]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
9b7b9cd3 | 21 | |
34dc7c2f | 22 | /* |
428870ff | 23 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
a448a255 | 24 | * Copyright (c) 2012, 2018 by Delphix. All rights reserved. |
5475aada | 25 | * Copyright (c) 2015 by Chunwei Chen. All rights reserved. |
9b7b9cd3 | 26 | * Copyright 2017 Nexenta Systems, Inc. |
34dc7c2f BB |
27 | */ |
28 | ||
29 | /* Portions Copyright 2007 Jeremy Teo */ | |
428870ff | 30 | /* Portions Copyright 2010 Robert Milkowski */ |
34dc7c2f | 31 | |
60101509 | 32 | |
34dc7c2f BB |
33 | #include <sys/types.h> |
34 | #include <sys/param.h> | |
35 | #include <sys/time.h> | |
34dc7c2f | 36 | #include <sys/sysmacros.h> |
34dc7c2f | 37 | #include <sys/vfs.h> |
34dc7c2f BB |
38 | #include <sys/file.h> |
39 | #include <sys/stat.h> | |
40 | #include <sys/kmem.h> | |
41 | #include <sys/taskq.h> | |
42 | #include <sys/uio.h> | |
43 | #include <sys/vmsystm.h> | |
44 | #include <sys/atomic.h> | |
34dc7c2f BB |
45 | #include <sys/pathname.h> |
46 | #include <sys/cmn_err.h> | |
47 | #include <sys/errno.h> | |
34dc7c2f BB |
48 | #include <sys/zfs_dir.h> |
49 | #include <sys/zfs_acl.h> | |
50 | #include <sys/zfs_ioctl.h> | |
51 | #include <sys/fs/zfs.h> | |
52 | #include <sys/dmu.h> | |
428870ff | 53 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
54 | #include <sys/spa.h> |
55 | #include <sys/txg.h> | |
56 | #include <sys/dbuf.h> | |
57 | #include <sys/zap.h> | |
428870ff | 58 | #include <sys/sa.h> |
34dc7c2f BB |
59 | #include <sys/policy.h> |
60 | #include <sys/sunddi.h> | |
b128c09f | 61 | #include <sys/sid.h> |
bcf30822 | 62 | #include <sys/mode.h> |
ebe7e575 | 63 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 64 | #include <sys/zfs_fuid.h> |
428870ff | 65 | #include <sys/zfs_sa.h> |
e5c39b95 | 66 | #include <sys/zfs_vnops.h> |
34dc7c2f | 67 | #include <sys/zfs_rlock.h> |
428870ff | 68 | #include <sys/cred.h> |
218b8eaf | 69 | #include <sys/zpl.h> |
1ce23dca | 70 | #include <sys/zil.h> |
9c5167d1 | 71 | #include <sys/sa_impl.h> |
34dc7c2f BB |
72 | |
73 | /* | |
74 | * Programming rules. | |
75 | * | |
76 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
77 | * properly lock its in-core state, create a DMU transaction, do the work, | |
78 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
79 | * and wait for the intent log to commit if it is a synchronous operation. | |
80 | * Moreover, the vnode ops must work in both normal and log replay context. | |
81 | * The ordering of events is important to avoid deadlocks and references | |
82 | * to freed memory. The example below illustrates the following Big Rules: | |
83 | * | |
84 | * (1) A check must be made in each zfs thread for a mounted file system. | |
0037b49e BB |
85 | * This is done avoiding races using ZFS_ENTER(zfsvfs). |
86 | * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes | |
34dc7c2f BB |
87 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros |
88 | * can return EIO from the calling function. | |
89 | * | |
3558fd73 | 90 | * (2) iput() should always be the last thing except for zil_commit() |
34dc7c2f BB |
91 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: |
92 | * First, if it's the last reference, the vnode/znode | |
93 | * can be freed, so the zp may point to freed memory. Second, the last | |
94 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
95 | * pushing cached pages (which acquires range locks) and syncing out | |
96 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
97 | * which could deadlock the system if you were already holding one. | |
0a50679c | 98 | * If you must call iput() within a tx then use zfs_iput_async(). |
34dc7c2f BB |
99 | * |
100 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
101 | * as they can span dmu_tx_assign() calls. | |
102 | * | |
384f8a09 MA |
103 | * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to |
104 | * dmu_tx_assign(). This is critical because we don't want to block | |
105 | * while holding locks. | |
106 | * | |
107 | * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This | |
108 | * reduces lock contention and CPU usage when we must wait (note that if | |
109 | * throughput is constrained by the storage, nearly every transaction | |
110 | * must wait). | |
111 | * | |
112 | * Note, in particular, that if a lock is sometimes acquired before | |
113 | * the tx assigns, and sometimes after (e.g. z_lock), then failing | |
114 | * to use a non-blocking assign can deadlock the system. The scenario: | |
34dc7c2f BB |
115 | * |
116 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
117 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
118 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
119 | * forever, because the previous txg can't quiesce until B's tx commits. | |
120 | * | |
0037b49e | 121 | * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, |
e8b96c60 | 122 | * then drop all locks, call dmu_tx_wait(), and try again. On subsequent |
0735ecb3 | 123 | * calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT, |
e8b96c60 MA |
124 | * to indicate that this operation has already called dmu_tx_wait(). |
125 | * This will ensure that we don't retry forever, waiting a short bit | |
126 | * each time. | |
34dc7c2f BB |
127 | * |
128 | * (5) If the operation succeeded, generate the intent log entry for it | |
129 | * before dropping locks. This ensures that the ordering of events | |
130 | * in the intent log matches the order in which they actually occurred. | |
d3cc8b15 | 131 | * During ZIL replay the zfs_log_* functions will update the sequence |
fb5f0bc8 | 132 | * number to indicate the zil transaction has replayed. |
34dc7c2f BB |
133 | * |
134 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
135 | * regardless of whether there were any errors. | |
136 | * | |
572e2857 | 137 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
138 | * to ensure that synchronous semantics are provided when necessary. |
139 | * | |
140 | * In general, this is how things should be ordered in each vnode op: | |
141 | * | |
0037b49e | 142 | * ZFS_ENTER(zfsvfs); // exit if unmounted |
34dc7c2f | 143 | * top: |
3558fd73 | 144 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
145 | * rw_enter(...); // grab any other locks you need |
146 | * tx = dmu_tx_create(...); // get DMU tx | |
147 | * dmu_tx_hold_*(); // hold each object you might modify | |
0735ecb3 | 148 | * error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
149 | * if (error) { |
150 | * rw_exit(...); // drop locks | |
151 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 152 | * iput(...); // release held vnodes |
fb5f0bc8 | 153 | * if (error == ERESTART) { |
e8b96c60 | 154 | * waited = B_TRUE; |
34dc7c2f BB |
155 | * dmu_tx_wait(tx); |
156 | * dmu_tx_abort(tx); | |
157 | * goto top; | |
158 | * } | |
159 | * dmu_tx_abort(tx); // abort DMU tx | |
0037b49e | 160 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
161 | * return (error); // really out of space |
162 | * } | |
163 | * error = do_real_work(); // do whatever this VOP does | |
164 | * if (error == 0) | |
165 | * zfs_log_*(...); // on success, make ZIL entry | |
166 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
167 | * rw_exit(...); // drop locks | |
168 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 169 | * iput(...); // release held vnodes |
572e2857 | 170 | * zil_commit(zilog, foid); // synchronous when necessary |
0037b49e | 171 | * ZFS_EXIT(zfsvfs); // finished in zfs |
34dc7c2f BB |
172 | * return (error); // done, report error |
173 | */ | |
174 | ||
126400a1 BB |
175 | /* |
176 | * Virus scanning is unsupported. It would be possible to add a hook | |
177 | * here to performance the required virus scan. This could be done | |
178 | * entirely in the kernel or potentially as an update to invoke a | |
179 | * scanning utility. | |
180 | */ | |
181 | static int | |
182 | zfs_vscan(struct inode *ip, cred_t *cr, int async) | |
183 | { | |
184 | return (0); | |
185 | } | |
186 | ||
187 | /* ARGSUSED */ | |
188 | int | |
189 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
190 | { | |
191 | znode_t *zp = ITOZ(ip); | |
0037b49e | 192 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 193 | |
0037b49e | 194 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
195 | ZFS_VERIFY_ZP(zp); |
196 | ||
197 | /* Honor ZFS_APPENDONLY file attribute */ | |
198 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
199 | ((flag & O_APPEND) == 0)) { | |
0037b49e | 200 | ZFS_EXIT(zfsvfs); |
2e528b49 | 201 | return (SET_ERROR(EPERM)); |
126400a1 BB |
202 | } |
203 | ||
204 | /* Virus scan eligible files on open */ | |
0037b49e | 205 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
206 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { |
207 | if (zfs_vscan(ip, cr, 0) != 0) { | |
0037b49e | 208 | ZFS_EXIT(zfsvfs); |
2e528b49 | 209 | return (SET_ERROR(EACCES)); |
126400a1 BB |
210 | } |
211 | } | |
212 | ||
213 | /* Keep a count of the synchronous opens in the znode */ | |
214 | if (flag & O_SYNC) | |
215 | atomic_inc_32(&zp->z_sync_cnt); | |
216 | ||
0037b49e | 217 | ZFS_EXIT(zfsvfs); |
126400a1 BB |
218 | return (0); |
219 | } | |
126400a1 BB |
220 | |
221 | /* ARGSUSED */ | |
222 | int | |
223 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
224 | { | |
225 | znode_t *zp = ITOZ(ip); | |
0037b49e | 226 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
126400a1 | 227 | |
0037b49e | 228 | ZFS_ENTER(zfsvfs); |
126400a1 BB |
229 | ZFS_VERIFY_ZP(zp); |
230 | ||
7dc71949 | 231 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 232 | if (flag & O_SYNC) |
7dc71949 | 233 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 | 234 | |
0037b49e | 235 | if (!zfs_has_ctldir(zp) && zfsvfs->z_vscan && S_ISREG(ip->i_mode) && |
126400a1 BB |
236 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) |
237 | VERIFY(zfs_vscan(ip, cr, 1) == 0); | |
238 | ||
0037b49e | 239 | ZFS_EXIT(zfsvfs); |
8780c539 | 240 | return (0); |
126400a1 | 241 | } |
126400a1 | 242 | |
802e7b5f | 243 | #if defined(SEEK_HOLE) && defined(SEEK_DATA) |
cf91b2b6 | 244 | /* |
802e7b5f LD |
245 | * Lseek support for finding holes (cmd == SEEK_HOLE) and |
246 | * data (cmd == SEEK_DATA). "off" is an in/out parameter. | |
cf91b2b6 MA |
247 | */ |
248 | static int | |
802e7b5f | 249 | zfs_holey_common(struct inode *ip, int cmd, loff_t *off) |
cf91b2b6 | 250 | { |
802e7b5f | 251 | znode_t *zp = ITOZ(ip); |
cf91b2b6 MA |
252 | uint64_t noff = (uint64_t)*off; /* new offset */ |
253 | uint64_t file_sz; | |
254 | int error; | |
255 | boolean_t hole; | |
256 | ||
257 | file_sz = zp->z_size; | |
258 | if (noff >= file_sz) { | |
2e528b49 | 259 | return (SET_ERROR(ENXIO)); |
cf91b2b6 MA |
260 | } |
261 | ||
802e7b5f | 262 | if (cmd == SEEK_HOLE) |
cf91b2b6 MA |
263 | hole = B_TRUE; |
264 | else | |
265 | hole = B_FALSE; | |
266 | ||
802e7b5f | 267 | error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff); |
cf91b2b6 | 268 | |
d97aa48f | 269 | if (error == ESRCH) |
2e528b49 | 270 | return (SET_ERROR(ENXIO)); |
d97aa48f | 271 | |
6e03ec4f DB |
272 | /* file was dirty, so fall back to using generic logic */ |
273 | if (error == EBUSY) { | |
274 | if (hole) | |
275 | *off = file_sz; | |
276 | ||
277 | return (0); | |
278 | } | |
66aca247 | 279 | |
d97aa48f MA |
280 | /* |
281 | * We could find a hole that begins after the logical end-of-file, | |
282 | * because dmu_offset_next() only works on whole blocks. If the | |
283 | * EOF falls mid-block, then indicate that the "virtual hole" | |
284 | * at the end of the file begins at the logical EOF, rather than | |
285 | * at the end of the last block. | |
286 | */ | |
287 | if (noff > file_sz) { | |
288 | ASSERT(hole); | |
289 | noff = file_sz; | |
cf91b2b6 MA |
290 | } |
291 | ||
292 | if (noff < *off) | |
293 | return (error); | |
294 | *off = noff; | |
295 | return (error); | |
296 | } | |
802e7b5f LD |
297 | |
298 | int | |
299 | zfs_holey(struct inode *ip, int cmd, loff_t *off) | |
300 | { | |
301 | znode_t *zp = ITOZ(ip); | |
0037b49e | 302 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
802e7b5f LD |
303 | int error; |
304 | ||
0037b49e | 305 | ZFS_ENTER(zfsvfs); |
802e7b5f LD |
306 | ZFS_VERIFY_ZP(zp); |
307 | ||
308 | error = zfs_holey_common(ip, cmd, off); | |
309 | ||
0037b49e | 310 | ZFS_EXIT(zfsvfs); |
802e7b5f LD |
311 | return (error); |
312 | } | |
802e7b5f | 313 | #endif /* SEEK_HOLE && SEEK_DATA */ |
cf91b2b6 | 314 | |
c0d35759 | 315 | #if defined(_KERNEL) |
34dc7c2f BB |
316 | /* |
317 | * When a file is memory mapped, we must keep the IO data synchronized | |
318 | * between the DMU cache and the memory mapped pages. What this means: | |
319 | * | |
320 | * On Write: If we find a memory mapped page, we write to *both* | |
321 | * the page and the dmu buffer. | |
34dc7c2f | 322 | */ |
d164b209 | 323 | static void |
c0d35759 BB |
324 | update_pages(struct inode *ip, int64_t start, int len, |
325 | objset_t *os, uint64_t oid) | |
34dc7c2f | 326 | { |
c0d35759 BB |
327 | struct address_space *mp = ip->i_mapping; |
328 | struct page *pp; | |
329 | uint64_t nbytes; | |
d164b209 | 330 | int64_t off; |
c0d35759 | 331 | void *pb; |
34dc7c2f | 332 | |
8b1899d3 BB |
333 | off = start & (PAGE_SIZE-1); |
334 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
335 | nbytes = MIN(PAGE_SIZE - off, len); | |
34dc7c2f | 336 | |
8b1899d3 | 337 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
338 | if (pp) { |
339 | if (mapping_writably_mapped(mp)) | |
340 | flush_dcache_page(pp); | |
34dc7c2f | 341 | |
c0d35759 BB |
342 | pb = kmap(pp); |
343 | (void) dmu_read(os, oid, start+off, nbytes, pb+off, | |
9babb374 | 344 | DMU_READ_PREFETCH); |
c0d35759 BB |
345 | kunmap(pp); |
346 | ||
347 | if (mapping_writably_mapped(mp)) | |
348 | flush_dcache_page(pp); | |
349 | ||
350 | mark_page_accessed(pp); | |
351 | SetPageUptodate(pp); | |
352 | ClearPageError(pp); | |
353 | unlock_page(pp); | |
8b1899d3 | 354 | put_page(pp); |
34dc7c2f | 355 | } |
c0d35759 | 356 | |
d164b209 | 357 | len -= nbytes; |
34dc7c2f | 358 | off = 0; |
34dc7c2f | 359 | } |
34dc7c2f BB |
360 | } |
361 | ||
362 | /* | |
363 | * When a file is memory mapped, we must keep the IO data synchronized | |
364 | * between the DMU cache and the memory mapped pages. What this means: | |
365 | * | |
366 | * On Read: We "read" preferentially from memory mapped pages, | |
367 | * else we default from the dmu buffer. | |
368 | * | |
369 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
d3cc8b15 | 370 | * the file is memory mapped. |
34dc7c2f BB |
371 | */ |
372 | static int | |
3558fd73 | 373 | mappedread(struct inode *ip, int nbytes, uio_t *uio) |
34dc7c2f | 374 | { |
c0d35759 BB |
375 | struct address_space *mp = ip->i_mapping; |
376 | struct page *pp; | |
3558fd73 | 377 | znode_t *zp = ITOZ(ip); |
34dc7c2f | 378 | int64_t start, off; |
c0d35759 | 379 | uint64_t bytes; |
34dc7c2f BB |
380 | int len = nbytes; |
381 | int error = 0; | |
c0d35759 | 382 | void *pb; |
34dc7c2f BB |
383 | |
384 | start = uio->uio_loffset; | |
8b1899d3 BB |
385 | off = start & (PAGE_SIZE-1); |
386 | for (start &= PAGE_MASK; len > 0; start += PAGE_SIZE) { | |
387 | bytes = MIN(PAGE_SIZE - off, len); | |
c0d35759 | 388 | |
8b1899d3 | 389 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
c0d35759 BB |
390 | if (pp) { |
391 | ASSERT(PageUptodate(pp)); | |
b2ab468d | 392 | unlock_page(pp); |
c0d35759 BB |
393 | |
394 | pb = kmap(pp); | |
395 | error = uiomove(pb + off, bytes, UIO_READ, uio); | |
396 | kunmap(pp); | |
397 | ||
398 | if (mapping_writably_mapped(mp)) | |
399 | flush_dcache_page(pp); | |
400 | ||
401 | mark_page_accessed(pp); | |
8b1899d3 | 402 | put_page(pp); |
34dc7c2f | 403 | } else { |
804e0504 MA |
404 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
405 | uio, bytes); | |
34dc7c2f | 406 | } |
c0d35759 | 407 | |
34dc7c2f BB |
408 | len -= bytes; |
409 | off = 0; | |
410 | if (error) | |
411 | break; | |
412 | } | |
413 | return (error); | |
414 | } | |
c0d35759 | 415 | #endif /* _KERNEL */ |
34dc7c2f | 416 | |
c409e464 | 417 | unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ |
a966c564 | 418 | unsigned long zfs_delete_blocks = DMU_MAX_DELETEBLKCNT; |
34dc7c2f BB |
419 | |
420 | /* | |
421 | * Read bytes from specified file into supplied buffer. | |
422 | * | |
3558fd73 | 423 | * IN: ip - inode of file to be read from. |
34dc7c2f BB |
424 | * uio - structure supplying read location, range info, |
425 | * and return buffer. | |
c0d35759 BB |
426 | * ioflag - FSYNC flags; used to provide FRSYNC semantics. |
427 | * O_DIRECT flag; used to bypass page cache. | |
34dc7c2f | 428 | * cr - credentials of caller. |
34dc7c2f BB |
429 | * |
430 | * OUT: uio - updated offset and range, buffer filled. | |
431 | * | |
d3cc8b15 | 432 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
433 | * |
434 | * Side Effects: | |
3558fd73 | 435 | * inode - atime updated if byte count > 0 |
34dc7c2f BB |
436 | */ |
437 | /* ARGSUSED */ | |
e5c39b95 | 438 | int |
3558fd73 | 439 | zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 440 | { |
a448a255 | 441 | int error = 0; |
2efea7c8 | 442 | boolean_t frsync = B_FALSE; |
34dc7c2f | 443 | |
a448a255 SD |
444 | znode_t *zp = ITOZ(ip); |
445 | zfsvfs_t *zfsvfs = ITOZSB(ip); | |
0037b49e | 446 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 447 | ZFS_VERIFY_ZP(zp); |
34dc7c2f | 448 | |
428870ff | 449 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
0037b49e | 450 | ZFS_EXIT(zfsvfs); |
2e528b49 | 451 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
452 | } |
453 | ||
454 | /* | |
455 | * Validate file offset | |
456 | */ | |
457 | if (uio->uio_loffset < (offset_t)0) { | |
0037b49e | 458 | ZFS_EXIT(zfsvfs); |
2e528b49 | 459 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
460 | } |
461 | ||
462 | /* | |
463 | * Fasttrack empty reads | |
464 | */ | |
465 | if (uio->uio_resid == 0) { | |
0037b49e | 466 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
467 | return (0); |
468 | } | |
469 | ||
2efea7c8 | 470 | #ifdef FRSYNC |
34dc7c2f BB |
471 | /* |
472 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
37699482 | 473 | * Only do this for non-snapshots. |
2efea7c8 GW |
474 | * |
475 | * Some platforms do not support FRSYNC and instead map it | |
476 | * to FSYNC, which results in unnecessary calls to zil_commit. We | |
477 | * only honor FRSYNC requests on platforms which support it. | |
34dc7c2f | 478 | */ |
2efea7c8 GW |
479 | frsync = !!(ioflag & FRSYNC); |
480 | #endif | |
37699482 | 481 | if (zfsvfs->z_log && |
2efea7c8 | 482 | (frsync || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)) |
0037b49e | 483 | zil_commit(zfsvfs->z_log, zp->z_id); |
34dc7c2f BB |
484 | |
485 | /* | |
486 | * Lock the range against changes. | |
487 | */ | |
5d43cc9a | 488 | locked_range_t *lr = rangelock_enter(&zp->z_rangelock, |
a448a255 | 489 | uio->uio_loffset, uio->uio_resid, RL_READER); |
34dc7c2f BB |
490 | |
491 | /* | |
492 | * If we are reading past end-of-file we can skip | |
493 | * to the end; but we might still need to set atime. | |
494 | */ | |
428870ff | 495 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
496 | error = 0; |
497 | goto out; | |
498 | } | |
499 | ||
428870ff | 500 | ASSERT(uio->uio_loffset < zp->z_size); |
a448a255 SD |
501 | ssize_t n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); |
502 | ssize_t start_resid = n; | |
428870ff | 503 | |
3558fd73 | 504 | #ifdef HAVE_UIO_ZEROCOPY |
a448a255 | 505 | xuio_t *xuio = NULL; |
428870ff BB |
506 | if ((uio->uio_extflg == UIO_XUIO) && |
507 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
508 | int nblk; | |
509 | int blksz = zp->z_blksz; | |
510 | uint64_t offset = uio->uio_loffset; | |
511 | ||
512 | xuio = (xuio_t *)uio; | |
513 | if ((ISP2(blksz))) { | |
514 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
515 | blksz)) / blksz; | |
516 | } else { | |
517 | ASSERT(offset + n <= blksz); | |
518 | nblk = 1; | |
519 | } | |
520 | (void) dmu_xuio_init(xuio, nblk); | |
521 | ||
3558fd73 | 522 | if (vn_has_cached_data(ip)) { |
428870ff BB |
523 | /* |
524 | * For simplicity, we always allocate a full buffer | |
525 | * even if we only expect to read a portion of a block. | |
526 | */ | |
527 | while (--nblk >= 0) { | |
528 | (void) dmu_xuio_add(xuio, | |
529 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
530 | blksz), 0, blksz); | |
531 | } | |
532 | } | |
533 | } | |
3558fd73 | 534 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
535 | |
536 | while (n > 0) { | |
a448a255 | 537 | ssize_t nbytes = MIN(n, zfs_read_chunk_size - |
34dc7c2f BB |
538 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); |
539 | ||
804e0504 | 540 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
3558fd73 | 541 | error = mappedread(ip, nbytes, uio); |
804e0504 MA |
542 | } else { |
543 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), | |
544 | uio, nbytes); | |
545 | } | |
c0d35759 | 546 | |
b128c09f BB |
547 | if (error) { |
548 | /* convert checksum errors into IO errors */ | |
549 | if (error == ECKSUM) | |
2e528b49 | 550 | error = SET_ERROR(EIO); |
34dc7c2f | 551 | break; |
b128c09f | 552 | } |
34dc7c2f BB |
553 | |
554 | n -= nbytes; | |
555 | } | |
a448a255 SD |
556 | |
557 | int64_t nread = start_resid - n; | |
558 | dataset_kstats_update_read_kstats(&zfsvfs->z_kstat, nread); | |
559 | task_io_account_read(nread); | |
34dc7c2f | 560 | out: |
5d43cc9a | 561 | rangelock_exit(lr); |
34dc7c2f | 562 | |
0037b49e | 563 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
564 | return (error); |
565 | } | |
566 | ||
34dc7c2f BB |
567 | /* |
568 | * Write the bytes to a file. | |
569 | * | |
3558fd73 | 570 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
571 | * uio - structure supplying write location, range info, |
572 | * and data buffer. | |
573 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 574 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 575 | * cr - credentials of caller. |
34dc7c2f BB |
576 | * |
577 | * OUT: uio - updated offset and range. | |
578 | * | |
579 | * RETURN: 0 if success | |
580 | * error code if failure | |
581 | * | |
582 | * Timestamps: | |
3558fd73 | 583 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 584 | */ |
428870ff | 585 | |
34dc7c2f | 586 | /* ARGSUSED */ |
e5c39b95 | 587 | int |
3558fd73 | 588 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 589 | { |
a448a255 SD |
590 | int error = 0; |
591 | ssize_t start_resid = uio->uio_resid; | |
34dc7c2f | 592 | |
34dc7c2f BB |
593 | /* |
594 | * Fasttrack empty write | |
595 | */ | |
a448a255 | 596 | ssize_t n = start_resid; |
34dc7c2f BB |
597 | if (n == 0) |
598 | return (0); | |
599 | ||
a448a255 | 600 | rlim64_t limit = uio->uio_limit; |
34dc7c2f BB |
601 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) |
602 | limit = MAXOFFSET_T; | |
603 | ||
a448a255 SD |
604 | znode_t *zp = ITOZ(ip); |
605 | zfsvfs_t *zfsvfs = ZTOZSB(zp); | |
0037b49e | 606 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 607 | ZFS_VERIFY_ZP(zp); |
b128c09f | 608 | |
a448a255 SD |
609 | sa_bulk_attr_t bulk[4]; |
610 | int count = 0; | |
611 | uint64_t mtime[2], ctime[2]; | |
0037b49e BB |
612 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
613 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
614 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
615 | &zp->z_size, 8); | |
616 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
428870ff BB |
617 | &zp->z_pflags, 8); |
618 | ||
f3c9dca0 MT |
619 | /* |
620 | * Callers might not be able to detect properly that we are read-only, | |
621 | * so check it explicitly here. | |
622 | */ | |
0037b49e BB |
623 | if (zfs_is_readonly(zfsvfs)) { |
624 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
625 | return (SET_ERROR(EROFS)); |
626 | } | |
627 | ||
b128c09f BB |
628 | /* |
629 | * If immutable or not appending then return EPERM | |
630 | */ | |
428870ff BB |
631 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
632 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
633 | (uio->uio_loffset < zp->z_size))) { | |
0037b49e | 634 | ZFS_EXIT(zfsvfs); |
2e528b49 | 635 | return (SET_ERROR(EPERM)); |
b128c09f BB |
636 | } |
637 | ||
428870ff BB |
638 | /* |
639 | * Validate file offset | |
640 | */ | |
a448a255 | 641 | offset_t woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; |
428870ff | 642 | if (woff < 0) { |
0037b49e | 643 | ZFS_EXIT(zfsvfs); |
2e528b49 | 644 | return (SET_ERROR(EINVAL)); |
428870ff BB |
645 | } |
646 | ||
a448a255 SD |
647 | int max_blksz = zfsvfs->z_max_blksz; |
648 | xuio_t *xuio = NULL; | |
649 | ||
34dc7c2f BB |
650 | /* |
651 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
652 | * don't hold up txg. | |
428870ff | 653 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 654 | */ |
9cac042c | 655 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
656 | if ((uio->uio_extflg == UIO_XUIO) && |
657 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
658 | xuio = (xuio_t *)uio; | |
659 | else | |
9cac042c | 660 | #endif |
779a6c0b | 661 | if (uio_prefaultpages(MIN(n, max_blksz), uio)) { |
662 | ZFS_EXIT(zfsvfs); | |
663 | return (SET_ERROR(EFAULT)); | |
664 | } | |
34dc7c2f BB |
665 | |
666 | /* | |
667 | * If in append mode, set the io offset pointer to eof. | |
668 | */ | |
5d43cc9a | 669 | locked_range_t *lr; |
34dc7c2f BB |
670 | if (ioflag & FAPPEND) { |
671 | /* | |
428870ff BB |
672 | * Obtain an appending range lock to guarantee file append |
673 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f | 674 | */ |
5d43cc9a MA |
675 | lr = rangelock_enter(&zp->z_rangelock, 0, n, RL_APPEND); |
676 | woff = lr->lr_offset; | |
677 | if (lr->lr_length == UINT64_MAX) { | |
428870ff BB |
678 | /* |
679 | * We overlocked the file because this write will cause | |
680 | * the file block size to increase. | |
681 | * Note that zp_size cannot change with this lock held. | |
682 | */ | |
683 | woff = zp->z_size; | |
34dc7c2f | 684 | } |
428870ff | 685 | uio->uio_loffset = woff; |
34dc7c2f | 686 | } else { |
34dc7c2f | 687 | /* |
428870ff BB |
688 | * Note that if the file block size will change as a result of |
689 | * this write, then this range lock will lock the entire file | |
690 | * so that we can re-write the block safely. | |
34dc7c2f | 691 | */ |
5d43cc9a | 692 | lr = rangelock_enter(&zp->z_rangelock, woff, n, RL_WRITER); |
34dc7c2f BB |
693 | } |
694 | ||
695 | if (woff >= limit) { | |
5d43cc9a | 696 | rangelock_exit(lr); |
0037b49e | 697 | ZFS_EXIT(zfsvfs); |
2e528b49 | 698 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
699 | } |
700 | ||
701 | if ((woff + n) > limit || woff > (limit - n)) | |
702 | n = limit - woff; | |
703 | ||
428870ff | 704 | /* Will this write extend the file length? */ |
a448a255 SD |
705 | int write_eof = (woff + n > zp->z_size); |
706 | ||
707 | uint64_t end_size = MAX(zp->z_size, woff + n); | |
708 | zilog_t *zilog = zfsvfs->z_log; | |
709 | #ifdef HAVE_UIO_ZEROCOPY | |
710 | int i_iov = 0; | |
711 | const iovec_t *iovp = uio->uio_iov; | |
712 | ASSERTV(int iovcnt = uio->uio_iovcnt); | |
713 | #endif | |
428870ff | 714 | |
34dc7c2f BB |
715 | |
716 | /* | |
717 | * Write the file in reasonable size chunks. Each chunk is written | |
718 | * in a separate transaction; this keeps the intent log records small | |
719 | * and allows us to do more fine-grained space accounting. | |
720 | */ | |
721 | while (n > 0) { | |
9babb374 | 722 | woff = uio->uio_loffset; |
a448a255 | 723 | |
9c5167d1 NF |
724 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
725 | KUID_TO_SUID(ip->i_uid)) || | |
726 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
727 | KGID_TO_SGID(ip->i_gid)) || | |
728 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
729 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
730 | zp->z_projid))) { | |
2e528b49 | 731 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
732 | break; |
733 | } | |
734 | ||
a448a255 SD |
735 | arc_buf_t *abuf = NULL; |
736 | const iovec_t *aiov = NULL; | |
737 | if (xuio) { | |
5a6765cf | 738 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 739 | ASSERT(i_iov < iovcnt); |
5475aada | 740 | ASSERT3U(uio->uio_segflg, !=, UIO_BVEC); |
428870ff BB |
741 | aiov = &iovp[i_iov]; |
742 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
743 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
744 | ASSERT((aiov->iov_base == abuf->b_data) || |
745 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
746 | aiov->iov_len == arc_buf_size(abuf))); | |
747 | i_iov++; | |
5a6765cf | 748 | #endif |
a448a255 | 749 | } else if (n >= max_blksz && woff >= zp->z_size && |
9babb374 BB |
750 | P2PHASE(woff, max_blksz) == 0 && |
751 | zp->z_blksz == max_blksz) { | |
428870ff BB |
752 | /* |
753 | * This write covers a full block. "Borrow" a buffer | |
754 | * from the dmu so that we can fill it before we enter | |
755 | * a transaction. This avoids the possibility of | |
756 | * holding up the transaction if the data copy hangs | |
757 | * up on a pagefault (e.g., from an NFS server mapping). | |
758 | */ | |
9babb374 BB |
759 | size_t cbytes; |
760 | ||
428870ff BB |
761 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
762 | max_blksz); | |
9babb374 BB |
763 | ASSERT(abuf != NULL); |
764 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
765 | if ((error = uiocopy(abuf->b_data, max_blksz, |
766 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
767 | dmu_return_arcbuf(abuf); |
768 | break; | |
769 | } | |
770 | ASSERT(cbytes == max_blksz); | |
771 | } | |
772 | ||
34dc7c2f BB |
773 | /* |
774 | * Start a transaction. | |
775 | */ | |
a448a255 | 776 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 777 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 778 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 779 | zfs_sa_upgrade_txholds(tx, zp); |
384f8a09 | 780 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 781 | if (error) { |
34dc7c2f | 782 | dmu_tx_abort(tx); |
9babb374 BB |
783 | if (abuf != NULL) |
784 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
785 | break; |
786 | } | |
787 | ||
788 | /* | |
5d43cc9a | 789 | * If rangelock_enter() over-locked we grow the blocksize |
34dc7c2f | 790 | * and then reduce the lock range. This will only happen |
5d43cc9a MA |
791 | * on the first iteration since rangelock_reduce() will |
792 | * shrink down lr_length to the appropriate size. | |
34dc7c2f | 793 | */ |
5d43cc9a | 794 | if (lr->lr_length == UINT64_MAX) { |
34dc7c2f BB |
795 | uint64_t new_blksz; |
796 | ||
797 | if (zp->z_blksz > max_blksz) { | |
f1512ee6 MA |
798 | /* |
799 | * File's blocksize is already larger than the | |
800 | * "recordsize" property. Only let it grow to | |
801 | * the next power of 2. | |
802 | */ | |
34dc7c2f | 803 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 MA |
804 | new_blksz = MIN(end_size, |
805 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
806 | } else { |
807 | new_blksz = MIN(end_size, max_blksz); | |
808 | } | |
809 | zfs_grow_blocksize(zp, new_blksz, tx); | |
5d43cc9a | 810 | rangelock_reduce(lr, woff, n); |
34dc7c2f BB |
811 | } |
812 | ||
813 | /* | |
814 | * XXX - should we really limit each write to z_max_blksz? | |
815 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
816 | */ | |
a448a255 | 817 | ssize_t nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); |
34dc7c2f | 818 | |
a448a255 | 819 | ssize_t tx_bytes; |
9babb374 BB |
820 | if (abuf == NULL) { |
821 | tx_bytes = uio->uio_resid; | |
779a6c0b | 822 | uio->uio_fault_disable = B_TRUE; |
428870ff BB |
823 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
824 | uio, nbytes, tx); | |
515ddf65 | 825 | uio->uio_fault_disable = B_FALSE; |
779a6c0b | 826 | if (error == EFAULT) { |
827 | dmu_tx_commit(tx); | |
828 | if (uio_prefaultpages(MIN(n, max_blksz), uio)) { | |
829 | break; | |
830 | } | |
831 | continue; | |
832 | } else if (error != 0) { | |
833 | dmu_tx_commit(tx); | |
834 | break; | |
835 | } | |
9babb374 BB |
836 | tx_bytes -= uio->uio_resid; |
837 | } else { | |
838 | tx_bytes = nbytes; | |
428870ff BB |
839 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
840 | /* | |
841 | * If this is not a full block write, but we are | |
842 | * extending the file past EOF and this data starts | |
843 | * block-aligned, use assign_arcbuf(). Otherwise, | |
844 | * write via dmu_write(). | |
845 | */ | |
846 | if (tx_bytes < max_blksz && (!write_eof || | |
847 | aiov->iov_base != abuf->b_data)) { | |
848 | ASSERT(xuio); | |
0037b49e | 849 | dmu_write(zfsvfs->z_os, zp->z_id, woff, |
94183a9d | 850 | /* cppcheck-suppress nullPointer */ |
428870ff BB |
851 | aiov->iov_len, aiov->iov_base, tx); |
852 | dmu_return_arcbuf(abuf); | |
853 | xuio_stat_wbuf_copied(); | |
854 | } else { | |
855 | ASSERT(xuio || tx_bytes == max_blksz); | |
305781da | 856 | error = dmu_assign_arcbuf_by_dbuf( |
440a3eb9 | 857 | sa_get_db(zp->z_sa_hdl), woff, abuf, tx); |
305781da TC |
858 | if (error != 0) { |
859 | dmu_return_arcbuf(abuf); | |
860 | dmu_tx_commit(tx); | |
861 | break; | |
862 | } | |
428870ff | 863 | } |
9babb374 BB |
864 | ASSERT(tx_bytes <= uio->uio_resid); |
865 | uioskip(uio, tx_bytes); | |
866 | } | |
0037b49e BB |
867 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
868 | update_pages(ip, woff, | |
869 | tx_bytes, zfsvfs->z_os, zp->z_id); | |
870 | } | |
34dc7c2f BB |
871 | |
872 | /* | |
873 | * If we made no progress, we're done. If we made even | |
874 | * partial progress, update the znode and ZIL accordingly. | |
875 | */ | |
876 | if (tx_bytes == 0) { | |
0037b49e | 877 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 878 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
879 | dmu_tx_commit(tx); |
880 | ASSERT(error != 0); | |
881 | break; | |
882 | } | |
883 | ||
884 | /* | |
885 | * Clear Set-UID/Set-GID bits on successful write if not | |
4e33ba4c | 886 | * privileged and at least one of the execute bits is set. |
34dc7c2f BB |
887 | * |
888 | * It would be nice to to this after all writes have | |
889 | * been done, but that would still expose the ISUID/ISGID | |
890 | * to another app after the partial write is committed. | |
891 | * | |
572e2857 BB |
892 | * Note: we don't call zfs_fuid_map_id() here because |
893 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
894 | */ |
895 | mutex_enter(&zp->z_acl_lock); | |
a448a255 | 896 | uint32_t uid = KUID_TO_SUID(ip->i_uid); |
428870ff | 897 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 898 | (S_IXUSR >> 6))) != 0 && |
428870ff | 899 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 900 | secpolicy_vnode_setid_retain(cr, |
2c6abf15 | 901 | ((zp->z_mode & S_ISUID) != 0 && uid == 0)) != 0) { |
428870ff BB |
902 | uint64_t newmode; |
903 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
12fa7f34 | 904 | ip->i_mode = newmode = zp->z_mode; |
0037b49e | 905 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), |
428870ff | 906 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
907 | } |
908 | mutex_exit(&zp->z_acl_lock); | |
909 | ||
0df9673f | 910 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
34dc7c2f BB |
911 | |
912 | /* | |
913 | * Update the file size (zp_size) if it has changed; | |
914 | * account for possible concurrent updates. | |
915 | */ | |
428870ff BB |
916 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
917 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 918 | uio->uio_loffset); |
428870ff BB |
919 | ASSERT(error == 0); |
920 | } | |
572e2857 BB |
921 | /* |
922 | * If we are replaying and eof is non zero then force | |
923 | * the file size to the specified eof. Note, there's no | |
924 | * concurrency during replay. | |
925 | */ | |
0037b49e BB |
926 | if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) |
927 | zp->z_size = zfsvfs->z_replay_eof; | |
572e2857 | 928 | |
428870ff BB |
929 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
930 | ||
119a394a ED |
931 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
932 | NULL, NULL); | |
34dc7c2f BB |
933 | dmu_tx_commit(tx); |
934 | ||
935 | if (error != 0) | |
936 | break; | |
937 | ASSERT(tx_bytes == nbytes); | |
938 | n -= nbytes; | |
572e2857 | 939 | |
779a6c0b | 940 | if (!xuio && n > 0) { |
941 | if (uio_prefaultpages(MIN(n, max_blksz), uio)) { | |
942 | error = EFAULT; | |
943 | break; | |
944 | } | |
945 | } | |
34dc7c2f BB |
946 | } |
947 | ||
2a53e2da | 948 | zfs_inode_update(zp); |
5d43cc9a | 949 | rangelock_exit(lr); |
34dc7c2f BB |
950 | |
951 | /* | |
952 | * If we're in replay mode, or we made no progress, return error. | |
953 | * Otherwise, it's at least a partial write, so it's successful. | |
954 | */ | |
0037b49e BB |
955 | if (zfsvfs->z_replay || uio->uio_resid == start_resid) { |
956 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
957 | return (error); |
958 | } | |
959 | ||
428870ff | 960 | if (ioflag & (FSYNC | FDSYNC) || |
0037b49e | 961 | zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 962 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 963 | |
a448a255 SD |
964 | int64_t nwritten = start_resid - uio->uio_resid; |
965 | dataset_kstats_update_write_kstats(&zfsvfs->z_kstat, nwritten); | |
966 | task_io_account_write(nwritten); | |
967 | ||
0037b49e | 968 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
969 | return (0); |
970 | } | |
971 | ||
7c502b0b | 972 | /* |
973 | * Drop a reference on the passed inode asynchronously. This ensures | |
974 | * that the caller will never drop the last reference on an inode in | |
975 | * the current context. Doing so while holding open a tx could result | |
976 | * in a deadlock if iput_final() re-enters the filesystem code. | |
977 | */ | |
0a50679c BB |
978 | void |
979 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 980 | { |
0a50679c BB |
981 | objset_t *os = ITOZSB(ip)->z_os; |
982 | ||
3558fd73 | 983 | ASSERT(atomic_read(&ip->i_count) > 0); |
0a50679c BB |
984 | ASSERT(os != NULL); |
985 | ||
3558fd73 | 986 | if (atomic_read(&ip->i_count) == 1) |
7c502b0b | 987 | VERIFY(taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
48d3eb40 | 988 | (task_func_t *)iput, ip, TQ_SLEEP) != TASKQID_INVALID); |
3558fd73 BB |
989 | else |
990 | iput(ip); | |
991 | } | |
992 | ||
900d09b2 | 993 | /* ARGSUSED */ |
34dc7c2f | 994 | void |
428870ff | 995 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 996 | { |
428870ff | 997 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
998 | |
999 | if (zgd->zgd_db) | |
1000 | dmu_buf_rele(zgd->zgd_db, zgd); | |
1001 | ||
5d43cc9a | 1002 | rangelock_exit(zgd->zgd_lr); |
34dc7c2f | 1003 | |
9babb374 BB |
1004 | /* |
1005 | * Release the vnode asynchronously as we currently have the | |
1006 | * txg stopped from syncing. | |
1007 | */ | |
0a50679c | 1008 | zfs_iput_async(ZTOI(zp)); |
428870ff | 1009 | |
34dc7c2f BB |
1010 | kmem_free(zgd, sizeof (zgd_t)); |
1011 | } | |
1012 | ||
45d1cae3 BB |
1013 | #ifdef DEBUG |
1014 | static int zil_fault_io = 0; | |
1015 | #endif | |
1016 | ||
34dc7c2f BB |
1017 | /* |
1018 | * Get data to generate a TX_WRITE intent log record. | |
1019 | */ | |
1020 | int | |
1ce23dca | 1021 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, struct lwb *lwb, zio_t *zio) |
34dc7c2f | 1022 | { |
0037b49e BB |
1023 | zfsvfs_t *zfsvfs = arg; |
1024 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 1025 | znode_t *zp; |
428870ff BB |
1026 | uint64_t object = lr->lr_foid; |
1027 | uint64_t offset = lr->lr_offset; | |
1028 | uint64_t size = lr->lr_length; | |
34dc7c2f | 1029 | dmu_buf_t *db; |
34dc7c2f | 1030 | zgd_t *zgd; |
34dc7c2f BB |
1031 | int error = 0; |
1032 | ||
1ce23dca PS |
1033 | ASSERT3P(lwb, !=, NULL); |
1034 | ASSERT3P(zio, !=, NULL); | |
1035 | ASSERT3U(size, !=, 0); | |
34dc7c2f BB |
1036 | |
1037 | /* | |
1038 | * Nothing to do if the file has been removed | |
1039 | */ | |
0037b49e | 1040 | if (zfs_zget(zfsvfs, object, &zp) != 0) |
2e528b49 | 1041 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 1042 | if (zp->z_unlinked) { |
9babb374 BB |
1043 | /* |
1044 | * Release the vnode asynchronously as we currently have the | |
1045 | * txg stopped from syncing. | |
1046 | */ | |
0a50679c | 1047 | zfs_iput_async(ZTOI(zp)); |
2e528b49 | 1048 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1049 | } |
1050 | ||
79c76d5b | 1051 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
1ce23dca | 1052 | zgd->zgd_lwb = lwb; |
428870ff BB |
1053 | zgd->zgd_private = zp; |
1054 | ||
34dc7c2f BB |
1055 | /* |
1056 | * Write records come in two flavors: immediate and indirect. | |
1057 | * For small writes it's cheaper to store the data with the | |
1058 | * log record (immediate); for large writes it's cheaper to | |
1059 | * sync the data and get a pointer to it (indirect) so that | |
1060 | * we don't have to write the data twice. | |
1061 | */ | |
1062 | if (buf != NULL) { /* immediate write */ | |
5d43cc9a MA |
1063 | zgd->zgd_lr = rangelock_enter(&zp->z_rangelock, |
1064 | offset, size, RL_READER); | |
34dc7c2f | 1065 | /* test for truncation needs to be done while range locked */ |
428870ff | 1066 | if (offset >= zp->z_size) { |
2e528b49 | 1067 | error = SET_ERROR(ENOENT); |
428870ff BB |
1068 | } else { |
1069 | error = dmu_read(os, object, offset, size, buf, | |
1070 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1071 | } |
428870ff | 1072 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1073 | } else { /* indirect write */ |
34dc7c2f BB |
1074 | /* |
1075 | * Have to lock the whole block to ensure when it's | |
f763c3d1 | 1076 | * written out and its checksum is being calculated |
34dc7c2f BB |
1077 | * that no one can change the data. We need to re-check |
1078 | * blocksize after we get the lock in case it's changed! | |
1079 | */ | |
1080 | for (;;) { | |
428870ff BB |
1081 | uint64_t blkoff; |
1082 | size = zp->z_blksz; | |
1083 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1084 | offset -= blkoff; | |
5d43cc9a MA |
1085 | zgd->zgd_lr = rangelock_enter(&zp->z_rangelock, |
1086 | offset, size, RL_READER); | |
428870ff | 1087 | if (zp->z_blksz == size) |
34dc7c2f | 1088 | break; |
428870ff | 1089 | offset += blkoff; |
5d43cc9a | 1090 | rangelock_exit(zgd->zgd_lr); |
34dc7c2f BB |
1091 | } |
1092 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1093 | if (lr->lr_offset >= zp->z_size) |
2e528b49 | 1094 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1095 | #ifdef DEBUG |
1096 | if (zil_fault_io) { | |
2e528b49 | 1097 | error = SET_ERROR(EIO); |
45d1cae3 | 1098 | zil_fault_io = 0; |
45d1cae3 | 1099 | } |
45d1cae3 | 1100 | #endif |
34dc7c2f | 1101 | if (error == 0) |
428870ff BB |
1102 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1103 | DMU_READ_NO_PREFETCH); | |
1104 | ||
1105 | if (error == 0) { | |
02dc43bc | 1106 | blkptr_t *bp = &lr->lr_blkptr; |
03c6040b | 1107 | |
428870ff BB |
1108 | zgd->zgd_db = db; |
1109 | zgd->zgd_bp = bp; | |
1110 | ||
1111 | ASSERT(db->db_offset == offset); | |
1112 | ASSERT(db->db_size == size); | |
1113 | ||
1114 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1115 | zfs_get_done, zgd); | |
61ca48ff | 1116 | ASSERT(error || lr->lr_length <= size); |
428870ff BB |
1117 | |
1118 | /* | |
1119 | * On success, we need to wait for the write I/O | |
1120 | * initiated by dmu_sync() to complete before we can | |
1121 | * release this dbuf. We will finish everything up | |
1122 | * in the zfs_get_done() callback. | |
1123 | */ | |
1124 | if (error == 0) | |
1125 | return (0); | |
1126 | ||
1127 | if (error == EALREADY) { | |
1128 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
a1d477c2 MA |
1129 | /* |
1130 | * TX_WRITE2 relies on the data previously | |
1131 | * written by the TX_WRITE that caused | |
1132 | * EALREADY. We zero out the BP because | |
900d09b2 | 1133 | * it is the old, currently-on-disk BP. |
a1d477c2 MA |
1134 | */ |
1135 | zgd->zgd_bp = NULL; | |
1136 | BP_ZERO(bp); | |
428870ff BB |
1137 | error = 0; |
1138 | } | |
1139 | } | |
34dc7c2f | 1140 | } |
428870ff BB |
1141 | |
1142 | zfs_get_done(zgd, error); | |
1143 | ||
34dc7c2f BB |
1144 | return (error); |
1145 | } | |
1146 | ||
1147 | /*ARGSUSED*/ | |
3558fd73 BB |
1148 | int |
1149 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1150 | { |
3558fd73 | 1151 | znode_t *zp = ITOZ(ip); |
0037b49e | 1152 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
1153 | int error; |
1154 | ||
0037b49e | 1155 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1156 | ZFS_VERIFY_ZP(zp); |
1157 | ||
1158 | if (flag & V_ACE_MASK) | |
1159 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1160 | else | |
1161 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1162 | ||
0037b49e | 1163 | ZFS_EXIT(zfsvfs); |
45d1cae3 BB |
1164 | return (error); |
1165 | } | |
45d1cae3 | 1166 | |
34dc7c2f BB |
1167 | /* |
1168 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1169 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1170 | * |
3558fd73 | 1171 | * IN: dip - inode of directory to search. |
34dc7c2f | 1172 | * nm - name of entry to lookup. |
34dc7c2f | 1173 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1174 | * cr - credentials of caller. |
34dc7c2f BB |
1175 | * direntflags - directory lookup flags |
1176 | * realpnp - returned pathname. | |
1177 | * | |
3558fd73 | 1178 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1179 | * |
d3cc8b15 | 1180 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1181 | * |
1182 | * Timestamps: | |
1183 | * NA | |
1184 | */ | |
1185 | /* ARGSUSED */ | |
e5c39b95 | 1186 | int |
3558fd73 BB |
1187 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1188 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1189 | { |
3558fd73 | 1190 | znode_t *zdp = ITOZ(dip); |
0037b49e | 1191 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
3558fd73 | 1192 | int error = 0; |
45d1cae3 | 1193 | |
9b7b9cd3 GM |
1194 | /* |
1195 | * Fast path lookup, however we must skip DNLC lookup | |
1196 | * for case folding or normalizing lookups because the | |
1197 | * DNLC code only stores the passed in name. This means | |
1198 | * creating 'a' and removing 'A' on a case insensitive | |
1199 | * file system would work, but DNLC still thinks 'a' | |
1200 | * exists and won't let you create it again on the next | |
1201 | * pass through fast path. | |
1202 | */ | |
45d1cae3 BB |
1203 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { |
1204 | ||
3558fd73 | 1205 | if (!S_ISDIR(dip->i_mode)) { |
2e528b49 | 1206 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1207 | } else if (zdp->z_sa_hdl == NULL) { |
2e528b49 | 1208 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1209 | } |
1210 | ||
1211 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1212 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1213 | if (!error) { | |
3558fd73 BB |
1214 | *ipp = dip; |
1215 | igrab(*ipp); | |
45d1cae3 BB |
1216 | return (0); |
1217 | } | |
1218 | return (error); | |
45d1cae3 BB |
1219 | } |
1220 | } | |
1221 | ||
0037b49e | 1222 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1223 | ZFS_VERIFY_ZP(zdp); |
1224 | ||
3558fd73 | 1225 | *ipp = NULL; |
34dc7c2f BB |
1226 | |
1227 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1228 | /* |
1229 | * We don't allow recursive attributes.. | |
1230 | * Maybe someday we will. | |
1231 | */ | |
428870ff | 1232 | if (zdp->z_pflags & ZFS_XATTR) { |
0037b49e | 1233 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1234 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1235 | } |
1236 | ||
3558fd73 | 1237 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
0037b49e | 1238 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1239 | return (error); |
1240 | } | |
1241 | ||
1242 | /* | |
1243 | * Do we have permission to get into attribute directory? | |
1244 | */ | |
1245 | ||
3558fd73 | 1246 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1247 | B_FALSE, cr))) { |
3558fd73 BB |
1248 | iput(*ipp); |
1249 | *ipp = NULL; | |
34dc7c2f BB |
1250 | } |
1251 | ||
0037b49e | 1252 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1253 | return (error); |
1254 | } | |
1255 | ||
3558fd73 | 1256 | if (!S_ISDIR(dip->i_mode)) { |
0037b49e | 1257 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1258 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1259 | } |
1260 | ||
1261 | /* | |
1262 | * Check accessibility of directory. | |
1263 | */ | |
1264 | ||
149e873a | 1265 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
0037b49e | 1266 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1267 | return (error); |
1268 | } | |
1269 | ||
0037b49e | 1270 | if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1271 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1272 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1273 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1274 | } |
1275 | ||
3558fd73 BB |
1276 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1277 | if ((error == 0) && (*ipp)) | |
1278 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1279 | |
0037b49e | 1280 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1281 | return (error); |
1282 | } | |
1283 | ||
1284 | /* | |
1285 | * Attempt to create a new entry in a directory. If the entry | |
1286 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1287 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1288 | * |
3558fd73 | 1289 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1290 | * name - name of new file entry. |
1291 | * vap - attributes of new file. | |
1292 | * excl - flag indicating exclusive or non-exclusive mode. | |
1293 | * mode - mode to open file with. | |
1294 | * cr - credentials of caller. | |
a35c1207 | 1295 | * flag - file flag. |
3558fd73 | 1296 | * vsecp - ACL to be set |
34dc7c2f | 1297 | * |
3558fd73 | 1298 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1299 | * |
d3cc8b15 | 1300 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1301 | * |
1302 | * Timestamps: | |
3558fd73 BB |
1303 | * dip - ctime|mtime updated if new entry created |
1304 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1305 | */ |
1306 | ||
1307 | /* ARGSUSED */ | |
e5c39b95 | 1308 | int |
3558fd73 BB |
1309 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1310 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1311 | { |
3558fd73 | 1312 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1313 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1314 | zilog_t *zilog; |
1315 | objset_t *os; | |
1316 | zfs_dirlock_t *dl; | |
1317 | dmu_tx_t *tx; | |
1318 | int error; | |
b128c09f | 1319 | uid_t uid; |
149e873a | 1320 | gid_t gid; |
428870ff | 1321 | zfs_acl_ids_t acl_ids; |
9babb374 | 1322 | boolean_t fuid_dirtied; |
428870ff | 1323 | boolean_t have_acl = B_FALSE; |
e8b96c60 | 1324 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1325 | |
1326 | /* | |
1327 | * If we have an ephemeral id, ACL, or XVATTR then | |
1328 | * make sure file system is at proper version | |
1329 | */ | |
1330 | ||
149e873a | 1331 | gid = crgetgid(cr); |
3558fd73 | 1332 | uid = crgetuid(cr); |
b128c09f | 1333 | |
0037b49e | 1334 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1335 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1336 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1337 | |
32dec7bd | 1338 | if (name == NULL) |
1339 | return (SET_ERROR(EINVAL)); | |
1340 | ||
0037b49e | 1341 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1342 | ZFS_VERIFY_ZP(dzp); |
0037b49e BB |
1343 | os = zfsvfs->z_os; |
1344 | zilog = zfsvfs->z_log; | |
34dc7c2f | 1345 | |
0037b49e | 1346 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1347 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1348 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1349 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1350 | } |
1351 | ||
5484965a | 1352 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1353 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1354 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1355 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1356 | return (error); |
1357 | } | |
1358 | } | |
34dc7c2f | 1359 | |
3558fd73 BB |
1360 | top: |
1361 | *ipp = NULL; | |
34dc7c2f BB |
1362 | if (*name == '\0') { |
1363 | /* | |
1364 | * Null component name refers to the directory itself. | |
1365 | */ | |
3558fd73 | 1366 | igrab(dip); |
34dc7c2f BB |
1367 | zp = dzp; |
1368 | dl = NULL; | |
1369 | error = 0; | |
1370 | } else { | |
3558fd73 | 1371 | /* possible igrab(zp) */ |
34dc7c2f BB |
1372 | int zflg = 0; |
1373 | ||
1374 | if (flag & FIGNORECASE) | |
1375 | zflg |= ZCILOOK; | |
1376 | ||
1377 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1378 | NULL, NULL); | |
1379 | if (error) { | |
572e2857 BB |
1380 | if (have_acl) |
1381 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1382 | if (strcmp(name, "..") == 0) |
2e528b49 | 1383 | error = SET_ERROR(EISDIR); |
0037b49e | 1384 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1385 | return (error); |
1386 | } | |
1387 | } | |
428870ff | 1388 | |
34dc7c2f BB |
1389 | if (zp == NULL) { |
1390 | uint64_t txtype; | |
9c5167d1 | 1391 | uint64_t projid = ZFS_DEFAULT_PROJID; |
34dc7c2f BB |
1392 | |
1393 | /* | |
1394 | * Create a new file object and update the directory | |
1395 | * to reference it. | |
1396 | */ | |
149e873a | 1397 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1398 | if (have_acl) |
1399 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1400 | goto out; |
1401 | } | |
1402 | ||
1403 | /* | |
1404 | * We only support the creation of regular files in | |
1405 | * extended attribute directories. | |
1406 | */ | |
428870ff | 1407 | |
3558fd73 | 1408 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1409 | if (have_acl) |
1410 | zfs_acl_ids_free(&acl_ids); | |
2e528b49 | 1411 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1412 | goto out; |
1413 | } | |
1414 | ||
428870ff BB |
1415 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1416 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1417 | goto out; |
428870ff BB |
1418 | have_acl = B_TRUE; |
1419 | ||
9c5167d1 NF |
1420 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
1421 | projid = zfs_inherit_projid(dzp); | |
1422 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
45d1cae3 | 1423 | zfs_acl_ids_free(&acl_ids); |
2e528b49 | 1424 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1425 | goto out; |
1426 | } | |
1427 | ||
34dc7c2f | 1428 | tx = dmu_tx_create(os); |
428870ff BB |
1429 | |
1430 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1431 | ZFS_SA_BASE_ATTR_SIZE); | |
1432 | ||
0037b49e | 1433 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 1434 | if (fuid_dirtied) |
0037b49e | 1435 | zfs_fuid_txhold(zfsvfs, tx); |
34dc7c2f | 1436 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1437 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
0037b49e | 1438 | if (!zfsvfs->z_use_sa && |
428870ff | 1439 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1440 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1441 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1442 | } |
599b8648 | 1443 | |
0735ecb3 PS |
1444 | error = dmu_tx_assign(tx, |
1445 | (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); | |
34dc7c2f BB |
1446 | if (error) { |
1447 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1448 | if (error == ERESTART) { |
e8b96c60 | 1449 | waited = B_TRUE; |
34dc7c2f BB |
1450 | dmu_tx_wait(tx); |
1451 | dmu_tx_abort(tx); | |
1452 | goto top; | |
1453 | } | |
428870ff | 1454 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1455 | dmu_tx_abort(tx); |
0037b49e | 1456 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1457 | return (error); |
1458 | } | |
428870ff | 1459 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 1460 | |
599b8648 CC |
1461 | error = zfs_link_create(dl, zp, tx, ZNEW); |
1462 | if (error != 0) { | |
1463 | /* | |
1464 | * Since, we failed to add the directory entry for it, | |
1465 | * delete the newly created dnode. | |
1466 | */ | |
1467 | zfs_znode_delete(zp, tx); | |
1468 | remove_inode_hash(ZTOI(zp)); | |
1469 | zfs_acl_ids_free(&acl_ids); | |
1470 | dmu_tx_commit(tx); | |
1471 | goto out; | |
1472 | } | |
1473 | ||
9babb374 | 1474 | if (fuid_dirtied) |
0037b49e | 1475 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 1476 | |
34dc7c2f BB |
1477 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); |
1478 | if (flag & FIGNORECASE) | |
1479 | txtype |= TX_CI; | |
1480 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1481 | vsecp, acl_ids.z_fuidp, vap); |
1482 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1483 | dmu_tx_commit(tx); |
1484 | } else { | |
1485 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1486 | ||
572e2857 BB |
1487 | if (have_acl) |
1488 | zfs_acl_ids_free(&acl_ids); | |
1489 | have_acl = B_FALSE; | |
1490 | ||
34dc7c2f BB |
1491 | /* |
1492 | * A directory entry already exists for this name. | |
1493 | */ | |
1494 | /* | |
1495 | * Can't truncate an existing file if in exclusive mode. | |
1496 | */ | |
3558fd73 | 1497 | if (excl) { |
2e528b49 | 1498 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1499 | goto out; |
1500 | } | |
1501 | /* | |
1502 | * Can't open a directory for writing. | |
1503 | */ | |
3558fd73 | 1504 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
2e528b49 | 1505 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1506 | goto out; |
1507 | } | |
1508 | /* | |
1509 | * Verify requested access to file. | |
1510 | */ | |
1511 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1512 | goto out; | |
1513 | } | |
1514 | ||
1515 | mutex_enter(&dzp->z_lock); | |
1516 | dzp->z_seq++; | |
1517 | mutex_exit(&dzp->z_lock); | |
1518 | ||
1519 | /* | |
1520 | * Truncate regular files if requested. | |
1521 | */ | |
3558fd73 BB |
1522 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1523 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f | 1524 | /* we can't hold any locks when calling zfs_freesp() */ |
609603a5 B |
1525 | if (dl) { |
1526 | zfs_dirent_unlock(dl); | |
1527 | dl = NULL; | |
1528 | } | |
34dc7c2f | 1529 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1530 | } |
1531 | } | |
1532 | out: | |
1533 | ||
1534 | if (dl) | |
1535 | zfs_dirent_unlock(dl); | |
1536 | ||
1537 | if (error) { | |
1538 | if (zp) | |
3558fd73 | 1539 | iput(ZTOI(zp)); |
34dc7c2f | 1540 | } else { |
960e08fe BB |
1541 | zfs_inode_update(dzp); |
1542 | zfs_inode_update(zp); | |
3558fd73 | 1543 | *ipp = ZTOI(zp); |
34dc7c2f | 1544 | } |
34dc7c2f | 1545 | |
0037b49e | 1546 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1547 | zil_commit(zilog, 0); |
428870ff | 1548 | |
0037b49e | 1549 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1550 | return (error); |
1551 | } | |
1552 | ||
ace1eae8 CC |
1553 | /* ARGSUSED */ |
1554 | int | |
1555 | zfs_tmpfile(struct inode *dip, vattr_t *vap, int excl, | |
1556 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
1557 | { | |
1558 | znode_t *zp = NULL, *dzp = ITOZ(dip); | |
0037b49e | 1559 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
ace1eae8 CC |
1560 | objset_t *os; |
1561 | dmu_tx_t *tx; | |
1562 | int error; | |
1563 | uid_t uid; | |
1564 | gid_t gid; | |
1565 | zfs_acl_ids_t acl_ids; | |
9c5167d1 | 1566 | uint64_t projid = ZFS_DEFAULT_PROJID; |
ace1eae8 CC |
1567 | boolean_t fuid_dirtied; |
1568 | boolean_t have_acl = B_FALSE; | |
1569 | boolean_t waited = B_FALSE; | |
1570 | ||
1571 | /* | |
1572 | * If we have an ephemeral id, ACL, or XVATTR then | |
1573 | * make sure file system is at proper version | |
1574 | */ | |
1575 | ||
1576 | gid = crgetgid(cr); | |
1577 | uid = crgetuid(cr); | |
1578 | ||
0037b49e | 1579 | if (zfsvfs->z_use_fuids == B_FALSE && |
ace1eae8 CC |
1580 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
1581 | return (SET_ERROR(EINVAL)); | |
1582 | ||
0037b49e | 1583 | ZFS_ENTER(zfsvfs); |
ace1eae8 | 1584 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1585 | os = zfsvfs->z_os; |
ace1eae8 CC |
1586 | |
1587 | if (vap->va_mask & ATTR_XVATTR) { | |
1588 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
1589 | crgetuid(cr), cr, vap->va_mode)) != 0) { | |
0037b49e | 1590 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1591 | return (error); |
1592 | } | |
1593 | } | |
1594 | ||
1595 | top: | |
1596 | *ipp = NULL; | |
1597 | ||
1598 | /* | |
1599 | * Create a new file object and update the directory | |
1600 | * to reference it. | |
1601 | */ | |
1602 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { | |
1603 | if (have_acl) | |
1604 | zfs_acl_ids_free(&acl_ids); | |
1605 | goto out; | |
1606 | } | |
1607 | ||
1608 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, | |
1609 | cr, vsecp, &acl_ids)) != 0) | |
1610 | goto out; | |
1611 | have_acl = B_TRUE; | |
1612 | ||
9c5167d1 NF |
1613 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) |
1614 | projid = zfs_inherit_projid(dzp); | |
1615 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, projid)) { | |
ace1eae8 CC |
1616 | zfs_acl_ids_free(&acl_ids); |
1617 | error = SET_ERROR(EDQUOT); | |
1618 | goto out; | |
1619 | } | |
1620 | ||
1621 | tx = dmu_tx_create(os); | |
1622 | ||
1623 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1624 | ZFS_SA_BASE_ATTR_SIZE); | |
0037b49e | 1625 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 1626 | |
0037b49e | 1627 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
ace1eae8 | 1628 | if (fuid_dirtied) |
0037b49e BB |
1629 | zfs_fuid_txhold(zfsvfs, tx); |
1630 | if (!zfsvfs->z_use_sa && | |
ace1eae8 CC |
1631 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
1632 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
1633 | 0, acl_ids.z_aclp->z_acl_bytes); | |
1634 | } | |
0735ecb3 | 1635 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
ace1eae8 CC |
1636 | if (error) { |
1637 | if (error == ERESTART) { | |
1638 | waited = B_TRUE; | |
1639 | dmu_tx_wait(tx); | |
1640 | dmu_tx_abort(tx); | |
1641 | goto top; | |
1642 | } | |
1643 | zfs_acl_ids_free(&acl_ids); | |
1644 | dmu_tx_abort(tx); | |
0037b49e | 1645 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1646 | return (error); |
1647 | } | |
1648 | zfs_mknode(dzp, vap, tx, cr, IS_TMPFILE, &zp, &acl_ids); | |
1649 | ||
1650 | if (fuid_dirtied) | |
0037b49e | 1651 | zfs_fuid_sync(zfsvfs, tx); |
ace1eae8 CC |
1652 | |
1653 | /* Add to unlinked set */ | |
1654 | zp->z_unlinked = 1; | |
1655 | zfs_unlinked_add(zp, tx); | |
1656 | zfs_acl_ids_free(&acl_ids); | |
1657 | dmu_tx_commit(tx); | |
1658 | out: | |
1659 | ||
1660 | if (error) { | |
1661 | if (zp) | |
1662 | iput(ZTOI(zp)); | |
1663 | } else { | |
1664 | zfs_inode_update(dzp); | |
1665 | zfs_inode_update(zp); | |
1666 | *ipp = ZTOI(zp); | |
1667 | } | |
1668 | ||
0037b49e | 1669 | ZFS_EXIT(zfsvfs); |
ace1eae8 CC |
1670 | return (error); |
1671 | } | |
1672 | ||
34dc7c2f BB |
1673 | /* |
1674 | * Remove an entry from a directory. | |
1675 | * | |
3558fd73 | 1676 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1677 | * name - name of entry to remove. |
1678 | * cr - credentials of caller. | |
34dc7c2f BB |
1679 | * |
1680 | * RETURN: 0 if success | |
1681 | * error code if failure | |
1682 | * | |
1683 | * Timestamps: | |
3558fd73 BB |
1684 | * dip - ctime|mtime |
1685 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1686 | */ |
428870ff BB |
1687 | |
1688 | uint64_t null_xattr = 0; | |
1689 | ||
34dc7c2f | 1690 | /*ARGSUSED*/ |
e5c39b95 | 1691 | int |
da5e151f | 1692 | zfs_remove(struct inode *dip, char *name, cred_t *cr, int flags) |
34dc7c2f | 1693 | { |
3558fd73 | 1694 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1695 | znode_t *xzp; |
3558fd73 | 1696 | struct inode *ip; |
0037b49e | 1697 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 1698 | zilog_t *zilog; |
a966c564 | 1699 | uint64_t acl_obj, xattr_obj; |
3558fd73 | 1700 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 1701 | uint64_t obj = 0; |
dfbc8630 | 1702 | uint64_t links; |
34dc7c2f BB |
1703 | zfs_dirlock_t *dl; |
1704 | dmu_tx_t *tx; | |
a966c564 K |
1705 | boolean_t may_delete_now, delete_now = FALSE; |
1706 | boolean_t unlinked, toobig = FALSE; | |
34dc7c2f BB |
1707 | uint64_t txtype; |
1708 | pathname_t *realnmp = NULL; | |
1709 | pathname_t realnm; | |
1710 | int error; | |
1711 | int zflg = ZEXISTS; | |
e8b96c60 | 1712 | boolean_t waited = B_FALSE; |
34dc7c2f | 1713 | |
32dec7bd | 1714 | if (name == NULL) |
1715 | return (SET_ERROR(EINVAL)); | |
1716 | ||
0037b49e | 1717 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1718 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1719 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
1720 | |
1721 | if (flags & FIGNORECASE) { | |
1722 | zflg |= ZCILOOK; | |
1723 | pn_alloc(&realnm); | |
1724 | realnmp = &realnm; | |
1725 | } | |
1726 | ||
1727 | top: | |
572e2857 BB |
1728 | xattr_obj = 0; |
1729 | xzp = NULL; | |
34dc7c2f BB |
1730 | /* |
1731 | * Attempt to lock directory; fail if entry doesn't exist. | |
1732 | */ | |
149e873a BB |
1733 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1734 | NULL, realnmp))) { | |
34dc7c2f BB |
1735 | if (realnmp) |
1736 | pn_free(realnmp); | |
0037b49e | 1737 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1738 | return (error); |
1739 | } | |
1740 | ||
3558fd73 | 1741 | ip = ZTOI(zp); |
34dc7c2f | 1742 | |
149e873a | 1743 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1744 | goto out; |
1745 | } | |
1746 | ||
1747 | /* | |
1748 | * Need to use rmdir for removing directories. | |
1749 | */ | |
3558fd73 | 1750 | if (S_ISDIR(ip->i_mode)) { |
2e528b49 | 1751 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1752 | goto out; |
1753 | } | |
1754 | ||
19d55079 MA |
1755 | mutex_enter(&zp->z_lock); |
1756 | may_delete_now = atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped); | |
1757 | mutex_exit(&zp->z_lock); | |
1758 | ||
34dc7c2f | 1759 | /* |
a966c564 K |
1760 | * We may delete the znode now, or we may put it in the unlinked set; |
1761 | * it depends on whether we're the last link, and on whether there are | |
1762 | * other holds on the inode. So we dmu_tx_hold() the right things to | |
1763 | * allow for either case. | |
34dc7c2f | 1764 | */ |
572e2857 | 1765 | obj = zp->z_id; |
0037b49e | 1766 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 1767 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1768 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1769 | zfs_sa_upgrade_txholds(tx, zp); | |
1770 | zfs_sa_upgrade_txholds(tx, dzp); | |
a966c564 K |
1771 | if (may_delete_now) { |
1772 | toobig = zp->z_size > zp->z_blksz * zfs_delete_blocks; | |
1773 | /* if the file is too big, only hold_free a token amount */ | |
1774 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1775 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1776 | } | |
34dc7c2f BB |
1777 | |
1778 | /* are there any extended attributes? */ | |
0037b49e | 1779 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1780 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1781 | if (error == 0 && xattr_obj) { |
0037b49e | 1782 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
c99c9001 | 1783 | ASSERT0(error); |
428870ff BB |
1784 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1785 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1786 | } |
1787 | ||
a966c564 K |
1788 | mutex_enter(&zp->z_lock); |
1789 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
1790 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); | |
1791 | mutex_exit(&zp->z_lock); | |
1792 | ||
34dc7c2f | 1793 | /* charge as an update -- would be nice not to charge at all */ |
0037b49e | 1794 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1795 | |
19d55079 | 1796 | /* |
1a04bab3 | 1797 | * Mark this transaction as typically resulting in a net free of space |
19d55079 | 1798 | */ |
1a04bab3 | 1799 | dmu_tx_mark_netfree(tx); |
19d55079 | 1800 | |
0735ecb3 | 1801 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
1802 | if (error) { |
1803 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1804 | if (error == ERESTART) { |
e8b96c60 | 1805 | waited = B_TRUE; |
34dc7c2f BB |
1806 | dmu_tx_wait(tx); |
1807 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
1808 | iput(ip); |
1809 | if (xzp) | |
1810 | iput(ZTOI(xzp)); | |
34dc7c2f BB |
1811 | goto top; |
1812 | } | |
1813 | if (realnmp) | |
1814 | pn_free(realnmp); | |
1815 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
1816 | iput(ip); |
1817 | if (xzp) | |
1818 | iput(ZTOI(xzp)); | |
0037b49e | 1819 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1820 | return (error); |
1821 | } | |
1822 | ||
1823 | /* | |
1824 | * Remove the directory entry. | |
1825 | */ | |
1826 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1827 | ||
1828 | if (error) { | |
1829 | dmu_tx_commit(tx); | |
1830 | goto out; | |
1831 | } | |
1832 | ||
1833 | if (unlinked) { | |
572e2857 BB |
1834 | /* |
1835 | * Hold z_lock so that we can make sure that the ACL obj | |
1836 | * hasn't changed. Could have been deleted due to | |
1837 | * zfs_sa_upgrade(). | |
1838 | */ | |
1839 | mutex_enter(&zp->z_lock); | |
0037b49e | 1840 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
428870ff | 1841 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
a966c564 K |
1842 | delete_now = may_delete_now && !toobig && |
1843 | atomic_read(&ip->i_count) == 1 && !(zp->z_is_mapped) && | |
1844 | xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == | |
1845 | acl_obj; | |
1846 | } | |
1847 | ||
1848 | if (delete_now) { | |
1849 | if (xattr_obj_unlinked) { | |
dfbc8630 | 1850 | ASSERT3U(ZTOI(xzp)->i_nlink, ==, 2); |
a966c564 K |
1851 | mutex_enter(&xzp->z_lock); |
1852 | xzp->z_unlinked = 1; | |
dfbc8630 CD |
1853 | clear_nlink(ZTOI(xzp)); |
1854 | links = 0; | |
0037b49e | 1855 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), |
dfbc8630 | 1856 | &links, sizeof (links), tx); |
a966c564 K |
1857 | ASSERT3U(error, ==, 0); |
1858 | mutex_exit(&xzp->z_lock); | |
1859 | zfs_unlinked_add(xzp, tx); | |
1860 | ||
1861 | if (zp->z_is_sa) | |
1862 | error = sa_remove(zp->z_sa_hdl, | |
0037b49e | 1863 | SA_ZPL_XATTR(zfsvfs), tx); |
a966c564 K |
1864 | else |
1865 | error = sa_update(zp->z_sa_hdl, | |
0037b49e | 1866 | SA_ZPL_XATTR(zfsvfs), &null_xattr, |
a966c564 K |
1867 | sizeof (uint64_t), tx); |
1868 | ASSERT0(error); | |
1869 | } | |
1870 | /* | |
1871 | * Add to the unlinked set because a new reference could be | |
1872 | * taken concurrently resulting in a deferred destruction. | |
1873 | */ | |
1874 | zfs_unlinked_add(zp, tx); | |
1875 | mutex_exit(&zp->z_lock); | |
a966c564 | 1876 | } else if (unlinked) { |
572e2857 | 1877 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1878 | zfs_unlinked_add(zp, tx); |
1879 | } | |
1880 | ||
1881 | txtype = TX_REMOVE; | |
1882 | if (flags & FIGNORECASE) | |
1883 | txtype |= TX_CI; | |
572e2857 | 1884 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1885 | |
1886 | dmu_tx_commit(tx); | |
1887 | out: | |
1888 | if (realnmp) | |
1889 | pn_free(realnmp); | |
1890 | ||
1891 | zfs_dirent_unlock(dl); | |
960e08fe | 1892 | zfs_inode_update(dzp); |
ea7e86d8 | 1893 | zfs_inode_update(zp); |
34dc7c2f | 1894 | |
ea7e86d8 BB |
1895 | if (delete_now) |
1896 | iput(ip); | |
1897 | else | |
a966c564 | 1898 | zfs_iput_async(ip); |
a966c564 K |
1899 | |
1900 | if (xzp) { | |
1901 | zfs_inode_update(xzp); | |
1902 | zfs_iput_async(ZTOI(xzp)); | |
1903 | } | |
428870ff | 1904 | |
0037b49e | 1905 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1906 | zil_commit(zilog, 0); |
34dc7c2f | 1907 | |
0037b49e | 1908 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1909 | return (error); |
1910 | } | |
1911 | ||
1912 | /* | |
3558fd73 | 1913 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1914 | * provided. Return a pointer to the inserted directory. |
1915 | * | |
3558fd73 | 1916 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1917 | * dirname - name of new directory. |
1918 | * vap - attributes of new directory. | |
1919 | * cr - credentials of caller. | |
34dc7c2f BB |
1920 | * vsecp - ACL to be set |
1921 | * | |
3558fd73 | 1922 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1923 | * |
1924 | * RETURN: 0 if success | |
1925 | * error code if failure | |
1926 | * | |
1927 | * Timestamps: | |
3558fd73 BB |
1928 | * dip - ctime|mtime updated |
1929 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1930 | */ |
1931 | /*ARGSUSED*/ | |
e5c39b95 | 1932 | int |
3558fd73 BB |
1933 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1934 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1935 | { |
3558fd73 | 1936 | znode_t *zp, *dzp = ITOZ(dip); |
0037b49e | 1937 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
1938 | zilog_t *zilog; |
1939 | zfs_dirlock_t *dl; | |
1940 | uint64_t txtype; | |
1941 | dmu_tx_t *tx; | |
1942 | int error; | |
34dc7c2f | 1943 | int zf = ZNEW; |
b128c09f BB |
1944 | uid_t uid; |
1945 | gid_t gid = crgetgid(cr); | |
428870ff | 1946 | zfs_acl_ids_t acl_ids; |
9babb374 | 1947 | boolean_t fuid_dirtied; |
e8b96c60 | 1948 | boolean_t waited = B_FALSE; |
34dc7c2f | 1949 | |
3558fd73 | 1950 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1951 | |
1952 | /* | |
1953 | * If we have an ephemeral id, ACL, or XVATTR then | |
1954 | * make sure file system is at proper version | |
1955 | */ | |
1956 | ||
3558fd73 | 1957 | uid = crgetuid(cr); |
0037b49e | 1958 | if (zfsvfs->z_use_fuids == B_FALSE && |
3558fd73 | 1959 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
2e528b49 | 1960 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1961 | |
32dec7bd | 1962 | if (dirname == NULL) |
1963 | return (SET_ERROR(EINVAL)); | |
1964 | ||
0037b49e | 1965 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1966 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 1967 | zilog = zfsvfs->z_log; |
34dc7c2f | 1968 | |
428870ff | 1969 | if (dzp->z_pflags & ZFS_XATTR) { |
0037b49e | 1970 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1971 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1972 | } |
1973 | ||
0037b49e | 1974 | if (zfsvfs->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1975 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 1976 | ZFS_EXIT(zfsvfs); |
2e528b49 | 1977 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1978 | } |
1979 | if (flags & FIGNORECASE) | |
1980 | zf |= ZCILOOK; | |
1981 | ||
5484965a | 1982 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1983 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 | 1984 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
0037b49e | 1985 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1986 | return (error); |
1987 | } | |
428870ff | 1988 | } |
34dc7c2f | 1989 | |
428870ff BB |
1990 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1991 | vsecp, &acl_ids)) != 0) { | |
0037b49e | 1992 | ZFS_EXIT(zfsvfs); |
428870ff BB |
1993 | return (error); |
1994 | } | |
34dc7c2f BB |
1995 | /* |
1996 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1997 | * |
1998 | * Existence is checked first to make sure we don't return | |
1999 | * EACCES instead of EEXIST which can cause some applications | |
2000 | * to fail. | |
34dc7c2f BB |
2001 | */ |
2002 | top: | |
3558fd73 | 2003 | *ipp = NULL; |
34dc7c2f | 2004 | |
149e873a BB |
2005 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
2006 | NULL, NULL))) { | |
428870ff | 2007 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 2008 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2009 | return (error); |
2010 | } | |
2011 | ||
149e873a | 2012 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 2013 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2014 | zfs_dirent_unlock(dl); |
0037b49e | 2015 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2016 | return (error); |
2017 | } | |
2018 | ||
9c5167d1 | 2019 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, zfs_inherit_projid(dzp))) { |
45d1cae3 | 2020 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 2021 | zfs_dirent_unlock(dl); |
0037b49e | 2022 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2023 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
2024 | } |
2025 | ||
34dc7c2f BB |
2026 | /* |
2027 | * Add a new entry to the directory. | |
2028 | */ | |
0037b49e | 2029 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f BB |
2030 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
2031 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
0037b49e | 2032 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 2033 | if (fuid_dirtied) |
0037b49e BB |
2034 | zfs_fuid_txhold(zfsvfs, tx); |
2035 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
2036 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
2037 | acl_ids.z_aclp->z_acl_bytes); | |
2038 | } | |
2039 | ||
2040 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
2041 | ZFS_SA_BASE_ATTR_SIZE); | |
2042 | ||
0735ecb3 | 2043 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
2044 | if (error) { |
2045 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2046 | if (error == ERESTART) { |
e8b96c60 | 2047 | waited = B_TRUE; |
34dc7c2f BB |
2048 | dmu_tx_wait(tx); |
2049 | dmu_tx_abort(tx); | |
2050 | goto top; | |
2051 | } | |
428870ff | 2052 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 2053 | dmu_tx_abort(tx); |
0037b49e | 2054 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2055 | return (error); |
2056 | } | |
2057 | ||
2058 | /* | |
2059 | * Create new node. | |
2060 | */ | |
428870ff | 2061 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f BB |
2062 | |
2063 | /* | |
2064 | * Now put new name in parent dir. | |
2065 | */ | |
599b8648 CC |
2066 | error = zfs_link_create(dl, zp, tx, ZNEW); |
2067 | if (error != 0) { | |
2068 | zfs_znode_delete(zp, tx); | |
2069 | remove_inode_hash(ZTOI(zp)); | |
2070 | goto out; | |
2071 | } | |
2072 | ||
2073 | if (fuid_dirtied) | |
2074 | zfs_fuid_sync(zfsvfs, tx); | |
34dc7c2f | 2075 | |
3558fd73 | 2076 | *ipp = ZTOI(zp); |
34dc7c2f BB |
2077 | |
2078 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
2079 | if (flags & FIGNORECASE) | |
2080 | txtype |= TX_CI; | |
9babb374 BB |
2081 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
2082 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 2083 | |
599b8648 | 2084 | out: |
9babb374 | 2085 | zfs_acl_ids_free(&acl_ids); |
428870ff | 2086 | |
34dc7c2f BB |
2087 | dmu_tx_commit(tx); |
2088 | ||
2089 | zfs_dirent_unlock(dl); | |
2090 | ||
0037b49e | 2091 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2092 | zil_commit(zilog, 0); |
428870ff | 2093 | |
599b8648 CC |
2094 | if (error != 0) { |
2095 | iput(ZTOI(zp)); | |
2096 | } else { | |
2097 | zfs_inode_update(dzp); | |
2098 | zfs_inode_update(zp); | |
2099 | } | |
0037b49e | 2100 | ZFS_EXIT(zfsvfs); |
599b8648 | 2101 | return (error); |
34dc7c2f BB |
2102 | } |
2103 | ||
2104 | /* | |
2105 | * Remove a directory subdir entry. If the current working | |
2106 | * directory is the same as the subdir to be removed, the | |
2107 | * remove will fail. | |
2108 | * | |
3558fd73 | 2109 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 2110 | * name - name of directory to be removed. |
3558fd73 | 2111 | * cwd - inode of current working directory. |
34dc7c2f | 2112 | * cr - credentials of caller. |
34dc7c2f BB |
2113 | * flags - case flags |
2114 | * | |
d3cc8b15 | 2115 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
2116 | * |
2117 | * Timestamps: | |
3558fd73 | 2118 | * dip - ctime|mtime updated |
34dc7c2f BB |
2119 | */ |
2120 | /*ARGSUSED*/ | |
e5c39b95 | 2121 | int |
3558fd73 BB |
2122 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
2123 | int flags) | |
34dc7c2f | 2124 | { |
3558fd73 | 2125 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 2126 | znode_t *zp; |
3558fd73 | 2127 | struct inode *ip; |
0037b49e | 2128 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f BB |
2129 | zilog_t *zilog; |
2130 | zfs_dirlock_t *dl; | |
2131 | dmu_tx_t *tx; | |
2132 | int error; | |
2133 | int zflg = ZEXISTS; | |
e8b96c60 | 2134 | boolean_t waited = B_FALSE; |
34dc7c2f | 2135 | |
32dec7bd | 2136 | if (name == NULL) |
2137 | return (SET_ERROR(EINVAL)); | |
2138 | ||
0037b49e | 2139 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2140 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 2141 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2142 | |
2143 | if (flags & FIGNORECASE) | |
2144 | zflg |= ZCILOOK; | |
2145 | top: | |
2146 | zp = NULL; | |
2147 | ||
2148 | /* | |
2149 | * Attempt to lock directory; fail if entry doesn't exist. | |
2150 | */ | |
149e873a BB |
2151 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
2152 | NULL, NULL))) { | |
0037b49e | 2153 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2154 | return (error); |
2155 | } | |
2156 | ||
3558fd73 | 2157 | ip = ZTOI(zp); |
34dc7c2f | 2158 | |
149e873a | 2159 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
2160 | goto out; |
2161 | } | |
2162 | ||
3558fd73 | 2163 | if (!S_ISDIR(ip->i_mode)) { |
2e528b49 | 2164 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
2165 | goto out; |
2166 | } | |
2167 | ||
3558fd73 | 2168 | if (ip == cwd) { |
2e528b49 | 2169 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
2170 | goto out; |
2171 | } | |
2172 | ||
34dc7c2f | 2173 | /* |
4e33ba4c | 2174 | * Grab a lock on the directory to make sure that no one is |
34dc7c2f BB |
2175 | * trying to add (or lookup) entries while we are removing it. |
2176 | */ | |
2177 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
2178 | ||
2179 | /* | |
2180 | * Grab a lock on the parent pointer to make sure we play well | |
2181 | * with the treewalk and directory rename code. | |
2182 | */ | |
2183 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
2184 | ||
0037b49e | 2185 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 2186 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 2187 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
0037b49e | 2188 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
2189 | zfs_sa_upgrade_txholds(tx, zp); |
2190 | zfs_sa_upgrade_txholds(tx, dzp); | |
db707ad0 | 2191 | dmu_tx_mark_netfree(tx); |
0735ecb3 | 2192 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
2193 | if (error) { |
2194 | rw_exit(&zp->z_parent_lock); | |
2195 | rw_exit(&zp->z_name_lock); | |
2196 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 2197 | if (error == ERESTART) { |
e8b96c60 | 2198 | waited = B_TRUE; |
34dc7c2f BB |
2199 | dmu_tx_wait(tx); |
2200 | dmu_tx_abort(tx); | |
ea7e86d8 | 2201 | iput(ip); |
34dc7c2f BB |
2202 | goto top; |
2203 | } | |
2204 | dmu_tx_abort(tx); | |
ea7e86d8 | 2205 | iput(ip); |
0037b49e | 2206 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2207 | return (error); |
2208 | } | |
2209 | ||
2210 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
2211 | ||
2212 | if (error == 0) { | |
2213 | uint64_t txtype = TX_RMDIR; | |
2214 | if (flags & FIGNORECASE) | |
2215 | txtype |= TX_CI; | |
572e2857 | 2216 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
2217 | } |
2218 | ||
2219 | dmu_tx_commit(tx); | |
2220 | ||
2221 | rw_exit(&zp->z_parent_lock); | |
2222 | rw_exit(&zp->z_name_lock); | |
2223 | out: | |
2224 | zfs_dirent_unlock(dl); | |
2225 | ||
59157910 BB |
2226 | zfs_inode_update(dzp); |
2227 | zfs_inode_update(zp); | |
3558fd73 | 2228 | iput(ip); |
34dc7c2f | 2229 | |
0037b49e | 2230 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2231 | zil_commit(zilog, 0); |
428870ff | 2232 | |
0037b49e | 2233 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2234 | return (error); |
2235 | } | |
2236 | ||
2237 | /* | |
2238 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 2239 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 2240 | * |
3558fd73 BB |
2241 | * IN: ip - inode of directory to read. |
2242 | * dirent - buffer for directory entries. | |
34dc7c2f | 2243 | * |
3558fd73 | 2244 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
2245 | * |
2246 | * RETURN: 0 if success | |
2247 | * error code if failure | |
2248 | * | |
2249 | * Timestamps: | |
3558fd73 | 2250 | * ip - atime updated |
34dc7c2f BB |
2251 | * |
2252 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2253 | * This allows us to use the low range for "special" directory entries: | |
2254 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2255 | * we use the offset 2 for the '.zfs' directory. | |
2256 | */ | |
2257 | /* ARGSUSED */ | |
3558fd73 | 2258 | int |
9464b959 | 2259 | zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) |
34dc7c2f | 2260 | { |
3558fd73 | 2261 | znode_t *zp = ITOZ(ip); |
0037b49e | 2262 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2263 | objset_t *os; |
34dc7c2f BB |
2264 | zap_cursor_t zc; |
2265 | zap_attribute_t zap; | |
34dc7c2f BB |
2266 | int error; |
2267 | uint8_t prefetch; | |
c12e3a59 | 2268 | uint8_t type; |
3558fd73 BB |
2269 | int done = 0; |
2270 | uint64_t parent; | |
c12e3a59 | 2271 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2272 | |
0037b49e | 2273 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2274 | ZFS_VERIFY_ZP(zp); |
2275 | ||
0037b49e | 2276 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
3558fd73 BB |
2277 | &parent, sizeof (parent))) != 0) |
2278 | goto out; | |
34dc7c2f BB |
2279 | |
2280 | /* | |
2281 | * Quit if directory has been removed (posix) | |
2282 | */ | |
3558fd73 BB |
2283 | if (zp->z_unlinked) |
2284 | goto out; | |
2285 | ||
c12e3a59 | 2286 | error = 0; |
0037b49e | 2287 | os = zfsvfs->z_os; |
c12e3a59 | 2288 | offset = ctx->pos; |
34dc7c2f BB |
2289 | prefetch = zp->z_zn_prefetch; |
2290 | ||
2291 | /* | |
2292 | * Initialize the iterator cursor. | |
2293 | */ | |
c12e3a59 | 2294 | if (offset <= 3) { |
34dc7c2f BB |
2295 | /* |
2296 | * Start iteration from the beginning of the directory. | |
2297 | */ | |
2298 | zap_cursor_init(&zc, os, zp->z_id); | |
2299 | } else { | |
2300 | /* | |
2301 | * The offset is a serialized cursor. | |
2302 | */ | |
c12e3a59 | 2303 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2304 | } |
2305 | ||
34dc7c2f BB |
2306 | /* |
2307 | * Transform to file-system independent format | |
2308 | */ | |
3558fd73 BB |
2309 | while (!done) { |
2310 | uint64_t objnum; | |
34dc7c2f BB |
2311 | /* |
2312 | * Special case `.', `..', and `.zfs'. | |
2313 | */ | |
c12e3a59 | 2314 | if (offset == 0) { |
34dc7c2f BB |
2315 | (void) strcpy(zap.za_name, "."); |
2316 | zap.za_normalization_conflict = 0; | |
2317 | objnum = zp->z_id; | |
c12e3a59 RY |
2318 | type = DT_DIR; |
2319 | } else if (offset == 1) { | |
34dc7c2f BB |
2320 | (void) strcpy(zap.za_name, ".."); |
2321 | zap.za_normalization_conflict = 0; | |
428870ff | 2322 | objnum = parent; |
c12e3a59 RY |
2323 | type = DT_DIR; |
2324 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2325 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2326 | zap.za_normalization_conflict = 0; | |
2327 | objnum = ZFSCTL_INO_ROOT; | |
c12e3a59 | 2328 | type = DT_DIR; |
34dc7c2f BB |
2329 | } else { |
2330 | /* | |
2331 | * Grab next entry. | |
2332 | */ | |
3558fd73 BB |
2333 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2334 | if (error == ENOENT) | |
34dc7c2f BB |
2335 | break; |
2336 | else | |
2337 | goto update; | |
2338 | } | |
2339 | ||
0c5dde49 BB |
2340 | /* |
2341 | * Allow multiple entries provided the first entry is | |
2342 | * the object id. Non-zpl consumers may safely make | |
2343 | * use of the additional space. | |
2344 | * | |
2345 | * XXX: This should be a feature flag for compatibility | |
2346 | */ | |
34dc7c2f | 2347 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2348 | zap.za_num_integers == 0) { |
34dc7c2f | 2349 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2350 | "entry, obj = %lld, offset = %lld, " |
2351 | "length = %d, num = %lld\n", | |
34dc7c2f | 2352 | (u_longlong_t)zp->z_id, |
c12e3a59 | 2353 | (u_longlong_t)offset, |
0c5dde49 BB |
2354 | zap.za_integer_length, |
2355 | (u_longlong_t)zap.za_num_integers); | |
2e528b49 | 2356 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2357 | goto update; |
2358 | } | |
2359 | ||
2360 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
c12e3a59 | 2361 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2362 | } |
0f37d0c8 | 2363 | |
9464b959 | 2364 | done = !zpl_dir_emit(ctx, zap.za_name, strlen(zap.za_name), |
c12e3a59 | 2365 | objnum, type); |
0f37d0c8 | 2366 | if (done) |
34dc7c2f | 2367 | break; |
34dc7c2f BB |
2368 | |
2369 | /* Prefetch znode */ | |
3558fd73 | 2370 | if (prefetch) { |
fcff0f35 PD |
2371 | dmu_prefetch(os, objnum, 0, 0, 0, |
2372 | ZIO_PRIORITY_SYNC_READ); | |
3558fd73 | 2373 | } |
34dc7c2f | 2374 | |
c12e3a59 RY |
2375 | /* |
2376 | * Move to the next entry, fill in the previous offset. | |
2377 | */ | |
2378 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2379 | zap_cursor_advance(&zc); |
c12e3a59 | 2380 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2381 | } else { |
c12e3a59 | 2382 | offset += 1; |
34dc7c2f | 2383 | } |
c12e3a59 | 2384 | ctx->pos = offset; |
34dc7c2f BB |
2385 | } |
2386 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2387 | ||
34dc7c2f BB |
2388 | update: |
2389 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2390 | if (error == ENOENT) |
2391 | error = 0; | |
3558fd73 | 2392 | out: |
0037b49e | 2393 | ZFS_EXIT(zfsvfs); |
34dc7c2f | 2394 | |
34dc7c2f BB |
2395 | return (error); |
2396 | } | |
2397 | ||
d5446cfc BB |
2398 | ulong_t zfs_fsync_sync_cnt = 4; |
2399 | ||
e5c39b95 | 2400 | int |
3558fd73 | 2401 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2402 | { |
3558fd73 | 2403 | znode_t *zp = ITOZ(ip); |
0037b49e | 2404 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 2405 | |
d5446cfc BB |
2406 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2407 | ||
0037b49e BB |
2408 | if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2409 | ZFS_ENTER(zfsvfs); | |
428870ff | 2410 | ZFS_VERIFY_ZP(zp); |
0037b49e BB |
2411 | zil_commit(zfsvfs->z_log, zp->z_id); |
2412 | ZFS_EXIT(zfsvfs); | |
428870ff | 2413 | } |
07012da6 CC |
2414 | tsd_set(zfs_fsyncer_key, NULL); |
2415 | ||
34dc7c2f BB |
2416 | return (0); |
2417 | } | |
2418 | ||
2419 | ||
2420 | /* | |
2421 | * Get the requested file attributes and place them in the provided | |
2422 | * vattr structure. | |
2423 | * | |
3558fd73 | 2424 | * IN: ip - inode of file. |
5484965a BB |
2425 | * vap - va_mask identifies requested attributes. |
2426 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2427 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2428 | * cr - credentials of caller. | |
34dc7c2f | 2429 | * |
5484965a BB |
2430 | * OUT: vap - attribute values. |
2431 | * | |
2432 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2433 | */ |
2434 | /* ARGSUSED */ | |
e5c39b95 | 2435 | int |
5484965a | 2436 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2437 | { |
3558fd73 | 2438 | znode_t *zp = ITOZ(ip); |
0037b49e | 2439 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
2440 | int error = 0; |
2441 | uint64_t links; | |
0df9673f | 2442 | uint64_t atime[2], mtime[2], ctime[2]; |
5484965a BB |
2443 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2444 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2445 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
0df9673f | 2446 | sa_bulk_attr_t bulk[3]; |
428870ff | 2447 | int count = 0; |
34dc7c2f | 2448 | |
0037b49e | 2449 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 2450 | ZFS_VERIFY_ZP(zp); |
428870ff | 2451 | |
5484965a | 2452 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2453 | |
0037b49e BB |
2454 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); |
2455 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
2456 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
428870ff BB |
2457 | |
2458 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
0037b49e | 2459 | ZFS_EXIT(zfsvfs); |
428870ff BB |
2460 | return (error); |
2461 | } | |
34dc7c2f | 2462 | |
34dc7c2f BB |
2463 | /* |
2464 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2465 | * Also, if we are the owner don't bother, since owner should | |
2466 | * always be allowed to read basic attributes of file. | |
2467 | */ | |
572e2857 | 2468 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2469 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2470 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2471 | skipaclchk, cr))) { | |
0037b49e | 2472 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2473 | return (error); |
2474 | } | |
2475 | } | |
2476 | ||
2477 | /* | |
2478 | * Return all attributes. It's cheaper to provide the answer | |
2479 | * than to determine whether we were asked the question. | |
2480 | */ | |
2481 | ||
9babb374 | 2482 | mutex_enter(&zp->z_lock); |
5484965a BB |
2483 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2484 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2485 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2486 | vap->va_nodeid = zp->z_id; |
0037b49e | 2487 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) |
dfbc8630 | 2488 | links = ZTOI(zp)->i_nlink + 1; |
34dc7c2f | 2489 | else |
dfbc8630 | 2490 | links = ZTOI(zp)->i_nlink; |
5484965a BB |
2491 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2492 | vap->va_size = i_size_read(ip); | |
2493 | vap->va_rdev = ip->i_rdev; | |
2494 | vap->va_seq = ip->i_generation; | |
2495 | ||
2496 | /* | |
2497 | * Add in any requested optional attributes and the create time. | |
2498 | * Also set the corresponding bits in the returned attribute bitmap. | |
2499 | */ | |
0037b49e | 2500 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { |
5484965a BB |
2501 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { |
2502 | xoap->xoa_archive = | |
2503 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2504 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2505 | } | |
2506 | ||
2507 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2508 | xoap->xoa_readonly = | |
2509 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2510 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2511 | } | |
2512 | ||
2513 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2514 | xoap->xoa_system = | |
2515 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2516 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2517 | } | |
2518 | ||
2519 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2520 | xoap->xoa_hidden = | |
2521 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2522 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2523 | } | |
2524 | ||
2525 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2526 | xoap->xoa_nounlink = | |
2527 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2528 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2529 | } | |
2530 | ||
2531 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2532 | xoap->xoa_immutable = | |
2533 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2534 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2535 | } | |
2536 | ||
2537 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2538 | xoap->xoa_appendonly = | |
2539 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2540 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2541 | } | |
2542 | ||
2543 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2544 | xoap->xoa_nodump = | |
2545 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2546 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2547 | } | |
2548 | ||
2549 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2550 | xoap->xoa_opaque = | |
2551 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2552 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2553 | } | |
2554 | ||
2555 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2556 | xoap->xoa_av_quarantined = | |
2557 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2558 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2559 | } | |
2560 | ||
2561 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2562 | xoap->xoa_av_modified = | |
2563 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2564 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2565 | } | |
2566 | ||
2567 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2568 | S_ISREG(ip->i_mode)) { | |
2569 | zfs_sa_get_scanstamp(zp, xvap); | |
2570 | } | |
34dc7c2f | 2571 | |
5484965a BB |
2572 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2573 | uint64_t times[2]; | |
2574 | ||
0037b49e | 2575 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), |
5484965a BB |
2576 | times, sizeof (times)); |
2577 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2578 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2579 | } | |
2580 | ||
2581 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2582 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2583 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2584 | } | |
2585 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
278f2236 | 2586 | xoap->xoa_generation = ip->i_generation; |
5484965a BB |
2587 | XVA_SET_RTN(xvap, XAT_GEN); |
2588 | } | |
2589 | ||
2590 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2591 | xoap->xoa_offline = | |
2592 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2593 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2594 | } | |
2595 | ||
2596 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2597 | xoap->xoa_sparse = | |
2598 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2599 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2600 | } | |
9c5167d1 NF |
2601 | |
2602 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { | |
2603 | xoap->xoa_projinherit = | |
2604 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0); | |
2605 | XVA_SET_RTN(xvap, XAT_PROJINHERIT); | |
2606 | } | |
2607 | ||
2608 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
2609 | xoap->xoa_projid = zp->z_projid; | |
2610 | XVA_SET_RTN(xvap, XAT_PROJID); | |
2611 | } | |
5484965a BB |
2612 | } |
2613 | ||
0df9673f | 2614 | ZFS_TIME_DECODE(&vap->va_atime, atime); |
5484965a BB |
2615 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); |
2616 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2617 | |
2618 | mutex_exit(&zp->z_lock); | |
2619 | ||
5484965a | 2620 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2621 | |
2622 | if (zp->z_blksz == 0) { | |
2623 | /* | |
2624 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2625 | */ | |
0037b49e | 2626 | vap->va_blksize = zfsvfs->z_max_blksz; |
34dc7c2f BB |
2627 | } |
2628 | ||
0037b49e | 2629 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2630 | return (0); |
2631 | } | |
2632 | ||
057e8eee BB |
2633 | /* |
2634 | * Get the basic file attributes and place them in the provided kstat | |
2635 | * structure. The inode is assumed to be the authoritative source | |
2636 | * for most of the attributes. However, the znode currently has the | |
2637 | * authoritative atime, blksize, and block count. | |
2638 | * | |
2639 | * IN: ip - inode of file. | |
2640 | * | |
2641 | * OUT: sp - kstat values. | |
2642 | * | |
2643 | * RETURN: 0 (always succeeds) | |
2644 | */ | |
2645 | /* ARGSUSED */ | |
2646 | int | |
2647 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2648 | { | |
2649 | znode_t *zp = ITOZ(ip); | |
0037b49e | 2650 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
b585bc4a BB |
2651 | uint32_t blksize; |
2652 | u_longlong_t nblocks; | |
057e8eee | 2653 | |
0037b49e | 2654 | ZFS_ENTER(zfsvfs); |
a7b125e9 GB |
2655 | ZFS_VERIFY_ZP(zp); |
2656 | ||
057e8eee BB |
2657 | mutex_enter(&zp->z_lock); |
2658 | ||
2659 | generic_fillattr(ip, sp); | |
97aa3ba4 TK |
2660 | /* |
2661 | * +1 link count for root inode with visible '.zfs' directory. | |
2662 | */ | |
2663 | if ((zp->z_id == zfsvfs->z_root) && zfs_show_ctldir(zp)) | |
2664 | if (sp->nlink < ZFS_LINK_MAX) | |
2665 | sp->nlink++; | |
057e8eee | 2666 | |
b585bc4a BB |
2667 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2668 | sp->blksize = blksize; | |
2669 | sp->blocks = nblocks; | |
2670 | ||
057e8eee BB |
2671 | if (unlikely(zp->z_blksz == 0)) { |
2672 | /* | |
2673 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2674 | */ | |
0037b49e | 2675 | sp->blksize = zfsvfs->z_max_blksz; |
057e8eee BB |
2676 | } |
2677 | ||
2678 | mutex_exit(&zp->z_lock); | |
2679 | ||
aa9b2708 AV |
2680 | /* |
2681 | * Required to prevent NFS client from detecting different inode | |
2682 | * numbers of snapshot root dentry before and after snapshot mount. | |
2683 | */ | |
0037b49e | 2684 | if (zfsvfs->z_issnap) { |
aa9b2708 AV |
2685 | if (ip->i_sb->s_root->d_inode == ip) |
2686 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
0037b49e | 2687 | dmu_objset_id(zfsvfs->z_os); |
aa9b2708 AV |
2688 | } |
2689 | ||
0037b49e | 2690 | ZFS_EXIT(zfsvfs); |
a7b125e9 | 2691 | |
057e8eee BB |
2692 | return (0); |
2693 | } | |
057e8eee | 2694 | |
9c5167d1 NF |
2695 | /* |
2696 | * For the operation of changing file's user/group/project, we need to | |
2697 | * handle not only the main object that is assigned to the file directly, | |
2698 | * but also the ones that are used by the file via hidden xattr directory. | |
2699 | * | |
2700 | * Because the xattr directory may contains many EA entries, as to it may | |
2701 | * be impossible to change all of them via the transaction of changing the | |
2702 | * main object's user/group/project attributes. Then we have to change them | |
2703 | * via other multiple independent transactions one by one. It may be not good | |
2704 | * solution, but we have no better idea yet. | |
2705 | */ | |
2706 | static int | |
2707 | zfs_setattr_dir(znode_t *dzp) | |
2708 | { | |
2709 | struct inode *dxip = ZTOI(dzp); | |
2710 | struct inode *xip = NULL; | |
2711 | zfsvfs_t *zfsvfs = ITOZSB(dxip); | |
2712 | objset_t *os = zfsvfs->z_os; | |
2713 | zap_cursor_t zc; | |
2714 | zap_attribute_t zap; | |
2715 | zfs_dirlock_t *dl; | |
2716 | znode_t *zp; | |
2717 | dmu_tx_t *tx = NULL; | |
2718 | uint64_t uid, gid; | |
2719 | sa_bulk_attr_t bulk[4]; | |
8cb34421 | 2720 | int count; |
9c5167d1 NF |
2721 | int err; |
2722 | ||
2723 | zap_cursor_init(&zc, os, dzp->z_id); | |
2724 | while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) { | |
8cb34421 | 2725 | count = 0; |
9c5167d1 NF |
2726 | if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { |
2727 | err = ENXIO; | |
2728 | break; | |
2729 | } | |
2730 | ||
2731 | err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp, | |
2732 | ZEXISTS, NULL, NULL); | |
2733 | if (err == ENOENT) | |
2734 | goto next; | |
2735 | if (err) | |
2736 | break; | |
2737 | ||
2738 | xip = ZTOI(zp); | |
2739 | if (KUID_TO_SUID(xip->i_uid) == KUID_TO_SUID(dxip->i_uid) && | |
2740 | KGID_TO_SGID(xip->i_gid) == KGID_TO_SGID(dxip->i_gid) && | |
2741 | zp->z_projid == dzp->z_projid) | |
2742 | goto next; | |
2743 | ||
2744 | tx = dmu_tx_create(os); | |
2745 | if (!(zp->z_pflags & ZFS_PROJID)) | |
2746 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
2747 | else | |
2748 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
2749 | ||
2750 | err = dmu_tx_assign(tx, TXG_WAIT); | |
2751 | if (err) | |
2752 | break; | |
2753 | ||
2754 | mutex_enter(&dzp->z_lock); | |
2755 | ||
2756 | if (KUID_TO_SUID(xip->i_uid) != KUID_TO_SUID(dxip->i_uid)) { | |
2757 | xip->i_uid = dxip->i_uid; | |
2758 | uid = zfs_uid_read(dxip); | |
2759 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, | |
2760 | &uid, sizeof (uid)); | |
2761 | } | |
2762 | ||
2763 | if (KGID_TO_SGID(xip->i_gid) != KGID_TO_SGID(dxip->i_gid)) { | |
2764 | xip->i_gid = dxip->i_gid; | |
2765 | gid = zfs_gid_read(dxip); | |
2766 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, | |
2767 | &gid, sizeof (gid)); | |
2768 | } | |
2769 | ||
2770 | if (zp->z_projid != dzp->z_projid) { | |
2771 | if (!(zp->z_pflags & ZFS_PROJID)) { | |
2772 | zp->z_pflags |= ZFS_PROJID; | |
2773 | SA_ADD_BULK_ATTR(bulk, count, | |
2774 | SA_ZPL_FLAGS(zfsvfs), NULL, &zp->z_pflags, | |
2775 | sizeof (zp->z_pflags)); | |
2776 | } | |
2777 | ||
2778 | zp->z_projid = dzp->z_projid; | |
2779 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PROJID(zfsvfs), | |
2780 | NULL, &zp->z_projid, sizeof (zp->z_projid)); | |
2781 | } | |
2782 | ||
2783 | mutex_exit(&dzp->z_lock); | |
2784 | ||
2785 | if (likely(count > 0)) { | |
2786 | err = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
2787 | dmu_tx_commit(tx); | |
2788 | } else { | |
2789 | dmu_tx_abort(tx); | |
2790 | } | |
2791 | tx = NULL; | |
2792 | if (err != 0 && err != ENOENT) | |
2793 | break; | |
2794 | ||
2795 | next: | |
2796 | if (xip) { | |
2797 | iput(xip); | |
2798 | xip = NULL; | |
2799 | zfs_dirent_unlock(dl); | |
2800 | } | |
2801 | zap_cursor_advance(&zc); | |
2802 | } | |
2803 | ||
2804 | if (tx) | |
2805 | dmu_tx_abort(tx); | |
2806 | if (xip) { | |
2807 | iput(xip); | |
2808 | zfs_dirent_unlock(dl); | |
2809 | } | |
2810 | zap_cursor_fini(&zc); | |
2811 | ||
2812 | return (err == ENOENT ? 0 : err); | |
2813 | } | |
2814 | ||
34dc7c2f BB |
2815 | /* |
2816 | * Set the file attributes to the values contained in the | |
2817 | * vattr structure. | |
2818 | * | |
3558fd73 | 2819 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2820 | * vap - new attribute values. |
5484965a | 2821 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2822 | * flags - ATTR_UTIME set if non-default time values provided. |
2823 | * - ATTR_NOACLCHECK (CIFS context only). | |
2824 | * cr - credentials of caller. | |
34dc7c2f BB |
2825 | * |
2826 | * RETURN: 0 if success | |
2827 | * error code if failure | |
2828 | * | |
2829 | * Timestamps: | |
3558fd73 | 2830 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2831 | */ |
2832 | /* ARGSUSED */ | |
e5c39b95 | 2833 | int |
5484965a | 2834 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2835 | { |
3558fd73 | 2836 | znode_t *zp = ITOZ(ip); |
0037b49e | 2837 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
9c5167d1 | 2838 | objset_t *os = zfsvfs->z_os; |
34dc7c2f BB |
2839 | zilog_t *zilog; |
2840 | dmu_tx_t *tx; | |
2841 | vattr_t oldva; | |
f4ea75d4 | 2842 | xvattr_t *tmpxvattr; |
5484965a | 2843 | uint_t mask = vap->va_mask; |
a117a6d6 | 2844 | uint_t saved_mask = 0; |
34dc7c2f BB |
2845 | int trim_mask = 0; |
2846 | uint64_t new_mode; | |
64aefee1 | 2847 | uint64_t new_kuid = 0, new_kgid = 0, new_uid, new_gid; |
572e2857 | 2848 | uint64_t xattr_obj; |
0df9673f | 2849 | uint64_t mtime[2], ctime[2], atime[2]; |
9c5167d1 | 2850 | uint64_t projid = ZFS_INVALID_PROJID; |
34dc7c2f BB |
2851 | znode_t *attrzp; |
2852 | int need_policy = FALSE; | |
9c5167d1 | 2853 | int err, err2 = 0; |
34dc7c2f | 2854 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2855 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2856 | xoptattr_t *xoap; | |
2857 | zfs_acl_t *aclp; | |
34dc7c2f | 2858 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2859 | boolean_t fuid_dirtied = B_FALSE; |
9c5167d1 | 2860 | boolean_t handle_eadir = B_FALSE; |
17c37660 | 2861 | sa_bulk_attr_t *bulk, *xattr_bulk; |
9c5167d1 | 2862 | int count = 0, xattr_count = 0, bulks = 8; |
34dc7c2f BB |
2863 | |
2864 | if (mask == 0) | |
2865 | return (0); | |
2866 | ||
0037b49e | 2867 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2868 | ZFS_VERIFY_ZP(zp); |
2869 | ||
9c5167d1 NF |
2870 | /* |
2871 | * If this is a xvattr_t, then get a pointer to the structure of | |
2872 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2873 | */ | |
2874 | xoap = xva_getxoptattr(xvap); | |
2875 | if (xoap != NULL && (mask & ATTR_XVATTR)) { | |
2876 | if (XVA_ISSET_REQ(xvap, XAT_PROJID)) { | |
2877 | if (!dmu_objset_projectquota_enabled(os) || | |
2878 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode))) { | |
2879 | ZFS_EXIT(zfsvfs); | |
2880 | return (SET_ERROR(ENOTSUP)); | |
2881 | } | |
2882 | ||
2883 | projid = xoap->xoa_projid; | |
2884 | if (unlikely(projid == ZFS_INVALID_PROJID)) { | |
2885 | ZFS_EXIT(zfsvfs); | |
2886 | return (SET_ERROR(EINVAL)); | |
2887 | } | |
2888 | ||
2889 | if (projid == zp->z_projid && zp->z_pflags & ZFS_PROJID) | |
2890 | projid = ZFS_INVALID_PROJID; | |
2891 | else | |
2892 | need_policy = TRUE; | |
2893 | } | |
2894 | ||
2895 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT) && | |
2705ebf0 NF |
2896 | (xoap->xoa_projinherit != |
2897 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) && | |
9c5167d1 NF |
2898 | (!dmu_objset_projectquota_enabled(os) || |
2899 | (!S_ISREG(ip->i_mode) && !S_ISDIR(ip->i_mode)))) { | |
2705ebf0 NF |
2900 | ZFS_EXIT(zfsvfs); |
2901 | return (SET_ERROR(ENOTSUP)); | |
9c5167d1 NF |
2902 | } |
2903 | } | |
2904 | ||
0037b49e | 2905 | zilog = zfsvfs->z_log; |
34dc7c2f BB |
2906 | |
2907 | /* | |
2908 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2909 | * that file system is at proper version level | |
2910 | */ | |
5484965a | 2911 | |
0037b49e | 2912 | if (zfsvfs->z_use_fuids == B_FALSE && |
5484965a BB |
2913 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2914 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2915 | (mask & ATTR_XVATTR))) { | |
0037b49e | 2916 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2917 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2918 | } |
2919 | ||
3558fd73 | 2920 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
0037b49e | 2921 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2922 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2923 | } |
2924 | ||
3558fd73 | 2925 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
0037b49e | 2926 | ZFS_EXIT(zfsvfs); |
2e528b49 | 2927 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2928 | } |
2929 | ||
d1d7e268 | 2930 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2931 | xva_init(tmpxvattr); |
5484965a | 2932 | |
9c5167d1 NF |
2933 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); |
2934 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * bulks, KM_SLEEP); | |
17c37660 | 2935 | |
5484965a BB |
2936 | /* |
2937 | * Immutable files can only alter immutable bit and atime | |
2938 | */ | |
2939 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2940 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2941 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
ecb2b7dc | 2942 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2943 | goto out3; |
5484965a BB |
2944 | } |
2945 | ||
3558fd73 | 2946 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
ecb2b7dc | 2947 | err = SET_ERROR(EPERM); |
f4ea75d4 | 2948 | goto out3; |
34dc7c2f BB |
2949 | } |
2950 | ||
5484965a BB |
2951 | /* |
2952 | * Verify timestamps doesn't overflow 32 bits. | |
2953 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2954 | * handle times greater than 2039. This check should be removed | |
2955 | * once large timestamps are fully supported. | |
2956 | */ | |
2957 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
d1d7e268 MK |
2958 | if (((mask & ATTR_ATIME) && |
2959 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2960 | ((mask & ATTR_MTIME) && | |
2961 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
ecb2b7dc | 2962 | err = SET_ERROR(EOVERFLOW); |
f4ea75d4 | 2963 | goto out3; |
5484965a BB |
2964 | } |
2965 | } | |
2966 | ||
34dc7c2f BB |
2967 | top: |
2968 | attrzp = NULL; | |
572e2857 | 2969 | aclp = NULL; |
34dc7c2f | 2970 | |
45d1cae3 | 2971 | /* Can this be moved to before the top label? */ |
0037b49e | 2972 | if (zfs_is_readonly(zfsvfs)) { |
ecb2b7dc | 2973 | err = SET_ERROR(EROFS); |
f4ea75d4 | 2974 | goto out3; |
34dc7c2f BB |
2975 | } |
2976 | ||
2977 | /* | |
2978 | * First validate permissions | |
2979 | */ | |
2980 | ||
3558fd73 | 2981 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2982 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2983 | if (err) |
2984 | goto out3; | |
2985 | ||
34dc7c2f BB |
2986 | /* |
2987 | * XXX - Note, we are not providing any open | |
2988 | * mode flags here (like FNDELAY), so we may | |
2989 | * block if there are locks present... this | |
2990 | * should be addressed in openat(). | |
2991 | */ | |
b128c09f | 2992 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2993 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2994 | if (err) |
2995 | goto out3; | |
428870ff | 2996 | } |
34dc7c2f | 2997 | |
5484965a BB |
2998 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2999 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
3000 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
3001 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
3002 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
3003 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
3004 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
3005 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
3006 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
3007 | skipaclchk, cr); | |
3008 | } | |
3009 | ||
3558fd73 BB |
3010 | if (mask & (ATTR_UID|ATTR_GID)) { |
3011 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
3012 | int take_owner; |
3013 | int take_group; | |
3014 | ||
3015 | /* | |
3016 | * NOTE: even if a new mode is being set, | |
3017 | * we may clear S_ISUID/S_ISGID bits. | |
3018 | */ | |
3019 | ||
3558fd73 | 3020 | if (!(mask & ATTR_MODE)) |
5484965a | 3021 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
3022 | |
3023 | /* | |
3024 | * Take ownership or chgrp to group we are a member of | |
3025 | */ | |
3026 | ||
5484965a | 3027 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 3028 | take_group = (mask & ATTR_GID) && |
0037b49e | 3029 | zfs_groupmember(zfsvfs, vap->va_gid, cr); |
34dc7c2f BB |
3030 | |
3031 | /* | |
5484965a | 3032 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
3033 | * take_group must both be set in order to allow taking |
3034 | * ownership. | |
3035 | * | |
3036 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
3037 | * | |
3038 | */ | |
3039 | ||
3558fd73 BB |
3040 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
3041 | take_owner && take_group) || | |
3042 | ((idmask == ATTR_UID) && take_owner) || | |
3043 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
3044 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
3045 | skipaclchk, cr) == 0) { | |
3046 | /* | |
3047 | * Remove setuid/setgid for non-privileged users | |
3048 | */ | |
5484965a | 3049 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 3050 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
3051 | } else { |
3052 | need_policy = TRUE; | |
3053 | } | |
3054 | } else { | |
3055 | need_policy = TRUE; | |
3056 | } | |
3057 | } | |
3058 | ||
3059 | mutex_enter(&zp->z_lock); | |
428870ff | 3060 | oldva.va_mode = zp->z_mode; |
572e2857 | 3061 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
3062 | if (mask & ATTR_XVATTR) { |
3063 | /* | |
3064 | * Update xvattr mask to include only those attributes | |
3065 | * that are actually changing. | |
3066 | * | |
3067 | * the bits will be restored prior to actually setting | |
3068 | * the attributes so the caller thinks they were set. | |
3069 | */ | |
3070 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
3071 | if (xoap->xoa_appendonly != | |
3072 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
3073 | need_policy = TRUE; | |
3074 | } else { | |
3075 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 3076 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
3077 | } |
3078 | } | |
3079 | ||
9c5167d1 NF |
3080 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { |
3081 | if (xoap->xoa_projinherit != | |
3082 | ((zp->z_pflags & ZFS_PROJINHERIT) != 0)) { | |
3083 | need_policy = TRUE; | |
3084 | } else { | |
3085 | XVA_CLR_REQ(xvap, XAT_PROJINHERIT); | |
3086 | XVA_SET_REQ(tmpxvattr, XAT_PROJINHERIT); | |
3087 | } | |
3088 | } | |
3089 | ||
5484965a BB |
3090 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { |
3091 | if (xoap->xoa_nounlink != | |
3092 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
3093 | need_policy = TRUE; | |
3094 | } else { | |
3095 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 3096 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
3097 | } |
3098 | } | |
3099 | ||
3100 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
3101 | if (xoap->xoa_immutable != | |
3102 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
3103 | need_policy = TRUE; | |
3104 | } else { | |
3105 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 3106 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
3107 | } |
3108 | } | |
3109 | ||
3110 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
3111 | if (xoap->xoa_nodump != | |
3112 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
3113 | need_policy = TRUE; | |
3114 | } else { | |
3115 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 3116 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
3117 | } |
3118 | } | |
3119 | ||
3120 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
3121 | if (xoap->xoa_av_modified != | |
3122 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
3123 | need_policy = TRUE; | |
3124 | } else { | |
3125 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 3126 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
3127 | } |
3128 | } | |
3129 | ||
3130 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
3131 | if ((!S_ISREG(ip->i_mode) && | |
3132 | xoap->xoa_av_quarantined) || | |
3133 | xoap->xoa_av_quarantined != | |
3134 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
3135 | need_policy = TRUE; | |
3136 | } else { | |
3137 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 3138 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
3139 | } |
3140 | } | |
3141 | ||
3142 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
3143 | mutex_exit(&zp->z_lock); | |
ecb2b7dc | 3144 | err = SET_ERROR(EPERM); |
f4ea75d4 | 3145 | goto out3; |
5484965a BB |
3146 | } |
3147 | ||
3148 | if (need_policy == FALSE && | |
3149 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
3150 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
3151 | need_policy = TRUE; | |
3152 | } | |
3153 | } | |
34dc7c2f BB |
3154 | |
3155 | mutex_exit(&zp->z_lock); | |
3156 | ||
3558fd73 | 3157 | if (mask & ATTR_MODE) { |
34dc7c2f | 3158 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 3159 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 3160 | &oldva, cr); |
f4ea75d4 BB |
3161 | if (err) |
3162 | goto out3; | |
3163 | ||
3558fd73 | 3164 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
3165 | } else { |
3166 | need_policy = TRUE; | |
3167 | } | |
3168 | } | |
3169 | ||
3170 | if (need_policy) { | |
3171 | /* | |
3172 | * If trim_mask is set then take ownership | |
3173 | * has been granted or write_acl is present and user | |
3174 | * has the ability to modify mode. In that case remove | |
3175 | * UID|GID and or MODE from mask so that | |
3176 | * secpolicy_vnode_setattr() doesn't revoke it. | |
3177 | */ | |
3178 | ||
3179 | if (trim_mask) { | |
5484965a BB |
3180 | saved_mask = vap->va_mask; |
3181 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 3182 | } |
5484965a | 3183 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 3184 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
3185 | if (err) |
3186 | goto out3; | |
34dc7c2f BB |
3187 | |
3188 | if (trim_mask) | |
5484965a | 3189 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
3190 | } |
3191 | ||
3192 | /* | |
3193 | * secpolicy_vnode_setattr, or take ownership may have | |
3194 | * changed va_mask | |
3195 | */ | |
5484965a | 3196 | mask = vap->va_mask; |
34dc7c2f | 3197 | |
9c5167d1 NF |
3198 | if ((mask & (ATTR_UID | ATTR_GID)) || projid != ZFS_INVALID_PROJID) { |
3199 | handle_eadir = B_TRUE; | |
0037b49e | 3200 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
572e2857 | 3201 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 3202 | |
572e2857 | 3203 | if (err == 0 && xattr_obj) { |
3558fd73 | 3204 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
3205 | if (err) |
3206 | goto out2; | |
3207 | } | |
3558fd73 | 3208 | if (mask & ATTR_UID) { |
0037b49e | 3209 | new_kuid = zfs_fuid_create(zfsvfs, |
5484965a | 3210 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
64aefee1 | 3211 | if (new_kuid != KUID_TO_SUID(ZTOI(zp)->i_uid) && |
9c5167d1 NF |
3212 | zfs_id_overquota(zfsvfs, DMU_USERUSED_OBJECT, |
3213 | new_kuid)) { | |
572e2857 | 3214 | if (attrzp) |
3558fd73 | 3215 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3216 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3217 | goto out2; |
3218 | } | |
3219 | } | |
3220 | ||
3558fd73 | 3221 | if (mask & ATTR_GID) { |
0037b49e BB |
3222 | new_kgid = zfs_fuid_create(zfsvfs, |
3223 | (uint64_t)vap->va_gid, cr, ZFS_GROUP, &fuidp); | |
64aefee1 | 3224 | if (new_kgid != KGID_TO_SGID(ZTOI(zp)->i_gid) && |
9c5167d1 NF |
3225 | zfs_id_overquota(zfsvfs, DMU_GROUPUSED_OBJECT, |
3226 | new_kgid)) { | |
572e2857 | 3227 | if (attrzp) |
3558fd73 | 3228 | iput(ZTOI(attrzp)); |
ecb2b7dc | 3229 | err = SET_ERROR(EDQUOT); |
428870ff BB |
3230 | goto out2; |
3231 | } | |
3232 | } | |
9c5167d1 NF |
3233 | |
3234 | if (projid != ZFS_INVALID_PROJID && | |
3235 | zfs_id_overquota(zfsvfs, DMU_PROJECTUSED_OBJECT, projid)) { | |
3236 | if (attrzp) | |
3237 | iput(ZTOI(attrzp)); | |
3238 | err = EDQUOT; | |
3239 | goto out2; | |
3240 | } | |
428870ff | 3241 | } |
9c5167d1 | 3242 | tx = dmu_tx_create(os); |
34dc7c2f | 3243 | |
3558fd73 | 3244 | if (mask & ATTR_MODE) { |
428870ff | 3245 | uint64_t pmode = zp->z_mode; |
572e2857 | 3246 | uint64_t acl_obj; |
5484965a | 3247 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 3248 | |
572e2857 | 3249 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 3250 | |
572e2857 BB |
3251 | mutex_enter(&zp->z_lock); |
3252 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
3253 | /* |
3254 | * Are we upgrading ACL from old V0 format | |
3255 | * to V1 format? | |
3256 | */ | |
0037b49e | 3257 | if (zfsvfs->z_version >= ZPL_VERSION_FUID && |
572e2857 | 3258 | zfs_znode_acl_version(zp) == |
34dc7c2f | 3259 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 3260 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
3261 | DMU_OBJECT_END); |
3262 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
3263 | 0, aclp->z_acl_bytes); | |
3264 | } else { | |
572e2857 | 3265 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
3266 | aclp->z_acl_bytes); |
3267 | } | |
428870ff | 3268 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
3269 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
3270 | 0, aclp->z_acl_bytes); | |
3271 | } | |
572e2857 | 3272 | mutex_exit(&zp->z_lock); |
428870ff BB |
3273 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3274 | } else { | |
9c5167d1 NF |
3275 | if (((mask & ATTR_XVATTR) && |
3276 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) || | |
3277 | (projid != ZFS_INVALID_PROJID && | |
3278 | !(zp->z_pflags & ZFS_PROJID))) | |
5484965a BB |
3279 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3280 | else | |
3281 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3282 | } |
3283 | ||
428870ff BB |
3284 | if (attrzp) { |
3285 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3286 | } |
3287 | ||
0037b49e | 3288 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
428870ff | 3289 | if (fuid_dirtied) |
0037b49e | 3290 | zfs_fuid_txhold(zfsvfs, tx); |
428870ff BB |
3291 | |
3292 | zfs_sa_upgrade_txholds(tx, zp); | |
3293 | ||
384f8a09 MA |
3294 | err = dmu_tx_assign(tx, TXG_WAIT); |
3295 | if (err) | |
9babb374 | 3296 | goto out; |
34dc7c2f | 3297 | |
428870ff | 3298 | count = 0; |
34dc7c2f BB |
3299 | /* |
3300 | * Set each attribute requested. | |
3301 | * We group settings according to the locks they need to acquire. | |
3302 | * | |
3303 | * Note: you cannot set ctime directly, although it will be | |
3304 | * updated as a side-effect of calling this function. | |
3305 | */ | |
3306 | ||
9c5167d1 NF |
3307 | if (projid != ZFS_INVALID_PROJID && !(zp->z_pflags & ZFS_PROJID)) { |
3308 | /* | |
3309 | * For the existed object that is upgraded from old system, | |
3310 | * its on-disk layout has no slot for the project ID attribute. | |
3311 | * But quota accounting logic needs to access related slots by | |
3312 | * offset directly. So we need to adjust old objects' layout | |
3313 | * to make the project ID to some unified and fixed offset. | |
3314 | */ | |
3315 | if (attrzp) | |
3316 | err = sa_add_projid(attrzp->z_sa_hdl, tx, projid); | |
3317 | if (err == 0) | |
3318 | err = sa_add_projid(zp->z_sa_hdl, tx, projid); | |
3319 | ||
3320 | if (unlikely(err == EEXIST)) | |
3321 | err = 0; | |
3322 | else if (err != 0) | |
3323 | goto out; | |
3324 | else | |
3325 | projid = ZFS_INVALID_PROJID; | |
3326 | } | |
572e2857 | 3327 | |
3558fd73 | 3328 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3329 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
3330 | mutex_enter(&zp->z_lock); |
3331 | ||
0037b49e | 3332 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff BB |
3333 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3334 | ||
3335 | if (attrzp) { | |
3558fd73 | 3336 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3337 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
3338 | mutex_enter(&attrzp->z_lock); |
3339 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3340 | SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, |
428870ff | 3341 | sizeof (attrzp->z_pflags)); |
9c5167d1 NF |
3342 | if (projid != ZFS_INVALID_PROJID) { |
3343 | attrzp->z_projid = projid; | |
3344 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3345 | SA_ZPL_PROJID(zfsvfs), NULL, &attrzp->z_projid, | |
3346 | sizeof (attrzp->z_projid)); | |
3347 | } | |
428870ff BB |
3348 | } |
3349 | ||
3558fd73 | 3350 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 3351 | |
3558fd73 | 3352 | if (mask & ATTR_UID) { |
64aefee1 NB |
3353 | ZTOI(zp)->i_uid = SUID_TO_KUID(new_kuid); |
3354 | new_uid = zfs_uid_read(ZTOI(zp)); | |
0037b49e | 3355 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 3356 | &new_uid, sizeof (new_uid)); |
428870ff BB |
3357 | if (attrzp) { |
3358 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3359 | SA_ZPL_UID(zfsvfs), NULL, &new_uid, |
428870ff | 3360 | sizeof (new_uid)); |
2c6abf15 | 3361 | ZTOI(attrzp)->i_uid = SUID_TO_KUID(new_uid); |
428870ff BB |
3362 | } |
3363 | } | |
3364 | ||
3558fd73 | 3365 | if (mask & ATTR_GID) { |
64aefee1 NB |
3366 | ZTOI(zp)->i_gid = SGID_TO_KGID(new_kgid); |
3367 | new_gid = zfs_gid_read(ZTOI(zp)); | |
0037b49e | 3368 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), |
428870ff | 3369 | NULL, &new_gid, sizeof (new_gid)); |
428870ff BB |
3370 | if (attrzp) { |
3371 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3372 | SA_ZPL_GID(zfsvfs), NULL, &new_gid, |
428870ff | 3373 | sizeof (new_gid)); |
64aefee1 | 3374 | ZTOI(attrzp)->i_gid = SGID_TO_KGID(new_kgid); |
428870ff BB |
3375 | } |
3376 | } | |
3558fd73 | 3377 | if (!(mask & ATTR_MODE)) { |
0037b49e | 3378 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), |
428870ff BB |
3379 | NULL, &new_mode, sizeof (new_mode)); |
3380 | new_mode = zp->z_mode; | |
3381 | } | |
3382 | err = zfs_acl_chown_setattr(zp); | |
3383 | ASSERT(err == 0); | |
3384 | if (attrzp) { | |
3385 | err = zfs_acl_chown_setattr(attrzp); | |
3386 | ASSERT(err == 0); | |
3387 | } | |
3388 | } | |
3389 | ||
3558fd73 | 3390 | if (mask & ATTR_MODE) { |
0037b49e | 3391 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 3392 | &new_mode, sizeof (new_mode)); |
12fa7f34 | 3393 | zp->z_mode = ZTOI(zp)->i_mode = new_mode; |
99c564bc | 3394 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 3395 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c99c9001 | 3396 | ASSERT0(err); |
572e2857 BB |
3397 | if (zp->z_acl_cached) |
3398 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
3399 | zp->z_acl_cached = aclp; |
3400 | aclp = NULL; | |
34dc7c2f BB |
3401 | } |
3402 | ||
704cd075 CC |
3403 | if ((mask & ATTR_ATIME) || zp->z_atime_dirty) { |
3404 | zp->z_atime_dirty = 0; | |
3405 | ZFS_TIME_ENCODE(&ip->i_atime, atime); | |
0037b49e | 3406 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
0df9673f | 3407 | &atime, sizeof (atime)); |
34dc7c2f BB |
3408 | } |
3409 | ||
99834d19 | 3410 | if (mask & (ATTR_MTIME | ATTR_SIZE)) { |
5484965a | 3411 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
6413c95f | 3412 | ZTOI(zp)->i_mtime = zpl_inode_timespec_trunc(vap->va_mtime, |
87f9371a NB |
3413 | ZTOI(zp)->i_sb->s_time_gran); |
3414 | ||
0037b49e | 3415 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
428870ff | 3416 | mtime, sizeof (mtime)); |
34dc7c2f BB |
3417 | } |
3418 | ||
99834d19 | 3419 | if (mask & (ATTR_CTIME | ATTR_SIZE)) { |
87f9371a | 3420 | ZFS_TIME_ENCODE(&vap->va_ctime, ctime); |
6413c95f | 3421 | ZTOI(zp)->i_ctime = zpl_inode_timespec_trunc(vap->va_ctime, |
87f9371a | 3422 | ZTOI(zp)->i_sb->s_time_gran); |
0037b49e | 3423 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
87f9371a | 3424 | ctime, sizeof (ctime)); |
428870ff | 3425 | } |
87f9371a | 3426 | |
9c5167d1 NF |
3427 | if (projid != ZFS_INVALID_PROJID) { |
3428 | zp->z_projid = projid; | |
3429 | SA_ADD_BULK_ATTR(bulk, count, | |
3430 | SA_ZPL_PROJID(zfsvfs), NULL, &zp->z_projid, | |
3431 | sizeof (zp->z_projid)); | |
3432 | } | |
3433 | ||
87f9371a NB |
3434 | if (attrzp && mask) { |
3435 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
0037b49e | 3436 | SA_ZPL_CTIME(zfsvfs), NULL, &ctime, |
87f9371a NB |
3437 | sizeof (ctime)); |
3438 | } | |
3439 | ||
34dc7c2f BB |
3440 | /* |
3441 | * Do this after setting timestamps to prevent timestamp | |
3442 | * update from toggling bit | |
3443 | */ | |
3444 | ||
5484965a BB |
3445 | if (xoap && (mask & ATTR_XVATTR)) { |
3446 | ||
3447 | /* | |
3448 | * restore trimmed off masks | |
3449 | * so that return masks can be set for caller. | |
3450 | */ | |
3451 | ||
f4ea75d4 | 3452 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
3453 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
3454 | } | |
f4ea75d4 | 3455 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
3456 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
3457 | } | |
f4ea75d4 | 3458 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
3459 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
3460 | } | |
f4ea75d4 | 3461 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
3462 | XVA_SET_REQ(xvap, XAT_NODUMP); |
3463 | } | |
f4ea75d4 | 3464 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
3465 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
3466 | } | |
f4ea75d4 | 3467 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
3468 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
3469 | } | |
9c5167d1 NF |
3470 | if (XVA_ISSET_REQ(tmpxvattr, XAT_PROJINHERIT)) { |
3471 | XVA_SET_REQ(xvap, XAT_PROJINHERIT); | |
3472 | } | |
5484965a BB |
3473 | |
3474 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3475 | ASSERT(S_ISREG(ip->i_mode)); | |
3476 | ||
3477 | zfs_xvattr_set(zp, xvap, tx); | |
3478 | } | |
3479 | ||
9babb374 | 3480 | if (fuid_dirtied) |
0037b49e | 3481 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 | 3482 | |
34dc7c2f | 3483 | if (mask != 0) |
5484965a | 3484 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3485 | |
34dc7c2f | 3486 | mutex_exit(&zp->z_lock); |
3558fd73 | 3487 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3488 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3489 | |
572e2857 | 3490 | if (attrzp) { |
3558fd73 | 3491 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3492 | mutex_exit(&attrzp->z_acl_lock); |
3493 | mutex_exit(&attrzp->z_lock); | |
3494 | } | |
9babb374 | 3495 | out: |
9c5167d1 | 3496 | if (err == 0 && xattr_count > 0) { |
428870ff BB |
3497 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, |
3498 | xattr_count, tx); | |
3499 | ASSERT(err2 == 0); | |
3500 | } | |
3501 | ||
45d1cae3 | 3502 | if (aclp) |
9babb374 | 3503 | zfs_acl_free(aclp); |
9babb374 BB |
3504 | |
3505 | if (fuidp) { | |
3506 | zfs_fuid_info_free(fuidp); | |
3507 | fuidp = NULL; | |
3508 | } | |
3509 | ||
428870ff | 3510 | if (err) { |
9babb374 | 3511 | dmu_tx_abort(tx); |
ea7e86d8 BB |
3512 | if (attrzp) |
3513 | iput(ZTOI(attrzp)); | |
428870ff BB |
3514 | if (err == ERESTART) |
3515 | goto top; | |
3516 | } else { | |
9c5167d1 NF |
3517 | if (count > 0) |
3518 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3519 | dmu_tx_commit(tx); |
9c5167d1 NF |
3520 | if (attrzp) { |
3521 | if (err2 == 0 && handle_eadir) | |
3522 | err2 = zfs_setattr_dir(attrzp); | |
ea7e86d8 | 3523 | iput(ZTOI(attrzp)); |
9c5167d1 | 3524 | } |
037849f8 | 3525 | zfs_inode_update(zp); |
428870ff BB |
3526 | } |
3527 | ||
428870ff | 3528 | out2: |
9c5167d1 | 3529 | if (os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3530 | zil_commit(zilog, 0); |
34dc7c2f | 3531 | |
f4ea75d4 | 3532 | out3: |
9c5167d1 NF |
3533 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * bulks); |
3534 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * bulks); | |
d1d7e268 | 3535 | kmem_free(tmpxvattr, sizeof (xvattr_t)); |
0037b49e | 3536 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3537 | return (err); |
3538 | } | |
3539 | ||
3540 | typedef struct zfs_zlock { | |
3541 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3542 | znode_t *zl_znode; /* znode we held */ | |
3543 | struct zfs_zlock *zl_next; /* next in list */ | |
3544 | } zfs_zlock_t; | |
3545 | ||
3546 | /* | |
3547 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3548 | */ | |
3549 | static void | |
3550 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3551 | { | |
3552 | zfs_zlock_t *zl; | |
3553 | ||
3554 | while ((zl = *zlpp) != NULL) { | |
3555 | if (zl->zl_znode != NULL) | |
ea7e86d8 | 3556 | zfs_iput_async(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3557 | rw_exit(zl->zl_rwlock); |
3558 | *zlpp = zl->zl_next; | |
3559 | kmem_free(zl, sizeof (*zl)); | |
3560 | } | |
3561 | } | |
3562 | ||
3563 | /* | |
3564 | * Search back through the directory tree, using the ".." entries. | |
3565 | * Lock each directory in the chain to prevent concurrent renames. | |
3566 | * Fail any attempt to move a directory into one of its own descendants. | |
3567 | * XXX - z_parent_lock can overlap with map or grow locks | |
3568 | */ | |
3569 | static int | |
3570 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3571 | { | |
3572 | zfs_zlock_t *zl; | |
3573 | znode_t *zp = tdzp; | |
3558fd73 | 3574 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3575 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3576 | krwlock_t *rwlp = &szp->z_parent_lock; |
3577 | krw_t rw = RW_WRITER; | |
3578 | ||
3579 | /* | |
3580 | * First pass write-locks szp and compares to zp->z_id. | |
3581 | * Later passes read-lock zp and compare to zp->z_parent. | |
3582 | */ | |
3583 | do { | |
3584 | if (!rw_tryenter(rwlp, rw)) { | |
3585 | /* | |
3586 | * Another thread is renaming in this path. | |
3587 | * Note that if we are a WRITER, we don't have any | |
3588 | * parent_locks held yet. | |
3589 | */ | |
3590 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3591 | /* | |
3592 | * Drop our locks and restart | |
3593 | */ | |
3594 | zfs_rename_unlock(&zl); | |
3595 | *zlpp = NULL; | |
3596 | zp = tdzp; | |
428870ff | 3597 | oidp = zp->z_id; |
34dc7c2f BB |
3598 | rwlp = &szp->z_parent_lock; |
3599 | rw = RW_WRITER; | |
3600 | continue; | |
3601 | } else { | |
3602 | /* | |
3603 | * Wait for other thread to drop its locks | |
3604 | */ | |
3605 | rw_enter(rwlp, rw); | |
3606 | } | |
3607 | } | |
3608 | ||
3609 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3610 | zl->zl_rwlock = rwlp; | |
3611 | zl->zl_znode = NULL; | |
3612 | zl->zl_next = *zlpp; | |
3613 | *zlpp = zl; | |
3614 | ||
428870ff | 3615 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
2e528b49 | 3616 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3617 | |
428870ff | 3618 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3619 | return (0); |
3620 | ||
3621 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3622 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3623 | if (error) |
3624 | return (error); | |
3625 | zl->zl_znode = zp; | |
3626 | } | |
3558fd73 | 3627 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3628 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3629 | rwlp = &zp->z_parent_lock; |
3630 | rw = RW_READER; | |
3631 | ||
3632 | } while (zp->z_id != sdzp->z_id); | |
3633 | ||
3634 | return (0); | |
3635 | } | |
3636 | ||
3637 | /* | |
3638 | * Move an entry from the provided source directory to the target | |
3639 | * directory. Change the entry name as indicated. | |
3640 | * | |
3558fd73 | 3641 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3642 | * snm - Old entry name. |
3558fd73 | 3643 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3644 | * tnm - New entry name. |
3645 | * cr - credentials of caller. | |
34dc7c2f BB |
3646 | * flags - case flags |
3647 | * | |
d3cc8b15 | 3648 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3649 | * |
3650 | * Timestamps: | |
3558fd73 | 3651 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3652 | */ |
3653 | /*ARGSUSED*/ | |
e5c39b95 | 3654 | int |
3558fd73 BB |
3655 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3656 | cred_t *cr, int flags) | |
34dc7c2f BB |
3657 | { |
3658 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 | 3659 | znode_t *sdzp = ITOZ(sdip); |
0037b49e | 3660 | zfsvfs_t *zfsvfs = ITOZSB(sdip); |
34dc7c2f | 3661 | zilog_t *zilog; |
34dc7c2f BB |
3662 | zfs_dirlock_t *sdl, *tdl; |
3663 | dmu_tx_t *tx; | |
3664 | zfs_zlock_t *zl; | |
3665 | int cmp, serr, terr; | |
3666 | int error = 0; | |
3667 | int zflg = 0; | |
e8b96c60 | 3668 | boolean_t waited = B_FALSE; |
34dc7c2f | 3669 | |
32dec7bd | 3670 | if (snm == NULL || tnm == NULL) |
3671 | return (SET_ERROR(EINVAL)); | |
3672 | ||
0037b49e | 3673 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 3674 | ZFS_VERIFY_ZP(sdzp); |
0037b49e | 3675 | zilog = zfsvfs->z_log; |
34dc7c2f | 3676 | |
812e91a7 MT |
3677 | tdzp = ITOZ(tdip); |
3678 | ZFS_VERIFY_ZP(tdzp); | |
3679 | ||
3680 | /* | |
3681 | * We check i_sb because snapshots and the ctldir must have different | |
3682 | * super blocks. | |
3683 | */ | |
c0ebc844 | 3684 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
0037b49e | 3685 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3686 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3687 | } |
3688 | ||
0037b49e | 3689 | if (zfsvfs->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3690 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 3691 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3692 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3693 | } |
3694 | ||
3695 | if (flags & FIGNORECASE) | |
3696 | zflg |= ZCILOOK; | |
3697 | ||
3698 | top: | |
3699 | szp = NULL; | |
3700 | tzp = NULL; | |
3701 | zl = NULL; | |
3702 | ||
3703 | /* | |
3704 | * This is to prevent the creation of links into attribute space | |
3705 | * by renaming a linked file into/outof an attribute directory. | |
3706 | * See the comment in zfs_link() for why this is considered bad. | |
3707 | */ | |
428870ff | 3708 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 3709 | ZFS_EXIT(zfsvfs); |
2e528b49 | 3710 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3711 | } |
3712 | ||
3713 | /* | |
3714 | * Lock source and target directory entries. To prevent deadlock, | |
3715 | * a lock ordering must be defined. We lock the directory with | |
3716 | * the smallest object id first, or if it's a tie, the one with | |
3717 | * the lexically first name. | |
3718 | */ | |
3719 | if (sdzp->z_id < tdzp->z_id) { | |
3720 | cmp = -1; | |
3721 | } else if (sdzp->z_id > tdzp->z_id) { | |
3722 | cmp = 1; | |
3723 | } else { | |
3724 | /* | |
3725 | * First compare the two name arguments without | |
3726 | * considering any case folding. | |
3727 | */ | |
0037b49e | 3728 | int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3729 | |
3730 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
0037b49e | 3731 | ASSERT(error == 0 || !zfsvfs->z_utf8); |
34dc7c2f BB |
3732 | if (cmp == 0) { |
3733 | /* | |
3734 | * POSIX: "If the old argument and the new argument | |
3735 | * both refer to links to the same existing file, | |
3736 | * the rename() function shall return successfully | |
3737 | * and perform no other action." | |
3738 | */ | |
0037b49e | 3739 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3740 | return (0); |
3741 | } | |
3742 | /* | |
3743 | * If the file system is case-folding, then we may | |
3744 | * have some more checking to do. A case-folding file | |
3745 | * system is either supporting mixed case sensitivity | |
3746 | * access or is completely case-insensitive. Note | |
3747 | * that the file system is always case preserving. | |
3748 | * | |
3749 | * In mixed sensitivity mode case sensitive behavior | |
3750 | * is the default. FIGNORECASE must be used to | |
3751 | * explicitly request case insensitive behavior. | |
3752 | * | |
3753 | * If the source and target names provided differ only | |
3754 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3755 | * we will treat this as a special case in the | |
3756 | * case-insensitive mode: as long as the source name | |
3757 | * is an exact match, we will allow this to proceed as | |
3758 | * a name-change request. | |
3759 | */ | |
0037b49e BB |
3760 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
3761 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3762 | flags & FIGNORECASE)) && |
0037b49e | 3763 | u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3764 | &error) == 0) { |
3765 | /* | |
3766 | * case preserving rename request, require exact | |
3767 | * name matches | |
3768 | */ | |
3769 | zflg |= ZCIEXACT; | |
3770 | zflg &= ~ZCILOOK; | |
3771 | } | |
3772 | } | |
3773 | ||
428870ff BB |
3774 | /* |
3775 | * If the source and destination directories are the same, we should | |
3776 | * grab the z_name_lock of that directory only once. | |
3777 | */ | |
3778 | if (sdzp == tdzp) { | |
3779 | zflg |= ZHAVELOCK; | |
3780 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3781 | } | |
3782 | ||
34dc7c2f BB |
3783 | if (cmp < 0) { |
3784 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3785 | ZEXISTS | zflg, NULL, NULL); | |
3786 | terr = zfs_dirent_lock(&tdl, | |
3787 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3788 | } else { | |
3789 | terr = zfs_dirent_lock(&tdl, | |
3790 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3791 | serr = zfs_dirent_lock(&sdl, | |
3792 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3793 | NULL, NULL); | |
3794 | } | |
3795 | ||
3796 | if (serr) { | |
3797 | /* | |
3798 | * Source entry invalid or not there. | |
3799 | */ | |
3800 | if (!terr) { | |
3801 | zfs_dirent_unlock(tdl); | |
3802 | if (tzp) | |
3558fd73 | 3803 | iput(ZTOI(tzp)); |
34dc7c2f | 3804 | } |
428870ff BB |
3805 | |
3806 | if (sdzp == tdzp) | |
3807 | rw_exit(&sdzp->z_name_lock); | |
3808 | ||
34dc7c2f BB |
3809 | if (strcmp(snm, "..") == 0) |
3810 | serr = EINVAL; | |
0037b49e | 3811 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3812 | return (serr); |
3813 | } | |
3814 | if (terr) { | |
3815 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3816 | iput(ZTOI(szp)); |
428870ff BB |
3817 | |
3818 | if (sdzp == tdzp) | |
3819 | rw_exit(&sdzp->z_name_lock); | |
3820 | ||
34dc7c2f BB |
3821 | if (strcmp(tnm, "..") == 0) |
3822 | terr = EINVAL; | |
0037b49e | 3823 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3824 | return (terr); |
3825 | } | |
3826 | ||
9c5167d1 NF |
3827 | /* |
3828 | * If we are using project inheritance, means if the directory has | |
3829 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
3830 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
3831 | * such case, we only allow renames into our tree when the project | |
3832 | * IDs are the same. | |
3833 | */ | |
3834 | if (tdzp->z_pflags & ZFS_PROJINHERIT && | |
3835 | tdzp->z_projid != szp->z_projid) { | |
3836 | error = SET_ERROR(EXDEV); | |
3837 | goto out; | |
3838 | } | |
3839 | ||
34dc7c2f BB |
3840 | /* |
3841 | * Must have write access at the source to remove the old entry | |
3842 | * and write access at the target to create the new entry. | |
3843 | * Note that if target and source are the same, this can be | |
3844 | * done in a single check. | |
3845 | */ | |
3846 | ||
149e873a | 3847 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3848 | goto out; |
3849 | ||
3558fd73 | 3850 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3851 | /* |
3852 | * Check to make sure rename is valid. | |
3853 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3854 | */ | |
149e873a | 3855 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3856 | goto out; |
3857 | } | |
3858 | ||
3859 | /* | |
3860 | * Does target exist? | |
3861 | */ | |
3862 | if (tzp) { | |
3863 | /* | |
3864 | * Source and target must be the same type. | |
3865 | */ | |
3558fd73 BB |
3866 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3867 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
2e528b49 | 3868 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3869 | goto out; |
3870 | } | |
3871 | } else { | |
3558fd73 | 3872 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
2e528b49 | 3873 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3874 | goto out; |
3875 | } | |
3876 | } | |
3877 | /* | |
3878 | * POSIX dictates that when the source and target | |
3879 | * entries refer to the same file object, rename | |
3880 | * must do nothing and exit without error. | |
3881 | */ | |
3882 | if (szp->z_id == tzp->z_id) { | |
3883 | error = 0; | |
3884 | goto out; | |
3885 | } | |
3886 | } | |
3887 | ||
0037b49e | 3888 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
3889 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3890 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3891 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3892 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3893 | if (sdzp != tdzp) { |
3894 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3895 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3896 | } | |
3897 | if (tzp) { | |
3898 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3899 | zfs_sa_upgrade_txholds(tx, tzp); | |
3900 | } | |
3901 | ||
3902 | zfs_sa_upgrade_txholds(tx, szp); | |
0037b49e | 3903 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
0735ecb3 | 3904 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
3905 | if (error) { |
3906 | if (zl != NULL) | |
3907 | zfs_rename_unlock(&zl); | |
3908 | zfs_dirent_unlock(sdl); | |
3909 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3910 | |
3911 | if (sdzp == tdzp) | |
3912 | rw_exit(&sdzp->z_name_lock); | |
3913 | ||
fb5f0bc8 | 3914 | if (error == ERESTART) { |
e8b96c60 | 3915 | waited = B_TRUE; |
34dc7c2f BB |
3916 | dmu_tx_wait(tx); |
3917 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3918 | iput(ZTOI(szp)); |
3919 | if (tzp) | |
3920 | iput(ZTOI(tzp)); | |
34dc7c2f BB |
3921 | goto top; |
3922 | } | |
3923 | dmu_tx_abort(tx); | |
ea7e86d8 BB |
3924 | iput(ZTOI(szp)); |
3925 | if (tzp) | |
3926 | iput(ZTOI(tzp)); | |
0037b49e | 3927 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
3928 | return (error); |
3929 | } | |
3930 | ||
3931 | if (tzp) /* Attempt to remove the existing target */ | |
3932 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3933 | ||
3934 | if (error == 0) { | |
3935 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3936 | if (error == 0) { | |
428870ff | 3937 | szp->z_pflags |= ZFS_AV_MODIFIED; |
9c5167d1 NF |
3938 | if (tdzp->z_pflags & ZFS_PROJINHERIT) |
3939 | szp->z_pflags |= ZFS_PROJINHERIT; | |
34dc7c2f | 3940 | |
0037b49e | 3941 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 3942 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c99c9001 | 3943 | ASSERT0(error); |
34dc7c2f | 3944 | |
428870ff BB |
3945 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3946 | if (error == 0) { | |
3947 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3948 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3949 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3950 | } else { |
3951 | /* | |
3952 | * At this point, we have successfully created | |
3953 | * the target name, but have failed to remove | |
3954 | * the source name. Since the create was done | |
3955 | * with the ZRENAMING flag, there are | |
3956 | * complications; for one, the link count is | |
3957 | * wrong. The easiest way to deal with this | |
3958 | * is to remove the newly created target, and | |
3959 | * return the original error. This must | |
3960 | * succeed; fortunately, it is very unlikely to | |
3961 | * fail, since we just created it. | |
3962 | */ | |
3963 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3964 | ZRENAMING, NULL), ==, 0); | |
3965 | } | |
599b8648 CC |
3966 | } else { |
3967 | /* | |
3968 | * If we had removed the existing target, subsequent | |
3969 | * call to zfs_link_create() to add back the same entry | |
3970 | * but, the new dnode (szp) should not fail. | |
3971 | */ | |
3972 | ASSERT(tzp == NULL); | |
34dc7c2f BB |
3973 | } |
3974 | } | |
3975 | ||
3976 | dmu_tx_commit(tx); | |
3977 | out: | |
3978 | if (zl != NULL) | |
3979 | zfs_rename_unlock(&zl); | |
3980 | ||
3981 | zfs_dirent_unlock(sdl); | |
3982 | zfs_dirent_unlock(tdl); | |
3983 | ||
960e08fe | 3984 | zfs_inode_update(sdzp); |
428870ff BB |
3985 | if (sdzp == tdzp) |
3986 | rw_exit(&sdzp->z_name_lock); | |
3987 | ||
960e08fe BB |
3988 | if (sdzp != tdzp) |
3989 | zfs_inode_update(tdzp); | |
428870ff | 3990 | |
960e08fe | 3991 | zfs_inode_update(szp); |
3558fd73 | 3992 | iput(ZTOI(szp)); |
960e08fe BB |
3993 | if (tzp) { |
3994 | zfs_inode_update(tzp); | |
3558fd73 | 3995 | iput(ZTOI(tzp)); |
960e08fe | 3996 | } |
34dc7c2f | 3997 | |
0037b49e | 3998 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3999 | zil_commit(zilog, 0); |
428870ff | 4000 | |
0037b49e | 4001 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4002 | return (error); |
4003 | } | |
4004 | ||
4005 | /* | |
4006 | * Insert the indicated symbolic reference entry into the directory. | |
4007 | * | |
3558fd73 | 4008 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
4009 | * link - Name for new symlink entry. |
4010 | * vap - Attributes of new entry. | |
4011 | * target - Target path of new symlink. | |
3558fd73 | 4012 | * |
34dc7c2f | 4013 | * cr - credentials of caller. |
34dc7c2f BB |
4014 | * flags - case flags |
4015 | * | |
d3cc8b15 | 4016 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4017 | * |
4018 | * Timestamps: | |
3558fd73 | 4019 | * dip - ctime|mtime updated |
34dc7c2f BB |
4020 | */ |
4021 | /*ARGSUSED*/ | |
e5c39b95 | 4022 | int |
3558fd73 BB |
4023 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
4024 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 4025 | { |
3558fd73 | 4026 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
4027 | zfs_dirlock_t *dl; |
4028 | dmu_tx_t *tx; | |
0037b49e | 4029 | zfsvfs_t *zfsvfs = ITOZSB(dip); |
34dc7c2f | 4030 | zilog_t *zilog; |
428870ff | 4031 | uint64_t len = strlen(link); |
34dc7c2f BB |
4032 | int error; |
4033 | int zflg = ZNEW; | |
9babb374 BB |
4034 | zfs_acl_ids_t acl_ids; |
4035 | boolean_t fuid_dirtied; | |
428870ff | 4036 | uint64_t txtype = TX_SYMLINK; |
e8b96c60 | 4037 | boolean_t waited = B_FALSE; |
34dc7c2f | 4038 | |
3558fd73 | 4039 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 4040 | |
32dec7bd | 4041 | if (name == NULL) |
4042 | return (SET_ERROR(EINVAL)); | |
4043 | ||
0037b49e | 4044 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4045 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 4046 | zilog = zfsvfs->z_log; |
34dc7c2f | 4047 | |
0037b49e | 4048 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 4049 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 4050 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4051 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
4052 | } |
4053 | if (flags & FIGNORECASE) | |
4054 | zflg |= ZCILOOK; | |
34dc7c2f BB |
4055 | |
4056 | if (len > MAXPATHLEN) { | |
0037b49e | 4057 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4058 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
4059 | } |
4060 | ||
428870ff BB |
4061 | if ((error = zfs_acl_ids_create(dzp, 0, |
4062 | vap, cr, NULL, &acl_ids)) != 0) { | |
0037b49e | 4063 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4064 | return (error); |
4065 | } | |
4066 | top: | |
3558fd73 BB |
4067 | *ipp = NULL; |
4068 | ||
34dc7c2f BB |
4069 | /* |
4070 | * Attempt to lock directory; fail if entry already exists. | |
4071 | */ | |
4072 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
4073 | if (error) { | |
428870ff | 4074 | zfs_acl_ids_free(&acl_ids); |
0037b49e | 4075 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4076 | return (error); |
4077 | } | |
4078 | ||
149e873a | 4079 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
4080 | zfs_acl_ids_free(&acl_ids); |
4081 | zfs_dirent_unlock(dl); | |
0037b49e | 4082 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4083 | return (error); |
4084 | } | |
4085 | ||
9c5167d1 | 4086 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids, ZFS_DEFAULT_PROJID)) { |
9babb374 BB |
4087 | zfs_acl_ids_free(&acl_ids); |
4088 | zfs_dirent_unlock(dl); | |
0037b49e | 4089 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4090 | return (SET_ERROR(EDQUOT)); |
9babb374 | 4091 | } |
0037b49e BB |
4092 | tx = dmu_tx_create(zfsvfs->z_os); |
4093 | fuid_dirtied = zfsvfs->z_fuid_dirty; | |
34dc7c2f | 4094 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 4095 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
4096 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
4097 | ZFS_SA_BASE_ATTR_SIZE + len); | |
4098 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
0037b49e | 4099 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
4100 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
4101 | acl_ids.z_aclp->z_acl_bytes); | |
4102 | } | |
9babb374 | 4103 | if (fuid_dirtied) |
0037b49e | 4104 | zfs_fuid_txhold(zfsvfs, tx); |
0735ecb3 | 4105 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
4106 | if (error) { |
4107 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4108 | if (error == ERESTART) { |
e8b96c60 | 4109 | waited = B_TRUE; |
34dc7c2f BB |
4110 | dmu_tx_wait(tx); |
4111 | dmu_tx_abort(tx); | |
4112 | goto top; | |
4113 | } | |
428870ff | 4114 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 4115 | dmu_tx_abort(tx); |
0037b49e | 4116 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4117 | return (error); |
4118 | } | |
4119 | ||
34dc7c2f BB |
4120 | /* |
4121 | * Create a new object for the symlink. | |
428870ff | 4122 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 4123 | */ |
428870ff | 4124 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 4125 | |
428870ff | 4126 | if (fuid_dirtied) |
0037b49e | 4127 | zfs_fuid_sync(zfsvfs, tx); |
34dc7c2f | 4128 | |
572e2857 | 4129 | mutex_enter(&zp->z_lock); |
428870ff | 4130 | if (zp->z_is_sa) |
0037b49e | 4131 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), |
428870ff BB |
4132 | link, len, tx); |
4133 | else | |
4134 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 4135 | mutex_exit(&zp->z_lock); |
34dc7c2f | 4136 | |
428870ff | 4137 | zp->z_size = len; |
0037b49e | 4138 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
428870ff | 4139 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
4140 | /* |
4141 | * Insert the new object into the directory. | |
4142 | */ | |
599b8648 CC |
4143 | error = zfs_link_create(dl, zp, tx, ZNEW); |
4144 | if (error != 0) { | |
4145 | zfs_znode_delete(zp, tx); | |
4146 | remove_inode_hash(ZTOI(zp)); | |
4147 | } else { | |
4148 | if (flags & FIGNORECASE) | |
4149 | txtype |= TX_CI; | |
4150 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
4f301661 | 4151 | |
599b8648 CC |
4152 | zfs_inode_update(dzp); |
4153 | zfs_inode_update(zp); | |
4154 | } | |
960e08fe | 4155 | |
9babb374 | 4156 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
4157 | |
4158 | dmu_tx_commit(tx); | |
4159 | ||
4160 | zfs_dirent_unlock(dl); | |
4161 | ||
599b8648 CC |
4162 | if (error == 0) { |
4163 | *ipp = ZTOI(zp); | |
34dc7c2f | 4164 | |
599b8648 CC |
4165 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
4166 | zil_commit(zilog, 0); | |
4167 | } else { | |
4168 | iput(ZTOI(zp)); | |
4169 | } | |
428870ff | 4170 | |
0037b49e | 4171 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4172 | return (error); |
4173 | } | |
4174 | ||
4175 | /* | |
4176 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 4177 | * the symbolic path referred to by ip. |
34dc7c2f | 4178 | * |
8b4f9a2d BB |
4179 | * IN: ip - inode of symbolic link |
4180 | * uio - structure to contain the link path. | |
4181 | * cr - credentials of caller. | |
34dc7c2f BB |
4182 | * |
4183 | * RETURN: 0 if success | |
4184 | * error code if failure | |
4185 | * | |
4186 | * Timestamps: | |
3558fd73 | 4187 | * ip - atime updated |
34dc7c2f BB |
4188 | */ |
4189 | /* ARGSUSED */ | |
e5c39b95 | 4190 | int |
8b4f9a2d | 4191 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 4192 | { |
3558fd73 | 4193 | znode_t *zp = ITOZ(ip); |
0037b49e | 4194 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4195 | int error; |
4196 | ||
0037b49e | 4197 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4198 | ZFS_VERIFY_ZP(zp); |
4199 | ||
572e2857 | 4200 | mutex_enter(&zp->z_lock); |
428870ff | 4201 | if (zp->z_is_sa) |
8b4f9a2d | 4202 | error = sa_lookup_uio(zp->z_sa_hdl, |
0037b49e | 4203 | SA_ZPL_SYMLINK(zfsvfs), uio); |
428870ff | 4204 | else |
8b4f9a2d | 4205 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 4206 | mutex_exit(&zp->z_lock); |
34dc7c2f | 4207 | |
0037b49e | 4208 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4209 | return (error); |
4210 | } | |
4211 | ||
4212 | /* | |
3558fd73 | 4213 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 4214 | * |
3558fd73 BB |
4215 | * IN: tdip - Directory to contain new entry. |
4216 | * sip - inode of new entry. | |
34dc7c2f BB |
4217 | * name - name of new entry. |
4218 | * cr - credentials of caller. | |
34dc7c2f BB |
4219 | * |
4220 | * RETURN: 0 if success | |
4221 | * error code if failure | |
4222 | * | |
4223 | * Timestamps: | |
3558fd73 BB |
4224 | * tdip - ctime|mtime updated |
4225 | * sip - ctime updated | |
34dc7c2f BB |
4226 | */ |
4227 | /* ARGSUSED */ | |
e5c39b95 | 4228 | int |
da5e151f BB |
4229 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr, |
4230 | int flags) | |
34dc7c2f | 4231 | { |
3558fd73 | 4232 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 4233 | znode_t *tzp, *szp; |
0037b49e | 4234 | zfsvfs_t *zfsvfs = ITOZSB(tdip); |
34dc7c2f BB |
4235 | zilog_t *zilog; |
4236 | zfs_dirlock_t *dl; | |
4237 | dmu_tx_t *tx; | |
34dc7c2f BB |
4238 | int error; |
4239 | int zf = ZNEW; | |
428870ff | 4240 | uint64_t parent; |
572e2857 | 4241 | uid_t owner; |
e8b96c60 | 4242 | boolean_t waited = B_FALSE; |
ace1eae8 CC |
4243 | boolean_t is_tmpfile = 0; |
4244 | uint64_t txg; | |
4245 | #ifdef HAVE_TMPFILE | |
4246 | is_tmpfile = (sip->i_nlink == 0 && (sip->i_state & I_LINKABLE)); | |
4247 | #endif | |
3558fd73 | 4248 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 4249 | |
32dec7bd | 4250 | if (name == NULL) |
4251 | return (SET_ERROR(EINVAL)); | |
4252 | ||
0037b49e | 4253 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 4254 | ZFS_VERIFY_ZP(dzp); |
0037b49e | 4255 | zilog = zfsvfs->z_log; |
34dc7c2f | 4256 | |
428870ff BB |
4257 | /* |
4258 | * POSIX dictates that we return EPERM here. | |
4259 | * Better choices include ENOTSUP or EISDIR. | |
4260 | */ | |
3558fd73 | 4261 | if (S_ISDIR(sip->i_mode)) { |
0037b49e | 4262 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4263 | return (SET_ERROR(EPERM)); |
428870ff BB |
4264 | } |
4265 | ||
812e91a7 MT |
4266 | szp = ITOZ(sip); |
4267 | ZFS_VERIFY_ZP(szp); | |
4268 | ||
9c5167d1 NF |
4269 | /* |
4270 | * If we are using project inheritance, means if the directory has | |
4271 | * ZFS_PROJINHERIT set, then its descendant directories will inherit | |
4272 | * not only the project ID, but also the ZFS_PROJINHERIT flag. Under | |
4273 | * such case, we only allow hard link creation in our tree when the | |
4274 | * project IDs are the same. | |
4275 | */ | |
4276 | if (dzp->z_pflags & ZFS_PROJINHERIT && dzp->z_projid != szp->z_projid) { | |
4277 | ZFS_EXIT(zfsvfs); | |
4278 | return (SET_ERROR(EXDEV)); | |
4279 | } | |
4280 | ||
812e91a7 MT |
4281 | /* |
4282 | * We check i_sb because snapshots and the ctldir must have different | |
4283 | * super blocks. | |
4284 | */ | |
c0ebc844 | 4285 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
0037b49e | 4286 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4287 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 4288 | } |
428870ff | 4289 | |
428870ff BB |
4290 | /* Prevent links to .zfs/shares files */ |
4291 | ||
0037b49e | 4292 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
428870ff | 4293 | &parent, sizeof (uint64_t))) != 0) { |
0037b49e | 4294 | ZFS_EXIT(zfsvfs); |
428870ff BB |
4295 | return (error); |
4296 | } | |
0037b49e BB |
4297 | if (parent == zfsvfs->z_shares_dir) { |
4298 | ZFS_EXIT(zfsvfs); | |
2e528b49 | 4299 | return (SET_ERROR(EPERM)); |
428870ff BB |
4300 | } |
4301 | ||
0037b49e | 4302 | if (zfsvfs->z_utf8 && u8_validate(name, |
34dc7c2f | 4303 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
0037b49e | 4304 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4305 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
4306 | } |
4307 | if (flags & FIGNORECASE) | |
4308 | zf |= ZCILOOK; | |
4309 | ||
34dc7c2f BB |
4310 | /* |
4311 | * We do not support links between attributes and non-attributes | |
4312 | * because of the potential security risk of creating links | |
4313 | * into "normal" file space in order to circumvent restrictions | |
4314 | * imposed in attribute space. | |
4315 | */ | |
428870ff | 4316 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
0037b49e | 4317 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4318 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4319 | } |
4320 | ||
0037b49e BB |
4321 | owner = zfs_fuid_map_id(zfsvfs, KUID_TO_SUID(sip->i_uid), |
4322 | cr, ZFS_OWNER); | |
572e2857 | 4323 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
0037b49e | 4324 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4325 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4326 | } |
4327 | ||
149e873a | 4328 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
0037b49e | 4329 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4330 | return (error); |
4331 | } | |
4332 | ||
428870ff | 4333 | top: |
34dc7c2f BB |
4334 | /* |
4335 | * Attempt to lock directory; fail if entry already exists. | |
4336 | */ | |
4337 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
4338 | if (error) { | |
0037b49e | 4339 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4340 | return (error); |
4341 | } | |
4342 | ||
0037b49e | 4343 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 4344 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 4345 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
ace1eae8 | 4346 | if (is_tmpfile) |
0037b49e | 4347 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
ace1eae8 | 4348 | |
428870ff BB |
4349 | zfs_sa_upgrade_txholds(tx, szp); |
4350 | zfs_sa_upgrade_txholds(tx, dzp); | |
0735ecb3 | 4351 | error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT); |
34dc7c2f BB |
4352 | if (error) { |
4353 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 4354 | if (error == ERESTART) { |
e8b96c60 | 4355 | waited = B_TRUE; |
34dc7c2f BB |
4356 | dmu_tx_wait(tx); |
4357 | dmu_tx_abort(tx); | |
4358 | goto top; | |
4359 | } | |
4360 | dmu_tx_abort(tx); | |
0037b49e | 4361 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4362 | return (error); |
4363 | } | |
ace1eae8 CC |
4364 | /* unmark z_unlinked so zfs_link_create will not reject */ |
4365 | if (is_tmpfile) | |
4366 | szp->z_unlinked = 0; | |
34dc7c2f BB |
4367 | error = zfs_link_create(dl, szp, tx, 0); |
4368 | ||
4369 | if (error == 0) { | |
4370 | uint64_t txtype = TX_LINK; | |
ace1eae8 CC |
4371 | /* |
4372 | * tmpfile is created to be in z_unlinkedobj, so remove it. | |
4373 | * Also, we don't log in ZIL, be cause all previous file | |
4374 | * operation on the tmpfile are ignored by ZIL. Instead we | |
4375 | * always wait for txg to sync to make sure all previous | |
4376 | * operation are sync safe. | |
4377 | */ | |
4378 | if (is_tmpfile) { | |
0037b49e BB |
4379 | VERIFY(zap_remove_int(zfsvfs->z_os, |
4380 | zfsvfs->z_unlinkedobj, szp->z_id, tx) == 0); | |
ace1eae8 CC |
4381 | } else { |
4382 | if (flags & FIGNORECASE) | |
4383 | txtype |= TX_CI; | |
4384 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); | |
4385 | } | |
4386 | } else if (is_tmpfile) { | |
4387 | /* restore z_unlinked since when linking failed */ | |
4388 | szp->z_unlinked = 1; | |
34dc7c2f | 4389 | } |
ace1eae8 | 4390 | txg = dmu_tx_get_txg(tx); |
34dc7c2f BB |
4391 | dmu_tx_commit(tx); |
4392 | ||
4393 | zfs_dirent_unlock(dl); | |
4394 | ||
0037b49e | 4395 | if (!is_tmpfile && zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4396 | zil_commit(zilog, 0); |
428870ff | 4397 | |
ace1eae8 | 4398 | if (is_tmpfile) |
0037b49e | 4399 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), txg); |
ace1eae8 | 4400 | |
960e08fe BB |
4401 | zfs_inode_update(dzp); |
4402 | zfs_inode_update(szp); | |
0037b49e | 4403 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4404 | return (error); |
4405 | } | |
4406 | ||
3c0e5c0f | 4407 | static void |
119a394a | 4408 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
4409 | { |
4410 | struct page *pp = arg; | |
4411 | ||
119a394a | 4412 | ClearPageError(pp); |
3c0e5c0f BB |
4413 | end_page_writeback(pp); |
4414 | } | |
4415 | ||
34dc7c2f | 4416 | /* |
3c0e5c0f BB |
4417 | * Push a page out to disk, once the page is on stable storage the |
4418 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 4419 | * |
3c0e5c0f BB |
4420 | * IN: ip - page mapped for inode. |
4421 | * pp - page to push (page is locked) | |
4422 | * wbc - writeback control data | |
34dc7c2f BB |
4423 | * |
4424 | * RETURN: 0 if success | |
4425 | * error code if failure | |
4426 | * | |
3c0e5c0f BB |
4427 | * Timestamps: |
4428 | * ip - ctime|mtime updated | |
34dc7c2f BB |
4429 | */ |
4430 | /* ARGSUSED */ | |
3c0e5c0f BB |
4431 | int |
4432 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 4433 | { |
3c0e5c0f | 4434 | znode_t *zp = ITOZ(ip); |
0037b49e | 4435 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
3c0e5c0f BB |
4436 | loff_t offset; |
4437 | loff_t pgoff; | |
4c837f0d | 4438 | unsigned int pglen; |
3c0e5c0f BB |
4439 | dmu_tx_t *tx; |
4440 | caddr_t va; | |
4441 | int err = 0; | |
4442 | uint64_t mtime[2], ctime[2]; | |
4443 | sa_bulk_attr_t bulk[3]; | |
4444 | int cnt = 0; | |
21a96fb6 | 4445 | struct address_space *mapping; |
3c0e5c0f | 4446 | |
0037b49e | 4447 | ZFS_ENTER(zfsvfs); |
4c837f0d | 4448 | ZFS_VERIFY_ZP(zp); |
d164b209 | 4449 | |
3c0e5c0f BB |
4450 | ASSERT(PageLocked(pp)); |
4451 | ||
d1d7e268 MK |
4452 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
4453 | offset = i_size_read(ip); /* File length in bytes */ | |
8b1899d3 BB |
4454 | pglen = MIN(PAGE_SIZE, /* Page length in bytes */ |
4455 | P2ROUNDUP(offset, PAGE_SIZE)-pgoff); | |
3c0e5c0f BB |
4456 | |
4457 | /* Page is beyond end of file */ | |
4458 | if (pgoff >= offset) { | |
4459 | unlock_page(pp); | |
0037b49e | 4460 | ZFS_EXIT(zfsvfs); |
3c0e5c0f BB |
4461 | return (0); |
4462 | } | |
4463 | ||
4464 | /* Truncate page length to end of file */ | |
4465 | if (pgoff + pglen > offset) | |
4466 | pglen = offset - pgoff; | |
4467 | ||
4468 | #if 0 | |
34dc7c2f | 4469 | /* |
3c0e5c0f BB |
4470 | * FIXME: Allow mmap writes past its quota. The correct fix |
4471 | * is to register a page_mkwrite() handler to count the page | |
4472 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 4473 | */ |
9c5167d1 NF |
4474 | if (zfs_id_overblockquota(zfsvfs, DMU_USERUSED_OBJECT, |
4475 | KUID_TO_SUID(ip->i_uid)) || | |
4476 | zfs_id_overblockquota(zfsvfs, DMU_GROUPUSED_OBJECT, | |
4477 | KGID_TO_SGID(ip->i_gid)) || | |
4478 | (zp->z_projid != ZFS_DEFAULT_PROJID && | |
4479 | zfs_id_overblockquota(zfsvfs, DMU_PROJECTUSED_OBJECT, | |
4480 | zp->z_projid))) { | |
9babb374 | 4481 | err = EDQUOT; |
9babb374 | 4482 | } |
3c0e5c0f BB |
4483 | #endif |
4484 | ||
d958324f BB |
4485 | /* |
4486 | * The ordering here is critical and must adhere to the following | |
4487 | * rules in order to avoid deadlocking in either zfs_read() or | |
4488 | * zfs_free_range() due to a lock inversion. | |
4489 | * | |
4490 | * 1) The page must be unlocked prior to acquiring the range lock. | |
4491 | * This is critical because zfs_read() calls find_lock_page() | |
4492 | * which may block on the page lock while holding the range lock. | |
4493 | * | |
4494 | * 2) Before setting or clearing write back on a page the range lock | |
4495 | * must be held in order to prevent a lock inversion with the | |
4496 | * zfs_free_range() function. | |
21a96fb6 CC |
4497 | * |
4498 | * This presents a problem because upon entering this function the | |
4499 | * page lock is already held. To safely acquire the range lock the | |
4500 | * page lock must be dropped. This creates a window where another | |
4501 | * process could truncate, invalidate, dirty, or write out the page. | |
4502 | * | |
4503 | * Therefore, after successfully reacquiring the range and page locks | |
4504 | * the current page state is checked. In the common case everything | |
4505 | * will be as is expected and it can be written out. However, if | |
4506 | * the page state has changed it must be handled accordingly. | |
d958324f | 4507 | */ |
21a96fb6 CC |
4508 | mapping = pp->mapping; |
4509 | redirty_page_for_writepage(wbc, pp); | |
d958324f | 4510 | unlock_page(pp); |
21a96fb6 | 4511 | |
5d43cc9a MA |
4512 | locked_range_t *lr = rangelock_enter(&zp->z_rangelock, |
4513 | pgoff, pglen, RL_WRITER); | |
21a96fb6 CC |
4514 | lock_page(pp); |
4515 | ||
4516 | /* Page mapping changed or it was no longer dirty, we're done */ | |
4517 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
4518 | unlock_page(pp); | |
5d43cc9a | 4519 | rangelock_exit(lr); |
0037b49e | 4520 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4521 | return (0); |
4522 | } | |
4523 | ||
4524 | /* Another process started write block if required */ | |
4525 | if (PageWriteback(pp)) { | |
4526 | unlock_page(pp); | |
5d43cc9a | 4527 | rangelock_exit(lr); |
21a96fb6 CC |
4528 | |
4529 | if (wbc->sync_mode != WB_SYNC_NONE) | |
4530 | wait_on_page_writeback(pp); | |
4531 | ||
0037b49e | 4532 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4533 | return (0); |
4534 | } | |
4535 | ||
4536 | /* Clear the dirty flag the required locks are held */ | |
4537 | if (!clear_page_dirty_for_io(pp)) { | |
4538 | unlock_page(pp); | |
5d43cc9a | 4539 | rangelock_exit(lr); |
0037b49e | 4540 | ZFS_EXIT(zfsvfs); |
21a96fb6 CC |
4541 | return (0); |
4542 | } | |
4543 | ||
4544 | /* | |
4545 | * Counterpart for redirty_page_for_writepage() above. This page | |
4546 | * was in fact not skipped and should not be counted as if it were. | |
4547 | */ | |
4548 | wbc->pages_skipped--; | |
3c0e5c0f | 4549 | set_page_writeback(pp); |
21a96fb6 | 4550 | unlock_page(pp); |
3c0e5c0f | 4551 | |
0037b49e | 4552 | tx = dmu_tx_create(zfsvfs->z_os); |
3c0e5c0f | 4553 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
4554 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4555 | zfs_sa_upgrade_txholds(tx, zp); | |
d958324f | 4556 | |
fb5f0bc8 | 4557 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 4558 | if (err != 0) { |
3c0e5c0f | 4559 | if (err == ERESTART) |
34dc7c2f | 4560 | dmu_tx_wait(tx); |
3c0e5c0f | 4561 | |
34dc7c2f | 4562 | dmu_tx_abort(tx); |
119a394a ED |
4563 | __set_page_dirty_nobuffers(pp); |
4564 | ClearPageError(pp); | |
4565 | end_page_writeback(pp); | |
5d43cc9a | 4566 | rangelock_exit(lr); |
0037b49e | 4567 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4568 | return (err); |
34dc7c2f BB |
4569 | } |
4570 | ||
dde471ef | 4571 | va = kmap(pp); |
8b1899d3 | 4572 | ASSERT3U(pglen, <=, PAGE_SIZE); |
0037b49e | 4573 | dmu_write(zfsvfs->z_os, zp->z_id, pgoff, pglen, va, tx); |
dde471ef | 4574 | kunmap(pp); |
34dc7c2f | 4575 | |
0037b49e BB |
4576 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
4577 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
4578 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zfsvfs), NULL, | |
4579 | &zp->z_pflags, 8); | |
428870ff | 4580 | |
d3aa3ea9 BB |
4581 | /* Preserve the mtime and ctime provided by the inode */ |
4582 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4583 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
4584 | zp->z_atime_dirty = 0; | |
4585 | zp->z_seq++; | |
4586 | ||
4587 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4588 | ||
0037b49e | 4589 | zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
119a394a | 4590 | zfs_putpage_commit_cb, pp); |
45d1cae3 | 4591 | dmu_tx_commit(tx); |
d3aa3ea9 | 4592 | |
5d43cc9a | 4593 | rangelock_exit(lr); |
34dc7c2f | 4594 | |
119a394a ED |
4595 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4596 | /* | |
4597 | * Note that this is rarely called under writepages(), because | |
4598 | * writepages() normally handles the entire commit for | |
4599 | * performance reasons. | |
4600 | */ | |
0037b49e | 4601 | zil_commit(zfsvfs->z_log, zp->z_id); |
2b286136 | 4602 | } |
3c0e5c0f | 4603 | |
0037b49e | 4604 | ZFS_EXIT(zfsvfs); |
3c0e5c0f | 4605 | return (err); |
34dc7c2f BB |
4606 | } |
4607 | ||
8780c539 BB |
4608 | /* |
4609 | * Update the system attributes when the inode has been dirtied. For the | |
023699cd | 4610 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
4611 | */ |
4612 | int | |
4613 | zfs_dirty_inode(struct inode *ip, int flags) | |
4614 | { | |
4615 | znode_t *zp = ITOZ(ip); | |
0037b49e | 4616 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
8780c539 | 4617 | dmu_tx_t *tx; |
023699cd MM |
4618 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
4619 | sa_bulk_attr_t bulk[4]; | |
704cd075 | 4620 | int error = 0; |
8780c539 BB |
4621 | int cnt = 0; |
4622 | ||
0037b49e | 4623 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
c944be5d BB |
4624 | return (0); |
4625 | ||
0037b49e | 4626 | ZFS_ENTER(zfsvfs); |
8780c539 BB |
4627 | ZFS_VERIFY_ZP(zp); |
4628 | ||
704cd075 CC |
4629 | #ifdef I_DIRTY_TIME |
4630 | /* | |
4631 | * This is the lazytime semantic indroduced in Linux 4.0 | |
4632 | * This flag will only be called from update_time when lazytime is set. | |
4633 | * (Note, I_DIRTY_SYNC will also set if not lazytime) | |
4634 | * Fortunately mtime and ctime are managed within ZFS itself, so we | |
4635 | * only need to dirty atime. | |
4636 | */ | |
4637 | if (flags == I_DIRTY_TIME) { | |
4638 | zp->z_atime_dirty = 1; | |
4639 | goto out; | |
4640 | } | |
4641 | #endif | |
4642 | ||
0037b49e | 4643 | tx = dmu_tx_create(zfsvfs->z_os); |
8780c539 BB |
4644 | |
4645 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4646 | zfs_sa_upgrade_txholds(tx, zp); | |
4647 | ||
4648 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4649 | if (error) { | |
4650 | dmu_tx_abort(tx); | |
4651 | goto out; | |
4652 | } | |
4653 | ||
4654 | mutex_enter(&zp->z_lock); | |
704cd075 CC |
4655 | zp->z_atime_dirty = 0; |
4656 | ||
0037b49e BB |
4657 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
4658 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
4659 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
4660 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
8780c539 | 4661 | |
023699cd | 4662 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4663 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4664 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4665 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
023699cd MM |
4666 | mode = ip->i_mode; |
4667 | ||
4668 | zp->z_mode = mode; | |
8780c539 BB |
4669 | |
4670 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4671 | mutex_exit(&zp->z_lock); | |
4672 | ||
4673 | dmu_tx_commit(tx); | |
4674 | out: | |
0037b49e | 4675 | ZFS_EXIT(zfsvfs); |
8780c539 BB |
4676 | return (error); |
4677 | } | |
8780c539 | 4678 | |
34dc7c2f BB |
4679 | /*ARGSUSED*/ |
4680 | void | |
c0d35759 | 4681 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4682 | { |
c0d35759 | 4683 | znode_t *zp = ITOZ(ip); |
0037b49e | 4684 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
0df9673f | 4685 | uint64_t atime[2]; |
34dc7c2f | 4686 | int error; |
cafbd2ac | 4687 | int need_unlock = 0; |
34dc7c2f | 4688 | |
cafbd2ac | 4689 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
0037b49e | 4690 | if (!RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)) { |
cafbd2ac | 4691 | need_unlock = 1; |
0037b49e | 4692 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); |
cafbd2ac | 4693 | } |
c0d35759 | 4694 | if (zp->z_sa_hdl == NULL) { |
cafbd2ac | 4695 | if (need_unlock) |
0037b49e | 4696 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
c0d35759 | 4697 | return; |
34dc7c2f BB |
4698 | } |
4699 | ||
4700 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
0037b49e | 4701 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f | 4702 | |
428870ff BB |
4703 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4704 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4705 | error = dmu_tx_assign(tx, TXG_WAIT); |
4706 | if (error) { | |
4707 | dmu_tx_abort(tx); | |
4708 | } else { | |
0df9673f | 4709 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
34dc7c2f | 4710 | mutex_enter(&zp->z_lock); |
0037b49e | 4711 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), |
0df9673f | 4712 | (void *)&atime, sizeof (atime), tx); |
34dc7c2f BB |
4713 | zp->z_atime_dirty = 0; |
4714 | mutex_exit(&zp->z_lock); | |
4715 | dmu_tx_commit(tx); | |
4716 | } | |
4717 | } | |
4718 | ||
4719 | zfs_zinactive(zp); | |
cafbd2ac | 4720 | if (need_unlock) |
0037b49e | 4721 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
34dc7c2f BB |
4722 | } |
4723 | ||
4724 | /* | |
4725 | * Bounds-check the seek operation. | |
4726 | * | |
3558fd73 | 4727 | * IN: ip - inode seeking within |
34dc7c2f BB |
4728 | * ooff - old file offset |
4729 | * noffp - pointer to new file offset | |
4730 | * ct - caller context | |
4731 | * | |
4732 | * RETURN: 0 if success | |
4733 | * EINVAL if new offset invalid | |
4734 | */ | |
4735 | /* ARGSUSED */ | |
3558fd73 | 4736 | int |
9623f736 | 4737 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4738 | { |
3558fd73 | 4739 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4740 | return (0); |
4741 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4742 | } | |
4743 | ||
34dc7c2f | 4744 | /* |
dde471ef | 4745 | * Fill pages with data from the disk. |
34dc7c2f BB |
4746 | */ |
4747 | static int | |
dde471ef | 4748 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4749 | { |
d1d7e268 | 4750 | znode_t *zp = ITOZ(ip); |
0037b49e | 4751 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
d1d7e268 | 4752 | objset_t *os; |
dde471ef | 4753 | struct page *cur_pp; |
d1d7e268 MK |
4754 | u_offset_t io_off, total; |
4755 | size_t io_len; | |
4756 | loff_t i_size; | |
4757 | unsigned page_idx; | |
4758 | int err; | |
34dc7c2f | 4759 | |
0037b49e | 4760 | os = zfsvfs->z_os; |
8b1899d3 | 4761 | io_len = nr_pages << PAGE_SHIFT; |
dde471ef PJ |
4762 | i_size = i_size_read(ip); |
4763 | io_off = page_offset(pl[0]); | |
4764 | ||
4765 | if (io_off + io_len > i_size) | |
4766 | io_len = i_size - io_off; | |
34dc7c2f BB |
4767 | |
4768 | /* | |
dde471ef | 4769 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4770 | */ |
dde471ef | 4771 | page_idx = 0; |
34dc7c2f | 4772 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4773 | caddr_t va; |
4774 | ||
540c3927 | 4775 | cur_pp = pl[page_idx++]; |
dde471ef | 4776 | va = kmap(cur_pp); |
9babb374 BB |
4777 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4778 | DMU_READ_PREFETCH); | |
dde471ef | 4779 | kunmap(cur_pp); |
34dc7c2f | 4780 | if (err) { |
b128c09f BB |
4781 | /* convert checksum errors into IO errors */ |
4782 | if (err == ECKSUM) | |
2e528b49 | 4783 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4784 | return (err); |
4785 | } | |
34dc7c2f | 4786 | } |
d164b209 | 4787 | |
34dc7c2f BB |
4788 | return (0); |
4789 | } | |
4790 | ||
4791 | /* | |
dde471ef | 4792 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4793 | * |
dde471ef PJ |
4794 | * IN: ip - inode of file to get data from. |
4795 | * pl - list of pages to read | |
4796 | * nr_pages - number of pages to read | |
34dc7c2f | 4797 | * |
d3cc8b15 | 4798 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4799 | * |
4800 | * Timestamps: | |
4801 | * vp - atime updated | |
4802 | */ | |
4803 | /* ARGSUSED */ | |
dde471ef PJ |
4804 | int |
4805 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4806 | { |
dde471ef | 4807 | znode_t *zp = ITOZ(ip); |
0037b49e | 4808 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
dde471ef | 4809 | int err; |
d164b209 | 4810 | |
d164b209 BB |
4811 | if (pl == NULL) |
4812 | return (0); | |
34dc7c2f | 4813 | |
0037b49e | 4814 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4815 | ZFS_VERIFY_ZP(zp); |
4816 | ||
dde471ef | 4817 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4818 | |
0037b49e | 4819 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4820 | return (err); |
4821 | } | |
4822 | ||
4823 | /* | |
e2e7aa2d | 4824 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4825 | * |
e2e7aa2d BB |
4826 | * IN: ip - inode of the file to mmap |
4827 | * off - file offset | |
4828 | * addrp - start address in memory region | |
4829 | * len - length of memory region | |
4830 | * vm_flags- address flags | |
34dc7c2f | 4831 | * |
e2e7aa2d BB |
4832 | * RETURN: 0 if success |
4833 | * error code if failure | |
34dc7c2f BB |
4834 | */ |
4835 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4836 | int |
4837 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4838 | unsigned long vm_flags) | |
34dc7c2f | 4839 | { |
e2e7aa2d | 4840 | znode_t *zp = ITOZ(ip); |
0037b49e | 4841 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4842 | |
0037b49e | 4843 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4844 | ZFS_VERIFY_ZP(zp); |
4845 | ||
e2e7aa2d | 4846 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4847 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
0037b49e | 4848 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4849 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4850 | } |
4851 | ||
e2e7aa2d | 4852 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4853 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
0037b49e | 4854 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4855 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4856 | } |
4857 | ||
34dc7c2f | 4858 | if (off < 0 || len > MAXOFFSET_T - off) { |
0037b49e | 4859 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4860 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4861 | } |
4862 | ||
0037b49e | 4863 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4864 | return (0); |
4865 | } | |
4866 | ||
3558fd73 BB |
4867 | /* |
4868 | * convoff - converts the given data (start, whence) to the | |
4869 | * given whence. | |
4870 | */ | |
4871 | int | |
4872 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4873 | { | |
5484965a | 4874 | vattr_t vap; |
3558fd73 BB |
4875 | int error; |
4876 | ||
126d0fa7 | 4877 | if ((lckdat->l_whence == SEEK_END) || (whence == SEEK_END)) { |
d95a5980 | 4878 | if ((error = zfs_getattr(ip, &vap, 0, CRED()))) |
3558fd73 BB |
4879 | return (error); |
4880 | } | |
4881 | ||
4882 | switch (lckdat->l_whence) { | |
126d0fa7 | 4883 | case SEEK_CUR: |
3558fd73 BB |
4884 | lckdat->l_start += offset; |
4885 | break; | |
126d0fa7 | 4886 | case SEEK_END: |
5484965a | 4887 | lckdat->l_start += vap.va_size; |
3558fd73 | 4888 | /* FALLTHRU */ |
126d0fa7 | 4889 | case SEEK_SET: |
3558fd73 BB |
4890 | break; |
4891 | default: | |
2e528b49 | 4892 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4893 | } |
4894 | ||
4895 | if (lckdat->l_start < 0) | |
2e528b49 | 4896 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4897 | |
4898 | switch (whence) { | |
126d0fa7 | 4899 | case SEEK_CUR: |
3558fd73 BB |
4900 | lckdat->l_start -= offset; |
4901 | break; | |
126d0fa7 | 4902 | case SEEK_END: |
5484965a | 4903 | lckdat->l_start -= vap.va_size; |
3558fd73 | 4904 | /* FALLTHRU */ |
126d0fa7 | 4905 | case SEEK_SET: |
3558fd73 BB |
4906 | break; |
4907 | default: | |
2e528b49 | 4908 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4909 | } |
4910 | ||
4911 | lckdat->l_whence = (short)whence; | |
4912 | return (0); | |
4913 | } | |
4914 | ||
34dc7c2f BB |
4915 | /* |
4916 | * Free or allocate space in a file. Currently, this function only | |
4917 | * supports the `F_FREESP' command. However, this command is somewhat | |
4918 | * misnamed, as its functionality includes the ability to allocate as | |
4919 | * well as free space. | |
4920 | * | |
3558fd73 | 4921 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4922 | * cmd - action to take (only F_FREESP supported). |
4923 | * bfp - section of file to free/alloc. | |
4924 | * flag - current file open mode flags. | |
4925 | * offset - current file offset. | |
a35c1207 | 4926 | * cr - credentials of caller. |
34dc7c2f | 4927 | * |
d3cc8b15 | 4928 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4929 | * |
4930 | * Timestamps: | |
3558fd73 | 4931 | * ip - ctime|mtime updated |
34dc7c2f BB |
4932 | */ |
4933 | /* ARGSUSED */ | |
e5c39b95 | 4934 | int |
3558fd73 BB |
4935 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4936 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4937 | { |
3558fd73 | 4938 | znode_t *zp = ITOZ(ip); |
0037b49e | 4939 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
4940 | uint64_t off, len; |
4941 | int error; | |
4942 | ||
0037b49e | 4943 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
4944 | ZFS_VERIFY_ZP(zp); |
4945 | ||
34dc7c2f | 4946 | if (cmd != F_FREESP) { |
0037b49e | 4947 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4948 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4949 | } |
4950 | ||
f3c9dca0 MT |
4951 | /* |
4952 | * Callers might not be able to detect properly that we are read-only, | |
4953 | * so check it explicitly here. | |
4954 | */ | |
0037b49e BB |
4955 | if (zfs_is_readonly(zfsvfs)) { |
4956 | ZFS_EXIT(zfsvfs); | |
f3c9dca0 MT |
4957 | return (SET_ERROR(EROFS)); |
4958 | } | |
4959 | ||
126d0fa7 | 4960 | if ((error = convoff(ip, bfp, SEEK_SET, offset))) { |
0037b49e | 4961 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4962 | return (error); |
4963 | } | |
4964 | ||
4965 | if (bfp->l_len < 0) { | |
0037b49e | 4966 | ZFS_EXIT(zfsvfs); |
2e528b49 | 4967 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4968 | } |
4969 | ||
aec69371 ED |
4970 | /* |
4971 | * Permissions aren't checked on Solaris because on this OS | |
4972 | * zfs_space() can only be called with an opened file handle. | |
4973 | * On Linux we can get here through truncate_range() which | |
4974 | * operates directly on inodes, so we need to check access rights. | |
4975 | */ | |
4976 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
0037b49e | 4977 | ZFS_EXIT(zfsvfs); |
aec69371 ED |
4978 | return (error); |
4979 | } | |
4980 | ||
34dc7c2f BB |
4981 | off = bfp->l_start; |
4982 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4983 | ||
b128c09f | 4984 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4985 | |
0037b49e | 4986 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
4987 | return (error); |
4988 | } | |
4989 | ||
4990 | /*ARGSUSED*/ | |
e5c39b95 | 4991 | int |
3558fd73 | 4992 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4993 | { |
3558fd73 | 4994 | znode_t *zp = ITOZ(ip); |
0037b49e | 4995 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 4996 | uint32_t gen; |
428870ff | 4997 | uint64_t gen64; |
34dc7c2f BB |
4998 | uint64_t object = zp->z_id; |
4999 | zfid_short_t *zfid; | |
428870ff | 5000 | int size, i, error; |
34dc7c2f | 5001 | |
0037b49e | 5002 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 5003 | ZFS_VERIFY_ZP(zp); |
428870ff | 5004 | |
0037b49e | 5005 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), |
428870ff | 5006 | &gen64, sizeof (uint64_t))) != 0) { |
0037b49e | 5007 | ZFS_EXIT(zfsvfs); |
428870ff BB |
5008 | return (error); |
5009 | } | |
5010 | ||
5011 | gen = (uint32_t)gen64; | |
34dc7c2f | 5012 | |
9b77d1c9 | 5013 | size = SHORT_FID_LEN; |
34dc7c2f BB |
5014 | |
5015 | zfid = (zfid_short_t *)fidp; | |
5016 | ||
5017 | zfid->zf_len = size; | |
5018 | ||
5019 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
5020 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
5021 | ||
5022 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
5023 | if (gen == 0) | |
5024 | gen = 1; | |
5025 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
5026 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
5027 | ||
0037b49e | 5028 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5029 | return (0); |
5030 | } | |
5031 | ||
34dc7c2f | 5032 | /*ARGSUSED*/ |
e5c39b95 | 5033 | int |
3558fd73 | 5034 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 5035 | { |
3558fd73 | 5036 | znode_t *zp = ITOZ(ip); |
0037b49e | 5037 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
5038 | int error; |
5039 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
5040 | ||
0037b49e | 5041 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
5042 | ZFS_VERIFY_ZP(zp); |
5043 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
0037b49e | 5044 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5045 | |
5046 | return (error); | |
5047 | } | |
5048 | ||
5049 | /*ARGSUSED*/ | |
e5c39b95 | 5050 | int |
3558fd73 | 5051 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 5052 | { |
3558fd73 | 5053 | znode_t *zp = ITOZ(ip); |
0037b49e | 5054 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f BB |
5055 | int error; |
5056 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
0037b49e | 5057 | zilog_t *zilog = zfsvfs->z_log; |
34dc7c2f | 5058 | |
0037b49e | 5059 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 5060 | ZFS_VERIFY_ZP(zp); |
428870ff | 5061 | |
34dc7c2f | 5062 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 5063 | |
0037b49e | 5064 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 5065 | zil_commit(zilog, 0); |
428870ff | 5066 | |
0037b49e | 5067 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
5068 | return (error); |
5069 | } | |
5070 | ||
3558fd73 | 5071 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
5072 | /* |
5073 | * Tunable, both must be a power of 2. | |
5074 | * | |
5075 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
5076 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 5077 | * an arcbuf for a partial block read |
428870ff BB |
5078 | */ |
5079 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
5080 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
5081 | ||
5082 | /*ARGSUSED*/ | |
5083 | static int | |
3558fd73 | 5084 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 5085 | { |
3558fd73 | 5086 | znode_t *zp = ITOZ(ip); |
0037b49e BB |
5087 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
5088 | int max_blksz = zfsvfs->z_max_blksz; | |
428870ff BB |
5089 | uio_t *uio = &xuio->xu_uio; |
5090 | ssize_t size = uio->uio_resid; | |
5091 | offset_t offset = uio->uio_loffset; | |
5092 | int blksz; | |
5093 | int fullblk, i; | |
5094 | arc_buf_t *abuf; | |
5095 | ssize_t maxsize; | |
5096 | int preamble, postamble; | |
5097 | ||
5098 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
2e528b49 | 5099 | return (SET_ERROR(EINVAL)); |
428870ff | 5100 | |
0037b49e | 5101 | ZFS_ENTER(zfsvfs); |
428870ff BB |
5102 | ZFS_VERIFY_ZP(zp); |
5103 | switch (ioflag) { | |
5104 | case UIO_WRITE: | |
5105 | /* | |
5106 | * Loan out an arc_buf for write if write size is bigger than | |
5107 | * max_blksz, and the file's block size is also max_blksz. | |
5108 | */ | |
5109 | blksz = max_blksz; | |
5110 | if (size < blksz || zp->z_blksz != blksz) { | |
0037b49e | 5111 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5112 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5113 | } |
5114 | /* | |
5115 | * Caller requests buffers for write before knowing where the | |
5116 | * write offset might be (e.g. NFS TCP write). | |
5117 | */ | |
5118 | if (offset == -1) { | |
5119 | preamble = 0; | |
5120 | } else { | |
5121 | preamble = P2PHASE(offset, blksz); | |
5122 | if (preamble) { | |
5123 | preamble = blksz - preamble; | |
5124 | size -= preamble; | |
5125 | } | |
5126 | } | |
5127 | ||
5128 | postamble = P2PHASE(size, blksz); | |
5129 | size -= postamble; | |
5130 | ||
5131 | fullblk = size / blksz; | |
5132 | (void) dmu_xuio_init(xuio, | |
5133 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
5134 | |
5135 | /* | |
5136 | * Have to fix iov base/len for partial buffers. They | |
5137 | * currently represent full arc_buf's. | |
5138 | */ | |
5139 | if (preamble) { | |
5140 | /* data begins 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, | |
5145 | blksz - preamble, preamble); | |
5146 | } | |
5147 | ||
5148 | for (i = 0; i < fullblk; i++) { | |
5149 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5150 | blksz); | |
5151 | ASSERT(abuf); | |
5152 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
5153 | } | |
5154 | ||
5155 | if (postamble) { | |
5156 | /* data ends in the middle of the arc_buf */ | |
5157 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5158 | blksz); | |
5159 | ASSERT(abuf); | |
5160 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
5161 | } | |
5162 | break; | |
5163 | case UIO_READ: | |
5164 | /* | |
5165 | * Loan out an arc_buf for read if the read size is larger than | |
5166 | * the current file block size. Block alignment is not | |
5167 | * considered. Partial arc_buf will be loaned out for read. | |
5168 | */ | |
5169 | blksz = zp->z_blksz; | |
5170 | if (blksz < zcr_blksz_min) | |
5171 | blksz = zcr_blksz_min; | |
5172 | if (blksz > zcr_blksz_max) | |
5173 | blksz = zcr_blksz_max; | |
5174 | /* avoid potential complexity of dealing with it */ | |
5175 | if (blksz > max_blksz) { | |
0037b49e | 5176 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5177 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5178 | } |
5179 | ||
5180 | maxsize = zp->z_size - uio->uio_loffset; | |
5181 | if (size > maxsize) | |
5182 | size = maxsize; | |
5183 | ||
3558fd73 | 5184 | if (size < blksz) { |
0037b49e | 5185 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5186 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5187 | } |
5188 | break; | |
5189 | default: | |
0037b49e | 5190 | ZFS_EXIT(zfsvfs); |
2e528b49 | 5191 | return (SET_ERROR(EINVAL)); |
428870ff BB |
5192 | } |
5193 | ||
5194 | uio->uio_extflg = UIO_XUIO; | |
5195 | XUIO_XUZC_RW(xuio) = ioflag; | |
0037b49e | 5196 | ZFS_EXIT(zfsvfs); |
428870ff BB |
5197 | return (0); |
5198 | } | |
5199 | ||
5200 | /*ARGSUSED*/ | |
5201 | static int | |
3558fd73 | 5202 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
5203 | { |
5204 | int i; | |
5205 | arc_buf_t *abuf; | |
5206 | int ioflag = XUIO_XUZC_RW(xuio); | |
5207 | ||
5208 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
5209 | ||
5210 | i = dmu_xuio_cnt(xuio); | |
5211 | while (i-- > 0) { | |
5212 | abuf = dmu_xuio_arcbuf(xuio, i); | |
5213 | /* | |
5214 | * if abuf == NULL, it must be a write buffer | |
5215 | * that has been returned in zfs_write(). | |
5216 | */ | |
5217 | if (abuf) | |
5218 | dmu_return_arcbuf(abuf); | |
5219 | ASSERT(abuf || ioflag == UIO_WRITE); | |
5220 | } | |
5221 | ||
5222 | dmu_xuio_fini(xuio); | |
5223 | return (0); | |
5224 | } | |
3558fd73 | 5225 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 | 5226 | |
93ce2b4c | 5227 | #if defined(_KERNEL) |
f298b24d BB |
5228 | EXPORT_SYMBOL(zfs_open); |
5229 | EXPORT_SYMBOL(zfs_close); | |
5230 | EXPORT_SYMBOL(zfs_read); | |
5231 | EXPORT_SYMBOL(zfs_write); | |
5232 | EXPORT_SYMBOL(zfs_access); | |
5233 | EXPORT_SYMBOL(zfs_lookup); | |
5234 | EXPORT_SYMBOL(zfs_create); | |
5235 | EXPORT_SYMBOL(zfs_tmpfile); | |
5236 | EXPORT_SYMBOL(zfs_remove); | |
5237 | EXPORT_SYMBOL(zfs_mkdir); | |
5238 | EXPORT_SYMBOL(zfs_rmdir); | |
5239 | EXPORT_SYMBOL(zfs_readdir); | |
5240 | EXPORT_SYMBOL(zfs_fsync); | |
5241 | EXPORT_SYMBOL(zfs_getattr); | |
5242 | EXPORT_SYMBOL(zfs_getattr_fast); | |
5243 | EXPORT_SYMBOL(zfs_setattr); | |
5244 | EXPORT_SYMBOL(zfs_rename); | |
5245 | EXPORT_SYMBOL(zfs_symlink); | |
5246 | EXPORT_SYMBOL(zfs_readlink); | |
5247 | EXPORT_SYMBOL(zfs_link); | |
5248 | EXPORT_SYMBOL(zfs_inactive); | |
5249 | EXPORT_SYMBOL(zfs_space); | |
5250 | EXPORT_SYMBOL(zfs_fid); | |
5251 | EXPORT_SYMBOL(zfs_getsecattr); | |
5252 | EXPORT_SYMBOL(zfs_setsecattr); | |
5253 | EXPORT_SYMBOL(zfs_getpage); | |
5254 | EXPORT_SYMBOL(zfs_putpage); | |
5255 | EXPORT_SYMBOL(zfs_dirty_inode); | |
5256 | EXPORT_SYMBOL(zfs_map); | |
5257 | ||
02730c33 | 5258 | /* CSTYLED */ |
a966c564 K |
5259 | module_param(zfs_delete_blocks, ulong, 0644); |
5260 | MODULE_PARM_DESC(zfs_delete_blocks, "Delete files larger than N blocks async"); | |
c409e464 BB |
5261 | module_param(zfs_read_chunk_size, long, 0644); |
5262 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
5263 | #endif |