]>
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; |
34dc7c2f | 442 | |
a448a255 SD |
443 | znode_t *zp = ITOZ(ip); |
444 | zfsvfs_t *zfsvfs = ITOZSB(ip); | |
0037b49e | 445 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 446 | ZFS_VERIFY_ZP(zp); |
34dc7c2f | 447 | |
428870ff | 448 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
0037b49e | 449 | ZFS_EXIT(zfsvfs); |
2e528b49 | 450 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
451 | } |
452 | ||
453 | /* | |
454 | * Validate file offset | |
455 | */ | |
456 | if (uio->uio_loffset < (offset_t)0) { | |
0037b49e | 457 | ZFS_EXIT(zfsvfs); |
2e528b49 | 458 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
459 | } |
460 | ||
461 | /* | |
462 | * Fasttrack empty reads | |
463 | */ | |
464 | if (uio->uio_resid == 0) { | |
0037b49e | 465 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
466 | return (0); |
467 | } | |
468 | ||
34dc7c2f BB |
469 | /* |
470 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
37699482 | 471 | * Only do this for non-snapshots. |
34dc7c2f | 472 | */ |
37699482 CC |
473 | if (zfsvfs->z_log && |
474 | (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)) | |
0037b49e | 475 | zil_commit(zfsvfs->z_log, zp->z_id); |
34dc7c2f BB |
476 | |
477 | /* | |
478 | * Lock the range against changes. | |
479 | */ | |
5d43cc9a | 480 | locked_range_t *lr = rangelock_enter(&zp->z_rangelock, |
a448a255 | 481 | uio->uio_loffset, uio->uio_resid, RL_READER); |
34dc7c2f BB |
482 | |
483 | /* | |
484 | * If we are reading past end-of-file we can skip | |
485 | * to the end; but we might still need to set atime. | |
486 | */ | |
428870ff | 487 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
488 | error = 0; |
489 | goto out; | |
490 | } | |
491 | ||
428870ff | 492 | ASSERT(uio->uio_loffset < zp->z_size); |
a448a255 SD |
493 | ssize_t n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); |
494 | ssize_t start_resid = n; | |
428870ff | 495 | |
3558fd73 | 496 | #ifdef HAVE_UIO_ZEROCOPY |
a448a255 | 497 | xuio_t *xuio = NULL; |
428870ff BB |
498 | if ((uio->uio_extflg == UIO_XUIO) && |
499 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
500 | int nblk; | |
501 | int blksz = zp->z_blksz; | |
502 | uint64_t offset = uio->uio_loffset; | |
503 | ||
504 | xuio = (xuio_t *)uio; | |
505 | if ((ISP2(blksz))) { | |
506 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
507 | blksz)) / blksz; | |
508 | } else { | |
509 | ASSERT(offset + n <= blksz); | |
510 | nblk = 1; | |
511 | } | |
512 | (void) dmu_xuio_init(xuio, nblk); | |
513 | ||
3558fd73 | 514 | if (vn_has_cached_data(ip)) { |
428870ff BB |
515 | /* |
516 | * For simplicity, we always allocate a full buffer | |
517 | * even if we only expect to read a portion of a block. | |
518 | */ | |
519 | while (--nblk >= 0) { | |
520 | (void) dmu_xuio_add(xuio, | |
521 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
522 | blksz), 0, blksz); | |
523 | } | |
524 | } | |
525 | } | |
3558fd73 | 526 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
527 | |
528 | while (n > 0) { | |
a448a255 | 529 | ssize_t nbytes = MIN(n, zfs_read_chunk_size - |
34dc7c2f BB |
530 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); |
531 | ||
804e0504 | 532 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
3558fd73 | 533 | error = mappedread(ip, nbytes, uio); |
804e0504 MA |
534 | } else { |
535 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), | |
536 | uio, nbytes); | |
537 | } | |
c0d35759 | 538 | |
b128c09f BB |
539 | if (error) { |
540 | /* convert checksum errors into IO errors */ | |
541 | if (error == ECKSUM) | |
2e528b49 | 542 | error = SET_ERROR(EIO); |
34dc7c2f | 543 | break; |
b128c09f | 544 | } |
34dc7c2f BB |
545 | |
546 | n -= nbytes; | |
547 | } | |
a448a255 SD |
548 | |
549 | int64_t nread = start_resid - n; | |
550 | dataset_kstats_update_read_kstats(&zfsvfs->z_kstat, nread); | |
551 | task_io_account_read(nread); | |
34dc7c2f | 552 | out: |
5d43cc9a | 553 | rangelock_exit(lr); |
34dc7c2f | 554 | |
0037b49e | 555 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
556 | return (error); |
557 | } | |
558 | ||
34dc7c2f BB |
559 | /* |
560 | * Write the bytes to a file. | |
561 | * | |
3558fd73 | 562 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
563 | * uio - structure supplying write location, range info, |
564 | * and data buffer. | |
565 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 566 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 567 | * cr - credentials of caller. |
34dc7c2f BB |
568 | * |
569 | * OUT: uio - updated offset and range. | |
570 | * | |
571 | * RETURN: 0 if success | |
572 | * error code if failure | |
573 | * | |
574 | * Timestamps: | |
3558fd73 | 575 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 576 | */ |
428870ff | 577 | |
34dc7c2f | 578 | /* ARGSUSED */ |
e5c39b95 | 579 | int |
3558fd73 | 580 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 581 | { |
a448a255 SD |
582 | int error = 0; |
583 | ssize_t start_resid = uio->uio_resid; | |
34dc7c2f | 584 | |
34dc7c2f BB |
585 | /* |
586 | * Fasttrack empty write | |
587 | */ | |
a448a255 | 588 | ssize_t n = start_resid; |
34dc7c2f BB |
589 | if (n == 0) |
590 | return (0); | |
591 | ||
a448a255 | 592 | rlim64_t limit = uio->uio_limit; |
34dc7c2f BB |
593 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) |
594 | limit = MAXOFFSET_T; | |
595 | ||
a448a255 SD |
596 | znode_t *zp = ITOZ(ip); |
597 | zfsvfs_t *zfsvfs = ZTOZSB(zp); | |
0037b49e | 598 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 599 | ZFS_VERIFY_ZP(zp); |
b128c09f | 600 | |
a448a255 SD |
601 | sa_bulk_attr_t bulk[4]; |
602 | int count = 0; | |
603 | uint64_t mtime[2], ctime[2]; | |
0037b49e BB |
604 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
605 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
606 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
607 | &zp->z_size, 8); | |
608 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
428870ff BB |
609 | &zp->z_pflags, 8); |
610 | ||
f3c9dca0 MT |
611 | /* |
612 | * Callers might not be able to detect properly that we are read-only, | |
613 | * so check it explicitly here. | |
614 | */ | |
0037b49e BB |
615 | if (zfs_is_readonly(zfsvfs)) { |
616 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
617 | return (SET_ERROR(EROFS)); |
618 | } | |
619 | ||
b128c09f BB |
620 | /* |
621 | * If immutable or not appending then return EPERM | |
622 | */ | |
428870ff BB |
623 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
624 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
625 | (uio->uio_loffset < zp->z_size))) { | |
0037b49e | 626 | ZFS_EXIT(zfsvfs); |
2e528b49 | 627 | return (SET_ERROR(EPERM)); |
b128c09f BB |
628 | } |
629 | ||
428870ff BB |
630 | /* |
631 | * Validate file offset | |
632 | */ | |
a448a255 | 633 | offset_t woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; |
428870ff | 634 | if (woff < 0) { |
0037b49e | 635 | ZFS_EXIT(zfsvfs); |
2e528b49 | 636 | return (SET_ERROR(EINVAL)); |
428870ff BB |
637 | } |
638 | ||
a448a255 SD |
639 | int max_blksz = zfsvfs->z_max_blksz; |
640 | xuio_t *xuio = NULL; | |
641 | ||
34dc7c2f BB |
642 | /* |
643 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
644 | * don't hold up txg. | |
428870ff | 645 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 646 | */ |
9cac042c | 647 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
648 | if ((uio->uio_extflg == UIO_XUIO) && |
649 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
650 | xuio = (xuio_t *)uio; | |
651 | else | |
9cac042c | 652 | #endif |
779a6c0b | 653 | if (uio_prefaultpages(MIN(n, max_blksz), uio)) { |
654 | ZFS_EXIT(zfsvfs); | |
655 | return (SET_ERROR(EFAULT)); | |
656 | } | |
34dc7c2f BB |
657 | |
658 | /* | |
659 | * If in append mode, set the io offset pointer to eof. | |
660 | */ | |
5d43cc9a | 661 | locked_range_t *lr; |
34dc7c2f BB |
662 | if (ioflag & FAPPEND) { |
663 | /* | |
428870ff BB |
664 | * Obtain an appending range lock to guarantee file append |
665 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f | 666 | */ |
5d43cc9a MA |
667 | lr = rangelock_enter(&zp->z_rangelock, 0, n, RL_APPEND); |
668 | woff = lr->lr_offset; | |
669 | if (lr->lr_length == UINT64_MAX) { | |
428870ff BB |
670 | /* |
671 | * We overlocked the file because this write will cause | |
672 | * the file block size to increase. | |
673 | * Note that zp_size cannot change with this lock held. | |
674 | */ | |
675 | woff = zp->z_size; | |
34dc7c2f | 676 | } |
428870ff | 677 | uio->uio_loffset = woff; |
34dc7c2f | 678 | } else { |
34dc7c2f | 679 | /* |
428870ff BB |
680 | * Note that if the file block size will change as a result of |
681 | * this write, then this range lock will lock the entire file | |
682 | * so that we can re-write the block safely. | |
34dc7c2f | 683 | */ |
5d43cc9a | 684 | lr = rangelock_enter(&zp->z_rangelock, woff, n, RL_WRITER); |
34dc7c2f BB |
685 | } |
686 | ||
687 | if (woff >= limit) { | |
5d43cc9a | 688 | rangelock_exit(lr); |
0037b49e | 689 | ZFS_EXIT(zfsvfs); |
2e528b49 | 690 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
691 | } |
692 | ||
693 | if ((woff + n) > limit || woff > (limit - n)) | |
694 | n = limit - woff; | |
695 | ||
428870ff | 696 | /* Will this write extend the file length? */ |
a448a255 SD |
697 | int write_eof = (woff + n > zp->z_size); |
698 | ||
699 | uint64_t end_size = MAX(zp->z_size, woff + n); | |
700 | zilog_t *zilog = zfsvfs->z_log; | |
701 | #ifdef HAVE_UIO_ZEROCOPY | |
702 | int i_iov = 0; | |
703 | const iovec_t *iovp = uio->uio_iov; | |
704 | ASSERTV(int iovcnt = uio->uio_iovcnt); | |
705 | #endif | |
428870ff | 706 | |
34dc7c2f BB |
707 | |
708 | /* | |
709 | * Write the file in reasonable size chunks. Each chunk is written | |
710 | * in a separate transaction; this keeps the intent log records small | |
711 | * and allows us to do more fine-grained space accounting. | |
712 | */ | |
713 | while (n > 0) { | |
9babb374 | 714 | woff = uio->uio_loffset; |
a448a255 | 715 | |
9c5167d1 NF |
716 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
717 | KUID_TO_SUID(ip->i_uid)) || | |
718 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
719 | KGID_TO_SGID(ip->i_gid)) || | |
720 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
721 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
722 | zp->z_projid))) { | |
2e528b49 | 723 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
724 | break; |
725 | } | |
726 | ||
a448a255 SD |
727 | arc_buf_t *abuf = NULL; |
728 | const iovec_t *aiov = NULL; | |
729 | if (xuio) { | |
5a6765cf | 730 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 731 | ASSERT(i_iov < iovcnt); |
5475aada | 732 | ASSERT3U(uio->uio_segflg, !=, UIO_BVEC); |
428870ff BB |
733 | aiov = &iovp[i_iov]; |
734 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
735 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
736 | ASSERT((aiov->iov_base == abuf->b_data) || |
737 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
738 | aiov->iov_len == arc_buf_size(abuf))); | |
739 | i_iov++; | |
5a6765cf | 740 | #endif |
a448a255 | 741 | } else if (n >= max_blksz && woff >= zp->z_size && |
9babb374 BB |
742 | P2PHASE(woff, max_blksz) == 0 && |
743 | zp->z_blksz == max_blksz) { | |
428870ff BB |
744 | /* |
745 | * This write covers a full block. "Borrow" a buffer | |
746 | * from the dmu so that we can fill it before we enter | |
747 | * a transaction. This avoids the possibility of | |
748 | * holding up the transaction if the data copy hangs | |
749 | * up on a pagefault (e.g., from an NFS server mapping). | |
750 | */ | |
9babb374 BB |
751 | size_t cbytes; |
752 | ||
428870ff BB |
753 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
754 | max_blksz); | |
9babb374 BB |
755 | ASSERT(abuf != NULL); |
756 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
757 | if ((error = uiocopy(abuf->b_data, max_blksz, |
758 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
759 | dmu_return_arcbuf(abuf); |
760 | break; | |
761 | } | |
762 | ASSERT(cbytes == max_blksz); | |
763 | } | |
764 | ||
34dc7c2f BB |
765 | /* |
766 | * Start a transaction. | |
767 | */ | |
a448a255 | 768 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 769 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 770 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 771 | zfs_sa_upgrade_txholds(tx, zp); |
384f8a09 | 772 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 773 | if (error) { |
34dc7c2f | 774 | dmu_tx_abort(tx); |
9babb374 BB |
775 | if (abuf != NULL) |
776 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
777 | break; |
778 | } | |
779 | ||
780 | /* | |
5d43cc9a | 781 | * If rangelock_enter() over-locked we grow the blocksize |
34dc7c2f | 782 | * and then reduce the lock range. This will only happen |
5d43cc9a MA |
783 | * on the first iteration since rangelock_reduce() will |
784 | * shrink down lr_length to the appropriate size. | |
34dc7c2f | 785 | */ |
5d43cc9a | 786 | if (lr->lr_length == UINT64_MAX) { |
34dc7c2f BB |
787 | uint64_t new_blksz; |
788 | ||
789 | if (zp->z_blksz > max_blksz) { | |
f1512ee6 MA |
790 | /* |
791 | * File's blocksize is already larger than the | |
792 | * "recordsize" property. Only let it grow to | |
793 | * the next power of 2. | |
794 | */ | |
34dc7c2f | 795 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 MA |
796 | new_blksz = MIN(end_size, |
797 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
798 | } else { |
799 | new_blksz = MIN(end_size, max_blksz); | |
800 | } | |
801 | zfs_grow_blocksize(zp, new_blksz, tx); | |
5d43cc9a | 802 | rangelock_reduce(lr, woff, n); |
34dc7c2f BB |
803 | } |
804 | ||
805 | /* | |
806 | * XXX - should we really limit each write to z_max_blksz? | |
807 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
808 | */ | |
a448a255 | 809 | ssize_t nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); |
34dc7c2f | 810 | |
a448a255 | 811 | ssize_t tx_bytes; |
9babb374 BB |
812 | if (abuf == NULL) { |
813 | tx_bytes = uio->uio_resid; | |
779a6c0b | 814 | uio->uio_fault_disable = B_TRUE; |
428870ff BB |
815 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
816 | uio, nbytes, tx); | |
779a6c0b | 817 | if (error == EFAULT) { |
818 | dmu_tx_commit(tx); | |
819 | if (uio_prefaultpages(MIN(n, max_blksz), uio)) { | |
820 | break; | |
821 | } | |
822 | continue; | |
823 | } else if (error != 0) { | |
824 | dmu_tx_commit(tx); | |
825 | break; | |
826 | } | |
9babb374 BB |
827 | tx_bytes -= uio->uio_resid; |
828 | } else { | |
829 | tx_bytes = nbytes; | |
428870ff BB |
830 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
831 | /* | |
832 | * If this is not a full block write, but we are | |
833 | * extending the file past EOF and this data starts | |
834 | * block-aligned, use assign_arcbuf(). Otherwise, | |
835 | * write via dmu_write(). | |
836 | */ | |
837 | if (tx_bytes < max_blksz && (!write_eof || | |
838 | aiov->iov_base != abuf->b_data)) { | |
839 | ASSERT(xuio); | |
0037b49e | 840 | dmu_write(zfsvfs->z_os, zp->z_id, woff, |
94183a9d | 841 | /* cppcheck-suppress nullPointer */ |
428870ff BB |
842 | aiov->iov_len, aiov->iov_base, tx); |
843 | dmu_return_arcbuf(abuf); | |
844 | xuio_stat_wbuf_copied(); | |
845 | } else { | |
846 | ASSERT(xuio || tx_bytes == max_blksz); | |
305781da | 847 | error = dmu_assign_arcbuf_by_dbuf( |
440a3eb9 | 848 | sa_get_db(zp->z_sa_hdl), woff, abuf, tx); |
305781da TC |
849 | if (error != 0) { |
850 | dmu_return_arcbuf(abuf); | |
851 | dmu_tx_commit(tx); | |
852 | break; | |
853 | } | |
428870ff | 854 | } |
9babb374 BB |
855 | ASSERT(tx_bytes <= uio->uio_resid); |
856 | uioskip(uio, tx_bytes); | |
857 | } | |
0037b49e BB |
858 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
859 | update_pages(ip, woff, | |
860 | tx_bytes, zfsvfs->z_os, zp->z_id); | |
861 | } | |
34dc7c2f BB |
862 | |
863 | /* | |
864 | * If we made no progress, we're done. If we made even | |
865 | * partial progress, update the znode and ZIL accordingly. | |
866 | */ | |
867 | if (tx_bytes == 0) { | |
0037b49e | 868 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 869 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
870 | dmu_tx_commit(tx); |
871 | ASSERT(error != 0); | |
872 | break; | |
873 | } | |
874 | ||
875 | /* | |
876 | * Clear Set-UID/Set-GID bits on successful write if not | |
4e33ba4c | 877 | * privileged and at least one of the execute bits is set. |
34dc7c2f BB |
878 | * |
879 | * It would be nice to to this after all writes have | |
880 | * been done, but that would still expose the ISUID/ISGID | |
881 | * to another app after the partial write is committed. | |
882 | * | |
572e2857 BB |
883 | * Note: we don't call zfs_fuid_map_id() here because |
884 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
885 | */ |
886 | mutex_enter(&zp->z_acl_lock); | |
a448a255 | 887 | uint32_t uid = KUID_TO_SUID(ip->i_uid); |
428870ff | 888 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 889 | (S_IXUSR >> 6))) != 0 && |
428870ff | 890 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 891 | secpolicy_vnode_setid_retain(cr, |
2c6abf15 | 892 | ((zp->z_mode & S_ISUID) != 0 && uid == 0)) != 0) { |
428870ff BB |
893 | uint64_t newmode; |
894 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
12fa7f34 | 895 | ip->i_mode = newmode = zp->z_mode; |
0037b49e | 896 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), |
428870ff | 897 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
898 | } |
899 | mutex_exit(&zp->z_acl_lock); | |
900 | ||
0df9673f | 901 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
34dc7c2f BB |
902 | |
903 | /* | |
904 | * Update the file size (zp_size) if it has changed; | |
905 | * account for possible concurrent updates. | |
906 | */ | |
428870ff BB |
907 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
908 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 909 | uio->uio_loffset); |
428870ff BB |
910 | ASSERT(error == 0); |
911 | } | |
572e2857 BB |
912 | /* |
913 | * If we are replaying and eof is non zero then force | |
914 | * the file size to the specified eof. Note, there's no | |
915 | * concurrency during replay. | |
916 | */ | |
0037b49e BB |
917 | if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) |
918 | zp->z_size = zfsvfs->z_replay_eof; | |
572e2857 | 919 | |
428870ff BB |
920 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
921 | ||
119a394a ED |
922 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
923 | NULL, NULL); | |
34dc7c2f BB |
924 | dmu_tx_commit(tx); |
925 | ||
926 | if (error != 0) | |
927 | break; | |
928 | ASSERT(tx_bytes == nbytes); | |
929 | n -= nbytes; | |
572e2857 | 930 | |
779a6c0b | 931 | if (!xuio && n > 0) { |
932 | if (uio_prefaultpages(MIN(n, max_blksz), uio)) { | |
933 | error = EFAULT; | |
934 | break; | |
935 | } | |
936 | } | |
34dc7c2f BB |
937 | } |
938 | ||
2a53e2da | 939 | zfs_inode_update(zp); |
5d43cc9a | 940 | rangelock_exit(lr); |
34dc7c2f BB |
941 | |
942 | /* | |
943 | * If we're in replay mode, or we made no progress, return error. | |
944 | * Otherwise, it's at least a partial write, so it's successful. | |
945 | */ | |
0037b49e BB |
946 | if (zfsvfs->z_replay || uio->uio_resid == start_resid) { |
947 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
948 | return (error); |
949 | } | |
950 | ||
428870ff | 951 | if (ioflag & (FSYNC | FDSYNC) || |
0037b49e | 952 | zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 953 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 954 | |
a448a255 SD |
955 | int64_t nwritten = start_resid - uio->uio_resid; |
956 | dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, nwritten); | |
957 | task_io_account_write(nwritten); | |
958 | ||
0037b49e | 959 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
960 | return (0); |
961 | } | |
962 | ||
7c502b0b | 963 | /* |
964 | * Drop a reference on the passed inode asynchronously. This ensures | |
965 | * that the caller will never drop the last reference on an inode in | |
966 | * the current context. Doing so while holding open a tx could result | |
967 | * in a deadlock if iput_final() re-enters the filesystem code. | |
968 | */ | |
0a50679c BB |
969 | void |
970 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 971 | { |
0a50679c BB |
972 | objset_t *os = ITOZSB(ip)->z_os; |
973 | ||
3558fd73 | 974 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
975 | ASSERT(os != NULL); |
976 | ||
3558fd73 | 977 | if (atomic_read(&ip->i_count) == 1) |
7c502b0b | 978 | VERIFY(taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
48d3eb40 | 979 | (task_func_t *)iput, ip, TQ_SLEEP) != TASKQID_INVALID); |
3558fd73 BB |
980 | else |
981 | iput(ip); | |
982 | } | |
983 | ||
900d09b2 | 984 | /* ARGSUSED */ |
34dc7c2f | 985 | void |
428870ff | 986 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 987 | { |
428870ff | 988 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
989 | |
990 | if (zgd->zgd_db) | |
991 | dmu_buf_rele(zgd->zgd_db, zgd); | |
992 | ||
5d43cc9a | 993 | rangelock_exit(zgd->zgd_lr); |
34dc7c2f | 994 | |
9babb374 BB |
995 | /* |
996 | * Release the vnode asynchronously as we currently have the | |
997 | * txg stopped from syncing. | |
998 | */ | |
0a50679c | 999 | zfs_iput_async(ZTOI(zp)); |
428870ff | 1000 | |
34dc7c2f BB |
1001 | kmem_free(zgd, sizeof (zgd_t)); |
1002 | } | |
1003 | ||
45d1cae3 BB |
1004 | #ifdef DEBUG |
1005 | static int zil_fault_io = 0; | |
1006 | #endif | |
1007 | ||
34dc7c2f BB |
1008 | /* |
1009 | * Get data to generate a TX_WRITE intent log record. | |
1010 | */ | |
1011 | int | |
1ce23dca | 1012 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, zio_t *zio) |
34dc7c2f | 1013 | { |
0037b49e BB |
1014 | zfsvfs_t *zfsvfs = arg; |
1015 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 1016 | znode_t *zp; |
428870ff BB |
1017 | uint64_t object = lr->lr_foid; |
1018 | uint64_t offset = lr->lr_offset; | |
1019 | uint64_t size = lr->lr_length; | |
34dc7c2f | 1020 | dmu_buf_t *db; |
34dc7c2f | 1021 | zgd_t *zgd; |
34dc7c2f BB |
1022 | int error = 0; |
1023 | ||
1ce23dca PS |
1024 | ASSERT3P(lwb, !=, NULL); |
1025 | ASSERT3P(zio, !=, NULL); | |
1026 | ASSERT3U(size, !=, 0); | |
34dc7c2f BB |
1027 | |
1028 | /* | |
1029 | * Nothing to do if the file has been removed | |
1030 | */ | |
0037b49e | 1031 | if (zfs_zget(zfsvfs, object, &zp) != 0) |
2e528b49 | 1032 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 1033 | if (zp->z_unlinked) { |
9babb374 BB |
1034 | /* |
1035 | * Release the vnode asynchronously as we currently have the | |
1036 | * txg stopped from syncing. | |
1037 | */ | |
0a50679c | 1038 | zfs_iput_async(ZTOI(zp)); |
2e528b49 | 1039 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1040 | } |
1041 | ||
79c76d5b | 1042 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
1ce23dca | 1043 | zgd->zgd_lwb = lwb; |
428870ff BB |
1044 | zgd->zgd_private = zp; |
1045 | ||
34dc7c2f BB |
1046 | /* |
1047 | * Write records come in two flavors: immediate and indirect. | |
1048 | * For small writes it's cheaper to store the data with the | |
1049 | * log record (immediate); for large writes it's cheaper to | |
1050 | * sync the data and get a pointer to it (indirect) so that | |
1051 | * we don't have to write the data twice. | |
1052 | */ | |
1053 | if (buf != NULL) { /* immediate write */ | |
5d43cc9a MA |
1054 | zgd->zgd_lr = rangelock_enter(&zp->z_rangelock, |
1055 | offset, size, RL_READER); | |
34dc7c2f | 1056 | /* test for truncation needs to be done while range locked */ |
428870ff | 1057 | if (offset >= zp->z_size) { |
2e528b49 | 1058 | error = SET_ERROR(ENOENT); |
428870ff BB |
1059 | } else { |
1060 | error = dmu_read(os, object, offset, size, buf, | |
1061 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1062 | } |
428870ff | 1063 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1064 | } else { /* indirect write */ |
34dc7c2f BB |
1065 | /* |
1066 | * Have to lock the whole block to ensure when it's | |
f763c3d1 | 1067 | * written out and its checksum is being calculated |
34dc7c2f BB |
1068 | * that no one can change the data. We need to re-check |
1069 | * blocksize after we get the lock in case it's changed! | |
1070 | */ | |
1071 | for (;;) { | |
428870ff BB |
1072 | uint64_t blkoff; |
1073 | size = zp->z_blksz; | |
1074 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1075 | offset -= blkoff; | |
5d43cc9a MA |
1076 | zgd->zgd_lr = rangelock_enter(&zp->z_rangelock, |
1077 | offset, size, RL_READER); | |
428870ff | 1078 | if (zp->z_blksz == size) |
34dc7c2f | 1079 | break; |
428870ff | 1080 | offset += blkoff; |
5d43cc9a | 1081 | rangelock_exit(zgd->zgd_lr); |
34dc7c2f BB |
1082 | } |
1083 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1084 | if (lr->lr_offset >= zp->z_size) |
2e528b49 | 1085 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1086 | #ifdef DEBUG |
1087 | if (zil_fault_io) { | |
2e528b49 | 1088 | error = SET_ERROR(EIO); |
45d1cae3 | 1089 | zil_fault_io = 0; |
45d1cae3 | 1090 | } |
45d1cae3 | 1091 | #endif |
34dc7c2f | 1092 | if (error == 0) |
428870ff BB |
1093 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1094 | DMU_READ_NO_PREFETCH); | |
1095 | ||
1096 | if (error == 0) { | |
02dc43bc | 1097 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 1098 | |
428870ff BB |
1099 | zgd->zgd_db = db; |
1100 | zgd->zgd_bp = bp; | |
1101 | ||
1102 | ASSERT(db->db_offset == offset); | |
1103 | ASSERT(db->db_size == size); | |
1104 | ||
1105 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1106 | zfs_get_done, zgd); | |
61ca48ff | 1107 | ASSERT(error || lr->lr_length <= size); |
428870ff BB |
1108 | |
1109 | /* | |
1110 | * On success, we need to wait for the write I/O | |
1111 | * initiated by dmu_sync() to complete before we can | |
1112 | * release this dbuf. We will finish everything up | |
1113 | * in the zfs_get_done() callback. | |
1114 | */ | |
1115 | if (error == 0) | |
1116 | return (0); | |
1117 | ||
1118 | if (error == EALREADY) { | |
1119 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
a1d477c2 MA |
1120 | /* |
1121 | * TX_WRITE2 relies on the data previously | |
1122 | * written by the TX_WRITE that caused | |
1123 | * EALREADY. We zero out the BP because | |
900d09b2 | 1124 | * it is the old, currently-on-disk BP. |
a1d477c2 MA |
1125 | */ |
1126 | zgd->zgd_bp = NULL; | |
1127 | BP_ZERO(bp); | |
428870ff BB |
1128 | error = 0; |
1129 | } | |
1130 | } | |
34dc7c2f | 1131 | } |
428870ff BB |
1132 | |
1133 | zfs_get_done(zgd, error); | |
1134 | ||
34dc7c2f BB |
1135 | return (error); |
1136 | } | |
1137 | ||
1138 | /*ARGSUSED*/ | |
3558fd73 BB |
1139 | int |
1140 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1141 | { |
3558fd73 | 1142 | znode_t *zp = ITOZ(ip); |
0037b49e | 1143 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
1144 | int error; |
1145 | ||
0037b49e | 1146 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1147 | ZFS_VERIFY_ZP(zp); |
1148 | ||
1149 | if (flag & V_ACE_MASK) | |
1150 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1151 | else | |
1152 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1153 | ||
0037b49e | 1154 | ZFS_EXIT(zfsvfs); |
45d1cae3 BB |
1155 | return (error); |
1156 | } | |
45d1cae3 | 1157 | |
34dc7c2f BB |
1158 | /* |
1159 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1160 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1161 | * |
3558fd73 | 1162 | * IN: dip - inode of directory to search. |
34dc7c2f | 1163 | * nm - name of entry to lookup. |
34dc7c2f | 1164 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1165 | * cr - credentials of caller. |
34dc7c2f BB |
1166 | * direntflags - directory lookup flags |
1167 | * realpnp - returned pathname. | |
1168 | * | |
3558fd73 | 1169 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1170 | * |
d3cc8b15 | 1171 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1172 | * |
1173 | * Timestamps: | |
1174 | * NA | |
1175 | */ | |
1176 | /* ARGSUSED */ | |
e5c39b95 | 1177 | int |
3558fd73 BB |
1178 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1179 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1180 | { |
3558fd73 | 1181 | znode_t *zdp = ITOZ(dip); |
0037b49e | 1182 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
3558fd73 | 1183 | int error = 0; |
45d1cae3 | 1184 | |
9b7b9cd3 GM |
1185 | /* |
1186 | * Fast path lookup, however we must skip DNLC lookup | |
1187 | * for case folding or normalizing lookups because the | |
1188 | * DNLC code only stores the passed in name. This means | |
1189 | * creating 'a' and removing 'A' on a case insensitive | |
1190 | * file system would work, but DNLC still thinks 'a' | |
1191 | * exists and won't let you create it again on the next | |
1192 | * pass through fast path. | |
1193 | */ | |
45d1cae3 BB |
1194 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { |
1195 | ||
3558fd73 | 1196 | if (!S_ISDIR(dip->i_mode)) { |
2e528b49 | 1197 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1198 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 1199 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1200 | } |
1201 | ||
1202 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1203 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1204 | if (!error) { | |
3558fd73 BB |
1205 | *ipp = dip; |
1206 | igrab(*ipp); | |
45d1cae3 BB |
1207 | return (0); |
1208 | } | |
1209 | return (error); | |
45d1cae3 BB |
1210 | } |
1211 | } | |
1212 | ||
0037b49e | 1213 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1214 | ZFS_VERIFY_ZP(zdp); |
1215 | ||
3558fd73 | 1216 | *ipp = NULL; |
34dc7c2f BB |
1217 | |
1218 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1219 | /* |
1220 | * We don't allow recursive attributes.. | |
1221 | * Maybe someday we will. | |
1222 | */ | |
428870ff | 1223 | if (zdp->z_pflags & ZFS_XATTR) { |
0037b49e | 1224 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1225 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1226 | } |
1227 | ||
3558fd73 | 1228 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
0037b49e | 1229 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1230 | return (error); |
1231 | } | |
1232 | ||
1233 | /* | |
1234 | * Do we have permission to get into attribute directory? | |
1235 | */ | |
1236 | ||
3558fd73 | 1237 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1238 | B_FALSE, cr))) { |
3558fd73 BB |
1239 | iput(*ipp); |
1240 | *ipp = NULL; | |
34dc7c2f BB |
1241 | } |
1242 | ||
0037b49e | 1243 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1244 | return (error); |
1245 | } | |
1246 | ||
3558fd73 | 1247 | if (!S_ISDIR(dip->i_mode)) { |
0037b49e | 1248 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1249 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1250 | } |
1251 | ||
1252 | /* | |
1253 | * Check accessibility of directory. | |
1254 | */ | |
1255 | ||
149e873a | 1256 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
0037b49e | 1257 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1258 | return (error); |
1259 | } | |
1260 | ||
0037b49e | 1261 | if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1262 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1263 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1264 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1265 | } |
1266 | ||
3558fd73 BB |
1267 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1268 | if ((error == 0) && (*ipp)) | |
1269 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1270 | |
0037b49e | 1271 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1272 | return (error); |
1273 | } | |
1274 | ||
1275 | /* | |
1276 | * Attempt to create a new entry in a directory. If the entry | |
1277 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1278 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1279 | * |
3558fd73 | 1280 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1281 | * name - name of new file entry. |
1282 | * vap - attributes of new file. | |
1283 | * excl - flag indicating exclusive or non-exclusive mode. | |
1284 | * mode - mode to open file with. | |
1285 | * cr - credentials of caller. | |
1286 | * flag - large file flag [UNUSED]. | |
3558fd73 | 1287 | * vsecp - ACL to be set |
34dc7c2f | 1288 | * |
3558fd73 | 1289 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1290 | * |
d3cc8b15 | 1291 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1292 | * |
1293 | * Timestamps: | |
3558fd73 BB |
1294 | * dip - ctime|mtime updated if new entry created |
1295 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1296 | */ |
1297 | ||
1298 | /* ARGSUSED */ | |
e5c39b95 | 1299 | int |
3558fd73 BB |
1300 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1301 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1302 | { |
3558fd73 | 1303 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1304 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1305 | zilog_t *zilog; |
1306 | objset_t *os; | |
1307 | zfs_dirlock_t *dl; | |
1308 | dmu_tx_t *tx; | |
1309 | int error; | |
b128c09f | 1310 | uid_t uid; |
149e873a | 1311 | gid_t gid; |
428870ff | 1312 | zfs_acl_ids_t acl_ids; |
9babb374 | 1313 | boolean_t fuid_dirtied; |
428870ff | 1314 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 1315 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1316 | |
1317 | /* | |
1318 | * If we have an ephemeral id, ACL, or XVATTR then | |
1319 | * make sure file system is at proper version | |
1320 | */ | |
1321 | ||
149e873a | 1322 | gid = crgetgid(cr); |
3558fd73 | 1323 | uid = crgetuid(cr); |
b128c09f | 1324 | |
0037b49e | 1325 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1326 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1327 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1328 | |
32dec7bd | 1329 | if (name == NULL) |
1330 | return (SET_ERROR(EINVAL)); | |
1331 | ||
0037b49e | 1332 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1333 | ZFS_VERIFY_ZP(dzp); |
0037b49e BB |
1334 | os = zfsvfs->z_os; |
1335 | zilog = zfsvfs->z_log; | |
34dc7c2f | 1336 | |
0037b49e | 1337 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1338 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1339 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1340 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1341 | } |
1342 | ||
5484965a | 1343 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1344 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1345 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1346 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1347 | return (error); |
1348 | } | |
1349 | } | |
34dc7c2f | 1350 | |
3558fd73 BB |
1351 | top: |
1352 | *ipp = NULL; | |
34dc7c2f BB |
1353 | if (*name == '\0') { |
1354 | /* | |
1355 | * Null component name refers to the directory itself. | |
1356 | */ | |
3558fd73 | 1357 | igrab(dip); |
34dc7c2f BB |
1358 | zp = dzp; |
1359 | dl = NULL; | |
1360 | error = 0; | |
1361 | } else { | |
3558fd73 | 1362 | /* possible igrab(zp) */ |
34dc7c2f BB |
1363 | int zflg = 0; |
1364 | ||
1365 | if (flag & FIGNORECASE) | |
1366 | zflg |= ZCILOOK; | |
1367 | ||
1368 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1369 | NULL, NULL); | |
1370 | if (error) { | |
572e2857 BB |
1371 | if (have_acl) |
1372 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1373 | if (strcmp(name, "..") == 0) |
2e528b49 | 1374 | error = SET_ERROR(EISDIR); |
0037b49e | 1375 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1376 | return (error); |
1377 | } | |
1378 | } | |
428870ff | 1379 | |
34dc7c2f BB |
1380 | if (zp == NULL) { |
1381 | uint64_t txtype; | |
9c5167d1 | 1382 | uint64_t projid = ZFS_DEFAULT_PROJID; |
34dc7c2f BB |
1383 | |
1384 | /* | |
1385 | * Create a new file object and update the directory | |
1386 | * to reference it. | |
1387 | */ | |
149e873a | 1388 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1389 | if (have_acl) |
1390 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1391 | goto out; |
1392 | } | |
1393 | ||
1394 | /* | |
1395 | * We only support the creation of regular files in | |
1396 | * extended attribute directories. | |
1397 | */ | |
428870ff | 1398 | |
3558fd73 | 1399 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1400 | if (have_acl) |
1401 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 1402 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1403 | goto out; |
1404 | } | |
1405 | ||
428870ff BB |
1406 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1407 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1408 | goto out; |
428870ff BB |
1409 | have_acl = B_TRUE; |
1410 | ||
9c5167d1 NF |
1411 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
1412 | projid = zfs_inherit_projid(dzp); | |
1413 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
45d1cae3 | 1414 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1415 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1416 | goto out; |
1417 | } | |
1418 | ||
34dc7c2f | 1419 | tx = dmu_tx_create(os); |
428870ff BB |
1420 | |
1421 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1422 | ZFS_SA_BASE_ATTR_SIZE); | |
1423 | ||
0037b49e | 1424 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1425 | if (fuid_dirtied) |
0037b49e | 1426 | zfs_fuid_txhold(zfsvfs, tx); |
34dc7c2f | 1427 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1428 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
0037b49e | 1429 | if (!zfsvfs->z_use_sa && |
428870ff | 1430 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1431 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1432 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1433 | } |
599b8648 | 1434 | |
0735ecb3 PS |
1435 | error = dmu_tx_assign(tx, |
1436 | (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); | |
34dc7c2f BB |
1437 | if (error) { |
1438 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1439 | if (error == ERESTART) { |
e8b96c60 | 1440 | waited = B_TRUE; |
34dc7c2f BB |
1441 | dmu_tx_wait(tx); |
1442 | dmu_tx_abort(tx); | |
1443 | goto top; | |
1444 | } | |
428870ff | 1445 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1446 | dmu_tx_abort(tx); |
0037b49e | 1447 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1448 | return (error); |
1449 | } | |
428870ff | 1450 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 1451 | |
599b8648 CC |
1452 | error = zfs_link_create(dl, zp, tx, ZNEW); |
1453 | if (error != 0) { | |
1454 | /* | |
1455 | * Since, we failed to add the directory entry for it, | |
1456 | * delete the newly created dnode. | |
1457 | */ | |
1458 | zfs_znode_delete(zp, tx); | |
1459 | remove_inode_hash(ZTOI(zp)); | |
1460 | zfs_acl_ids_free(&acl_ids); | |
1461 | dmu_tx_commit(tx); | |
1462 | goto out; | |
1463 | } | |
1464 | ||
9babb374 | 1465 | if (fuid_dirtied) |
0037b49e | 1466 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 1467 | |
34dc7c2f BB |
1468 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); |
1469 | if (flag & FIGNORECASE) | |
1470 | txtype |= TX_CI; | |
1471 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1472 | vsecp, acl_ids.z_fuidp, vap); |
1473 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1474 | dmu_tx_commit(tx); |
1475 | } else { | |
1476 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1477 | ||
572e2857 BB |
1478 | if (have_acl) |
1479 | zfs_acl_ids_free(&acl_ids); | |
1480 | have_acl = B_FALSE; | |
1481 | ||
34dc7c2f BB |
1482 | /* |
1483 | * A directory entry already exists for this name. | |
1484 | */ | |
1485 | /* | |
1486 | * Can't truncate an existing file if in exclusive mode. | |
1487 | */ | |
3558fd73 | 1488 | if (excl) { |
2e528b49 | 1489 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1490 | goto out; |
1491 | } | |
1492 | /* | |
1493 | * Can't open a directory for writing. | |
1494 | */ | |
3558fd73 | 1495 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1496 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1497 | goto out; |
1498 | } | |
1499 | /* | |
1500 | * Verify requested access to file. | |
1501 | */ | |
1502 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1503 | goto out; | |
1504 | } | |
1505 | ||
1506 | mutex_enter(&dzp->z_lock); | |
1507 | dzp->z_seq++; | |
1508 | mutex_exit(&dzp->z_lock); | |
1509 | ||
1510 | /* | |
1511 | * Truncate regular files if requested. | |
1512 | */ | |
3558fd73 BB |
1513 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1514 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 1515 | /* we can't hold any locks when calling zfs_freesp() */ |
609603a5 B |
1516 | if (dl) { |
1517 | zfs_dirent_unlock(dl); | |
1518 | dl = NULL; | |
1519 | } | |
34dc7c2f | 1520 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1521 | } |
1522 | } | |
1523 | out: | |
1524 | ||
1525 | if (dl) | |
1526 | zfs_dirent_unlock(dl); | |
1527 | ||
1528 | if (error) { | |
1529 | if (zp) | |
3558fd73 | 1530 | iput(ZTOI(zp)); |
34dc7c2f | 1531 | } else { |
960e08fe BB |
1532 | zfs_inode_update(dzp); |
1533 | zfs_inode_update(zp); | |
3558fd73 | 1534 | *ipp = ZTOI(zp); |
34dc7c2f | 1535 | } |
34dc7c2f | 1536 | |
0037b49e | 1537 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1538 | zil_commit(zilog, 0); |
428870ff | 1539 | |
0037b49e | 1540 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1541 | return (error); |
1542 | } | |
1543 | ||
ace1eae8 CC |
1544 | /* ARGSUSED */ |
1545 | int | |
1546 | zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, | |
1547 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
1548 | { | |
1549 | znode_t *zp = NULL, *dzp = ITOZ(dip); | |
0037b49e | 1550 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
ace1eae8 CC |
1551 | objset_t *os; |
1552 | dmu_tx_t *tx; | |
1553 | int error; | |
1554 | uid_t uid; | |
1555 | gid_t gid; | |
1556 | zfs_acl_ids_t acl_ids; | |
9c5167d1 | 1557 | uint64_t projid = ZFS_DEFAULT_PROJID; |
ace1eae8 CC |
1558 | boolean_t fuid_dirtied; |
1559 | boolean_t have_acl = B_FALSE; | |
1560 | boolean_t waited = B_FALSE; | |
1561 | ||
1562 | /* | |
1563 | * If we have an ephemeral id, ACL, or XVATTR then | |
1564 | * make sure file system is at proper version | |
1565 | */ | |
1566 | ||
1567 | gid = crgetgid(cr); | |
1568 | uid = crgetuid(cr); | |
1569 | ||
0037b49e | 1570 | if (zfsvfs->z_use_fuids == B_FALSE && |
ace1eae8 CC |
1571 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
1572 | return (SET_ERROR(EINVAL)); | |
1573 | ||
0037b49e | 1574 | ZFS_ENTER(zfsvfs); |
ace1eae8 | 1575 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1576 | os = zfsvfs->z_os; |
ace1eae8 CC |
1577 | |
1578 | if (vap->va_mask & ATTR_XVATTR) { | |
1579 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
1580 | crgetuid(cr), cr, vap->va_mode)) != 0) { | |
0037b49e | 1581 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1582 | return (error); |
1583 | } | |
1584 | } | |
1585 | ||
1586 | top: | |
1587 | *ipp = NULL; | |
1588 | ||
1589 | /* | |
1590 | * Create a new file object and update the directory | |
1591 | * to reference it. | |
1592 | */ | |
1593 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { | |
1594 | if (have_acl) | |
1595 | zfs_acl_ids_free(&acl_ids); | |
1596 | goto out; | |
1597 | } | |
1598 | ||
1599 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, | |
1600 | cr, vsecp, &acl_ids)) != 0) | |
1601 | goto out; | |
1602 | have_acl = B_TRUE; | |
1603 | ||
9c5167d1 NF |
1604 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
1605 | projid = zfs_inherit_projid(dzp); | |
1606 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
ace1eae8 CC |
1607 | zfs_acl_ids_free(&acl_ids); |
1608 | error = SET_ERROR(EDQUOT); | |
1609 | goto out; | |
1610 | } | |
1611 | ||
1612 | tx = dmu_tx_create(os); | |
1613 | ||
1614 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1615 | ZFS_SA_BASE_ATTR_SIZE); | |
0037b49e | 1616 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 1617 | |
0037b49e | 1618 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
ace1eae8 | 1619 | if (fuid_dirtied) |
0037b49e BB |
1620 | zfs_fuid_txhold(zfsvfs, tx); |
1621 | if (!zfsvfs->z_use_sa && | |
ace1eae8 CC |
1622 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
1623 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
1624 | 0, acl_ids.z_aclp->z_acl_bytes); | |
1625 | } | |
0735ecb3 | 1626 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
ace1eae8 CC |
1627 | if (error) { |
1628 | if (error == ERESTART) { | |
1629 | waited = B_TRUE; | |
1630 | dmu_tx_wait(tx); | |
1631 | dmu_tx_abort(tx); | |
1632 | goto top; | |
1633 | } | |
1634 | zfs_acl_ids_free(&acl_ids); | |
1635 | dmu_tx_abort(tx); | |
0037b49e | 1636 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1637 | return (error); |
1638 | } | |
1639 | zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); | |
1640 | ||
1641 | if (fuid_dirtied) | |
0037b49e | 1642 | zfs_fuid_sync(zfsvfs, tx); |
ace1eae8 CC |
1643 | |
1644 | /* Add to unlinked set */ | |
1645 | zp->z_unlinked = 1; | |
1646 | zfs_unlinked_add(zp, tx); | |
1647 | zfs_acl_ids_free(&acl_ids); | |
1648 | dmu_tx_commit(tx); | |
1649 | out: | |
1650 | ||
1651 | if (error) { | |
1652 | if (zp) | |
1653 | iput(ZTOI(zp)); | |
1654 | } else { | |
1655 | zfs_inode_update(dzp); | |
1656 | zfs_inode_update(zp); | |
1657 | *ipp = ZTOI(zp); | |
1658 | } | |
1659 | ||
0037b49e | 1660 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1661 | return (error); |
1662 | } | |
1663 | ||
34dc7c2f BB |
1664 | /* |
1665 | * Remove an entry from a directory. | |
1666 | * | |
3558fd73 | 1667 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1668 | * name - name of entry to remove. |
1669 | * cr - credentials of caller. | |
34dc7c2f BB |
1670 | * |
1671 | * RETURN: 0 if success | |
1672 | * error code if failure | |
1673 | * | |
1674 | * Timestamps: | |
3558fd73 BB |
1675 | * dip - ctime|mtime |
1676 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1677 | */ |
428870ff BB |
1678 | |
1679 | uint64_t null_xattr = 0; | |
1680 | ||
34dc7c2f | 1681 | /*ARGSUSED*/ |
e5c39b95 | 1682 | int |
da5e151f | 1683 | zfs_remove(struct inode *dip, char *name, cred_t *cr, int flags) |
34dc7c2f | 1684 | { |
3558fd73 | 1685 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1686 | znode_t *xzp; |
3558fd73 | 1687 | struct inode *ip; |
0037b49e | 1688 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 1689 | zilog_t *zilog; |
a966c564 | 1690 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 1691 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 1692 | uint64_t obj = 0; |
dfbc8630 | 1693 | uint64_t links; |
34dc7c2f BB |
1694 | zfs_dirlock_t *dl; |
1695 | dmu_tx_t *tx; | |
a966c564 K |
1696 | boolean_t may_delete_now, delete_now = FALSE; |
1697 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
1698 | uint64_t txtype; |
1699 | pathname_t *realnmp = NULL; | |
1700 | pathname_t realnm; | |
1701 | int error; | |
1702 | int zflg = ZEXISTS; | |
e8b96c60 | 1703 | boolean_t waited = B_FALSE; |
34dc7c2f | 1704 | |
32dec7bd | 1705 | if (name == NULL) |
1706 | return (SET_ERROR(EINVAL)); | |
1707 | ||
0037b49e | 1708 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1709 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1710 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1711 | |
1712 | if (flags & FIGNORECASE) { | |
1713 | zflg |= ZCILOOK; | |
1714 | pn_alloc(&realnm); | |
1715 | realnmp = &realnm; | |
1716 | } | |
1717 | ||
1718 | top: | |
572e2857 BB |
1719 | xattr_obj = 0; |
1720 | xzp = NULL; | |
34dc7c2f BB |
1721 | /* |
1722 | * Attempt to lock directory; fail if entry doesn't exist. | |
1723 | */ | |
149e873a BB |
1724 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1725 | NULL, realnmp))) { | |
34dc7c2f BB |
1726 | if (realnmp) |
1727 | pn_free(realnmp); | |
0037b49e | 1728 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1729 | return (error); |
1730 | } | |
1731 | ||
3558fd73 | 1732 | ip = ZTOI(zp); |
34dc7c2f | 1733 | |
149e873a | 1734 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1735 | goto out; |
1736 | } | |
1737 | ||
1738 | /* | |
1739 | * Need to use rmdir for removing directories. | |
1740 | */ | |
3558fd73 | 1741 | if (S_ISDIR(ip->i_mode)) { |
2e528b49 | 1742 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1743 | goto out; |
1744 | } | |
1745 | ||
19d55079 MA |
1746 | mutex_enter(&zp->z_lock); |
1747 | may_delete_now = atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped); | |
1748 | mutex_exit(&zp->z_lock); | |
1749 | ||
34dc7c2f | 1750 | /* |
a966c564 K |
1751 | * We may delete the znode now, or we may put it in the unlinked set; |
1752 | * it depends on whether we're the last link, and on whether there are | |
1753 | * other holds on the inode. So we dmu_tx_hold() the right things to | |
1754 | * allow for either case. | |
34dc7c2f | 1755 | */ |
572e2857 | 1756 | obj = zp->z_id; |
0037b49e | 1757 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 1758 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1759 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1760 | zfs_sa_upgrade_txholds(tx, zp); | |
1761 | zfs_sa_upgrade_txholds(tx, dzp); | |
a966c564 K |
1762 | if (may_delete_now) { |
1763 | toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; | |
1764 | /* if the file is too big, only hold_free a token amount */ | |
1765 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1766 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1767 | } | |
34dc7c2f BB |
1768 | |
1769 | /* are there any extended attributes? */ | |
0037b49e | 1770 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1771 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1772 | if (error == 0 && xattr_obj) { |
0037b49e | 1773 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
c99c9001 | 1774 | ASSERT0(error); |
428870ff BB |
1775 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1776 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1777 | } |
1778 | ||
a966c564 K |
1779 | mutex_enter(&zp->z_lock); |
1780 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
1781 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
1782 | mutex_exit(&zp->z_lock); | |
1783 | ||
34dc7c2f | 1784 | /* charge as an update -- would be nice not to charge at all */ |
0037b49e | 1785 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1786 | |
19d55079 | 1787 | /* |
1a04bab3 | 1788 | * Mark this transaction as typically resulting in a net free of space |
19d55079 | 1789 | */ |
1a04bab3 | 1790 | dmu_tx_mark_netfree(tx); |
19d55079 | 1791 | |
0735ecb3 | 1792 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
1793 | if (error) { |
1794 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1795 | if (error == ERESTART) { |
e8b96c60 | 1796 | waited = B_TRUE; |
34dc7c2f BB |
1797 | dmu_tx_wait(tx); |
1798 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
1799 | iput(ip); |
1800 | if (xzp) | |
1801 | iput(ZTOI(xzp)); | |
34dc7c2f BB |
1802 | goto top; |
1803 | } | |
1804 | if (realnmp) | |
1805 | pn_free(realnmp); | |
1806 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
1807 | iput(ip); |
1808 | if (xzp) | |
1809 | iput(ZTOI(xzp)); | |
0037b49e | 1810 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1811 | return (error); |
1812 | } | |
1813 | ||
1814 | /* | |
1815 | * Remove the directory entry. | |
1816 | */ | |
1817 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1818 | ||
1819 | if (error) { | |
1820 | dmu_tx_commit(tx); | |
1821 | goto out; | |
1822 | } | |
1823 | ||
1824 | if (unlinked) { | |
572e2857 BB |
1825 | /* |
1826 | * Hold z_lock so that we can make sure that the ACL obj | |
1827 | * hasn't changed. Could have been deleted due to | |
1828 | * zfs_sa_upgrade(). | |
1829 | */ | |
1830 | mutex_enter(&zp->z_lock); | |
0037b49e | 1831 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1832 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
a966c564 K |
1833 | delete_now = may_delete_now && !toobig && |
1834 | atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped) && | |
1835 | xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == | |
1836 | acl_obj; | |
1837 | } | |
1838 | ||
1839 | if (delete_now) { | |
1840 | if (xattr_obj_unlinked) { | |
dfbc8630 | 1841 | ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); |
a966c564 K |
1842 | mutex_enter(&xzp->z_lock); |
1843 | xzp->z_unlinked = 1; | |
dfbc8630 CD |
1844 | clear_nlink(ZTOI(xzp)); |
1845 | links = 0; | |
0037b49e | 1846 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 1847 | &links, sizeof (links), tx); |
a966c564 K |
1848 | ASSERT3U(error, ==, 0); |
1849 | mutex_exit(&xzp->z_lock); | |
1850 | zfs_unlinked_add(xzp, tx); | |
1851 | ||
1852 | if (zp->z_is_sa) | |
1853 | error = sa_remove(zp->z_sa_hdl, | |
0037b49e | 1854 | SA_ZPL_XATTR(zfsvfs), tx); |
a966c564 K |
1855 | else |
1856 | error = sa_update(zp->z_sa_hdl, | |
0037b49e | 1857 | SA_ZPL_XATTR(zfsvfs), &null_xattr, |
a966c564 K |
1858 | sizeof (uint64_t), tx); |
1859 | ASSERT0(error); | |
1860 | } | |
1861 | /* | |
1862 | * Add to the unlinked set because a new reference could be | |
1863 | * taken concurrently resulting in a deferred destruction. | |
1864 | */ | |
1865 | zfs_unlinked_add(zp, tx); | |
1866 | mutex_exit(&zp->z_lock); | |
a966c564 | 1867 | } else if (unlinked) { |
572e2857 | 1868 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1869 | zfs_unlinked_add(zp, tx); |
1870 | } | |
1871 | ||
1872 | txtype = TX_REMOVE; | |
1873 | if (flags & FIGNORECASE) | |
1874 | txtype |= TX_CI; | |
572e2857 | 1875 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1876 | |
1877 | dmu_tx_commit(tx); | |
1878 | out: | |
1879 | if (realnmp) | |
1880 | pn_free(realnmp); | |
1881 | ||
1882 | zfs_dirent_unlock(dl); | |
960e08fe | 1883 | zfs_inode_update(dzp); |
ea7e86d8 | 1884 | zfs_inode_update(zp); |
34dc7c2f | 1885 | |
ea7e86d8 BB |
1886 | if (delete_now) |
1887 | iput(ip); | |
1888 | else | |
a966c564 | 1889 | zfs_iput_async(ip); |
a966c564 K |
1890 | |
1891 | if (xzp) { | |
1892 | zfs_inode_update(xzp); | |
1893 | zfs_iput_async(ZTOI(xzp)); | |
1894 | } | |
428870ff | 1895 | |
0037b49e | 1896 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1897 | zil_commit(zilog, 0); |
34dc7c2f | 1898 | |
0037b49e | 1899 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1900 | return (error); |
1901 | } | |
1902 | ||
1903 | /* | |
3558fd73 | 1904 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1905 | * provided. Return a pointer to the inserted directory. |
1906 | * | |
3558fd73 | 1907 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1908 | * dirname - name of new directory. |
1909 | * vap - attributes of new directory. | |
1910 | * cr - credentials of caller. | |
34dc7c2f BB |
1911 | * vsecp - ACL to be set |
1912 | * | |
3558fd73 | 1913 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1914 | * |
1915 | * RETURN: 0 if success | |
1916 | * error code if failure | |
1917 | * | |
1918 | * Timestamps: | |
3558fd73 BB |
1919 | * dip - ctime|mtime updated |
1920 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1921 | */ |
1922 | /*ARGSUSED*/ | |
e5c39b95 | 1923 | int |
3558fd73 BB |
1924 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1925 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1926 | { |
3558fd73 | 1927 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1928 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1929 | zilog_t *zilog; |
1930 | zfs_dirlock_t *dl; | |
1931 | uint64_t txtype; | |
1932 | dmu_tx_t *tx; | |
1933 | int error; | |
34dc7c2f | 1934 | int zf = ZNEW; |
b128c09f BB |
1935 | uid_t uid; |
1936 | gid_t gid = crgetgid(cr); | |
428870ff | 1937 | zfs_acl_ids_t acl_ids; |
9babb374 | 1938 | boolean_t fuid_dirtied; |
e8b96c60 | 1939 | boolean_t waited = B_FALSE; |
34dc7c2f | 1940 | |
3558fd73 | 1941 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1942 | |
1943 | /* | |
1944 | * If we have an ephemeral id, ACL, or XVATTR then | |
1945 | * make sure file system is at proper version | |
1946 | */ | |
1947 | ||
3558fd73 | 1948 | uid = crgetuid(cr); |
0037b49e | 1949 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1950 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1951 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1952 | |
32dec7bd | 1953 | if (dirname == NULL) |
1954 | return (SET_ERROR(EINVAL)); | |
1955 | ||
0037b49e | 1956 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1957 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1958 | zilog = zfsvfs->z_log; |
34dc7c2f | 1959 | |
428870ff | 1960 | if (dzp->z_pflags & ZFS_XATTR) { |
0037b49e | 1961 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1962 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1963 | } |
1964 | ||
0037b49e | 1965 | if (zfsvfs->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1966 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1967 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1968 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1969 | } |
1970 | if (flags & FIGNORECASE) | |
1971 | zf |= ZCILOOK; | |
1972 | ||
5484965a | 1973 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1974 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1975 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1976 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1977 | return (error); |
1978 | } | |
428870ff | 1979 | } |
34dc7c2f | 1980 | |
428870ff BB |
1981 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1982 | vsecp, &acl_ids)) != 0) { | |
0037b49e | 1983 | ZFS_EXIT(zfsvfs); |
428870ff BB |
1984 | return (error); |
1985 | } | |
34dc7c2f BB |
1986 | /* |
1987 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1988 | * |
1989 | * Existence is checked first to make sure we don't return | |
1990 | * EACCES instead of EEXIST which can cause some applications | |
1991 | * to fail. | |
34dc7c2f BB |
1992 | */ |
1993 | top: | |
3558fd73 | 1994 | *ipp = NULL; |
34dc7c2f | 1995 | |
149e873a BB |
1996 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1997 | NULL, NULL))) { | |
428870ff | 1998 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 1999 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2000 | return (error); |
2001 | } | |
2002 | ||
149e873a | 2003 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 2004 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2005 | zfs_dirent_unlock(dl); |
0037b49e | 2006 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2007 | return (error); |
2008 | } | |
2009 | ||
9c5167d1 | 2010 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) { |
45d1cae3 | 2011 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 2012 | zfs_dirent_unlock(dl); |
0037b49e | 2013 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2014 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
2015 | } |
2016 | ||
34dc7c2f BB |
2017 | /* |
2018 | * Add a new entry to the directory. | |
2019 | */ | |
0037b49e | 2020 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f BB |
2021 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
2022 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
0037b49e | 2023 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 2024 | if (fuid_dirtied) |
0037b49e BB |
2025 | zfs_fuid_txhold(zfsvfs, tx); |
2026 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
2027 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
2028 | acl_ids.z_aclp->z_acl_bytes); | |
2029 | } | |
2030 | ||
2031 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
2032 | ZFS_SA_BASE_ATTR_SIZE); | |
2033 | ||
0735ecb3 | 2034 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
2035 | if (error) { |
2036 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2037 | if (error == ERESTART) { |
e8b96c60 | 2038 | waited = B_TRUE; |
34dc7c2f BB |
2039 | dmu_tx_wait(tx); |
2040 | dmu_tx_abort(tx); | |
2041 | goto top; | |
2042 | } | |
428870ff | 2043 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2044 | dmu_tx_abort(tx); |
0037b49e | 2045 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2046 | return (error); |
2047 | } | |
2048 | ||
2049 | /* | |
2050 | * Create new node. | |
2051 | */ | |
428870ff | 2052 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f BB |
2053 | |
2054 | /* | |
2055 | * Now put new name in parent dir. | |
2056 | */ | |
599b8648 CC |
2057 | error = zfs_link_create(dl, zp, tx, ZNEW); |
2058 | if (error != 0) { | |
2059 | zfs_znode_delete(zp, tx); | |
2060 | remove_inode_hash(ZTOI(zp)); | |
2061 | goto out; | |
2062 | } | |
2063 | ||
2064 | if (fuid_dirtied) | |
2065 | zfs_fuid_sync(zfsvfs, tx); | |
34dc7c2f | 2066 | |
3558fd73 | 2067 | *ipp = ZTOI(zp); |
34dc7c2f BB |
2068 | |
2069 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
2070 | if (flags & FIGNORECASE) | |
2071 | txtype |= TX_CI; | |
9babb374 BB |
2072 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
2073 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 2074 | |
599b8648 | 2075 | out: |
9babb374 | 2076 | zfs_acl_ids_free(&acl_ids); |
428870ff | 2077 | |
34dc7c2f BB |
2078 | dmu_tx_commit(tx); |
2079 | ||
2080 | zfs_dirent_unlock(dl); | |
2081 | ||
0037b49e | 2082 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2083 | zil_commit(zilog, 0); |
428870ff | 2084 | |
599b8648 CC |
2085 | if (error != 0) { |
2086 | iput(ZTOI(zp)); | |
2087 | } else { | |
2088 | zfs_inode_update(dzp); | |
2089 | zfs_inode_update(zp); | |
2090 | } | |
0037b49e | 2091 | ZFS_EXIT(zfsvfs); |
599b8648 | 2092 | return (error); |
34dc7c2f BB |
2093 | } |
2094 | ||
2095 | /* | |
2096 | * Remove a directory subdir entry. If the current working | |
2097 | * directory is the same as the subdir to be removed, the | |
2098 | * remove will fail. | |
2099 | * | |
3558fd73 | 2100 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 2101 | * name - name of directory to be removed. |
3558fd73 | 2102 | * cwd - inode of current working directory. |
34dc7c2f | 2103 | * cr - credentials of caller. |
34dc7c2f BB |
2104 | * flags - case flags |
2105 | * | |
d3cc8b15 | 2106 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
2107 | * |
2108 | * Timestamps: | |
3558fd73 | 2109 | * dip - ctime|mtime updated |
34dc7c2f BB |
2110 | */ |
2111 | /*ARGSUSED*/ | |
e5c39b95 | 2112 | int |
3558fd73 BB |
2113 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
2114 | int flags) | |
34dc7c2f | 2115 | { |
3558fd73 | 2116 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 2117 | znode_t *zp; |
3558fd73 | 2118 | struct inode *ip; |
0037b49e | 2119 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
2120 | zilog_t *zilog; |
2121 | zfs_dirlock_t *dl; | |
2122 | dmu_tx_t *tx; | |
2123 | int error; | |
2124 | int zflg = ZEXISTS; | |
e8b96c60 | 2125 | boolean_t waited = B_FALSE; |
34dc7c2f | 2126 | |
32dec7bd | 2127 | if (name == NULL) |
2128 | return (SET_ERROR(EINVAL)); | |
2129 | ||
0037b49e | 2130 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2131 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 2132 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2133 | |
2134 | if (flags & FIGNORECASE) | |
2135 | zflg |= ZCILOOK; | |
2136 | top: | |
2137 | zp = NULL; | |
2138 | ||
2139 | /* | |
2140 | * Attempt to lock directory; fail if entry doesn't exist. | |
2141 | */ | |
149e873a BB |
2142 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
2143 | NULL, NULL))) { | |
0037b49e | 2144 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2145 | return (error); |
2146 | } | |
2147 | ||
3558fd73 | 2148 | ip = ZTOI(zp); |
34dc7c2f | 2149 | |
149e873a | 2150 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
2151 | goto out; |
2152 | } | |
2153 | ||
3558fd73 | 2154 | if (!S_ISDIR(ip->i_mode)) { |
2e528b49 | 2155 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
2156 | goto out; |
2157 | } | |
2158 | ||
3558fd73 | 2159 | if (ip == cwd) { |
2e528b49 | 2160 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
2161 | goto out; |
2162 | } | |
2163 | ||
34dc7c2f | 2164 | /* |
4e33ba4c | 2165 | * Grab a lock on the directory to make sure that no one is |
34dc7c2f BB |
2166 | * trying to add (or lookup) entries while we are removing it. |
2167 | */ | |
2168 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
2169 | ||
2170 | /* | |
2171 | * Grab a lock on the parent pointer to make sure we play well | |
2172 | * with the treewalk and directory rename code. | |
2173 | */ | |
2174 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
2175 | ||
0037b49e | 2176 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 2177 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 2178 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
0037b49e | 2179 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
2180 | zfs_sa_upgrade_txholds(tx, zp); |
2181 | zfs_sa_upgrade_txholds(tx, dzp); | |
db707ad0 | 2182 | dmu_tx_mark_netfree(tx); |
0735ecb3 | 2183 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
2184 | if (error) { |
2185 | rw_exit(&zp->z_parent_lock); | |
2186 | rw_exit(&zp->z_name_lock); | |
2187 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2188 | if (error == ERESTART) { |
e8b96c60 | 2189 | waited = B_TRUE; |
34dc7c2f BB |
2190 | dmu_tx_wait(tx); |
2191 | dmu_tx_abort(tx); | |
ea7e86d8 | 2192 | iput(ip); |
34dc7c2f BB |
2193 | goto top; |
2194 | } | |
2195 | dmu_tx_abort(tx); | |
ea7e86d8 | 2196 | iput(ip); |
0037b49e | 2197 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2198 | return (error); |
2199 | } | |
2200 | ||
2201 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
2202 | ||
2203 | if (error == 0) { | |
2204 | uint64_t txtype = TX_RMDIR; | |
2205 | if (flags & FIGNORECASE) | |
2206 | txtype |= TX_CI; | |
572e2857 | 2207 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
2208 | } |
2209 | ||
2210 | dmu_tx_commit(tx); | |
2211 | ||
2212 | rw_exit(&zp->z_parent_lock); | |
2213 | rw_exit(&zp->z_name_lock); | |
2214 | out: | |
2215 | zfs_dirent_unlock(dl); | |
2216 | ||
59157910 BB |
2217 | zfs_inode_update(dzp); |
2218 | zfs_inode_update(zp); | |
3558fd73 | 2219 | iput(ip); |
34dc7c2f | 2220 | |
0037b49e | 2221 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2222 | zil_commit(zilog, 0); |
428870ff | 2223 | |
0037b49e | 2224 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2225 | return (error); |
2226 | } | |
2227 | ||
2228 | /* | |
2229 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 2230 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 2231 | * |
3558fd73 BB |
2232 | * IN: ip - inode of directory to read. |
2233 | * dirent - buffer for directory entries. | |
34dc7c2f | 2234 | * |
3558fd73 | 2235 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
2236 | * |
2237 | * RETURN: 0 if success | |
2238 | * error code if failure | |
2239 | * | |
2240 | * Timestamps: | |
3558fd73 | 2241 | * ip - atime updated |
34dc7c2f BB |
2242 | * |
2243 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2244 | * This allows us to use the low range for "special" directory entries: | |
2245 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2246 | * we use the offset 2 for the '.zfs' directory. | |
2247 | */ | |
2248 | /* ARGSUSED */ | |
3558fd73 | 2249 | int |
9464b959 | 2250 | zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) |
34dc7c2f | 2251 | { |
3558fd73 | 2252 | znode_t *zp = ITOZ(ip); |
0037b49e | 2253 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2254 | objset_t *os; |
34dc7c2f BB |
2255 | zap_cursor_t zc; |
2256 | zap_attribute_t zap; | |
34dc7c2f BB |
2257 | int error; |
2258 | uint8_t prefetch; | |
c12e3a59 | 2259 | uint8_t type; |
3558fd73 BB |
2260 | int done = 0; |
2261 | uint64_t parent; | |
c12e3a59 | 2262 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2263 | |
0037b49e | 2264 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2265 | ZFS_VERIFY_ZP(zp); |
2266 | ||
0037b49e | 2267 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
3558fd73 BB |
2268 | &parent, sizeof (parent))) != 0) |
2269 | goto out; | |
34dc7c2f BB |
2270 | |
2271 | /* | |
2272 | * Quit if directory has been removed (posix) | |
2273 | */ | |
3558fd73 BB |
2274 | if (zp->z_unlinked) |
2275 | goto out; | |
2276 | ||
c12e3a59 | 2277 | error = 0; |
0037b49e | 2278 | os = zfsvfs->z_os; |
c12e3a59 | 2279 | offset = ctx->pos; |
34dc7c2f BB |
2280 | prefetch = zp->z_zn_prefetch; |
2281 | ||
2282 | /* | |
2283 | * Initialize the iterator cursor. | |
2284 | */ | |
c12e3a59 | 2285 | if (offset <= 3) { |
34dc7c2f BB |
2286 | /* |
2287 | * Start iteration from the beginning of the directory. | |
2288 | */ | |
2289 | zap_cursor_init(&zc, os, zp->z_id); | |
2290 | } else { | |
2291 | /* | |
2292 | * The offset is a serialized cursor. | |
2293 | */ | |
c12e3a59 | 2294 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2295 | } |
2296 | ||
34dc7c2f BB |
2297 | /* |
2298 | * Transform to file-system independent format | |
2299 | */ | |
3558fd73 BB |
2300 | while (!done) { |
2301 | uint64_t objnum; | |
34dc7c2f BB |
2302 | /* |
2303 | * Special case `.', `..', and `.zfs'. | |
2304 | */ | |
c12e3a59 | 2305 | if (offset == 0) { |
34dc7c2f BB |
2306 | (void) strcpy(zap.za_name, "."); |
2307 | zap.za_normalization_conflict = 0; | |
2308 | objnum = zp->z_id; | |
c12e3a59 RY |
2309 | type = DT_DIR; |
2310 | } else if (offset == 1) { | |
34dc7c2f BB |
2311 | (void) strcpy(zap.za_name, ".."); |
2312 | zap.za_normalization_conflict = 0; | |
428870ff | 2313 | objnum = parent; |
c12e3a59 RY |
2314 | type = DT_DIR; |
2315 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2316 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2317 | zap.za_normalization_conflict = 0; | |
2318 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 2319 | type = DT_DIR; |
34dc7c2f BB |
2320 | } else { |
2321 | /* | |
2322 | * Grab next entry. | |
2323 | */ | |
3558fd73 BB |
2324 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2325 | if (error == ENOENT) | |
34dc7c2f BB |
2326 | break; |
2327 | else | |
2328 | goto update; | |
2329 | } | |
2330 | ||
0c5dde49 BB |
2331 | /* |
2332 | * Allow multiple entries provided the first entry is | |
2333 | * the object id. Non-zpl consumers may safely make | |
2334 | * use of the additional space. | |
2335 | * | |
2336 | * XXX: This should be a feature flag for compatibility | |
2337 | */ | |
34dc7c2f | 2338 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2339 | zap.za_num_integers == 0) { |
34dc7c2f | 2340 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2341 | "entry, obj = %lld, offset = %lld, " |
2342 | "length = %d, num = %lld\n", | |
34dc7c2f | 2343 | (u_longlong_t)zp->z_id, |
c12e3a59 | 2344 | (u_longlong_t)offset, |
0c5dde49 BB |
2345 | zap.za_integer_length, |
2346 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 2347 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2348 | goto update; |
2349 | } | |
2350 | ||
2351 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 2352 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2353 | } |
0f37d0c8 | 2354 | |
9464b959 | 2355 | done = !zpl_dir_emit(ctx, zap.za_name, strlen(zap.za_name), |
c12e3a59 | 2356 | objnum, type); |
0f37d0c8 | 2357 | if (done) |
34dc7c2f | 2358 | break; |
34dc7c2f BB |
2359 | |
2360 | /* Prefetch znode */ | |
3558fd73 | 2361 | if (prefetch) { |
fcff0f35 PD |
2362 | dmu_prefetch(os, objnum, 0, 0, 0, |
2363 | ZIO_PRIORITY_SYNC_READ); | |
3558fd73 | 2364 | } |
34dc7c2f | 2365 | |
c12e3a59 RY |
2366 | /* |
2367 | * Move to the next entry, fill in the previous offset. | |
2368 | */ | |
2369 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2370 | zap_cursor_advance(&zc); |
c12e3a59 | 2371 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2372 | } else { |
c12e3a59 | 2373 | offset += 1; |
34dc7c2f | 2374 | } |
c12e3a59 | 2375 | ctx->pos = offset; |
34dc7c2f BB |
2376 | } |
2377 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2378 | ||
34dc7c2f BB |
2379 | update: |
2380 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2381 | if (error == ENOENT) |
2382 | error = 0; | |
3558fd73 | 2383 | out: |
0037b49e | 2384 | ZFS_EXIT(zfsvfs); |
34dc7c2f | 2385 | |
34dc7c2f BB |
2386 | return (error); |
2387 | } | |
2388 | ||
d5446cfc BB |
2389 | ulong_t zfs_fsync_sync_cnt = 4; |
2390 | ||
e5c39b95 | 2391 | int |
3558fd73 | 2392 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2393 | { |
3558fd73 | 2394 | znode_t *zp = ITOZ(ip); |
0037b49e | 2395 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2396 | |
d5446cfc BB |
2397 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2398 | ||
0037b49e BB |
2399 | if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2400 | ZFS_ENTER(zfsvfs); | |
428870ff | 2401 | ZFS_VERIFY_ZP(zp); |
0037b49e BB |
2402 | zil_commit(zfsvfs->z_log, zp->z_id); |
2403 | ZFS_EXIT(zfsvfs); | |
428870ff | 2404 | } |
07012da6 CC |
2405 | tsd_set(zfs_fsyncer_key, NULL); |
2406 | ||
34dc7c2f BB |
2407 | return (0); |
2408 | } | |
2409 | ||
2410 | ||
2411 | /* | |
2412 | * Get the requested file attributes and place them in the provided | |
2413 | * vattr structure. | |
2414 | * | |
3558fd73 | 2415 | * IN: ip - inode of file. |
5484965a BB |
2416 | * vap - va_mask identifies requested attributes. |
2417 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2418 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2419 | * cr - credentials of caller. | |
34dc7c2f | 2420 | * |
5484965a BB |
2421 | * OUT: vap - attribute values. |
2422 | * | |
2423 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2424 | */ |
2425 | /* ARGSUSED */ | |
e5c39b95 | 2426 | int |
5484965a | 2427 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2428 | { |
3558fd73 | 2429 | znode_t *zp = ITOZ(ip); |
0037b49e | 2430 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
2431 | int error = 0; |
2432 | uint64_t links; | |
0df9673f | 2433 | uint64_t atime[2], mtime[2], ctime[2]; |
5484965a BB |
2434 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2435 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2436 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
0df9673f | 2437 | sa_bulk_attr_t bulk[3]; |
428870ff | 2438 | int count = 0; |
34dc7c2f | 2439 | |
0037b49e | 2440 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2441 | ZFS_VERIFY_ZP(zp); |
428870ff | 2442 | |
5484965a | 2443 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2444 | |
0037b49e BB |
2445 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); |
2446 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
2447 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
428870ff BB |
2448 | |
2449 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
0037b49e | 2450 | ZFS_EXIT(zfsvfs); |
428870ff BB |
2451 | return (error); |
2452 | } | |
34dc7c2f | 2453 | |
34dc7c2f BB |
2454 | /* |
2455 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2456 | * Also, if we are the owner don't bother, since owner should | |
2457 | * always be allowed to read basic attributes of file. | |
2458 | */ | |
572e2857 | 2459 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2460 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2461 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2462 | skipaclchk, cr))) { | |
0037b49e | 2463 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2464 | return (error); |
2465 | } | |
2466 | } | |
2467 | ||
2468 | /* | |
2469 | * Return all attributes. It's cheaper to provide the answer | |
2470 | * than to determine whether we were asked the question. | |
2471 | */ | |
2472 | ||
9babb374 | 2473 | mutex_enter(&zp->z_lock); |
5484965a BB |
2474 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2475 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2476 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2477 | vap->va_nodeid = zp->z_id; |
0037b49e | 2478 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) |
dfbc8630 | 2479 | links = ZTOI(zp)->i_nlink + 1; |
34dc7c2f | 2480 | else |
dfbc8630 | 2481 | links = ZTOI(zp)->i_nlink; |
5484965a BB |
2482 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2483 | vap->va_size = i_size_read(ip); | |
2484 | vap->va_rdev = ip->i_rdev; | |
2485 | vap->va_seq = ip->i_generation; | |
2486 | ||
2487 | /* | |
2488 | * Add in any requested optional attributes and the create time. | |
2489 | * Also set the corresponding bits in the returned attribute bitmap. | |
2490 | */ | |
0037b49e | 2491 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { |
5484965a BB |
2492 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { |
2493 | xoap->xoa_archive = | |
2494 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2495 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2496 | } | |
2497 | ||
2498 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2499 | xoap->xoa_readonly = | |
2500 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2501 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2502 | } | |
2503 | ||
2504 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2505 | xoap->xoa_system = | |
2506 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2507 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2508 | } | |
2509 | ||
2510 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2511 | xoap->xoa_hidden = | |
2512 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2513 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2514 | } | |
2515 | ||
2516 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2517 | xoap->xoa_nounlink = | |
2518 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2519 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2520 | } | |
2521 | ||
2522 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2523 | xoap->xoa_immutable = | |
2524 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2525 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2526 | } | |
2527 | ||
2528 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2529 | xoap->xoa_appendonly = | |
2530 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2531 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2532 | } | |
2533 | ||
2534 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2535 | xoap->xoa_nodump = | |
2536 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2537 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2538 | } | |
2539 | ||
2540 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2541 | xoap->xoa_opaque = | |
2542 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2543 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2544 | } | |
2545 | ||
2546 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2547 | xoap->xoa_av_quarantined = | |
2548 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2549 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2550 | } | |
2551 | ||
2552 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2553 | xoap->xoa_av_modified = | |
2554 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2555 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2556 | } | |
2557 | ||
2558 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2559 | S_ISREG(ip->i_mode)) { | |
2560 | zfs_sa_get_scanstamp(zp, xvap); | |
2561 | } | |
34dc7c2f | 2562 | |
5484965a BB |
2563 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2564 | uint64_t times[2]; | |
2565 | ||
0037b49e | 2566 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), |
5484965a BB |
2567 | times, sizeof (times)); |
2568 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2569 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2570 | } | |
2571 | ||
2572 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2573 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2574 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2575 | } | |
2576 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
278f2236 | 2577 | xoap->xoa_generation = ip->i_generation; |
5484965a BB |
2578 | XVA_SET_RTN(xvap, XAT_GEN); |
2579 | } | |
2580 | ||
2581 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2582 | xoap->xoa_offline = | |
2583 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2584 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2585 | } | |
2586 | ||
2587 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2588 | xoap->xoa_sparse = | |
2589 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2590 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2591 | } | |
9c5167d1 NF |
2592 | |
2593 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { | |
2594 | xoap->xoa_projinherit = | |
2595 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0); | |
2596 | XVA_SET_RTN(xvap, XAT_PROJINHERIT); | |
2597 | } | |
2598 | ||
2599 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
2600 | xoap->xoa_projid = zp->z_projid; | |
2601 | XVA_SET_RTN(xvap, XAT_PROJID); | |
2602 | } | |
5484965a BB |
2603 | } |
2604 | ||
0df9673f | 2605 | ZFS_TIME_DECODE(&vap->va_atime, atime); |
5484965a BB |
2606 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); |
2607 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2608 | |
2609 | mutex_exit(&zp->z_lock); | |
2610 | ||
5484965a | 2611 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2612 | |
2613 | if (zp->z_blksz == 0) { | |
2614 | /* | |
2615 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2616 | */ | |
0037b49e | 2617 | vap->va_blksize = zfsvfs->z_max_blksz; |
34dc7c2f BB |
2618 | } |
2619 | ||
0037b49e | 2620 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2621 | return (0); |
2622 | } | |
2623 | ||
057e8eee BB |
2624 | /* |
2625 | * Get the basic file attributes and place them in the provided kstat | |
2626 | * structure. The inode is assumed to be the authoritative source | |
2627 | * for most of the attributes. However, the znode currently has the | |
2628 | * authoritative atime, blksize, and block count. | |
2629 | * | |
2630 | * IN: ip - inode of file. | |
2631 | * | |
2632 | * OUT: sp - kstat values. | |
2633 | * | |
2634 | * RETURN: 0 (always succeeds) | |
2635 | */ | |
2636 | /* ARGSUSED */ | |
2637 | int | |
2638 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2639 | { | |
2640 | znode_t *zp = ITOZ(ip); | |
0037b49e | 2641 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
b585bc4a BB |
2642 | uint32_t blksize; |
2643 | u_longlong_t nblocks; | |
057e8eee | 2644 | |
0037b49e | 2645 | ZFS_ENTER(zfsvfs); |
a7b125e9 GB |
2646 | ZFS_VERIFY_ZP(zp); |
2647 | ||
057e8eee BB |
2648 | mutex_enter(&zp->z_lock); |
2649 | ||
2650 | generic_fillattr(ip, sp); | |
057e8eee | 2651 | |
b585bc4a BB |
2652 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2653 | sp->blksize = blksize; | |
2654 | sp->blocks = nblocks; | |
2655 | ||
057e8eee BB |
2656 | if (unlikely(zp->z_blksz == 0)) { |
2657 | /* | |
2658 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2659 | */ | |
0037b49e | 2660 | sp->blksize = zfsvfs->z_max_blksz; |
057e8eee BB |
2661 | } |
2662 | ||
2663 | mutex_exit(&zp->z_lock); | |
2664 | ||
aa9b2708 AV |
2665 | /* |
2666 | * Required to prevent NFS client from detecting different inode | |
2667 | * numbers of snapshot root dentry before and after snapshot mount. | |
2668 | */ | |
0037b49e | 2669 | if (zfsvfs->z_issnap) { |
aa9b2708 AV |
2670 | if (ip->i_sb->s_root->d_inode == ip) |
2671 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
0037b49e | 2672 | dmu_objset_id(zfsvfs->z_os); |
aa9b2708 AV |
2673 | } |
2674 | ||
0037b49e | 2675 | ZFS_EXIT(zfsvfs); |
a7b125e9 | 2676 | |
057e8eee BB |
2677 | return (0); |
2678 | } | |
057e8eee | 2679 | |
9c5167d1 NF |
2680 | /* |
2681 | * For the operation of changing file's user/group/project, we need to | |
2682 | * handle not only the main object that is assigned to the file directly, | |
2683 | * but also the ones that are used by the file via hidden xattr directory. | |
2684 | * | |
2685 | * Because the xattr directory may contains many EA entries, as to it may | |
2686 | * be impossible to change all of them via the transaction of changing the | |
2687 | * main object's user/group/project attributes. Then we have to change them | |
2688 | * via other multiple independent transactions one by one. It may be not good | |
2689 | * solution, but we have no better idea yet. | |
2690 | */ | |
2691 | static int | |
2692 | zfs_setattr_dir(znode_t *dzp) | |
2693 | { | |
2694 | struct inode *dxip = ZTOI(dzp); | |
2695 | struct inode *xip = NULL; | |
2696 | zfsvfs_t *zfsvfs = ITOZSB(dxip); | |
2697 | objset_t *os = zfsvfs->z_os; | |
2698 | zap_cursor_t zc; | |
2699 | zap_attribute_t zap; | |
2700 | zfs_dirlock_t *dl; | |
2701 | znode_t *zp; | |
2702 | dmu_tx_t *tx = NULL; | |
2703 | uint64_t uid, gid; | |
2704 | sa_bulk_attr_t bulk[4]; | |
2705 | int count = 0; | |
2706 | int err; | |
2707 | ||
2708 | zap_cursor_init(&zc, os, dzp->z_id); | |
2709 | while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) { | |
2710 | if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { | |
2711 | err = ENXIO; | |
2712 | break; | |
2713 | } | |
2714 | ||
2715 | err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp, | |
2716 | ZEXISTS, NULL, NULL); | |
2717 | if (err == ENOENT) | |
2718 | goto next; | |
2719 | if (err) | |
2720 | break; | |
2721 | ||
2722 | xip = ZTOI(zp); | |
2723 | if (KUID_TO_SUID(xip->i_uid) == KUID_TO_SUID(dxip->i_uid) && | |
2724 | KGID_TO_SGID(xip->i_gid) == KGID_TO_SGID(dxip->i_gid) && | |
2725 | zp->z_projid == dzp->z_projid) | |
2726 | goto next; | |
2727 | ||
2728 | tx = dmu_tx_create(os); | |
2729 | if (!(zp->z_pflags & ZFS_PROJID)) | |
2730 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
2731 | else | |
2732 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
2733 | ||
2734 | err = dmu_tx_assign(tx, TXG_WAIT); | |
2735 | if (err) | |
2736 | break; | |
2737 | ||
2738 | mutex_enter(&dzp->z_lock); | |
2739 | ||
2740 | if (KUID_TO_SUID(xip->i_uid) != KUID_TO_SUID(dxip->i_uid)) { | |
2741 | xip->i_uid = dxip->i_uid; | |
2742 | uid = zfs_uid_read(dxip); | |
2743 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, | |
2744 | &uid, sizeof (uid)); | |
2745 | } | |
2746 | ||
2747 | if (KGID_TO_SGID(xip->i_gid) != KGID_TO_SGID(dxip->i_gid)) { | |
2748 | xip->i_gid = dxip->i_gid; | |
2749 | gid = zfs_gid_read(dxip); | |
2750 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, | |
2751 | &gid, sizeof (gid)); | |
2752 | } | |
2753 | ||
2754 | if (zp->z_projid != dzp->z_projid) { | |
2755 | if (!(zp->z_pflags & ZFS_PROJID)) { | |
2756 | zp->z_pflags |= ZFS_PROJID; | |
2757 | SA_ADD_BULK_ATTR(bulk, count, | |
2758 | SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, | |
2759 | sizeof (zp->z_pflags)); | |
2760 | } | |
2761 | ||
2762 | zp->z_projid = dzp->z_projid; | |
2763 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs), | |
2764 | NULL, &zp->z_projid, sizeof (zp->z_projid)); | |
2765 | } | |
2766 | ||
2767 | mutex_exit(&dzp->z_lock); | |
2768 | ||
2769 | if (likely(count > 0)) { | |
2770 | err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
2771 | dmu_tx_commit(tx); | |
2772 | } else { | |
2773 | dmu_tx_abort(tx); | |
2774 | } | |
2775 | tx = NULL; | |
2776 | if (err != 0 && err != ENOENT) | |
2777 | break; | |
2778 | ||
2779 | next: | |
2780 | if (xip) { | |
2781 | iput(xip); | |
2782 | xip = NULL; | |
2783 | zfs_dirent_unlock(dl); | |
2784 | } | |
2785 | zap_cursor_advance(&zc); | |
2786 | } | |
2787 | ||
2788 | if (tx) | |
2789 | dmu_tx_abort(tx); | |
2790 | if (xip) { | |
2791 | iput(xip); | |
2792 | zfs_dirent_unlock(dl); | |
2793 | } | |
2794 | zap_cursor_fini(&zc); | |
2795 | ||
2796 | return (err == ENOENT ? 0 : err); | |
2797 | } | |
2798 | ||
34dc7c2f BB |
2799 | /* |
2800 | * Set the file attributes to the values contained in the | |
2801 | * vattr structure. | |
2802 | * | |
3558fd73 | 2803 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2804 | * vap - new attribute values. |
5484965a | 2805 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2806 | * flags - ATTR_UTIME set if non-default time values provided. |
2807 | * - ATTR_NOACLCHECK (CIFS context only). | |
2808 | * cr - credentials of caller. | |
34dc7c2f BB |
2809 | * |
2810 | * RETURN: 0 if success | |
2811 | * error code if failure | |
2812 | * | |
2813 | * Timestamps: | |
3558fd73 | 2814 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2815 | */ |
2816 | /* ARGSUSED */ | |
e5c39b95 | 2817 | int |
5484965a | 2818 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2819 | { |
3558fd73 | 2820 | znode_t *zp = ITOZ(ip); |
0037b49e | 2821 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
9c5167d1 | 2822 | objset_t *os = zfsvfs->z_os; |
34dc7c2f BB |
2823 | zilog_t *zilog; |
2824 | dmu_tx_t *tx; | |
2825 | vattr_t oldva; | |
f4ea75d4 | 2826 | xvattr_t *tmpxvattr; |
5484965a | 2827 | uint_t mask = vap->va_mask; |
a117a6d6 | 2828 | uint_t saved_mask = 0; |
34dc7c2f BB |
2829 | int trim_mask = 0; |
2830 | uint64_t new_mode; | |
64aefee1 | 2831 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 2832 | uint64_t xattr_obj; |
0df9673f | 2833 | uint64_t mtime[2], ctime[2], atime[2]; |
9c5167d1 | 2834 | uint64_t projid = ZFS_INVALID_PROJID; |
34dc7c2f BB |
2835 | znode_t *attrzp; |
2836 | int need_policy = FALSE; | |
9c5167d1 | 2837 | int err, err2 = 0; |
34dc7c2f | 2838 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2839 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2840 | xoptattr_t *xoap; | |
2841 | zfs_acl_t *aclp; | |
34dc7c2f | 2842 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2843 | boolean_t fuid_dirtied = B_FALSE; |
9c5167d1 | 2844 | boolean_t handle_eadir = B_FALSE; |
17c37660 | 2845 | sa_bulk_attr_t *bulk, *xattr_bulk; |
9c5167d1 | 2846 | int count = 0, xattr_count = 0, bulks = 8; |
34dc7c2f BB |
2847 | |
2848 | if (mask == 0) | |
2849 | return (0); | |
2850 | ||
0037b49e | 2851 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2852 | ZFS_VERIFY_ZP(zp); |
2853 | ||
9c5167d1 NF |
2854 | /* |
2855 | * If this is a xvattr_t, then get a pointer to the structure of | |
2856 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2857 | */ | |
2858 | xoap = xva_getxoptattr(xvap); | |
2859 | if (xoap != NULL && (mask & ATTR_XVATTR)) { | |
2860 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
2861 | if (!dmu_objset_projectquota_enabled(os) || | |
2862 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode))) { | |
2863 | ZFS_EXIT(zfsvfs); | |
2864 | return (SET_ERROR(ENOTSUP)); | |
2865 | } | |
2866 | ||
2867 | projid = xoap->xoa_projid; | |
2868 | if (unlikely(projid == ZFS_INVALID_PROJID)) { | |
2869 | ZFS_EXIT(zfsvfs); | |
2870 | return (SET_ERROR(EINVAL)); | |
2871 | } | |
2872 | ||
2873 | if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID) | |
2874 | projid = ZFS_INVALID_PROJID; | |
2875 | else | |
2876 | need_policy = TRUE; | |
2877 | } | |
2878 | ||
2879 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) && | |
2705ebf0 NF |
2880 | (xoap->xoa_projinherit != |
2881 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) && | |
9c5167d1 NF |
2882 | (!dmu_objset_projectquota_enabled(os) || |
2883 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode)))) { | |
2705ebf0 NF |
2884 | ZFS_EXIT(zfsvfs); |
2885 | return (SET_ERROR(ENOTSUP)); | |
9c5167d1 NF |
2886 | } |
2887 | } | |
2888 | ||
0037b49e | 2889 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2890 | |
2891 | /* | |
2892 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2893 | * that file system is at proper version level | |
2894 | */ | |
5484965a | 2895 | |
0037b49e | 2896 | if (zfsvfs->z_use_fuids == B_FALSE && |
5484965a BB |
2897 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2898 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2899 | (mask & ATTR_XVATTR))) { | |
0037b49e | 2900 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2901 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2902 | } |
2903 | ||
3558fd73 | 2904 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
0037b49e | 2905 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2906 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2907 | } |
2908 | ||
3558fd73 | 2909 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
0037b49e | 2910 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2911 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2912 | } |
2913 | ||
d1d7e268 | 2914 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2915 | xva_init(tmpxvattr); |
5484965a | 2916 | |
9c5167d1 NF |
2917 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); |
2918 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); | |
17c37660 | 2919 | |
5484965a BB |
2920 | /* |
2921 | * Immutable files can only alter immutable bit and atime | |
2922 | */ | |
2923 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2924 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2925 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
ecb2b7dc | 2926 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2927 | goto out3; |
5484965a BB |
2928 | } |
2929 | ||
3558fd73 | 2930 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
ecb2b7dc | 2931 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2932 | goto out3; |
34dc7c2f BB |
2933 | } |
2934 | ||
5484965a BB |
2935 | /* |
2936 | * Verify timestamps doesn't overflow 32 bits. | |
2937 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2938 | * handle times greater than 2039. This check should be removed | |
2939 | * once large timestamps are fully supported. | |
2940 | */ | |
2941 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
2942 | if (((mask & ATTR_ATIME) && |
2943 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2944 | ((mask & ATTR_MTIME) && | |
2945 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
ecb2b7dc | 2946 | err = SET_ERROR(EOVERFLOW); |
f4ea75d4 | 2947 | goto out3; |
5484965a BB |
2948 | } |
2949 | } | |
2950 | ||
34dc7c2f BB |
2951 | top: |
2952 | attrzp = NULL; | |
572e2857 | 2953 | aclp = NULL; |
34dc7c2f | 2954 | |
45d1cae3 | 2955 | /* Can this be moved to before the top label? */ |
0037b49e | 2956 | if (zfs_is_readonly(zfsvfs)) { |
ecb2b7dc | 2957 | err = SET_ERROR(EROFS); |
f4ea75d4 | 2958 | goto out3; |
34dc7c2f BB |
2959 | } |
2960 | ||
2961 | /* | |
2962 | * First validate permissions | |
2963 | */ | |
2964 | ||
3558fd73 | 2965 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2966 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2967 | if (err) |
2968 | goto out3; | |
2969 | ||
34dc7c2f BB |
2970 | /* |
2971 | * XXX - Note, we are not providing any open | |
2972 | * mode flags here (like FNDELAY), so we may | |
2973 | * block if there are locks present... this | |
2974 | * should be addressed in openat(). | |
2975 | */ | |
b128c09f | 2976 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2977 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2978 | if (err) |
2979 | goto out3; | |
428870ff | 2980 | } |
34dc7c2f | 2981 | |
5484965a BB |
2982 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2983 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2984 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2985 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2986 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2987 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2988 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2989 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2990 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2991 | skipaclchk, cr); | |
2992 | } | |
2993 | ||
3558fd73 BB |
2994 | if (mask & (ATTR_UID|ATTR_GID)) { |
2995 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2996 | int take_owner; |
2997 | int take_group; | |
2998 | ||
2999 | /* | |
3000 | * NOTE: even if a new mode is being set, | |
3001 | * we may clear S_ISUID/S_ISGID bits. | |
3002 | */ | |
3003 | ||
3558fd73 | 3004 | if (!(mask & ATTR_MODE)) |
5484965a | 3005 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
3006 | |
3007 | /* | |
3008 | * Take ownership or chgrp to group we are a member of | |
3009 | */ | |
3010 | ||
5484965a | 3011 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 3012 | take_group = (mask & ATTR_GID) && |
0037b49e | 3013 | zfs_groupmember(zfsvfs, vap->va_gid, cr); |
34dc7c2f BB |
3014 | |
3015 | /* | |
5484965a | 3016 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
3017 | * take_group must both be set in order to allow taking |
3018 | * ownership. | |
3019 | * | |
3020 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
3021 | * | |
3022 | */ | |
3023 | ||
3558fd73 BB |
3024 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
3025 | take_owner && take_group) || | |
3026 | ((idmask == ATTR_UID) && take_owner) || | |
3027 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
3028 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
3029 | skipaclchk, cr) == 0) { | |
3030 | /* | |
3031 | * Remove setuid/setgid for non-privileged users | |
3032 | */ | |
5484965a | 3033 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 3034 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
3035 | } else { |
3036 | need_policy = TRUE; | |
3037 | } | |
3038 | } else { | |
3039 | need_policy = TRUE; | |
3040 | } | |
3041 | } | |
3042 | ||
3043 | mutex_enter(&zp->z_lock); | |
428870ff | 3044 | oldva.va_mode = zp->z_mode; |
572e2857 | 3045 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
3046 | if (mask & ATTR_XVATTR) { |
3047 | /* | |
3048 | * Update xvattr mask to include only those attributes | |
3049 | * that are actually changing. | |
3050 | * | |
3051 | * the bits will be restored prior to actually setting | |
3052 | * the attributes so the caller thinks they were set. | |
3053 | */ | |
3054 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
3055 | if (xoap->xoa_appendonly != | |
3056 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
3057 | need_policy = TRUE; | |
3058 | } else { | |
3059 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 3060 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
3061 | } |
3062 | } | |
3063 | ||
9c5167d1 NF |
3064 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { |
3065 | if (xoap->xoa_projinherit != | |
3066 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) { | |
3067 | need_policy = TRUE; | |
3068 | } else { | |
3069 | XVA_CLR_REQ(xvap, XAT_PROJINHERIT); | |
3070 | XVA_SET_REQ(tmpxvattr, XAT_PROJINHERIT); | |
3071 | } | |
3072 | } | |
3073 | ||
5484965a BB |
3074 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { |
3075 | if (xoap->xoa_nounlink != | |
3076 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
3077 | need_policy = TRUE; | |
3078 | } else { | |
3079 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 3080 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
3081 | } |
3082 | } | |
3083 | ||
3084 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
3085 | if (xoap->xoa_immutable != | |
3086 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
3087 | need_policy = TRUE; | |
3088 | } else { | |
3089 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 3090 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
3091 | } |
3092 | } | |
3093 | ||
3094 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
3095 | if (xoap->xoa_nodump != | |
3096 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
3097 | need_policy = TRUE; | |
3098 | } else { | |
3099 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 3100 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
3101 | } |
3102 | } | |
3103 | ||
3104 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
3105 | if (xoap->xoa_av_modified != | |
3106 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
3107 | need_policy = TRUE; | |
3108 | } else { | |
3109 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 3110 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
3111 | } |
3112 | } | |
3113 | ||
3114 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
3115 | if ((!S_ISREG(ip->i_mode) && | |
3116 | xoap->xoa_av_quarantined) || | |
3117 | xoap->xoa_av_quarantined != | |
3118 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
3119 | need_policy = TRUE; | |
3120 | } else { | |
3121 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 3122 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
3123 | } |
3124 | } | |
3125 | ||
3126 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
3127 | mutex_exit(&zp->z_lock); | |
ecb2b7dc | 3128 | err = SET_ERROR(EPERM); |
f4ea75d4 | 3129 | goto out3; |
5484965a BB |
3130 | } |
3131 | ||
3132 | if (need_policy == FALSE && | |
3133 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
3134 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
3135 | need_policy = TRUE; | |
3136 | } | |
3137 | } | |
34dc7c2f BB |
3138 | |
3139 | mutex_exit(&zp->z_lock); | |
3140 | ||
3558fd73 | 3141 | if (mask & ATTR_MODE) { |
34dc7c2f | 3142 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 3143 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 3144 | &oldva, cr); |
f4ea75d4 BB |
3145 | if (err) |
3146 | goto out3; | |
3147 | ||
3558fd73 | 3148 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
3149 | } else { |
3150 | need_policy = TRUE; | |
3151 | } | |
3152 | } | |
3153 | ||
3154 | if (need_policy) { | |
3155 | /* | |
3156 | * If trim_mask is set then take ownership | |
3157 | * has been granted or write_acl is present and user | |
3158 | * has the ability to modify mode. In that case remove | |
3159 | * UID|GID and or MODE from mask so that | |
3160 | * secpolicy_vnode_setattr() doesn't revoke it. | |
3161 | */ | |
3162 | ||
3163 | if (trim_mask) { | |
5484965a BB |
3164 | saved_mask = vap->va_mask; |
3165 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 3166 | } |
5484965a | 3167 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 3168 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
3169 | if (err) |
3170 | goto out3; | |
34dc7c2f BB |
3171 | |
3172 | if (trim_mask) | |
5484965a | 3173 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
3174 | } |
3175 | ||
3176 | /* | |
3177 | * secpolicy_vnode_setattr, or take ownership may have | |
3178 | * changed va_mask | |
3179 | */ | |
5484965a | 3180 | mask = vap->va_mask; |
34dc7c2f | 3181 | |
9c5167d1 NF |
3182 | if ((mask & (ATTR_UID | ATTR_GID)) || projid != ZFS_INVALID_PROJID) { |
3183 | handle_eadir = B_TRUE; | |
0037b49e | 3184 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
572e2857 | 3185 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 3186 | |
572e2857 | 3187 | if (err == 0 && xattr_obj) { |
3558fd73 | 3188 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
3189 | if (err) |
3190 | goto out2; | |
3191 | } | |
3558fd73 | 3192 | if (mask & ATTR_UID) { |
0037b49e | 3193 | new_kuid = zfs_fuid_create(zfsvfs, |
5484965a | 3194 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
64aefee1 | 3195 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
9c5167d1 NF |
3196 | zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT, |
3197 | new_kuid)) { | |
572e2857 | 3198 | if (attrzp) |
3558fd73 | 3199 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3200 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3201 | goto out2; |
3202 | } | |
3203 | } | |
3204 | ||
3558fd73 | 3205 | if (mask & ATTR_GID) { |
0037b49e BB |
3206 | new_kgid = zfs_fuid_create(zfsvfs, |
3207 | (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); | |
64aefee1 | 3208 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && |
9c5167d1 NF |
3209 | zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT, |
3210 | new_kgid)) { | |
572e2857 | 3211 | if (attrzp) |
3558fd73 | 3212 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3213 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3214 | goto out2; |
3215 | } | |
3216 | } | |
9c5167d1 NF |
3217 | |
3218 | if (projid != ZFS_INVALID_PROJID && | |
3219 | zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) { | |
3220 | if (attrzp) | |
3221 | iput(ZTOI(attrzp)); | |
3222 | err = EDQUOT; | |
3223 | goto out2; | |
3224 | } | |
428870ff | 3225 | } |
9c5167d1 | 3226 | tx = dmu_tx_create(os); |
34dc7c2f | 3227 | |
3558fd73 | 3228 | if (mask & ATTR_MODE) { |
428870ff | 3229 | uint64_t pmode = zp->z_mode; |
572e2857 | 3230 | uint64_t acl_obj; |
5484965a | 3231 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 3232 | |
572e2857 | 3233 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 3234 | |
572e2857 BB |
3235 | mutex_enter(&zp->z_lock); |
3236 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
3237 | /* |
3238 | * Are we upgrading ACL from old V0 format | |
3239 | * to V1 format? | |
3240 | */ | |
0037b49e | 3241 | if (zfsvfs->z_version >= ZPL_VERSION_FUID && |
572e2857 | 3242 | zfs_znode_acl_version(zp) == |
34dc7c2f | 3243 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 3244 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
3245 | DMU_OBJECT_END); |
3246 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
3247 | 0, aclp->z_acl_bytes); | |
3248 | } else { | |
572e2857 | 3249 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
3250 | aclp->z_acl_bytes); |
3251 | } | |
428870ff | 3252 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
3253 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
3254 | 0, aclp->z_acl_bytes); | |
3255 | } | |
572e2857 | 3256 | mutex_exit(&zp->z_lock); |
428870ff BB |
3257 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3258 | } else { | |
9c5167d1 NF |
3259 | if (((mask & ATTR_XVATTR) && |
3260 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) || | |
3261 | (projid != ZFS_INVALID_PROJID && | |
3262 | !(zp->z_pflags & ZFS_PROJID))) | |
5484965a BB |
3263 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3264 | else | |
3265 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3266 | } |
3267 | ||
428870ff BB |
3268 | if (attrzp) { |
3269 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3270 | } |
3271 | ||
0037b49e | 3272 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
428870ff | 3273 | if (fuid_dirtied) |
0037b49e | 3274 | zfs_fuid_txhold(zfsvfs, tx); |
428870ff BB |
3275 | |
3276 | zfs_sa_upgrade_txholds(tx, zp); | |
3277 | ||
384f8a09 MA |
3278 | err = dmu_tx_assign(tx, TXG_WAIT); |
3279 | if (err) | |
9babb374 | 3280 | goto out; |
34dc7c2f | 3281 | |
428870ff | 3282 | count = 0; |
34dc7c2f BB |
3283 | /* |
3284 | * Set each attribute requested. | |
3285 | * We group settings according to the locks they need to acquire. | |
3286 | * | |
3287 | * Note: you cannot set ctime directly, although it will be | |
3288 | * updated as a side-effect of calling this function. | |
3289 | */ | |
3290 | ||
9c5167d1 NF |
3291 | if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) { |
3292 | /* | |
3293 | * For the existed object that is upgraded from old system, | |
3294 | * its on-disk layout has no slot for the project ID attribute. | |
3295 | * But quota accounting logic needs to access related slots by | |
3296 | * offset directly. So we need to adjust old objects' layout | |
3297 | * to make the project ID to some unified and fixed offset. | |
3298 | */ | |
3299 | if (attrzp) | |
3300 | err = sa_add_projid(attrzp->z_sa_hdl, tx, projid); | |
3301 | if (err == 0) | |
3302 | err = sa_add_projid(zp->z_sa_hdl, tx, projid); | |
3303 | ||
3304 | if (unlikely(err == EEXIST)) | |
3305 | err = 0; | |
3306 | else if (err != 0) | |
3307 | goto out; | |
3308 | else | |
3309 | projid = ZFS_INVALID_PROJID; | |
3310 | } | |
572e2857 | 3311 | |
3558fd73 | 3312 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3313 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
3314 | mutex_enter(&zp->z_lock); |
3315 | ||
0037b49e | 3316 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
3317 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3318 | ||
3319 | if (attrzp) { | |
3558fd73 | 3320 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3321 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
3322 | mutex_enter(&attrzp->z_lock); |
3323 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3324 | SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, |
428870ff | 3325 | sizeof (attrzp->z_pflags)); |
9c5167d1 NF |
3326 | if (projid != ZFS_INVALID_PROJID) { |
3327 | attrzp->z_projid = projid; | |
3328 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3329 | SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid, | |
3330 | sizeof (attrzp->z_projid)); | |
3331 | } | |
428870ff BB |
3332 | } |
3333 | ||
3558fd73 | 3334 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 3335 | |
3558fd73 | 3336 | if (mask & ATTR_UID) { |
64aefee1 NB |
3337 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
3338 | new_uid = zfs_uid_read(ZTOI(zp)); | |
0037b49e | 3339 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 3340 | &new_uid, sizeof (new_uid)); |
428870ff BB |
3341 | if (attrzp) { |
3342 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3343 | SA_ZPL_UID(zfsvfs), NULL, &new_uid, |
428870ff | 3344 | sizeof (new_uid)); |
2c6abf15 | 3345 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
3346 | } |
3347 | } | |
3348 | ||
3558fd73 | 3349 | if (mask & ATTR_GID) { |
64aefee1 NB |
3350 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
3351 | new_gid = zfs_gid_read(ZTOI(zp)); | |
0037b49e | 3352 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), |
428870ff | 3353 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
3354 | if (attrzp) { |
3355 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3356 | SA_ZPL_GID(zfsvfs), NULL, &new_gid, |
428870ff | 3357 | sizeof (new_gid)); |
64aefee1 | 3358 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
3359 | } |
3360 | } | |
3558fd73 | 3361 | if (!(mask & ATTR_MODE)) { |
0037b49e | 3362 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), |
428870ff BB |
3363 | NULL, &new_mode, sizeof (new_mode)); |
3364 | new_mode = zp->z_mode; | |
3365 | } | |
3366 | err = zfs_acl_chown_setattr(zp); | |
3367 | ASSERT(err == 0); | |
3368 | if (attrzp) { | |
3369 | err = zfs_acl_chown_setattr(attrzp); | |
3370 | ASSERT(err == 0); | |
3371 | } | |
3372 | } | |
3373 | ||
3558fd73 | 3374 | if (mask & ATTR_MODE) { |
0037b49e | 3375 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 3376 | &new_mode, sizeof (new_mode)); |
12fa7f34 | 3377 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 3378 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 3379 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 3380 | ASSERT0(err); |
572e2857 BB |
3381 | if (zp->z_acl_cached) |
3382 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
3383 | zp->z_acl_cached = aclp; |
3384 | aclp = NULL; | |
34dc7c2f BB |
3385 | } |
3386 | ||
704cd075 CC |
3387 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
3388 | zp->z_atime_dirty = 0; | |
3389 | ZFS_TIME_ENCODE(&ip->i_atime, atime); | |
0037b49e | 3390 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
0df9673f | 3391 | &atime, sizeof (atime)); |
34dc7c2f BB |
3392 | } |
3393 | ||
99834d19 | 3394 | if (mask & (ATTR_MTIME | ATTR_SIZE)) { |
5484965a | 3395 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
6413c95f | 3396 | ZTOI(zp)->i_mtime = zpl_inode_timespec_trunc(vap->va_mtime, |
87f9371a NB |
3397 | ZTOI(zp)->i_sb->s_time_gran); |
3398 | ||
0037b49e | 3399 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 3400 | mtime, sizeof (mtime)); |
34dc7c2f BB |
3401 | } |
3402 | ||
99834d19 | 3403 | if (mask & (ATTR_CTIME | ATTR_SIZE)) { |
87f9371a | 3404 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); |
6413c95f | 3405 | ZTOI(zp)->i_ctime = zpl_inode_timespec_trunc(vap->va_ctime, |
87f9371a | 3406 | ZTOI(zp)->i_sb->s_time_gran); |
0037b49e | 3407 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
87f9371a | 3408 | ctime, sizeof (ctime)); |
428870ff | 3409 | } |
87f9371a | 3410 | |
9c5167d1 NF |
3411 | if (projid != ZFS_INVALID_PROJID) { |
3412 | zp->z_projid = projid; | |
3413 | SA_ADD_BULK_ATTR(bulk, count, | |
3414 | SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid, | |
3415 | sizeof (zp->z_projid)); | |
3416 | } | |
3417 | ||
87f9371a NB |
3418 | if (attrzp && mask) { |
3419 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3420 | SA_ZPL_CTIME(zfsvfs), NULL, &ctime, |
87f9371a NB |
3421 | sizeof (ctime)); |
3422 | } | |
3423 | ||
34dc7c2f BB |
3424 | /* |
3425 | * Do this after setting timestamps to prevent timestamp | |
3426 | * update from toggling bit | |
3427 | */ | |
3428 | ||
5484965a BB |
3429 | if (xoap && (mask & ATTR_XVATTR)) { |
3430 | ||
3431 | /* | |
3432 | * restore trimmed off masks | |
3433 | * so that return masks can be set for caller. | |
3434 | */ | |
3435 | ||
f4ea75d4 | 3436 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
3437 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
3438 | } | |
f4ea75d4 | 3439 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
3440 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
3441 | } | |
f4ea75d4 | 3442 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
3443 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
3444 | } | |
f4ea75d4 | 3445 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
3446 | XVA_SET_REQ(xvap, XAT_NODUMP); |
3447 | } | |
f4ea75d4 | 3448 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
3449 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
3450 | } | |
f4ea75d4 | 3451 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
3452 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
3453 | } | |
9c5167d1 NF |
3454 | if (XVA_ISSET_REQ(tmpxvattr, XAT_PROJINHERIT)) { |
3455 | XVA_SET_REQ(xvap, XAT_PROJINHERIT); | |
3456 | } | |
5484965a BB |
3457 | |
3458 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3459 | ASSERT(S_ISREG(ip->i_mode)); | |
3460 | ||
3461 | zfs_xvattr_set(zp, xvap, tx); | |
3462 | } | |
3463 | ||
9babb374 | 3464 | if (fuid_dirtied) |
0037b49e | 3465 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 3466 | |
34dc7c2f | 3467 | if (mask != 0) |
5484965a | 3468 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3469 | |
34dc7c2f | 3470 | mutex_exit(&zp->z_lock); |
3558fd73 | 3471 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3472 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3473 | |
572e2857 | 3474 | if (attrzp) { |
3558fd73 | 3475 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3476 | mutex_exit(&attrzp->z_acl_lock); |
3477 | mutex_exit(&attrzp->z_lock); | |
3478 | } | |
9babb374 | 3479 | out: |
9c5167d1 | 3480 | if (err == 0 && xattr_count > 0) { |
428870ff BB |
3481 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, |
3482 | xattr_count, tx); | |
3483 | ASSERT(err2 == 0); | |
3484 | } | |
3485 | ||
45d1cae3 | 3486 | if (aclp) |
9babb374 | 3487 | zfs_acl_free(aclp); |
9babb374 BB |
3488 | |
3489 | if (fuidp) { | |
3490 | zfs_fuid_info_free(fuidp); | |
3491 | fuidp = NULL; | |
3492 | } | |
3493 | ||
428870ff | 3494 | if (err) { |
9babb374 | 3495 | dmu_tx_abort(tx); |
ea7e86d8 BB |
3496 | if (attrzp) |
3497 | iput(ZTOI(attrzp)); | |
428870ff BB |
3498 | if (err == ERESTART) |
3499 | goto top; | |
3500 | } else { | |
9c5167d1 NF |
3501 | if (count > 0) |
3502 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3503 | dmu_tx_commit(tx); |
9c5167d1 NF |
3504 | if (attrzp) { |
3505 | if (err2 == 0 && handle_eadir) | |
3506 | err2 = zfs_setattr_dir(attrzp); | |
ea7e86d8 | 3507 | iput(ZTOI(attrzp)); |
9c5167d1 | 3508 | } |
037849f8 | 3509 | zfs_inode_update(zp); |
428870ff BB |
3510 | } |
3511 | ||
428870ff | 3512 | out2: |
9c5167d1 | 3513 | if (os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3514 | zil_commit(zilog, 0); |
34dc7c2f | 3515 | |
f4ea75d4 | 3516 | out3: |
9c5167d1 NF |
3517 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks); |
3518 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * bulks); | |
d1d7e268 | 3519 | kmem_free(tmpxvattr, sizeof (xvattr_t)); |
0037b49e | 3520 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3521 | return (err); |
3522 | } | |
3523 | ||
3524 | typedef struct zfs_zlock { | |
3525 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3526 | znode_t *zl_znode; /* znode we held */ | |
3527 | struct zfs_zlock *zl_next; /* next in list */ | |
3528 | } zfs_zlock_t; | |
3529 | ||
3530 | /* | |
3531 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3532 | */ | |
3533 | static void | |
3534 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3535 | { | |
3536 | zfs_zlock_t *zl; | |
3537 | ||
3538 | while ((zl = *zlpp) != NULL) { | |
3539 | if (zl->zl_znode != NULL) | |
ea7e86d8 | 3540 | zfs_iput_async(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3541 | rw_exit(zl->zl_rwlock); |
3542 | *zlpp = zl->zl_next; | |
3543 | kmem_free(zl, sizeof (*zl)); | |
3544 | } | |
3545 | } | |
3546 | ||
3547 | /* | |
3548 | * Search back through the directory tree, using the ".." entries. | |
3549 | * Lock each directory in the chain to prevent concurrent renames. | |
3550 | * Fail any attempt to move a directory into one of its own descendants. | |
3551 | * XXX - z_parent_lock can overlap with map or grow locks | |
3552 | */ | |
3553 | static int | |
3554 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3555 | { | |
3556 | zfs_zlock_t *zl; | |
3557 | znode_t *zp = tdzp; | |
3558fd73 | 3558 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3559 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3560 | krwlock_t *rwlp = &szp->z_parent_lock; |
3561 | krw_t rw = RW_WRITER; | |
3562 | ||
3563 | /* | |
3564 | * First pass write-locks szp and compares to zp->z_id. | |
3565 | * Later passes read-lock zp and compare to zp->z_parent. | |
3566 | */ | |
3567 | do { | |
3568 | if (!rw_tryenter(rwlp, rw)) { | |
3569 | /* | |
3570 | * Another thread is renaming in this path. | |
3571 | * Note that if we are a WRITER, we don't have any | |
3572 | * parent_locks held yet. | |
3573 | */ | |
3574 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3575 | /* | |
3576 | * Drop our locks and restart | |
3577 | */ | |
3578 | zfs_rename_unlock(&zl); | |
3579 | *zlpp = NULL; | |
3580 | zp = tdzp; | |
428870ff | 3581 | oidp = zp->z_id; |
34dc7c2f BB |
3582 | rwlp = &szp->z_parent_lock; |
3583 | rw = RW_WRITER; | |
3584 | continue; | |
3585 | } else { | |
3586 | /* | |
3587 | * Wait for other thread to drop its locks | |
3588 | */ | |
3589 | rw_enter(rwlp, rw); | |
3590 | } | |
3591 | } | |
3592 | ||
3593 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3594 | zl->zl_rwlock = rwlp; | |
3595 | zl->zl_znode = NULL; | |
3596 | zl->zl_next = *zlpp; | |
3597 | *zlpp = zl; | |
3598 | ||
428870ff | 3599 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 3600 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3601 | |
428870ff | 3602 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3603 | return (0); |
3604 | ||
3605 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3606 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3607 | if (error) |
3608 | return (error); | |
3609 | zl->zl_znode = zp; | |
3610 | } | |
3558fd73 | 3611 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3612 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3613 | rwlp = &zp->z_parent_lock; |
3614 | rw = RW_READER; | |
3615 | ||
3616 | } while (zp->z_id != sdzp->z_id); | |
3617 | ||
3618 | return (0); | |
3619 | } | |
3620 | ||
3621 | /* | |
3622 | * Move an entry from the provided source directory to the target | |
3623 | * directory. Change the entry name as indicated. | |
3624 | * | |
3558fd73 | 3625 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3626 | * snm - Old entry name. |
3558fd73 | 3627 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3628 | * tnm - New entry name. |
3629 | * cr - credentials of caller. | |
34dc7c2f BB |
3630 | * flags - case flags |
3631 | * | |
d3cc8b15 | 3632 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3633 | * |
3634 | * Timestamps: | |
3558fd73 | 3635 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3636 | */ |
3637 | /*ARGSUSED*/ | |
e5c39b95 | 3638 | int |
3558fd73 BB |
3639 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3640 | cred_t *cr, int flags) | |
34dc7c2f BB |
3641 | { |
3642 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 | 3643 | znode_t *sdzp = ITOZ(sdip); |
0037b49e | 3644 | zfsvfs_t *zfsvfs = ITOZSB(sdip); |
34dc7c2f | 3645 | zilog_t *zilog; |
34dc7c2f BB |
3646 | zfs_dirlock_t *sdl, *tdl; |
3647 | dmu_tx_t *tx; | |
3648 | zfs_zlock_t *zl; | |
3649 | int cmp, serr, terr; | |
3650 | int error = 0; | |
3651 | int zflg = 0; | |
e8b96c60 | 3652 | boolean_t waited = B_FALSE; |
34dc7c2f | 3653 | |
32dec7bd | 3654 | if (snm == NULL || tnm == NULL) |
3655 | return (SET_ERROR(EINVAL)); | |
3656 | ||
0037b49e | 3657 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 3658 | ZFS_VERIFY_ZP(sdzp); |
0037b49e | 3659 | zilog = zfsvfs->z_log; |
34dc7c2f | 3660 | |
812e91a7 MT |
3661 | tdzp = ITOZ(tdip); |
3662 | ZFS_VERIFY_ZP(tdzp); | |
3663 | ||
3664 | /* | |
3665 | * We check i_sb because snapshots and the ctldir must have different | |
3666 | * super blocks. | |
3667 | */ | |
c0ebc844 | 3668 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
0037b49e | 3669 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3670 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3671 | } |
3672 | ||
0037b49e | 3673 | if (zfsvfs->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3674 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 3675 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3676 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3677 | } |
3678 | ||
3679 | if (flags & FIGNORECASE) | |
3680 | zflg |= ZCILOOK; | |
3681 | ||
3682 | top: | |
3683 | szp = NULL; | |
3684 | tzp = NULL; | |
3685 | zl = NULL; | |
3686 | ||
3687 | /* | |
3688 | * This is to prevent the creation of links into attribute space | |
3689 | * by renaming a linked file into/outof an attribute directory. | |
3690 | * See the comment in zfs_link() for why this is considered bad. | |
3691 | */ | |
428870ff | 3692 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 3693 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3694 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3695 | } |
3696 | ||
3697 | /* | |
3698 | * Lock source and target directory entries. To prevent deadlock, | |
3699 | * a lock ordering must be defined. We lock the directory with | |
3700 | * the smallest object id first, or if it's a tie, the one with | |
3701 | * the lexically first name. | |
3702 | */ | |
3703 | if (sdzp->z_id < tdzp->z_id) { | |
3704 | cmp = -1; | |
3705 | } else if (sdzp->z_id > tdzp->z_id) { | |
3706 | cmp = 1; | |
3707 | } else { | |
3708 | /* | |
3709 | * First compare the two name arguments without | |
3710 | * considering any case folding. | |
3711 | */ | |
0037b49e | 3712 | int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3713 | |
3714 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
0037b49e | 3715 | ASSERT(error == 0 || !zfsvfs->z_utf8); |
34dc7c2f BB |
3716 | if (cmp == 0) { |
3717 | /* | |
3718 | * POSIX: "If the old argument and the new argument | |
3719 | * both refer to links to the same existing file, | |
3720 | * the rename() function shall return successfully | |
3721 | * and perform no other action." | |
3722 | */ | |
0037b49e | 3723 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3724 | return (0); |
3725 | } | |
3726 | /* | |
3727 | * If the file system is case-folding, then we may | |
3728 | * have some more checking to do. A case-folding file | |
3729 | * system is either supporting mixed case sensitivity | |
3730 | * access or is completely case-insensitive. Note | |
3731 | * that the file system is always case preserving. | |
3732 | * | |
3733 | * In mixed sensitivity mode case sensitive behavior | |
3734 | * is the default. FIGNORECASE must be used to | |
3735 | * explicitly request case insensitive behavior. | |
3736 | * | |
3737 | * If the source and target names provided differ only | |
3738 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3739 | * we will treat this as a special case in the | |
3740 | * case-insensitive mode: as long as the source name | |
3741 | * is an exact match, we will allow this to proceed as | |
3742 | * a name-change request. | |
3743 | */ | |
0037b49e BB |
3744 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
3745 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3746 | flags & FIGNORECASE)) && |
0037b49e | 3747 | u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3748 | &error) == 0) { |
3749 | /* | |
3750 | * case preserving rename request, require exact | |
3751 | * name matches | |
3752 | */ | |
3753 | zflg |= ZCIEXACT; | |
3754 | zflg &= ~ZCILOOK; | |
3755 | } | |
3756 | } | |
3757 | ||
428870ff BB |
3758 | /* |
3759 | * If the source and destination directories are the same, we should | |
3760 | * grab the z_name_lock of that directory only once. | |
3761 | */ | |
3762 | if (sdzp == tdzp) { | |
3763 | zflg |= ZHAVELOCK; | |
3764 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3765 | } | |
3766 | ||
34dc7c2f BB |
3767 | if (cmp < 0) { |
3768 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3769 | ZEXISTS | zflg, NULL, NULL); | |
3770 | terr = zfs_dirent_lock(&tdl, | |
3771 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3772 | } else { | |
3773 | terr = zfs_dirent_lock(&tdl, | |
3774 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3775 | serr = zfs_dirent_lock(&sdl, | |
3776 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3777 | NULL, NULL); | |
3778 | } | |
3779 | ||
3780 | if (serr) { | |
3781 | /* | |
3782 | * Source entry invalid or not there. | |
3783 | */ | |
3784 | if (!terr) { | |
3785 | zfs_dirent_unlock(tdl); | |
3786 | if (tzp) | |
3558fd73 | 3787 | iput(ZTOI(tzp)); |
34dc7c2f | 3788 | } |
428870ff BB |
3789 | |
3790 | if (sdzp == tdzp) | |
3791 | rw_exit(&sdzp->z_name_lock); | |
3792 | ||
34dc7c2f BB |
3793 | if (strcmp(snm, "..") == 0) |
3794 | serr = EINVAL; | |
0037b49e | 3795 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3796 | return (serr); |
3797 | } | |
3798 | if (terr) { | |
3799 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3800 | iput(ZTOI(szp)); |
428870ff BB |
3801 | |
3802 | if (sdzp == tdzp) | |
3803 | rw_exit(&sdzp->z_name_lock); | |
3804 | ||
34dc7c2f BB |
3805 | if (strcmp(tnm, "..") == 0) |
3806 | terr = EINVAL; | |
0037b49e | 3807 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3808 | return (terr); |
3809 | } | |
3810 | ||
9c5167d1 NF |
3811 | /* |
3812 | * If we are using project inheritance, means if the directory has | |
3813 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
3814 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
3815 | * such case, we only allow renames into our tree when the project | |
3816 | * IDs are the same. | |
3817 | */ | |
3818 | if (tdzp->z_pflags & ZFS_PROJINHERIT && | |
3819 | tdzp->z_projid != szp->z_projid) { | |
3820 | error = SET_ERROR(EXDEV); | |
3821 | goto out; | |
3822 | } | |
3823 | ||
34dc7c2f BB |
3824 | /* |
3825 | * Must have write access at the source to remove the old entry | |
3826 | * and write access at the target to create the new entry. | |
3827 | * Note that if target and source are the same, this can be | |
3828 | * done in a single check. | |
3829 | */ | |
3830 | ||
149e873a | 3831 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3832 | goto out; |
3833 | ||
3558fd73 | 3834 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3835 | /* |
3836 | * Check to make sure rename is valid. | |
3837 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3838 | */ | |
149e873a | 3839 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3840 | goto out; |
3841 | } | |
3842 | ||
3843 | /* | |
3844 | * Does target exist? | |
3845 | */ | |
3846 | if (tzp) { | |
3847 | /* | |
3848 | * Source and target must be the same type. | |
3849 | */ | |
3558fd73 BB |
3850 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3851 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
2e528b49 | 3852 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3853 | goto out; |
3854 | } | |
3855 | } else { | |
3558fd73 | 3856 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
2e528b49 | 3857 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3858 | goto out; |
3859 | } | |
3860 | } | |
3861 | /* | |
3862 | * POSIX dictates that when the source and target | |
3863 | * entries refer to the same file object, rename | |
3864 | * must do nothing and exit without error. | |
3865 | */ | |
3866 | if (szp->z_id == tzp->z_id) { | |
3867 | error = 0; | |
3868 | goto out; | |
3869 | } | |
3870 | } | |
3871 | ||
0037b49e | 3872 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
3873 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3874 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3875 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3876 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3877 | if (sdzp != tdzp) { |
3878 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3879 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3880 | } | |
3881 | if (tzp) { | |
3882 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3883 | zfs_sa_upgrade_txholds(tx, tzp); | |
3884 | } | |
3885 | ||
3886 | zfs_sa_upgrade_txholds(tx, szp); | |
0037b49e | 3887 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
0735ecb3 | 3888 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
3889 | if (error) { |
3890 | if (zl != NULL) | |
3891 | zfs_rename_unlock(&zl); | |
3892 | zfs_dirent_unlock(sdl); | |
3893 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3894 | |
3895 | if (sdzp == tdzp) | |
3896 | rw_exit(&sdzp->z_name_lock); | |
3897 | ||
fb5f0bc8 | 3898 | if (error == ERESTART) { |
e8b96c60 | 3899 | waited = B_TRUE; |
34dc7c2f BB |
3900 | dmu_tx_wait(tx); |
3901 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3902 | iput(ZTOI(szp)); |
3903 | if (tzp) | |
3904 | iput(ZTOI(tzp)); | |
34dc7c2f BB |
3905 | goto top; |
3906 | } | |
3907 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3908 | iput(ZTOI(szp)); |
3909 | if (tzp) | |
3910 | iput(ZTOI(tzp)); | |
0037b49e | 3911 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3912 | return (error); |
3913 | } | |
3914 | ||
3915 | if (tzp) /* Attempt to remove the existing target */ | |
3916 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3917 | ||
3918 | if (error == 0) { | |
3919 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3920 | if (error == 0) { | |
428870ff | 3921 | szp->z_pflags |= ZFS_AV_MODIFIED; |
9c5167d1 NF |
3922 | if (tdzp->z_pflags & ZFS_PROJINHERIT) |
3923 | szp->z_pflags |= ZFS_PROJINHERIT; | |
34dc7c2f | 3924 | |
0037b49e | 3925 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 3926 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c99c9001 | 3927 | ASSERT0(error); |
34dc7c2f | 3928 | |
428870ff BB |
3929 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3930 | if (error == 0) { | |
3931 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3932 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3933 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3934 | } else { |
3935 | /* | |
3936 | * At this point, we have successfully created | |
3937 | * the target name, but have failed to remove | |
3938 | * the source name. Since the create was done | |
3939 | * with the ZRENAMING flag, there are | |
3940 | * complications; for one, the link count is | |
3941 | * wrong. The easiest way to deal with this | |
3942 | * is to remove the newly created target, and | |
3943 | * return the original error. This must | |
3944 | * succeed; fortunately, it is very unlikely to | |
3945 | * fail, since we just created it. | |
3946 | */ | |
3947 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3948 | ZRENAMING, NULL), ==, 0); | |
3949 | } | |
599b8648 CC |
3950 | } else { |
3951 | /* | |
3952 | * If we had removed the existing target, subsequent | |
3953 | * call to zfs_link_create() to add back the same entry | |
3954 | * but, the new dnode (szp) should not fail. | |
3955 | */ | |
3956 | ASSERT(tzp == NULL); | |
34dc7c2f BB |
3957 | } |
3958 | } | |
3959 | ||
3960 | dmu_tx_commit(tx); | |
3961 | out: | |
3962 | if (zl != NULL) | |
3963 | zfs_rename_unlock(&zl); | |
3964 | ||
3965 | zfs_dirent_unlock(sdl); | |
3966 | zfs_dirent_unlock(tdl); | |
3967 | ||
960e08fe | 3968 | zfs_inode_update(sdzp); |
428870ff BB |
3969 | if (sdzp == tdzp) |
3970 | rw_exit(&sdzp->z_name_lock); | |
3971 | ||
960e08fe BB |
3972 | if (sdzp != tdzp) |
3973 | zfs_inode_update(tdzp); | |
428870ff | 3974 | |
960e08fe | 3975 | zfs_inode_update(szp); |
3558fd73 | 3976 | iput(ZTOI(szp)); |
960e08fe BB |
3977 | if (tzp) { |
3978 | zfs_inode_update(tzp); | |
3558fd73 | 3979 | iput(ZTOI(tzp)); |
960e08fe | 3980 | } |
34dc7c2f | 3981 | |
0037b49e | 3982 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3983 | zil_commit(zilog, 0); |
428870ff | 3984 | |
0037b49e | 3985 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3986 | return (error); |
3987 | } | |
3988 | ||
3989 | /* | |
3990 | * Insert the indicated symbolic reference entry into the directory. | |
3991 | * | |
3558fd73 | 3992 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
3993 | * link - Name for new symlink entry. |
3994 | * vap - Attributes of new entry. | |
3995 | * target - Target path of new symlink. | |
3558fd73 | 3996 | * |
34dc7c2f | 3997 | * cr - credentials of caller. |
34dc7c2f BB |
3998 | * flags - case flags |
3999 | * | |
d3cc8b15 | 4000 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4001 | * |
4002 | * Timestamps: | |
3558fd73 | 4003 | * dip - ctime|mtime updated |
34dc7c2f BB |
4004 | */ |
4005 | /*ARGSUSED*/ | |
e5c39b95 | 4006 | int |
3558fd73 BB |
4007 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
4008 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 4009 | { |
3558fd73 | 4010 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
4011 | zfs_dirlock_t *dl; |
4012 | dmu_tx_t *tx; | |
0037b49e | 4013 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 4014 | zilog_t *zilog; |
428870ff | 4015 | uint64_t len = strlen(link); |
34dc7c2f BB |
4016 | int error; |
4017 | int zflg = ZNEW; | |
9babb374 BB |
4018 | zfs_acl_ids_t acl_ids; |
4019 | boolean_t fuid_dirtied; | |
428870ff | 4020 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 4021 | boolean_t waited = B_FALSE; |
34dc7c2f | 4022 | |
3558fd73 | 4023 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 4024 | |
32dec7bd | 4025 | if (name == NULL) |
4026 | return (SET_ERROR(EINVAL)); | |
4027 | ||
0037b49e | 4028 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4029 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 4030 | zilog = zfsvfs->z_log; |
34dc7c2f | 4031 | |
0037b49e | 4032 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 4033 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 4034 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4035 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
4036 | } |
4037 | if (flags & FIGNORECASE) | |
4038 | zflg |= ZCILOOK; | |
34dc7c2f BB |
4039 | |
4040 | if (len > MAXPATHLEN) { | |
0037b49e | 4041 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4042 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
4043 | } |
4044 | ||
428870ff BB |
4045 | if ((error = zfs_acl_ids_create(dzp, 0, |
4046 | vap, cr, NULL, &acl_ids)) != 0) { | |
0037b49e | 4047 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4048 | return (error); |
4049 | } | |
4050 | top: | |
3558fd73 BB |
4051 | *ipp = NULL; |
4052 | ||
34dc7c2f BB |
4053 | /* |
4054 | * Attempt to lock directory; fail if entry already exists. | |
4055 | */ | |
4056 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
4057 | if (error) { | |
428870ff | 4058 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 4059 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4060 | return (error); |
4061 | } | |
4062 | ||
149e873a | 4063 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
4064 | zfs_acl_ids_free(&acl_ids); |
4065 | zfs_dirent_unlock(dl); | |
0037b49e | 4066 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4067 | return (error); |
4068 | } | |
4069 | ||
9c5167d1 | 4070 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, ZFS_DEFAULT_PROJID)) { |
9babb374 BB |
4071 | zfs_acl_ids_free(&acl_ids); |
4072 | zfs_dirent_unlock(dl); | |
0037b49e | 4073 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4074 | return (SET_ERROR(EDQUOT)); |
9babb374 | 4075 | } |
0037b49e BB |
4076 | tx = dmu_tx_create(zfsvfs->z_os); |
4077 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
34dc7c2f | 4078 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 4079 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
4080 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
4081 | ZFS_SA_BASE_ATTR_SIZE + len); | |
4082 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
0037b49e | 4083 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
4084 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
4085 | acl_ids.z_aclp->z_acl_bytes); | |
4086 | } | |
9babb374 | 4087 | if (fuid_dirtied) |
0037b49e | 4088 | zfs_fuid_txhold(zfsvfs, tx); |
0735ecb3 | 4089 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
4090 | if (error) { |
4091 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4092 | if (error == ERESTART) { |
e8b96c60 | 4093 | waited = B_TRUE; |
34dc7c2f BB |
4094 | dmu_tx_wait(tx); |
4095 | dmu_tx_abort(tx); | |
4096 | goto top; | |
4097 | } | |
428870ff | 4098 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 4099 | dmu_tx_abort(tx); |
0037b49e | 4100 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4101 | return (error); |
4102 | } | |
4103 | ||
34dc7c2f BB |
4104 | /* |
4105 | * Create a new object for the symlink. | |
428870ff | 4106 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 4107 | */ |
428870ff | 4108 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 4109 | |
428870ff | 4110 | if (fuid_dirtied) |
0037b49e | 4111 | zfs_fuid_sync(zfsvfs, tx); |
34dc7c2f | 4112 | |
572e2857 | 4113 | mutex_enter(&zp->z_lock); |
428870ff | 4114 | if (zp->z_is_sa) |
0037b49e | 4115 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), |
428870ff BB |
4116 | link, len, tx); |
4117 | else | |
4118 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 4119 | mutex_exit(&zp->z_lock); |
34dc7c2f | 4120 | |
428870ff | 4121 | zp->z_size = len; |
0037b49e | 4122 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 4123 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
4124 | /* |
4125 | * Insert the new object into the directory. | |
4126 | */ | |
599b8648 CC |
4127 | error = zfs_link_create(dl, zp, tx, ZNEW); |
4128 | if (error != 0) { | |
4129 | zfs_znode_delete(zp, tx); | |
4130 | remove_inode_hash(ZTOI(zp)); | |
4131 | } else { | |
4132 | if (flags & FIGNORECASE) | |
4133 | txtype |= TX_CI; | |
4134 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
4f301661 | 4135 | |
599b8648 CC |
4136 | zfs_inode_update(dzp); |
4137 | zfs_inode_update(zp); | |
4138 | } | |
960e08fe | 4139 | |
9babb374 | 4140 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
4141 | |
4142 | dmu_tx_commit(tx); | |
4143 | ||
4144 | zfs_dirent_unlock(dl); | |
4145 | ||
599b8648 CC |
4146 | if (error == 0) { |
4147 | *ipp = ZTOI(zp); | |
34dc7c2f | 4148 | |
599b8648 CC |
4149 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
4150 | zil_commit(zilog, 0); | |
4151 | } else { | |
4152 | iput(ZTOI(zp)); | |
4153 | } | |
428870ff | 4154 | |
0037b49e | 4155 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4156 | return (error); |
4157 | } | |
4158 | ||
4159 | /* | |
4160 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 4161 | * the symbolic path referred to by ip. |
34dc7c2f | 4162 | * |
8b4f9a2d BB |
4163 | * IN: ip - inode of symbolic link |
4164 | * uio - structure to contain the link path. | |
4165 | * cr - credentials of caller. | |
34dc7c2f BB |
4166 | * |
4167 | * RETURN: 0 if success | |
4168 | * error code if failure | |
4169 | * | |
4170 | * Timestamps: | |
3558fd73 | 4171 | * ip - atime updated |
34dc7c2f BB |
4172 | */ |
4173 | /* ARGSUSED */ | |
e5c39b95 | 4174 | int |
8b4f9a2d | 4175 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 4176 | { |
3558fd73 | 4177 | znode_t *zp = ITOZ(ip); |
0037b49e | 4178 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4179 | int error; |
4180 | ||
0037b49e | 4181 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4182 | ZFS_VERIFY_ZP(zp); |
4183 | ||
572e2857 | 4184 | mutex_enter(&zp->z_lock); |
428870ff | 4185 | if (zp->z_is_sa) |
8b4f9a2d | 4186 | error = sa_lookup_uio(zp->z_sa_hdl, |
0037b49e | 4187 | SA_ZPL_SYMLINK(zfsvfs), uio); |
428870ff | 4188 | else |
8b4f9a2d | 4189 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 4190 | mutex_exit(&zp->z_lock); |
34dc7c2f | 4191 | |
0037b49e | 4192 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4193 | return (error); |
4194 | } | |
4195 | ||
4196 | /* | |
3558fd73 | 4197 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 4198 | * |
3558fd73 BB |
4199 | * IN: tdip - Directory to contain new entry. |
4200 | * sip - inode of new entry. | |
34dc7c2f BB |
4201 | * name - name of new entry. |
4202 | * cr - credentials of caller. | |
34dc7c2f BB |
4203 | * |
4204 | * RETURN: 0 if success | |
4205 | * error code if failure | |
4206 | * | |
4207 | * Timestamps: | |
3558fd73 BB |
4208 | * tdip - ctime|mtime updated |
4209 | * sip - ctime updated | |
34dc7c2f BB |
4210 | */ |
4211 | /* ARGSUSED */ | |
e5c39b95 | 4212 | int |
da5e151f BB |
4213 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr, |
4214 | int flags) | |
34dc7c2f | 4215 | { |
3558fd73 | 4216 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 4217 | znode_t *tzp, *szp; |
0037b49e | 4218 | zfsvfs_t *zfsvfs = ITOZSB(tdip); |
34dc7c2f BB |
4219 | zilog_t *zilog; |
4220 | zfs_dirlock_t *dl; | |
4221 | dmu_tx_t *tx; | |
34dc7c2f BB |
4222 | int error; |
4223 | int zf = ZNEW; | |
428870ff | 4224 | uint64_t parent; |
572e2857 | 4225 | uid_t owner; |
e8b96c60 | 4226 | boolean_t waited = B_FALSE; |
ace1eae8 CC |
4227 | boolean_t is_tmpfile = 0; |
4228 | uint64_t txg; | |
4229 | #ifdef HAVE_TMPFILE | |
4230 | is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); | |
4231 | #endif | |
3558fd73 | 4232 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 4233 | |
32dec7bd | 4234 | if (name == NULL) |
4235 | return (SET_ERROR(EINVAL)); | |
4236 | ||
0037b49e | 4237 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4238 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 4239 | zilog = zfsvfs->z_log; |
34dc7c2f | 4240 | |
428870ff BB |
4241 | /* |
4242 | * POSIX dictates that we return EPERM here. | |
4243 | * Better choices include ENOTSUP or EISDIR. | |
4244 | */ | |
3558fd73 | 4245 | if (S_ISDIR(sip->i_mode)) { |
0037b49e | 4246 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4247 | return (SET_ERROR(EPERM)); |
428870ff BB |
4248 | } |
4249 | ||
812e91a7 MT |
4250 | szp = ITOZ(sip); |
4251 | ZFS_VERIFY_ZP(szp); | |
4252 | ||
9c5167d1 NF |
4253 | /* |
4254 | * If we are using project inheritance, means if the directory has | |
4255 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
4256 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
4257 | * such case, we only allow hard link creation in our tree when the | |
4258 | * project IDs are the same. | |
4259 | */ | |
4260 | if (dzp->z_pflags & ZFS_PROJINHERIT && dzp->z_projid != szp->z_projid) { | |
4261 | ZFS_EXIT(zfsvfs); | |
4262 | return (SET_ERROR(EXDEV)); | |
4263 | } | |
4264 | ||
812e91a7 MT |
4265 | /* |
4266 | * We check i_sb because snapshots and the ctldir must have different | |
4267 | * super blocks. | |
4268 | */ | |
c0ebc844 | 4269 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
0037b49e | 4270 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4271 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 4272 | } |
428870ff | 4273 | |
428870ff BB |
4274 | /* Prevent links to .zfs/shares files */ |
4275 | ||
0037b49e | 4276 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff | 4277 | &parent, sizeof (uint64_t))) != 0) { |
0037b49e | 4278 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4279 | return (error); |
4280 | } | |
0037b49e BB |
4281 | if (parent == zfsvfs->z_shares_dir) { |
4282 | ZFS_EXIT(zfsvfs); | |
2e528b49 | 4283 | return (SET_ERROR(EPERM)); |
428870ff BB |
4284 | } |
4285 | ||
0037b49e | 4286 | if (zfsvfs->z_utf8 && u8_validate(name, |
34dc7c2f | 4287 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 4288 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4289 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
4290 | } |
4291 | if (flags & FIGNORECASE) | |
4292 | zf |= ZCILOOK; | |
4293 | ||
34dc7c2f BB |
4294 | /* |
4295 | * We do not support links between attributes and non-attributes | |
4296 | * because of the potential security risk of creating links | |
4297 | * into "normal" file space in order to circumvent restrictions | |
4298 | * imposed in attribute space. | |
4299 | */ | |
428870ff | 4300 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 4301 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4302 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4303 | } |
4304 | ||
0037b49e BB |
4305 | owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid), |
4306 | cr, ZFS_OWNER); | |
572e2857 | 4307 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
0037b49e | 4308 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4309 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4310 | } |
4311 | ||
149e873a | 4312 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
0037b49e | 4313 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4314 | return (error); |
4315 | } | |
4316 | ||
428870ff | 4317 | top: |
34dc7c2f BB |
4318 | /* |
4319 | * Attempt to lock directory; fail if entry already exists. | |
4320 | */ | |
4321 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
4322 | if (error) { | |
0037b49e | 4323 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4324 | return (error); |
4325 | } | |
4326 | ||
0037b49e | 4327 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 4328 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 4329 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
ace1eae8 | 4330 | if (is_tmpfile) |
0037b49e | 4331 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 4332 | |
428870ff BB |
4333 | zfs_sa_upgrade_txholds(tx, szp); |
4334 | zfs_sa_upgrade_txholds(tx, dzp); | |
0735ecb3 | 4335 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
4336 | if (error) { |
4337 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4338 | if (error == ERESTART) { |
e8b96c60 | 4339 | waited = B_TRUE; |
34dc7c2f BB |
4340 | dmu_tx_wait(tx); |
4341 | dmu_tx_abort(tx); | |
4342 | goto top; | |
4343 | } | |
4344 | dmu_tx_abort(tx); | |
0037b49e | 4345 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4346 | return (error); |
4347 | } | |
ace1eae8 CC |
4348 | /* unmark z_unlinked so zfs_link_create will not reject */ |
4349 | if (is_tmpfile) | |
4350 | szp->z_unlinked = 0; | |
34dc7c2f BB |
4351 | error = zfs_link_create(dl, szp, tx, 0); |
4352 | ||
4353 | if (error == 0) { | |
4354 | uint64_t txtype = TX_LINK; | |
ace1eae8 CC |
4355 | /* |
4356 | * tmpfile is created to be in z_unlinkedobj, so remove it. | |
4357 | * Also, we don't log in ZIL, be cause all previous file | |
4358 | * operation on the tmpfile are ignored by ZIL. Instead we | |
4359 | * always wait for txg to sync to make sure all previous | |
4360 | * operation are sync safe. | |
4361 | */ | |
4362 | if (is_tmpfile) { | |
0037b49e BB |
4363 | VERIFY(zap_remove_int(zfsvfs->z_os, |
4364 | zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); | |
ace1eae8 CC |
4365 | } else { |
4366 | if (flags & FIGNORECASE) | |
4367 | txtype |= TX_CI; | |
4368 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); | |
4369 | } | |
4370 | } else if (is_tmpfile) { | |
4371 | /* restore z_unlinked since when linking failed */ | |
4372 | szp->z_unlinked = 1; | |
34dc7c2f | 4373 | } |
ace1eae8 | 4374 | txg = dmu_tx_get_txg(tx); |
34dc7c2f BB |
4375 | dmu_tx_commit(tx); |
4376 | ||
4377 | zfs_dirent_unlock(dl); | |
4378 | ||
0037b49e | 4379 | if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4380 | zil_commit(zilog, 0); |
428870ff | 4381 | |
ace1eae8 | 4382 | if (is_tmpfile) |
0037b49e | 4383 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg); |
ace1eae8 | 4384 | |
960e08fe BB |
4385 | zfs_inode_update(dzp); |
4386 | zfs_inode_update(szp); | |
0037b49e | 4387 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4388 | return (error); |
4389 | } | |
4390 | ||
3c0e5c0f | 4391 | static void |
119a394a | 4392 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
4393 | { |
4394 | struct page *pp = arg; | |
4395 | ||
119a394a | 4396 | ClearPageError(pp); |
3c0e5c0f BB |
4397 | end_page_writeback(pp); |
4398 | } | |
4399 | ||
34dc7c2f | 4400 | /* |
3c0e5c0f BB |
4401 | * Push a page out to disk, once the page is on stable storage the |
4402 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 4403 | * |
3c0e5c0f BB |
4404 | * IN: ip - page mapped for inode. |
4405 | * pp - page to push (page is locked) | |
4406 | * wbc - writeback control data | |
34dc7c2f BB |
4407 | * |
4408 | * RETURN: 0 if success | |
4409 | * error code if failure | |
4410 | * | |
3c0e5c0f BB |
4411 | * Timestamps: |
4412 | * ip - ctime|mtime updated | |
34dc7c2f BB |
4413 | */ |
4414 | /* ARGSUSED */ | |
3c0e5c0f BB |
4415 | int |
4416 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 4417 | { |
3c0e5c0f | 4418 | znode_t *zp = ITOZ(ip); |
0037b49e | 4419 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
3c0e5c0f BB |
4420 | loff_t offset; |
4421 | loff_t pgoff; | |
4c837f0d | 4422 | unsigned int pglen; |
3c0e5c0f BB |
4423 | dmu_tx_t *tx; |
4424 | caddr_t va; | |
4425 | int err = 0; | |
4426 | uint64_t mtime[2], ctime[2]; | |
4427 | sa_bulk_attr_t bulk[3]; | |
4428 | int cnt = 0; | |
21a96fb6 | 4429 | struct address_space *mapping; |
3c0e5c0f | 4430 | |
0037b49e | 4431 | ZFS_ENTER(zfsvfs); |
4c837f0d | 4432 | ZFS_VERIFY_ZP(zp); |
d164b209 | 4433 | |
3c0e5c0f BB |
4434 | ASSERT(PageLocked(pp)); |
4435 | ||
d1d7e268 MK |
4436 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
4437 | offset = i_size_read(ip); /* File length in bytes */ | |
8b1899d3 BB |
4438 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
4439 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
4440 | |
4441 | /* Page is beyond end of file */ | |
4442 | if (pgoff >= offset) { | |
4443 | unlock_page(pp); | |
0037b49e | 4444 | ZFS_EXIT(zfsvfs); |
3c0e5c0f BB |
4445 | return (0); |
4446 | } | |
4447 | ||
4448 | /* Truncate page length to end of file */ | |
4449 | if (pgoff + pglen > offset) | |
4450 | pglen = offset - pgoff; | |
4451 | ||
4452 | #if 0 | |
34dc7c2f | 4453 | /* |
3c0e5c0f BB |
4454 | * FIXME: Allow mmap writes past its quota. The correct fix |
4455 | * is to register a page_mkwrite() handler to count the page | |
4456 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 4457 | */ |
9c5167d1 NF |
4458 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
4459 | KUID_TO_SUID(ip->i_uid)) || | |
4460 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
4461 | KGID_TO_SGID(ip->i_gid)) || | |
4462 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
4463 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
4464 | zp->z_projid))) { | |
9babb374 | 4465 | err = EDQUOT; |
9babb374 | 4466 | } |
3c0e5c0f BB |
4467 | #endif |
4468 | ||
d958324f BB |
4469 | /* |
4470 | * The ordering here is critical and must adhere to the following | |
4471 | * rules in order to avoid deadlocking in either zfs_read() or | |
4472 | * zfs_free_range() due to a lock inversion. | |
4473 | * | |
4474 | * 1) The page must be unlocked prior to acquiring the range lock. | |
4475 | * This is critical because zfs_read() calls find_lock_page() | |
4476 | * which may block on the page lock while holding the range lock. | |
4477 | * | |
4478 | * 2) Before setting or clearing write back on a page the range lock | |
4479 | * must be held in order to prevent a lock inversion with the | |
4480 | * zfs_free_range() function. | |
21a96fb6 CC |
4481 | * |
4482 | * This presents a problem because upon entering this function the | |
4483 | * page lock is already held. To safely acquire the range lock the | |
4484 | * page lock must be dropped. This creates a window where another | |
4485 | * process could truncate, invalidate, dirty, or write out the page. | |
4486 | * | |
4487 | * Therefore, after successfully reacquiring the range and page locks | |
4488 | * the current page state is checked. In the common case everything | |
4489 | * will be as is expected and it can be written out. However, if | |
4490 | * the page state has changed it must be handled accordingly. | |
d958324f | 4491 | */ |
21a96fb6 CC |
4492 | mapping = pp->mapping; |
4493 | redirty_page_for_writepage(wbc, pp); | |
d958324f | 4494 | unlock_page(pp); |
21a96fb6 | 4495 | |
5d43cc9a MA |
4496 | locked_range_t *lr = rangelock_enter(&zp->z_rangelock, |
4497 | pgoff, pglen, RL_WRITER); | |
21a96fb6 CC |
4498 | lock_page(pp); |
4499 | ||
4500 | /* Page mapping changed or it was no longer dirty, we're done */ | |
4501 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
4502 | unlock_page(pp); | |
5d43cc9a | 4503 | rangelock_exit(lr); |
0037b49e | 4504 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4505 | return (0); |
4506 | } | |
4507 | ||
4508 | /* Another process started write block if required */ | |
4509 | if (PageWriteback(pp)) { | |
4510 | unlock_page(pp); | |
5d43cc9a | 4511 | rangelock_exit(lr); |
21a96fb6 CC |
4512 | |
4513 | if (wbc->sync_mode != WB_SYNC_NONE) | |
4514 | wait_on_page_writeback(pp); | |
4515 | ||
0037b49e | 4516 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4517 | return (0); |
4518 | } | |
4519 | ||
4520 | /* Clear the dirty flag the required locks are held */ | |
4521 | if (!clear_page_dirty_for_io(pp)) { | |
4522 | unlock_page(pp); | |
5d43cc9a | 4523 | rangelock_exit(lr); |
0037b49e | 4524 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4525 | return (0); |
4526 | } | |
4527 | ||
4528 | /* | |
4529 | * Counterpart for redirty_page_for_writepage() above. This page | |
4530 | * was in fact not skipped and should not be counted as if it were. | |
4531 | */ | |
4532 | wbc->pages_skipped--; | |
3c0e5c0f | 4533 | set_page_writeback(pp); |
21a96fb6 | 4534 | unlock_page(pp); |
3c0e5c0f | 4535 | |
0037b49e | 4536 | tx = dmu_tx_create(zfsvfs->z_os); |
3c0e5c0f | 4537 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
4538 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4539 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 4540 | |
fb5f0bc8 | 4541 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 4542 | if (err != 0) { |
3c0e5c0f | 4543 | if (err == ERESTART) |
34dc7c2f | 4544 | dmu_tx_wait(tx); |
3c0e5c0f | 4545 | |
34dc7c2f | 4546 | dmu_tx_abort(tx); |
119a394a ED |
4547 | __set_page_dirty_nobuffers(pp); |
4548 | ClearPageError(pp); | |
4549 | end_page_writeback(pp); | |
5d43cc9a | 4550 | rangelock_exit(lr); |
0037b49e | 4551 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4552 | return (err); |
34dc7c2f BB |
4553 | } |
4554 | ||
dde471ef | 4555 | va = kmap(pp); |
8b1899d3 | 4556 | ASSERT3U(pglen, <=, PAGE_SIZE); |
0037b49e | 4557 | dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 4558 | kunmap(pp); |
34dc7c2f | 4559 | |
0037b49e BB |
4560 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
4561 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
4562 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, | |
4563 | &zp->z_pflags, 8); | |
428870ff | 4564 | |
d3aa3ea9 BB |
4565 | /* Preserve the mtime and ctime provided by the inode */ |
4566 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4567 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
4568 | zp->z_atime_dirty = 0; | |
4569 | zp->z_seq++; | |
4570 | ||
4571 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4572 | ||
0037b49e | 4573 | zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
119a394a | 4574 | zfs_putpage_commit_cb, pp); |
45d1cae3 | 4575 | dmu_tx_commit(tx); |
d3aa3ea9 | 4576 | |
5d43cc9a | 4577 | rangelock_exit(lr); |
34dc7c2f | 4578 | |
119a394a ED |
4579 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4580 | /* | |
4581 | * Note that this is rarely called under writepages(), because | |
4582 | * writepages() normally handles the entire commit for | |
4583 | * performance reasons. | |
4584 | */ | |
0037b49e | 4585 | zil_commit(zfsvfs->z_log, zp->z_id); |
2b286136 | 4586 | } |
3c0e5c0f | 4587 | |
0037b49e | 4588 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4589 | return (err); |
34dc7c2f BB |
4590 | } |
4591 | ||
8780c539 BB |
4592 | /* |
4593 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 4594 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
4595 | */ |
4596 | int | |
4597 | zfs_dirty_inode(struct inode *ip, int flags) | |
4598 | { | |
4599 | znode_t *zp = ITOZ(ip); | |
0037b49e | 4600 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
8780c539 | 4601 | dmu_tx_t *tx; |
023699cd MM |
4602 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
4603 | sa_bulk_attr_t bulk[4]; | |
704cd075 | 4604 | int error = 0; |
8780c539 BB |
4605 | int cnt = 0; |
4606 | ||
0037b49e | 4607 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
c944be5d BB |
4608 | return (0); |
4609 | ||
0037b49e | 4610 | ZFS_ENTER(zfsvfs); |
8780c539 BB |
4611 | ZFS_VERIFY_ZP(zp); |
4612 | ||
704cd075 CC |
4613 | #ifdef I_DIRTY_TIME |
4614 | /* | |
4615 | * This is the lazytime semantic indroduced in Linux 4.0 | |
4616 | * This flag will only be called from update_time when lazytime is set. | |
4617 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
4618 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
4619 | * only need to dirty atime. | |
4620 | */ | |
4621 | if (flags == I_DIRTY_TIME) { | |
4622 | zp->z_atime_dirty = 1; | |
4623 | goto out; | |
4624 | } | |
4625 | #endif | |
4626 | ||
0037b49e | 4627 | tx = dmu_tx_create(zfsvfs->z_os); |
8780c539 BB |
4628 | |
4629 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4630 | zfs_sa_upgrade_txholds(tx, zp); | |
4631 | ||
4632 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4633 | if (error) { | |
4634 | dmu_tx_abort(tx); | |
4635 | goto out; | |
4636 | } | |
4637 | ||
4638 | mutex_enter(&zp->z_lock); | |
704cd075 CC |
4639 | zp->z_atime_dirty = 0; |
4640 | ||
0037b49e BB |
4641 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
4642 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
4643 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
4644 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
8780c539 | 4645 | |
023699cd | 4646 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4647 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4648 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4649 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
023699cd MM |
4650 | mode = ip->i_mode; |
4651 | ||
4652 | zp->z_mode = mode; | |
8780c539 BB |
4653 | |
4654 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4655 | mutex_exit(&zp->z_lock); | |
4656 | ||
4657 | dmu_tx_commit(tx); | |
4658 | out: | |
0037b49e | 4659 | ZFS_EXIT(zfsvfs); |
8780c539 BB |
4660 | return (error); |
4661 | } | |
8780c539 | 4662 | |
34dc7c2f BB |
4663 | /*ARGSUSED*/ |
4664 | void | |
c0d35759 | 4665 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4666 | { |
c0d35759 | 4667 | znode_t *zp = ITOZ(ip); |
0037b49e | 4668 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
0df9673f | 4669 | uint64_t atime[2]; |
34dc7c2f | 4670 | int error; |
cafbd2ac | 4671 | int need_unlock = 0; |
34dc7c2f | 4672 | |
cafbd2ac | 4673 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
0037b49e | 4674 | if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) { |
cafbd2ac | 4675 | need_unlock = 1; |
0037b49e | 4676 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); |
cafbd2ac | 4677 | } |
c0d35759 | 4678 | if (zp->z_sa_hdl == NULL) { |
cafbd2ac | 4679 | if (need_unlock) |
0037b49e | 4680 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
c0d35759 | 4681 | return; |
34dc7c2f BB |
4682 | } |
4683 | ||
4684 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
0037b49e | 4685 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 4686 | |
428870ff BB |
4687 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4688 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4689 | error = dmu_tx_assign(tx, TXG_WAIT); |
4690 | if (error) { | |
4691 | dmu_tx_abort(tx); | |
4692 | } else { | |
0df9673f | 4693 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
34dc7c2f | 4694 | mutex_enter(&zp->z_lock); |
0037b49e | 4695 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), |
0df9673f | 4696 | (void *)&atime, sizeof (atime), tx); |
34dc7c2f BB |
4697 | zp->z_atime_dirty = 0; |
4698 | mutex_exit(&zp->z_lock); | |
4699 | dmu_tx_commit(tx); | |
4700 | } | |
4701 | } | |
4702 | ||
4703 | zfs_zinactive(zp); | |
cafbd2ac | 4704 | if (need_unlock) |
0037b49e | 4705 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
34dc7c2f BB |
4706 | } |
4707 | ||
4708 | /* | |
4709 | * Bounds-check the seek operation. | |
4710 | * | |
3558fd73 | 4711 | * IN: ip - inode seeking within |
34dc7c2f BB |
4712 | * ooff - old file offset |
4713 | * noffp - pointer to new file offset | |
4714 | * ct - caller context | |
4715 | * | |
4716 | * RETURN: 0 if success | |
4717 | * EINVAL if new offset invalid | |
4718 | */ | |
4719 | /* ARGSUSED */ | |
3558fd73 | 4720 | int |
9623f736 | 4721 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4722 | { |
3558fd73 | 4723 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4724 | return (0); |
4725 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4726 | } | |
4727 | ||
34dc7c2f | 4728 | /* |
dde471ef | 4729 | * Fill pages with data from the disk. |
34dc7c2f BB |
4730 | */ |
4731 | static int | |
dde471ef | 4732 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4733 | { |
d1d7e268 | 4734 | znode_t *zp = ITOZ(ip); |
0037b49e | 4735 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
d1d7e268 | 4736 | objset_t *os; |
dde471ef | 4737 | struct page *cur_pp; |
d1d7e268 MK |
4738 | u_offset_t io_off, total; |
4739 | size_t io_len; | |
4740 | loff_t i_size; | |
4741 | unsigned page_idx; | |
4742 | int err; | |
34dc7c2f | 4743 | |
0037b49e | 4744 | os = zfsvfs->z_os; |
8b1899d3 | 4745 | io_len = nr_pages << PAGE_SHIFT; |
dde471ef PJ |
4746 | i_size = i_size_read(ip); |
4747 | io_off = page_offset(pl[0]); | |
4748 | ||
4749 | if (io_off + io_len > i_size) | |
4750 | io_len = i_size - io_off; | |
34dc7c2f BB |
4751 | |
4752 | /* | |
dde471ef | 4753 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4754 | */ |
dde471ef | 4755 | page_idx = 0; |
34dc7c2f | 4756 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4757 | caddr_t va; |
4758 | ||
540c3927 | 4759 | cur_pp = pl[page_idx++]; |
dde471ef | 4760 | va = kmap(cur_pp); |
9babb374 BB |
4761 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4762 | DMU_READ_PREFETCH); | |
dde471ef | 4763 | kunmap(cur_pp); |
34dc7c2f | 4764 | if (err) { |
b128c09f BB |
4765 | /* convert checksum errors into IO errors */ |
4766 | if (err == ECKSUM) | |
2e528b49 | 4767 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4768 | return (err); |
4769 | } | |
34dc7c2f | 4770 | } |
d164b209 | 4771 | |
34dc7c2f BB |
4772 | return (0); |
4773 | } | |
4774 | ||
4775 | /* | |
dde471ef | 4776 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4777 | * |
dde471ef PJ |
4778 | * IN: ip - inode of file to get data from. |
4779 | * pl - list of pages to read | |
4780 | * nr_pages - number of pages to read | |
34dc7c2f | 4781 | * |
d3cc8b15 | 4782 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4783 | * |
4784 | * Timestamps: | |
4785 | * vp - atime updated | |
4786 | */ | |
4787 | /* ARGSUSED */ | |
dde471ef PJ |
4788 | int |
4789 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4790 | { |
dde471ef | 4791 | znode_t *zp = ITOZ(ip); |
0037b49e | 4792 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
dde471ef | 4793 | int err; |
d164b209 | 4794 | |
d164b209 BB |
4795 | if (pl == NULL) |
4796 | return (0); | |
34dc7c2f | 4797 | |
0037b49e | 4798 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4799 | ZFS_VERIFY_ZP(zp); |
4800 | ||
dde471ef | 4801 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4802 | |
0037b49e | 4803 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4804 | return (err); |
4805 | } | |
4806 | ||
4807 | /* | |
e2e7aa2d | 4808 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4809 | * |
e2e7aa2d BB |
4810 | * IN: ip - inode of the file to mmap |
4811 | * off - file offset | |
4812 | * addrp - start address in memory region | |
4813 | * len - length of memory region | |
4814 | * vm_flags- address flags | |
34dc7c2f | 4815 | * |
e2e7aa2d BB |
4816 | * RETURN: 0 if success |
4817 | * error code if failure | |
34dc7c2f BB |
4818 | */ |
4819 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4820 | int |
4821 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4822 | unsigned long vm_flags) | |
34dc7c2f | 4823 | { |
e2e7aa2d | 4824 | znode_t *zp = ITOZ(ip); |
0037b49e | 4825 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4826 | |
0037b49e | 4827 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4828 | ZFS_VERIFY_ZP(zp); |
4829 | ||
e2e7aa2d | 4830 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4831 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
0037b49e | 4832 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4833 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4834 | } |
4835 | ||
e2e7aa2d | 4836 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4837 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
0037b49e | 4838 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4839 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4840 | } |
4841 | ||
34dc7c2f | 4842 | if (off < 0 || len > MAXOFFSET_T - off) { |
0037b49e | 4843 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4844 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4845 | } |
4846 | ||
0037b49e | 4847 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4848 | return (0); |
4849 | } | |
4850 | ||
3558fd73 BB |
4851 | /* |
4852 | * convoff - converts the given data (start, whence) to the | |
4853 | * given whence. | |
4854 | */ | |
4855 | int | |
4856 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4857 | { | |
5484965a | 4858 | vattr_t vap; |
3558fd73 BB |
4859 | int error; |
4860 | ||
4861 | if ((lckdat->l_whence == 2) || (whence == 2)) { | |
d95a5980 | 4862 | if ((error = zfs_getattr(ip, &vap, 0, CRED()))) |
3558fd73 BB |
4863 | return (error); |
4864 | } | |
4865 | ||
4866 | switch (lckdat->l_whence) { | |
4867 | case 1: | |
4868 | lckdat->l_start += offset; | |
4869 | break; | |
4870 | case 2: | |
5484965a | 4871 | lckdat->l_start += vap.va_size; |
3558fd73 BB |
4872 | /* FALLTHRU */ |
4873 | case 0: | |
4874 | break; | |
4875 | default: | |
2e528b49 | 4876 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4877 | } |
4878 | ||
4879 | if (lckdat->l_start < 0) | |
2e528b49 | 4880 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4881 | |
4882 | switch (whence) { | |
4883 | case 1: | |
4884 | lckdat->l_start -= offset; | |
4885 | break; | |
4886 | case 2: | |
5484965a | 4887 | lckdat->l_start -= vap.va_size; |
3558fd73 BB |
4888 | /* FALLTHRU */ |
4889 | case 0: | |
4890 | break; | |
4891 | default: | |
2e528b49 | 4892 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4893 | } |
4894 | ||
4895 | lckdat->l_whence = (short)whence; | |
4896 | return (0); | |
4897 | } | |
4898 | ||
34dc7c2f BB |
4899 | /* |
4900 | * Free or allocate space in a file. Currently, this function only | |
4901 | * supports the `F_FREESP' command. However, this command is somewhat | |
4902 | * misnamed, as its functionality includes the ability to allocate as | |
4903 | * well as free space. | |
4904 | * | |
3558fd73 | 4905 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4906 | * cmd - action to take (only F_FREESP supported). |
4907 | * bfp - section of file to free/alloc. | |
4908 | * flag - current file open mode flags. | |
4909 | * offset - current file offset. | |
4910 | * cr - credentials of caller [UNUSED]. | |
34dc7c2f | 4911 | * |
d3cc8b15 | 4912 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4913 | * |
4914 | * Timestamps: | |
3558fd73 | 4915 | * ip - ctime|mtime updated |
34dc7c2f BB |
4916 | */ |
4917 | /* ARGSUSED */ | |
e5c39b95 | 4918 | int |
3558fd73 BB |
4919 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4920 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4921 | { |
3558fd73 | 4922 | znode_t *zp = ITOZ(ip); |
0037b49e | 4923 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4924 | uint64_t off, len; |
4925 | int error; | |
4926 | ||
0037b49e | 4927 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4928 | ZFS_VERIFY_ZP(zp); |
4929 | ||
34dc7c2f | 4930 | if (cmd != F_FREESP) { |
0037b49e | 4931 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4932 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4933 | } |
4934 | ||
f3c9dca0 MT |
4935 | /* |
4936 | * Callers might not be able to detect properly that we are read-only, | |
4937 | * so check it explicitly here. | |
4938 | */ | |
0037b49e BB |
4939 | if (zfs_is_readonly(zfsvfs)) { |
4940 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
4941 | return (SET_ERROR(EROFS)); |
4942 | } | |
4943 | ||
3558fd73 | 4944 | if ((error = convoff(ip, bfp, 0, offset))) { |
0037b49e | 4945 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4946 | return (error); |
4947 | } | |
4948 | ||
4949 | if (bfp->l_len < 0) { | |
0037b49e | 4950 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4951 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4952 | } |
4953 | ||
aec69371 ED |
4954 | /* |
4955 | * Permissions aren't checked on Solaris because on this OS | |
4956 | * zfs_space() can only be called with an opened file handle. | |
4957 | * On Linux we can get here through truncate_range() which | |
4958 | * operates directly on inodes, so we need to check access rights. | |
4959 | */ | |
4960 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
0037b49e | 4961 | ZFS_EXIT(zfsvfs); |
aec69371 ED |
4962 | return (error); |
4963 | } | |
4964 | ||
34dc7c2f BB |
4965 | off = bfp->l_start; |
4966 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4967 | ||
b128c09f | 4968 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4969 | |
0037b49e | 4970 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4971 | return (error); |
4972 | } | |
4973 | ||
4974 | /*ARGSUSED*/ | |
e5c39b95 | 4975 | int |
3558fd73 | 4976 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4977 | { |
3558fd73 | 4978 | znode_t *zp = ITOZ(ip); |
0037b49e | 4979 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4980 | uint32_t gen; |
428870ff | 4981 | uint64_t gen64; |
34dc7c2f BB |
4982 | uint64_t object = zp->z_id; |
4983 | zfid_short_t *zfid; | |
428870ff | 4984 | int size, i, error; |
34dc7c2f | 4985 | |
0037b49e | 4986 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4987 | ZFS_VERIFY_ZP(zp); |
428870ff | 4988 | |
0037b49e | 4989 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), |
428870ff | 4990 | &gen64, sizeof (uint64_t))) != 0) { |
0037b49e | 4991 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4992 | return (error); |
4993 | } | |
4994 | ||
4995 | gen = (uint32_t)gen64; | |
34dc7c2f | 4996 | |
9b77d1c9 | 4997 | size = SHORT_FID_LEN; |
34dc7c2f BB |
4998 | |
4999 | zfid = (zfid_short_t *)fidp; | |
5000 | ||
5001 | zfid->zf_len = size; | |
5002 | ||
5003 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
5004 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
5005 | ||
5006 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
5007 | if (gen == 0) | |
5008 | gen = 1; | |
5009 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
5010 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
5011 | ||
0037b49e | 5012 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5013 | return (0); |
5014 | } | |
5015 | ||
34dc7c2f | 5016 | /*ARGSUSED*/ |
e5c39b95 | 5017 | int |
3558fd73 | 5018 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 5019 | { |
3558fd73 | 5020 | znode_t *zp = ITOZ(ip); |
0037b49e | 5021 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
5022 | int error; |
5023 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
5024 | ||
0037b49e | 5025 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
5026 | ZFS_VERIFY_ZP(zp); |
5027 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
0037b49e | 5028 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5029 | |
5030 | return (error); | |
5031 | } | |
5032 | ||
5033 | /*ARGSUSED*/ | |
e5c39b95 | 5034 | int |
3558fd73 | 5035 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 5036 | { |
3558fd73 | 5037 | znode_t *zp = ITOZ(ip); |
0037b49e | 5038 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
5039 | int error; |
5040 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
0037b49e | 5041 | zilog_t *zilog = zfsvfs->z_log; |
34dc7c2f | 5042 | |
0037b49e | 5043 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 5044 | ZFS_VERIFY_ZP(zp); |
428870ff | 5045 | |
34dc7c2f | 5046 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 5047 | |
0037b49e | 5048 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 5049 | zil_commit(zilog, 0); |
428870ff | 5050 | |
0037b49e | 5051 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5052 | return (error); |
5053 | } | |
5054 | ||
3558fd73 | 5055 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
5056 | /* |
5057 | * Tunable, both must be a power of 2. | |
5058 | * | |
5059 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
5060 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 5061 | * an arcbuf for a partial block read |
428870ff BB |
5062 | */ |
5063 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
5064 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
5065 | ||
5066 | /*ARGSUSED*/ | |
5067 | static int | |
3558fd73 | 5068 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 5069 | { |
3558fd73 | 5070 | znode_t *zp = ITOZ(ip); |
0037b49e BB |
5071 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
5072 | int max_blksz = zfsvfs->z_max_blksz; | |
428870ff BB |
5073 | uio_t *uio = &xuio->xu_uio; |
5074 | ssize_t size = uio->uio_resid; | |
5075 | offset_t offset = uio->uio_loffset; | |
5076 | int blksz; | |
5077 | int fullblk, i; | |
5078 | arc_buf_t *abuf; | |
5079 | ssize_t maxsize; | |
5080 | int preamble, postamble; | |
5081 | ||
5082 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
2e528b49 | 5083 | return (SET_ERROR(EINVAL)); |
428870ff | 5084 | |
0037b49e | 5085 | ZFS_ENTER(zfsvfs); |
428870ff BB |
5086 | ZFS_VERIFY_ZP(zp); |
5087 | switch (ioflag) { | |
5088 | case UIO_WRITE: | |
5089 | /* | |
5090 | * Loan out an arc_buf for write if write size is bigger than | |
5091 | * max_blksz, and the file's block size is also max_blksz. | |
5092 | */ | |
5093 | blksz = max_blksz; | |
5094 | if (size < blksz || zp->z_blksz != blksz) { | |
0037b49e | 5095 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5096 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5097 | } |
5098 | /* | |
5099 | * Caller requests buffers for write before knowing where the | |
5100 | * write offset might be (e.g. NFS TCP write). | |
5101 | */ | |
5102 | if (offset == -1) { | |
5103 | preamble = 0; | |
5104 | } else { | |
5105 | preamble = P2PHASE(offset, blksz); | |
5106 | if (preamble) { | |
5107 | preamble = blksz - preamble; | |
5108 | size -= preamble; | |
5109 | } | |
5110 | } | |
5111 | ||
5112 | postamble = P2PHASE(size, blksz); | |
5113 | size -= postamble; | |
5114 | ||
5115 | fullblk = size / blksz; | |
5116 | (void) dmu_xuio_init(xuio, | |
5117 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
5118 | |
5119 | /* | |
5120 | * Have to fix iov base/len for partial buffers. They | |
5121 | * currently represent full arc_buf's. | |
5122 | */ | |
5123 | if (preamble) { | |
5124 | /* data begins in the middle of the arc_buf */ | |
5125 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5126 | blksz); | |
5127 | ASSERT(abuf); | |
5128 | (void) dmu_xuio_add(xuio, abuf, | |
5129 | blksz - preamble, preamble); | |
5130 | } | |
5131 | ||
5132 | for (i = 0; i < fullblk; i++) { | |
5133 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5134 | blksz); | |
5135 | ASSERT(abuf); | |
5136 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
5137 | } | |
5138 | ||
5139 | if (postamble) { | |
5140 | /* data ends in the middle of the arc_buf */ | |
5141 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5142 | blksz); | |
5143 | ASSERT(abuf); | |
5144 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
5145 | } | |
5146 | break; | |
5147 | case UIO_READ: | |
5148 | /* | |
5149 | * Loan out an arc_buf for read if the read size is larger than | |
5150 | * the current file block size. Block alignment is not | |
5151 | * considered. Partial arc_buf will be loaned out for read. | |
5152 | */ | |
5153 | blksz = zp->z_blksz; | |
5154 | if (blksz < zcr_blksz_min) | |
5155 | blksz = zcr_blksz_min; | |
5156 | if (blksz > zcr_blksz_max) | |
5157 | blksz = zcr_blksz_max; | |
5158 | /* avoid potential complexity of dealing with it */ | |
5159 | if (blksz > max_blksz) { | |
0037b49e | 5160 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5161 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5162 | } |
5163 | ||
5164 | maxsize = zp->z_size - uio->uio_loffset; | |
5165 | if (size > maxsize) | |
5166 | size = maxsize; | |
5167 | ||
3558fd73 | 5168 | if (size < blksz) { |
0037b49e | 5169 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5170 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5171 | } |
5172 | break; | |
5173 | default: | |
0037b49e | 5174 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5175 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5176 | } |
5177 | ||
5178 | uio->uio_extflg = UIO_XUIO; | |
5179 | XUIO_XUZC_RW(xuio) = ioflag; | |
0037b49e | 5180 | ZFS_EXIT(zfsvfs); |
428870ff BB |
5181 | return (0); |
5182 | } | |
5183 | ||
5184 | /*ARGSUSED*/ | |
5185 | static int | |
3558fd73 | 5186 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
5187 | { |
5188 | int i; | |
5189 | arc_buf_t *abuf; | |
5190 | int ioflag = XUIO_XUZC_RW(xuio); | |
5191 | ||
5192 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
5193 | ||
5194 | i = dmu_xuio_cnt(xuio); | |
5195 | while (i-- > 0) { | |
5196 | abuf = dmu_xuio_arcbuf(xuio, i); | |
5197 | /* | |
5198 | * if abuf == NULL, it must be a write buffer | |
5199 | * that has been returned in zfs_write(). | |
5200 | */ | |
5201 | if (abuf) | |
5202 | dmu_return_arcbuf(abuf); | |
5203 | ASSERT(abuf || ioflag == UIO_WRITE); | |
5204 | } | |
5205 | ||
5206 | dmu_xuio_fini(xuio); | |
5207 | return (0); | |
5208 | } | |
3558fd73 | 5209 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 | 5210 | |
93ce2b4c | 5211 | #if defined(_KERNEL) |
f298b24d BB |
5212 | EXPORT_SYMBOL(zfs_open); |
5213 | EXPORT_SYMBOL(zfs_close); | |
5214 | EXPORT_SYMBOL(zfs_read); | |
5215 | EXPORT_SYMBOL(zfs_write); | |
5216 | EXPORT_SYMBOL(zfs_access); | |
5217 | EXPORT_SYMBOL(zfs_lookup); | |
5218 | EXPORT_SYMBOL(zfs_create); | |
5219 | EXPORT_SYMBOL(zfs_tmpfile); | |
5220 | EXPORT_SYMBOL(zfs_remove); | |
5221 | EXPORT_SYMBOL(zfs_mkdir); | |
5222 | EXPORT_SYMBOL(zfs_rmdir); | |
5223 | EXPORT_SYMBOL(zfs_readdir); | |
5224 | EXPORT_SYMBOL(zfs_fsync); | |
5225 | EXPORT_SYMBOL(zfs_getattr); | |
5226 | EXPORT_SYMBOL(zfs_getattr_fast); | |
5227 | EXPORT_SYMBOL(zfs_setattr); | |
5228 | EXPORT_SYMBOL(zfs_rename); | |
5229 | EXPORT_SYMBOL(zfs_symlink); | |
5230 | EXPORT_SYMBOL(zfs_readlink); | |
5231 | EXPORT_SYMBOL(zfs_link); | |
5232 | EXPORT_SYMBOL(zfs_inactive); | |
5233 | EXPORT_SYMBOL(zfs_space); | |
5234 | EXPORT_SYMBOL(zfs_fid); | |
5235 | EXPORT_SYMBOL(zfs_getsecattr); | |
5236 | EXPORT_SYMBOL(zfs_setsecattr); | |
5237 | EXPORT_SYMBOL(zfs_getpage); | |
5238 | EXPORT_SYMBOL(zfs_putpage); | |
5239 | EXPORT_SYMBOL(zfs_dirty_inode); | |
5240 | EXPORT_SYMBOL(zfs_map); | |
5241 | ||
02730c33 | 5242 | /* CSTYLED */ |
a966c564 K |
5243 | module_param(zfs_delete_blocks, ulong, 0644); |
5244 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
c409e464 BB |
5245 | module_param(zfs_read_chunk_size, long, 0644); |
5246 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
5247 | #endif |