dc7c2f BB |
601 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) |
602 | limit = MAXOFFSET_T; | |
603 | ||
a448a255 SD |
604 | znode_t *zp = ITOZ(ip); |
605 | zfsvfs_t *zfsvfs = ZTOZSB(zp); | |
0037b49e | 606 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 607 | ZFS_VERIFY_ZP(zp); |
b128c09f | 608 | |
a448a255 SD |
609 | sa_bulk_attr_t bulk[4]; |
610 | int count = 0; | |
611 | uint64_t mtime[2], ctime[2]; | |
0037b49e BB |
612 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
613 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
614 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
615 | &zp->z_size, 8); | |
616 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
428870ff BB |
617 | &zp->z_pflags, 8); |
618 | ||
f3c9dca0 MT |
619 | /* |
620 | * Callers might not be able to detect properly that we are read-only, | |
621 | * so check it explicitly here. | |
622 | */ | |
0037b49e BB |
623 | if (zfs_is_readonly(zfsvfs)) { |
624 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
625 | return (SET_ERROR(EROFS)); |
626 | } | |
627 | ||
b128c09f BB |
628 | /* |
629 | * If immutable or not appending then return EPERM | |
630 | */ | |
428870ff BB |
631 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
632 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
633 | (uio->uio_loffset < zp->z_size))) { | |
0037b49e | 634 | ZFS_EXIT(zfsvfs); |
2e528b49 | 635 | return (SET_ERROR(EPERM)); |
b128c09f BB |
636 | } |
637 | ||
428870ff BB |
638 | /* |
639 | * Validate file offset | |
640 | */ | |
a448a255 | 641 | offset_t woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; |
428870ff | 642 | if (woff < 0) { |
0037b49e | 643 | ZFS_EXIT(zfsvfs); |
2e528b49 | 644 | return (SET_ERROR(EINVAL)); |
428870ff BB |
645 | } |
646 | ||
a448a255 SD |
647 | int max_blksz = zfsvfs->z_max_blksz; |
648 | xuio_t *xuio = NULL; | |
649 | ||
34dc7c2f BB |
650 | /* |
651 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
652 | * don't hold up txg. | |
428870ff | 653 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 654 | */ |
9cac042c | 655 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
656 | if ((uio->uio_extflg == UIO_XUIO) && |
657 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
658 | xuio = (xuio_t *)uio; | |
659 | else | |
9cac042c | 660 | #endif |
779a6c0b | 661 | if (uio_prefaultpages(MIN(n, max_blksz), uio)) { |
662 | ZFS_EXIT(zfsvfs); | |
663 | return (SET_ERROR(EFAULT)); | |
664 | } | |
34dc7c2f BB |
665 | |
666 | /* | |
667 | * If in append mode, set the io offset pointer to eof. | |
668 | */ | |
5d43cc9a | 669 | locked_range_t *lr; |
34dc7c2f BB |
670 | if (ioflag & FAPPEND) { |
671 | /* | |
428870ff BB |
672 | * Obtain an appending range lock to guarantee file append |
673 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f | 674 | */ |
5d43cc9a MA |
675 | lr = rangelock_enter(&zp->z_rangelock, 0, n, RL_APPEND); |
676 | woff = lr->lr_offset; | |
677 | if (lr->lr_length == UINT64_MAX) { | |
428870ff BB |
678 | /* |
679 | * We overlocked the file because this write will cause | |
680 | * the file block size to increase. | |
681 | * Note that zp_size cannot change with this lock held. | |
682 | */ | |
683 | woff = zp->z_size; | |
34dc7c2f | 684 | } |
428870ff | 685 | uio->uio_loffset = woff; |
34dc7c2f | 686 | } else { |
34dc7c2f | 687 | /* |
428870ff BB |
688 | * Note that if the file block size will change as a result of |
689 | * this write, then this range lock will lock the entire file | |
690 | * so that we can re-write the block safely. | |
34dc7c2f | 691 | */ |
5d43cc9a | 692 | lr = rangelock_enter(&zp->z_rangelock, woff, n, RL_WRITER); |
34dc7c2f BB |
693 | } |
694 | ||
695 | if (woff >= limit) { | |
5d43cc9a | 696 | rangelock_exit(lr); |
0037b49e | 697 | ZFS_EXIT(zfsvfs); |
2e528b49 | 698 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
699 | } |
700 | ||
701 | if ((woff + n) > limit || woff > (limit - n)) | |
702 | n = limit - woff; | |
703 | ||
428870ff | 704 | /* Will this write extend the file length? */ |
a448a255 SD |
705 | int write_eof = (woff + n > zp->z_size); |
706 | ||
707 | uint64_t end_size = MAX(zp->z_size, woff + n); | |
708 | zilog_t *zilog = zfsvfs->z_log; | |
709 | #ifdef HAVE_UIO_ZEROCOPY | |
710 | int i_iov = 0; | |
711 | const iovec_t *iovp = uio->uio_iov; | |
712 | ASSERTV(int iovcnt = uio->uio_iovcnt); | |
713 | #endif | |
428870ff | 714 | |
34dc7c2f BB |
715 | |
716 | /* | |
717 | * Write the file in reasonable size chunks. Each chunk is written | |
718 | * in a separate transaction; this keeps the intent log records small | |
719 | * and allows us to do more fine-grained space accounting. | |
720 | */ | |
721 | while (n > 0) { | |
9babb374 | 722 | woff = uio->uio_loffset; |
a448a255 | 723 | |
9c5167d1 NF |
724 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
725 | KUID_TO_SUID(ip->i_uid)) || | |
726 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
727 | KGID_TO_SGID(ip->i_gid)) || | |
728 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
729 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
730 | zp->z_projid))) { | |
2e528b49 | 731 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
732 | break; |
733 | } | |
734 | ||
a448a255 SD |
735 | arc_buf_t *abuf = NULL; |
736 | const iovec_t *aiov = NULL; | |
737 | if (xuio) { | |
5a6765cf | 738 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 739 | ASSERT(i_iov < iovcnt); |
5475aada | 740 | ASSERT3U(uio->uio_segflg, !=, UIO_BVEC); |
428870ff BB |
741 | aiov = &iovp[i_iov]; |
742 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
743 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
744 | ASSERT((aiov->iov_base == abuf->b_data) || |
745 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
746 | aiov->iov_len == arc_buf_size(abuf))); | |
747 | i_iov++; | |
5a6765cf | 748 | #endif |
a448a255 | 749 | } else if (n >= max_blksz && woff >= zp->z_size && |
9babb374 BB |
750 | P2PHASE(woff, max_blksz) == 0 && |
751 | zp->z_blksz == max_blksz) { | |
428870ff BB |
752 | /* |
753 | * This write covers a full block. "Borrow" a buffer | |
754 | * from the dmu so that we can fill it before we enter | |
755 | * a transaction. This avoids the possibility of | |
756 | * holding up the transaction if the data copy hangs | |
757 | * up on a pagefault (e.g., from an NFS server mapping). | |
758 | */ | |
9babb374 BB |
759 | size_t cbytes; |
760 | ||
428870ff BB |
761 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
762 | max_blksz); | |
9babb374 BB |
763 | ASSERT(abuf != NULL); |
764 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
765 | if ((error = uiocopy(abuf->b_data, max_blksz, |
766 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
767 | dmu_return_arcbuf(abuf); |
768 | break; | |
769 | } | |
770 | ASSERT(cbytes == max_blksz); | |
771 | } | |
772 | ||
34dc7c2f BB |
773 | /* |
774 | * Start a transaction. | |
775 | */ | |
a448a255 | 776 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 777 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
0eb8ba6a MA |
778 | dmu_buf_impl_t *db = (dmu_buf_impl_t *)sa_get_db(zp->z_sa_hdl); |
779 | DB_DNODE_ENTER(db); | |
780 | dmu_tx_hold_write_by_dnode(tx, DB_DNODE(db), woff, | |
781 | MIN(n, max_blksz)); | |
782 | DB_DNODE_EXIT(db); | |
428870ff | 783 | zfs_sa_upgrade_txholds(tx, zp); |
384f8a09 | 784 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 785 | if (error) { |
34dc7c2f | 786 | dmu_tx_abort(tx); |
9babb374 BB |
787 | if (abuf != NULL) |
788 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
789 | break; |
790 | } | |
791 | ||
792 | /* | |
5d43cc9a | 793 | * If rangelock_enter() over-locked we grow the blocksize |
34dc7c2f | 794 | * and then reduce the lock range. This will only happen |
5d43cc9a MA |
795 | * on the first iteration since rangelock_reduce() will |
796 | * shrink down lr_length to the appropriate size. | |
34dc7c2f | 797 | */ |
5d43cc9a | 798 | if (lr->lr_length == UINT64_MAX) { |
34dc7c2f BB |
799 | uint64_t new_blksz; |
800 | ||
801 | if (zp->z_blksz > max_blksz) { | |
f1512ee6 MA |
802 | /* |
803 | * File's blocksize is already larger than the | |
804 | * "recordsize" property. Only let it grow to | |
805 | * the next power of 2. | |
806 | */ | |
34dc7c2f | 807 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 MA |
808 | new_blksz = MIN(end_size, |
809 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
810 | } else { |
811 | new_blksz = MIN(end_size, max_blksz); | |
812 | } | |
813 | zfs_grow_blocksize(zp, new_blksz, tx); | |
5d43cc9a | 814 | rangelock_reduce(lr, woff, n); |
34dc7c2f BB |
815 | } |
816 | ||
817 | /* | |
818 | * XXX - should we really limit each write to z_max_blksz? | |
819 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
820 | */ | |
a448a255 | 821 | ssize_t nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); |
34dc7c2f | 822 | |
a448a255 | 823 | ssize_t tx_bytes; |
9babb374 BB |
824 | if (abuf == NULL) { |
825 | tx_bytes = uio->uio_resid; | |
779a6c0b | 826 | uio->uio_fault_disable = B_TRUE; |
428870ff BB |
827 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
828 | uio, nbytes, tx); | |
515ddf65 | 829 | uio->uio_fault_disable = B_FALSE; |
779a6c0b | 830 | if (error == EFAULT) { |
831 | dmu_tx_commit(tx); | |
832 | if (uio_prefaultpages(MIN(n, max_blksz), uio)) { | |
833 | break; | |
834 | } | |
835 | continue; | |
836 | } else if (error != 0) { | |
837 | dmu_tx_commit(tx); | |
838 | break; | |
839 | } | |
9babb374 BB |
840 | tx_bytes -= uio->uio_resid; |
841 | } else { | |
842 | tx_bytes = nbytes; | |
428870ff BB |
843 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
844 | /* | |
845 | * If this is not a full block write, but we are | |
846 | * extending the file past EOF and this data starts | |
847 | * block-aligned, use assign_arcbuf(). Otherwise, | |
848 | * write via dmu_write(). | |
849 | */ | |
850 | if (tx_bytes < max_blksz && (!write_eof || | |
851 | aiov->iov_base != abuf->b_data)) { | |
852 | ASSERT(xuio); | |
0037b49e | 853 | dmu_write(zfsvfs->z_os, zp->z_id, woff, |
94183a9d | 854 | /* cppcheck-suppress nullPointer */ |
428870ff BB |
855 | aiov->iov_len, aiov->iov_base, tx); |
856 | dmu_return_arcbuf(abuf); | |
857 | xuio_stat_wbuf_copied(); | |
858 | } else { | |
859 | ASSERT(xuio || tx_bytes == max_blksz); | |
305781da | 860 | error = dmu_assign_arcbuf_by_dbuf( |
440a3eb9 | 861 | sa_get_db(zp->z_sa_hdl), woff, abuf, tx); |
305781da TC |
862 | if (error != 0) { |
863 | dmu_return_arcbuf(abuf); | |
864 | dmu_tx_commit(tx); | |
865 | break; | |
866 | } | |
428870ff | 867 | } |
9babb374 BB |
868 | ASSERT(tx_bytes <= uio->uio_resid); |
869 | uioskip(uio, tx_bytes); | |
870 | } | |
0037b49e BB |
871 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
872 | update_pages(ip, woff, | |
873 | tx_bytes, zfsvfs->z_os, zp->z_id); | |
874 | } | |
34dc7c2f BB |
875 | |
876 | /* | |
877 | * If we made no progress, we're done. If we made even | |
878 | * partial progress, update the znode and ZIL accordingly. | |
879 | */ | |
880 | if (tx_bytes == 0) { | |
0037b49e | 881 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 882 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
883 | dmu_tx_commit(tx); |
884 | ASSERT(error != 0); | |
885 | break; | |
886 | } | |
887 | ||
888 | /* | |
889 | * Clear Set-UID/Set-GID bits on successful write if not | |
4e33ba4c | 890 | * privileged and at least one of the execute bits is set. |
34dc7c2f BB |
891 | * |
892 | * It would be nice to to this after all writes have | |
893 | * been done, but that would still expose the ISUID/ISGID | |
894 | * to another app after the partial write is committed. | |
895 | * | |
572e2857 BB |
896 | * Note: we don't call zfs_fuid_map_id() here because |
897 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
898 | */ |
899 | mutex_enter(&zp->z_acl_lock); | |
a448a255 | 900 | uint32_t uid = KUID_TO_SUID(ip->i_uid); |
428870ff | 901 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 902 | (S_IXUSR >> 6))) != 0 && |
428870ff | 903 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 904 | secpolicy_vnode_setid_retain(cr, |
2c6abf15 | 905 | ((zp->z_mode & S_ISUID) != 0 && uid == 0)) != 0) { |
428870ff BB |
906 | uint64_t newmode; |
907 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
12fa7f34 | 908 | ip->i_mode = newmode = zp->z_mode; |
0037b49e | 909 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), |
428870ff | 910 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
911 | } |
912 | mutex_exit(&zp->z_acl_lock); | |
913 | ||
0df9673f | 914 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
34dc7c2f BB |
915 | |
916 | /* | |
917 | * Update the file size (zp_size) if it has changed; | |
918 | * account for possible concurrent updates. | |
919 | */ | |
428870ff BB |
920 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
921 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 922 | uio->uio_loffset); |
428870ff BB |
923 | ASSERT(error == 0); |
924 | } | |
572e2857 BB |
925 | /* |
926 | * If we are replaying and eof is non zero then force | |
927 | * the file size to the specified eof. Note, there's no | |
928 | * concurrency during replay. | |
929 | */ | |
0037b49e BB |
930 | if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) |
931 | zp->z_size = zfsvfs->z_replay_eof; | |
572e2857 | 932 | |
428870ff BB |
933 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
934 | ||
119a394a ED |
935 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
936 | NULL, NULL); | |
34dc7c2f BB |
937 | dmu_tx_commit(tx); |
938 | ||
939 | if (error != 0) | |
940 | break; | |
941 | ASSERT(tx_bytes == nbytes); | |
942 | n -= nbytes; | |
572e2857 | 943 | |
779a6c0b | 944 | if (!xuio && n > 0) { |
945 | if (uio_prefaultpages(MIN(n, max_blksz), uio)) { | |
946 | error = EFAULT; | |
947 | break; | |
948 | } | |
949 | } | |
34dc7c2f BB |
950 | } |
951 | ||
2a53e2da | 952 | zfs_inode_update(zp); |
5d43cc9a | 953 | rangelock_exit(lr); |
34dc7c2f BB |
954 | |
955 | /* | |
956 | * If we're in replay mode, or we made no progress, return error. | |
957 | * Otherwise, it's at least a partial write, so it's successful. | |
958 | */ | |
0037b49e BB |
959 | if (zfsvfs->z_replay || uio->uio_resid == start_resid) { |
960 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
961 | return (error); |
962 | } | |
963 | ||
428870ff | 964 | if (ioflag & (FSYNC | FDSYNC) || |
0037b49e | 965 | zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 966 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 967 | |
a448a255 SD |
968 | int64_t nwritten = start_resid - uio->uio_resid; |
969 | dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, nwritten); | |
970 | task_io_account_write(nwritten); | |
971 | ||
0037b49e | 972 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
973 | return (0); |
974 | } | |
975 | ||
7c502b0b | 976 | /* |
977 | * Drop a reference on the passed inode asynchronously. This ensures | |
978 | * that the caller will never drop the last reference on an inode in | |
979 | * the current context. Doing so while holding open a tx could result | |
980 | * in a deadlock if iput_final() re-enters the filesystem code. | |
981 | */ | |
0a50679c BB |
982 | void |
983 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 984 | { |
0a50679c BB |
985 | objset_t *os = ITOZSB(ip)->z_os; |
986 | ||
3558fd73 | 987 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
988 | ASSERT(os != NULL); |
989 | ||
3558fd73 | 990 | if (atomic_read(&ip->i_count) == 1) |
7c502b0b | 991 | VERIFY(taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
48d3eb40 | 992 | (task_func_t *)iput, ip, TQ_SLEEP) != TASKQID_INVALID); |
3558fd73 BB |
993 | else |
994 | iput(ip); | |
995 | } | |
996 | ||
900d09b2 | 997 | /* ARGSUSED */ |
34dc7c2f | 998 | void |
428870ff | 999 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 1000 | { |
428870ff | 1001 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
1002 | |
1003 | if (zgd->zgd_db) | |
1004 | dmu_buf_rele(zgd->zgd_db, zgd); | |
1005 | ||
5d43cc9a | 1006 | rangelock_exit(zgd->zgd_lr); |
34dc7c2f | 1007 | |
9babb374 BB |
1008 | /* |
1009 | * Release the vnode asynchronously as we currently have the | |
1010 | * txg stopped from syncing. | |
1011 | */ | |
0a50679c | 1012 | zfs_iput_async(ZTOI(zp)); |
428870ff | 1013 | |
34dc7c2f BB |
1014 | kmem_free(zgd, sizeof (zgd_t)); |
1015 | } | |
1016 | ||
45d1cae3 BB |
1017 | #ifdef DEBUG |
1018 | static int zil_fault_io = 0; | |
1019 | #endif | |
1020 | ||
34dc7c2f BB |
1021 | /* |
1022 | * Get data to generate a TX_WRITE intent log record. | |
1023 | */ | |
1024 | int | |
1ce23dca | 1025 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, zio_t *zio) |
34dc7c2f | 1026 | { |
0037b49e BB |
1027 | zfsvfs_t *zfsvfs = arg; |
1028 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 1029 | znode_t *zp; |
428870ff BB |
1030 | uint64_t object = lr->lr_foid; |
1031 | uint64_t offset = lr->lr_offset; | |
1032 | uint64_t size = lr->lr_length; | |
34dc7c2f | 1033 | dmu_buf_t *db; |
34dc7c2f | 1034 | zgd_t *zgd; |
34dc7c2f BB |
1035 | int error = 0; |
1036 | ||
1ce23dca PS |
1037 | ASSERT3P(lwb, !=, NULL); |
1038 | ASSERT3P(zio, !=, NULL); | |
1039 | ASSERT3U(size, !=, 0); | |
34dc7c2f BB |
1040 | |
1041 | /* | |
1042 | * Nothing to do if the file has been removed | |
1043 | */ | |
0037b49e | 1044 | if (zfs_zget(zfsvfs, object, &zp) != 0) |
2e528b49 | 1045 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 1046 | if (zp->z_unlinked) { |
9babb374 BB |
1047 | /* |
1048 | * Release the vnode asynchronously as we currently have the | |
1049 | * txg stopped from syncing. | |
1050 | */ | |
0a50679c | 1051 | zfs_iput_async(ZTOI(zp)); |
2e528b49 | 1052 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1053 | } |
1054 | ||
0eb8ba6a | 1055 | zgd = kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
1ce23dca | 1056 | zgd->zgd_lwb = lwb; |
428870ff BB |
1057 | zgd->zgd_private = zp; |
1058 | ||
34dc7c2f BB |
1059 | /* |
1060 | * Write records come in two flavors: immediate and indirect. | |
1061 | * For small writes it's cheaper to store the data with the | |
1062 | * log record (immediate); for large writes it's cheaper to | |
1063 | * sync the data and get a pointer to it (indirect) so that | |
1064 | * we don't have to write the data twice. | |
1065 | */ | |
1066 | if (buf != NULL) { /* immediate write */ | |
5d43cc9a MA |
1067 | zgd->zgd_lr = rangelock_enter(&zp->z_rangelock, |
1068 | offset, size, RL_READER); | |
34dc7c2f | 1069 | /* test for truncation needs to be done while range locked */ |
428870ff | 1070 | if (offset >= zp->z_size) { |
2e528b49 | 1071 | error = SET_ERROR(ENOENT); |
428870ff BB |
1072 | } else { |
1073 | error = dmu_read(os, object, offset, size, buf, | |
1074 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1075 | } |
428870ff | 1076 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1077 | } else { /* indirect write */ |
34dc7c2f BB |
1078 | /* |
1079 | * Have to lock the whole block to ensure when it's | |
f763c3d1 | 1080 | * written out and its checksum is being calculated |
34dc7c2f BB |
1081 | * that no one can change the data. We need to re-check |
1082 | * blocksize after we get the lock in case it's changed! | |
1083 | */ | |
1084 | for (;;) { | |
428870ff BB |
1085 | uint64_t blkoff; |
1086 | size = zp->z_blksz; | |
1087 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1088 | offset -= blkoff; | |
5d43cc9a MA |
1089 | zgd->zgd_lr = rangelock_enter(&zp->z_rangelock, |
1090 | offset, size, RL_READER); | |
428870ff | 1091 | if (zp->z_blksz == size) |
34dc7c2f | 1092 | break; |
428870ff | 1093 | offset += blkoff; |
5d43cc9a | 1094 | rangelock_exit(zgd->zgd_lr); |
34dc7c2f BB |
1095 | } |
1096 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1097 | if (lr->lr_offset >= zp->z_size) |
2e528b49 | 1098 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1099 | #ifdef DEBUG |
1100 | if (zil_fault_io) { | |
2e528b49 | 1101 | error = SET_ERROR(EIO); |
45d1cae3 | 1102 | zil_fault_io = 0; |
45d1cae3 | 1103 | } |
45d1cae3 | 1104 | #endif |
34dc7c2f | 1105 | if (error == 0) |
428870ff BB |
1106 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1107 | DMU_READ_NO_PREFETCH); | |
1108 | ||
1109 | if (error == 0) { | |
02dc43bc | 1110 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 1111 | |
428870ff BB |
1112 | zgd->zgd_db = db; |
1113 | zgd->zgd_bp = bp; | |
1114 | ||
1115 | ASSERT(db->db_offset == offset); | |
1116 | ASSERT(db->db_size == size); | |
1117 | ||
1118 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1119 | zfs_get_done, zgd); | |
61ca48ff | 1120 | ASSERT(error || lr->lr_length <= size); |
428870ff BB |
1121 | |
1122 | /* | |
1123 | * On success, we need to wait for the write I/O | |
1124 | * initiated by dmu_sync() to complete before we can | |
1125 | * release this dbuf. We will finish everything up | |
1126 | * in the zfs_get_done() callback. | |
1127 | */ | |
1128 | if (error == 0) | |
1129 | return (0); | |
1130 | ||
1131 | if (error == EALREADY) { | |
1132 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
a1d477c2 MA |
1133 | /* |
1134 | * TX_WRITE2 relies on the data previously | |
1135 | * written by the TX_WRITE that caused | |
1136 | * EALREADY. We zero out the BP because | |
900d09b2 | 1137 | * it is the old, currently-on-disk BP. |
a1d477c2 MA |
1138 | */ |
1139 | zgd->zgd_bp = NULL; | |
1140 | BP_ZERO(bp); | |
428870ff BB |
1141 | error = 0; |
1142 | } | |
1143 | } | |
34dc7c2f | 1144 | } |
428870ff BB |
1145 | |
1146 | zfs_get_done(zgd, error); | |
1147 | ||
34dc7c2f BB |
1148 | return (error); |
1149 | } | |
1150 | ||
1151 | /*ARGSUSED*/ | |
3558fd73 BB |
1152 | int |
1153 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1154 | { |
3558fd73 | 1155 | znode_t *zp = ITOZ(ip); |
0037b49e | 1156 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
1157 | int error; |
1158 | ||
0037b49e | 1159 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1160 | ZFS_VERIFY_ZP(zp); |
1161 | ||
1162 | if (flag & V_ACE_MASK) | |
1163 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1164 | else | |
1165 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1166 | ||
0037b49e | 1167 | ZFS_EXIT(zfsvfs); |
45d1cae3 BB |
1168 | return (error); |
1169 | } | |
45d1cae3 | 1170 | |
34dc7c2f BB |
1171 | /* |
1172 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1173 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1174 | * |
3558fd73 | 1175 | * IN: dip - inode of directory to search. |
34dc7c2f | 1176 | * nm - name of entry to lookup. |
34dc7c2f | 1177 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1178 | * cr - credentials of caller. |
34dc7c2f BB |
1179 | * direntflags - directory lookup flags |
1180 | * realpnp - returned pathname. | |
1181 | * | |
3558fd73 | 1182 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1183 | * |
d3cc8b15 | 1184 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1185 | * |
1186 | * Timestamps: | |
1187 | * NA | |
1188 | */ | |
1189 | /* ARGSUSED */ | |
e5c39b95 | 1190 | int |
3558fd73 BB |
1191 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1192 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1193 | { |
3558fd73 | 1194 | znode_t *zdp = ITOZ(dip); |
0037b49e | 1195 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
3558fd73 | 1196 | int error = 0; |
45d1cae3 | 1197 | |
9b7b9cd3 GM |
1198 | /* |
1199 | * Fast path lookup, however we must skip DNLC lookup | |
1200 | * for case folding or normalizing lookups because the | |
1201 | * DNLC code only stores the passed in name. This means | |
1202 | * creating 'a' and removing 'A' on a case insensitive | |
1203 | * file system would work, but DNLC still thinks 'a' | |
1204 | * exists and won't let you create it again on the next | |
1205 | * pass through fast path. | |
1206 | */ | |
45d1cae3 BB |
1207 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { |
1208 | ||
3558fd73 | 1209 | if (!S_ISDIR(dip->i_mode)) { |
2e528b49 | 1210 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1211 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 1212 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1213 | } |
1214 | ||
1215 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1216 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1217 | if (!error) { | |
3558fd73 BB |
1218 | *ipp = dip; |
1219 | igrab(*ipp); | |
45d1cae3 BB |
1220 | return (0); |
1221 | } | |
1222 | return (error); | |
45d1cae3 BB |
1223 | } |
1224 | } | |
1225 | ||
0037b49e | 1226 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1227 | ZFS_VERIFY_ZP(zdp); |
1228 | ||
3558fd73 | 1229 | *ipp = NULL; |
34dc7c2f BB |
1230 | |
1231 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1232 | /* |
1233 | * We don't allow recursive attributes.. | |
1234 | * Maybe someday we will. | |
1235 | */ | |
428870ff | 1236 | if (zdp->z_pflags & ZFS_XATTR) { |
0037b49e | 1237 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1238 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1239 | } |
1240 | ||
3558fd73 | 1241 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
0037b49e | 1242 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1243 | return (error); |
1244 | } | |
1245 | ||
1246 | /* | |
1247 | * Do we have permission to get into attribute directory? | |
1248 | */ | |
1249 | ||
3558fd73 | 1250 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1251 | B_FALSE, cr))) { |
3558fd73 BB |
1252 | iput(*ipp); |
1253 | *ipp = NULL; | |
34dc7c2f BB |
1254 | } |
1255 | ||
0037b49e | 1256 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1257 | return (error); |
1258 | } | |
1259 | ||
3558fd73 | 1260 | if (!S_ISDIR(dip->i_mode)) { |
0037b49e | 1261 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1262 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1263 | } |
1264 | ||
1265 | /* | |
1266 | * Check accessibility of directory. | |
1267 | */ | |
1268 | ||
149e873a | 1269 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
0037b49e | 1270 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1271 | return (error); |
1272 | } | |
1273 | ||
0037b49e | 1274 | if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1275 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1276 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1277 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1278 | } |
1279 | ||
3558fd73 BB |
1280 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1281 | if ((error == 0) && (*ipp)) | |
1282 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1283 | |
0037b49e | 1284 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1285 | return (error); |
1286 | } | |
1287 | ||
1288 | /* | |
1289 | * Attempt to create a new entry in a directory. If the entry | |
1290 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1291 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1292 | * |
3558fd73 | 1293 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1294 | * name - name of new file entry. |
1295 | * vap - attributes of new file. | |
1296 | * excl - flag indicating exclusive or non-exclusive mode. | |
1297 | * mode - mode to open file with. | |
1298 | * cr - credentials of caller. | |
a35c1207 | 1299 | * flag - file flag. |
3558fd73 | 1300 | * vsecp - ACL to be set |
34dc7c2f | 1301 | * |
3558fd73 | 1302 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1303 | * |
d3cc8b15 | 1304 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1305 | * |
1306 | * Timestamps: | |
3558fd73 BB |
1307 | * dip - ctime|mtime updated if new entry created |
1308 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1309 | */ |
1310 | ||
1311 | /* ARGSUSED */ | |
e5c39b95 | 1312 | int |
3558fd73 BB |
1313 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1314 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1315 | { |
3558fd73 | 1316 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1317 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1318 | zilog_t *zilog; |
1319 | objset_t *os; | |
1320 | zfs_dirlock_t *dl; | |
1321 | dmu_tx_t *tx; | |
1322 | int error; | |
b128c09f | 1323 | uid_t uid; |
149e873a | 1324 | gid_t gid; |
428870ff | 1325 | zfs_acl_ids_t acl_ids; |
9babb374 | 1326 | boolean_t fuid_dirtied; |
428870ff | 1327 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 1328 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1329 | |
1330 | /* | |
1331 | * If we have an ephemeral id, ACL, or XVATTR then | |
1332 | * make sure file system is at proper version | |
1333 | */ | |
1334 | ||
149e873a | 1335 | gid = crgetgid(cr); |
3558fd73 | 1336 | uid = crgetuid(cr); |
b128c09f | 1337 | |
0037b49e | 1338 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1339 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1340 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1341 | |
32dec7bd | 1342 | if (name == NULL) |
1343 | return (SET_ERROR(EINVAL)); | |
1344 | ||
0037b49e | 1345 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1346 | ZFS_VERIFY_ZP(dzp); |
0037b49e BB |
1347 | os = zfsvfs->z_os; |
1348 | zilog = zfsvfs->z_log; | |
34dc7c2f | 1349 | |
0037b49e | 1350 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1351 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1352 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1353 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1354 | } |
1355 | ||
5484965a | 1356 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1357 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1358 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1359 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1360 | return (error); |
1361 | } | |
1362 | } | |
34dc7c2f | 1363 | |
3558fd73 BB |
1364 | top: |
1365 | *ipp = NULL; | |
34dc7c2f BB |
1366 | if (*name == '\0') { |
1367 | /* | |
1368 | * Null component name refers to the directory itself. | |
1369 | */ | |
3558fd73 | 1370 | igrab(dip); |
34dc7c2f BB |
1371 | zp = dzp; |
1372 | dl = NULL; | |
1373 | error = 0; | |
1374 | } else { | |
3558fd73 | 1375 | /* possible igrab(zp) */ |
34dc7c2f BB |
1376 | int zflg = 0; |
1377 | ||
1378 | if (flag & FIGNORECASE) | |
1379 | zflg |= ZCILOOK; | |
1380 | ||
1381 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1382 | NULL, NULL); | |
1383 | if (error) { | |
572e2857 BB |
1384 | if (have_acl) |
1385 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1386 | if (strcmp(name, "..") == 0) |
2e528b49 | 1387 | error = SET_ERROR(EISDIR); |
0037b49e | 1388 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1389 | return (error); |
1390 | } | |
1391 | } | |
428870ff | 1392 | |
34dc7c2f BB |
1393 | if (zp == NULL) { |
1394 | uint64_t txtype; | |
9c5167d1 | 1395 | uint64_t projid = ZFS_DEFAULT_PROJID; |
34dc7c2f BB |
1396 | |
1397 | /* | |
1398 | * Create a new file object and update the directory | |
1399 | * to reference it. | |
1400 | */ | |
149e873a | 1401 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1402 | if (have_acl) |
1403 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1404 | goto out; |
1405 | } | |
1406 | ||
1407 | /* | |
1408 | * We only support the creation of regular files in | |
1409 | * extended attribute directories. | |
1410 | */ | |
428870ff | 1411 | |
3558fd73 | 1412 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1413 | if (have_acl) |
1414 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 1415 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1416 | goto out; |
1417 | } | |
1418 | ||
428870ff BB |
1419 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1420 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1421 | goto out; |
428870ff BB |
1422 | have_acl = B_TRUE; |
1423 | ||
9c5167d1 NF |
1424 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
1425 | projid = zfs_inherit_projid(dzp); | |
1426 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
45d1cae3 | 1427 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1428 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1429 | goto out; |
1430 | } | |
1431 | ||
34dc7c2f | 1432 | tx = dmu_tx_create(os); |
428870ff BB |
1433 | |
1434 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1435 | ZFS_SA_BASE_ATTR_SIZE); | |
1436 | ||
0037b49e | 1437 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1438 | if (fuid_dirtied) |
0037b49e | 1439 | zfs_fuid_txhold(zfsvfs, tx); |
34dc7c2f | 1440 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1441 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
0037b49e | 1442 | if (!zfsvfs->z_use_sa && |
428870ff | 1443 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1444 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1445 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1446 | } |
599b8648 | 1447 | |
0735ecb3 PS |
1448 | error = dmu_tx_assign(tx, |
1449 | (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); | |
34dc7c2f BB |
1450 | if (error) { |
1451 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1452 | if (error == ERESTART) { |
e8b96c60 | 1453 | waited = B_TRUE; |
34dc7c2f BB |
1454 | dmu_tx_wait(tx); |
1455 | dmu_tx_abort(tx); | |
1456 | goto top; | |
1457 | } | |
428870ff | 1458 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1459 | dmu_tx_abort(tx); |
0037b49e | 1460 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1461 | return (error); |
1462 | } | |
428870ff | 1463 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 1464 | |
599b8648 CC |
1465 | error = zfs_link_create(dl, zp, tx, ZNEW); |
1466 | if (error != 0) { | |
1467 | /* | |
1468 | * Since, we failed to add the directory entry for it, | |
1469 | * delete the newly created dnode. | |
1470 | */ | |
1471 | zfs_znode_delete(zp, tx); | |
1472 | remove_inode_hash(ZTOI(zp)); | |
1473 | zfs_acl_ids_free(&acl_ids); | |
1474 | dmu_tx_commit(tx); | |
1475 | goto out; | |
1476 | } | |
1477 | ||
9babb374 | 1478 | if (fuid_dirtied) |
0037b49e | 1479 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 1480 | |
34dc7c2f BB |
1481 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); |
1482 | if (flag & FIGNORECASE) | |
1483 | txtype |= TX_CI; | |
1484 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1485 | vsecp, acl_ids.z_fuidp, vap); |
1486 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1487 | dmu_tx_commit(tx); |
1488 | } else { | |
1489 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1490 | ||
572e2857 BB |
1491 | if (have_acl) |
1492 | zfs_acl_ids_free(&acl_ids); | |
1493 | have_acl = B_FALSE; | |
1494 | ||
34dc7c2f BB |
1495 | /* |
1496 | * A directory entry already exists for this name. | |
1497 | */ | |
1498 | /* | |
1499 | * Can't truncate an existing file if in exclusive mode. | |
1500 | */ | |
3558fd73 | 1501 | if (excl) { |
2e528b49 | 1502 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1503 | goto out; |
1504 | } | |
1505 | /* | |
1506 | * Can't open a directory for writing. | |
1507 | */ | |
3558fd73 | 1508 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1509 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1510 | goto out; |
1511 | } | |
1512 | /* | |
1513 | * Verify requested access to file. | |
1514 | */ | |
1515 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1516 | goto out; | |
1517 | } | |
1518 | ||
1519 | mutex_enter(&dzp->z_lock); | |
1520 | dzp->z_seq++; | |
1521 | mutex_exit(&dzp->z_lock); | |
1522 | ||
1523 | /* | |
1524 | * Truncate regular files if requested. | |
1525 | */ | |
3558fd73 BB |
1526 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1527 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 1528 | /* we can't hold any locks when calling zfs_freesp() */ |
609603a5 B |
1529 | if (dl) { |
1530 | zfs_dirent_unlock(dl); | |
1531 | dl = NULL; | |
1532 | } | |
34dc7c2f | 1533 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1534 | } |
1535 | } | |
1536 | out: | |
1537 | ||
1538 | if (dl) | |
1539 | zfs_dirent_unlock(dl); | |
1540 | ||
1541 | if (error) { | |
1542 | if (zp) | |
3558fd73 | 1543 | iput(ZTOI(zp)); |
34dc7c2f | 1544 | } else { |
960e08fe BB |
1545 | zfs_inode_update(dzp); |
1546 | zfs_inode_update(zp); | |
3558fd73 | 1547 | *ipp = ZTOI(zp); |
34dc7c2f | 1548 | } |
34dc7c2f | 1549 | |
0037b49e | 1550 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1551 | zil_commit(zilog, 0); |
428870ff | 1552 | |
0037b49e | 1553 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1554 | return (error); |
1555 | } | |
1556 | ||
ace1eae8 CC |
1557 | /* ARGSUSED */ |
1558 | int | |
1559 | zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, | |
1560 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
1561 | { | |
1562 | znode_t *zp = NULL, *dzp = ITOZ(dip); | |
0037b49e | 1563 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
ace1eae8 CC |
1564 | objset_t *os; |
1565 | dmu_tx_t *tx; | |
1566 | int error; | |
1567 | uid_t uid; | |
1568 | gid_t gid; | |
1569 | zfs_acl_ids_t acl_ids; | |
9c5167d1 | 1570 | uint64_t projid = ZFS_DEFAULT_PROJID; |
ace1eae8 CC |
1571 | boolean_t fuid_dirtied; |
1572 | boolean_t have_acl = B_FALSE; | |
1573 | boolean_t waited = B_FALSE; | |
1574 | ||
1575 | /* | |
1576 | * If we have an ephemeral id, ACL, or XVATTR then | |
1577 | * make sure file system is at proper version | |
1578 | */ | |
1579 | ||
1580 | gid = crgetgid(cr); | |
1581 | uid = crgetuid(cr); | |
1582 | ||
0037b49e | 1583 | if (zfsvfs->z_use_fuids == B_FALSE && |
ace1eae8 CC |
1584 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
1585 | return (SET_ERROR(EINVAL)); | |
1586 | ||
0037b49e | 1587 | ZFS_ENTER(zfsvfs); |
ace1eae8 | 1588 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1589 | os = zfsvfs->z_os; |
ace1eae8 CC |
1590 | |
1591 | if (vap->va_mask & ATTR_XVATTR) { | |
1592 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
1593 | crgetuid(cr), cr, vap->va_mode)) != 0) { | |
0037b49e | 1594 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1595 | return (error); |
1596 | } | |
1597 | } | |
1598 | ||
1599 | top: | |
1600 | *ipp = NULL; | |
1601 | ||
1602 | /* | |
1603 | * Create a new file object and update the directory | |
1604 | * to reference it. | |
1605 | */ | |
1606 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { | |
1607 | if (have_acl) | |
1608 | zfs_acl_ids_free(&acl_ids); | |
1609 | goto out; | |
1610 | } | |
1611 | ||
1612 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, | |
1613 | cr, vsecp, &acl_ids)) != 0) | |
1614 | goto out; | |
1615 | have_acl = B_TRUE; | |
1616 | ||
9c5167d1 NF |
1617 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
1618 | projid = zfs_inherit_projid(dzp); | |
1619 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
ace1eae8 CC |
1620 | zfs_acl_ids_free(&acl_ids); |
1621 | error = SET_ERROR(EDQUOT); | |
1622 | goto out; | |
1623 | } | |
1624 | ||
1625 | tx = dmu_tx_create(os); | |
1626 | ||
1627 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1628 | ZFS_SA_BASE_ATTR_SIZE); | |
0037b49e | 1629 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 1630 | |
0037b49e | 1631 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
ace1eae8 | 1632 | if (fuid_dirtied) |
0037b49e BB |
1633 | zfs_fuid_txhold(zfsvfs, tx); |
1634 | if (!zfsvfs->z_use_sa && | |
ace1eae8 CC |
1635 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
1636 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
1637 | 0, acl_ids.z_aclp->z_acl_bytes); | |
1638 | } | |
0735ecb3 | 1639 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
ace1eae8 CC |
1640 | if (error) { |
1641 | if (error == ERESTART) { | |
1642 | waited = B_TRUE; | |
1643 | dmu_tx_wait(tx); | |
1644 | dmu_tx_abort(tx); | |
1645 | goto top; | |
1646 | } | |
1647 | zfs_acl_ids_free(&acl_ids); | |
1648 | dmu_tx_abort(tx); | |
0037b49e | 1649 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1650 | return (error); |
1651 | } | |
1652 | zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); | |
1653 | ||
1654 | if (fuid_dirtied) | |
0037b49e | 1655 | zfs_fuid_sync(zfsvfs, tx); |
ace1eae8 CC |
1656 | |
1657 | /* Add to unlinked set */ | |
a43570c5 | 1658 | zp->z_unlinked = B_TRUE; |
ace1eae8 CC |
1659 | zfs_unlinked_add(zp, tx); |
1660 | zfs_acl_ids_free(&acl_ids); | |
1661 | dmu_tx_commit(tx); | |
1662 | out: | |
1663 | ||
1664 | if (error) { | |
1665 | if (zp) | |
1666 | iput(ZTOI(zp)); | |
1667 | } else { | |
1668 | zfs_inode_update(dzp); | |
1669 | zfs_inode_update(zp); | |
1670 | *ipp = ZTOI(zp); | |
1671 | } | |
1672 | ||
0037b49e | 1673 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1674 | return (error); |
1675 | } | |
1676 | ||
34dc7c2f BB |
1677 | /* |
1678 | * Remove an entry from a directory. | |
1679 | * | |
3558fd73 | 1680 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1681 | * name - name of entry to remove. |
1682 | * cr - credentials of caller. | |
841a7a98 | 1683 | * flags - case flags. |
34dc7c2f BB |
1684 | * |
1685 | * RETURN: 0 if success | |
1686 | * error code if failure | |
1687 | * | |
1688 | * Timestamps: | |
3558fd73 BB |
1689 | * dip - ctime|mtime |
1690 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1691 | */ |
428870ff BB |
1692 | |
1693 | uint64_t null_xattr = 0; | |
1694 | ||
34dc7c2f | 1695 | /*ARGSUSED*/ |
e5c39b95 | 1696 | int |
da5e151f | 1697 | zfs_remove(struct inode *dip, char *name, cred_t *cr, int flags) |
34dc7c2f | 1698 | { |
3558fd73 | 1699 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1700 | znode_t *xzp; |
3558fd73 | 1701 | struct inode *ip; |
0037b49e | 1702 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 1703 | zilog_t *zilog; |
a966c564 | 1704 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 1705 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 1706 | uint64_t obj = 0; |
dfbc8630 | 1707 | uint64_t links; |
34dc7c2f BB |
1708 | zfs_dirlock_t *dl; |
1709 | dmu_tx_t *tx; | |
a966c564 K |
1710 | boolean_t may_delete_now, delete_now = FALSE; |
1711 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
1712 | uint64_t txtype; |
1713 | pathname_t *realnmp = NULL; | |
1714 | pathname_t realnm; | |
1715 | int error; | |
1716 | int zflg = ZEXISTS; | |
e8b96c60 | 1717 | boolean_t waited = B_FALSE; |
34dc7c2f | 1718 | |
32dec7bd | 1719 | if (name == NULL) |
1720 | return (SET_ERROR(EINVAL)); | |
1721 | ||
0037b49e | 1722 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1723 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1724 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1725 | |
1726 | if (flags & FIGNORECASE) { | |
1727 | zflg |= ZCILOOK; | |
1728 | pn_alloc(&realnm); | |
1729 | realnmp = &realnm; | |
1730 | } | |
1731 | ||
1732 | top: | |
572e2857 BB |
1733 | xattr_obj = 0; |
1734 | xzp = NULL; | |
34dc7c2f BB |
1735 | /* |
1736 | * Attempt to lock directory; fail if entry doesn't exist. | |
1737 | */ | |
149e873a BB |
1738 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1739 | NULL, realnmp))) { | |
34dc7c2f BB |
1740 | if (realnmp) |
1741 | pn_free(realnmp); | |
0037b49e | 1742 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1743 | return (error); |
1744 | } | |
1745 | ||
3558fd73 | 1746 | ip = ZTOI(zp); |
34dc7c2f | 1747 | |
149e873a | 1748 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1749 | goto out; |
1750 | } | |
1751 | ||
1752 | /* | |
1753 | * Need to use rmdir for removing directories. | |
1754 | */ | |
3558fd73 | 1755 | if (S_ISDIR(ip->i_mode)) { |
2e528b49 | 1756 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1757 | goto out; |
1758 | } | |
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 | 1856 | mutex_enter(&xzp->z_lock); |
a43570c5 | 1857 | xzp->z_unlinked = B_TRUE; |
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; | |
8e556c5e | 1889 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj, unlinked); |
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. | |
841a7a98 | 1925 | * flags - case flags. |
34dc7c2f BB |
1926 | * vsecp - ACL to be set |
1927 | * | |
3558fd73 | 1928 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1929 | * |
1930 | * RETURN: 0 if success | |
1931 | * error code if failure | |
1932 | * | |
1933 | * Timestamps: | |
3558fd73 BB |
1934 | * dip - ctime|mtime updated |
1935 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1936 | */ |
1937 | /*ARGSUSED*/ | |
e5c39b95 | 1938 | int |
3558fd73 BB |
1939 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1940 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1941 | { |
3558fd73 | 1942 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1943 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1944 | zilog_t *zilog; |
1945 | zfs_dirlock_t *dl; | |
1946 | uint64_t txtype; | |
1947 | dmu_tx_t *tx; | |
1948 | int error; | |
34dc7c2f | 1949 | int zf = ZNEW; |
b128c09f BB |
1950 | uid_t uid; |
1951 | gid_t gid = crgetgid(cr); | |
428870ff | 1952 | zfs_acl_ids_t acl_ids; |
9babb374 | 1953 | boolean_t fuid_dirtied; |
e8b96c60 | 1954 | boolean_t waited = B_FALSE; |
34dc7c2f | 1955 | |
3558fd73 | 1956 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1957 | |
1958 | /* | |
1959 | * If we have an ephemeral id, ACL, or XVATTR then | |
1960 | * make sure file system is at proper version | |
1961 | */ | |
1962 | ||
3558fd73 | 1963 | uid = crgetuid(cr); |
0037b49e | 1964 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1965 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1966 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1967 | |
32dec7bd | 1968 | if (dirname == NULL) |
1969 | return (SET_ERROR(EINVAL)); | |
1970 | ||
0037b49e | 1971 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1972 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1973 | zilog = zfsvfs->z_log; |
34dc7c2f | 1974 | |
428870ff | 1975 | if (dzp->z_pflags & ZFS_XATTR) { |
0037b49e | 1976 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1977 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1978 | } |
1979 | ||
0037b49e | 1980 | if (zfsvfs->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1981 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1982 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1983 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1984 | } |
1985 | if (flags & FIGNORECASE) | |
1986 | zf |= ZCILOOK; | |
1987 | ||
5484965a | 1988 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1989 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1990 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1991 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1992 | return (error); |
1993 | } | |
428870ff | 1994 | } |
34dc7c2f | 1995 | |
428870ff BB |
1996 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1997 | vsecp, &acl_ids)) != 0) { | |
0037b49e | 1998 | ZFS_EXIT(zfsvfs); |
428870ff BB |
1999 | return (error); |
2000 | } | |
34dc7c2f BB |
2001 | /* |
2002 | * First make sure the new directory doesn't exist. | |
428870ff BB |
2003 | * |
2004 | * Existence is checked first to make sure we don't return | |
2005 | * EACCES instead of EEXIST which can cause some applications | |
2006 | * to fail. | |
34dc7c2f BB |
2007 | */ |
2008 | top: | |
3558fd73 | 2009 | *ipp = NULL; |
34dc7c2f | 2010 | |
149e873a BB |
2011 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
2012 | NULL, NULL))) { | |
428870ff | 2013 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 2014 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2015 | return (error); |
2016 | } | |
2017 | ||
149e873a | 2018 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 2019 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2020 | zfs_dirent_unlock(dl); |
0037b49e | 2021 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2022 | return (error); |
2023 | } | |
2024 | ||
9c5167d1 | 2025 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) { |
45d1cae3 | 2026 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 2027 | zfs_dirent_unlock(dl); |
0037b49e | 2028 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2029 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
2030 | } |
2031 | ||
34dc7c2f BB |
2032 | /* |
2033 | * Add a new entry to the directory. | |
2034 | */ | |
0037b49e | 2035 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f BB |
2036 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
2037 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
0037b49e | 2038 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 2039 | if (fuid_dirtied) |
0037b49e BB |
2040 | zfs_fuid_txhold(zfsvfs, tx); |
2041 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
2042 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
2043 | acl_ids.z_aclp->z_acl_bytes); | |
2044 | } | |
2045 | ||
2046 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
2047 | ZFS_SA_BASE_ATTR_SIZE); | |
2048 | ||
0735ecb3 | 2049 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
2050 | if (error) { |
2051 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2052 | if (error == ERESTART) { |
e8b96c60 | 2053 | waited = B_TRUE; |
34dc7c2f BB |
2054 | dmu_tx_wait(tx); |
2055 | dmu_tx_abort(tx); | |
2056 | goto top; | |
2057 | } | |
428870ff | 2058 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2059 | dmu_tx_abort(tx); |
0037b49e | 2060 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2061 | return (error); |
2062 | } | |
2063 | ||
2064 | /* | |
2065 | * Create new node. | |
2066 | */ | |
428870ff | 2067 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f BB |
2068 | |
2069 | /* | |
2070 | * Now put new name in parent dir. | |
2071 | */ | |
599b8648 CC |
2072 | error = zfs_link_create(dl, zp, tx, ZNEW); |
2073 | if (error != 0) { | |
2074 | zfs_znode_delete(zp, tx); | |
2075 | remove_inode_hash(ZTOI(zp)); | |
2076 | goto out; | |
2077 | } | |
2078 | ||
2079 | if (fuid_dirtied) | |
2080 | zfs_fuid_sync(zfsvfs, tx); | |
34dc7c2f | 2081 | |
3558fd73 | 2082 | *ipp = ZTOI(zp); |
34dc7c2f BB |
2083 | |
2084 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
2085 | if (flags & FIGNORECASE) | |
2086 | txtype |= TX_CI; | |
9babb374 BB |
2087 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
2088 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 2089 | |
599b8648 | 2090 | out: |
9babb374 | 2091 | zfs_acl_ids_free(&acl_ids); |
428870ff | 2092 | |
34dc7c2f BB |
2093 | dmu_tx_commit(tx); |
2094 | ||
2095 | zfs_dirent_unlock(dl); | |
2096 | ||
0037b49e | 2097 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2098 | zil_commit(zilog, 0); |
428870ff | 2099 | |
599b8648 CC |
2100 | if (error != 0) { |
2101 | iput(ZTOI(zp)); | |
2102 | } else { | |
2103 | zfs_inode_update(dzp); | |
2104 | zfs_inode_update(zp); | |
2105 | } | |
0037b49e | 2106 | ZFS_EXIT(zfsvfs); |
599b8648 | 2107 | return (error); |
34dc7c2f BB |
2108 | } |
2109 | ||
2110 | /* | |
2111 | * Remove a directory subdir entry. If the current working | |
2112 | * directory is the same as the subdir to be removed, the | |
2113 | * remove will fail. | |
2114 | * | |
3558fd73 | 2115 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 2116 | * name - name of directory to be removed. |
3558fd73 | 2117 | * cwd - inode of current working directory. |
34dc7c2f | 2118 | * cr - credentials of caller. |
34dc7c2f BB |
2119 | * flags - case flags |
2120 | * | |
d3cc8b15 | 2121 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
2122 | * |
2123 | * Timestamps: | |
3558fd73 | 2124 | * dip - ctime|mtime updated |
34dc7c2f BB |
2125 | */ |
2126 | /*ARGSUSED*/ | |
e5c39b95 | 2127 | int |
3558fd73 BB |
2128 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
2129 | int flags) | |
34dc7c2f | 2130 | { |
3558fd73 | 2131 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 2132 | znode_t *zp; |
3558fd73 | 2133 | struct inode *ip; |
0037b49e | 2134 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
2135 | zilog_t *zilog; |
2136 | zfs_dirlock_t *dl; | |
2137 | dmu_tx_t *tx; | |
2138 | int error; | |
2139 | int zflg = ZEXISTS; | |
e8b96c60 | 2140 | boolean_t waited = B_FALSE; |
34dc7c2f | 2141 | |
32dec7bd | 2142 | if (name == NULL) |
2143 | return (SET_ERROR(EINVAL)); | |
2144 | ||
0037b49e | 2145 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2146 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 2147 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2148 | |
2149 | if (flags & FIGNORECASE) | |
2150 | zflg |= ZCILOOK; | |
2151 | top: | |
2152 | zp = NULL; | |
2153 | ||
2154 | /* | |
2155 | * Attempt to lock directory; fail if entry doesn't exist. | |
2156 | */ | |
149e873a BB |
2157 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
2158 | NULL, NULL))) { | |
0037b49e | 2159 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2160 | return (error); |
2161 | } | |
2162 | ||
3558fd73 | 2163 | ip = ZTOI(zp); |
34dc7c2f | 2164 | |
149e873a | 2165 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
2166 | goto out; |
2167 | } | |
2168 | ||
3558fd73 | 2169 | if (!S_ISDIR(ip->i_mode)) { |
2e528b49 | 2170 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
2171 | goto out; |
2172 | } | |
2173 | ||
3558fd73 | 2174 | if (ip == cwd) { |
2e528b49 | 2175 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
2176 | goto out; |
2177 | } | |
2178 | ||
34dc7c2f | 2179 | /* |
4e33ba4c | 2180 | * Grab a lock on the directory to make sure that no one is |
34dc7c2f BB |
2181 | * trying to add (or lookup) entries while we are removing it. |
2182 | */ | |
2183 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
2184 | ||
2185 | /* | |
2186 | * Grab a lock on the parent pointer to make sure we play well | |
2187 | * with the treewalk and directory rename code. | |
2188 | */ | |
2189 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
2190 | ||
0037b49e | 2191 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 2192 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 2193 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
0037b49e | 2194 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
2195 | zfs_sa_upgrade_txholds(tx, zp); |
2196 | zfs_sa_upgrade_txholds(tx, dzp); | |
db707ad0 | 2197 | dmu_tx_mark_netfree(tx); |
0735ecb3 | 2198 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
2199 | if (error) { |
2200 | rw_exit(&zp->z_parent_lock); | |
2201 | rw_exit(&zp->z_name_lock); | |
2202 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2203 | if (error == ERESTART) { |
e8b96c60 | 2204 | waited = B_TRUE; |
34dc7c2f BB |
2205 | dmu_tx_wait(tx); |
2206 | dmu_tx_abort(tx); | |
ea7e86d8 | 2207 | iput(ip); |
34dc7c2f BB |
2208 | goto top; |
2209 | } | |
2210 | dmu_tx_abort(tx); | |
ea7e86d8 | 2211 | iput(ip); |
0037b49e | 2212 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2213 | return (error); |
2214 | } | |
2215 | ||
2216 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
2217 | ||
2218 | if (error == 0) { | |
2219 | uint64_t txtype = TX_RMDIR; | |
2220 | if (flags & FIGNORECASE) | |
2221 | txtype |= TX_CI; | |
8e556c5e CC |
2222 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT, |
2223 | B_FALSE); | |
34dc7c2f BB |
2224 | } |
2225 | ||
2226 | dmu_tx_commit(tx); | |
2227 | ||
2228 | rw_exit(&zp->z_parent_lock); | |
2229 | rw_exit(&zp->z_name_lock); | |
2230 | out: | |
2231 | zfs_dirent_unlock(dl); | |
2232 | ||
59157910 BB |
2233 | zfs_inode_update(dzp); |
2234 | zfs_inode_update(zp); | |
3558fd73 | 2235 | iput(ip); |
34dc7c2f | 2236 | |
0037b49e | 2237 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2238 | zil_commit(zilog, 0); |
428870ff | 2239 | |
0037b49e | 2240 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2241 | return (error); |
2242 | } | |
2243 | ||
2244 | /* | |
841a7a98 TK |
2245 | * Read directory entries from the given directory cursor position and emit |
2246 | * name and position for each entry. | |
34dc7c2f | 2247 | * |
3558fd73 | 2248 | * IN: ip - inode of directory to read. |
841a7a98 TK |
2249 | * ctx - directory entry context. |
2250 | * cr - credentials of caller. | |
34dc7c2f BB |
2251 | * |
2252 | * RETURN: 0 if success | |
2253 | * error code if failure | |
2254 | * | |
2255 | * Timestamps: | |
3558fd73 | 2256 | * ip - atime updated |
34dc7c2f BB |
2257 | * |
2258 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2259 | * This allows us to use the low range for "special" directory entries: | |
2260 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2261 | * we use the offset 2 for the '.zfs' directory. | |
2262 | */ | |
2263 | /* ARGSUSED */ | |
3558fd73 | 2264 | int |
9464b959 | 2265 | zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) |
34dc7c2f | 2266 | { |
3558fd73 | 2267 | znode_t *zp = ITOZ(ip); |
0037b49e | 2268 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2269 | objset_t *os; |
34dc7c2f BB |
2270 | zap_cursor_t zc; |
2271 | zap_attribute_t zap; | |
34dc7c2f BB |
2272 | int error; |
2273 | uint8_t prefetch; | |
c12e3a59 | 2274 | uint8_t type; |
3558fd73 BB |
2275 | int done = 0; |
2276 | uint64_t parent; | |
c12e3a59 | 2277 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2278 | |
0037b49e | 2279 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2280 | ZFS_VERIFY_ZP(zp); |
2281 | ||
0037b49e | 2282 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
3558fd73 BB |
2283 | &parent, sizeof (parent))) != 0) |
2284 | goto out; | |
34dc7c2f BB |
2285 | |
2286 | /* | |
2287 | * Quit if directory has been removed (posix) | |
2288 | */ | |
3558fd73 BB |
2289 | if (zp->z_unlinked) |
2290 | goto out; | |
2291 | ||
c12e3a59 | 2292 | error = 0; |
0037b49e | 2293 | os = zfsvfs->z_os; |
c12e3a59 | 2294 | offset = ctx->pos; |
34dc7c2f BB |
2295 | prefetch = zp->z_zn_prefetch; |
2296 | ||
2297 | /* | |
2298 | * Initialize the iterator cursor. | |
2299 | */ | |
c12e3a59 | 2300 | if (offset <= 3) { |
34dc7c2f BB |
2301 | /* |
2302 | * Start iteration from the beginning of the directory. | |
2303 | */ | |
2304 | zap_cursor_init(&zc, os, zp->z_id); | |
2305 | } else { | |
2306 | /* | |
2307 | * The offset is a serialized cursor. | |
2308 | */ | |
c12e3a59 | 2309 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2310 | } |
2311 | ||
34dc7c2f BB |
2312 | /* |
2313 | * Transform to file-system independent format | |
2314 | */ | |
3558fd73 BB |
2315 | while (!done) { |
2316 | uint64_t objnum; | |
34dc7c2f BB |
2317 | /* |
2318 | * Special case `.', `..', and `.zfs'. | |
2319 | */ | |
c12e3a59 | 2320 | if (offset == 0) { |
34dc7c2f BB |
2321 | (void) strcpy(zap.za_name, "."); |
2322 | zap.za_normalization_conflict = 0; | |
2323 | objnum = zp->z_id; | |
c12e3a59 RY |
2324 | type = DT_DIR; |
2325 | } else if (offset == 1) { | |
34dc7c2f BB |
2326 | (void) strcpy(zap.za_name, ".."); |
2327 | zap.za_normalization_conflict = 0; | |
428870ff | 2328 | objnum = parent; |
c12e3a59 RY |
2329 | type = DT_DIR; |
2330 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2331 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2332 | zap.za_normalization_conflict = 0; | |
2333 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 2334 | type = DT_DIR; |
34dc7c2f BB |
2335 | } else { |
2336 | /* | |
2337 | * Grab next entry. | |
2338 | */ | |
3558fd73 BB |
2339 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2340 | if (error == ENOENT) | |
34dc7c2f BB |
2341 | break; |
2342 | else | |
2343 | goto update; | |
2344 | } | |
2345 | ||
0c5dde49 BB |
2346 | /* |
2347 | * Allow multiple entries provided the first entry is | |
2348 | * the object id. Non-zpl consumers may safely make | |
2349 | * use of the additional space. | |
2350 | * | |
2351 | * XXX: This should be a feature flag for compatibility | |
2352 | */ | |
34dc7c2f | 2353 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2354 | zap.za_num_integers == 0) { |
34dc7c2f | 2355 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2356 | "entry, obj = %lld, offset = %lld, " |
2357 | "length = %d, num = %lld\n", | |
34dc7c2f | 2358 | (u_longlong_t)zp->z_id, |
c12e3a59 | 2359 | (u_longlong_t)offset, |
0c5dde49 BB |
2360 | zap.za_integer_length, |
2361 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 2362 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2363 | goto update; |
2364 | } | |
2365 | ||
2366 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 2367 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2368 | } |
0f37d0c8 | 2369 | |
9464b959 | 2370 | done = !zpl_dir_emit(ctx, zap.za_name, strlen(zap.za_name), |
c12e3a59 | 2371 | objnum, type); |
0f37d0c8 | 2372 | if (done) |
34dc7c2f | 2373 | break; |
34dc7c2f BB |
2374 | |
2375 | /* Prefetch znode */ | |
3558fd73 | 2376 | if (prefetch) { |
fcff0f35 PD |
2377 | dmu_prefetch(os, objnum, 0, 0, 0, |
2378 | ZIO_PRIORITY_SYNC_READ); | |
3558fd73 | 2379 | } |
34dc7c2f | 2380 | |
c12e3a59 RY |
2381 | /* |
2382 | * Move to the next entry, fill in the previous offset. | |
2383 | */ | |
2384 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2385 | zap_cursor_advance(&zc); |
c12e3a59 | 2386 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2387 | } else { |
c12e3a59 | 2388 | offset += 1; |
34dc7c2f | 2389 | } |
c12e3a59 | 2390 | ctx->pos = offset; |
34dc7c2f BB |
2391 | } |
2392 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2393 | ||
34dc7c2f BB |
2394 | update: |
2395 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2396 | if (error == ENOENT) |
2397 | error = 0; | |
3558fd73 | 2398 | out: |
0037b49e | 2399 | ZFS_EXIT(zfsvfs); |
34dc7c2f | 2400 | |
34dc7c2f BB |
2401 | return (error); |
2402 | } | |
2403 | ||
d5446cfc BB |
2404 | ulong_t zfs_fsync_sync_cnt = 4; |
2405 | ||
e5c39b95 | 2406 | int |
3558fd73 | 2407 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2408 | { |
3558fd73 | 2409 | znode_t *zp = ITOZ(ip); |
0037b49e | 2410 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2411 | |
d5446cfc BB |
2412 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2413 | ||
0037b49e BB |
2414 | if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2415 | ZFS_ENTER(zfsvfs); | |
428870ff | 2416 | ZFS_VERIFY_ZP(zp); |
0037b49e BB |
2417 | zil_commit(zfsvfs->z_log, zp->z_id); |
2418 | ZFS_EXIT(zfsvfs); | |
428870ff | 2419 | } |
07012da6 CC |
2420 | tsd_set(zfs_fsyncer_key, NULL); |
2421 | ||
34dc7c2f BB |
2422 | return (0); |
2423 | } | |
2424 | ||
2425 | ||
2426 | /* | |
2427 | * Get the requested file attributes and place them in the provided | |
2428 | * vattr structure. | |
2429 | * | |
3558fd73 | 2430 | * IN: ip - inode of file. |
5484965a BB |
2431 | * vap - va_mask identifies requested attributes. |
2432 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2433 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2434 | * cr - credentials of caller. | |
34dc7c2f | 2435 | * |
5484965a BB |
2436 | * OUT: vap - attribute values. |
2437 | * | |
2438 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2439 | */ |
2440 | /* ARGSUSED */ | |
e5c39b95 | 2441 | int |
5484965a | 2442 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2443 | { |
3558fd73 | 2444 | znode_t *zp = ITOZ(ip); |
0037b49e | 2445 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
2446 | int error = 0; |
2447 | uint64_t links; | |
0df9673f | 2448 | uint64_t atime[2], mtime[2], ctime[2]; |
5484965a BB |
2449 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2450 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2451 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
0df9673f | 2452 | sa_bulk_attr_t bulk[3]; |
428870ff | 2453 | int count = 0; |
34dc7c2f | 2454 | |
0037b49e | 2455 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2456 | ZFS_VERIFY_ZP(zp); |
428870ff | 2457 | |
5484965a | 2458 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2459 | |
0037b49e BB |
2460 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); |
2461 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
2462 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
428870ff BB |
2463 | |
2464 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
0037b49e | 2465 | ZFS_EXIT(zfsvfs); |
428870ff BB |
2466 | return (error); |
2467 | } | |
34dc7c2f | 2468 | |
34dc7c2f BB |
2469 | /* |
2470 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2471 | * Also, if we are the owner don't bother, since owner should | |
2472 | * always be allowed to read basic attributes of file. | |
2473 | */ | |
572e2857 | 2474 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2475 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2476 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2477 | skipaclchk, cr))) { | |
0037b49e | 2478 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2479 | return (error); |
2480 | } | |
2481 | } | |
2482 | ||
2483 | /* | |
2484 | * Return all attributes. It's cheaper to provide the answer | |
2485 | * than to determine whether we were asked the question. | |
2486 | */ | |
2487 | ||
9babb374 | 2488 | mutex_enter(&zp->z_lock); |
5484965a BB |
2489 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2490 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2491 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2492 | vap->va_nodeid = zp->z_id; |
0037b49e | 2493 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) |
dfbc8630 | 2494 | links = ZTOI(zp)->i_nlink + 1; |
34dc7c2f | 2495 | else |
dfbc8630 | 2496 | links = ZTOI(zp)->i_nlink; |
5484965a BB |
2497 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2498 | vap->va_size = i_size_read(ip); | |
2499 | vap->va_rdev = ip->i_rdev; | |
2500 | vap->va_seq = ip->i_generation; | |
2501 | ||
2502 | /* | |
2503 | * Add in any requested optional attributes and the create time. | |
2504 | * Also set the corresponding bits in the returned attribute bitmap. | |
2505 | */ | |
0037b49e | 2506 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { |
5484965a BB |
2507 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { |
2508 | xoap->xoa_archive = | |
2509 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2510 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2511 | } | |
2512 | ||
2513 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2514 | xoap->xoa_readonly = | |
2515 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2516 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2517 | } | |
2518 | ||
2519 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2520 | xoap->xoa_system = | |
2521 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2522 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2523 | } | |
2524 | ||
2525 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2526 | xoap->xoa_hidden = | |
2527 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2528 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2529 | } | |
2530 | ||
2531 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2532 | xoap->xoa_nounlink = | |
2533 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2534 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2535 | } | |
2536 | ||
2537 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2538 | xoap->xoa_immutable = | |
2539 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2540 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2541 | } | |
2542 | ||
2543 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2544 | xoap->xoa_appendonly = | |
2545 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2546 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2547 | } | |
2548 | ||
2549 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2550 | xoap->xoa_nodump = | |
2551 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2552 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2553 | } | |
2554 | ||
2555 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2556 | xoap->xoa_opaque = | |
2557 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2558 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2559 | } | |
2560 | ||
2561 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2562 | xoap->xoa_av_quarantined = | |
2563 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2564 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2565 | } | |
2566 | ||
2567 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2568 | xoap->xoa_av_modified = | |
2569 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2570 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2571 | } | |
2572 | ||
2573 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2574 | S_ISREG(ip->i_mode)) { | |
2575 | zfs_sa_get_scanstamp(zp, xvap); | |
2576 | } | |
34dc7c2f | 2577 | |
5484965a BB |
2578 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2579 | uint64_t times[2]; | |
2580 | ||
0037b49e | 2581 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), |
5484965a BB |
2582 | times, sizeof (times)); |
2583 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2584 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2585 | } | |
2586 | ||
2587 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2588 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2589 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2590 | } | |
2591 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
278f2236 | 2592 | xoap->xoa_generation = ip->i_generation; |
5484965a BB |
2593 | XVA_SET_RTN(xvap, XAT_GEN); |
2594 | } | |
2595 | ||
2596 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2597 | xoap->xoa_offline = | |
2598 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2599 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2600 | } | |
2601 | ||
2602 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2603 | xoap->xoa_sparse = | |
2604 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2605 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2606 | } | |
9c5167d1 NF |
2607 | |
2608 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { | |
2609 | xoap->xoa_projinherit = | |
2610 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0); | |
2611 | XVA_SET_RTN(xvap, XAT_PROJINHERIT); | |
2612 | } | |
2613 | ||
2614 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
2615 | xoap->xoa_projid = zp->z_projid; | |
2616 | XVA_SET_RTN(xvap, XAT_PROJID); | |
2617 | } | |
5484965a BB |
2618 | } |
2619 | ||
0df9673f | 2620 | ZFS_TIME_DECODE(&vap->va_atime, atime); |
5484965a BB |
2621 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); |
2622 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2623 | |
2624 | mutex_exit(&zp->z_lock); | |
2625 | ||
5484965a | 2626 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2627 | |
2628 | if (zp->z_blksz == 0) { | |
2629 | /* | |
2630 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2631 | */ | |
0037b49e | 2632 | vap->va_blksize = zfsvfs->z_max_blksz; |
34dc7c2f BB |
2633 | } |
2634 | ||
0037b49e | 2635 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2636 | return (0); |
2637 | } | |
2638 | ||
057e8eee BB |
2639 | /* |
2640 | * Get the basic file attributes and place them in the provided kstat | |
2641 | * structure. The inode is assumed to be the authoritative source | |
2642 | * for most of the attributes. However, the znode currently has the | |
2643 | * authoritative atime, blksize, and block count. | |
2644 | * | |
2645 | * IN: ip - inode of file. | |
2646 | * | |
2647 | * OUT: sp - kstat values. | |
2648 | * | |
2649 | * RETURN: 0 (always succeeds) | |
2650 | */ | |
2651 | /* ARGSUSED */ | |
2652 | int | |
2653 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2654 | { | |
2655 | znode_t *zp = ITOZ(ip); | |
0037b49e | 2656 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
b585bc4a BB |
2657 | uint32_t blksize; |
2658 | u_longlong_t nblocks; | |
057e8eee | 2659 | |
0037b49e | 2660 | ZFS_ENTER(zfsvfs); |
a7b125e9 GB |
2661 | ZFS_VERIFY_ZP(zp); |
2662 | ||
057e8eee BB |
2663 | mutex_enter(&zp->z_lock); |
2664 | ||
2665 | generic_fillattr(ip, sp); | |
97aa3ba4 TK |
2666 | /* |
2667 | * +1 link count for root inode with visible '.zfs' directory. | |
2668 | */ | |
2669 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) | |
2670 | if (sp->nlink < ZFS_LINK_MAX) | |
2671 | sp->nlink++; | |
057e8eee | 2672 | |
b585bc4a BB |
2673 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2674 | sp->blksize = blksize; | |
2675 | sp->blocks = nblocks; | |
2676 | ||
057e8eee BB |
2677 | if (unlikely(zp->z_blksz == 0)) { |
2678 | /* | |
2679 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2680 | */ | |
0037b49e | 2681 | sp->blksize = zfsvfs->z_max_blksz; |
057e8eee BB |
2682 | } |
2683 | ||
2684 | mutex_exit(&zp->z_lock); | |
2685 | ||
aa9b2708 AV |
2686 | /* |
2687 | * Required to prevent NFS client from detecting different inode | |
2688 | * numbers of snapshot root dentry before and after snapshot mount. | |
2689 | */ | |
0037b49e | 2690 | if (zfsvfs->z_issnap) { |
aa9b2708 AV |
2691 | if (ip->i_sb->s_root->d_inode == ip) |
2692 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
0037b49e | 2693 | dmu_objset_id(zfsvfs->z_os); |
aa9b2708 AV |
2694 | } |
2695 | ||
0037b49e | 2696 | ZFS_EXIT(zfsvfs); |
a7b125e9 | 2697 | |
057e8eee BB |
2698 | return (0); |
2699 | } | |
057e8eee | 2700 | |
9c5167d1 NF |
2701 | /* |
2702 | * For the operation of changing file's user/group/project, we need to | |
2703 | * handle not only the main object that is assigned to the file directly, | |
2704 | * but also the ones that are used by the file via hidden xattr directory. | |
2705 | * | |
2706 | * Because the xattr directory may contains many EA entries, as to it may | |
2707 | * be impossible to change all of them via the transaction of changing the | |
2708 | * main object's user/group/project attributes. Then we have to change them | |
2709 | * via other multiple independent transactions one by one. It may be not good | |
2710 | * solution, but we have no better idea yet. | |
2711 | */ | |
2712 | static int | |
2713 | zfs_setattr_dir(znode_t *dzp) | |
2714 | { | |
2715 | struct inode *dxip = ZTOI(dzp); | |
2716 | struct inode *xip = NULL; | |
2717 | zfsvfs_t *zfsvfs = ITOZSB(dxip); | |
2718 | objset_t *os = zfsvfs->z_os; | |
2719 | zap_cursor_t zc; | |
2720 | zap_attribute_t zap; | |
2721 | zfs_dirlock_t *dl; | |
2722 | znode_t *zp; | |
2723 | dmu_tx_t *tx = NULL; | |
2724 | uint64_t uid, gid; | |
2725 | sa_bulk_attr_t bulk[4]; | |
8cb34421 | 2726 | int count; |
9c5167d1 NF |
2727 | int err; |
2728 | ||
2729 | zap_cursor_init(&zc, os, dzp->z_id); | |
2730 | while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) { | |
8cb34421 | 2731 | count = 0; |
9c5167d1 NF |
2732 | if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { |
2733 | err = ENXIO; | |
2734 | break; | |
2735 | } | |
2736 | ||
2737 | err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp, | |
2738 | ZEXISTS, NULL, NULL); | |
2739 | if (err == ENOENT) | |
2740 | goto next; | |
2741 | if (err) | |
2742 | break; | |
2743 | ||
2744 | xip = ZTOI(zp); | |
2745 | if (KUID_TO_SUID(xip->i_uid) == KUID_TO_SUID(dxip->i_uid) && | |
2746 | KGID_TO_SGID(xip->i_gid) == KGID_TO_SGID(dxip->i_gid) && | |
2747 | zp->z_projid == dzp->z_projid) | |
2748 | goto next; | |
2749 | ||
2750 | tx = dmu_tx_create(os); | |
2751 | if (!(zp->z_pflags & ZFS_PROJID)) | |
2752 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
2753 | else | |
2754 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
2755 | ||
2756 | err = dmu_tx_assign(tx, TXG_WAIT); | |
2757 | if (err) | |
2758 | break; | |
2759 | ||
2760 | mutex_enter(&dzp->z_lock); | |
2761 | ||
2762 | if (KUID_TO_SUID(xip->i_uid) != KUID_TO_SUID(dxip->i_uid)) { | |
2763 | xip->i_uid = dxip->i_uid; | |
2764 | uid = zfs_uid_read(dxip); | |
2765 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, | |
2766 | &uid, sizeof (uid)); | |
2767 | } | |
2768 | ||
2769 | if (KGID_TO_SGID(xip->i_gid) != KGID_TO_SGID(dxip->i_gid)) { | |
2770 | xip->i_gid = dxip->i_gid; | |
2771 | gid = zfs_gid_read(dxip); | |
2772 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, | |
2773 | &gid, sizeof (gid)); | |
2774 | } | |
2775 | ||
2776 | if (zp->z_projid != dzp->z_projid) { | |
2777 | if (!(zp->z_pflags & ZFS_PROJID)) { | |
2778 | zp->z_pflags |= ZFS_PROJID; | |
2779 | SA_ADD_BULK_ATTR(bulk, count, | |
2780 | SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, | |
2781 | sizeof (zp->z_pflags)); | |
2782 | } | |
2783 | ||
2784 | zp->z_projid = dzp->z_projid; | |
2785 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs), | |
2786 | NULL, &zp->z_projid, sizeof (zp->z_projid)); | |
2787 | } | |
2788 | ||
2789 | mutex_exit(&dzp->z_lock); | |
2790 | ||
2791 | if (likely(count > 0)) { | |
2792 | err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
2793 | dmu_tx_commit(tx); | |
2794 | } else { | |
2795 | dmu_tx_abort(tx); | |
2796 | } | |
2797 | tx = NULL; | |
2798 | if (err != 0 && err != ENOENT) | |
2799 | break; | |
2800 | ||
2801 | next: | |
2802 | if (xip) { | |
2803 | iput(xip); | |
2804 | xip = NULL; | |
2805 | zfs_dirent_unlock(dl); | |
2806 | } | |
2807 | zap_cursor_advance(&zc); | |
2808 | } | |
2809 | ||
2810 | if (tx) | |
2811 | dmu_tx_abort(tx); | |
2812 | if (xip) { | |
2813 | iput(xip); | |
2814 | zfs_dirent_unlock(dl); | |
2815 | } | |
2816 | zap_cursor_fini(&zc); | |
2817 | ||
2818 | return (err == ENOENT ? 0 : err); | |
2819 | } | |
2820 | ||
34dc7c2f BB |
2821 | /* |
2822 | * Set the file attributes to the values contained in the | |
2823 | * vattr structure. | |
2824 | * | |
3558fd73 | 2825 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2826 | * vap - new attribute values. |
5484965a | 2827 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2828 | * flags - ATTR_UTIME set if non-default time values provided. |
2829 | * - ATTR_NOACLCHECK (CIFS context only). | |
2830 | * cr - credentials of caller. | |
34dc7c2f BB |
2831 | * |
2832 | * RETURN: 0 if success | |
2833 | * error code if failure | |
2834 | * | |
2835 | * Timestamps: | |
3558fd73 | 2836 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2837 | */ |
2838 | /* ARGSUSED */ | |
e5c39b95 | 2839 | int |
5484965a | 2840 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2841 | { |
3558fd73 | 2842 | znode_t *zp = ITOZ(ip); |
0037b49e | 2843 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
9c5167d1 | 2844 | objset_t *os = zfsvfs->z_os; |
34dc7c2f BB |
2845 | zilog_t *zilog; |
2846 | dmu_tx_t *tx; | |
2847 | vattr_t oldva; | |
f4ea75d4 | 2848 | xvattr_t *tmpxvattr; |
5484965a | 2849 | uint_t mask = vap->va_mask; |
a117a6d6 | 2850 | uint_t saved_mask = 0; |
34dc7c2f BB |
2851 | int trim_mask = 0; |
2852 | uint64_t new_mode; | |
64aefee1 | 2853 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 2854 | uint64_t xattr_obj; |
0df9673f | 2855 | uint64_t mtime[2], ctime[2], atime[2]; |
9c5167d1 | 2856 | uint64_t projid = ZFS_INVALID_PROJID; |
34dc7c2f BB |
2857 | znode_t *attrzp; |
2858 | int need_policy = FALSE; | |
9c5167d1 | 2859 | int err, err2 = 0; |
34dc7c2f | 2860 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2861 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2862 | xoptattr_t *xoap; | |
2863 | zfs_acl_t *aclp; | |
34dc7c2f | 2864 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2865 | boolean_t fuid_dirtied = B_FALSE; |
9c5167d1 | 2866 | boolean_t handle_eadir = B_FALSE; |
17c37660 | 2867 | sa_bulk_attr_t *bulk, *xattr_bulk; |
9c5167d1 | 2868 | int count = 0, xattr_count = 0, bulks = 8; |
34dc7c2f BB |
2869 | |
2870 | if (mask == 0) | |
2871 | return (0); | |
2872 | ||
0037b49e | 2873 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2874 | ZFS_VERIFY_ZP(zp); |
2875 | ||
9c5167d1 NF |
2876 | /* |
2877 | * If this is a xvattr_t, then get a pointer to the structure of | |
2878 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2879 | */ | |
2880 | xoap = xva_getxoptattr(xvap); | |
2881 | if (xoap != NULL && (mask & ATTR_XVATTR)) { | |
2882 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
2883 | if (!dmu_objset_projectquota_enabled(os) || | |
2884 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode))) { | |
2885 | ZFS_EXIT(zfsvfs); | |
2886 | return (SET_ERROR(ENOTSUP)); | |
2887 | } | |
2888 | ||
2889 | projid = xoap->xoa_projid; | |
2890 | if (unlikely(projid == ZFS_INVALID_PROJID)) { | |
2891 | ZFS_EXIT(zfsvfs); | |
2892 | return (SET_ERROR(EINVAL)); | |
2893 | } | |
2894 | ||
2895 | if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID) | |
2896 | projid = ZFS_INVALID_PROJID; | |
2897 | else | |
2898 | need_policy = TRUE; | |
2899 | } | |
2900 | ||
2901 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) && | |
2705ebf0 NF |
2902 | (xoap->xoa_projinherit != |
2903 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) && | |
9c5167d1 NF |
2904 | (!dmu_objset_projectquota_enabled(os) || |
2905 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode)))) { | |
2705ebf0 NF |
2906 | ZFS_EXIT(zfsvfs); |
2907 | return (SET_ERROR(ENOTSUP)); | |
9c5167d1 NF |
2908 | } |
2909 | } | |
2910 | ||
0037b49e | 2911 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2912 | |
2913 | /* | |
2914 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2915 | * that file system is at proper version level | |
2916 | */ | |
5484965a | 2917 | |
0037b49e | 2918 | if (zfsvfs->z_use_fuids == B_FALSE && |
5484965a BB |
2919 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2920 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2921 | (mask & ATTR_XVATTR))) { | |
0037b49e | 2922 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2923 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2924 | } |
2925 | ||
3558fd73 | 2926 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
0037b49e | 2927 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2928 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2929 | } |
2930 | ||
3558fd73 | 2931 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
0037b49e | 2932 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2933 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2934 | } |
2935 | ||
d1d7e268 | 2936 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2937 | xva_init(tmpxvattr); |
5484965a | 2938 | |
9c5167d1 NF |
2939 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); |
2940 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); | |
17c37660 | 2941 | |
5484965a BB |
2942 | /* |
2943 | * Immutable files can only alter immutable bit and atime | |
2944 | */ | |
2945 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2946 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2947 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
ecb2b7dc | 2948 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2949 | goto out3; |
5484965a BB |
2950 | } |
2951 | ||
3558fd73 | 2952 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
ecb2b7dc | 2953 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2954 | goto out3; |
34dc7c2f BB |
2955 | } |
2956 | ||
5484965a BB |
2957 | /* |
2958 | * Verify timestamps doesn't overflow 32 bits. | |
2959 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2960 | * handle times greater than 2039. This check should be removed | |
2961 | * once large timestamps are fully supported. | |
2962 | */ | |
2963 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
2964 | if (((mask & ATTR_ATIME) && |
2965 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2966 | ((mask & ATTR_MTIME) && | |
2967 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
ecb2b7dc | 2968 | err = SET_ERROR(EOVERFLOW); |
f4ea75d4 | 2969 | goto out3; |
5484965a BB |
2970 | } |
2971 | } | |
2972 | ||
34dc7c2f BB |
2973 | top: |
2974 | attrzp = NULL; | |
572e2857 | 2975 | aclp = NULL; |
34dc7c2f | 2976 | |
45d1cae3 | 2977 | /* Can this be moved to before the top label? */ |
0037b49e | 2978 | if (zfs_is_readonly(zfsvfs)) { |
ecb2b7dc | 2979 | err = SET_ERROR(EROFS); |
f4ea75d4 | 2980 | goto out3; |
34dc7c2f BB |
2981 | } |
2982 | ||
2983 | /* | |
2984 | * First validate permissions | |
2985 | */ | |
2986 | ||
3558fd73 | 2987 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2988 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2989 | if (err) |
2990 | goto out3; | |
2991 | ||
34dc7c2f BB |
2992 | /* |
2993 | * XXX - Note, we are not providing any open | |
2994 | * mode flags here (like FNDELAY), so we may | |
2995 | * block if there are locks present... this | |
2996 | * should be addressed in openat(). | |
2997 | */ | |
b128c09f | 2998 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2999 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
3000 | if (err) |
3001 | goto out3; | |
428870ff | 3002 | } |
34dc7c2f | 3003 | |
5484965a BB |
3004 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
3005 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
3006 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
3007 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
3008 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
3009 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
3010 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
3011 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
3012 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
3013 | skipaclchk, cr); | |
3014 | } | |
3015 | ||
3558fd73 BB |
3016 | if (mask & (ATTR_UID|ATTR_GID)) { |
3017 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
3018 | int take_owner; |
3019 | int take_group; | |
3020 | ||
3021 | /* | |
3022 | * NOTE: even if a new mode is being set, | |
3023 | * we may clear S_ISUID/S_ISGID bits. | |
3024 | */ | |
3025 | ||
3558fd73 | 3026 | if (!(mask & ATTR_MODE)) |
5484965a | 3027 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
3028 | |
3029 | /* | |
3030 | * Take ownership or chgrp to group we are a member of | |
3031 | */ | |
3032 | ||
5484965a | 3033 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 3034 | take_group = (mask & ATTR_GID) && |
0037b49e | 3035 | zfs_groupmember(zfsvfs, vap->va_gid, cr); |
34dc7c2f BB |
3036 | |
3037 | /* | |
5484965a | 3038 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
3039 | * take_group must both be set in order to allow taking |
3040 | * ownership. | |
3041 | * | |
3042 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
3043 | * | |
3044 | */ | |
3045 | ||
3558fd73 BB |
3046 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
3047 | take_owner && take_group) || | |
3048 | ((idmask == ATTR_UID) && take_owner) || | |
3049 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
3050 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
3051 | skipaclchk, cr) == 0) { | |
3052 | /* | |
3053 | * Remove setuid/setgid for non-privileged users | |
3054 | */ | |
5484965a | 3055 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 3056 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
3057 | } else { |
3058 | need_policy = TRUE; | |
3059 | } | |
3060 | } else { | |
3061 | need_policy = TRUE; | |
3062 | } | |
3063 | } | |
3064 | ||
3065 | mutex_enter(&zp->z_lock); | |
428870ff | 3066 | oldva.va_mode = zp->z_mode; |
572e2857 | 3067 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
3068 | if (mask & ATTR_XVATTR) { |
3069 | /* | |
3070 | * Update xvattr mask to include only those attributes | |
3071 | * that are actually changing. | |
3072 | * | |
3073 | * the bits will be restored prior to actually setting | |
3074 | * the attributes so the caller thinks they were set. | |
3075 | */ | |
3076 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
3077 | if (xoap->xoa_appendonly != | |
3078 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
3079 | need_policy = TRUE; | |
3080 | } else { | |
3081 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 3082 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
3083 | } |
3084 | } | |
3085 | ||
9c5167d1 NF |
3086 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { |
3087 | if (xoap->xoa_projinherit != | |
3088 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) { | |
3089 | need_policy = TRUE; | |
3090 | } else { | |
3091 | XVA_CLR_REQ(xvap, XAT_PROJINHERIT); | |
3092 | XVA_SET_REQ(tmpxvattr, XAT_PROJINHERIT); | |
3093 | } | |
3094 | } | |
3095 | ||
5484965a BB |
3096 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { |
3097 | if (xoap->xoa_nounlink != | |
3098 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
3099 | need_policy = TRUE; | |
3100 | } else { | |
3101 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 3102 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
3103 | } |
3104 | } | |
3105 | ||
3106 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
3107 | if (xoap->xoa_immutable != | |
3108 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
3109 | need_policy = TRUE; | |
3110 | } else { | |
3111 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 3112 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
3113 | } |
3114 | } | |
3115 | ||
3116 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
3117 | if (xoap->xoa_nodump != | |
3118 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
3119 | need_policy = TRUE; | |
3120 | } else { | |
3121 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 3122 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
3123 | } |
3124 | } | |
3125 | ||
3126 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
3127 | if (xoap->xoa_av_modified != | |
3128 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
3129 | need_policy = TRUE; | |
3130 | } else { | |
3131 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 3132 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
3133 | } |
3134 | } | |
3135 | ||
3136 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
3137 | if ((!S_ISREG(ip->i_mode) && | |
3138 | xoap->xoa_av_quarantined) || | |
3139 | xoap->xoa_av_quarantined != | |
3140 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
3141 | need_policy = TRUE; | |
3142 | } else { | |
3143 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 3144 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
3145 | } |
3146 | } | |
3147 | ||
3148 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
3149 | mutex_exit(&zp->z_lock); | |
ecb2b7dc | 3150 | err = SET_ERROR(EPERM); |
f4ea75d4 | 3151 | goto out3; |
5484965a BB |
3152 | } |
3153 | ||
3154 | if (need_policy == FALSE && | |
3155 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
3156 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
3157 | need_policy = TRUE; | |
3158 | } | |
3159 | } | |
34dc7c2f BB |
3160 | |
3161 | mutex_exit(&zp->z_lock); | |
3162 | ||
3558fd73 | 3163 | if (mask & ATTR_MODE) { |
34dc7c2f | 3164 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 3165 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 3166 | &oldva, cr); |
f4ea75d4 BB |
3167 | if (err) |
3168 | goto out3; | |
3169 | ||
3558fd73 | 3170 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
3171 | } else { |
3172 | need_policy = TRUE; | |
3173 | } | |
3174 | } | |
3175 | ||
3176 | if (need_policy) { | |
3177 | /* | |
3178 | * If trim_mask is set then take ownership | |
3179 | * has been granted or write_acl is present and user | |
3180 | * has the ability to modify mode. In that case remove | |
3181 | * UID|GID and or MODE from mask so that | |
3182 | * secpolicy_vnode_setattr() doesn't revoke it. | |
3183 | */ | |
3184 | ||
3185 | if (trim_mask) { | |
5484965a BB |
3186 | saved_mask = vap->va_mask; |
3187 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 3188 | } |
5484965a | 3189 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 3190 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
3191 | if (err) |
3192 | goto out3; | |
34dc7c2f BB |
3193 | |
3194 | if (trim_mask) | |
5484965a | 3195 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
3196 | } |
3197 | ||
3198 | /* | |
3199 | * secpolicy_vnode_setattr, or take ownership may have | |
3200 | * changed va_mask | |
3201 | */ | |
5484965a | 3202 | mask = vap->va_mask; |
34dc7c2f | 3203 | |
9c5167d1 NF |
3204 | if ((mask & (ATTR_UID | ATTR_GID)) || projid != ZFS_INVALID_PROJID) { |
3205 | handle_eadir = B_TRUE; | |
0037b49e | 3206 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
572e2857 | 3207 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 3208 | |
572e2857 | 3209 | if (err == 0 && xattr_obj) { |
3558fd73 | 3210 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
3211 | if (err) |
3212 | goto out2; | |
3213 | } | |
3558fd73 | 3214 | if (mask & ATTR_UID) { |
0037b49e | 3215 | new_kuid = zfs_fuid_create(zfsvfs, |
5484965a | 3216 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
64aefee1 | 3217 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
9c5167d1 NF |
3218 | zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT, |
3219 | new_kuid)) { | |
572e2857 | 3220 | if (attrzp) |
3558fd73 | 3221 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3222 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3223 | goto out2; |
3224 | } | |
3225 | } | |
3226 | ||
3558fd73 | 3227 | if (mask & ATTR_GID) { |
0037b49e BB |
3228 | new_kgid = zfs_fuid_create(zfsvfs, |
3229 | (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); | |
64aefee1 | 3230 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && |
9c5167d1 NF |
3231 | zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT, |
3232 | new_kgid)) { | |
572e2857 | 3233 | if (attrzp) |
3558fd73 | 3234 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3235 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3236 | goto out2; |
3237 | } | |
3238 | } | |
9c5167d1 NF |
3239 | |
3240 | if (projid != ZFS_INVALID_PROJID && | |
3241 | zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) { | |
3242 | if (attrzp) | |
3243 | iput(ZTOI(attrzp)); | |
3244 | err = EDQUOT; | |
3245 | goto out2; | |
3246 | } | |
428870ff | 3247 | } |
9c5167d1 | 3248 | tx = dmu_tx_create(os); |
34dc7c2f | 3249 | |
3558fd73 | 3250 | if (mask & ATTR_MODE) { |
428870ff | 3251 | uint64_t pmode = zp->z_mode; |
572e2857 | 3252 | uint64_t acl_obj; |
5484965a | 3253 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 3254 | |
572e2857 | 3255 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 3256 | |
572e2857 BB |
3257 | mutex_enter(&zp->z_lock); |
3258 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
3259 | /* |
3260 | * Are we upgrading ACL from old V0 format | |
3261 | * to V1 format? | |
3262 | */ | |
0037b49e | 3263 | if (zfsvfs->z_version >= ZPL_VERSION_FUID && |
572e2857 | 3264 | zfs_znode_acl_version(zp) == |
34dc7c2f | 3265 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 3266 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
3267 | DMU_OBJECT_END); |
3268 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
3269 | 0, aclp->z_acl_bytes); | |
3270 | } else { | |
572e2857 | 3271 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
3272 | aclp->z_acl_bytes); |
3273 | } | |
428870ff | 3274 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
3275 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
3276 | 0, aclp->z_acl_bytes); | |
3277 | } | |
572e2857 | 3278 | mutex_exit(&zp->z_lock); |
428870ff BB |
3279 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3280 | } else { | |
9c5167d1 NF |
3281 | if (((mask & ATTR_XVATTR) && |
3282 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) || | |
3283 | (projid != ZFS_INVALID_PROJID && | |
3284 | !(zp->z_pflags & ZFS_PROJID))) | |
5484965a BB |
3285 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3286 | else | |
3287 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3288 | } |
3289 | ||
428870ff BB |
3290 | if (attrzp) { |
3291 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3292 | } |
3293 | ||
0037b49e | 3294 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
428870ff | 3295 | if (fuid_dirtied) |
0037b49e | 3296 | zfs_fuid_txhold(zfsvfs, tx); |
428870ff BB |
3297 | |
3298 | zfs_sa_upgrade_txholds(tx, zp); | |
3299 | ||
384f8a09 MA |
3300 | err = dmu_tx_assign(tx, TXG_WAIT); |
3301 | if (err) | |
9babb374 | 3302 | goto out; |
34dc7c2f | 3303 | |
428870ff | 3304 | count = 0; |
34dc7c2f BB |
3305 | /* |
3306 | * Set each attribute requested. | |
3307 | * We group settings according to the locks they need to acquire. | |
3308 | * | |
3309 | * Note: you cannot set ctime directly, although it will be | |
3310 | * updated as a side-effect of calling this function. | |
3311 | */ | |
3312 | ||
9c5167d1 NF |
3313 | if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) { |
3314 | /* | |
3315 | * For the existed object that is upgraded from old system, | |
3316 | * its on-disk layout has no slot for the project ID attribute. | |
3317 | * But quota accounting logic needs to access related slots by | |
3318 | * offset directly. So we need to adjust old objects' layout | |
3319 | * to make the project ID to some unified and fixed offset. | |
3320 | */ | |
3321 | if (attrzp) | |
3322 | err = sa_add_projid(attrzp->z_sa_hdl, tx, projid); | |
3323 | if (err == 0) | |
3324 | err = sa_add_projid(zp->z_sa_hdl, tx, projid); | |
3325 | ||
3326 | if (unlikely(err == EEXIST)) | |
3327 | err = 0; | |
3328 | else if (err != 0) | |
3329 | goto out; | |
3330 | else | |
3331 | projid = ZFS_INVALID_PROJID; | |
3332 | } | |
572e2857 | 3333 | |
3558fd73 | 3334 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3335 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
3336 | mutex_enter(&zp->z_lock); |
3337 | ||
0037b49e | 3338 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
3339 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3340 | ||
3341 | if (attrzp) { | |
3558fd73 | 3342 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3343 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
3344 | mutex_enter(&attrzp->z_lock); |
3345 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3346 | SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, |
428870ff | 3347 | sizeof (attrzp->z_pflags)); |
9c5167d1 NF |
3348 | if (projid != ZFS_INVALID_PROJID) { |
3349 | attrzp->z_projid = projid; | |
3350 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3351 | SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid, | |
3352 | sizeof (attrzp->z_projid)); | |
3353 | } | |
428870ff BB |
3354 | } |
3355 | ||
3558fd73 | 3356 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 3357 | |
3558fd73 | 3358 | if (mask & ATTR_UID) { |
64aefee1 NB |
3359 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
3360 | new_uid = zfs_uid_read(ZTOI(zp)); | |
0037b49e | 3361 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 3362 | &new_uid, sizeof (new_uid)); |
428870ff BB |
3363 | if (attrzp) { |
3364 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3365 | SA_ZPL_UID(zfsvfs), NULL, &new_uid, |
428870ff | 3366 | sizeof (new_uid)); |
2c6abf15 | 3367 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
3368 | } |
3369 | } | |
3370 | ||
3558fd73 | 3371 | if (mask & ATTR_GID) { |
64aefee1 NB |
3372 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
3373 | new_gid = zfs_gid_read(ZTOI(zp)); | |
0037b49e | 3374 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), |
428870ff | 3375 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
3376 | if (attrzp) { |
3377 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3378 | SA_ZPL_GID(zfsvfs), NULL, &new_gid, |
428870ff | 3379 | sizeof (new_gid)); |
64aefee1 | 3380 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
3381 | } |
3382 | } | |
3558fd73 | 3383 | if (!(mask & ATTR_MODE)) { |
0037b49e | 3384 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), |
428870ff BB |
3385 | NULL, &new_mode, sizeof (new_mode)); |
3386 | new_mode = zp->z_mode; | |
3387 | } | |
3388 | err = zfs_acl_chown_setattr(zp); | |
3389 | ASSERT(err == 0); | |
3390 | if (attrzp) { | |
3391 | err = zfs_acl_chown_setattr(attrzp); | |
3392 | ASSERT(err == 0); | |
3393 | } | |
3394 | } | |
3395 | ||
3558fd73 | 3396 | if (mask & ATTR_MODE) { |
0037b49e | 3397 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 3398 | &new_mode, sizeof (new_mode)); |
12fa7f34 | 3399 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 3400 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 3401 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 3402 | ASSERT0(err); |
572e2857 BB |
3403 | if (zp->z_acl_cached) |
3404 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
3405 | zp->z_acl_cached = aclp; |
3406 | aclp = NULL; | |
34dc7c2f BB |
3407 | } |
3408 | ||
704cd075 | 3409 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
a43570c5 | 3410 | zp->z_atime_dirty = B_FALSE; |
704cd075 | 3411 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
0037b49e | 3412 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
0df9673f | 3413 | &atime, sizeof (atime)); |
34dc7c2f BB |
3414 | } |
3415 | ||
99834d19 | 3416 | if (mask & (ATTR_MTIME | ATTR_SIZE)) { |
5484965a | 3417 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
6413c95f | 3418 | ZTOI(zp)->i_mtime = zpl_inode_timespec_trunc(vap->va_mtime, |
87f9371a NB |
3419 | ZTOI(zp)->i_sb->s_time_gran); |
3420 | ||
0037b49e | 3421 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 3422 | mtime, sizeof (mtime)); |
34dc7c2f BB |
3423 | } |
3424 | ||
99834d19 | 3425 | if (mask & (ATTR_CTIME | ATTR_SIZE)) { |
87f9371a | 3426 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); |
6413c95f | 3427 | ZTOI(zp)->i_ctime = zpl_inode_timespec_trunc(vap->va_ctime, |
87f9371a | 3428 | ZTOI(zp)->i_sb->s_time_gran); |
0037b49e | 3429 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
87f9371a | 3430 | ctime, sizeof (ctime)); |
428870ff | 3431 | } |
87f9371a | 3432 | |
9c5167d1 NF |
3433 | if (projid != ZFS_INVALID_PROJID) { |
3434 | zp->z_projid = projid; | |
3435 | SA_ADD_BULK_ATTR(bulk, count, | |
3436 | SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid, | |
3437 | sizeof (zp->z_projid)); | |
3438 | } | |
3439 | ||
87f9371a NB |
3440 | if (attrzp && mask) { |
3441 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3442 | SA_ZPL_CTIME(zfsvfs), NULL, &ctime, |
87f9371a NB |
3443 | sizeof (ctime)); |
3444 | } | |
3445 | ||
34dc7c2f BB |
3446 | /* |
3447 | * Do this after setting timestamps to prevent timestamp | |
3448 | * update from toggling bit | |
3449 | */ | |
3450 | ||
5484965a BB |
3451 | if (xoap && (mask & ATTR_XVATTR)) { |
3452 | ||
3453 | /* | |
3454 | * restore trimmed off masks | |
3455 | * so that return masks can be set for caller. | |
3456 | */ | |
3457 | ||
f4ea75d4 | 3458 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
3459 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
3460 | } | |
f4ea75d4 | 3461 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
3462 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
3463 | } | |
f4ea75d4 | 3464 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
3465 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
3466 | } | |
f4ea75d4 | 3467 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
3468 | XVA_SET_REQ(xvap, XAT_NODUMP); |
3469 | } | |
f4ea75d4 | 3470 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
3471 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
3472 | } | |
f4ea75d4 | 3473 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
3474 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
3475 | } | |
9c5167d1 NF |
3476 | if (XVA_ISSET_REQ(tmpxvattr, XAT_PROJINHERIT)) { |
3477 | XVA_SET_REQ(xvap, XAT_PROJINHERIT); | |
3478 | } | |
5484965a BB |
3479 | |
3480 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3481 | ASSERT(S_ISREG(ip->i_mode)); | |
3482 | ||
3483 | zfs_xvattr_set(zp, xvap, tx); | |
3484 | } | |
3485 | ||
9babb374 | 3486 | if (fuid_dirtied) |
0037b49e | 3487 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 3488 | |
34dc7c2f | 3489 | if (mask != 0) |
5484965a | 3490 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3491 | |
34dc7c2f | 3492 | mutex_exit(&zp->z_lock); |
3558fd73 | 3493 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3494 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3495 | |
572e2857 | 3496 | if (attrzp) { |
3558fd73 | 3497 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3498 | mutex_exit(&attrzp->z_acl_lock); |
3499 | mutex_exit(&attrzp->z_lock); | |
3500 | } | |
9babb374 | 3501 | out: |
9c5167d1 | 3502 | if (err == 0 && xattr_count > 0) { |
428870ff BB |
3503 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, |
3504 | xattr_count, tx); | |
3505 | ASSERT(err2 == 0); | |
3506 | } | |
3507 | ||
45d1cae3 | 3508 | if (aclp) |
9babb374 | 3509 | zfs_acl_free(aclp); |
9babb374 BB |
3510 | |
3511 | if (fuidp) { | |
3512 | zfs_fuid_info_free(fuidp); | |
3513 | fuidp = NULL; | |
3514 | } | |
3515 | ||
428870ff | 3516 | if (err) { |
9babb374 | 3517 | dmu_tx_abort(tx); |
ea7e86d8 BB |
3518 | if (attrzp) |
3519 | iput(ZTOI(attrzp)); | |
428870ff BB |
3520 | if (err == ERESTART) |
3521 | goto top; | |
3522 | } else { | |
9c5167d1 NF |
3523 | if (count > 0) |
3524 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3525 | dmu_tx_commit(tx); |
9c5167d1 NF |
3526 | if (attrzp) { |
3527 | if (err2 == 0 && handle_eadir) | |
3528 | err2 = zfs_setattr_dir(attrzp); | |
ea7e86d8 | 3529 | iput(ZTOI(attrzp)); |
9c5167d1 | 3530 | } |
037849f8 | 3531 | zfs_inode_update(zp); |
428870ff BB |
3532 | } |
3533 | ||
428870ff | 3534 | out2: |
9c5167d1 | 3535 | if (os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3536 | zil_commit(zilog, 0); |
34dc7c2f | 3537 | |
f4ea75d4 | 3538 | out3: |
9c5167d1 NF |
3539 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks); |
3540 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * bulks); | |
d1d7e268 | 3541 | kmem_free(tmpxvattr, sizeof (xvattr_t)); |
0037b49e | 3542 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3543 | return (err); |
3544 | } | |
3545 | ||
3546 | typedef struct zfs_zlock { | |
3547 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3548 | znode_t *zl_znode; /* znode we held */ | |
3549 | struct zfs_zlock *zl_next; /* next in list */ | |
3550 | } zfs_zlock_t; | |
3551 | ||
3552 | /* | |
3553 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3554 | */ | |
3555 | static void | |
3556 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3557 | { | |
3558 | zfs_zlock_t *zl; | |
3559 | ||
3560 | while ((zl = *zlpp) != NULL) { | |
3561 | if (zl->zl_znode != NULL) | |
ea7e86d8 | 3562 | zfs_iput_async(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3563 | rw_exit(zl->zl_rwlock); |
3564 | *zlpp = zl->zl_next; | |
3565 | kmem_free(zl, sizeof (*zl)); | |
3566 | } | |
3567 | } | |
3568 | ||
3569 | /* | |
3570 | * Search back through the directory tree, using the ".." entries. | |
3571 | * Lock each directory in the chain to prevent concurrent renames. | |
3572 | * Fail any attempt to move a directory into one of its own descendants. | |
3573 | * XXX - z_parent_lock can overlap with map or grow locks | |
3574 | */ | |
3575 | static int | |
3576 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3577 | { | |
3578 | zfs_zlock_t *zl; | |
3579 | znode_t *zp = tdzp; | |
3558fd73 | 3580 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3581 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3582 | krwlock_t *rwlp = &szp->z_parent_lock; |
3583 | krw_t rw = RW_WRITER; | |
3584 | ||
3585 | /* | |
3586 | * First pass write-locks szp and compares to zp->z_id. | |
3587 | * Later passes read-lock zp and compare to zp->z_parent. | |
3588 | */ | |
3589 | do { | |
3590 | if (!rw_tryenter(rwlp, rw)) { | |
3591 | /* | |
3592 | * Another thread is renaming in this path. | |
3593 | * Note that if we are a WRITER, we don't have any | |
3594 | * parent_locks held yet. | |
3595 | */ | |
3596 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3597 | /* | |
3598 | * Drop our locks and restart | |
3599 | */ | |
3600 | zfs_rename_unlock(&zl); | |
3601 | *zlpp = NULL; | |
3602 | zp = tdzp; | |
428870ff | 3603 | oidp = zp->z_id; |
34dc7c2f BB |
3604 | rwlp = &szp->z_parent_lock; |
3605 | rw = RW_WRITER; | |
3606 | continue; | |
3607 | } else { | |
3608 | /* | |
3609 | * Wait for other thread to drop its locks | |
3610 | */ | |
3611 | rw_enter(rwlp, rw); | |
3612 | } | |
3613 | } | |
3614 | ||
3615 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3616 | zl->zl_rwlock = rwlp; | |
3617 | zl->zl_znode = NULL; | |
3618 | zl->zl_next = *zlpp; | |
3619 | *zlpp = zl; | |
3620 | ||
428870ff | 3621 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 3622 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3623 | |
428870ff | 3624 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3625 | return (0); |
3626 | ||
3627 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3628 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3629 | if (error) |
3630 | return (error); | |
3631 | zl->zl_znode = zp; | |
3632 | } | |
3558fd73 | 3633 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3634 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3635 | rwlp = &zp->z_parent_lock; |
3636 | rw = RW_READER; | |
3637 | ||
3638 | } while (zp->z_id != sdzp->z_id); | |
3639 | ||
3640 | return (0); | |
3641 | } | |
3642 | ||
3643 | /* | |
3644 | * Move an entry from the provided source directory to the target | |
3645 | * directory. Change the entry name as indicated. | |
3646 | * | |
3558fd73 | 3647 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3648 | * snm - Old entry name. |
3558fd73 | 3649 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3650 | * tnm - New entry name. |
3651 | * cr - credentials of caller. | |
34dc7c2f BB |
3652 | * flags - case flags |
3653 | * | |
d3cc8b15 | 3654 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3655 | * |
3656 | * Timestamps: | |
3558fd73 | 3657 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3658 | */ |
3659 | /*ARGSUSED*/ | |
e5c39b95 | 3660 | int |
3558fd73 BB |
3661 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3662 | cred_t *cr, int flags) | |
34dc7c2f BB |
3663 | { |
3664 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 | 3665 | znode_t *sdzp = ITOZ(sdip); |
0037b49e | 3666 | zfsvfs_t *zfsvfs = ITOZSB(sdip); |
34dc7c2f | 3667 | zilog_t *zilog; |
34dc7c2f BB |
3668 | zfs_dirlock_t *sdl, *tdl; |
3669 | dmu_tx_t *tx; | |
3670 | zfs_zlock_t *zl; | |
3671 | int cmp, serr, terr; | |
3672 | int error = 0; | |
3673 | int zflg = 0; | |
e8b96c60 | 3674 | boolean_t waited = B_FALSE; |
34dc7c2f | 3675 | |
32dec7bd | 3676 | if (snm == NULL || tnm == NULL) |
3677 | return (SET_ERROR(EINVAL)); | |
3678 | ||
0037b49e | 3679 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 3680 | ZFS_VERIFY_ZP(sdzp); |
0037b49e | 3681 | zilog = zfsvfs->z_log; |
34dc7c2f | 3682 | |
812e91a7 MT |
3683 | tdzp = ITOZ(tdip); |
3684 | ZFS_VERIFY_ZP(tdzp); | |
3685 | ||
3686 | /* | |
3687 | * We check i_sb because snapshots and the ctldir must have different | |
3688 | * super blocks. | |
3689 | */ | |
c0ebc844 | 3690 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
0037b49e | 3691 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3692 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3693 | } |
3694 | ||
0037b49e | 3695 | if (zfsvfs->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3696 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 3697 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3698 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3699 | } |
3700 | ||
3701 | if (flags & FIGNORECASE) | |
3702 | zflg |= ZCILOOK; | |
3703 | ||
3704 | top: | |
3705 | szp = NULL; | |
3706 | tzp = NULL; | |
3707 | zl = NULL; | |
3708 | ||
3709 | /* | |
3710 | * This is to prevent the creation of links into attribute space | |
3711 | * by renaming a linked file into/outof an attribute directory. | |
3712 | * See the comment in zfs_link() for why this is considered bad. | |
3713 | */ | |
428870ff | 3714 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 3715 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3716 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3717 | } |
3718 | ||
3719 | /* | |
3720 | * Lock source and target directory entries. To prevent deadlock, | |
3721 | * a lock ordering must be defined. We lock the directory with | |
3722 | * the smallest object id first, or if it's a tie, the one with | |
3723 | * the lexically first name. | |
3724 | */ | |
3725 | if (sdzp->z_id < tdzp->z_id) { | |
3726 | cmp = -1; | |
3727 | } else if (sdzp->z_id > tdzp->z_id) { | |
3728 | cmp = 1; | |
3729 | } else { | |
3730 | /* | |
3731 | * First compare the two name arguments without | |
3732 | * considering any case folding. | |
3733 | */ | |
0037b49e | 3734 | int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3735 | |
3736 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
0037b49e | 3737 | ASSERT(error == 0 || !zfsvfs->z_utf8); |
34dc7c2f BB |
3738 | if (cmp == 0) { |
3739 | /* | |
3740 | * POSIX: "If the old argument and the new argument | |
3741 | * both refer to links to the same existing file, | |
3742 | * the rename() function shall return successfully | |
3743 | * and perform no other action." | |
3744 | */ | |
0037b49e | 3745 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3746 | return (0); |
3747 | } | |
3748 | /* | |
3749 | * If the file system is case-folding, then we may | |
3750 | * have some more checking to do. A case-folding file | |
3751 | * system is either supporting mixed case sensitivity | |
3752 | * access or is completely case-insensitive. Note | |
3753 | * that the file system is always case preserving. | |
3754 | * | |
3755 | * In mixed sensitivity mode case sensitive behavior | |
3756 | * is the default. FIGNORECASE must be used to | |
3757 | * explicitly request case insensitive behavior. | |
3758 | * | |
3759 | * If the source and target names provided differ only | |
3760 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3761 | * we will treat this as a special case in the | |
3762 | * case-insensitive mode: as long as the source name | |
3763 | * is an exact match, we will allow this to proceed as | |
3764 | * a name-change request. | |
3765 | */ | |
0037b49e BB |
3766 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
3767 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3768 | flags & FIGNORECASE)) && |
0037b49e | 3769 | u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3770 | &error) == 0) { |
3771 | /* | |
3772 | * case preserving rename request, require exact | |
3773 | * name matches | |
3774 | */ | |
3775 | zflg |= ZCIEXACT; | |
3776 | zflg &= ~ZCILOOK; | |
3777 | } | |
3778 | } | |
3779 | ||
428870ff BB |
3780 | /* |
3781 | * If the source and destination directories are the same, we should | |
3782 | * grab the z_name_lock of that directory only once. | |
3783 | */ | |
3784 | if (sdzp == tdzp) { | |
3785 | zflg |= ZHAVELOCK; | |
3786 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3787 | } | |
3788 | ||
34dc7c2f BB |
3789 | if (cmp < 0) { |
3790 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3791 | ZEXISTS | zflg, NULL, NULL); | |
3792 | terr = zfs_dirent_lock(&tdl, | |
3793 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3794 | } else { | |
3795 | terr = zfs_dirent_lock(&tdl, | |
3796 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3797 | serr = zfs_dirent_lock(&sdl, | |
3798 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3799 | NULL, NULL); | |
3800 | } | |
3801 | ||
3802 | if (serr) { | |
3803 | /* | |
3804 | * Source entry invalid or not there. | |
3805 | */ | |
3806 | if (!terr) { | |
3807 | zfs_dirent_unlock(tdl); | |
3808 | if (tzp) | |
3558fd73 | 3809 | iput(ZTOI(tzp)); |
34dc7c2f | 3810 | } |
428870ff BB |
3811 | |
3812 | if (sdzp == tdzp) | |
3813 | rw_exit(&sdzp->z_name_lock); | |
3814 | ||
34dc7c2f BB |
3815 | if (strcmp(snm, "..") == 0) |
3816 | serr = EINVAL; | |
0037b49e | 3817 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3818 | return (serr); |
3819 | } | |
3820 | if (terr) { | |
3821 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3822 | iput(ZTOI(szp)); |
428870ff BB |
3823 | |
3824 | if (sdzp == tdzp) | |
3825 | rw_exit(&sdzp->z_name_lock); | |
3826 | ||
34dc7c2f BB |
3827 | if (strcmp(tnm, "..") == 0) |
3828 | terr = EINVAL; | |
0037b49e | 3829 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3830 | return (terr); |
3831 | } | |
3832 | ||
9c5167d1 NF |
3833 | /* |
3834 | * If we are using project inheritance, means if the directory has | |
3835 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
3836 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
3837 | * such case, we only allow renames into our tree when the project | |
3838 | * IDs are the same. | |
3839 | */ | |
3840 | if (tdzp->z_pflags & ZFS_PROJINHERIT && | |
3841 | tdzp->z_projid != szp->z_projid) { | |
3842 | error = SET_ERROR(EXDEV); | |
3843 | goto out; | |
3844 | } | |
3845 | ||
34dc7c2f BB |
3846 | /* |
3847 | * Must have write access at the source to remove the old entry | |
3848 | * and write access at the target to create the new entry. | |
3849 | * Note that if target and source are the same, this can be | |
3850 | * done in a single check. | |
3851 | */ | |
3852 | ||
149e873a | 3853 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3854 | goto out; |
3855 | ||
3558fd73 | 3856 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3857 | /* |
3858 | * Check to make sure rename is valid. | |
3859 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3860 | */ | |
149e873a | 3861 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3862 | goto out; |
3863 | } | |
3864 | ||
3865 | /* | |
3866 | * Does target exist? | |
3867 | */ | |
3868 | if (tzp) { | |
3869 | /* | |
3870 | * Source and target must be the same type. | |
3871 | */ | |
3558fd73 BB |
3872 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3873 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
2e528b49 | 3874 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3875 | goto out; |
3876 | } | |
3877 | } else { | |
3558fd73 | 3878 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
2e528b49 | 3879 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3880 | goto out; |
3881 | } | |
3882 | } | |
3883 | /* | |
3884 | * POSIX dictates that when the source and target | |
3885 | * entries refer to the same file object, rename | |
3886 | * must do nothing and exit without error. | |
3887 | */ | |
3888 | if (szp->z_id == tzp->z_id) { | |
3889 | error = 0; | |
3890 | goto out; | |
3891 | } | |
3892 | } | |
3893 | ||
0037b49e | 3894 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
3895 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3896 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3897 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3898 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3899 | if (sdzp != tdzp) { |
3900 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3901 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3902 | } | |
3903 | if (tzp) { | |
3904 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3905 | zfs_sa_upgrade_txholds(tx, tzp); | |
3906 | } | |
3907 | ||
3908 | zfs_sa_upgrade_txholds(tx, szp); | |
0037b49e | 3909 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
0735ecb3 | 3910 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
3911 | if (error) { |
3912 | if (zl != NULL) | |
3913 | zfs_rename_unlock(&zl); | |
3914 | zfs_dirent_unlock(sdl); | |
3915 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3916 | |
3917 | if (sdzp == tdzp) | |
3918 | rw_exit(&sdzp->z_name_lock); | |
3919 | ||
fb5f0bc8 | 3920 | if (error == ERESTART) { |
e8b96c60 | 3921 | waited = B_TRUE; |
34dc7c2f BB |
3922 | dmu_tx_wait(tx); |
3923 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3924 | iput(ZTOI(szp)); |
3925 | if (tzp) | |
3926 | iput(ZTOI(tzp)); | |
34dc7c2f BB |
3927 | goto top; |
3928 | } | |
3929 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3930 | iput(ZTOI(szp)); |
3931 | if (tzp) | |
3932 | iput(ZTOI(tzp)); | |
0037b49e | 3933 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3934 | return (error); |
3935 | } | |
3936 | ||
3937 | if (tzp) /* Attempt to remove the existing target */ | |
3938 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3939 | ||
3940 | if (error == 0) { | |
3941 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3942 | if (error == 0) { | |
428870ff | 3943 | szp->z_pflags |= ZFS_AV_MODIFIED; |
9c5167d1 NF |
3944 | if (tdzp->z_pflags & ZFS_PROJINHERIT) |
3945 | szp->z_pflags |= ZFS_PROJINHERIT; | |
34dc7c2f | 3946 | |
0037b49e | 3947 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 3948 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c99c9001 | 3949 | ASSERT0(error); |
34dc7c2f | 3950 | |
428870ff BB |
3951 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3952 | if (error == 0) { | |
3953 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3954 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3955 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3956 | } else { |
3957 | /* | |
3958 | * At this point, we have successfully created | |
3959 | * the target name, but have failed to remove | |
3960 | * the source name. Since the create was done | |
3961 | * with the ZRENAMING flag, there are | |
3962 | * complications; for one, the link count is | |
3963 | * wrong. The easiest way to deal with this | |
3964 | * is to remove the newly created target, and | |
3965 | * return the original error. This must | |
3966 | * succeed; fortunately, it is very unlikely to | |
3967 | * fail, since we just created it. | |
3968 | */ | |
3969 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3970 | ZRENAMING, NULL), ==, 0); | |
3971 | } | |
599b8648 CC |
3972 | } else { |
3973 | /* | |
3974 | * If we had removed the existing target, subsequent | |
3975 | * call to zfs_link_create() to add back the same entry | |
3976 | * but, the new dnode (szp) should not fail. | |
3977 | */ | |
3978 | ASSERT(tzp == NULL); | |
34dc7c2f BB |
3979 | } |
3980 | } | |
3981 | ||
3982 | dmu_tx_commit(tx); | |
3983 | out: | |
3984 | if (zl != NULL) | |
3985 | zfs_rename_unlock(&zl); | |
3986 | ||
3987 | zfs_dirent_unlock(sdl); | |
3988 | zfs_dirent_unlock(tdl); | |
3989 | ||
960e08fe | 3990 | zfs_inode_update(sdzp); |
428870ff BB |
3991 | if (sdzp == tdzp) |
3992 | rw_exit(&sdzp->z_name_lock); | |
3993 | ||
960e08fe BB |
3994 | if (sdzp != tdzp) |
3995 | zfs_inode_update(tdzp); | |
428870ff | 3996 | |
960e08fe | 3997 | zfs_inode_update(szp); |
3558fd73 | 3998 | iput(ZTOI(szp)); |
960e08fe BB |
3999 | if (tzp) { |
4000 | zfs_inode_update(tzp); | |
3558fd73 | 4001 | iput(ZTOI(tzp)); |
960e08fe | 4002 | } |
34dc7c2f | 4003 | |
0037b49e | 4004 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4005 | zil_commit(zilog, 0); |
428870ff | 4006 | |
0037b49e | 4007 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4008 | return (error); |
4009 | } | |
4010 | ||
4011 | /* | |
4012 | * Insert the indicated symbolic reference entry into the directory. | |
4013 | * | |
3558fd73 | 4014 | * IN: dip - Directory to contain new symbolic link. |
841a7a98 | 4015 | * name - Name of directory entry in dip. |
34dc7c2f | 4016 | * vap - Attributes of new entry. |
841a7a98 | 4017 | * link - Name for new symlink entry. |
34dc7c2f | 4018 | * cr - credentials of caller. |
34dc7c2f BB |
4019 | * flags - case flags |
4020 | * | |
841a7a98 TK |
4021 | * OUT: ipp - Inode for new symbolic link. |
4022 | * | |
d3cc8b15 | 4023 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4024 | * |
4025 | * Timestamps: | |
3558fd73 | 4026 | * dip - ctime|mtime updated |
34dc7c2f BB |
4027 | */ |
4028 | /*ARGSUSED*/ | |
e5c39b95 | 4029 | int |
3558fd73 BB |
4030 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
4031 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 4032 | { |
3558fd73 | 4033 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
4034 | zfs_dirlock_t *dl; |
4035 | dmu_tx_t *tx; | |
0037b49e | 4036 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 4037 | zilog_t *zilog; |
428870ff | 4038 | uint64_t len = strlen(link); |
34dc7c2f BB |
4039 | int error; |
4040 | int zflg = ZNEW; | |
9babb374 BB |
4041 | zfs_acl_ids_t acl_ids; |
4042 | boolean_t fuid_dirtied; | |
428870ff | 4043 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 4044 | boolean_t waited = B_FALSE; |
34dc7c2f | 4045 | |
3558fd73 | 4046 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 4047 | |
32dec7bd | 4048 | if (name == NULL) |
4049 | return (SET_ERROR(EINVAL)); | |
4050 | ||
0037b49e | 4051 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4052 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 4053 | zilog = zfsvfs->z_log; |
34dc7c2f | 4054 | |
0037b49e | 4055 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 4056 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 4057 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4058 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
4059 | } |
4060 | if (flags & FIGNORECASE) | |
4061 | zflg |= ZCILOOK; | |
34dc7c2f BB |
4062 | |
4063 | if (len > MAXPATHLEN) { | |
0037b49e | 4064 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4065 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
4066 | } |
4067 | ||
428870ff BB |
4068 | if ((error = zfs_acl_ids_create(dzp, 0, |
4069 | vap, cr, NULL, &acl_ids)) != 0) { | |
0037b49e | 4070 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4071 | return (error); |
4072 | } | |
4073 | top: | |
3558fd73 BB |
4074 | *ipp = NULL; |
4075 | ||
34dc7c2f BB |
4076 | /* |
4077 | * Attempt to lock directory; fail if entry already exists. | |
4078 | */ | |
4079 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
4080 | if (error) { | |
428870ff | 4081 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 4082 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4083 | return (error); |
4084 | } | |
4085 | ||
149e873a | 4086 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
4087 | zfs_acl_ids_free(&acl_ids); |
4088 | zfs_dirent_unlock(dl); | |
0037b49e | 4089 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4090 | return (error); |
4091 | } | |
4092 | ||
9c5167d1 | 4093 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, ZFS_DEFAULT_PROJID)) { |
9babb374 BB |
4094 | zfs_acl_ids_free(&acl_ids); |
4095 | zfs_dirent_unlock(dl); | |
0037b49e | 4096 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4097 | return (SET_ERROR(EDQUOT)); |
9babb374 | 4098 | } |
0037b49e BB |
4099 | tx = dmu_tx_create(zfsvfs->z_os); |
4100 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
34dc7c2f | 4101 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 4102 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
4103 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
4104 | ZFS_SA_BASE_ATTR_SIZE + len); | |
4105 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
0037b49e | 4106 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
4107 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
4108 | acl_ids.z_aclp->z_acl_bytes); | |
4109 | } | |
9babb374 | 4110 | if (fuid_dirtied) |
0037b49e | 4111 | zfs_fuid_txhold(zfsvfs, tx); |
0735ecb3 | 4112 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
4113 | if (error) { |
4114 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4115 | if (error == ERESTART) { |
e8b96c60 | 4116 | waited = B_TRUE; |
34dc7c2f BB |
4117 | dmu_tx_wait(tx); |
4118 | dmu_tx_abort(tx); | |
4119 | goto top; | |
4120 | } | |
428870ff | 4121 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 4122 | dmu_tx_abort(tx); |
0037b49e | 4123 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4124 | return (error); |
4125 | } | |
4126 | ||
34dc7c2f BB |
4127 | /* |
4128 | * Create a new object for the symlink. | |
428870ff | 4129 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 4130 | */ |
428870ff | 4131 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 4132 | |
428870ff | 4133 | if (fuid_dirtied) |
0037b49e | 4134 | zfs_fuid_sync(zfsvfs, tx); |
34dc7c2f | 4135 | |
572e2857 | 4136 | mutex_enter(&zp->z_lock); |
428870ff | 4137 | if (zp->z_is_sa) |
0037b49e | 4138 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), |
428870ff BB |
4139 | link, len, tx); |
4140 | else | |
4141 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 4142 | mutex_exit(&zp->z_lock); |
34dc7c2f | 4143 | |
428870ff | 4144 | zp->z_size = len; |
0037b49e | 4145 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 4146 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
4147 | /* |
4148 | * Insert the new object into the directory. | |
4149 | */ | |
599b8648 CC |
4150 | error = zfs_link_create(dl, zp, tx, ZNEW); |
4151 | if (error != 0) { | |
4152 | zfs_znode_delete(zp, tx); | |
4153 | remove_inode_hash(ZTOI(zp)); | |
4154 | } else { | |
4155 | if (flags & FIGNORECASE) | |
4156 | txtype |= TX_CI; | |
4157 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
4f301661 | 4158 | |
599b8648 CC |
4159 | zfs_inode_update(dzp); |
4160 | zfs_inode_update(zp); | |
4161 | } | |
960e08fe | 4162 | |
9babb374 | 4163 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
4164 | |
4165 | dmu_tx_commit(tx); | |
4166 | ||
4167 | zfs_dirent_unlock(dl); | |
4168 | ||
599b8648 CC |
4169 | if (error == 0) { |
4170 | *ipp = ZTOI(zp); | |
34dc7c2f | 4171 | |
599b8648 CC |
4172 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
4173 | zil_commit(zilog, 0); | |
4174 | } else { | |
4175 | iput(ZTOI(zp)); | |
4176 | } | |
428870ff | 4177 | |
0037b49e | 4178 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4179 | return (error); |
4180 | } | |
4181 | ||
4182 | /* | |
4183 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 4184 | * the symbolic path referred to by ip. |
34dc7c2f | 4185 | * |
8b4f9a2d BB |
4186 | * IN: ip - inode of symbolic link |
4187 | * uio - structure to contain the link path. | |
4188 | * cr - credentials of caller. | |
34dc7c2f BB |
4189 | * |
4190 | * RETURN: 0 if success | |
4191 | * error code if failure | |
4192 | * | |
4193 | * Timestamps: | |
3558fd73 | 4194 | * ip - atime updated |
34dc7c2f BB |
4195 | */ |
4196 | /* ARGSUSED */ | |
e5c39b95 | 4197 | int |
8b4f9a2d | 4198 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 4199 | { |
3558fd73 | 4200 | znode_t *zp = ITOZ(ip); |
0037b49e | 4201 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4202 | int error; |
4203 | ||
0037b49e | 4204 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4205 | ZFS_VERIFY_ZP(zp); |
4206 | ||
572e2857 | 4207 | mutex_enter(&zp->z_lock); |
428870ff | 4208 | if (zp->z_is_sa) |
8b4f9a2d | 4209 | error = sa_lookup_uio(zp->z_sa_hdl, |
0037b49e | 4210 | SA_ZPL_SYMLINK(zfsvfs), uio); |
428870ff | 4211 | else |
8b4f9a2d | 4212 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 4213 | mutex_exit(&zp->z_lock); |
34dc7c2f | 4214 | |
0037b49e | 4215 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4216 | return (error); |
4217 | } | |
4218 | ||
4219 | /* | |
3558fd73 | 4220 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 4221 | * |
3558fd73 BB |
4222 | * IN: tdip - Directory to contain new entry. |
4223 | * sip - inode of new entry. | |
34dc7c2f BB |
4224 | * name - name of new entry. |
4225 | * cr - credentials of caller. | |
841a7a98 | 4226 | * flags - case flags. |
34dc7c2f BB |
4227 | * |
4228 | * RETURN: 0 if success | |
4229 | * error code if failure | |
4230 | * | |
4231 | * Timestamps: | |
3558fd73 BB |
4232 | * tdip - ctime|mtime updated |
4233 | * sip - ctime updated | |
34dc7c2f BB |
4234 | */ |
4235 | /* ARGSUSED */ | |
e5c39b95 | 4236 | int |
da5e151f BB |
4237 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr, |
4238 | int flags) | |
34dc7c2f | 4239 | { |
3558fd73 | 4240 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 4241 | znode_t *tzp, *szp; |
0037b49e | 4242 | zfsvfs_t *zfsvfs = ITOZSB(tdip); |
34dc7c2f BB |
4243 | zilog_t *zilog; |
4244 | zfs_dirlock_t *dl; | |
4245 | dmu_tx_t *tx; | |
34dc7c2f BB |
4246 | int error; |
4247 | int zf = ZNEW; | |
428870ff | 4248 | uint64_t parent; |
572e2857 | 4249 | uid_t owner; |
e8b96c60 | 4250 | boolean_t waited = B_FALSE; |
ace1eae8 CC |
4251 | boolean_t is_tmpfile = 0; |
4252 | uint64_t txg; | |
4253 | #ifdef HAVE_TMPFILE | |
4254 | is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); | |
4255 | #endif | |
3558fd73 | 4256 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 4257 | |
32dec7bd | 4258 | if (name == NULL) |
4259 | return (SET_ERROR(EINVAL)); | |
4260 | ||
0037b49e | 4261 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4262 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 4263 | zilog = zfsvfs->z_log; |
34dc7c2f | 4264 | |
428870ff BB |
4265 | /* |
4266 | * POSIX dictates that we return EPERM here. | |
4267 | * Better choices include ENOTSUP or EISDIR. | |
4268 | */ | |
3558fd73 | 4269 | if (S_ISDIR(sip->i_mode)) { |
0037b49e | 4270 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4271 | return (SET_ERROR(EPERM)); |
428870ff BB |
4272 | } |
4273 | ||
812e91a7 MT |
4274 | szp = ITOZ(sip); |
4275 | ZFS_VERIFY_ZP(szp); | |
4276 | ||
9c5167d1 NF |
4277 | /* |
4278 | * If we are using project inheritance, means if the directory has | |
4279 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
4280 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
4281 | * such case, we only allow hard link creation in our tree when the | |
4282 | * project IDs are the same. | |
4283 | */ | |
4284 | if (dzp->z_pflags & ZFS_PROJINHERIT && dzp->z_projid != szp->z_projid) { | |
4285 | ZFS_EXIT(zfsvfs); | |
4286 | return (SET_ERROR(EXDEV)); | |
4287 | } | |
4288 | ||
812e91a7 MT |
4289 | /* |
4290 | * We check i_sb because snapshots and the ctldir must have different | |
4291 | * super blocks. | |
4292 | */ | |
c0ebc844 | 4293 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
0037b49e | 4294 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4295 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 4296 | } |
428870ff | 4297 | |
428870ff BB |
4298 | /* Prevent links to .zfs/shares files */ |
4299 | ||
0037b49e | 4300 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff | 4301 | &parent, sizeof (uint64_t))) != 0) { |
0037b49e | 4302 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4303 | return (error); |
4304 | } | |
0037b49e BB |
4305 | if (parent == zfsvfs->z_shares_dir) { |
4306 | ZFS_EXIT(zfsvfs); | |
2e528b49 | 4307 | return (SET_ERROR(EPERM)); |
428870ff BB |
4308 | } |
4309 | ||
0037b49e | 4310 | if (zfsvfs->z_utf8 && u8_validate(name, |
34dc7c2f | 4311 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 4312 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4313 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
4314 | } |
4315 | if (flags & FIGNORECASE) | |
4316 | zf |= ZCILOOK; | |
4317 | ||
34dc7c2f BB |
4318 | /* |
4319 | * We do not support links between attributes and non-attributes | |
4320 | * because of the potential security risk of creating links | |
4321 | * into "normal" file space in order to circumvent restrictions | |
4322 | * imposed in attribute space. | |
4323 | */ | |
428870ff | 4324 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 4325 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4326 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4327 | } |
4328 | ||
0037b49e BB |
4329 | owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid), |
4330 | cr, ZFS_OWNER); | |
572e2857 | 4331 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
0037b49e | 4332 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4333 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4334 | } |
4335 | ||
149e873a | 4336 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
0037b49e | 4337 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4338 | return (error); |
4339 | } | |
4340 | ||
428870ff | 4341 | top: |
34dc7c2f BB |
4342 | /* |
4343 | * Attempt to lock directory; fail if entry already exists. | |
4344 | */ | |
4345 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
4346 | if (error) { | |
0037b49e | 4347 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4348 | return (error); |
4349 | } | |
4350 | ||
0037b49e | 4351 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 4352 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 4353 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
ace1eae8 | 4354 | if (is_tmpfile) |
0037b49e | 4355 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 4356 | |
428870ff BB |
4357 | zfs_sa_upgrade_txholds(tx, szp); |
4358 | zfs_sa_upgrade_txholds(tx, dzp); | |
0735ecb3 | 4359 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
4360 | if (error) { |
4361 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4362 | if (error == ERESTART) { |
e8b96c60 | 4363 | waited = B_TRUE; |
34dc7c2f BB |
4364 | dmu_tx_wait(tx); |
4365 | dmu_tx_abort(tx); | |
4366 | goto top; | |
4367 | } | |
4368 | dmu_tx_abort(tx); | |
0037b49e | 4369 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4370 | return (error); |
4371 | } | |
ace1eae8 CC |
4372 | /* unmark z_unlinked so zfs_link_create will not reject */ |
4373 | if (is_tmpfile) | |
a43570c5 | 4374 | szp->z_unlinked = B_FALSE; |
34dc7c2f BB |
4375 | error = zfs_link_create(dl, szp, tx, 0); |
4376 | ||
4377 | if (error == 0) { | |
4378 | uint64_t txtype = TX_LINK; | |
ace1eae8 CC |
4379 | /* |
4380 | * tmpfile is created to be in z_unlinkedobj, so remove it. | |
4381 | * Also, we don't log in ZIL, be cause all previous file | |
4382 | * operation on the tmpfile are ignored by ZIL. Instead we | |
4383 | * always wait for txg to sync to make sure all previous | |
4384 | * operation are sync safe. | |
4385 | */ | |
4386 | if (is_tmpfile) { | |
0037b49e BB |
4387 | VERIFY(zap_remove_int(zfsvfs->z_os, |
4388 | zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); | |
ace1eae8 CC |
4389 | } else { |
4390 | if (flags & FIGNORECASE) | |
4391 | txtype |= TX_CI; | |
4392 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); | |
4393 | } | |
4394 | } else if (is_tmpfile) { | |
4395 | /* restore z_unlinked since when linking failed */ | |
a43570c5 | 4396 | szp->z_unlinked = B_TRUE; |
34dc7c2f | 4397 | } |
ace1eae8 | 4398 | txg = dmu_tx_get_txg(tx); |
34dc7c2f BB |
4399 | dmu_tx_commit(tx); |
4400 | ||
4401 | zfs_dirent_unlock(dl); | |
4402 | ||
0037b49e | 4403 | if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4404 | zil_commit(zilog, 0); |
428870ff | 4405 | |
ace1eae8 | 4406 | if (is_tmpfile) |
0037b49e | 4407 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg); |
ace1eae8 | 4408 | |
960e08fe BB |
4409 | zfs_inode_update(dzp); |
4410 | zfs_inode_update(szp); | |
0037b49e | 4411 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4412 | return (error); |
4413 | } | |
4414 | ||
3c0e5c0f | 4415 | static void |
119a394a | 4416 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
4417 | { |
4418 | struct page *pp = arg; | |
4419 | ||
119a394a | 4420 | ClearPageError(pp); |
3c0e5c0f BB |
4421 | end_page_writeback(pp); |
4422 | } | |
4423 | ||
34dc7c2f | 4424 | /* |
3c0e5c0f BB |
4425 | * Push a page out to disk, once the page is on stable storage the |
4426 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 4427 | * |
3c0e5c0f BB |
4428 | * IN: ip - page mapped for inode. |
4429 | * pp - page to push (page is locked) | |
4430 | * wbc - writeback control data | |
34dc7c2f BB |
4431 | * |
4432 | * RETURN: 0 if success | |
4433 | * error code if failure | |
4434 | * | |
3c0e5c0f BB |
4435 | * Timestamps: |
4436 | * ip - ctime|mtime updated | |
34dc7c2f BB |
4437 | */ |
4438 | /* ARGSUSED */ | |
3c0e5c0f BB |
4439 | int |
4440 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 4441 | { |
3c0e5c0f | 4442 | znode_t *zp = ITOZ(ip); |
0037b49e | 4443 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
3c0e5c0f BB |
4444 | loff_t offset; |
4445 | loff_t pgoff; | |
4c837f0d | 4446 | unsigned int pglen; |
3c0e5c0f BB |
4447 | dmu_tx_t *tx; |
4448 | caddr_t va; | |
4449 | int err = 0; | |
4450 | uint64_t mtime[2], ctime[2]; | |
4451 | sa_bulk_attr_t bulk[3]; | |
4452 | int cnt = 0; | |
21a96fb6 | 4453 | struct address_space *mapping; |
3c0e5c0f | 4454 | |
0037b49e | 4455 | ZFS_ENTER(zfsvfs); |
4c837f0d | 4456 | ZFS_VERIFY_ZP(zp); |
d164b209 | 4457 | |
3c0e5c0f BB |
4458 | ASSERT(PageLocked(pp)); |
4459 | ||
d1d7e268 MK |
4460 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
4461 | offset = i_size_read(ip); /* File length in bytes */ | |
8b1899d3 BB |
4462 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
4463 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
4464 | |
4465 | /* Page is beyond end of file */ | |
4466 | if (pgoff >= offset) { | |
4467 | unlock_page(pp); | |
0037b49e | 4468 | ZFS_EXIT(zfsvfs); |
3c0e5c0f BB |
4469 | return (0); |
4470 | } | |
4471 | ||
4472 | /* Truncate page length to end of file */ | |
4473 | if (pgoff + pglen > offset) | |
4474 | pglen = offset - pgoff; | |
4475 | ||
4476 | #if 0 | |
34dc7c2f | 4477 | /* |
3c0e5c0f BB |
4478 | * FIXME: Allow mmap writes past its quota. The correct fix |
4479 | * is to register a page_mkwrite() handler to count the page | |
4480 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 4481 | */ |
9c5167d1 NF |
4482 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
4483 | KUID_TO_SUID(ip->i_uid)) || | |
4484 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
4485 | KGID_TO_SGID(ip->i_gid)) || | |
4486 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
4487 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
4488 | zp->z_projid))) { | |
9babb374 | 4489 | err = EDQUOT; |
9babb374 | 4490 | } |
3c0e5c0f BB |
4491 | #endif |
4492 | ||
d958324f BB |
4493 | /* |
4494 | * The ordering here is critical and must adhere to the following | |
4495 | * rules in order to avoid deadlocking in either zfs_read() or | |
4496 | * zfs_free_range() due to a lock inversion. | |
4497 | * | |
4498 | * 1) The page must be unlocked prior to acquiring the range lock. | |
4499 | * This is critical because zfs_read() calls find_lock_page() | |
4500 | * which may block on the page lock while holding the range lock. | |
4501 | * | |
4502 | * 2) Before setting or clearing write back on a page the range lock | |
4503 | * must be held in order to prevent a lock inversion with the | |
4504 | * zfs_free_range() function. | |
21a96fb6 CC |
4505 | * |
4506 | * This presents a problem because upon entering this function the | |
4507 | * page lock is already held. To safely acquire the range lock the | |
4508 | * page lock must be dropped. This creates a window where another | |
4509 | * process could truncate, invalidate, dirty, or write out the page. | |
4510 | * | |
4511 | * Therefore, after successfully reacquiring the range and page locks | |
4512 | * the current page state is checked. In the common case everything | |
4513 | * will be as is expected and it can be written out. However, if | |
4514 | * the page state has changed it must be handled accordingly. | |
d958324f | 4515 | */ |
21a96fb6 CC |
4516 | mapping = pp->mapping; |
4517 | redirty_page_for_writepage(wbc, pp); | |
d958324f | 4518 | unlock_page(pp); |
21a96fb6 | 4519 | |
5d43cc9a MA |
4520 | locked_range_t *lr = rangelock_enter(&zp->z_rangelock, |
4521 | pgoff, pglen, RL_WRITER); | |
21a96fb6 CC |
4522 | lock_page(pp); |
4523 | ||
4524 | /* Page mapping changed or it was no longer dirty, we're done */ | |
4525 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
4526 | unlock_page(pp); | |
5d43cc9a | 4527 | rangelock_exit(lr); |
0037b49e | 4528 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4529 | return (0); |
4530 | } | |
4531 | ||
4532 | /* Another process started write block if required */ | |
4533 | if (PageWriteback(pp)) { | |
4534 | unlock_page(pp); | |
5d43cc9a | 4535 | rangelock_exit(lr); |
21a96fb6 | 4536 | |
bfd5a709 TK |
4537 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4538 | if (PageWriteback(pp)) | |
4539 | wait_on_page_bit(pp, PG_writeback); | |
4540 | } | |
21a96fb6 | 4541 | |
0037b49e | 4542 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4543 | return (0); |
4544 | } | |
4545 | ||
4546 | /* Clear the dirty flag the required locks are held */ | |
4547 | if (!clear_page_dirty_for_io(pp)) { | |
4548 | unlock_page(pp); | |
5d43cc9a | 4549 | rangelock_exit(lr); |
0037b49e | 4550 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4551 | return (0); |
4552 | } | |
4553 | ||
4554 | /* | |
4555 | * Counterpart for redirty_page_for_writepage() above. This page | |
4556 | * was in fact not skipped and should not be counted as if it were. | |
4557 | */ | |
4558 | wbc->pages_skipped--; | |
3c0e5c0f | 4559 | set_page_writeback(pp); |
21a96fb6 | 4560 | unlock_page(pp); |
3c0e5c0f | 4561 | |
0037b49e | 4562 | tx = dmu_tx_create(zfsvfs->z_os); |
3c0e5c0f | 4563 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
4564 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4565 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 4566 | |
fb5f0bc8 | 4567 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 4568 | if (err != 0) { |
3c0e5c0f | 4569 | if (err == ERESTART) |
34dc7c2f | 4570 | dmu_tx_wait(tx); |
3c0e5c0f | 4571 | |
34dc7c2f | 4572 | dmu_tx_abort(tx); |
119a394a ED |
4573 | __set_page_dirty_nobuffers(pp); |
4574 | ClearPageError(pp); | |
4575 | end_page_writeback(pp); | |
5d43cc9a | 4576 | rangelock_exit(lr); |
0037b49e | 4577 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4578 | return (err); |
34dc7c2f BB |
4579 | } |
4580 | ||
dde471ef | 4581 | va = kmap(pp); |
8b1899d3 | 4582 | ASSERT3U(pglen, <=, PAGE_SIZE); |
0037b49e | 4583 | dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 4584 | kunmap(pp); |
34dc7c2f | 4585 | |
0037b49e BB |
4586 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
4587 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
4588 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, | |
4589 | &zp->z_pflags, 8); | |
428870ff | 4590 | |
d3aa3ea9 BB |
4591 | /* Preserve the mtime and ctime provided by the inode */ |
4592 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4593 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
a43570c5 | 4594 | zp->z_atime_dirty = B_FALSE; |
d3aa3ea9 BB |
4595 | zp->z_seq++; |
4596 | ||
4597 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4598 | ||
0037b49e | 4599 | zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
119a394a | 4600 | zfs_putpage_commit_cb, pp); |
45d1cae3 | 4601 | dmu_tx_commit(tx); |
d3aa3ea9 | 4602 | |
5d43cc9a | 4603 | rangelock_exit(lr); |
34dc7c2f | 4604 | |
119a394a ED |
4605 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4606 | /* | |
4607 | * Note that this is rarely called under writepages(), because | |
4608 | * writepages() normally handles the entire commit for | |
4609 | * performance reasons. | |
4610 | */ | |
0037b49e | 4611 | zil_commit(zfsvfs->z_log, zp->z_id); |
2b286136 | 4612 | } |
3c0e5c0f | 4613 | |
0037b49e | 4614 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4615 | return (err); |
34dc7c2f BB |
4616 | } |
4617 | ||
8780c539 BB |
4618 | /* |
4619 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 4620 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
4621 | */ |
4622 | int | |
4623 | zfs_dirty_inode(struct inode *ip, int flags) | |
4624 | { | |
4625 | znode_t *zp = ITOZ(ip); | |
0037b49e | 4626 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
8780c539 | 4627 | dmu_tx_t *tx; |
023699cd MM |
4628 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
4629 | sa_bulk_attr_t bulk[4]; | |
704cd075 | 4630 | int error = 0; |
8780c539 BB |
4631 | int cnt = 0; |
4632 | ||
0037b49e | 4633 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
c944be5d BB |
4634 | return (0); |
4635 | ||
0037b49e | 4636 | ZFS_ENTER(zfsvfs); |
8780c539 BB |
4637 | ZFS_VERIFY_ZP(zp); |
4638 | ||
704cd075 CC |
4639 | #ifdef I_DIRTY_TIME |
4640 | /* | |
4641 | * This is the lazytime semantic indroduced in Linux 4.0 | |
4642 | * This flag will only be called from update_time when lazytime is set. | |
4643 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
4644 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
4645 | * only need to dirty atime. | |
4646 | */ | |
4647 | if (flags == I_DIRTY_TIME) { | |
a43570c5 | 4648 | zp->z_atime_dirty = B_TRUE; |
704cd075 CC |
4649 | goto out; |
4650 | } | |
4651 | #endif | |
4652 | ||
0037b49e | 4653 | tx = dmu_tx_create(zfsvfs->z_os); |
8780c539 BB |
4654 | |
4655 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4656 | zfs_sa_upgrade_txholds(tx, zp); | |
4657 | ||
4658 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4659 | if (error) { | |
4660 | dmu_tx_abort(tx); | |
4661 | goto out; | |
4662 | } | |
4663 | ||
4664 | mutex_enter(&zp->z_lock); | |
a43570c5 | 4665 | zp->z_atime_dirty = B_FALSE; |
704cd075 | 4666 | |
0037b49e BB |
4667 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
4668 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
4669 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
4670 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
8780c539 | 4671 | |
023699cd | 4672 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4673 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4674 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4675 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
023699cd MM |
4676 | mode = ip->i_mode; |
4677 | ||
4678 | zp->z_mode = mode; | |
8780c539 BB |
4679 | |
4680 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4681 | mutex_exit(&zp->z_lock); | |
4682 | ||
4683 | dmu_tx_commit(tx); | |
4684 | out: | |
0037b49e | 4685 | ZFS_EXIT(zfsvfs); |
8780c539 BB |
4686 | return (error); |
4687 | } | |
8780c539 | 4688 | |
34dc7c2f BB |
4689 | /*ARGSUSED*/ |
4690 | void | |
c0d35759 | 4691 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4692 | { |
c0d35759 | 4693 | znode_t *zp = ITOZ(ip); |
0037b49e | 4694 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
0df9673f | 4695 | uint64_t atime[2]; |
34dc7c2f | 4696 | int error; |
cafbd2ac | 4697 | int need_unlock = 0; |
34dc7c2f | 4698 | |
cafbd2ac | 4699 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
0037b49e | 4700 | if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) { |
cafbd2ac | 4701 | need_unlock = 1; |
0037b49e | 4702 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); |
cafbd2ac | 4703 | } |
c0d35759 | 4704 | if (zp->z_sa_hdl == NULL) { |
cafbd2ac | 4705 | if (need_unlock) |
0037b49e | 4706 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
c0d35759 | 4707 | return; |
34dc7c2f BB |
4708 | } |
4709 | ||
a43570c5 | 4710 | if (zp->z_atime_dirty && zp->z_unlinked == B_FALSE) { |
0037b49e | 4711 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 4712 | |
428870ff BB |
4713 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4714 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4715 | error = dmu_tx_assign(tx, TXG_WAIT); |
4716 | if (error) { | |
4717 | dmu_tx_abort(tx); | |
4718 | } else { | |
0df9673f | 4719 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
34dc7c2f | 4720 | mutex_enter(&zp->z_lock); |
0037b49e | 4721 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), |
0df9673f | 4722 | (void *)&atime, sizeof (atime), tx); |
a43570c5 | 4723 | zp->z_atime_dirty = B_FALSE; |
34dc7c2f BB |
4724 | mutex_exit(&zp->z_lock); |
4725 | dmu_tx_commit(tx); | |
4726 | } | |
4727 | } | |
4728 | ||
4729 | zfs_zinactive(zp); | |
cafbd2ac | 4730 | if (need_unlock) |
0037b49e | 4731 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
34dc7c2f BB |
4732 | } |
4733 | ||
4734 | /* | |
4735 | * Bounds-check the seek operation. | |
4736 | * | |
3558fd73 | 4737 | * IN: ip - inode seeking within |
34dc7c2f BB |
4738 | * ooff - old file offset |
4739 | * noffp - pointer to new file offset | |
34dc7c2f BB |
4740 | * |
4741 | * RETURN: 0 if success | |
4742 | * EINVAL if new offset invalid | |
4743 | */ | |
4744 | /* ARGSUSED */ | |
3558fd73 | 4745 | int |
9623f736 | 4746 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4747 | { |
3558fd73 | 4748 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4749 | return (0); |
4750 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4751 | } | |
4752 | ||
34dc7c2f | 4753 | /* |
dde471ef | 4754 | * Fill pages with data from the disk. |
34dc7c2f BB |
4755 | */ |
4756 | static int | |
dde471ef | 4757 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4758 | { |
d1d7e268 | 4759 | znode_t *zp = ITOZ(ip); |
0037b49e | 4760 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
d1d7e268 | 4761 | objset_t *os; |
dde471ef | 4762 | struct page *cur_pp; |
d1d7e268 MK |
4763 | u_offset_t io_off, total; |
4764 | size_t io_len; | |
4765 | loff_t i_size; | |
4766 | unsigned page_idx; | |
4767 | int err; | |
34dc7c2f | 4768 | |
0037b49e | 4769 | os = zfsvfs->z_os; |
8b1899d3 | 4770 | io_len = nr_pages << PAGE_SHIFT; |
dde471ef PJ |
4771 | i_size = i_size_read(ip); |
4772 | io_off = page_offset(pl[0]); | |
4773 | ||
4774 | if (io_off + io_len > i_size) | |
4775 | io_len = i_size - io_off; | |
34dc7c2f BB |
4776 | |
4777 | /* | |
dde471ef | 4778 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4779 | */ |
dde471ef | 4780 | page_idx = 0; |
34dc7c2f | 4781 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4782 | caddr_t va; |
4783 | ||
540c3927 | 4784 | cur_pp = pl[page_idx++]; |
dde471ef | 4785 | va = kmap(cur_pp); |
9babb374 BB |
4786 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4787 | DMU_READ_PREFETCH); | |
dde471ef | 4788 | kunmap(cur_pp); |
34dc7c2f | 4789 | if (err) { |
b128c09f BB |
4790 | /* convert checksum errors into IO errors */ |
4791 | if (err == ECKSUM) | |
2e528b49 | 4792 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4793 | return (err); |
4794 | } | |
34dc7c2f | 4795 | } |
d164b209 | 4796 | |
34dc7c2f BB |
4797 | return (0); |
4798 | } | |
4799 | ||
4800 | /* | |
dde471ef | 4801 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4802 | * |
dde471ef PJ |
4803 | * IN: ip - inode of file to get data from. |
4804 | * pl - list of pages to read | |
4805 | * nr_pages - number of pages to read | |
34dc7c2f | 4806 | * |
d3cc8b15 | 4807 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4808 | * |
4809 | * Timestamps: | |
4810 | * vp - atime updated | |
4811 | */ | |
4812 | /* ARGSUSED */ | |
dde471ef PJ |
4813 | int |
4814 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4815 | { |
dde471ef | 4816 | znode_t *zp = ITOZ(ip); |
0037b49e | 4817 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
dde471ef | 4818 | int err; |
d164b209 | 4819 | |
d164b209 BB |
4820 | if (pl == NULL) |
4821 | return (0); | |
34dc7c2f | 4822 | |
0037b49e | 4823 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4824 | ZFS_VERIFY_ZP(zp); |
4825 | ||
dde471ef | 4826 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4827 | |
0037b49e | 4828 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4829 | return (err); |
4830 | } | |
4831 | ||
4832 | /* | |
e2e7aa2d | 4833 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4834 | * |
e2e7aa2d BB |
4835 | * IN: ip - inode of the file to mmap |
4836 | * off - file offset | |
4837 | * addrp - start address in memory region | |
4838 | * len - length of memory region | |
4839 | * vm_flags- address flags | |
34dc7c2f | 4840 | * |
e2e7aa2d BB |
4841 | * RETURN: 0 if success |
4842 | * error code if failure | |
34dc7c2f BB |
4843 | */ |
4844 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4845 | int |
4846 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4847 | unsigned long vm_flags) | |
34dc7c2f | 4848 | { |
e2e7aa2d | 4849 | znode_t *zp = ITOZ(ip); |
0037b49e | 4850 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4851 | |
0037b49e | 4852 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4853 | ZFS_VERIFY_ZP(zp); |
4854 | ||
e2e7aa2d | 4855 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4856 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
0037b49e | 4857 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4858 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4859 | } |
4860 | ||
e2e7aa2d | 4861 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4862 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
0037b49e | 4863 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4864 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4865 | } |
4866 | ||
34dc7c2f | 4867 | if (off < 0 || len > MAXOFFSET_T - off) { |
0037b49e | 4868 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4869 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4870 | } |
4871 | ||
0037b49e | 4872 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4873 | return (0); |
4874 | } | |
4875 | ||
3558fd73 BB |
4876 | /* |
4877 | * convoff - converts the given data (start, whence) to the | |
4878 | * given whence. | |
4879 | */ | |
4880 | int | |
4881 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4882 | { | |
5484965a | 4883 | vattr_t vap; |
3558fd73 BB |
4884 | int error; |
4885 | ||
126d0fa7 | 4886 | if ((lckdat->l_whence == SEEK_END) || (whence == SEEK_END)) { |
d95a5980 | 4887 | if ((error = zfs_getattr(ip, &vap, 0, CRED()))) |
3558fd73 BB |
4888 | return (error); |
4889 | } | |
4890 | ||
4891 | switch (lckdat->l_whence) { | |
126d0fa7 | 4892 | case SEEK_CUR: |
3558fd73 BB |
4893 | lckdat->l_start += offset; |
4894 | break; | |
126d0fa7 | 4895 | case SEEK_END: |
5484965a | 4896 | lckdat->l_start += vap.va_size; |
3558fd73 | 4897 | /* FALLTHRU */ |
126d0fa7 | 4898 | case SEEK_SET: |
3558fd73 BB |
4899 | break; |
4900 | default: | |
2e528b49 | 4901 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4902 | } |
4903 | ||
4904 | if (lckdat->l_start < 0) | |
2e528b49 | 4905 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4906 | |
4907 | switch (whence) { | |
126d0fa7 | 4908 | case SEEK_CUR: |
3558fd73 BB |
4909 | lckdat->l_start -= offset; |
4910 | break; | |
126d0fa7 | 4911 | case SEEK_END: |
5484965a | 4912 | lckdat->l_start -= vap.va_size; |
3558fd73 | 4913 | /* FALLTHRU */ |
126d0fa7 | 4914 | case SEEK_SET: |
3558fd73 BB |
4915 | break; |
4916 | default: | |
2e528b49 | 4917 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4918 | } |
4919 | ||
4920 | lckdat->l_whence = (short)whence; | |
4921 | return (0); | |
4922 | } | |
4923 | ||
34dc7c2f BB |
4924 | /* |
4925 | * Free or allocate space in a file. Currently, this function only | |
4926 | * supports the `F_FREESP' command. However, this command is somewhat | |
4927 | * misnamed, as its functionality includes the ability to allocate as | |
4928 | * well as free space. | |
4929 | * | |
3558fd73 | 4930 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4931 | * cmd - action to take (only F_FREESP supported). |
4932 | * bfp - section of file to free/alloc. | |
4933 | * flag - current file open mode flags. | |
4934 | * offset - current file offset. | |
a35c1207 | 4935 | * cr - credentials of caller. |
34dc7c2f | 4936 | * |
d3cc8b15 | 4937 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4938 | * |
4939 | * Timestamps: | |
3558fd73 | 4940 | * ip - ctime|mtime updated |
34dc7c2f BB |
4941 | */ |
4942 | /* ARGSUSED */ | |
e5c39b95 | 4943 | int |
3558fd73 BB |
4944 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4945 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4946 | { |
3558fd73 | 4947 | znode_t *zp = ITOZ(ip); |
0037b49e | 4948 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4949 | uint64_t off, len; |
4950 | int error; | |
4951 | ||
0037b49e | 4952 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4953 | ZFS_VERIFY_ZP(zp); |
4954 | ||
34dc7c2f | 4955 | if (cmd != F_FREESP) { |
0037b49e | 4956 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4957 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4958 | } |
4959 | ||
f3c9dca0 MT |
4960 | /* |
4961 | * Callers might not be able to detect properly that we are read-only, | |
4962 | * so check it explicitly here. | |
4963 | */ | |
0037b49e BB |
4964 | if (zfs_is_readonly(zfsvfs)) { |
4965 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
4966 | return (SET_ERROR(EROFS)); |
4967 | } | |
4968 | ||
126d0fa7 | 4969 | if ((error = convoff(ip, bfp, SEEK_SET, offset))) { |
0037b49e | 4970 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4971 | return (error); |
4972 | } | |
4973 | ||
4974 | if (bfp->l_len < 0) { | |
0037b49e | 4975 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4976 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4977 | } |
4978 | ||
aec69371 ED |
4979 | /* |
4980 | * Permissions aren't checked on Solaris because on this OS | |
4981 | * zfs_space() can only be called with an opened file handle. | |
4982 | * On Linux we can get here through truncate_range() which | |
4983 | * operates directly on inodes, so we need to check access rights. | |
4984 | */ | |
4985 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
0037b49e | 4986 | ZFS_EXIT(zfsvfs); |
aec69371 ED |
4987 | return (error); |
4988 | } | |
4989 | ||
34dc7c2f BB |
4990 | off = bfp->l_start; |
4991 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4992 | ||
b128c09f | 4993 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4994 | |
0037b49e | 4995 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4996 | return (error); |
4997 | } | |
4998 | ||
4999 | /*ARGSUSED*/ | |
e5c39b95 | 5000 | int |
3558fd73 | 5001 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 5002 | { |
3558fd73 | 5003 | znode_t *zp = ITOZ(ip); |
0037b49e | 5004 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 5005 | uint32_t gen; |
428870ff | 5006 | uint64_t gen64; |
34dc7c2f BB |
5007 | uint64_t object = zp->z_id; |
5008 | zfid_short_t *zfid; | |
428870ff | 5009 | int size, i, error; |
34dc7c2f | 5010 | |
0037b49e | 5011 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 5012 | ZFS_VERIFY_ZP(zp); |
428870ff | 5013 | |
0037b49e | 5014 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), |
428870ff | 5015 | &gen64, sizeof (uint64_t))) != 0) { |
0037b49e | 5016 | ZFS_EXIT(zfsvfs); |
428870ff BB |
5017 | return (error); |
5018 | } | |
5019 | ||
5020 | gen = (uint32_t)gen64; | |
34dc7c2f | 5021 | |
9b77d1c9 | 5022 | size = SHORT_FID_LEN; |
34dc7c2f BB |
5023 | |
5024 | zfid = (zfid_short_t *)fidp; | |
5025 | ||
5026 | zfid->zf_len = size; | |
5027 | ||
5028 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
5029 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
5030 | ||
5031 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
5032 | if (gen == 0) | |
5033 | gen = 1; | |
5034 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
5035 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
5036 | ||
0037b49e | 5037 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5038 | return (0); |
5039 | } | |
5040 | ||
34dc7c2f | 5041 | /*ARGSUSED*/ |
e5c39b95 | 5042 | int |
3558fd73 | 5043 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 5044 | { |
3558fd73 | 5045 | znode_t *zp = ITOZ(ip); |
0037b49e | 5046 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
5047 | int error; |
5048 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
5049 | ||
0037b49e | 5050 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
5051 | ZFS_VERIFY_ZP(zp); |
5052 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
0037b49e | 5053 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5054 | |
5055 | return (error); | |
5056 | } | |
5057 | ||
5058 | /*ARGSUSED*/ | |
e5c39b95 | 5059 | int |
3558fd73 | 5060 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 5061 | { |
3558fd73 | 5062 | znode_t *zp = ITOZ(ip); |
0037b49e | 5063 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
5064 | int error; |
5065 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
0037b49e | 5066 | zilog_t *zilog = zfsvfs->z_log; |
34dc7c2f | 5067 | |
0037b49e | 5068 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 5069 | ZFS_VERIFY_ZP(zp); |
428870ff | 5070 | |
34dc7c2f | 5071 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 5072 | |
0037b49e | 5073 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 5074 | zil_commit(zilog, 0); |
428870ff | 5075 | |
0037b49e | 5076 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5077 | return (error); |
5078 | } | |
5079 | ||
3558fd73 | 5080 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 5081 | /* |
f79121d1 TK |
5082 | * The smallest read we may consider to loan out an arcbuf. |
5083 | * This must be a power of 2. | |
428870ff BB |
5084 | */ |
5085 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
f79121d1 TK |
5086 | /* |
5087 | * If set to less than the file block size, allow loaning out of an | |
5088 | * arcbuf for a partial block read. This must be a power of 2. | |
5089 | */ | |
428870ff BB |
5090 | int zcr_blksz_max = (1 << 17); /* 128K */ |
5091 | ||
5092 | /*ARGSUSED*/ | |
5093 | static int | |
3558fd73 | 5094 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 5095 | { |
3558fd73 | 5096 | znode_t *zp = ITOZ(ip); |
0037b49e BB |
5097 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
5098 | int max_blksz = zfsvfs->z_max_blksz; | |
428870ff BB |
5099 | uio_t *uio = &xuio->xu_uio; |
5100 | ssize_t size = uio->uio_resid; | |
5101 | offset_t offset = uio->uio_loffset; | |
5102 | int blksz; | |
5103 | int fullblk, i; | |
5104 | arc_buf_t *abuf; | |
5105 | ssize_t maxsize; | |
5106 | int preamble, postamble; | |
5107 | ||
5108 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
2e528b49 | 5109 | return (SET_ERROR(EINVAL)); |
428870ff | 5110 | |
0037b49e | 5111 | ZFS_ENTER(zfsvfs); |
428870ff BB |
5112 | ZFS_VERIFY_ZP(zp); |
5113 | switch (ioflag) { | |
5114 | case UIO_WRITE: | |
5115 | /* | |
5116 | * Loan out an arc_buf for write if write size is bigger than | |
5117 | * max_blksz, and the file's block size is also max_blksz. | |
5118 | */ | |
5119 | blksz = max_blksz; | |
5120 | if (size < blksz || zp->z_blksz != blksz) { | |
0037b49e | 5121 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5122 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5123 | } |
5124 | /* | |
5125 | * Caller requests buffers for write before knowing where the | |
5126 | * write offset might be (e.g. NFS TCP write). | |
5127 | */ | |
5128 | if (offset == -1) { | |
5129 | preamble = 0; | |
5130 | } else { | |
5131 | preamble = P2PHASE(offset, blksz); | |
5132 | if (preamble) { | |
5133 | preamble = blksz - preamble; | |
5134 | size -= preamble; | |
5135 | } | |
5136 | } | |
5137 | ||
5138 | postamble = P2PHASE(size, blksz); | |
5139 | size -= postamble; | |
5140 | ||
5141 | fullblk = size / blksz; | |
5142 | (void) dmu_xuio_init(xuio, | |
5143 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
5144 | |
5145 | /* | |
5146 | * Have to fix iov base/len for partial buffers. They | |
5147 | * currently represent full arc_buf's. | |
5148 | */ | |
5149 | if (preamble) { | |
5150 | /* data begins in the middle of the arc_buf */ | |
5151 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5152 | blksz); | |
5153 | ASSERT(abuf); | |
5154 | (void) dmu_xuio_add(xuio, abuf, | |
5155 | blksz - preamble, preamble); | |
5156 | } | |
5157 | ||
5158 | for (i = 0; i < fullblk; i++) { | |
5159 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5160 | blksz); | |
5161 | ASSERT(abuf); | |
5162 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
5163 | } | |
5164 | ||
5165 | if (postamble) { | |
5166 | /* data ends in the middle of the arc_buf */ | |
5167 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5168 | blksz); | |
5169 | ASSERT(abuf); | |
5170 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
5171 | } | |
5172 | break; | |
5173 | case UIO_READ: | |
5174 | /* | |
5175 | * Loan out an arc_buf for read if the read size is larger than | |
5176 | * the current file block size. Block alignment is not | |
5177 | * considered. Partial arc_buf will be loaned out for read. | |
5178 | */ | |
5179 | blksz = zp->z_blksz; | |
5180 | if (blksz < zcr_blksz_min) | |
5181 | blksz = zcr_blksz_min; | |
5182 | if (blksz > zcr_blksz_max) | |
5183 | blksz = zcr_blksz_max; | |
5184 | /* avoid potential complexity of dealing with it */ | |
5185 | if (blksz > max_blksz) { | |
0037b49e | 5186 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5187 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5188 | } |
5189 | ||
5190 | maxsize = zp->z_size - uio->uio_loffset; | |
5191 | if (size > maxsize) | |
5192 | size = maxsize; | |
5193 | ||
3558fd73 | 5194 | if (size < blksz) { |
0037b49e | 5195 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5196 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5197 | } |
5198 | break; | |
5199 | default: | |
0037b49e | 5200 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5201 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5202 | } |
5203 | ||
5204 | uio->uio_extflg = UIO_XUIO; | |
5205 | XUIO_XUZC_RW(xuio) = ioflag; | |
0037b49e | 5206 | ZFS_EXIT(zfsvfs); |
428870ff BB |
5207 | return (0); |
5208 | } | |
5209 | ||
5210 | /*ARGSUSED*/ | |
5211 | static int | |
3558fd73 | 5212 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
5213 | { |
5214 | int i; | |
5215 | arc_buf_t *abuf; | |
5216 | int ioflag = XUIO_XUZC_RW(xuio); | |
5217 | ||
5218 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
5219 | ||
5220 | i = dmu_xuio_cnt(xuio); | |
5221 | while (i-- > 0) { | |
5222 | abuf = dmu_xuio_arcbuf(xuio, i); | |
5223 | /* | |
5224 | * if abuf == NULL, it must be a write buffer | |
5225 | * that has been returned in zfs_write(). | |
5226 | */ | |
5227 | if (abuf) | |
5228 | dmu_return_arcbuf(abuf); | |
5229 | ASSERT(abuf || ioflag == UIO_WRITE); | |
5230 | } | |
5231 | ||
5232 | dmu_xuio_fini(xuio); | |
5233 | return (0); | |
5234 | } | |
3558fd73 | 5235 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 | 5236 | |
93ce2b4c | 5237 | #if defined(_KERNEL) |
f298b24d BB |
5238 | EXPORT_SYMBOL(zfs_open); |
5239 | EXPORT_SYMBOL(zfs_close); | |
5240 | EXPORT_SYMBOL(zfs_read); | |
5241 | EXPORT_SYMBOL(zfs_write); | |
5242 | EXPORT_SYMBOL(zfs_access); | |
5243 | EXPORT_SYMBOL(zfs_lookup); | |
5244 | EXPORT_SYMBOL(zfs_create); | |
5245 | EXPORT_SYMBOL(zfs_tmpfile); | |
5246 | EXPORT_SYMBOL(zfs_remove); | |
5247 | EXPORT_SYMBOL(zfs_mkdir); | |
5248 | EXPORT_SYMBOL(zfs_rmdir); | |
5249 | EXPORT_SYMBOL(zfs_readdir); | |
5250 | EXPORT_SYMBOL(zfs_fsync); | |
5251 | EXPORT_SYMBOL(zfs_getattr); | |
5252 | EXPORT_SYMBOL(zfs_getattr_fast); | |
5253 | EXPORT_SYMBOL(zfs_setattr); | |
5254 | EXPORT_SYMBOL(zfs_rename); | |
5255 | EXPORT_SYMBOL(zfs_symlink); | |
5256 | EXPORT_SYMBOL(zfs_readlink); | |
5257 | EXPORT_SYMBOL(zfs_link); | |
5258 | EXPORT_SYMBOL(zfs_inactive); | |
5259 | EXPORT_SYMBOL(zfs_space); | |
5260 | EXPORT_SYMBOL(zfs_fid); | |
5261 | EXPORT_SYMBOL(zfs_getsecattr); | |
5262 | EXPORT_SYMBOL(zfs_setsecattr); | |
5263 | EXPORT_SYMBOL(zfs_getpage); | |
5264 | EXPORT_SYMBOL(zfs_putpage); | |
5265 | EXPORT_SYMBOL(zfs_dirty_inode); | |
5266 | EXPORT_SYMBOL(zfs_map); | |
5267 | ||
09276fde | 5268 | /* BEGIN CSTYLED */ |
a966c564 K |
5269 | module_param(zfs_delete_blocks, ulong, 0644); |
5270 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
09276fde | 5271 | module_param(zfs_read_chunk_size, ulong, 0644); |
c409e464 | 5272 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); |
09276fde TK |
5273 | /* END CSTYLED */ |
5274 | ||
c409e464 | 5275 | #endif |