]>
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 | */ | |
21 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
a08ee875 | 23 | * Copyright (c) 2013 by Delphix. All rights reserved. |
e10b0808 | 24 | * Copyright (c) 2015 by Chunwei Chen. All rights reserved. |
34dc7c2f BB |
25 | */ |
26 | ||
27 | /* Portions Copyright 2007 Jeremy Teo */ | |
428870ff | 28 | /* Portions Copyright 2010 Robert Milkowski */ |
34dc7c2f | 29 | |
60101509 | 30 | |
34dc7c2f BB |
31 | #include <sys/types.h> |
32 | #include <sys/param.h> | |
33 | #include <sys/time.h> | |
34 | #include <sys/systm.h> | |
35 | #include <sys/sysmacros.h> | |
36 | #include <sys/resource.h> | |
37 | #include <sys/vfs.h> | |
38 | #include <sys/vfs_opreg.h> | |
34dc7c2f BB |
39 | #include <sys/file.h> |
40 | #include <sys/stat.h> | |
41 | #include <sys/kmem.h> | |
42 | #include <sys/taskq.h> | |
43 | #include <sys/uio.h> | |
44 | #include <sys/vmsystm.h> | |
45 | #include <sys/atomic.h> | |
34dc7c2f | 46 | #include <vm/pvn.h> |
34dc7c2f BB |
47 | #include <sys/pathname.h> |
48 | #include <sys/cmn_err.h> | |
49 | #include <sys/errno.h> | |
50 | #include <sys/unistd.h> | |
51 | #include <sys/zfs_dir.h> | |
52 | #include <sys/zfs_acl.h> | |
53 | #include <sys/zfs_ioctl.h> | |
54 | #include <sys/fs/zfs.h> | |
55 | #include <sys/dmu.h> | |
428870ff | 56 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
57 | #include <sys/spa.h> |
58 | #include <sys/txg.h> | |
59 | #include <sys/dbuf.h> | |
60 | #include <sys/zap.h> | |
428870ff | 61 | #include <sys/sa.h> |
34dc7c2f BB |
62 | #include <sys/dirent.h> |
63 | #include <sys/policy.h> | |
64 | #include <sys/sunddi.h> | |
b128c09f | 65 | #include <sys/sid.h> |
bcf30822 | 66 | #include <sys/mode.h> |
34dc7c2f | 67 | #include "fs/fs_subr.h" |
ebe7e575 | 68 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 69 | #include <sys/zfs_fuid.h> |
428870ff | 70 | #include <sys/zfs_sa.h> |
e5c39b95 | 71 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
72 | #include <sys/dnlc.h> |
73 | #include <sys/zfs_rlock.h> | |
74 | #include <sys/extdirent.h> | |
75 | #include <sys/kidmap.h> | |
428870ff | 76 | #include <sys/cred.h> |
34dc7c2f | 77 | #include <sys/attr.h> |
218b8eaf | 78 | #include <sys/zpl.h> |
34dc7c2f BB |
79 | |
80 | /* | |
81 | * Programming rules. | |
82 | * | |
83 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
84 | * properly lock its in-core state, create a DMU transaction, do the work, | |
85 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
86 | * and wait for the intent log to commit if it is a synchronous operation. | |
87 | * Moreover, the vnode ops must work in both normal and log replay context. | |
88 | * The ordering of events is important to avoid deadlocks and references | |
89 | * to freed memory. The example below illustrates the following Big Rules: | |
90 | * | |
91 | * (1) A check must be made in each zfs thread for a mounted file system. | |
3558fd73 BB |
92 | * This is done avoiding races using ZFS_ENTER(zsb). |
93 | * A ZFS_EXIT(zsb) is needed before all returns. Any znodes | |
34dc7c2f BB |
94 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros |
95 | * can return EIO from the calling function. | |
96 | * | |
3558fd73 | 97 | * (2) iput() should always be the last thing except for zil_commit() |
34dc7c2f BB |
98 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: |
99 | * First, if it's the last reference, the vnode/znode | |
100 | * can be freed, so the zp may point to freed memory. Second, the last | |
101 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
102 | * pushing cached pages (which acquires range locks) and syncing out | |
103 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
104 | * which could deadlock the system if you were already holding one. | |
ea04106b | 105 | * If you must call iput() within a tx then use zfs_iput_async(). |
34dc7c2f BB |
106 | * |
107 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
108 | * as they can span dmu_tx_assign() calls. | |
109 | * | |
a08ee875 LG |
110 | * (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to |
111 | * dmu_tx_assign(). This is critical because we don't want to block | |
112 | * while holding locks. | |
113 | * | |
114 | * If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT. This | |
115 | * reduces lock contention and CPU usage when we must wait (note that if | |
116 | * throughput is constrained by the storage, nearly every transaction | |
117 | * must wait). | |
118 | * | |
119 | * Note, in particular, that if a lock is sometimes acquired before | |
120 | * the tx assigns, and sometimes after (e.g. z_lock), then failing | |
121 | * to use a non-blocking assign can deadlock the system. The scenario: | |
34dc7c2f BB |
122 | * |
123 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
124 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
125 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
126 | * forever, because the previous txg can't quiesce until B's tx commits. | |
127 | * | |
3558fd73 | 128 | * If dmu_tx_assign() returns ERESTART and zsb->z_assign is TXG_NOWAIT, |
a08ee875 LG |
129 | * then drop all locks, call dmu_tx_wait(), and try again. On subsequent |
130 | * calls to dmu_tx_assign(), pass TXG_WAITED rather than TXG_NOWAIT, | |
131 | * to indicate that this operation has already called dmu_tx_wait(). | |
132 | * This will ensure that we don't retry forever, waiting a short bit | |
133 | * each time. | |
34dc7c2f BB |
134 | * |
135 | * (5) If the operation succeeded, generate the intent log entry for it | |
136 | * before dropping locks. This ensures that the ordering of events | |
137 | * in the intent log matches the order in which they actually occurred. | |
a08ee875 | 138 | * During ZIL replay the zfs_log_* functions will update the sequence |
fb5f0bc8 | 139 | * number to indicate the zil transaction has replayed. |
34dc7c2f BB |
140 | * |
141 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
142 | * regardless of whether there were any errors. | |
143 | * | |
572e2857 | 144 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
145 | * to ensure that synchronous semantics are provided when necessary. |
146 | * | |
147 | * In general, this is how things should be ordered in each vnode op: | |
148 | * | |
3558fd73 | 149 | * ZFS_ENTER(zsb); // exit if unmounted |
34dc7c2f | 150 | * top: |
3558fd73 | 151 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
152 | * rw_enter(...); // grab any other locks you need |
153 | * tx = dmu_tx_create(...); // get DMU tx | |
154 | * dmu_tx_hold_*(); // hold each object you might modify | |
a08ee875 | 155 | * error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
156 | * if (error) { |
157 | * rw_exit(...); // drop locks | |
158 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 159 | * iput(...); // release held vnodes |
fb5f0bc8 | 160 | * if (error == ERESTART) { |
a08ee875 | 161 | * waited = B_TRUE; |
34dc7c2f BB |
162 | * dmu_tx_wait(tx); |
163 | * dmu_tx_abort(tx); | |
164 | * goto top; | |
165 | * } | |
166 | * dmu_tx_abort(tx); // abort DMU tx | |
3558fd73 | 167 | * ZFS_EXIT(zsb); // finished in zfs |
34dc7c2f BB |
168 | * return (error); // really out of space |
169 | * } | |
170 | * error = do_real_work(); // do whatever this VOP does | |
171 | * if (error == 0) | |
172 | * zfs_log_*(...); // on success, make ZIL entry | |
173 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
174 | * rw_exit(...); // drop locks | |
175 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 176 | * iput(...); // release held vnodes |
572e2857 | 177 | * zil_commit(zilog, foid); // synchronous when necessary |
3558fd73 | 178 | * ZFS_EXIT(zsb); // finished in zfs |
34dc7c2f BB |
179 | * return (error); // done, report error |
180 | */ | |
181 | ||
126400a1 BB |
182 | /* |
183 | * Virus scanning is unsupported. It would be possible to add a hook | |
184 | * here to performance the required virus scan. This could be done | |
185 | * entirely in the kernel or potentially as an update to invoke a | |
186 | * scanning utility. | |
187 | */ | |
188 | static int | |
189 | zfs_vscan(struct inode *ip, cred_t *cr, int async) | |
190 | { | |
191 | return (0); | |
192 | } | |
193 | ||
194 | /* ARGSUSED */ | |
195 | int | |
196 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
197 | { | |
198 | znode_t *zp = ITOZ(ip); | |
199 | zfs_sb_t *zsb = ITOZSB(ip); | |
200 | ||
201 | ZFS_ENTER(zsb); | |
202 | ZFS_VERIFY_ZP(zp); | |
203 | ||
204 | /* Honor ZFS_APPENDONLY file attribute */ | |
205 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
206 | ((flag & O_APPEND) == 0)) { | |
207 | ZFS_EXIT(zsb); | |
a08ee875 | 208 | return (SET_ERROR(EPERM)); |
126400a1 BB |
209 | } |
210 | ||
211 | /* Virus scan eligible files on open */ | |
212 | if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && | |
213 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { | |
214 | if (zfs_vscan(ip, cr, 0) != 0) { | |
215 | ZFS_EXIT(zsb); | |
a08ee875 | 216 | return (SET_ERROR(EACCES)); |
126400a1 BB |
217 | } |
218 | } | |
219 | ||
220 | /* Keep a count of the synchronous opens in the znode */ | |
221 | if (flag & O_SYNC) | |
222 | atomic_inc_32(&zp->z_sync_cnt); | |
223 | ||
224 | ZFS_EXIT(zsb); | |
225 | return (0); | |
226 | } | |
227 | EXPORT_SYMBOL(zfs_open); | |
228 | ||
229 | /* ARGSUSED */ | |
230 | int | |
231 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
232 | { | |
233 | znode_t *zp = ITOZ(ip); | |
234 | zfs_sb_t *zsb = ITOZSB(ip); | |
235 | ||
236 | ZFS_ENTER(zsb); | |
237 | ZFS_VERIFY_ZP(zp); | |
238 | ||
a08ee875 | 239 | /* Decrement the synchronous opens in the znode */ |
126400a1 | 240 | if (flag & O_SYNC) |
a08ee875 | 241 | atomic_dec_32(&zp->z_sync_cnt); |
126400a1 BB |
242 | |
243 | if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && | |
244 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) | |
245 | VERIFY(zfs_vscan(ip, cr, 1) == 0); | |
246 | ||
247 | ZFS_EXIT(zsb); | |
8780c539 | 248 | return (0); |
126400a1 BB |
249 | } |
250 | EXPORT_SYMBOL(zfs_close); | |
251 | ||
c06d4368 AX |
252 | #if defined(SEEK_HOLE) && defined(SEEK_DATA) |
253 | /* | |
254 | * Lseek support for finding holes (cmd == SEEK_HOLE) and | |
255 | * data (cmd == SEEK_DATA). "off" is an in/out parameter. | |
256 | */ | |
257 | static int | |
258 | zfs_holey_common(struct inode *ip, int cmd, loff_t *off) | |
259 | { | |
260 | znode_t *zp = ITOZ(ip); | |
261 | uint64_t noff = (uint64_t)*off; /* new offset */ | |
262 | uint64_t file_sz; | |
263 | int error; | |
264 | boolean_t hole; | |
265 | ||
266 | file_sz = zp->z_size; | |
267 | if (noff >= file_sz) { | |
a08ee875 | 268 | return (SET_ERROR(ENXIO)); |
c06d4368 AX |
269 | } |
270 | ||
271 | if (cmd == SEEK_HOLE) | |
272 | hole = B_TRUE; | |
273 | else | |
274 | hole = B_FALSE; | |
275 | ||
276 | error = dmu_offset_next(ZTOZSB(zp)->z_os, zp->z_id, hole, &noff); | |
277 | ||
ea04106b | 278 | if (error == ESRCH) |
a08ee875 | 279 | return (SET_ERROR(ENXIO)); |
ea04106b AX |
280 | |
281 | /* | |
282 | * We could find a hole that begins after the logical end-of-file, | |
283 | * because dmu_offset_next() only works on whole blocks. If the | |
284 | * EOF falls mid-block, then indicate that the "virtual hole" | |
285 | * at the end of the file begins at the logical EOF, rather than | |
286 | * at the end of the last block. | |
287 | */ | |
288 | if (noff > file_sz) { | |
289 | ASSERT(hole); | |
290 | noff = file_sz; | |
c06d4368 AX |
291 | } |
292 | ||
293 | if (noff < *off) | |
294 | return (error); | |
295 | *off = noff; | |
296 | return (error); | |
297 | } | |
298 | ||
299 | int | |
300 | zfs_holey(struct inode *ip, int cmd, loff_t *off) | |
301 | { | |
302 | znode_t *zp = ITOZ(ip); | |
303 | zfs_sb_t *zsb = ITOZSB(ip); | |
304 | int error; | |
305 | ||
306 | ZFS_ENTER(zsb); | |
307 | ZFS_VERIFY_ZP(zp); | |
308 | ||
309 | error = zfs_holey_common(ip, cmd, off); | |
310 | ||
311 | ZFS_EXIT(zsb); | |
312 | return (error); | |
313 | } | |
314 | EXPORT_SYMBOL(zfs_holey); | |
315 | #endif /* SEEK_HOLE && SEEK_DATA */ | |
316 | ||
c0d35759 | 317 | #if defined(_KERNEL) |
34dc7c2f BB |
318 | /* |
319 | * When a file is memory mapped, we must keep the IO data synchronized | |
320 | * between the DMU cache and the memory mapped pages. What this means: | |
321 | * | |
322 | * On Write: If we find a memory mapped page, we write to *both* | |
323 | * the page and the dmu buffer. | |
34dc7c2f | 324 | */ |
d164b209 | 325 | static void |
c0d35759 BB |
326 | update_pages(struct inode *ip, int64_t start, int len, |
327 | objset_t *os, uint64_t oid) | |
34dc7c2f | 328 | { |
c0d35759 BB |
329 | struct address_space *mp = ip->i_mapping; |
330 | struct page *pp; | |
331 | uint64_t nbytes; | |
d164b209 | 332 | int64_t off; |
c0d35759 | 333 | void *pb; |
34dc7c2f | 334 | |
c0d35759 BB |
335 | off = start & (PAGE_CACHE_SIZE-1); |
336 | for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) { | |
337 | nbytes = MIN(PAGE_CACHE_SIZE - off, len); | |
34dc7c2f | 338 | |
c0d35759 BB |
339 | pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT); |
340 | if (pp) { | |
341 | if (mapping_writably_mapped(mp)) | |
342 | flush_dcache_page(pp); | |
34dc7c2f | 343 | |
c0d35759 BB |
344 | pb = kmap(pp); |
345 | (void) dmu_read(os, oid, start+off, nbytes, pb+off, | |
9babb374 | 346 | DMU_READ_PREFETCH); |
c0d35759 BB |
347 | kunmap(pp); |
348 | ||
349 | if (mapping_writably_mapped(mp)) | |
350 | flush_dcache_page(pp); | |
351 | ||
352 | mark_page_accessed(pp); | |
353 | SetPageUptodate(pp); | |
354 | ClearPageError(pp); | |
355 | unlock_page(pp); | |
356 | page_cache_release(pp); | |
34dc7c2f | 357 | } |
c0d35759 | 358 | |
d164b209 | 359 | len -= nbytes; |
34dc7c2f | 360 | off = 0; |
34dc7c2f | 361 | } |
34dc7c2f BB |
362 | } |
363 | ||
364 | /* | |
365 | * When a file is memory mapped, we must keep the IO data synchronized | |
366 | * between the DMU cache and the memory mapped pages. What this means: | |
367 | * | |
368 | * On Read: We "read" preferentially from memory mapped pages, | |
369 | * else we default from the dmu buffer. | |
370 | * | |
371 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
a08ee875 | 372 | * the file is memory mapped. |
34dc7c2f BB |
373 | */ |
374 | static int | |
3558fd73 | 375 | mappedread(struct inode *ip, int nbytes, uio_t *uio) |
34dc7c2f | 376 | { |
c0d35759 BB |
377 | struct address_space *mp = ip->i_mapping; |
378 | struct page *pp; | |
3558fd73 | 379 | znode_t *zp = ITOZ(ip); |
34dc7c2f | 380 | int64_t start, off; |
c0d35759 | 381 | uint64_t bytes; |
34dc7c2f BB |
382 | int len = nbytes; |
383 | int error = 0; | |
c0d35759 | 384 | void *pb; |
34dc7c2f BB |
385 | |
386 | start = uio->uio_loffset; | |
c0d35759 BB |
387 | off = start & (PAGE_CACHE_SIZE-1); |
388 | for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) { | |
389 | bytes = MIN(PAGE_CACHE_SIZE - off, len); | |
390 | ||
391 | pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT); | |
392 | if (pp) { | |
393 | ASSERT(PageUptodate(pp)); | |
394 | ||
395 | pb = kmap(pp); | |
396 | error = uiomove(pb + off, bytes, UIO_READ, uio); | |
397 | kunmap(pp); | |
398 | ||
399 | if (mapping_writably_mapped(mp)) | |
400 | flush_dcache_page(pp); | |
401 | ||
402 | mark_page_accessed(pp); | |
403 | unlock_page(pp); | |
404 | page_cache_release(pp); | |
34dc7c2f | 405 | } else { |
e10b0808 AX |
406 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
407 | uio, bytes); | |
34dc7c2f | 408 | } |
c0d35759 | 409 | |
34dc7c2f BB |
410 | len -= bytes; |
411 | off = 0; | |
412 | if (error) | |
413 | break; | |
414 | } | |
415 | return (error); | |
416 | } | |
c0d35759 | 417 | #endif /* _KERNEL */ |
34dc7c2f | 418 | |
c409e464 | 419 | unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ |
34dc7c2f BB |
420 | |
421 | /* | |
422 | * Read bytes from specified file into supplied buffer. | |
423 | * | |
3558fd73 | 424 | * IN: ip - inode of file to be read from. |
34dc7c2f BB |
425 | * uio - structure supplying read location, range info, |
426 | * and return buffer. | |
c0d35759 BB |
427 | * ioflag - FSYNC flags; used to provide FRSYNC semantics. |
428 | * O_DIRECT flag; used to bypass page cache. | |
34dc7c2f | 429 | * cr - credentials of caller. |
34dc7c2f BB |
430 | * |
431 | * OUT: uio - updated offset and range, buffer filled. | |
432 | * | |
a08ee875 | 433 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
434 | * |
435 | * Side Effects: | |
3558fd73 | 436 | * inode - atime updated if byte count > 0 |
34dc7c2f BB |
437 | */ |
438 | /* ARGSUSED */ | |
e5c39b95 | 439 | int |
3558fd73 | 440 | zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 441 | { |
3558fd73 BB |
442 | znode_t *zp = ITOZ(ip); |
443 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 444 | ssize_t n, nbytes; |
149e873a | 445 | int error = 0; |
34dc7c2f | 446 | rl_t *rl; |
3558fd73 | 447 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 448 | xuio_t *xuio = NULL; |
3558fd73 | 449 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f | 450 | |
3558fd73 | 451 | ZFS_ENTER(zsb); |
34dc7c2f | 452 | ZFS_VERIFY_ZP(zp); |
34dc7c2f | 453 | |
428870ff | 454 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
3558fd73 | 455 | ZFS_EXIT(zsb); |
a08ee875 | 456 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
457 | } |
458 | ||
459 | /* | |
460 | * Validate file offset | |
461 | */ | |
462 | if (uio->uio_loffset < (offset_t)0) { | |
3558fd73 | 463 | ZFS_EXIT(zsb); |
a08ee875 | 464 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
465 | } |
466 | ||
467 | /* | |
468 | * Fasttrack empty reads | |
469 | */ | |
470 | if (uio->uio_resid == 0) { | |
3558fd73 | 471 | ZFS_EXIT(zsb); |
34dc7c2f BB |
472 | return (0); |
473 | } | |
474 | ||
34dc7c2f BB |
475 | /* |
476 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
477 | */ | |
3558fd73 BB |
478 | if (ioflag & FRSYNC || zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
479 | zil_commit(zsb->z_log, zp->z_id); | |
34dc7c2f BB |
480 | |
481 | /* | |
482 | * Lock the range against changes. | |
483 | */ | |
484 | rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER); | |
485 | ||
486 | /* | |
487 | * If we are reading past end-of-file we can skip | |
488 | * to the end; but we might still need to set atime. | |
489 | */ | |
428870ff | 490 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
491 | error = 0; |
492 | goto out; | |
493 | } | |
494 | ||
428870ff BB |
495 | ASSERT(uio->uio_loffset < zp->z_size); |
496 | n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); | |
497 | ||
3558fd73 | 498 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
499 | if ((uio->uio_extflg == UIO_XUIO) && |
500 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
501 | int nblk; | |
502 | int blksz = zp->z_blksz; | |
503 | uint64_t offset = uio->uio_loffset; | |
504 | ||
505 | xuio = (xuio_t *)uio; | |
506 | if ((ISP2(blksz))) { | |
507 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
508 | blksz)) / blksz; | |
509 | } else { | |
510 | ASSERT(offset + n <= blksz); | |
511 | nblk = 1; | |
512 | } | |
513 | (void) dmu_xuio_init(xuio, nblk); | |
514 | ||
3558fd73 | 515 | if (vn_has_cached_data(ip)) { |
428870ff BB |
516 | /* |
517 | * For simplicity, we always allocate a full buffer | |
518 | * even if we only expect to read a portion of a block. | |
519 | */ | |
520 | while (--nblk >= 0) { | |
521 | (void) dmu_xuio_add(xuio, | |
522 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
523 | blksz), 0, blksz); | |
524 | } | |
525 | } | |
526 | } | |
3558fd73 | 527 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
528 | |
529 | while (n > 0) { | |
530 | nbytes = MIN(n, zfs_read_chunk_size - | |
531 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); | |
532 | ||
e10b0808 | 533 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) { |
3558fd73 | 534 | error = mappedread(ip, nbytes, uio); |
e10b0808 AX |
535 | } else { |
536 | error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl), | |
537 | uio, nbytes); | |
538 | } | |
c0d35759 | 539 | |
b128c09f BB |
540 | if (error) { |
541 | /* convert checksum errors into IO errors */ | |
542 | if (error == ECKSUM) | |
a08ee875 | 543 | error = SET_ERROR(EIO); |
34dc7c2f | 544 | break; |
b128c09f | 545 | } |
34dc7c2f BB |
546 | |
547 | n -= nbytes; | |
548 | } | |
34dc7c2f BB |
549 | out: |
550 | zfs_range_unlock(rl); | |
551 | ||
3558fd73 | 552 | ZFS_ACCESSTIME_STAMP(zsb, zp); |
3558fd73 | 553 | ZFS_EXIT(zsb); |
34dc7c2f BB |
554 | return (error); |
555 | } | |
e5c39b95 | 556 | EXPORT_SYMBOL(zfs_read); |
34dc7c2f | 557 | |
34dc7c2f BB |
558 | /* |
559 | * Write the bytes to a file. | |
560 | * | |
3558fd73 | 561 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
562 | * uio - structure supplying write location, range info, |
563 | * and data buffer. | |
564 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 565 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 566 | * cr - credentials of caller. |
34dc7c2f BB |
567 | * |
568 | * OUT: uio - updated offset and range. | |
569 | * | |
570 | * RETURN: 0 if success | |
571 | * error code if failure | |
572 | * | |
573 | * Timestamps: | |
3558fd73 | 574 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 575 | */ |
428870ff | 576 | |
34dc7c2f | 577 | /* ARGSUSED */ |
e5c39b95 | 578 | int |
3558fd73 | 579 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 580 | { |
3558fd73 BB |
581 | znode_t *zp = ITOZ(ip); |
582 | rlim64_t limit = uio->uio_limit; | |
34dc7c2f BB |
583 | ssize_t start_resid = uio->uio_resid; |
584 | ssize_t tx_bytes; | |
585 | uint64_t end_size; | |
586 | dmu_tx_t *tx; | |
3558fd73 | 587 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
588 | zilog_t *zilog; |
589 | offset_t woff; | |
590 | ssize_t n, nbytes; | |
591 | rl_t *rl; | |
3558fd73 BB |
592 | int max_blksz = zsb->z_max_blksz; |
593 | int error = 0; | |
9babb374 | 594 | arc_buf_t *abuf; |
e10b0808 | 595 | const iovec_t *aiov = NULL; |
428870ff BB |
596 | xuio_t *xuio = NULL; |
597 | int i_iov = 0; | |
e10b0808 | 598 | const iovec_t *iovp = uio->uio_iov; |
428870ff BB |
599 | int write_eof; |
600 | int count = 0; | |
601 | sa_bulk_attr_t bulk[4]; | |
602 | uint64_t mtime[2], ctime[2]; | |
3558fd73 | 603 | ASSERTV(int iovcnt = uio->uio_iovcnt); |
34dc7c2f | 604 | |
34dc7c2f BB |
605 | /* |
606 | * Fasttrack empty write | |
607 | */ | |
608 | n = start_resid; | |
609 | if (n == 0) | |
610 | return (0); | |
611 | ||
612 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) | |
613 | limit = MAXOFFSET_T; | |
614 | ||
3558fd73 | 615 | ZFS_ENTER(zsb); |
34dc7c2f | 616 | ZFS_VERIFY_ZP(zp); |
b128c09f | 617 | |
3558fd73 BB |
618 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
619 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
620 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); | |
621 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, | |
428870ff BB |
622 | &zp->z_pflags, 8); |
623 | ||
b128c09f BB |
624 | /* |
625 | * If immutable or not appending then return EPERM | |
626 | */ | |
428870ff BB |
627 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
628 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
629 | (uio->uio_loffset < zp->z_size))) { | |
3558fd73 | 630 | ZFS_EXIT(zsb); |
a08ee875 | 631 | return (SET_ERROR(EPERM)); |
b128c09f BB |
632 | } |
633 | ||
3558fd73 | 634 | zilog = zsb->z_log; |
34dc7c2f | 635 | |
428870ff BB |
636 | /* |
637 | * Validate file offset | |
638 | */ | |
639 | woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; | |
640 | if (woff < 0) { | |
3558fd73 | 641 | ZFS_EXIT(zsb); |
a08ee875 | 642 | return (SET_ERROR(EINVAL)); |
428870ff BB |
643 | } |
644 | ||
34dc7c2f BB |
645 | /* |
646 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
647 | * don't hold up txg. | |
428870ff | 648 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 649 | */ |
a08ee875 | 650 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
651 | if ((uio->uio_extflg == UIO_XUIO) && |
652 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
653 | xuio = (xuio_t *)uio; | |
654 | else | |
a08ee875 | 655 | #endif |
572e2857 | 656 | uio_prefaultpages(MIN(n, max_blksz), uio); |
34dc7c2f BB |
657 | |
658 | /* | |
659 | * If in append mode, set the io offset pointer to eof. | |
660 | */ | |
661 | if (ioflag & FAPPEND) { | |
662 | /* | |
428870ff BB |
663 | * Obtain an appending range lock to guarantee file append |
664 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f BB |
665 | */ |
666 | rl = zfs_range_lock(zp, 0, n, RL_APPEND); | |
428870ff | 667 | woff = rl->r_off; |
34dc7c2f | 668 | if (rl->r_len == UINT64_MAX) { |
428870ff BB |
669 | /* |
670 | * We overlocked the file because this write will cause | |
671 | * the file block size to increase. | |
672 | * Note that zp_size cannot change with this lock held. | |
673 | */ | |
674 | woff = zp->z_size; | |
34dc7c2f | 675 | } |
428870ff | 676 | uio->uio_loffset = woff; |
34dc7c2f | 677 | } else { |
34dc7c2f | 678 | /* |
428870ff BB |
679 | * Note that if the file block size will change as a result of |
680 | * this write, then this range lock will lock the entire file | |
681 | * so that we can re-write the block safely. | |
34dc7c2f BB |
682 | */ |
683 | rl = zfs_range_lock(zp, woff, n, RL_WRITER); | |
684 | } | |
685 | ||
686 | if (woff >= limit) { | |
687 | zfs_range_unlock(rl); | |
3558fd73 | 688 | ZFS_EXIT(zsb); |
a08ee875 | 689 | return (SET_ERROR(EFBIG)); |
34dc7c2f BB |
690 | } |
691 | ||
692 | if ((woff + n) > limit || woff > (limit - n)) | |
693 | n = limit - woff; | |
694 | ||
428870ff BB |
695 | /* Will this write extend the file length? */ |
696 | write_eof = (woff + n > zp->z_size); | |
697 | ||
698 | end_size = MAX(zp->z_size, woff + n); | |
34dc7c2f BB |
699 | |
700 | /* | |
701 | * Write the file in reasonable size chunks. Each chunk is written | |
702 | * in a separate transaction; this keeps the intent log records small | |
703 | * and allows us to do more fine-grained space accounting. | |
704 | */ | |
705 | while (n > 0) { | |
9babb374 BB |
706 | abuf = NULL; |
707 | woff = uio->uio_loffset; | |
3558fd73 BB |
708 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
709 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 BB |
710 | if (abuf != NULL) |
711 | dmu_return_arcbuf(abuf); | |
a08ee875 | 712 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
713 | break; |
714 | } | |
715 | ||
428870ff BB |
716 | if (xuio && abuf == NULL) { |
717 | ASSERT(i_iov < iovcnt); | |
e10b0808 | 718 | ASSERT3U(uio->uio_segflg, !=, UIO_BVEC); |
428870ff BB |
719 | aiov = &iovp[i_iov]; |
720 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
721 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
722 | ASSERT((aiov->iov_base == abuf->b_data) || |
723 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
724 | aiov->iov_len == arc_buf_size(abuf))); | |
725 | i_iov++; | |
726 | } else if (abuf == NULL && n >= max_blksz && | |
727 | woff >= zp->z_size && | |
9babb374 BB |
728 | P2PHASE(woff, max_blksz) == 0 && |
729 | zp->z_blksz == max_blksz) { | |
428870ff BB |
730 | /* |
731 | * This write covers a full block. "Borrow" a buffer | |
732 | * from the dmu so that we can fill it before we enter | |
733 | * a transaction. This avoids the possibility of | |
734 | * holding up the transaction if the data copy hangs | |
735 | * up on a pagefault (e.g., from an NFS server mapping). | |
736 | */ | |
9babb374 BB |
737 | size_t cbytes; |
738 | ||
428870ff BB |
739 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
740 | max_blksz); | |
9babb374 BB |
741 | ASSERT(abuf != NULL); |
742 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
743 | if ((error = uiocopy(abuf->b_data, max_blksz, |
744 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
745 | dmu_return_arcbuf(abuf); |
746 | break; | |
747 | } | |
748 | ASSERT(cbytes == max_blksz); | |
749 | } | |
750 | ||
34dc7c2f BB |
751 | /* |
752 | * Start a transaction. | |
753 | */ | |
3558fd73 | 754 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 755 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 756 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 757 | zfs_sa_upgrade_txholds(tx, zp); |
a08ee875 | 758 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 759 | if (error) { |
34dc7c2f | 760 | dmu_tx_abort(tx); |
9babb374 BB |
761 | if (abuf != NULL) |
762 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
763 | break; |
764 | } | |
765 | ||
766 | /* | |
767 | * If zfs_range_lock() over-locked we grow the blocksize | |
768 | * and then reduce the lock range. This will only happen | |
769 | * on the first iteration since zfs_range_reduce() will | |
770 | * shrink down r_len to the appropriate size. | |
771 | */ | |
772 | if (rl->r_len == UINT64_MAX) { | |
773 | uint64_t new_blksz; | |
774 | ||
775 | if (zp->z_blksz > max_blksz) { | |
e10b0808 AX |
776 | /* |
777 | * File's blocksize is already larger than the | |
778 | * "recordsize" property. Only let it grow to | |
779 | * the next power of 2. | |
780 | */ | |
34dc7c2f | 781 | ASSERT(!ISP2(zp->z_blksz)); |
e10b0808 AX |
782 | new_blksz = MIN(end_size, |
783 | 1 << highbit64(zp->z_blksz)); | |
34dc7c2f BB |
784 | } else { |
785 | new_blksz = MIN(end_size, max_blksz); | |
786 | } | |
787 | zfs_grow_blocksize(zp, new_blksz, tx); | |
788 | zfs_range_reduce(rl, woff, n); | |
789 | } | |
790 | ||
791 | /* | |
792 | * XXX - should we really limit each write to z_max_blksz? | |
793 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
794 | */ | |
795 | nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); | |
34dc7c2f | 796 | |
9babb374 BB |
797 | if (abuf == NULL) { |
798 | tx_bytes = uio->uio_resid; | |
428870ff BB |
799 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
800 | uio, nbytes, tx); | |
9babb374 BB |
801 | tx_bytes -= uio->uio_resid; |
802 | } else { | |
803 | tx_bytes = nbytes; | |
428870ff BB |
804 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
805 | /* | |
806 | * If this is not a full block write, but we are | |
807 | * extending the file past EOF and this data starts | |
808 | * block-aligned, use assign_arcbuf(). Otherwise, | |
809 | * write via dmu_write(). | |
810 | */ | |
811 | if (tx_bytes < max_blksz && (!write_eof || | |
812 | aiov->iov_base != abuf->b_data)) { | |
813 | ASSERT(xuio); | |
3558fd73 | 814 | dmu_write(zsb->z_os, zp->z_id, woff, |
428870ff BB |
815 | aiov->iov_len, aiov->iov_base, tx); |
816 | dmu_return_arcbuf(abuf); | |
817 | xuio_stat_wbuf_copied(); | |
818 | } else { | |
819 | ASSERT(xuio || tx_bytes == max_blksz); | |
820 | dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl), | |
821 | woff, abuf, tx); | |
822 | } | |
9babb374 BB |
823 | ASSERT(tx_bytes <= uio->uio_resid); |
824 | uioskip(uio, tx_bytes); | |
825 | } | |
c0d35759 BB |
826 | |
827 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) | |
828 | update_pages(ip, woff, tx_bytes, zsb->z_os, zp->z_id); | |
34dc7c2f BB |
829 | |
830 | /* | |
831 | * If we made no progress, we're done. If we made even | |
832 | * partial progress, update the znode and ZIL accordingly. | |
833 | */ | |
834 | if (tx_bytes == 0) { | |
3558fd73 | 835 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 836 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
837 | dmu_tx_commit(tx); |
838 | ASSERT(error != 0); | |
839 | break; | |
840 | } | |
841 | ||
842 | /* | |
843 | * Clear Set-UID/Set-GID bits on successful write if not | |
844 | * privileged and at least one of the excute bits is set. | |
845 | * | |
846 | * It would be nice to to this after all writes have | |
847 | * been done, but that would still expose the ISUID/ISGID | |
848 | * to another app after the partial write is committed. | |
849 | * | |
572e2857 BB |
850 | * Note: we don't call zfs_fuid_map_id() here because |
851 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
852 | */ |
853 | mutex_enter(&zp->z_acl_lock); | |
428870ff | 854 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 855 | (S_IXUSR >> 6))) != 0 && |
428870ff | 856 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 857 | secpolicy_vnode_setid_retain(cr, |
428870ff BB |
858 | (zp->z_mode & S_ISUID) != 0 && zp->z_uid == 0) != 0) { |
859 | uint64_t newmode; | |
860 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
861 | newmode = zp->z_mode; | |
3558fd73 | 862 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zsb), |
428870ff | 863 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
864 | } |
865 | mutex_exit(&zp->z_acl_lock); | |
866 | ||
428870ff BB |
867 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, |
868 | B_TRUE); | |
34dc7c2f BB |
869 | |
870 | /* | |
871 | * Update the file size (zp_size) if it has changed; | |
872 | * account for possible concurrent updates. | |
873 | */ | |
428870ff BB |
874 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
875 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 876 | uio->uio_loffset); |
428870ff BB |
877 | ASSERT(error == 0); |
878 | } | |
572e2857 BB |
879 | /* |
880 | * If we are replaying and eof is non zero then force | |
881 | * the file size to the specified eof. Note, there's no | |
882 | * concurrency during replay. | |
883 | */ | |
3558fd73 BB |
884 | if (zsb->z_replay && zsb->z_replay_eof != 0) |
885 | zp->z_size = zsb->z_replay_eof; | |
572e2857 | 886 | |
428870ff BB |
887 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
888 | ||
a08ee875 LG |
889 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag, |
890 | NULL, NULL); | |
34dc7c2f BB |
891 | dmu_tx_commit(tx); |
892 | ||
893 | if (error != 0) | |
894 | break; | |
895 | ASSERT(tx_bytes == nbytes); | |
896 | n -= nbytes; | |
572e2857 BB |
897 | |
898 | if (!xuio && n > 0) | |
899 | uio_prefaultpages(MIN(n, max_blksz), uio); | |
34dc7c2f BB |
900 | } |
901 | ||
e10b0808 | 902 | zfs_inode_update(zp); |
34dc7c2f BB |
903 | zfs_range_unlock(rl); |
904 | ||
905 | /* | |
906 | * If we're in replay mode, or we made no progress, return error. | |
907 | * Otherwise, it's at least a partial write, so it's successful. | |
908 | */ | |
3558fd73 BB |
909 | if (zsb->z_replay || uio->uio_resid == start_resid) { |
910 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
911 | return (error); |
912 | } | |
913 | ||
428870ff | 914 | if (ioflag & (FSYNC | FDSYNC) || |
3558fd73 | 915 | zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 916 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 917 | |
3558fd73 | 918 | ZFS_EXIT(zsb); |
34dc7c2f BB |
919 | return (0); |
920 | } | |
e5c39b95 | 921 | EXPORT_SYMBOL(zfs_write); |
34dc7c2f | 922 | |
ea04106b AX |
923 | void |
924 | zfs_iput_async(struct inode *ip) | |
3558fd73 | 925 | { |
ea04106b AX |
926 | objset_t *os = ITOZSB(ip)->z_os; |
927 | ||
3558fd73 | 928 | ASSERT(atomic_read(&ip->i_count) > 0); |
ea04106b AX |
929 | ASSERT(os != NULL); |
930 | ||
3558fd73 | 931 | if (atomic_read(&ip->i_count) == 1) |
ea04106b AX |
932 | taskq_dispatch(dsl_pool_iput_taskq(dmu_objset_pool(os)), |
933 | (task_func_t *)iput, ip, TQ_SLEEP); | |
3558fd73 BB |
934 | else |
935 | iput(ip); | |
936 | } | |
937 | ||
34dc7c2f | 938 | void |
428870ff | 939 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 940 | { |
428870ff | 941 | znode_t *zp = zgd->zgd_private; |
428870ff BB |
942 | |
943 | if (zgd->zgd_db) | |
944 | dmu_buf_rele(zgd->zgd_db, zgd); | |
945 | ||
946 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f | 947 | |
9babb374 BB |
948 | /* |
949 | * Release the vnode asynchronously as we currently have the | |
950 | * txg stopped from syncing. | |
951 | */ | |
ea04106b | 952 | zfs_iput_async(ZTOI(zp)); |
428870ff BB |
953 | |
954 | if (error == 0 && zgd->zgd_bp) | |
955 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
956 | ||
34dc7c2f BB |
957 | kmem_free(zgd, sizeof (zgd_t)); |
958 | } | |
959 | ||
45d1cae3 BB |
960 | #ifdef DEBUG |
961 | static int zil_fault_io = 0; | |
962 | #endif | |
963 | ||
34dc7c2f BB |
964 | /* |
965 | * Get data to generate a TX_WRITE intent log record. | |
966 | */ | |
967 | int | |
968 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
969 | { | |
3558fd73 BB |
970 | zfs_sb_t *zsb = arg; |
971 | objset_t *os = zsb->z_os; | |
34dc7c2f | 972 | znode_t *zp; |
428870ff BB |
973 | uint64_t object = lr->lr_foid; |
974 | uint64_t offset = lr->lr_offset; | |
975 | uint64_t size = lr->lr_length; | |
976 | blkptr_t *bp = &lr->lr_blkptr; | |
34dc7c2f | 977 | dmu_buf_t *db; |
34dc7c2f | 978 | zgd_t *zgd; |
34dc7c2f BB |
979 | int error = 0; |
980 | ||
428870ff BB |
981 | ASSERT(zio != NULL); |
982 | ASSERT(size != 0); | |
34dc7c2f BB |
983 | |
984 | /* | |
985 | * Nothing to do if the file has been removed | |
986 | */ | |
3558fd73 | 987 | if (zfs_zget(zsb, object, &zp) != 0) |
a08ee875 | 988 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 989 | if (zp->z_unlinked) { |
9babb374 BB |
990 | /* |
991 | * Release the vnode asynchronously as we currently have the | |
992 | * txg stopped from syncing. | |
993 | */ | |
ea04106b | 994 | zfs_iput_async(ZTOI(zp)); |
a08ee875 | 995 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
996 | } |
997 | ||
ea04106b | 998 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
3558fd73 | 999 | zgd->zgd_zilog = zsb->z_log; |
428870ff BB |
1000 | zgd->zgd_private = zp; |
1001 | ||
34dc7c2f BB |
1002 | /* |
1003 | * Write records come in two flavors: immediate and indirect. | |
1004 | * For small writes it's cheaper to store the data with the | |
1005 | * log record (immediate); for large writes it's cheaper to | |
1006 | * sync the data and get a pointer to it (indirect) so that | |
1007 | * we don't have to write the data twice. | |
1008 | */ | |
1009 | if (buf != NULL) { /* immediate write */ | |
428870ff | 1010 | zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER); |
34dc7c2f | 1011 | /* test for truncation needs to be done while range locked */ |
428870ff | 1012 | if (offset >= zp->z_size) { |
a08ee875 | 1013 | error = SET_ERROR(ENOENT); |
428870ff BB |
1014 | } else { |
1015 | error = dmu_read(os, object, offset, size, buf, | |
1016 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1017 | } |
428870ff | 1018 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1019 | } else { /* indirect write */ |
34dc7c2f BB |
1020 | /* |
1021 | * Have to lock the whole block to ensure when it's | |
1022 | * written out and it's checksum is being calculated | |
1023 | * that no one can change the data. We need to re-check | |
1024 | * blocksize after we get the lock in case it's changed! | |
1025 | */ | |
1026 | for (;;) { | |
428870ff BB |
1027 | uint64_t blkoff; |
1028 | size = zp->z_blksz; | |
1029 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1030 | offset -= blkoff; | |
1031 | zgd->zgd_rl = zfs_range_lock(zp, offset, size, | |
1032 | RL_READER); | |
1033 | if (zp->z_blksz == size) | |
34dc7c2f | 1034 | break; |
428870ff BB |
1035 | offset += blkoff; |
1036 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f BB |
1037 | } |
1038 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1039 | if (lr->lr_offset >= zp->z_size) |
a08ee875 | 1040 | error = SET_ERROR(ENOENT); |
45d1cae3 BB |
1041 | #ifdef DEBUG |
1042 | if (zil_fault_io) { | |
a08ee875 | 1043 | error = SET_ERROR(EIO); |
45d1cae3 | 1044 | zil_fault_io = 0; |
45d1cae3 | 1045 | } |
45d1cae3 | 1046 | #endif |
34dc7c2f | 1047 | if (error == 0) |
428870ff BB |
1048 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1049 | DMU_READ_NO_PREFETCH); | |
1050 | ||
1051 | if (error == 0) { | |
a08ee875 LG |
1052 | blkptr_t *obp = dmu_buf_get_blkptr(db); |
1053 | if (obp) { | |
1054 | ASSERT(BP_IS_HOLE(bp)); | |
1055 | *bp = *obp; | |
1056 | } | |
1057 | ||
428870ff BB |
1058 | zgd->zgd_db = db; |
1059 | zgd->zgd_bp = bp; | |
1060 | ||
1061 | ASSERT(db->db_offset == offset); | |
1062 | ASSERT(db->db_size == size); | |
1063 | ||
1064 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1065 | zfs_get_done, zgd); | |
1066 | ASSERT(error || lr->lr_length <= zp->z_blksz); | |
1067 | ||
1068 | /* | |
1069 | * On success, we need to wait for the write I/O | |
1070 | * initiated by dmu_sync() to complete before we can | |
1071 | * release this dbuf. We will finish everything up | |
1072 | * in the zfs_get_done() callback. | |
1073 | */ | |
1074 | if (error == 0) | |
1075 | return (0); | |
1076 | ||
1077 | if (error == EALREADY) { | |
1078 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
1079 | error = 0; | |
1080 | } | |
1081 | } | |
34dc7c2f | 1082 | } |
428870ff BB |
1083 | |
1084 | zfs_get_done(zgd, error); | |
1085 | ||
34dc7c2f BB |
1086 | return (error); |
1087 | } | |
1088 | ||
1089 | /*ARGSUSED*/ | |
3558fd73 BB |
1090 | int |
1091 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1092 | { |
3558fd73 BB |
1093 | znode_t *zp = ITOZ(ip); |
1094 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
1095 | int error; |
1096 | ||
3558fd73 | 1097 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1098 | ZFS_VERIFY_ZP(zp); |
1099 | ||
1100 | if (flag & V_ACE_MASK) | |
1101 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1102 | else | |
1103 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1104 | ||
3558fd73 | 1105 | ZFS_EXIT(zsb); |
45d1cae3 BB |
1106 | return (error); |
1107 | } | |
3558fd73 | 1108 | EXPORT_SYMBOL(zfs_access); |
45d1cae3 | 1109 | |
34dc7c2f BB |
1110 | /* |
1111 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1112 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1113 | * |
3558fd73 | 1114 | * IN: dip - inode of directory to search. |
34dc7c2f | 1115 | * nm - name of entry to lookup. |
34dc7c2f | 1116 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1117 | * cr - credentials of caller. |
34dc7c2f BB |
1118 | * direntflags - directory lookup flags |
1119 | * realpnp - returned pathname. | |
1120 | * | |
3558fd73 | 1121 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f | 1122 | * |
a08ee875 | 1123 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1124 | * |
1125 | * Timestamps: | |
1126 | * NA | |
1127 | */ | |
1128 | /* ARGSUSED */ | |
e5c39b95 | 1129 | int |
3558fd73 BB |
1130 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1131 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1132 | { |
3558fd73 BB |
1133 | znode_t *zdp = ITOZ(dip); |
1134 | zfs_sb_t *zsb = ITOZSB(dip); | |
1135 | int error = 0; | |
45d1cae3 BB |
1136 | |
1137 | /* fast path */ | |
1138 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { | |
1139 | ||
3558fd73 | 1140 | if (!S_ISDIR(dip->i_mode)) { |
a08ee875 | 1141 | return (SET_ERROR(ENOTDIR)); |
428870ff | 1142 | } else if (zdp->z_sa_hdl == NULL) { |
a08ee875 | 1143 | return (SET_ERROR(EIO)); |
45d1cae3 BB |
1144 | } |
1145 | ||
1146 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1147 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1148 | if (!error) { | |
3558fd73 BB |
1149 | *ipp = dip; |
1150 | igrab(*ipp); | |
45d1cae3 BB |
1151 | return (0); |
1152 | } | |
1153 | return (error); | |
3558fd73 | 1154 | #ifdef HAVE_DNLC |
45d1cae3 BB |
1155 | } else { |
1156 | vnode_t *tvp = dnlc_lookup(dvp, nm); | |
1157 | ||
1158 | if (tvp) { | |
1159 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1160 | if (error) { | |
3558fd73 | 1161 | iput(tvp); |
45d1cae3 BB |
1162 | return (error); |
1163 | } | |
1164 | if (tvp == DNLC_NO_VNODE) { | |
3558fd73 | 1165 | iput(tvp); |
a08ee875 | 1166 | return (SET_ERROR(ENOENT)); |
45d1cae3 BB |
1167 | } else { |
1168 | *vpp = tvp; | |
1169 | return (specvp_check(vpp, cr)); | |
1170 | } | |
1171 | } | |
3558fd73 | 1172 | #endif /* HAVE_DNLC */ |
45d1cae3 BB |
1173 | } |
1174 | } | |
1175 | ||
3558fd73 | 1176 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1177 | ZFS_VERIFY_ZP(zdp); |
1178 | ||
3558fd73 | 1179 | *ipp = NULL; |
34dc7c2f BB |
1180 | |
1181 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1182 | /* |
1183 | * We don't allow recursive attributes.. | |
1184 | * Maybe someday we will. | |
1185 | */ | |
428870ff | 1186 | if (zdp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1187 | ZFS_EXIT(zsb); |
a08ee875 | 1188 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1189 | } |
1190 | ||
3558fd73 BB |
1191 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
1192 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1193 | return (error); |
1194 | } | |
1195 | ||
1196 | /* | |
1197 | * Do we have permission to get into attribute directory? | |
1198 | */ | |
1199 | ||
3558fd73 | 1200 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1201 | B_FALSE, cr))) { |
3558fd73 BB |
1202 | iput(*ipp); |
1203 | *ipp = NULL; | |
34dc7c2f BB |
1204 | } |
1205 | ||
3558fd73 | 1206 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1207 | return (error); |
1208 | } | |
1209 | ||
3558fd73 BB |
1210 | if (!S_ISDIR(dip->i_mode)) { |
1211 | ZFS_EXIT(zsb); | |
a08ee875 | 1212 | return (SET_ERROR(ENOTDIR)); |
34dc7c2f BB |
1213 | } |
1214 | ||
1215 | /* | |
1216 | * Check accessibility of directory. | |
1217 | */ | |
1218 | ||
149e873a | 1219 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
3558fd73 | 1220 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1221 | return (error); |
1222 | } | |
1223 | ||
3558fd73 | 1224 | if (zsb->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1225 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1226 | ZFS_EXIT(zsb); |
a08ee875 | 1227 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1228 | } |
1229 | ||
3558fd73 BB |
1230 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1231 | if ((error == 0) && (*ipp)) | |
1232 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1233 | |
3558fd73 | 1234 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1235 | return (error); |
1236 | } | |
e5c39b95 | 1237 | EXPORT_SYMBOL(zfs_lookup); |
34dc7c2f BB |
1238 | |
1239 | /* | |
1240 | * Attempt to create a new entry in a directory. If the entry | |
1241 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1242 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1243 | * |
3558fd73 | 1244 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1245 | * name - name of new file entry. |
1246 | * vap - attributes of new file. | |
1247 | * excl - flag indicating exclusive or non-exclusive mode. | |
1248 | * mode - mode to open file with. | |
1249 | * cr - credentials of caller. | |
1250 | * flag - large file flag [UNUSED]. | |
3558fd73 | 1251 | * vsecp - ACL to be set |
34dc7c2f | 1252 | * |
3558fd73 | 1253 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f | 1254 | * |
a08ee875 | 1255 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1256 | * |
1257 | * Timestamps: | |
3558fd73 BB |
1258 | * dip - ctime|mtime updated if new entry created |
1259 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1260 | */ |
1261 | ||
1262 | /* ARGSUSED */ | |
e5c39b95 | 1263 | int |
3558fd73 BB |
1264 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1265 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1266 | { |
3558fd73 BB |
1267 | znode_t *zp, *dzp = ITOZ(dip); |
1268 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1269 | zilog_t *zilog; |
1270 | objset_t *os; | |
1271 | zfs_dirlock_t *dl; | |
1272 | dmu_tx_t *tx; | |
1273 | int error; | |
b128c09f | 1274 | uid_t uid; |
149e873a | 1275 | gid_t gid; |
428870ff | 1276 | zfs_acl_ids_t acl_ids; |
9babb374 | 1277 | boolean_t fuid_dirtied; |
428870ff | 1278 | boolean_t have_acl = B_FALSE; |
a08ee875 | 1279 | boolean_t waited = B_FALSE; |
34dc7c2f BB |
1280 | |
1281 | /* | |
1282 | * If we have an ephemeral id, ACL, or XVATTR then | |
1283 | * make sure file system is at proper version | |
1284 | */ | |
1285 | ||
149e873a | 1286 | gid = crgetgid(cr); |
3558fd73 | 1287 | uid = crgetuid(cr); |
b128c09f | 1288 | |
3558fd73 BB |
1289 | if (zsb->z_use_fuids == B_FALSE && |
1290 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
a08ee875 | 1291 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1292 | |
3558fd73 | 1293 | ZFS_ENTER(zsb); |
34dc7c2f | 1294 | ZFS_VERIFY_ZP(dzp); |
3558fd73 BB |
1295 | os = zsb->z_os; |
1296 | zilog = zsb->z_log; | |
34dc7c2f | 1297 | |
3558fd73 | 1298 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1299 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1300 | ZFS_EXIT(zsb); |
a08ee875 | 1301 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1302 | } |
1303 | ||
5484965a | 1304 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1305 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1306 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1307 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1308 | return (error); |
1309 | } | |
1310 | } | |
34dc7c2f | 1311 | |
3558fd73 BB |
1312 | top: |
1313 | *ipp = NULL; | |
34dc7c2f BB |
1314 | if (*name == '\0') { |
1315 | /* | |
1316 | * Null component name refers to the directory itself. | |
1317 | */ | |
3558fd73 | 1318 | igrab(dip); |
34dc7c2f BB |
1319 | zp = dzp; |
1320 | dl = NULL; | |
1321 | error = 0; | |
1322 | } else { | |
3558fd73 | 1323 | /* possible igrab(zp) */ |
34dc7c2f BB |
1324 | int zflg = 0; |
1325 | ||
1326 | if (flag & FIGNORECASE) | |
1327 | zflg |= ZCILOOK; | |
1328 | ||
1329 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1330 | NULL, NULL); | |
1331 | if (error) { | |
572e2857 BB |
1332 | if (have_acl) |
1333 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f | 1334 | if (strcmp(name, "..") == 0) |
a08ee875 | 1335 | error = SET_ERROR(EISDIR); |
3558fd73 | 1336 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1337 | return (error); |
1338 | } | |
1339 | } | |
428870ff | 1340 | |
34dc7c2f BB |
1341 | if (zp == NULL) { |
1342 | uint64_t txtype; | |
1343 | ||
1344 | /* | |
1345 | * Create a new file object and update the directory | |
1346 | * to reference it. | |
1347 | */ | |
149e873a | 1348 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1349 | if (have_acl) |
1350 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1351 | goto out; |
1352 | } | |
1353 | ||
1354 | /* | |
1355 | * We only support the creation of regular files in | |
1356 | * extended attribute directories. | |
1357 | */ | |
428870ff | 1358 | |
3558fd73 | 1359 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1360 | if (have_acl) |
1361 | zfs_acl_ids_free(&acl_ids); | |
a08ee875 | 1362 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1363 | goto out; |
1364 | } | |
1365 | ||
428870ff BB |
1366 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1367 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1368 | goto out; |
428870ff BB |
1369 | have_acl = B_TRUE; |
1370 | ||
3558fd73 | 1371 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1372 | zfs_acl_ids_free(&acl_ids); |
a08ee875 | 1373 | error = SET_ERROR(EDQUOT); |
9babb374 BB |
1374 | goto out; |
1375 | } | |
1376 | ||
34dc7c2f | 1377 | tx = dmu_tx_create(os); |
428870ff BB |
1378 | |
1379 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1380 | ZFS_SA_BASE_ATTR_SIZE); | |
1381 | ||
3558fd73 | 1382 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1383 | if (fuid_dirtied) |
3558fd73 | 1384 | zfs_fuid_txhold(zsb, tx); |
34dc7c2f | 1385 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1386 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
3558fd73 | 1387 | if (!zsb->z_use_sa && |
428870ff | 1388 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1389 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1390 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1391 | } |
a08ee875 | 1392 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1393 | if (error) { |
1394 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1395 | if (error == ERESTART) { |
a08ee875 | 1396 | waited = B_TRUE; |
34dc7c2f BB |
1397 | dmu_tx_wait(tx); |
1398 | dmu_tx_abort(tx); | |
1399 | goto top; | |
1400 | } | |
428870ff | 1401 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1402 | dmu_tx_abort(tx); |
3558fd73 | 1403 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1404 | return (error); |
1405 | } | |
428870ff | 1406 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 BB |
1407 | |
1408 | if (fuid_dirtied) | |
3558fd73 | 1409 | zfs_fuid_sync(zsb, tx); |
9babb374 | 1410 | |
34dc7c2f BB |
1411 | (void) zfs_link_create(dl, zp, tx, ZNEW); |
1412 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); | |
1413 | if (flag & FIGNORECASE) | |
1414 | txtype |= TX_CI; | |
1415 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1416 | vsecp, acl_ids.z_fuidp, vap); |
1417 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1418 | dmu_tx_commit(tx); |
1419 | } else { | |
1420 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1421 | ||
572e2857 BB |
1422 | if (have_acl) |
1423 | zfs_acl_ids_free(&acl_ids); | |
1424 | have_acl = B_FALSE; | |
1425 | ||
34dc7c2f BB |
1426 | /* |
1427 | * A directory entry already exists for this name. | |
1428 | */ | |
1429 | /* | |
1430 | * Can't truncate an existing file if in exclusive mode. | |
1431 | */ | |
3558fd73 | 1432 | if (excl) { |
a08ee875 | 1433 | error = SET_ERROR(EEXIST); |
34dc7c2f BB |
1434 | goto out; |
1435 | } | |
1436 | /* | |
1437 | * Can't open a directory for writing. | |
1438 | */ | |
3558fd73 | 1439 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
a08ee875 | 1440 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
1441 | goto out; |
1442 | } | |
1443 | /* | |
1444 | * Verify requested access to file. | |
1445 | */ | |
1446 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1447 | goto out; | |
1448 | } | |
1449 | ||
1450 | mutex_enter(&dzp->z_lock); | |
1451 | dzp->z_seq++; | |
1452 | mutex_exit(&dzp->z_lock); | |
1453 | ||
1454 | /* | |
1455 | * Truncate regular files if requested. | |
1456 | */ | |
3558fd73 BB |
1457 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1458 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f BB |
1459 | /* we can't hold any locks when calling zfs_freesp() */ |
1460 | zfs_dirent_unlock(dl); | |
1461 | dl = NULL; | |
34dc7c2f | 1462 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1463 | } |
1464 | } | |
1465 | out: | |
1466 | ||
1467 | if (dl) | |
1468 | zfs_dirent_unlock(dl); | |
1469 | ||
1470 | if (error) { | |
1471 | if (zp) | |
3558fd73 | 1472 | iput(ZTOI(zp)); |
34dc7c2f | 1473 | } else { |
960e08fe BB |
1474 | zfs_inode_update(dzp); |
1475 | zfs_inode_update(zp); | |
3558fd73 | 1476 | *ipp = ZTOI(zp); |
34dc7c2f | 1477 | } |
34dc7c2f | 1478 | |
3558fd73 | 1479 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1480 | zil_commit(zilog, 0); |
428870ff | 1481 | |
3558fd73 | 1482 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1483 | return (error); |
1484 | } | |
e5c39b95 | 1485 | EXPORT_SYMBOL(zfs_create); |
34dc7c2f BB |
1486 | |
1487 | /* | |
1488 | * Remove an entry from a directory. | |
1489 | * | |
3558fd73 | 1490 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1491 | * name - name of entry to remove. |
1492 | * cr - credentials of caller. | |
34dc7c2f BB |
1493 | * |
1494 | * RETURN: 0 if success | |
1495 | * error code if failure | |
1496 | * | |
1497 | * Timestamps: | |
3558fd73 BB |
1498 | * dip - ctime|mtime |
1499 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1500 | */ |
428870ff BB |
1501 | |
1502 | uint64_t null_xattr = 0; | |
1503 | ||
34dc7c2f | 1504 | /*ARGSUSED*/ |
e5c39b95 | 1505 | int |
3558fd73 | 1506 | zfs_remove(struct inode *dip, char *name, cred_t *cr) |
34dc7c2f | 1507 | { |
3558fd73 | 1508 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1509 | znode_t *xzp; |
3558fd73 BB |
1510 | struct inode *ip; |
1511 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f | 1512 | zilog_t *zilog; |
3558fd73 BB |
1513 | uint64_t xattr_obj; |
1514 | uint64_t xattr_obj_unlinked = 0; | |
572e2857 | 1515 | uint64_t obj = 0; |
34dc7c2f BB |
1516 | zfs_dirlock_t *dl; |
1517 | dmu_tx_t *tx; | |
3558fd73 | 1518 | boolean_t unlinked; |
34dc7c2f BB |
1519 | uint64_t txtype; |
1520 | pathname_t *realnmp = NULL; | |
3558fd73 | 1521 | #ifdef HAVE_PN_UTILS |
34dc7c2f | 1522 | pathname_t realnm; |
3558fd73 | 1523 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
1524 | int error; |
1525 | int zflg = ZEXISTS; | |
a08ee875 | 1526 | boolean_t waited = B_FALSE; |
34dc7c2f | 1527 | |
3558fd73 | 1528 | ZFS_ENTER(zsb); |
34dc7c2f | 1529 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1530 | zilog = zsb->z_log; |
34dc7c2f | 1531 | |
3558fd73 | 1532 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1533 | if (flags & FIGNORECASE) { |
1534 | zflg |= ZCILOOK; | |
1535 | pn_alloc(&realnm); | |
1536 | realnmp = &realnm; | |
1537 | } | |
3558fd73 | 1538 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
1539 | |
1540 | top: | |
572e2857 BB |
1541 | xattr_obj = 0; |
1542 | xzp = NULL; | |
34dc7c2f BB |
1543 | /* |
1544 | * Attempt to lock directory; fail if entry doesn't exist. | |
1545 | */ | |
149e873a BB |
1546 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1547 | NULL, realnmp))) { | |
3558fd73 | 1548 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1549 | if (realnmp) |
1550 | pn_free(realnmp); | |
3558fd73 BB |
1551 | #endif /* HAVE_PN_UTILS */ |
1552 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1553 | return (error); |
1554 | } | |
1555 | ||
3558fd73 | 1556 | ip = ZTOI(zp); |
34dc7c2f | 1557 | |
149e873a | 1558 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1559 | goto out; |
1560 | } | |
1561 | ||
1562 | /* | |
1563 | * Need to use rmdir for removing directories. | |
1564 | */ | |
3558fd73 | 1565 | if (S_ISDIR(ip->i_mode)) { |
a08ee875 | 1566 | error = SET_ERROR(EPERM); |
34dc7c2f BB |
1567 | goto out; |
1568 | } | |
1569 | ||
3558fd73 | 1570 | #ifdef HAVE_DNLC |
34dc7c2f BB |
1571 | if (realnmp) |
1572 | dnlc_remove(dvp, realnmp->pn_buf); | |
1573 | else | |
1574 | dnlc_remove(dvp, name); | |
3558fd73 | 1575 | #endif /* HAVE_DNLC */ |
34dc7c2f BB |
1576 | |
1577 | /* | |
3558fd73 BB |
1578 | * We never delete the znode and always place it in the unlinked |
1579 | * set. The dentry cache will always hold the last reference and | |
1580 | * is responsible for safely freeing the znode. | |
34dc7c2f | 1581 | */ |
572e2857 | 1582 | obj = zp->z_id; |
3558fd73 | 1583 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 1584 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1585 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1586 | zfs_sa_upgrade_txholds(tx, zp); | |
1587 | zfs_sa_upgrade_txholds(tx, dzp); | |
34dc7c2f BB |
1588 | |
1589 | /* are there any extended attributes? */ | |
3558fd73 | 1590 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1591 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1592 | if (error == 0 && xattr_obj) { |
3558fd73 | 1593 | error = zfs_zget(zsb, xattr_obj, &xzp); |
c06d4368 | 1594 | ASSERT0(error); |
428870ff BB |
1595 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
1596 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1597 | } |
1598 | ||
34dc7c2f | 1599 | /* charge as an update -- would be nice not to charge at all */ |
3558fd73 | 1600 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1601 | |
a08ee875 | 1602 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1603 | if (error) { |
1604 | zfs_dirent_unlock(dl); | |
3558fd73 | 1605 | iput(ip); |
572e2857 | 1606 | if (xzp) |
3558fd73 | 1607 | iput(ZTOI(xzp)); |
fb5f0bc8 | 1608 | if (error == ERESTART) { |
a08ee875 | 1609 | waited = B_TRUE; |
34dc7c2f BB |
1610 | dmu_tx_wait(tx); |
1611 | dmu_tx_abort(tx); | |
1612 | goto top; | |
1613 | } | |
3558fd73 | 1614 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1615 | if (realnmp) |
1616 | pn_free(realnmp); | |
3558fd73 | 1617 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f | 1618 | dmu_tx_abort(tx); |
3558fd73 | 1619 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1620 | return (error); |
1621 | } | |
1622 | ||
1623 | /* | |
1624 | * Remove the directory entry. | |
1625 | */ | |
1626 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1627 | ||
1628 | if (error) { | |
1629 | dmu_tx_commit(tx); | |
1630 | goto out; | |
1631 | } | |
1632 | ||
1633 | if (unlinked) { | |
572e2857 BB |
1634 | /* |
1635 | * Hold z_lock so that we can make sure that the ACL obj | |
1636 | * hasn't changed. Could have been deleted due to | |
1637 | * zfs_sa_upgrade(). | |
1638 | */ | |
1639 | mutex_enter(&zp->z_lock); | |
3558fd73 | 1640 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1641 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
572e2857 | 1642 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1643 | zfs_unlinked_add(zp, tx); |
1644 | } | |
1645 | ||
1646 | txtype = TX_REMOVE; | |
3558fd73 | 1647 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1648 | if (flags & FIGNORECASE) |
1649 | txtype |= TX_CI; | |
3558fd73 | 1650 | #endif /* HAVE_PN_UTILS */ |
572e2857 | 1651 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1652 | |
1653 | dmu_tx_commit(tx); | |
1654 | out: | |
3558fd73 | 1655 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1656 | if (realnmp) |
1657 | pn_free(realnmp); | |
3558fd73 | 1658 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
1659 | |
1660 | zfs_dirent_unlock(dl); | |
960e08fe BB |
1661 | zfs_inode_update(dzp); |
1662 | zfs_inode_update(zp); | |
1663 | if (xzp) | |
1664 | zfs_inode_update(xzp); | |
34dc7c2f | 1665 | |
3558fd73 | 1666 | iput(ip); |
428870ff | 1667 | if (xzp) |
3558fd73 | 1668 | iput(ZTOI(xzp)); |
428870ff | 1669 | |
3558fd73 | 1670 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1671 | zil_commit(zilog, 0); |
34dc7c2f | 1672 | |
3558fd73 | 1673 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1674 | return (error); |
1675 | } | |
e5c39b95 | 1676 | EXPORT_SYMBOL(zfs_remove); |
34dc7c2f BB |
1677 | |
1678 | /* | |
3558fd73 | 1679 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1680 | * provided. Return a pointer to the inserted directory. |
1681 | * | |
3558fd73 | 1682 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1683 | * dirname - name of new directory. |
1684 | * vap - attributes of new directory. | |
1685 | * cr - credentials of caller. | |
34dc7c2f BB |
1686 | * vsecp - ACL to be set |
1687 | * | |
3558fd73 | 1688 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1689 | * |
1690 | * RETURN: 0 if success | |
1691 | * error code if failure | |
1692 | * | |
1693 | * Timestamps: | |
3558fd73 BB |
1694 | * dip - ctime|mtime updated |
1695 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1696 | */ |
1697 | /*ARGSUSED*/ | |
e5c39b95 | 1698 | int |
3558fd73 BB |
1699 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1700 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1701 | { |
3558fd73 BB |
1702 | znode_t *zp, *dzp = ITOZ(dip); |
1703 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1704 | zilog_t *zilog; |
1705 | zfs_dirlock_t *dl; | |
1706 | uint64_t txtype; | |
1707 | dmu_tx_t *tx; | |
1708 | int error; | |
34dc7c2f | 1709 | int zf = ZNEW; |
b128c09f BB |
1710 | uid_t uid; |
1711 | gid_t gid = crgetgid(cr); | |
428870ff | 1712 | zfs_acl_ids_t acl_ids; |
9babb374 | 1713 | boolean_t fuid_dirtied; |
a08ee875 | 1714 | boolean_t waited = B_FALSE; |
34dc7c2f | 1715 | |
3558fd73 | 1716 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1717 | |
1718 | /* | |
1719 | * If we have an ephemeral id, ACL, or XVATTR then | |
1720 | * make sure file system is at proper version | |
1721 | */ | |
1722 | ||
3558fd73 BB |
1723 | uid = crgetuid(cr); |
1724 | if (zsb->z_use_fuids == B_FALSE && | |
1725 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
a08ee875 | 1726 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1727 | |
3558fd73 | 1728 | ZFS_ENTER(zsb); |
34dc7c2f | 1729 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1730 | zilog = zsb->z_log; |
34dc7c2f | 1731 | |
428870ff | 1732 | if (dzp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1733 | ZFS_EXIT(zsb); |
a08ee875 | 1734 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1735 | } |
1736 | ||
3558fd73 | 1737 | if (zsb->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1738 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1739 | ZFS_EXIT(zsb); |
a08ee875 | 1740 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
1741 | } |
1742 | if (flags & FIGNORECASE) | |
1743 | zf |= ZCILOOK; | |
1744 | ||
5484965a | 1745 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1746 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1747 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1748 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1749 | return (error); |
1750 | } | |
428870ff | 1751 | } |
34dc7c2f | 1752 | |
428870ff BB |
1753 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1754 | vsecp, &acl_ids)) != 0) { | |
3558fd73 | 1755 | ZFS_EXIT(zsb); |
428870ff BB |
1756 | return (error); |
1757 | } | |
34dc7c2f BB |
1758 | /* |
1759 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1760 | * |
1761 | * Existence is checked first to make sure we don't return | |
1762 | * EACCES instead of EEXIST which can cause some applications | |
1763 | * to fail. | |
34dc7c2f BB |
1764 | */ |
1765 | top: | |
3558fd73 | 1766 | *ipp = NULL; |
34dc7c2f | 1767 | |
149e873a BB |
1768 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1769 | NULL, NULL))) { | |
428870ff | 1770 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 1771 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1772 | return (error); |
1773 | } | |
1774 | ||
149e873a | 1775 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 1776 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1777 | zfs_dirent_unlock(dl); |
3558fd73 | 1778 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1779 | return (error); |
1780 | } | |
1781 | ||
3558fd73 | 1782 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1783 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 1784 | zfs_dirent_unlock(dl); |
3558fd73 | 1785 | ZFS_EXIT(zsb); |
a08ee875 | 1786 | return (SET_ERROR(EDQUOT)); |
9babb374 BB |
1787 | } |
1788 | ||
34dc7c2f BB |
1789 | /* |
1790 | * Add a new entry to the directory. | |
1791 | */ | |
3558fd73 | 1792 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f BB |
1793 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
1794 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
3558fd73 | 1795 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1796 | if (fuid_dirtied) |
3558fd73 BB |
1797 | zfs_fuid_txhold(zsb, tx); |
1798 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
1799 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
1800 | acl_ids.z_aclp->z_acl_bytes); | |
1801 | } | |
1802 | ||
1803 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1804 | ZFS_SA_BASE_ATTR_SIZE); | |
1805 | ||
a08ee875 | 1806 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1807 | if (error) { |
1808 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1809 | if (error == ERESTART) { |
a08ee875 | 1810 | waited = B_TRUE; |
34dc7c2f BB |
1811 | dmu_tx_wait(tx); |
1812 | dmu_tx_abort(tx); | |
1813 | goto top; | |
1814 | } | |
428870ff | 1815 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1816 | dmu_tx_abort(tx); |
3558fd73 | 1817 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1818 | return (error); |
1819 | } | |
1820 | ||
1821 | /* | |
1822 | * Create new node. | |
1823 | */ | |
428870ff | 1824 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f | 1825 | |
9babb374 | 1826 | if (fuid_dirtied) |
3558fd73 | 1827 | zfs_fuid_sync(zsb, tx); |
428870ff | 1828 | |
34dc7c2f BB |
1829 | /* |
1830 | * Now put new name in parent dir. | |
1831 | */ | |
1832 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
1833 | ||
3558fd73 | 1834 | *ipp = ZTOI(zp); |
34dc7c2f BB |
1835 | |
1836 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
1837 | if (flags & FIGNORECASE) | |
1838 | txtype |= TX_CI; | |
9babb374 BB |
1839 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
1840 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 1841 | |
9babb374 | 1842 | zfs_acl_ids_free(&acl_ids); |
428870ff | 1843 | |
34dc7c2f BB |
1844 | dmu_tx_commit(tx); |
1845 | ||
1846 | zfs_dirent_unlock(dl); | |
1847 | ||
3558fd73 | 1848 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1849 | zil_commit(zilog, 0); |
428870ff | 1850 | |
960e08fe BB |
1851 | zfs_inode_update(dzp); |
1852 | zfs_inode_update(zp); | |
3558fd73 | 1853 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1854 | return (0); |
1855 | } | |
e5c39b95 | 1856 | EXPORT_SYMBOL(zfs_mkdir); |
34dc7c2f BB |
1857 | |
1858 | /* | |
1859 | * Remove a directory subdir entry. If the current working | |
1860 | * directory is the same as the subdir to be removed, the | |
1861 | * remove will fail. | |
1862 | * | |
3558fd73 | 1863 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 1864 | * name - name of directory to be removed. |
3558fd73 | 1865 | * cwd - inode of current working directory. |
34dc7c2f | 1866 | * cr - credentials of caller. |
34dc7c2f BB |
1867 | * flags - case flags |
1868 | * | |
a08ee875 | 1869 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
1870 | * |
1871 | * Timestamps: | |
3558fd73 | 1872 | * dip - ctime|mtime updated |
34dc7c2f BB |
1873 | */ |
1874 | /*ARGSUSED*/ | |
e5c39b95 | 1875 | int |
3558fd73 BB |
1876 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
1877 | int flags) | |
34dc7c2f | 1878 | { |
3558fd73 | 1879 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 1880 | znode_t *zp; |
3558fd73 BB |
1881 | struct inode *ip; |
1882 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1883 | zilog_t *zilog; |
1884 | zfs_dirlock_t *dl; | |
1885 | dmu_tx_t *tx; | |
1886 | int error; | |
1887 | int zflg = ZEXISTS; | |
a08ee875 | 1888 | boolean_t waited = B_FALSE; |
34dc7c2f | 1889 | |
3558fd73 | 1890 | ZFS_ENTER(zsb); |
34dc7c2f | 1891 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1892 | zilog = zsb->z_log; |
34dc7c2f BB |
1893 | |
1894 | if (flags & FIGNORECASE) | |
1895 | zflg |= ZCILOOK; | |
1896 | top: | |
1897 | zp = NULL; | |
1898 | ||
1899 | /* | |
1900 | * Attempt to lock directory; fail if entry doesn't exist. | |
1901 | */ | |
149e873a BB |
1902 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1903 | NULL, NULL))) { | |
3558fd73 | 1904 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1905 | return (error); |
1906 | } | |
1907 | ||
3558fd73 | 1908 | ip = ZTOI(zp); |
34dc7c2f | 1909 | |
149e873a | 1910 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1911 | goto out; |
1912 | } | |
1913 | ||
3558fd73 | 1914 | if (!S_ISDIR(ip->i_mode)) { |
a08ee875 | 1915 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
1916 | goto out; |
1917 | } | |
1918 | ||
3558fd73 | 1919 | if (ip == cwd) { |
a08ee875 | 1920 | error = SET_ERROR(EINVAL); |
34dc7c2f BB |
1921 | goto out; |
1922 | } | |
1923 | ||
34dc7c2f BB |
1924 | /* |
1925 | * Grab a lock on the directory to make sure that noone is | |
1926 | * trying to add (or lookup) entries while we are removing it. | |
1927 | */ | |
1928 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
1929 | ||
1930 | /* | |
1931 | * Grab a lock on the parent pointer to make sure we play well | |
1932 | * with the treewalk and directory rename code. | |
1933 | */ | |
1934 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
1935 | ||
3558fd73 | 1936 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 1937 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 1938 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3558fd73 | 1939 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
1940 | zfs_sa_upgrade_txholds(tx, zp); |
1941 | zfs_sa_upgrade_txholds(tx, dzp); | |
a08ee875 | 1942 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
1943 | if (error) { |
1944 | rw_exit(&zp->z_parent_lock); | |
1945 | rw_exit(&zp->z_name_lock); | |
1946 | zfs_dirent_unlock(dl); | |
3558fd73 | 1947 | iput(ip); |
fb5f0bc8 | 1948 | if (error == ERESTART) { |
a08ee875 | 1949 | waited = B_TRUE; |
34dc7c2f BB |
1950 | dmu_tx_wait(tx); |
1951 | dmu_tx_abort(tx); | |
1952 | goto top; | |
1953 | } | |
1954 | dmu_tx_abort(tx); | |
3558fd73 | 1955 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1956 | return (error); |
1957 | } | |
1958 | ||
1959 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
1960 | ||
1961 | if (error == 0) { | |
1962 | uint64_t txtype = TX_RMDIR; | |
1963 | if (flags & FIGNORECASE) | |
1964 | txtype |= TX_CI; | |
572e2857 | 1965 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
1966 | } |
1967 | ||
1968 | dmu_tx_commit(tx); | |
1969 | ||
1970 | rw_exit(&zp->z_parent_lock); | |
1971 | rw_exit(&zp->z_name_lock); | |
1972 | out: | |
1973 | zfs_dirent_unlock(dl); | |
1974 | ||
59157910 BB |
1975 | zfs_inode_update(dzp); |
1976 | zfs_inode_update(zp); | |
3558fd73 | 1977 | iput(ip); |
34dc7c2f | 1978 | |
3558fd73 | 1979 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1980 | zil_commit(zilog, 0); |
428870ff | 1981 | |
3558fd73 | 1982 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1983 | return (error); |
1984 | } | |
e5c39b95 | 1985 | EXPORT_SYMBOL(zfs_rmdir); |
34dc7c2f BB |
1986 | |
1987 | /* | |
1988 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 1989 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 1990 | * |
3558fd73 BB |
1991 | * IN: ip - inode of directory to read. |
1992 | * dirent - buffer for directory entries. | |
34dc7c2f | 1993 | * |
3558fd73 | 1994 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
1995 | * |
1996 | * RETURN: 0 if success | |
1997 | * error code if failure | |
1998 | * | |
1999 | * Timestamps: | |
3558fd73 | 2000 | * ip - atime updated |
34dc7c2f BB |
2001 | * |
2002 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2003 | * This allows us to use the low range for "special" directory entries: | |
2004 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2005 | * we use the offset 2 for the '.zfs' directory. | |
2006 | */ | |
2007 | /* ARGSUSED */ | |
3558fd73 | 2008 | int |
c06d4368 | 2009 | zfs_readdir(struct inode *ip, struct dir_context *ctx, cred_t *cr) |
34dc7c2f | 2010 | { |
3558fd73 BB |
2011 | znode_t *zp = ITOZ(ip); |
2012 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 2013 | objset_t *os; |
34dc7c2f BB |
2014 | zap_cursor_t zc; |
2015 | zap_attribute_t zap; | |
34dc7c2f BB |
2016 | int error; |
2017 | uint8_t prefetch; | |
a08ee875 | 2018 | uint8_t type; |
3558fd73 BB |
2019 | int done = 0; |
2020 | uint64_t parent; | |
a08ee875 | 2021 | uint64_t offset; /* must be unsigned; checks for < 1 */ |
34dc7c2f | 2022 | |
3558fd73 | 2023 | ZFS_ENTER(zsb); |
34dc7c2f BB |
2024 | ZFS_VERIFY_ZP(zp); |
2025 | ||
3558fd73 BB |
2026 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
2027 | &parent, sizeof (parent))) != 0) | |
2028 | goto out; | |
34dc7c2f BB |
2029 | |
2030 | /* | |
2031 | * Quit if directory has been removed (posix) | |
2032 | */ | |
3558fd73 BB |
2033 | if (zp->z_unlinked) |
2034 | goto out; | |
2035 | ||
a08ee875 | 2036 | error = 0; |
3558fd73 | 2037 | os = zsb->z_os; |
a08ee875 | 2038 | offset = ctx->pos; |
34dc7c2f BB |
2039 | prefetch = zp->z_zn_prefetch; |
2040 | ||
2041 | /* | |
2042 | * Initialize the iterator cursor. | |
2043 | */ | |
a08ee875 | 2044 | if (offset <= 3) { |
34dc7c2f BB |
2045 | /* |
2046 | * Start iteration from the beginning of the directory. | |
2047 | */ | |
2048 | zap_cursor_init(&zc, os, zp->z_id); | |
2049 | } else { | |
2050 | /* | |
2051 | * The offset is a serialized cursor. | |
2052 | */ | |
a08ee875 | 2053 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); |
34dc7c2f BB |
2054 | } |
2055 | ||
34dc7c2f BB |
2056 | /* |
2057 | * Transform to file-system independent format | |
2058 | */ | |
3558fd73 BB |
2059 | while (!done) { |
2060 | uint64_t objnum; | |
34dc7c2f BB |
2061 | /* |
2062 | * Special case `.', `..', and `.zfs'. | |
2063 | */ | |
a08ee875 | 2064 | if (offset == 0) { |
34dc7c2f BB |
2065 | (void) strcpy(zap.za_name, "."); |
2066 | zap.za_normalization_conflict = 0; | |
2067 | objnum = zp->z_id; | |
a08ee875 LG |
2068 | type = DT_DIR; |
2069 | } else if (offset == 1) { | |
34dc7c2f BB |
2070 | (void) strcpy(zap.za_name, ".."); |
2071 | zap.za_normalization_conflict = 0; | |
428870ff | 2072 | objnum = parent; |
a08ee875 LG |
2073 | type = DT_DIR; |
2074 | } else if (offset == 2 && zfs_show_ctldir(zp)) { | |
34dc7c2f BB |
2075 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2076 | zap.za_normalization_conflict = 0; | |
2077 | objnum = ZFSCTL_INO_ROOT; | |
a08ee875 | 2078 | type = DT_DIR; |
34dc7c2f BB |
2079 | } else { |
2080 | /* | |
2081 | * Grab next entry. | |
2082 | */ | |
3558fd73 BB |
2083 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2084 | if (error == ENOENT) | |
34dc7c2f BB |
2085 | break; |
2086 | else | |
2087 | goto update; | |
2088 | } | |
2089 | ||
0c5dde49 BB |
2090 | /* |
2091 | * Allow multiple entries provided the first entry is | |
2092 | * the object id. Non-zpl consumers may safely make | |
2093 | * use of the additional space. | |
2094 | * | |
2095 | * XXX: This should be a feature flag for compatibility | |
2096 | */ | |
34dc7c2f | 2097 | if (zap.za_integer_length != 8 || |
0c5dde49 | 2098 | zap.za_num_integers == 0) { |
34dc7c2f | 2099 | cmn_err(CE_WARN, "zap_readdir: bad directory " |
0c5dde49 BB |
2100 | "entry, obj = %lld, offset = %lld, " |
2101 | "length = %d, num = %lld\n", | |
34dc7c2f | 2102 | (u_longlong_t)zp->z_id, |
a08ee875 | 2103 | (u_longlong_t)offset, |
0c5dde49 BB |
2104 | zap.za_integer_length, |
2105 | (u_longlong_t)zap.za_num_integers); | |
a08ee875 | 2106 | error = SET_ERROR(ENXIO); |
34dc7c2f BB |
2107 | goto update; |
2108 | } | |
2109 | ||
2110 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
a08ee875 | 2111 | type = ZFS_DIRENT_TYPE(zap.za_first_integer); |
34dc7c2f | 2112 | } |
c06d4368 AX |
2113 | |
2114 | done = !dir_emit(ctx, zap.za_name, strlen(zap.za_name), | |
a08ee875 | 2115 | objnum, type); |
c06d4368 | 2116 | if (done) |
34dc7c2f | 2117 | break; |
34dc7c2f BB |
2118 | |
2119 | /* Prefetch znode */ | |
3558fd73 | 2120 | if (prefetch) { |
34dc7c2f | 2121 | dmu_prefetch(os, objnum, 0, 0); |
3558fd73 | 2122 | } |
34dc7c2f | 2123 | |
a08ee875 LG |
2124 | /* |
2125 | * Move to the next entry, fill in the previous offset. | |
2126 | */ | |
2127 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
34dc7c2f | 2128 | zap_cursor_advance(&zc); |
a08ee875 | 2129 | offset = zap_cursor_serialize(&zc); |
34dc7c2f | 2130 | } else { |
a08ee875 | 2131 | offset += 1; |
34dc7c2f | 2132 | } |
a08ee875 | 2133 | ctx->pos = offset; |
34dc7c2f BB |
2134 | } |
2135 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2136 | ||
34dc7c2f BB |
2137 | update: |
2138 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2139 | if (error == ENOENT) |
2140 | error = 0; | |
2141 | ||
3558fd73 | 2142 | ZFS_ACCESSTIME_STAMP(zsb, zp); |
3558fd73 BB |
2143 | |
2144 | out: | |
2145 | ZFS_EXIT(zsb); | |
34dc7c2f | 2146 | |
34dc7c2f BB |
2147 | return (error); |
2148 | } | |
3558fd73 | 2149 | EXPORT_SYMBOL(zfs_readdir); |
34dc7c2f | 2150 | |
d5446cfc BB |
2151 | ulong_t zfs_fsync_sync_cnt = 4; |
2152 | ||
e5c39b95 | 2153 | int |
3558fd73 | 2154 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2155 | { |
3558fd73 BB |
2156 | znode_t *zp = ITOZ(ip); |
2157 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 2158 | |
d5446cfc BB |
2159 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); |
2160 | ||
3558fd73 BB |
2161 | if (zsb->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2162 | ZFS_ENTER(zsb); | |
428870ff | 2163 | ZFS_VERIFY_ZP(zp); |
3558fd73 BB |
2164 | zil_commit(zsb->z_log, zp->z_id); |
2165 | ZFS_EXIT(zsb); | |
428870ff | 2166 | } |
e10b0808 AX |
2167 | tsd_set(zfs_fsyncer_key, NULL); |
2168 | ||
34dc7c2f BB |
2169 | return (0); |
2170 | } | |
e5c39b95 | 2171 | EXPORT_SYMBOL(zfs_fsync); |
34dc7c2f BB |
2172 | |
2173 | ||
2174 | /* | |
2175 | * Get the requested file attributes and place them in the provided | |
2176 | * vattr structure. | |
2177 | * | |
3558fd73 | 2178 | * IN: ip - inode of file. |
5484965a BB |
2179 | * vap - va_mask identifies requested attributes. |
2180 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2181 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2182 | * cr - credentials of caller. | |
34dc7c2f | 2183 | * |
5484965a BB |
2184 | * OUT: vap - attribute values. |
2185 | * | |
2186 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2187 | */ |
2188 | /* ARGSUSED */ | |
e5c39b95 | 2189 | int |
5484965a | 2190 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2191 | { |
3558fd73 BB |
2192 | znode_t *zp = ITOZ(ip); |
2193 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2194 | int error = 0; |
2195 | uint64_t links; | |
428870ff | 2196 | uint64_t mtime[2], ctime[2]; |
5484965a BB |
2197 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2198 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2199 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff BB |
2200 | sa_bulk_attr_t bulk[2]; |
2201 | int count = 0; | |
34dc7c2f | 2202 | |
3558fd73 | 2203 | ZFS_ENTER(zsb); |
34dc7c2f | 2204 | ZFS_VERIFY_ZP(zp); |
428870ff | 2205 | |
5484965a | 2206 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2207 | |
3558fd73 BB |
2208 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
2209 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
428870ff BB |
2210 | |
2211 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
3558fd73 | 2212 | ZFS_EXIT(zsb); |
428870ff BB |
2213 | return (error); |
2214 | } | |
34dc7c2f | 2215 | |
34dc7c2f BB |
2216 | /* |
2217 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2218 | * Also, if we are the owner don't bother, since owner should | |
2219 | * always be allowed to read basic attributes of file. | |
2220 | */ | |
572e2857 | 2221 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2222 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2223 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2224 | skipaclchk, cr))) { | |
3558fd73 | 2225 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2226 | return (error); |
2227 | } | |
2228 | } | |
2229 | ||
2230 | /* | |
2231 | * Return all attributes. It's cheaper to provide the answer | |
2232 | * than to determine whether we were asked the question. | |
2233 | */ | |
2234 | ||
9babb374 | 2235 | mutex_enter(&zp->z_lock); |
5484965a BB |
2236 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2237 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2238 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2239 | vap->va_nodeid = zp->z_id; |
3558fd73 | 2240 | if ((zp->z_id == zsb->z_root) && zfs_show_ctldir(zp)) |
428870ff | 2241 | links = zp->z_links + 1; |
34dc7c2f | 2242 | else |
428870ff | 2243 | links = zp->z_links; |
5484965a BB |
2244 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2245 | vap->va_size = i_size_read(ip); | |
2246 | vap->va_rdev = ip->i_rdev; | |
2247 | vap->va_seq = ip->i_generation; | |
2248 | ||
2249 | /* | |
2250 | * Add in any requested optional attributes and the create time. | |
2251 | * Also set the corresponding bits in the returned attribute bitmap. | |
2252 | */ | |
2253 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zsb->z_use_fuids) { | |
2254 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
2255 | xoap->xoa_archive = | |
2256 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2257 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2258 | } | |
2259 | ||
2260 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2261 | xoap->xoa_readonly = | |
2262 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2263 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2264 | } | |
2265 | ||
2266 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2267 | xoap->xoa_system = | |
2268 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2269 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2270 | } | |
2271 | ||
2272 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2273 | xoap->xoa_hidden = | |
2274 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2275 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2276 | } | |
2277 | ||
2278 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2279 | xoap->xoa_nounlink = | |
2280 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2281 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2282 | } | |
2283 | ||
2284 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2285 | xoap->xoa_immutable = | |
2286 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2287 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2288 | } | |
2289 | ||
2290 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2291 | xoap->xoa_appendonly = | |
2292 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2293 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2294 | } | |
2295 | ||
2296 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2297 | xoap->xoa_nodump = | |
2298 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2299 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2300 | } | |
2301 | ||
2302 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2303 | xoap->xoa_opaque = | |
2304 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2305 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2306 | } | |
2307 | ||
2308 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2309 | xoap->xoa_av_quarantined = | |
2310 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2311 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2312 | } | |
2313 | ||
2314 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2315 | xoap->xoa_av_modified = | |
2316 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2317 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2318 | } | |
2319 | ||
2320 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2321 | S_ISREG(ip->i_mode)) { | |
2322 | zfs_sa_get_scanstamp(zp, xvap); | |
2323 | } | |
34dc7c2f | 2324 | |
5484965a BB |
2325 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2326 | uint64_t times[2]; | |
2327 | ||
2328 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zsb), | |
2329 | times, sizeof (times)); | |
2330 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2331 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2332 | } | |
2333 | ||
2334 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2335 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2336 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2337 | } | |
2338 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
2339 | xoap->xoa_generation = zp->z_gen; | |
2340 | XVA_SET_RTN(xvap, XAT_GEN); | |
2341 | } | |
2342 | ||
2343 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2344 | xoap->xoa_offline = | |
2345 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2346 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2347 | } | |
2348 | ||
2349 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2350 | xoap->xoa_sparse = | |
2351 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2352 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2353 | } | |
2354 | } | |
2355 | ||
2356 | ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime); | |
2357 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); | |
2358 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2359 | |
2360 | mutex_exit(&zp->z_lock); | |
2361 | ||
5484965a | 2362 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2363 | |
2364 | if (zp->z_blksz == 0) { | |
2365 | /* | |
2366 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2367 | */ | |
5484965a | 2368 | vap->va_blksize = zsb->z_max_blksz; |
34dc7c2f BB |
2369 | } |
2370 | ||
3558fd73 | 2371 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2372 | return (0); |
2373 | } | |
e5c39b95 | 2374 | EXPORT_SYMBOL(zfs_getattr); |
34dc7c2f | 2375 | |
057e8eee BB |
2376 | /* |
2377 | * Get the basic file attributes and place them in the provided kstat | |
2378 | * structure. The inode is assumed to be the authoritative source | |
2379 | * for most of the attributes. However, the znode currently has the | |
2380 | * authoritative atime, blksize, and block count. | |
2381 | * | |
2382 | * IN: ip - inode of file. | |
2383 | * | |
2384 | * OUT: sp - kstat values. | |
2385 | * | |
2386 | * RETURN: 0 (always succeeds) | |
2387 | */ | |
2388 | /* ARGSUSED */ | |
2389 | int | |
2390 | zfs_getattr_fast(struct inode *ip, struct kstat *sp) | |
2391 | { | |
2392 | znode_t *zp = ITOZ(ip); | |
2393 | zfs_sb_t *zsb = ITOZSB(ip); | |
a08ee875 LG |
2394 | uint32_t blksize; |
2395 | u_longlong_t nblocks; | |
057e8eee | 2396 | |
a7b125e9 GB |
2397 | ZFS_ENTER(zsb); |
2398 | ZFS_VERIFY_ZP(zp); | |
2399 | ||
057e8eee BB |
2400 | mutex_enter(&zp->z_lock); |
2401 | ||
2402 | generic_fillattr(ip, sp); | |
2403 | ZFS_TIME_DECODE(&sp->atime, zp->z_atime); | |
2404 | ||
a08ee875 LG |
2405 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); |
2406 | sp->blksize = blksize; | |
2407 | sp->blocks = nblocks; | |
2408 | ||
057e8eee BB |
2409 | if (unlikely(zp->z_blksz == 0)) { |
2410 | /* | |
2411 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2412 | */ | |
2413 | sp->blksize = zsb->z_max_blksz; | |
2414 | } | |
2415 | ||
2416 | mutex_exit(&zp->z_lock); | |
2417 | ||
e10b0808 AX |
2418 | /* |
2419 | * Required to prevent NFS client from detecting different inode | |
2420 | * numbers of snapshot root dentry before and after snapshot mount. | |
2421 | */ | |
2422 | if (zsb->z_issnap) { | |
2423 | if (ip->i_sb->s_root->d_inode == ip) | |
2424 | sp->ino = ZFSCTL_INO_SNAPDIRS - | |
2425 | dmu_objset_id(zsb->z_os); | |
2426 | } | |
2427 | ||
a7b125e9 GB |
2428 | ZFS_EXIT(zsb); |
2429 | ||
057e8eee BB |
2430 | return (0); |
2431 | } | |
2432 | EXPORT_SYMBOL(zfs_getattr_fast); | |
2433 | ||
34dc7c2f BB |
2434 | /* |
2435 | * Set the file attributes to the values contained in the | |
2436 | * vattr structure. | |
2437 | * | |
3558fd73 | 2438 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2439 | * vap - new attribute values. |
5484965a | 2440 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2441 | * flags - ATTR_UTIME set if non-default time values provided. |
2442 | * - ATTR_NOACLCHECK (CIFS context only). | |
2443 | * cr - credentials of caller. | |
34dc7c2f BB |
2444 | * |
2445 | * RETURN: 0 if success | |
2446 | * error code if failure | |
2447 | * | |
2448 | * Timestamps: | |
3558fd73 | 2449 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2450 | */ |
2451 | /* ARGSUSED */ | |
e5c39b95 | 2452 | int |
5484965a | 2453 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2454 | { |
3558fd73 BB |
2455 | znode_t *zp = ITOZ(ip); |
2456 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2457 | zilog_t *zilog; |
2458 | dmu_tx_t *tx; | |
2459 | vattr_t oldva; | |
f4ea75d4 | 2460 | xvattr_t *tmpxvattr; |
5484965a | 2461 | uint_t mask = vap->va_mask; |
a08ee875 | 2462 | uint_t saved_mask = 0; |
34dc7c2f BB |
2463 | int trim_mask = 0; |
2464 | uint64_t new_mode; | |
9babb374 | 2465 | uint64_t new_uid, new_gid; |
572e2857 | 2466 | uint64_t xattr_obj; |
428870ff | 2467 | uint64_t mtime[2], ctime[2]; |
34dc7c2f BB |
2468 | znode_t *attrzp; |
2469 | int need_policy = FALSE; | |
428870ff | 2470 | int err, err2; |
34dc7c2f | 2471 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2472 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2473 | xoptattr_t *xoap; | |
2474 | zfs_acl_t *aclp; | |
34dc7c2f | 2475 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2476 | boolean_t fuid_dirtied = B_FALSE; |
17c37660 | 2477 | sa_bulk_attr_t *bulk, *xattr_bulk; |
428870ff | 2478 | int count = 0, xattr_count = 0; |
34dc7c2f BB |
2479 | |
2480 | if (mask == 0) | |
2481 | return (0); | |
2482 | ||
3558fd73 | 2483 | ZFS_ENTER(zsb); |
34dc7c2f BB |
2484 | ZFS_VERIFY_ZP(zp); |
2485 | ||
3558fd73 | 2486 | zilog = zsb->z_log; |
34dc7c2f BB |
2487 | |
2488 | /* | |
2489 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2490 | * that file system is at proper version level | |
2491 | */ | |
5484965a | 2492 | |
3558fd73 | 2493 | if (zsb->z_use_fuids == B_FALSE && |
5484965a BB |
2494 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2495 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2496 | (mask & ATTR_XVATTR))) { | |
3558fd73 | 2497 | ZFS_EXIT(zsb); |
a08ee875 | 2498 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2499 | } |
2500 | ||
3558fd73 BB |
2501 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
2502 | ZFS_EXIT(zsb); | |
a08ee875 | 2503 | return (SET_ERROR(EISDIR)); |
34dc7c2f BB |
2504 | } |
2505 | ||
3558fd73 BB |
2506 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
2507 | ZFS_EXIT(zsb); | |
a08ee875 | 2508 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
2509 | } |
2510 | ||
5484965a BB |
2511 | /* |
2512 | * If this is an xvattr_t, then get a pointer to the structure of | |
2513 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2514 | */ | |
2515 | xoap = xva_getxoptattr(xvap); | |
2516 | ||
a08ee875 | 2517 | tmpxvattr = kmem_alloc(sizeof (xvattr_t), KM_SLEEP); |
f4ea75d4 | 2518 | xva_init(tmpxvattr); |
5484965a | 2519 | |
a08ee875 LG |
2520 | bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); |
2521 | xattr_bulk = kmem_alloc(sizeof (sa_bulk_attr_t) * 7, KM_SLEEP); | |
17c37660 | 2522 | |
5484965a BB |
2523 | /* |
2524 | * Immutable files can only alter immutable bit and atime | |
2525 | */ | |
2526 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2527 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2528 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
f4ea75d4 BB |
2529 | err = EPERM; |
2530 | goto out3; | |
5484965a BB |
2531 | } |
2532 | ||
3558fd73 | 2533 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
f4ea75d4 BB |
2534 | err = EPERM; |
2535 | goto out3; | |
34dc7c2f BB |
2536 | } |
2537 | ||
5484965a BB |
2538 | /* |
2539 | * Verify timestamps doesn't overflow 32 bits. | |
2540 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2541 | * handle times greater than 2039. This check should be removed | |
2542 | * once large timestamps are fully supported. | |
2543 | */ | |
2544 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
a08ee875 LG |
2545 | if (((mask & ATTR_ATIME) && |
2546 | TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2547 | ((mask & ATTR_MTIME) && | |
2548 | TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
f4ea75d4 BB |
2549 | err = EOVERFLOW; |
2550 | goto out3; | |
5484965a BB |
2551 | } |
2552 | } | |
2553 | ||
34dc7c2f BB |
2554 | top: |
2555 | attrzp = NULL; | |
572e2857 | 2556 | aclp = NULL; |
34dc7c2f | 2557 | |
45d1cae3 | 2558 | /* Can this be moved to before the top label? */ |
2cf7f52b | 2559 | if (zfs_is_readonly(zsb)) { |
f4ea75d4 BB |
2560 | err = EROFS; |
2561 | goto out3; | |
34dc7c2f BB |
2562 | } |
2563 | ||
2564 | /* | |
2565 | * First validate permissions | |
2566 | */ | |
2567 | ||
3558fd73 | 2568 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2569 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2570 | if (err) |
2571 | goto out3; | |
2572 | ||
34dc7c2f BB |
2573 | /* |
2574 | * XXX - Note, we are not providing any open | |
2575 | * mode flags here (like FNDELAY), so we may | |
2576 | * block if there are locks present... this | |
2577 | * should be addressed in openat(). | |
2578 | */ | |
b128c09f | 2579 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2580 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2581 | if (err) |
2582 | goto out3; | |
428870ff | 2583 | } |
34dc7c2f | 2584 | |
5484965a BB |
2585 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2586 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2587 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2588 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2589 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2590 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2591 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2592 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2593 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2594 | skipaclchk, cr); | |
2595 | } | |
2596 | ||
3558fd73 BB |
2597 | if (mask & (ATTR_UID|ATTR_GID)) { |
2598 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2599 | int take_owner; |
2600 | int take_group; | |
2601 | ||
2602 | /* | |
2603 | * NOTE: even if a new mode is being set, | |
2604 | * we may clear S_ISUID/S_ISGID bits. | |
2605 | */ | |
2606 | ||
3558fd73 | 2607 | if (!(mask & ATTR_MODE)) |
5484965a | 2608 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2609 | |
2610 | /* | |
2611 | * Take ownership or chgrp to group we are a member of | |
2612 | */ | |
2613 | ||
5484965a | 2614 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 2615 | take_group = (mask & ATTR_GID) && |
5484965a | 2616 | zfs_groupmember(zsb, vap->va_gid, cr); |
34dc7c2f BB |
2617 | |
2618 | /* | |
5484965a | 2619 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
2620 | * take_group must both be set in order to allow taking |
2621 | * ownership. | |
2622 | * | |
2623 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2624 | * | |
2625 | */ | |
2626 | ||
3558fd73 BB |
2627 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
2628 | take_owner && take_group) || | |
2629 | ((idmask == ATTR_UID) && take_owner) || | |
2630 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
2631 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
2632 | skipaclchk, cr) == 0) { | |
2633 | /* | |
2634 | * Remove setuid/setgid for non-privileged users | |
2635 | */ | |
5484965a | 2636 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 2637 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
2638 | } else { |
2639 | need_policy = TRUE; | |
2640 | } | |
2641 | } else { | |
2642 | need_policy = TRUE; | |
2643 | } | |
2644 | } | |
2645 | ||
2646 | mutex_enter(&zp->z_lock); | |
428870ff | 2647 | oldva.va_mode = zp->z_mode; |
572e2857 | 2648 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
2649 | if (mask & ATTR_XVATTR) { |
2650 | /* | |
2651 | * Update xvattr mask to include only those attributes | |
2652 | * that are actually changing. | |
2653 | * | |
2654 | * the bits will be restored prior to actually setting | |
2655 | * the attributes so the caller thinks they were set. | |
2656 | */ | |
2657 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2658 | if (xoap->xoa_appendonly != | |
2659 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
2660 | need_policy = TRUE; | |
2661 | } else { | |
2662 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 2663 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
2664 | } |
2665 | } | |
2666 | ||
2667 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2668 | if (xoap->xoa_nounlink != | |
2669 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
2670 | need_policy = TRUE; | |
2671 | } else { | |
2672 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 2673 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
2674 | } |
2675 | } | |
2676 | ||
2677 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2678 | if (xoap->xoa_immutable != | |
2679 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
2680 | need_policy = TRUE; | |
2681 | } else { | |
2682 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 2683 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
2684 | } |
2685 | } | |
2686 | ||
2687 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2688 | if (xoap->xoa_nodump != | |
2689 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
2690 | need_policy = TRUE; | |
2691 | } else { | |
2692 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 2693 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
2694 | } |
2695 | } | |
2696 | ||
2697 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2698 | if (xoap->xoa_av_modified != | |
2699 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
2700 | need_policy = TRUE; | |
2701 | } else { | |
2702 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 2703 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
2704 | } |
2705 | } | |
2706 | ||
2707 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2708 | if ((!S_ISREG(ip->i_mode) && | |
2709 | xoap->xoa_av_quarantined) || | |
2710 | xoap->xoa_av_quarantined != | |
2711 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
2712 | need_policy = TRUE; | |
2713 | } else { | |
2714 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 2715 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
2716 | } |
2717 | } | |
2718 | ||
2719 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2720 | mutex_exit(&zp->z_lock); | |
f4ea75d4 BB |
2721 | err = EPERM; |
2722 | goto out3; | |
5484965a BB |
2723 | } |
2724 | ||
2725 | if (need_policy == FALSE && | |
2726 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2727 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
2728 | need_policy = TRUE; | |
2729 | } | |
2730 | } | |
34dc7c2f BB |
2731 | |
2732 | mutex_exit(&zp->z_lock); | |
2733 | ||
3558fd73 | 2734 | if (mask & ATTR_MODE) { |
34dc7c2f | 2735 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 2736 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 2737 | &oldva, cr); |
f4ea75d4 BB |
2738 | if (err) |
2739 | goto out3; | |
2740 | ||
3558fd73 | 2741 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
2742 | } else { |
2743 | need_policy = TRUE; | |
2744 | } | |
2745 | } | |
2746 | ||
2747 | if (need_policy) { | |
2748 | /* | |
2749 | * If trim_mask is set then take ownership | |
2750 | * has been granted or write_acl is present and user | |
2751 | * has the ability to modify mode. In that case remove | |
2752 | * UID|GID and or MODE from mask so that | |
2753 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2754 | */ | |
2755 | ||
2756 | if (trim_mask) { | |
5484965a BB |
2757 | saved_mask = vap->va_mask; |
2758 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 2759 | } |
5484965a | 2760 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 2761 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
2762 | if (err) |
2763 | goto out3; | |
34dc7c2f BB |
2764 | |
2765 | if (trim_mask) | |
5484965a | 2766 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
2767 | } |
2768 | ||
2769 | /* | |
2770 | * secpolicy_vnode_setattr, or take ownership may have | |
2771 | * changed va_mask | |
2772 | */ | |
5484965a | 2773 | mask = vap->va_mask; |
34dc7c2f | 2774 | |
3558fd73 BB |
2775 | if ((mask & (ATTR_UID | ATTR_GID))) { |
2776 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), | |
572e2857 | 2777 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 2778 | |
572e2857 | 2779 | if (err == 0 && xattr_obj) { |
3558fd73 | 2780 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
2781 | if (err) |
2782 | goto out2; | |
2783 | } | |
3558fd73 BB |
2784 | if (mask & ATTR_UID) { |
2785 | new_uid = zfs_fuid_create(zsb, | |
5484965a | 2786 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
572e2857 | 2787 | if (new_uid != zp->z_uid && |
3558fd73 | 2788 | zfs_fuid_overquota(zsb, B_FALSE, new_uid)) { |
572e2857 | 2789 | if (attrzp) |
3558fd73 | 2790 | iput(ZTOI(attrzp)); |
428870ff BB |
2791 | err = EDQUOT; |
2792 | goto out2; | |
2793 | } | |
2794 | } | |
2795 | ||
3558fd73 | 2796 | if (mask & ATTR_GID) { |
5484965a | 2797 | new_gid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid, |
428870ff BB |
2798 | cr, ZFS_GROUP, &fuidp); |
2799 | if (new_gid != zp->z_gid && | |
3558fd73 | 2800 | zfs_fuid_overquota(zsb, B_TRUE, new_gid)) { |
572e2857 | 2801 | if (attrzp) |
3558fd73 | 2802 | iput(ZTOI(attrzp)); |
428870ff BB |
2803 | err = EDQUOT; |
2804 | goto out2; | |
2805 | } | |
2806 | } | |
2807 | } | |
3558fd73 | 2808 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 2809 | |
3558fd73 | 2810 | if (mask & ATTR_MODE) { |
428870ff | 2811 | uint64_t pmode = zp->z_mode; |
572e2857 | 2812 | uint64_t acl_obj; |
5484965a | 2813 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 2814 | |
572e2857 | 2815 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 2816 | |
572e2857 BB |
2817 | mutex_enter(&zp->z_lock); |
2818 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
2819 | /* |
2820 | * Are we upgrading ACL from old V0 format | |
2821 | * to V1 format? | |
2822 | */ | |
3558fd73 | 2823 | if (zsb->z_version >= ZPL_VERSION_FUID && |
572e2857 | 2824 | zfs_znode_acl_version(zp) == |
34dc7c2f | 2825 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 2826 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
2827 | DMU_OBJECT_END); |
2828 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
2829 | 0, aclp->z_acl_bytes); | |
2830 | } else { | |
572e2857 | 2831 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
2832 | aclp->z_acl_bytes); |
2833 | } | |
428870ff | 2834 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
2835 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
2836 | 0, aclp->z_acl_bytes); | |
2837 | } | |
572e2857 | 2838 | mutex_exit(&zp->z_lock); |
428870ff BB |
2839 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
2840 | } else { | |
5484965a BB |
2841 | if ((mask & ATTR_XVATTR) && |
2842 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
2843 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
2844 | else | |
2845 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2846 | } |
2847 | ||
428870ff BB |
2848 | if (attrzp) { |
2849 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2850 | } |
2851 | ||
3558fd73 | 2852 | fuid_dirtied = zsb->z_fuid_dirty; |
428870ff | 2853 | if (fuid_dirtied) |
3558fd73 | 2854 | zfs_fuid_txhold(zsb, tx); |
428870ff BB |
2855 | |
2856 | zfs_sa_upgrade_txholds(tx, zp); | |
2857 | ||
a08ee875 LG |
2858 | err = dmu_tx_assign(tx, TXG_WAIT); |
2859 | if (err) | |
9babb374 | 2860 | goto out; |
34dc7c2f | 2861 | |
428870ff | 2862 | count = 0; |
34dc7c2f BB |
2863 | /* |
2864 | * Set each attribute requested. | |
2865 | * We group settings according to the locks they need to acquire. | |
2866 | * | |
2867 | * Note: you cannot set ctime directly, although it will be | |
2868 | * updated as a side-effect of calling this function. | |
2869 | */ | |
2870 | ||
572e2857 | 2871 | |
3558fd73 | 2872 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2873 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
2874 | mutex_enter(&zp->z_lock); |
2875 | ||
3558fd73 | 2876 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff BB |
2877 | &zp->z_pflags, sizeof (zp->z_pflags)); |
2878 | ||
2879 | if (attrzp) { | |
3558fd73 | 2880 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2881 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
2882 | mutex_enter(&attrzp->z_lock); |
2883 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2884 | SA_ZPL_FLAGS(zsb), NULL, &attrzp->z_pflags, |
428870ff BB |
2885 | sizeof (attrzp->z_pflags)); |
2886 | } | |
2887 | ||
3558fd73 | 2888 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 2889 | |
3558fd73 BB |
2890 | if (mask & ATTR_UID) { |
2891 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, | |
428870ff | 2892 | &new_uid, sizeof (new_uid)); |
572e2857 | 2893 | zp->z_uid = new_uid; |
428870ff BB |
2894 | if (attrzp) { |
2895 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2896 | SA_ZPL_UID(zsb), NULL, &new_uid, |
428870ff | 2897 | sizeof (new_uid)); |
572e2857 | 2898 | attrzp->z_uid = new_uid; |
428870ff BB |
2899 | } |
2900 | } | |
2901 | ||
3558fd73 BB |
2902 | if (mask & ATTR_GID) { |
2903 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), | |
428870ff | 2904 | NULL, &new_gid, sizeof (new_gid)); |
572e2857 | 2905 | zp->z_gid = new_gid; |
428870ff BB |
2906 | if (attrzp) { |
2907 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2908 | SA_ZPL_GID(zsb), NULL, &new_gid, |
428870ff | 2909 | sizeof (new_gid)); |
572e2857 | 2910 | attrzp->z_gid = new_gid; |
428870ff BB |
2911 | } |
2912 | } | |
3558fd73 BB |
2913 | if (!(mask & ATTR_MODE)) { |
2914 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), | |
428870ff BB |
2915 | NULL, &new_mode, sizeof (new_mode)); |
2916 | new_mode = zp->z_mode; | |
2917 | } | |
2918 | err = zfs_acl_chown_setattr(zp); | |
2919 | ASSERT(err == 0); | |
2920 | if (attrzp) { | |
2921 | err = zfs_acl_chown_setattr(attrzp); | |
2922 | ASSERT(err == 0); | |
2923 | } | |
2924 | } | |
2925 | ||
3558fd73 BB |
2926 | if (mask & ATTR_MODE) { |
2927 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, | |
428870ff BB |
2928 | &new_mode, sizeof (new_mode)); |
2929 | zp->z_mode = new_mode; | |
99c564bc | 2930 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 2931 | err = zfs_aclset_common(zp, aclp, cr, tx); |
c06d4368 | 2932 | ASSERT0(err); |
572e2857 BB |
2933 | if (zp->z_acl_cached) |
2934 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
2935 | zp->z_acl_cached = aclp; |
2936 | aclp = NULL; | |
34dc7c2f BB |
2937 | } |
2938 | ||
34dc7c2f | 2939 | |
3558fd73 | 2940 | if (mask & ATTR_ATIME) { |
5484965a | 2941 | ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime); |
3558fd73 | 2942 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, |
428870ff | 2943 | &zp->z_atime, sizeof (zp->z_atime)); |
34dc7c2f BB |
2944 | } |
2945 | ||
3558fd73 | 2946 | if (mask & ATTR_MTIME) { |
5484965a | 2947 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
3558fd73 | 2948 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, |
428870ff | 2949 | mtime, sizeof (mtime)); |
34dc7c2f BB |
2950 | } |
2951 | ||
b128c09f | 2952 | /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */ |
3558fd73 BB |
2953 | if (mask & ATTR_SIZE && !(mask & ATTR_MTIME)) { |
2954 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), | |
428870ff | 2955 | NULL, mtime, sizeof (mtime)); |
3558fd73 | 2956 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
2957 | &ctime, sizeof (ctime)); |
2958 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, | |
2959 | B_TRUE); | |
2960 | } else if (mask != 0) { | |
3558fd73 | 2961 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
2962 | &ctime, sizeof (ctime)); |
2963 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, | |
2964 | B_TRUE); | |
2965 | if (attrzp) { | |
2966 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2967 | SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
2968 | &ctime, sizeof (ctime)); |
2969 | zfs_tstamp_update_setup(attrzp, STATE_CHANGED, | |
2970 | mtime, ctime, B_TRUE); | |
2971 | } | |
2972 | } | |
34dc7c2f BB |
2973 | /* |
2974 | * Do this after setting timestamps to prevent timestamp | |
2975 | * update from toggling bit | |
2976 | */ | |
2977 | ||
5484965a BB |
2978 | if (xoap && (mask & ATTR_XVATTR)) { |
2979 | ||
2980 | /* | |
2981 | * restore trimmed off masks | |
2982 | * so that return masks can be set for caller. | |
2983 | */ | |
2984 | ||
f4ea75d4 | 2985 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
2986 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
2987 | } | |
f4ea75d4 | 2988 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
2989 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
2990 | } | |
f4ea75d4 | 2991 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
2992 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
2993 | } | |
f4ea75d4 | 2994 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
2995 | XVA_SET_REQ(xvap, XAT_NODUMP); |
2996 | } | |
f4ea75d4 | 2997 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
2998 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
2999 | } | |
f4ea75d4 | 3000 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
3001 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
3002 | } | |
3003 | ||
3004 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3005 | ASSERT(S_ISREG(ip->i_mode)); | |
3006 | ||
3007 | zfs_xvattr_set(zp, xvap, tx); | |
3008 | } | |
3009 | ||
9babb374 | 3010 | if (fuid_dirtied) |
3558fd73 | 3011 | zfs_fuid_sync(zsb, tx); |
9babb374 | 3012 | |
34dc7c2f | 3013 | if (mask != 0) |
5484965a | 3014 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 3015 | |
34dc7c2f | 3016 | mutex_exit(&zp->z_lock); |
3558fd73 | 3017 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 3018 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 3019 | |
572e2857 | 3020 | if (attrzp) { |
3558fd73 | 3021 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
3022 | mutex_exit(&attrzp->z_acl_lock); |
3023 | mutex_exit(&attrzp->z_lock); | |
3024 | } | |
9babb374 | 3025 | out: |
428870ff BB |
3026 | if (err == 0 && attrzp) { |
3027 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, | |
3028 | xattr_count, tx); | |
3029 | ASSERT(err2 == 0); | |
3030 | } | |
3031 | ||
34dc7c2f | 3032 | if (attrzp) |
3558fd73 | 3033 | iput(ZTOI(attrzp)); |
45d1cae3 | 3034 | if (aclp) |
9babb374 | 3035 | zfs_acl_free(aclp); |
9babb374 BB |
3036 | |
3037 | if (fuidp) { | |
3038 | zfs_fuid_info_free(fuidp); | |
3039 | fuidp = NULL; | |
3040 | } | |
3041 | ||
428870ff | 3042 | if (err) { |
9babb374 | 3043 | dmu_tx_abort(tx); |
428870ff BB |
3044 | if (err == ERESTART) |
3045 | goto top; | |
3046 | } else { | |
3047 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3048 | dmu_tx_commit(tx); |
037849f8 | 3049 | zfs_inode_update(zp); |
428870ff BB |
3050 | } |
3051 | ||
428870ff | 3052 | out2: |
3558fd73 | 3053 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3054 | zil_commit(zilog, 0); |
34dc7c2f | 3055 | |
f4ea75d4 | 3056 | out3: |
a08ee875 LG |
3057 | kmem_free(xattr_bulk, sizeof (sa_bulk_attr_t) * 7); |
3058 | kmem_free(bulk, sizeof (sa_bulk_attr_t) * 7); | |
3059 | kmem_free(tmpxvattr, sizeof (xvattr_t)); | |
3558fd73 | 3060 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3061 | return (err); |
3062 | } | |
e5c39b95 | 3063 | EXPORT_SYMBOL(zfs_setattr); |
34dc7c2f BB |
3064 | |
3065 | typedef struct zfs_zlock { | |
3066 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3067 | znode_t *zl_znode; /* znode we held */ | |
3068 | struct zfs_zlock *zl_next; /* next in list */ | |
3069 | } zfs_zlock_t; | |
3070 | ||
3071 | /* | |
3072 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3073 | */ | |
3074 | static void | |
3075 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3076 | { | |
3077 | zfs_zlock_t *zl; | |
3078 | ||
3079 | while ((zl = *zlpp) != NULL) { | |
3080 | if (zl->zl_znode != NULL) | |
3558fd73 | 3081 | iput(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
3082 | rw_exit(zl->zl_rwlock); |
3083 | *zlpp = zl->zl_next; | |
3084 | kmem_free(zl, sizeof (*zl)); | |
3085 | } | |
3086 | } | |
3087 | ||
3088 | /* | |
3089 | * Search back through the directory tree, using the ".." entries. | |
3090 | * Lock each directory in the chain to prevent concurrent renames. | |
3091 | * Fail any attempt to move a directory into one of its own descendants. | |
3092 | * XXX - z_parent_lock can overlap with map or grow locks | |
3093 | */ | |
3094 | static int | |
3095 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3096 | { | |
3097 | zfs_zlock_t *zl; | |
3098 | znode_t *zp = tdzp; | |
3558fd73 | 3099 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 3100 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3101 | krwlock_t *rwlp = &szp->z_parent_lock; |
3102 | krw_t rw = RW_WRITER; | |
3103 | ||
3104 | /* | |
3105 | * First pass write-locks szp and compares to zp->z_id. | |
3106 | * Later passes read-lock zp and compare to zp->z_parent. | |
3107 | */ | |
3108 | do { | |
3109 | if (!rw_tryenter(rwlp, rw)) { | |
3110 | /* | |
3111 | * Another thread is renaming in this path. | |
3112 | * Note that if we are a WRITER, we don't have any | |
3113 | * parent_locks held yet. | |
3114 | */ | |
3115 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3116 | /* | |
3117 | * Drop our locks and restart | |
3118 | */ | |
3119 | zfs_rename_unlock(&zl); | |
3120 | *zlpp = NULL; | |
3121 | zp = tdzp; | |
428870ff | 3122 | oidp = zp->z_id; |
34dc7c2f BB |
3123 | rwlp = &szp->z_parent_lock; |
3124 | rw = RW_WRITER; | |
3125 | continue; | |
3126 | } else { | |
3127 | /* | |
3128 | * Wait for other thread to drop its locks | |
3129 | */ | |
3130 | rw_enter(rwlp, rw); | |
3131 | } | |
3132 | } | |
3133 | ||
3134 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3135 | zl->zl_rwlock = rwlp; | |
3136 | zl->zl_znode = NULL; | |
3137 | zl->zl_next = *zlpp; | |
3138 | *zlpp = zl; | |
3139 | ||
428870ff | 3140 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
a08ee875 | 3141 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 3142 | |
428870ff | 3143 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3144 | return (0); |
3145 | ||
3146 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3147 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3148 | if (error) |
3149 | return (error); | |
3150 | zl->zl_znode = zp; | |
3151 | } | |
3558fd73 | 3152 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3153 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3154 | rwlp = &zp->z_parent_lock; |
3155 | rw = RW_READER; | |
3156 | ||
3157 | } while (zp->z_id != sdzp->z_id); | |
3158 | ||
3159 | return (0); | |
3160 | } | |
3161 | ||
3162 | /* | |
3163 | * Move an entry from the provided source directory to the target | |
3164 | * directory. Change the entry name as indicated. | |
3165 | * | |
3558fd73 | 3166 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3167 | * snm - Old entry name. |
3558fd73 | 3168 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3169 | * tnm - New entry name. |
3170 | * cr - credentials of caller. | |
34dc7c2f BB |
3171 | * flags - case flags |
3172 | * | |
a08ee875 | 3173 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3174 | * |
3175 | * Timestamps: | |
3558fd73 | 3176 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3177 | */ |
3178 | /*ARGSUSED*/ | |
e5c39b95 | 3179 | int |
3558fd73 BB |
3180 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3181 | cred_t *cr, int flags) | |
34dc7c2f BB |
3182 | { |
3183 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 BB |
3184 | znode_t *sdzp = ITOZ(sdip); |
3185 | zfs_sb_t *zsb = ITOZSB(sdip); | |
34dc7c2f | 3186 | zilog_t *zilog; |
34dc7c2f BB |
3187 | zfs_dirlock_t *sdl, *tdl; |
3188 | dmu_tx_t *tx; | |
3189 | zfs_zlock_t *zl; | |
3190 | int cmp, serr, terr; | |
3191 | int error = 0; | |
3192 | int zflg = 0; | |
a08ee875 | 3193 | boolean_t waited = B_FALSE; |
34dc7c2f | 3194 | |
3558fd73 | 3195 | ZFS_ENTER(zsb); |
34dc7c2f | 3196 | ZFS_VERIFY_ZP(sdzp); |
3558fd73 | 3197 | zilog = zsb->z_log; |
34dc7c2f | 3198 | |
a08ee875 | 3199 | if (tdip->i_sb != sdip->i_sb || zfsctl_is_node(tdip)) { |
3558fd73 | 3200 | ZFS_EXIT(zsb); |
a08ee875 | 3201 | return (SET_ERROR(EXDEV)); |
34dc7c2f BB |
3202 | } |
3203 | ||
3558fd73 | 3204 | tdzp = ITOZ(tdip); |
34dc7c2f | 3205 | ZFS_VERIFY_ZP(tdzp); |
3558fd73 | 3206 | if (zsb->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3207 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3208 | ZFS_EXIT(zsb); |
a08ee875 | 3209 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3210 | } |
3211 | ||
3212 | if (flags & FIGNORECASE) | |
3213 | zflg |= ZCILOOK; | |
3214 | ||
3215 | top: | |
3216 | szp = NULL; | |
3217 | tzp = NULL; | |
3218 | zl = NULL; | |
3219 | ||
3220 | /* | |
3221 | * This is to prevent the creation of links into attribute space | |
3222 | * by renaming a linked file into/outof an attribute directory. | |
3223 | * See the comment in zfs_link() for why this is considered bad. | |
3224 | */ | |
428870ff | 3225 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 3226 | ZFS_EXIT(zsb); |
a08ee875 | 3227 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3228 | } |
3229 | ||
3230 | /* | |
3231 | * Lock source and target directory entries. To prevent deadlock, | |
3232 | * a lock ordering must be defined. We lock the directory with | |
3233 | * the smallest object id first, or if it's a tie, the one with | |
3234 | * the lexically first name. | |
3235 | */ | |
3236 | if (sdzp->z_id < tdzp->z_id) { | |
3237 | cmp = -1; | |
3238 | } else if (sdzp->z_id > tdzp->z_id) { | |
3239 | cmp = 1; | |
3240 | } else { | |
3241 | /* | |
3242 | * First compare the two name arguments without | |
3243 | * considering any case folding. | |
3244 | */ | |
3558fd73 | 3245 | int nofold = (zsb->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3246 | |
3247 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
3558fd73 | 3248 | ASSERT(error == 0 || !zsb->z_utf8); |
34dc7c2f BB |
3249 | if (cmp == 0) { |
3250 | /* | |
3251 | * POSIX: "If the old argument and the new argument | |
3252 | * both refer to links to the same existing file, | |
3253 | * the rename() function shall return successfully | |
3254 | * and perform no other action." | |
3255 | */ | |
3558fd73 | 3256 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3257 | return (0); |
3258 | } | |
3259 | /* | |
3260 | * If the file system is case-folding, then we may | |
3261 | * have some more checking to do. A case-folding file | |
3262 | * system is either supporting mixed case sensitivity | |
3263 | * access or is completely case-insensitive. Note | |
3264 | * that the file system is always case preserving. | |
3265 | * | |
3266 | * In mixed sensitivity mode case sensitive behavior | |
3267 | * is the default. FIGNORECASE must be used to | |
3268 | * explicitly request case insensitive behavior. | |
3269 | * | |
3270 | * If the source and target names provided differ only | |
3271 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3272 | * we will treat this as a special case in the | |
3273 | * case-insensitive mode: as long as the source name | |
3274 | * is an exact match, we will allow this to proceed as | |
3275 | * a name-change request. | |
3276 | */ | |
3558fd73 BB |
3277 | if ((zsb->z_case == ZFS_CASE_INSENSITIVE || |
3278 | (zsb->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3279 | flags & FIGNORECASE)) && |
3558fd73 | 3280 | u8_strcmp(snm, tnm, 0, zsb->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3281 | &error) == 0) { |
3282 | /* | |
3283 | * case preserving rename request, require exact | |
3284 | * name matches | |
3285 | */ | |
3286 | zflg |= ZCIEXACT; | |
3287 | zflg &= ~ZCILOOK; | |
3288 | } | |
3289 | } | |
3290 | ||
428870ff BB |
3291 | /* |
3292 | * If the source and destination directories are the same, we should | |
3293 | * grab the z_name_lock of that directory only once. | |
3294 | */ | |
3295 | if (sdzp == tdzp) { | |
3296 | zflg |= ZHAVELOCK; | |
3297 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3298 | } | |
3299 | ||
34dc7c2f BB |
3300 | if (cmp < 0) { |
3301 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3302 | ZEXISTS | zflg, NULL, NULL); | |
3303 | terr = zfs_dirent_lock(&tdl, | |
3304 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3305 | } else { | |
3306 | terr = zfs_dirent_lock(&tdl, | |
3307 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3308 | serr = zfs_dirent_lock(&sdl, | |
3309 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3310 | NULL, NULL); | |
3311 | } | |
3312 | ||
3313 | if (serr) { | |
3314 | /* | |
3315 | * Source entry invalid or not there. | |
3316 | */ | |
3317 | if (!terr) { | |
3318 | zfs_dirent_unlock(tdl); | |
3319 | if (tzp) | |
3558fd73 | 3320 | iput(ZTOI(tzp)); |
34dc7c2f | 3321 | } |
428870ff BB |
3322 | |
3323 | if (sdzp == tdzp) | |
3324 | rw_exit(&sdzp->z_name_lock); | |
3325 | ||
34dc7c2f BB |
3326 | if (strcmp(snm, "..") == 0) |
3327 | serr = EINVAL; | |
3558fd73 | 3328 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3329 | return (serr); |
3330 | } | |
3331 | if (terr) { | |
3332 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3333 | iput(ZTOI(szp)); |
428870ff BB |
3334 | |
3335 | if (sdzp == tdzp) | |
3336 | rw_exit(&sdzp->z_name_lock); | |
3337 | ||
34dc7c2f BB |
3338 | if (strcmp(tnm, "..") == 0) |
3339 | terr = EINVAL; | |
3558fd73 | 3340 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3341 | return (terr); |
3342 | } | |
3343 | ||
3344 | /* | |
3345 | * Must have write access at the source to remove the old entry | |
3346 | * and write access at the target to create the new entry. | |
3347 | * Note that if target and source are the same, this can be | |
3348 | * done in a single check. | |
3349 | */ | |
3350 | ||
149e873a | 3351 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3352 | goto out; |
3353 | ||
3558fd73 | 3354 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3355 | /* |
3356 | * Check to make sure rename is valid. | |
3357 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3358 | */ | |
149e873a | 3359 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3360 | goto out; |
3361 | } | |
3362 | ||
3363 | /* | |
3364 | * Does target exist? | |
3365 | */ | |
3366 | if (tzp) { | |
3367 | /* | |
3368 | * Source and target must be the same type. | |
3369 | */ | |
3558fd73 BB |
3370 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3371 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
a08ee875 | 3372 | error = SET_ERROR(ENOTDIR); |
34dc7c2f BB |
3373 | goto out; |
3374 | } | |
3375 | } else { | |
3558fd73 | 3376 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
a08ee875 | 3377 | error = SET_ERROR(EISDIR); |
34dc7c2f BB |
3378 | goto out; |
3379 | } | |
3380 | } | |
3381 | /* | |
3382 | * POSIX dictates that when the source and target | |
3383 | * entries refer to the same file object, rename | |
3384 | * must do nothing and exit without error. | |
3385 | */ | |
3386 | if (szp->z_id == tzp->z_id) { | |
3387 | error = 0; | |
3388 | goto out; | |
3389 | } | |
3390 | } | |
3391 | ||
3558fd73 | 3392 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
3393 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3394 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3395 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3396 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3397 | if (sdzp != tdzp) { |
3398 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3399 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3400 | } | |
3401 | if (tzp) { | |
3402 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3403 | zfs_sa_upgrade_txholds(tx, tzp); | |
3404 | } | |
3405 | ||
3406 | zfs_sa_upgrade_txholds(tx, szp); | |
3558fd73 | 3407 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
a08ee875 | 3408 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3409 | if (error) { |
3410 | if (zl != NULL) | |
3411 | zfs_rename_unlock(&zl); | |
3412 | zfs_dirent_unlock(sdl); | |
3413 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3414 | |
3415 | if (sdzp == tdzp) | |
3416 | rw_exit(&sdzp->z_name_lock); | |
3417 | ||
3558fd73 | 3418 | iput(ZTOI(szp)); |
34dc7c2f | 3419 | if (tzp) |
3558fd73 | 3420 | iput(ZTOI(tzp)); |
fb5f0bc8 | 3421 | if (error == ERESTART) { |
a08ee875 | 3422 | waited = B_TRUE; |
34dc7c2f BB |
3423 | dmu_tx_wait(tx); |
3424 | dmu_tx_abort(tx); | |
3425 | goto top; | |
3426 | } | |
3427 | dmu_tx_abort(tx); | |
3558fd73 | 3428 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3429 | return (error); |
3430 | } | |
3431 | ||
3432 | if (tzp) /* Attempt to remove the existing target */ | |
3433 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3434 | ||
3435 | if (error == 0) { | |
3436 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3437 | if (error == 0) { | |
428870ff | 3438 | szp->z_pflags |= ZFS_AV_MODIFIED; |
34dc7c2f | 3439 | |
3558fd73 | 3440 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zsb), |
428870ff | 3441 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
c06d4368 | 3442 | ASSERT0(error); |
34dc7c2f | 3443 | |
428870ff BB |
3444 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3445 | if (error == 0) { | |
3446 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3447 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3448 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3449 | } else { |
3450 | /* | |
3451 | * At this point, we have successfully created | |
3452 | * the target name, but have failed to remove | |
3453 | * the source name. Since the create was done | |
3454 | * with the ZRENAMING flag, there are | |
3455 | * complications; for one, the link count is | |
3456 | * wrong. The easiest way to deal with this | |
3457 | * is to remove the newly created target, and | |
3458 | * return the original error. This must | |
3459 | * succeed; fortunately, it is very unlikely to | |
3460 | * fail, since we just created it. | |
3461 | */ | |
3462 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3463 | ZRENAMING, NULL), ==, 0); | |
3464 | } | |
34dc7c2f BB |
3465 | } |
3466 | } | |
3467 | ||
3468 | dmu_tx_commit(tx); | |
3469 | out: | |
3470 | if (zl != NULL) | |
3471 | zfs_rename_unlock(&zl); | |
3472 | ||
3473 | zfs_dirent_unlock(sdl); | |
3474 | zfs_dirent_unlock(tdl); | |
3475 | ||
960e08fe | 3476 | zfs_inode_update(sdzp); |
428870ff BB |
3477 | if (sdzp == tdzp) |
3478 | rw_exit(&sdzp->z_name_lock); | |
3479 | ||
960e08fe BB |
3480 | if (sdzp != tdzp) |
3481 | zfs_inode_update(tdzp); | |
428870ff | 3482 | |
960e08fe | 3483 | zfs_inode_update(szp); |
3558fd73 | 3484 | iput(ZTOI(szp)); |
960e08fe BB |
3485 | if (tzp) { |
3486 | zfs_inode_update(tzp); | |
3558fd73 | 3487 | iput(ZTOI(tzp)); |
960e08fe | 3488 | } |
34dc7c2f | 3489 | |
3558fd73 | 3490 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3491 | zil_commit(zilog, 0); |
428870ff | 3492 | |
3558fd73 | 3493 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3494 | return (error); |
3495 | } | |
e5c39b95 | 3496 | EXPORT_SYMBOL(zfs_rename); |
34dc7c2f BB |
3497 | |
3498 | /* | |
3499 | * Insert the indicated symbolic reference entry into the directory. | |
3500 | * | |
3558fd73 | 3501 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
3502 | * link - Name for new symlink entry. |
3503 | * vap - Attributes of new entry. | |
3504 | * target - Target path of new symlink. | |
3558fd73 | 3505 | * |
34dc7c2f | 3506 | * cr - credentials of caller. |
34dc7c2f BB |
3507 | * flags - case flags |
3508 | * | |
a08ee875 | 3509 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
3510 | * |
3511 | * Timestamps: | |
3558fd73 | 3512 | * dip - ctime|mtime updated |
34dc7c2f BB |
3513 | */ |
3514 | /*ARGSUSED*/ | |
e5c39b95 | 3515 | int |
3558fd73 BB |
3516 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
3517 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 3518 | { |
3558fd73 | 3519 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
3520 | zfs_dirlock_t *dl; |
3521 | dmu_tx_t *tx; | |
3558fd73 | 3522 | zfs_sb_t *zsb = ITOZSB(dip); |
34dc7c2f | 3523 | zilog_t *zilog; |
428870ff | 3524 | uint64_t len = strlen(link); |
34dc7c2f BB |
3525 | int error; |
3526 | int zflg = ZNEW; | |
9babb374 BB |
3527 | zfs_acl_ids_t acl_ids; |
3528 | boolean_t fuid_dirtied; | |
428870ff | 3529 | uint64_t txtype = TX_SYMLINK; |
a08ee875 | 3530 | boolean_t waited = B_FALSE; |
34dc7c2f | 3531 | |
3558fd73 | 3532 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 3533 | |
3558fd73 | 3534 | ZFS_ENTER(zsb); |
34dc7c2f | 3535 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3536 | zilog = zsb->z_log; |
34dc7c2f | 3537 | |
3558fd73 | 3538 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 3539 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3540 | ZFS_EXIT(zsb); |
a08ee875 | 3541 | return (SET_ERROR(EILSEQ)); |
34dc7c2f BB |
3542 | } |
3543 | if (flags & FIGNORECASE) | |
3544 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3545 | |
3546 | if (len > MAXPATHLEN) { | |
3558fd73 | 3547 | ZFS_EXIT(zsb); |
a08ee875 | 3548 | return (SET_ERROR(ENAMETOOLONG)); |
34dc7c2f BB |
3549 | } |
3550 | ||
428870ff BB |
3551 | if ((error = zfs_acl_ids_create(dzp, 0, |
3552 | vap, cr, NULL, &acl_ids)) != 0) { | |
3558fd73 | 3553 | ZFS_EXIT(zsb); |
428870ff BB |
3554 | return (error); |
3555 | } | |
3556 | top: | |
3558fd73 BB |
3557 | *ipp = NULL; |
3558 | ||
34dc7c2f BB |
3559 | /* |
3560 | * Attempt to lock directory; fail if entry already exists. | |
3561 | */ | |
3562 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3563 | if (error) { | |
428870ff | 3564 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 3565 | ZFS_EXIT(zsb); |
428870ff BB |
3566 | return (error); |
3567 | } | |
3568 | ||
149e873a | 3569 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
3570 | zfs_acl_ids_free(&acl_ids); |
3571 | zfs_dirent_unlock(dl); | |
3558fd73 | 3572 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3573 | return (error); |
3574 | } | |
3575 | ||
3558fd73 | 3576 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
9babb374 BB |
3577 | zfs_acl_ids_free(&acl_ids); |
3578 | zfs_dirent_unlock(dl); | |
3558fd73 | 3579 | ZFS_EXIT(zsb); |
a08ee875 | 3580 | return (SET_ERROR(EDQUOT)); |
9babb374 | 3581 | } |
3558fd73 BB |
3582 | tx = dmu_tx_create(zsb->z_os); |
3583 | fuid_dirtied = zsb->z_fuid_dirty; | |
34dc7c2f | 3584 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3585 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3586 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3587 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3588 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
3558fd73 | 3589 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
3590 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
3591 | acl_ids.z_aclp->z_acl_bytes); | |
3592 | } | |
9babb374 | 3593 | if (fuid_dirtied) |
3558fd73 | 3594 | zfs_fuid_txhold(zsb, tx); |
a08ee875 | 3595 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3596 | if (error) { |
3597 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3598 | if (error == ERESTART) { |
a08ee875 | 3599 | waited = B_TRUE; |
34dc7c2f BB |
3600 | dmu_tx_wait(tx); |
3601 | dmu_tx_abort(tx); | |
3602 | goto top; | |
3603 | } | |
428870ff | 3604 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 3605 | dmu_tx_abort(tx); |
3558fd73 | 3606 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3607 | return (error); |
3608 | } | |
3609 | ||
34dc7c2f BB |
3610 | /* |
3611 | * Create a new object for the symlink. | |
428870ff | 3612 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 3613 | */ |
428870ff | 3614 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3615 | |
428870ff | 3616 | if (fuid_dirtied) |
3558fd73 | 3617 | zfs_fuid_sync(zsb, tx); |
34dc7c2f | 3618 | |
572e2857 | 3619 | mutex_enter(&zp->z_lock); |
428870ff | 3620 | if (zp->z_is_sa) |
3558fd73 | 3621 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zsb), |
428870ff BB |
3622 | link, len, tx); |
3623 | else | |
3624 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3625 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3626 | |
428870ff | 3627 | zp->z_size = len; |
3558fd73 | 3628 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 3629 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
3630 | /* |
3631 | * Insert the new object into the directory. | |
3632 | */ | |
3633 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
428870ff BB |
3634 | |
3635 | if (flags & FIGNORECASE) | |
3636 | txtype |= TX_CI; | |
3637 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
9babb374 | 3638 | |
960e08fe BB |
3639 | zfs_inode_update(dzp); |
3640 | zfs_inode_update(zp); | |
3641 | ||
9babb374 | 3642 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3643 | |
3644 | dmu_tx_commit(tx); | |
3645 | ||
3646 | zfs_dirent_unlock(dl); | |
3647 | ||
3558fd73 | 3648 | *ipp = ZTOI(zp); |
34dc7c2f | 3649 | |
3558fd73 | 3650 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3651 | zil_commit(zilog, 0); |
428870ff | 3652 | |
3558fd73 | 3653 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3654 | return (error); |
3655 | } | |
e5c39b95 | 3656 | EXPORT_SYMBOL(zfs_symlink); |
34dc7c2f BB |
3657 | |
3658 | /* | |
3659 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 3660 | * the symbolic path referred to by ip. |
34dc7c2f | 3661 | * |
8b4f9a2d BB |
3662 | * IN: ip - inode of symbolic link |
3663 | * uio - structure to contain the link path. | |
3664 | * cr - credentials of caller. | |
34dc7c2f BB |
3665 | * |
3666 | * RETURN: 0 if success | |
3667 | * error code if failure | |
3668 | * | |
3669 | * Timestamps: | |
3558fd73 | 3670 | * ip - atime updated |
34dc7c2f BB |
3671 | */ |
3672 | /* ARGSUSED */ | |
e5c39b95 | 3673 | int |
8b4f9a2d | 3674 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 3675 | { |
3558fd73 BB |
3676 | znode_t *zp = ITOZ(ip); |
3677 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
3678 | int error; |
3679 | ||
3558fd73 | 3680 | ZFS_ENTER(zsb); |
34dc7c2f BB |
3681 | ZFS_VERIFY_ZP(zp); |
3682 | ||
572e2857 | 3683 | mutex_enter(&zp->z_lock); |
428870ff | 3684 | if (zp->z_is_sa) |
8b4f9a2d BB |
3685 | error = sa_lookup_uio(zp->z_sa_hdl, |
3686 | SA_ZPL_SYMLINK(zsb), uio); | |
428870ff | 3687 | else |
8b4f9a2d | 3688 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 3689 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3690 | |
3558fd73 | 3691 | ZFS_ACCESSTIME_STAMP(zsb, zp); |
3558fd73 | 3692 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3693 | return (error); |
3694 | } | |
8b4f9a2d | 3695 | EXPORT_SYMBOL(zfs_readlink); |
34dc7c2f BB |
3696 | |
3697 | /* | |
3558fd73 | 3698 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 3699 | * |
3558fd73 BB |
3700 | * IN: tdip - Directory to contain new entry. |
3701 | * sip - inode of new entry. | |
34dc7c2f BB |
3702 | * name - name of new entry. |
3703 | * cr - credentials of caller. | |
34dc7c2f BB |
3704 | * |
3705 | * RETURN: 0 if success | |
3706 | * error code if failure | |
3707 | * | |
3708 | * Timestamps: | |
3558fd73 BB |
3709 | * tdip - ctime|mtime updated |
3710 | * sip - ctime updated | |
34dc7c2f BB |
3711 | */ |
3712 | /* ARGSUSED */ | |
e5c39b95 | 3713 | int |
3558fd73 | 3714 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr) |
34dc7c2f | 3715 | { |
3558fd73 | 3716 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 3717 | znode_t *tzp, *szp; |
3558fd73 | 3718 | zfs_sb_t *zsb = ITOZSB(tdip); |
34dc7c2f BB |
3719 | zilog_t *zilog; |
3720 | zfs_dirlock_t *dl; | |
3721 | dmu_tx_t *tx; | |
34dc7c2f BB |
3722 | int error; |
3723 | int zf = ZNEW; | |
428870ff | 3724 | uint64_t parent; |
572e2857 | 3725 | uid_t owner; |
a08ee875 | 3726 | boolean_t waited = B_FALSE; |
34dc7c2f | 3727 | |
3558fd73 | 3728 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 3729 | |
3558fd73 | 3730 | ZFS_ENTER(zsb); |
34dc7c2f | 3731 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3732 | zilog = zsb->z_log; |
34dc7c2f | 3733 | |
428870ff BB |
3734 | /* |
3735 | * POSIX dictates that we return EPERM here. | |
3736 | * Better choices include ENOTSUP or EISDIR. | |
3737 | */ | |
3558fd73 BB |
3738 | if (S_ISDIR(sip->i_mode)) { |
3739 | ZFS_EXIT(zsb); | |
a08ee875 | 3740 | return (SET_ERROR(EPERM)); |
428870ff BB |
3741 | } |
3742 | ||
a08ee875 | 3743 | if (sip->i_sb != tdip->i_sb || zfsctl_is_node(sip)) { |
3558fd73 | 3744 | ZFS_EXIT(zsb); |
a08ee875 | 3745 | return (SET_ERROR(EXDEV)); |
34dc7c2f | 3746 | } |
428870ff | 3747 | |
3558fd73 | 3748 | szp = ITOZ(sip); |
34dc7c2f BB |
3749 | ZFS_VERIFY_ZP(szp); |
3750 | ||
428870ff BB |
3751 | /* Prevent links to .zfs/shares files */ |
3752 | ||
3558fd73 | 3753 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
428870ff | 3754 | &parent, sizeof (uint64_t))) != 0) { |
3558fd73 | 3755 | ZFS_EXIT(zsb); |
428870ff BB |
3756 | return (error); |
3757 | } | |
3558fd73 BB |
3758 | if (parent == zsb->z_shares_dir) { |
3759 | ZFS_EXIT(zsb); | |
a08ee875 | 3760 | return (SET_ERROR(EPERM)); |
428870ff BB |
3761 | } |
3762 | ||
3558fd73 | 3763 | if (zsb->z_utf8 && u8_validate(name, |
34dc7c2f | 3764 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3765 | ZFS_EXIT(zsb); |
a08ee875 | 3766 | return (SET_ERROR(EILSEQ)); |
34dc7c2f | 3767 | } |
3558fd73 | 3768 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
3769 | if (flags & FIGNORECASE) |
3770 | zf |= ZCILOOK; | |
3558fd73 | 3771 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f | 3772 | |
34dc7c2f BB |
3773 | /* |
3774 | * We do not support links between attributes and non-attributes | |
3775 | * because of the potential security risk of creating links | |
3776 | * into "normal" file space in order to circumvent restrictions | |
3777 | * imposed in attribute space. | |
3778 | */ | |
428870ff | 3779 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 3780 | ZFS_EXIT(zsb); |
a08ee875 | 3781 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
3782 | } |
3783 | ||
3558fd73 | 3784 | owner = zfs_fuid_map_id(zsb, szp->z_uid, cr, ZFS_OWNER); |
572e2857 | 3785 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
3558fd73 | 3786 | ZFS_EXIT(zsb); |
a08ee875 | 3787 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
3788 | } |
3789 | ||
149e873a | 3790 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
3558fd73 | 3791 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3792 | return (error); |
3793 | } | |
3794 | ||
428870ff | 3795 | top: |
34dc7c2f BB |
3796 | /* |
3797 | * Attempt to lock directory; fail if entry already exists. | |
3798 | */ | |
3799 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
3800 | if (error) { | |
3558fd73 | 3801 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3802 | return (error); |
3803 | } | |
3804 | ||
3558fd73 | 3805 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 3806 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 3807 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3808 | zfs_sa_upgrade_txholds(tx, szp); |
3809 | zfs_sa_upgrade_txholds(tx, dzp); | |
a08ee875 | 3810 | error = dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT); |
34dc7c2f BB |
3811 | if (error) { |
3812 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3813 | if (error == ERESTART) { |
a08ee875 | 3814 | waited = B_TRUE; |
34dc7c2f BB |
3815 | dmu_tx_wait(tx); |
3816 | dmu_tx_abort(tx); | |
3817 | goto top; | |
3818 | } | |
3819 | dmu_tx_abort(tx); | |
3558fd73 | 3820 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3821 | return (error); |
3822 | } | |
3823 | ||
3824 | error = zfs_link_create(dl, szp, tx, 0); | |
3825 | ||
3826 | if (error == 0) { | |
3827 | uint64_t txtype = TX_LINK; | |
3558fd73 | 3828 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
3829 | if (flags & FIGNORECASE) |
3830 | txtype |= TX_CI; | |
3558fd73 | 3831 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
3832 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); |
3833 | } | |
3834 | ||
3835 | dmu_tx_commit(tx); | |
3836 | ||
3837 | zfs_dirent_unlock(dl); | |
3838 | ||
3558fd73 | 3839 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3840 | zil_commit(zilog, 0); |
428870ff | 3841 | |
960e08fe BB |
3842 | zfs_inode_update(dzp); |
3843 | zfs_inode_update(szp); | |
3558fd73 | 3844 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3845 | return (error); |
3846 | } | |
e5c39b95 | 3847 | EXPORT_SYMBOL(zfs_link); |
34dc7c2f | 3848 | |
3c0e5c0f | 3849 | static void |
a08ee875 | 3850 | zfs_putpage_commit_cb(void *arg) |
3c0e5c0f BB |
3851 | { |
3852 | struct page *pp = arg; | |
3853 | ||
a08ee875 | 3854 | ClearPageError(pp); |
3c0e5c0f BB |
3855 | end_page_writeback(pp); |
3856 | } | |
3857 | ||
34dc7c2f | 3858 | /* |
3c0e5c0f BB |
3859 | * Push a page out to disk, once the page is on stable storage the |
3860 | * registered commit callback will be run as notification of completion. | |
34dc7c2f | 3861 | * |
3c0e5c0f BB |
3862 | * IN: ip - page mapped for inode. |
3863 | * pp - page to push (page is locked) | |
3864 | * wbc - writeback control data | |
34dc7c2f BB |
3865 | * |
3866 | * RETURN: 0 if success | |
3867 | * error code if failure | |
3868 | * | |
3c0e5c0f BB |
3869 | * Timestamps: |
3870 | * ip - ctime|mtime updated | |
34dc7c2f BB |
3871 | */ |
3872 | /* ARGSUSED */ | |
3c0e5c0f BB |
3873 | int |
3874 | zfs_putpage(struct inode *ip, struct page *pp, struct writeback_control *wbc) | |
34dc7c2f | 3875 | { |
3c0e5c0f BB |
3876 | znode_t *zp = ITOZ(ip); |
3877 | zfs_sb_t *zsb = ITOZSB(ip); | |
3878 | loff_t offset; | |
3879 | loff_t pgoff; | |
4c837f0d BB |
3880 | unsigned int pglen; |
3881 | rl_t *rl; | |
3c0e5c0f BB |
3882 | dmu_tx_t *tx; |
3883 | caddr_t va; | |
3884 | int err = 0; | |
3885 | uint64_t mtime[2], ctime[2]; | |
3886 | sa_bulk_attr_t bulk[3]; | |
3887 | int cnt = 0; | |
e10b0808 | 3888 | struct address_space *mapping; |
3c0e5c0f | 3889 | |
4c837f0d BB |
3890 | ZFS_ENTER(zsb); |
3891 | ZFS_VERIFY_ZP(zp); | |
d164b209 | 3892 | |
3c0e5c0f BB |
3893 | ASSERT(PageLocked(pp)); |
3894 | ||
a08ee875 LG |
3895 | pgoff = page_offset(pp); /* Page byte-offset in file */ |
3896 | offset = i_size_read(ip); /* File length in bytes */ | |
3897 | pglen = MIN(PAGE_CACHE_SIZE, /* Page length in bytes */ | |
3c0e5c0f BB |
3898 | P2ROUNDUP(offset, PAGE_CACHE_SIZE)-pgoff); |
3899 | ||
3900 | /* Page is beyond end of file */ | |
3901 | if (pgoff >= offset) { | |
3902 | unlock_page(pp); | |
4c837f0d | 3903 | ZFS_EXIT(zsb); |
3c0e5c0f BB |
3904 | return (0); |
3905 | } | |
3906 | ||
3907 | /* Truncate page length to end of file */ | |
3908 | if (pgoff + pglen > offset) | |
3909 | pglen = offset - pgoff; | |
3910 | ||
3911 | #if 0 | |
34dc7c2f | 3912 | /* |
3c0e5c0f BB |
3913 | * FIXME: Allow mmap writes past its quota. The correct fix |
3914 | * is to register a page_mkwrite() handler to count the page | |
3915 | * against its quota when it is about to be dirtied. | |
34dc7c2f | 3916 | */ |
dde471ef PJ |
3917 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
3918 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 | 3919 | err = EDQUOT; |
9babb374 | 3920 | } |
3c0e5c0f BB |
3921 | #endif |
3922 | ||
ea04106b AX |
3923 | /* |
3924 | * The ordering here is critical and must adhere to the following | |
3925 | * rules in order to avoid deadlocking in either zfs_read() or | |
3926 | * zfs_free_range() due to a lock inversion. | |
3927 | * | |
3928 | * 1) The page must be unlocked prior to acquiring the range lock. | |
3929 | * This is critical because zfs_read() calls find_lock_page() | |
3930 | * which may block on the page lock while holding the range lock. | |
3931 | * | |
3932 | * 2) Before setting or clearing write back on a page the range lock | |
3933 | * must be held in order to prevent a lock inversion with the | |
3934 | * zfs_free_range() function. | |
e10b0808 AX |
3935 | * |
3936 | * This presents a problem because upon entering this function the | |
3937 | * page lock is already held. To safely acquire the range lock the | |
3938 | * page lock must be dropped. This creates a window where another | |
3939 | * process could truncate, invalidate, dirty, or write out the page. | |
3940 | * | |
3941 | * Therefore, after successfully reacquiring the range and page locks | |
3942 | * the current page state is checked. In the common case everything | |
3943 | * will be as is expected and it can be written out. However, if | |
3944 | * the page state has changed it must be handled accordingly. | |
ea04106b | 3945 | */ |
e10b0808 AX |
3946 | mapping = pp->mapping; |
3947 | redirty_page_for_writepage(wbc, pp); | |
3c0e5c0f | 3948 | unlock_page(pp); |
e10b0808 | 3949 | |
4c837f0d | 3950 | rl = zfs_range_lock(zp, pgoff, pglen, RL_WRITER); |
e10b0808 AX |
3951 | lock_page(pp); |
3952 | ||
3953 | /* Page mapping changed or it was no longer dirty, we're done */ | |
3954 | if (unlikely((mapping != pp->mapping) || !PageDirty(pp))) { | |
3955 | unlock_page(pp); | |
3956 | zfs_range_unlock(rl); | |
3957 | ZFS_EXIT(zsb); | |
3958 | return (0); | |
3959 | } | |
3960 | ||
3961 | /* Another process started write block if required */ | |
3962 | if (PageWriteback(pp)) { | |
3963 | unlock_page(pp); | |
3964 | zfs_range_unlock(rl); | |
3965 | ||
3966 | if (wbc->sync_mode != WB_SYNC_NONE) | |
3967 | wait_on_page_writeback(pp); | |
3968 | ||
3969 | ZFS_EXIT(zsb); | |
3970 | return (0); | |
3971 | } | |
3972 | ||
3973 | /* Clear the dirty flag the required locks are held */ | |
3974 | if (!clear_page_dirty_for_io(pp)) { | |
3975 | unlock_page(pp); | |
3976 | zfs_range_unlock(rl); | |
3977 | ZFS_EXIT(zsb); | |
3978 | return (0); | |
3979 | } | |
3980 | ||
3981 | /* | |
3982 | * Counterpart for redirty_page_for_writepage() above. This page | |
3983 | * was in fact not skipped and should not be counted as if it were. | |
3984 | */ | |
3985 | wbc->pages_skipped--; | |
ea04106b | 3986 | set_page_writeback(pp); |
e10b0808 | 3987 | unlock_page(pp); |
3c0e5c0f | 3988 | |
ea04106b | 3989 | tx = dmu_tx_create(zsb->z_os); |
3c0e5c0f | 3990 | dmu_tx_hold_write(tx, zp->z_id, pgoff, pglen); |
428870ff BB |
3991 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3992 | zfs_sa_upgrade_txholds(tx, zp); | |
ea04106b | 3993 | |
fb5f0bc8 | 3994 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 3995 | if (err != 0) { |
3c0e5c0f | 3996 | if (err == ERESTART) |
34dc7c2f | 3997 | dmu_tx_wait(tx); |
3c0e5c0f | 3998 | |
34dc7c2f | 3999 | dmu_tx_abort(tx); |
a08ee875 LG |
4000 | __set_page_dirty_nobuffers(pp); |
4001 | ClearPageError(pp); | |
4002 | end_page_writeback(pp); | |
4c837f0d BB |
4003 | zfs_range_unlock(rl); |
4004 | ZFS_EXIT(zsb); | |
3c0e5c0f | 4005 | return (err); |
34dc7c2f BB |
4006 | } |
4007 | ||
dde471ef | 4008 | va = kmap(pp); |
3c0e5c0f BB |
4009 | ASSERT3U(pglen, <=, PAGE_CACHE_SIZE); |
4010 | dmu_write(zsb->z_os, zp->z_id, pgoff, pglen, va, tx); | |
dde471ef | 4011 | kunmap(pp); |
34dc7c2f | 4012 | |
3c0e5c0f BB |
4013 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
4014 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
4015 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_FLAGS(zsb), NULL, &zp->z_pflags, 8); | |
428870ff | 4016 | |
d3aa3ea9 BB |
4017 | /* Preserve the mtime and ctime provided by the inode */ |
4018 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4019 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
4020 | zp->z_atime_dirty = 0; | |
4021 | zp->z_seq++; | |
4022 | ||
4023 | err = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4024 | ||
a08ee875 LG |
4025 | zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, pgoff, pglen, 0, |
4026 | zfs_putpage_commit_cb, pp); | |
45d1cae3 | 4027 | dmu_tx_commit(tx); |
d3aa3ea9 | 4028 | |
4c837f0d | 4029 | zfs_range_unlock(rl); |
34dc7c2f | 4030 | |
a08ee875 LG |
4031 | if (wbc->sync_mode != WB_SYNC_NONE) { |
4032 | /* | |
4033 | * Note that this is rarely called under writepages(), because | |
4034 | * writepages() normally handles the entire commit for | |
4035 | * performance reasons. | |
4036 | */ | |
ea04106b AX |
4037 | if (zsb->z_log != NULL) |
4038 | zil_commit(zsb->z_log, zp->z_id); | |
2b286136 | 4039 | } |
3c0e5c0f | 4040 | |
4c837f0d | 4041 | ZFS_EXIT(zsb); |
3c0e5c0f | 4042 | return (err); |
34dc7c2f BB |
4043 | } |
4044 | ||
8780c539 BB |
4045 | /* |
4046 | * Update the system attributes when the inode has been dirtied. For the | |
a08ee875 | 4047 | * moment we only update the mode, atime, mtime, and ctime. |
8780c539 BB |
4048 | */ |
4049 | int | |
4050 | zfs_dirty_inode(struct inode *ip, int flags) | |
4051 | { | |
4052 | znode_t *zp = ITOZ(ip); | |
4053 | zfs_sb_t *zsb = ITOZSB(ip); | |
4054 | dmu_tx_t *tx; | |
a08ee875 LG |
4055 | uint64_t mode, atime[2], mtime[2], ctime[2]; |
4056 | sa_bulk_attr_t bulk[4]; | |
8780c539 BB |
4057 | int error; |
4058 | int cnt = 0; | |
4059 | ||
ea04106b AX |
4060 | if (zfs_is_readonly(zsb) || dmu_objset_is_snapshot(zsb->z_os)) |
4061 | return (0); | |
4062 | ||
8780c539 BB |
4063 | ZFS_ENTER(zsb); |
4064 | ZFS_VERIFY_ZP(zp); | |
4065 | ||
4066 | tx = dmu_tx_create(zsb->z_os); | |
4067 | ||
4068 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4069 | zfs_sa_upgrade_txholds(tx, zp); | |
4070 | ||
4071 | error = dmu_tx_assign(tx, TXG_WAIT); | |
4072 | if (error) { | |
4073 | dmu_tx_abort(tx); | |
4074 | goto out; | |
4075 | } | |
4076 | ||
4077 | mutex_enter(&zp->z_lock); | |
a08ee875 | 4078 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MODE(zsb), NULL, &mode, 8); |
8780c539 BB |
4079 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_ATIME(zsb), NULL, &atime, 16); |
4080 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); | |
4081 | SA_ADD_BULK_ATTR(bulk, cnt, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
4082 | ||
a08ee875 | 4083 | /* Preserve the mode, mtime and ctime provided by the inode */ |
8780c539 BB |
4084 | ZFS_TIME_ENCODE(&ip->i_atime, atime); |
4085 | ZFS_TIME_ENCODE(&ip->i_mtime, mtime); | |
4086 | ZFS_TIME_ENCODE(&ip->i_ctime, ctime); | |
a08ee875 LG |
4087 | mode = ip->i_mode; |
4088 | ||
4089 | zp->z_mode = mode; | |
8780c539 BB |
4090 | zp->z_atime_dirty = 0; |
4091 | ||
4092 | error = sa_bulk_update(zp->z_sa_hdl, bulk, cnt, tx); | |
4093 | mutex_exit(&zp->z_lock); | |
4094 | ||
4095 | dmu_tx_commit(tx); | |
4096 | out: | |
4097 | ZFS_EXIT(zsb); | |
4098 | return (error); | |
4099 | } | |
4100 | EXPORT_SYMBOL(zfs_dirty_inode); | |
4101 | ||
34dc7c2f BB |
4102 | /*ARGSUSED*/ |
4103 | void | |
c0d35759 | 4104 | zfs_inactive(struct inode *ip) |
34dc7c2f | 4105 | { |
c0d35759 BB |
4106 | znode_t *zp = ITOZ(ip); |
4107 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4108 | int error; |
e10b0808 | 4109 | int need_unlock = 0; |
34dc7c2f | 4110 | |
e10b0808 AX |
4111 | /* Only read lock if we haven't already write locked, e.g. rollback */ |
4112 | if (!RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)) { | |
4113 | need_unlock = 1; | |
4114 | rw_enter(&zsb->z_teardown_inactive_lock, RW_READER); | |
ebe7e575 | 4115 | } |
c0d35759 | 4116 | if (zp->z_sa_hdl == NULL) { |
e10b0808 AX |
4117 | if (need_unlock) |
4118 | rw_exit(&zsb->z_teardown_inactive_lock); | |
c0d35759 | 4119 | return; |
34dc7c2f BB |
4120 | } |
4121 | ||
4122 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
c0d35759 | 4123 | dmu_tx_t *tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 4124 | |
428870ff BB |
4125 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4126 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4127 | error = dmu_tx_assign(tx, TXG_WAIT); |
4128 | if (error) { | |
4129 | dmu_tx_abort(tx); | |
4130 | } else { | |
34dc7c2f | 4131 | mutex_enter(&zp->z_lock); |
3558fd73 | 4132 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), |
428870ff | 4133 | (void *)&zp->z_atime, sizeof (zp->z_atime), tx); |
34dc7c2f BB |
4134 | zp->z_atime_dirty = 0; |
4135 | mutex_exit(&zp->z_lock); | |
4136 | dmu_tx_commit(tx); | |
4137 | } | |
4138 | } | |
4139 | ||
4140 | zfs_zinactive(zp); | |
e10b0808 AX |
4141 | if (need_unlock) |
4142 | rw_exit(&zsb->z_teardown_inactive_lock); | |
34dc7c2f | 4143 | } |
e5c39b95 | 4144 | EXPORT_SYMBOL(zfs_inactive); |
34dc7c2f BB |
4145 | |
4146 | /* | |
4147 | * Bounds-check the seek operation. | |
4148 | * | |
3558fd73 | 4149 | * IN: ip - inode seeking within |
34dc7c2f BB |
4150 | * ooff - old file offset |
4151 | * noffp - pointer to new file offset | |
4152 | * ct - caller context | |
4153 | * | |
4154 | * RETURN: 0 if success | |
4155 | * EINVAL if new offset invalid | |
4156 | */ | |
4157 | /* ARGSUSED */ | |
3558fd73 | 4158 | int |
9623f736 | 4159 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 4160 | { |
3558fd73 | 4161 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
4162 | return (0); |
4163 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4164 | } | |
3558fd73 | 4165 | EXPORT_SYMBOL(zfs_seek); |
34dc7c2f | 4166 | |
34dc7c2f | 4167 | /* |
dde471ef | 4168 | * Fill pages with data from the disk. |
34dc7c2f BB |
4169 | */ |
4170 | static int | |
dde471ef | 4171 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 4172 | { |
a08ee875 LG |
4173 | znode_t *zp = ITOZ(ip); |
4174 | zfs_sb_t *zsb = ITOZSB(ip); | |
4175 | objset_t *os; | |
dde471ef | 4176 | struct page *cur_pp; |
a08ee875 LG |
4177 | u_offset_t io_off, total; |
4178 | size_t io_len; | |
4179 | loff_t i_size; | |
4180 | unsigned page_idx; | |
4181 | int err; | |
34dc7c2f | 4182 | |
a08ee875 | 4183 | os = zsb->z_os; |
dde471ef PJ |
4184 | io_len = nr_pages << PAGE_CACHE_SHIFT; |
4185 | i_size = i_size_read(ip); | |
4186 | io_off = page_offset(pl[0]); | |
4187 | ||
4188 | if (io_off + io_len > i_size) | |
4189 | io_len = i_size - io_off; | |
34dc7c2f BB |
4190 | |
4191 | /* | |
dde471ef | 4192 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 4193 | */ |
dde471ef PJ |
4194 | page_idx = 0; |
4195 | cur_pp = pl[0]; | |
34dc7c2f | 4196 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
4197 | caddr_t va; |
4198 | ||
dde471ef | 4199 | va = kmap(cur_pp); |
9babb374 BB |
4200 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4201 | DMU_READ_PREFETCH); | |
dde471ef | 4202 | kunmap(cur_pp); |
34dc7c2f | 4203 | if (err) { |
b128c09f BB |
4204 | /* convert checksum errors into IO errors */ |
4205 | if (err == ECKSUM) | |
a08ee875 | 4206 | err = SET_ERROR(EIO); |
34dc7c2f BB |
4207 | return (err); |
4208 | } | |
dde471ef | 4209 | cur_pp = pl[++page_idx]; |
34dc7c2f | 4210 | } |
d164b209 | 4211 | |
34dc7c2f BB |
4212 | return (0); |
4213 | } | |
4214 | ||
4215 | /* | |
dde471ef | 4216 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 4217 | * |
dde471ef PJ |
4218 | * IN: ip - inode of file to get data from. |
4219 | * pl - list of pages to read | |
4220 | * nr_pages - number of pages to read | |
34dc7c2f | 4221 | * |
a08ee875 | 4222 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4223 | * |
4224 | * Timestamps: | |
4225 | * vp - atime updated | |
4226 | */ | |
4227 | /* ARGSUSED */ | |
dde471ef PJ |
4228 | int |
4229 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 4230 | { |
dde471ef PJ |
4231 | znode_t *zp = ITOZ(ip); |
4232 | zfs_sb_t *zsb = ITOZSB(ip); | |
4233 | int err; | |
d164b209 | 4234 | |
d164b209 BB |
4235 | if (pl == NULL) |
4236 | return (0); | |
34dc7c2f | 4237 | |
dde471ef | 4238 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4239 | ZFS_VERIFY_ZP(zp); |
4240 | ||
dde471ef | 4241 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 4242 | |
dde471ef PJ |
4243 | if (!err) |
4244 | ZFS_ACCESSTIME_STAMP(zsb, zp); | |
34dc7c2f | 4245 | |
dde471ef | 4246 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4247 | return (err); |
4248 | } | |
dde471ef | 4249 | EXPORT_SYMBOL(zfs_getpage); |
34dc7c2f BB |
4250 | |
4251 | /* | |
e2e7aa2d | 4252 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 4253 | * |
e2e7aa2d BB |
4254 | * IN: ip - inode of the file to mmap |
4255 | * off - file offset | |
4256 | * addrp - start address in memory region | |
4257 | * len - length of memory region | |
4258 | * vm_flags- address flags | |
34dc7c2f | 4259 | * |
e2e7aa2d BB |
4260 | * RETURN: 0 if success |
4261 | * error code if failure | |
34dc7c2f BB |
4262 | */ |
4263 | /*ARGSUSED*/ | |
e2e7aa2d BB |
4264 | int |
4265 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
4266 | unsigned long vm_flags) | |
34dc7c2f | 4267 | { |
e2e7aa2d BB |
4268 | znode_t *zp = ITOZ(ip); |
4269 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4270 | |
e2e7aa2d | 4271 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4272 | ZFS_VERIFY_ZP(zp); |
4273 | ||
e2e7aa2d | 4274 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4275 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
e2e7aa2d | 4276 | ZFS_EXIT(zsb); |
a08ee875 | 4277 | return (SET_ERROR(EPERM)); |
34dc7c2f BB |
4278 | } |
4279 | ||
e2e7aa2d | 4280 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4281 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
e2e7aa2d | 4282 | ZFS_EXIT(zsb); |
a08ee875 | 4283 | return (SET_ERROR(EACCES)); |
34dc7c2f BB |
4284 | } |
4285 | ||
34dc7c2f | 4286 | if (off < 0 || len > MAXOFFSET_T - off) { |
e2e7aa2d | 4287 | ZFS_EXIT(zsb); |
a08ee875 | 4288 | return (SET_ERROR(ENXIO)); |
34dc7c2f BB |
4289 | } |
4290 | ||
e2e7aa2d | 4291 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4292 | return (0); |
4293 | } | |
e2e7aa2d | 4294 | EXPORT_SYMBOL(zfs_map); |
34dc7c2f | 4295 | |
3558fd73 BB |
4296 | /* |
4297 | * convoff - converts the given data (start, whence) to the | |
4298 | * given whence. | |
4299 | */ | |
4300 | int | |
4301 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4302 | { | |
5484965a | 4303 | vattr_t vap; |
3558fd73 BB |
4304 | int error; |
4305 | ||
4306 | if ((lckdat->l_whence == 2) || (whence == 2)) { | |
5484965a | 4307 | if ((error = zfs_getattr(ip, &vap, 0, CRED()) != 0)) |
3558fd73 BB |
4308 | return (error); |
4309 | } | |
4310 | ||
4311 | switch (lckdat->l_whence) { | |
4312 | case 1: | |
4313 | lckdat->l_start += offset; | |
4314 | break; | |
4315 | case 2: | |
5484965a | 4316 | lckdat->l_start += vap.va_size; |
3558fd73 BB |
4317 | /* FALLTHRU */ |
4318 | case 0: | |
4319 | break; | |
4320 | default: | |
a08ee875 | 4321 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4322 | } |
4323 | ||
4324 | if (lckdat->l_start < 0) | |
a08ee875 | 4325 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4326 | |
4327 | switch (whence) { | |
4328 | case 1: | |
4329 | lckdat->l_start -= offset; | |
4330 | break; | |
4331 | case 2: | |
5484965a | 4332 | lckdat->l_start -= vap.va_size; |
3558fd73 BB |
4333 | /* FALLTHRU */ |
4334 | case 0: | |
4335 | break; | |
4336 | default: | |
a08ee875 | 4337 | return (SET_ERROR(EINVAL)); |
3558fd73 BB |
4338 | } |
4339 | ||
4340 | lckdat->l_whence = (short)whence; | |
4341 | return (0); | |
4342 | } | |
4343 | ||
34dc7c2f BB |
4344 | /* |
4345 | * Free or allocate space in a file. Currently, this function only | |
4346 | * supports the `F_FREESP' command. However, this command is somewhat | |
4347 | * misnamed, as its functionality includes the ability to allocate as | |
4348 | * well as free space. | |
4349 | * | |
3558fd73 | 4350 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4351 | * cmd - action to take (only F_FREESP supported). |
4352 | * bfp - section of file to free/alloc. | |
4353 | * flag - current file open mode flags. | |
4354 | * offset - current file offset. | |
4355 | * cr - credentials of caller [UNUSED]. | |
34dc7c2f | 4356 | * |
a08ee875 | 4357 | * RETURN: 0 on success, error code on failure. |
34dc7c2f BB |
4358 | * |
4359 | * Timestamps: | |
3558fd73 | 4360 | * ip - ctime|mtime updated |
34dc7c2f BB |
4361 | */ |
4362 | /* ARGSUSED */ | |
e5c39b95 | 4363 | int |
3558fd73 BB |
4364 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4365 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4366 | { |
3558fd73 BB |
4367 | znode_t *zp = ITOZ(ip); |
4368 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4369 | uint64_t off, len; |
4370 | int error; | |
4371 | ||
3558fd73 | 4372 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4373 | ZFS_VERIFY_ZP(zp); |
4374 | ||
34dc7c2f | 4375 | if (cmd != F_FREESP) { |
3558fd73 | 4376 | ZFS_EXIT(zsb); |
a08ee875 | 4377 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4378 | } |
4379 | ||
3558fd73 BB |
4380 | if ((error = convoff(ip, bfp, 0, offset))) { |
4381 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
4382 | return (error); |
4383 | } | |
4384 | ||
4385 | if (bfp->l_len < 0) { | |
3558fd73 | 4386 | ZFS_EXIT(zsb); |
a08ee875 | 4387 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
4388 | } |
4389 | ||
aec69371 ED |
4390 | /* |
4391 | * Permissions aren't checked on Solaris because on this OS | |
4392 | * zfs_space() can only be called with an opened file handle. | |
4393 | * On Linux we can get here through truncate_range() which | |
4394 | * operates directly on inodes, so we need to check access rights. | |
4395 | */ | |
4396 | if ((error = zfs_zaccess(zp, ACE_WRITE_DATA, 0, B_FALSE, cr))) { | |
4397 | ZFS_EXIT(zsb); | |
4398 | return (error); | |
4399 | } | |
4400 | ||
34dc7c2f BB |
4401 | off = bfp->l_start; |
4402 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4403 | ||
b128c09f | 4404 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4405 | |
3558fd73 | 4406 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4407 | return (error); |
4408 | } | |
e5c39b95 | 4409 | EXPORT_SYMBOL(zfs_space); |
34dc7c2f BB |
4410 | |
4411 | /*ARGSUSED*/ | |
e5c39b95 | 4412 | int |
3558fd73 | 4413 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4414 | { |
3558fd73 BB |
4415 | znode_t *zp = ITOZ(ip); |
4416 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4417 | uint32_t gen; |
428870ff | 4418 | uint64_t gen64; |
34dc7c2f BB |
4419 | uint64_t object = zp->z_id; |
4420 | zfid_short_t *zfid; | |
428870ff | 4421 | int size, i, error; |
34dc7c2f | 4422 | |
3558fd73 | 4423 | ZFS_ENTER(zsb); |
34dc7c2f | 4424 | ZFS_VERIFY_ZP(zp); |
428870ff | 4425 | |
3558fd73 | 4426 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), |
428870ff | 4427 | &gen64, sizeof (uint64_t))) != 0) { |
3558fd73 | 4428 | ZFS_EXIT(zsb); |
428870ff BB |
4429 | return (error); |
4430 | } | |
4431 | ||
4432 | gen = (uint32_t)gen64; | |
34dc7c2f | 4433 | |
3558fd73 | 4434 | size = (zsb->z_parent != zsb) ? LONG_FID_LEN : SHORT_FID_LEN; |
34dc7c2f BB |
4435 | if (fidp->fid_len < size) { |
4436 | fidp->fid_len = size; | |
3558fd73 | 4437 | ZFS_EXIT(zsb); |
a08ee875 | 4438 | return (SET_ERROR(ENOSPC)); |
34dc7c2f BB |
4439 | } |
4440 | ||
4441 | zfid = (zfid_short_t *)fidp; | |
4442 | ||
4443 | zfid->zf_len = size; | |
4444 | ||
4445 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4446 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4447 | ||
4448 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4449 | if (gen == 0) | |
4450 | gen = 1; | |
4451 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4452 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4453 | ||
4454 | if (size == LONG_FID_LEN) { | |
3558fd73 | 4455 | uint64_t objsetid = dmu_objset_id(zsb->z_os); |
34dc7c2f BB |
4456 | zfid_long_t *zlfid; |
4457 | ||
4458 | zlfid = (zfid_long_t *)fidp; | |
4459 | ||
4460 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
4461 | zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); | |
4462 | ||
4463 | /* XXX - this should be the generation number for the objset */ | |
4464 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
4465 | zlfid->zf_setgen[i] = 0; | |
4466 | } | |
4467 | ||
3558fd73 | 4468 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4469 | return (0); |
4470 | } | |
e5c39b95 | 4471 | EXPORT_SYMBOL(zfs_fid); |
34dc7c2f | 4472 | |
34dc7c2f | 4473 | /*ARGSUSED*/ |
e5c39b95 | 4474 | int |
3558fd73 | 4475 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4476 | { |
3558fd73 BB |
4477 | znode_t *zp = ITOZ(ip); |
4478 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4479 | int error; |
4480 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
4481 | ||
3558fd73 | 4482 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4483 | ZFS_VERIFY_ZP(zp); |
4484 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
3558fd73 | 4485 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4486 | |
4487 | return (error); | |
4488 | } | |
e5c39b95 | 4489 | EXPORT_SYMBOL(zfs_getsecattr); |
34dc7c2f BB |
4490 | |
4491 | /*ARGSUSED*/ | |
e5c39b95 | 4492 | int |
3558fd73 | 4493 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4494 | { |
3558fd73 BB |
4495 | znode_t *zp = ITOZ(ip); |
4496 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4497 | int error; |
4498 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
3558fd73 | 4499 | zilog_t *zilog = zsb->z_log; |
34dc7c2f | 4500 | |
3558fd73 | 4501 | ZFS_ENTER(zsb); |
34dc7c2f | 4502 | ZFS_VERIFY_ZP(zp); |
428870ff | 4503 | |
34dc7c2f | 4504 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 4505 | |
3558fd73 | 4506 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4507 | zil_commit(zilog, 0); |
428870ff | 4508 | |
3558fd73 | 4509 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4510 | return (error); |
4511 | } | |
e5c39b95 | 4512 | EXPORT_SYMBOL(zfs_setsecattr); |
34dc7c2f | 4513 | |
3558fd73 | 4514 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
4515 | /* |
4516 | * Tunable, both must be a power of 2. | |
4517 | * | |
4518 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
4519 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 4520 | * an arcbuf for a partial block read |
428870ff BB |
4521 | */ |
4522 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
4523 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
4524 | ||
4525 | /*ARGSUSED*/ | |
4526 | static int | |
3558fd73 | 4527 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 4528 | { |
3558fd73 BB |
4529 | znode_t *zp = ITOZ(ip); |
4530 | zfs_sb_t *zsb = ITOZSB(ip); | |
4531 | int max_blksz = zsb->z_max_blksz; | |
428870ff BB |
4532 | uio_t *uio = &xuio->xu_uio; |
4533 | ssize_t size = uio->uio_resid; | |
4534 | offset_t offset = uio->uio_loffset; | |
4535 | int blksz; | |
4536 | int fullblk, i; | |
4537 | arc_buf_t *abuf; | |
4538 | ssize_t maxsize; | |
4539 | int preamble, postamble; | |
4540 | ||
4541 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
a08ee875 | 4542 | return (SET_ERROR(EINVAL)); |
428870ff | 4543 | |
3558fd73 | 4544 | ZFS_ENTER(zsb); |
428870ff BB |
4545 | ZFS_VERIFY_ZP(zp); |
4546 | switch (ioflag) { | |
4547 | case UIO_WRITE: | |
4548 | /* | |
4549 | * Loan out an arc_buf for write if write size is bigger than | |
4550 | * max_blksz, and the file's block size is also max_blksz. | |
4551 | */ | |
4552 | blksz = max_blksz; | |
4553 | if (size < blksz || zp->z_blksz != blksz) { | |
3558fd73 | 4554 | ZFS_EXIT(zsb); |
a08ee875 | 4555 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4556 | } |
4557 | /* | |
4558 | * Caller requests buffers for write before knowing where the | |
4559 | * write offset might be (e.g. NFS TCP write). | |
4560 | */ | |
4561 | if (offset == -1) { | |
4562 | preamble = 0; | |
4563 | } else { | |
4564 | preamble = P2PHASE(offset, blksz); | |
4565 | if (preamble) { | |
4566 | preamble = blksz - preamble; | |
4567 | size -= preamble; | |
4568 | } | |
4569 | } | |
4570 | ||
4571 | postamble = P2PHASE(size, blksz); | |
4572 | size -= postamble; | |
4573 | ||
4574 | fullblk = size / blksz; | |
4575 | (void) dmu_xuio_init(xuio, | |
4576 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
4577 | |
4578 | /* | |
4579 | * Have to fix iov base/len for partial buffers. They | |
4580 | * currently represent full arc_buf's. | |
4581 | */ | |
4582 | if (preamble) { | |
4583 | /* data begins in the middle of the arc_buf */ | |
4584 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4585 | blksz); | |
4586 | ASSERT(abuf); | |
4587 | (void) dmu_xuio_add(xuio, abuf, | |
4588 | blksz - preamble, preamble); | |
4589 | } | |
4590 | ||
4591 | for (i = 0; i < fullblk; i++) { | |
4592 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4593 | blksz); | |
4594 | ASSERT(abuf); | |
4595 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
4596 | } | |
4597 | ||
4598 | if (postamble) { | |
4599 | /* data ends in the middle of the arc_buf */ | |
4600 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4601 | blksz); | |
4602 | ASSERT(abuf); | |
4603 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
4604 | } | |
4605 | break; | |
4606 | case UIO_READ: | |
4607 | /* | |
4608 | * Loan out an arc_buf for read if the read size is larger than | |
4609 | * the current file block size. Block alignment is not | |
4610 | * considered. Partial arc_buf will be loaned out for read. | |
4611 | */ | |
4612 | blksz = zp->z_blksz; | |
4613 | if (blksz < zcr_blksz_min) | |
4614 | blksz = zcr_blksz_min; | |
4615 | if (blksz > zcr_blksz_max) | |
4616 | blksz = zcr_blksz_max; | |
4617 | /* avoid potential complexity of dealing with it */ | |
4618 | if (blksz > max_blksz) { | |
3558fd73 | 4619 | ZFS_EXIT(zsb); |
a08ee875 | 4620 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4621 | } |
4622 | ||
4623 | maxsize = zp->z_size - uio->uio_loffset; | |
4624 | if (size > maxsize) | |
4625 | size = maxsize; | |
4626 | ||
3558fd73 BB |
4627 | if (size < blksz) { |
4628 | ZFS_EXIT(zsb); | |
a08ee875 | 4629 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4630 | } |
4631 | break; | |
4632 | default: | |
3558fd73 | 4633 | ZFS_EXIT(zsb); |
a08ee875 | 4634 | return (SET_ERROR(EINVAL)); |
428870ff BB |
4635 | } |
4636 | ||
4637 | uio->uio_extflg = UIO_XUIO; | |
4638 | XUIO_XUZC_RW(xuio) = ioflag; | |
3558fd73 | 4639 | ZFS_EXIT(zsb); |
428870ff BB |
4640 | return (0); |
4641 | } | |
4642 | ||
4643 | /*ARGSUSED*/ | |
4644 | static int | |
3558fd73 | 4645 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
4646 | { |
4647 | int i; | |
4648 | arc_buf_t *abuf; | |
4649 | int ioflag = XUIO_XUZC_RW(xuio); | |
4650 | ||
4651 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
4652 | ||
4653 | i = dmu_xuio_cnt(xuio); | |
4654 | while (i-- > 0) { | |
4655 | abuf = dmu_xuio_arcbuf(xuio, i); | |
4656 | /* | |
4657 | * if abuf == NULL, it must be a write buffer | |
4658 | * that has been returned in zfs_write(). | |
4659 | */ | |
4660 | if (abuf) | |
4661 | dmu_return_arcbuf(abuf); | |
4662 | ASSERT(abuf || ioflag == UIO_WRITE); | |
4663 | } | |
4664 | ||
4665 | dmu_xuio_fini(xuio); | |
4666 | return (0); | |
4667 | } | |
3558fd73 | 4668 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 BB |
4669 | |
4670 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
4671 | module_param(zfs_read_chunk_size, long, 0644); | |
4672 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
4673 | #endif |