]>
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. |
34dc7c2f BB |
23 | */ |
24 | ||
25 | /* Portions Copyright 2007 Jeremy Teo */ | |
428870ff | 26 | /* Portions Copyright 2010 Robert Milkowski */ |
34dc7c2f | 27 | |
60101509 | 28 | |
34dc7c2f BB |
29 | #include <sys/types.h> |
30 | #include <sys/param.h> | |
31 | #include <sys/time.h> | |
32 | #include <sys/systm.h> | |
33 | #include <sys/sysmacros.h> | |
34 | #include <sys/resource.h> | |
35 | #include <sys/vfs.h> | |
36 | #include <sys/vfs_opreg.h> | |
34dc7c2f BB |
37 | #include <sys/file.h> |
38 | #include <sys/stat.h> | |
39 | #include <sys/kmem.h> | |
40 | #include <sys/taskq.h> | |
41 | #include <sys/uio.h> | |
42 | #include <sys/vmsystm.h> | |
43 | #include <sys/atomic.h> | |
34dc7c2f | 44 | #include <vm/pvn.h> |
34dc7c2f BB |
45 | #include <sys/pathname.h> |
46 | #include <sys/cmn_err.h> | |
47 | #include <sys/errno.h> | |
48 | #include <sys/unistd.h> | |
49 | #include <sys/zfs_dir.h> | |
50 | #include <sys/zfs_acl.h> | |
51 | #include <sys/zfs_ioctl.h> | |
52 | #include <sys/fs/zfs.h> | |
53 | #include <sys/dmu.h> | |
428870ff | 54 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
55 | #include <sys/spa.h> |
56 | #include <sys/txg.h> | |
57 | #include <sys/dbuf.h> | |
58 | #include <sys/zap.h> | |
428870ff | 59 | #include <sys/sa.h> |
34dc7c2f BB |
60 | #include <sys/dirent.h> |
61 | #include <sys/policy.h> | |
62 | #include <sys/sunddi.h> | |
b128c09f | 63 | #include <sys/sid.h> |
bcf30822 | 64 | #include <sys/mode.h> |
34dc7c2f | 65 | #include "fs/fs_subr.h" |
34dc7c2f | 66 | #include <sys/zfs_fuid.h> |
428870ff | 67 | #include <sys/zfs_sa.h> |
e5c39b95 | 68 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
69 | #include <sys/dnlc.h> |
70 | #include <sys/zfs_rlock.h> | |
71 | #include <sys/extdirent.h> | |
72 | #include <sys/kidmap.h> | |
428870ff | 73 | #include <sys/cred.h> |
34dc7c2f | 74 | #include <sys/attr.h> |
218b8eaf | 75 | #include <sys/zpl.h> |
34dc7c2f BB |
76 | |
77 | /* | |
78 | * Programming rules. | |
79 | * | |
80 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
81 | * properly lock its in-core state, create a DMU transaction, do the work, | |
82 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
83 | * and wait for the intent log to commit if it is a synchronous operation. | |
84 | * Moreover, the vnode ops must work in both normal and log replay context. | |
85 | * The ordering of events is important to avoid deadlocks and references | |
86 | * to freed memory. The example below illustrates the following Big Rules: | |
87 | * | |
88 | * (1) A check must be made in each zfs thread for a mounted file system. | |
3558fd73 BB |
89 | * This is done avoiding races using ZFS_ENTER(zsb). |
90 | * A ZFS_EXIT(zsb) is needed before all returns. Any znodes | |
34dc7c2f BB |
91 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros |
92 | * can return EIO from the calling function. | |
93 | * | |
3558fd73 | 94 | * (2) iput() should always be the last thing except for zil_commit() |
34dc7c2f BB |
95 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: |
96 | * First, if it's the last reference, the vnode/znode | |
97 | * can be freed, so the zp may point to freed memory. Second, the last | |
98 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
99 | * pushing cached pages (which acquires range locks) and syncing out | |
100 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
101 | * which could deadlock the system if you were already holding one. | |
3558fd73 | 102 | * If you must call iput() within a tx then use iput_ASYNC(). |
34dc7c2f BB |
103 | * |
104 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
105 | * as they can span dmu_tx_assign() calls. | |
106 | * | |
fb5f0bc8 | 107 | * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign(). |
34dc7c2f BB |
108 | * This is critical because we don't want to block while holding locks. |
109 | * Note, in particular, that if a lock is sometimes acquired before | |
110 | * the tx assigns, and sometimes after (e.g. z_lock), then failing to | |
111 | * use a non-blocking assign can deadlock the system. The scenario: | |
112 | * | |
113 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
114 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
115 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
116 | * forever, because the previous txg can't quiesce until B's tx commits. | |
117 | * | |
3558fd73 | 118 | * If dmu_tx_assign() returns ERESTART and zsb->z_assign is TXG_NOWAIT, |
34dc7c2f BB |
119 | * then drop all locks, call dmu_tx_wait(), and try again. |
120 | * | |
121 | * (5) If the operation succeeded, generate the intent log entry for it | |
122 | * before dropping locks. This ensures that the ordering of events | |
123 | * in the intent log matches the order in which they actually occurred. | |
fb5f0bc8 BB |
124 | * During ZIL replay the zfs_log_* functions will update the sequence |
125 | * number to indicate the zil transaction has replayed. | |
34dc7c2f BB |
126 | * |
127 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
128 | * regardless of whether there were any errors. | |
129 | * | |
572e2857 | 130 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
131 | * to ensure that synchronous semantics are provided when necessary. |
132 | * | |
133 | * In general, this is how things should be ordered in each vnode op: | |
134 | * | |
3558fd73 | 135 | * ZFS_ENTER(zsb); // exit if unmounted |
34dc7c2f | 136 | * top: |
3558fd73 | 137 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may igrab()) |
34dc7c2f BB |
138 | * rw_enter(...); // grab any other locks you need |
139 | * tx = dmu_tx_create(...); // get DMU tx | |
140 | * dmu_tx_hold_*(); // hold each object you might modify | |
fb5f0bc8 | 141 | * error = dmu_tx_assign(tx, TXG_NOWAIT); // try to assign |
34dc7c2f BB |
142 | * if (error) { |
143 | * rw_exit(...); // drop locks | |
144 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 145 | * iput(...); // release held vnodes |
fb5f0bc8 | 146 | * if (error == ERESTART) { |
34dc7c2f BB |
147 | * dmu_tx_wait(tx); |
148 | * dmu_tx_abort(tx); | |
149 | * goto top; | |
150 | * } | |
151 | * dmu_tx_abort(tx); // abort DMU tx | |
3558fd73 | 152 | * ZFS_EXIT(zsb); // finished in zfs |
34dc7c2f BB |
153 | * return (error); // really out of space |
154 | * } | |
155 | * error = do_real_work(); // do whatever this VOP does | |
156 | * if (error == 0) | |
157 | * zfs_log_*(...); // on success, make ZIL entry | |
158 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
159 | * rw_exit(...); // drop locks | |
160 | * zfs_dirent_unlock(dl); // unlock directory entry | |
3558fd73 | 161 | * iput(...); // release held vnodes |
572e2857 | 162 | * zil_commit(zilog, foid); // synchronous when necessary |
3558fd73 | 163 | * ZFS_EXIT(zsb); // finished in zfs |
34dc7c2f BB |
164 | * return (error); // done, report error |
165 | */ | |
166 | ||
126400a1 BB |
167 | /* |
168 | * Virus scanning is unsupported. It would be possible to add a hook | |
169 | * here to performance the required virus scan. This could be done | |
170 | * entirely in the kernel or potentially as an update to invoke a | |
171 | * scanning utility. | |
172 | */ | |
173 | static int | |
174 | zfs_vscan(struct inode *ip, cred_t *cr, int async) | |
175 | { | |
176 | return (0); | |
177 | } | |
178 | ||
179 | /* ARGSUSED */ | |
180 | int | |
181 | zfs_open(struct inode *ip, int mode, int flag, cred_t *cr) | |
182 | { | |
183 | znode_t *zp = ITOZ(ip); | |
184 | zfs_sb_t *zsb = ITOZSB(ip); | |
185 | ||
186 | ZFS_ENTER(zsb); | |
187 | ZFS_VERIFY_ZP(zp); | |
188 | ||
189 | /* Honor ZFS_APPENDONLY file attribute */ | |
190 | if ((mode & FMODE_WRITE) && (zp->z_pflags & ZFS_APPENDONLY) && | |
191 | ((flag & O_APPEND) == 0)) { | |
192 | ZFS_EXIT(zsb); | |
193 | return (EPERM); | |
194 | } | |
195 | ||
196 | /* Virus scan eligible files on open */ | |
197 | if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && | |
198 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { | |
199 | if (zfs_vscan(ip, cr, 0) != 0) { | |
200 | ZFS_EXIT(zsb); | |
201 | return (EACCES); | |
202 | } | |
203 | } | |
204 | ||
205 | /* Keep a count of the synchronous opens in the znode */ | |
206 | if (flag & O_SYNC) | |
207 | atomic_inc_32(&zp->z_sync_cnt); | |
208 | ||
209 | ZFS_EXIT(zsb); | |
210 | return (0); | |
211 | } | |
212 | EXPORT_SYMBOL(zfs_open); | |
213 | ||
214 | /* ARGSUSED */ | |
215 | int | |
216 | zfs_close(struct inode *ip, int flag, cred_t *cr) | |
217 | { | |
218 | znode_t *zp = ITOZ(ip); | |
219 | zfs_sb_t *zsb = ITOZSB(ip); | |
220 | ||
221 | ZFS_ENTER(zsb); | |
222 | ZFS_VERIFY_ZP(zp); | |
223 | ||
e814770f BB |
224 | /* |
225 | * Zero the synchronous opens in the znode. Under Linux the | |
226 | * zfs_close() hook is not symmetric with zfs_open(), it is | |
227 | * only called once when the last reference is dropped. | |
228 | */ | |
126400a1 BB |
229 | if (flag & O_SYNC) |
230 | zp->z_sync_cnt = 0; | |
231 | ||
232 | if (!zfs_has_ctldir(zp) && zsb->z_vscan && S_ISREG(ip->i_mode) && | |
233 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) | |
234 | VERIFY(zfs_vscan(ip, cr, 1) == 0); | |
235 | ||
236 | ZFS_EXIT(zsb); | |
237 | return (0); | |
238 | } | |
239 | EXPORT_SYMBOL(zfs_close); | |
240 | ||
c0d35759 | 241 | #if defined(_KERNEL) |
34dc7c2f BB |
242 | /* |
243 | * When a file is memory mapped, we must keep the IO data synchronized | |
244 | * between the DMU cache and the memory mapped pages. What this means: | |
245 | * | |
246 | * On Write: If we find a memory mapped page, we write to *both* | |
247 | * the page and the dmu buffer. | |
34dc7c2f | 248 | */ |
d164b209 | 249 | static void |
c0d35759 BB |
250 | update_pages(struct inode *ip, int64_t start, int len, |
251 | objset_t *os, uint64_t oid) | |
34dc7c2f | 252 | { |
c0d35759 BB |
253 | struct address_space *mp = ip->i_mapping; |
254 | struct page *pp; | |
255 | uint64_t nbytes; | |
d164b209 | 256 | int64_t off; |
c0d35759 | 257 | void *pb; |
34dc7c2f | 258 | |
c0d35759 BB |
259 | off = start & (PAGE_CACHE_SIZE-1); |
260 | for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) { | |
261 | nbytes = MIN(PAGE_CACHE_SIZE - off, len); | |
34dc7c2f | 262 | |
c0d35759 BB |
263 | pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT); |
264 | if (pp) { | |
265 | if (mapping_writably_mapped(mp)) | |
266 | flush_dcache_page(pp); | |
34dc7c2f | 267 | |
c0d35759 BB |
268 | pb = kmap(pp); |
269 | (void) dmu_read(os, oid, start+off, nbytes, pb+off, | |
9babb374 | 270 | DMU_READ_PREFETCH); |
c0d35759 BB |
271 | kunmap(pp); |
272 | ||
273 | if (mapping_writably_mapped(mp)) | |
274 | flush_dcache_page(pp); | |
275 | ||
276 | mark_page_accessed(pp); | |
277 | SetPageUptodate(pp); | |
278 | ClearPageError(pp); | |
279 | unlock_page(pp); | |
280 | page_cache_release(pp); | |
34dc7c2f | 281 | } |
c0d35759 | 282 | |
d164b209 | 283 | len -= nbytes; |
34dc7c2f | 284 | off = 0; |
34dc7c2f | 285 | } |
34dc7c2f BB |
286 | } |
287 | ||
288 | /* | |
289 | * When a file is memory mapped, we must keep the IO data synchronized | |
290 | * between the DMU cache and the memory mapped pages. What this means: | |
291 | * | |
292 | * On Read: We "read" preferentially from memory mapped pages, | |
293 | * else we default from the dmu buffer. | |
294 | * | |
295 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
296 | * the file is memory mapped. | |
297 | */ | |
298 | static int | |
3558fd73 | 299 | mappedread(struct inode *ip, int nbytes, uio_t *uio) |
34dc7c2f | 300 | { |
c0d35759 BB |
301 | struct address_space *mp = ip->i_mapping; |
302 | struct page *pp; | |
3558fd73 BB |
303 | znode_t *zp = ITOZ(ip); |
304 | objset_t *os = ITOZSB(ip)->z_os; | |
34dc7c2f | 305 | int64_t start, off; |
c0d35759 | 306 | uint64_t bytes; |
34dc7c2f BB |
307 | int len = nbytes; |
308 | int error = 0; | |
c0d35759 | 309 | void *pb; |
34dc7c2f BB |
310 | |
311 | start = uio->uio_loffset; | |
c0d35759 BB |
312 | off = start & (PAGE_CACHE_SIZE-1); |
313 | for (start &= PAGE_CACHE_MASK; len > 0; start += PAGE_CACHE_SIZE) { | |
314 | bytes = MIN(PAGE_CACHE_SIZE - off, len); | |
315 | ||
316 | pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT); | |
317 | if (pp) { | |
318 | ASSERT(PageUptodate(pp)); | |
319 | ||
320 | pb = kmap(pp); | |
321 | error = uiomove(pb + off, bytes, UIO_READ, uio); | |
322 | kunmap(pp); | |
323 | ||
324 | if (mapping_writably_mapped(mp)) | |
325 | flush_dcache_page(pp); | |
326 | ||
327 | mark_page_accessed(pp); | |
328 | unlock_page(pp); | |
329 | page_cache_release(pp); | |
34dc7c2f BB |
330 | } else { |
331 | error = dmu_read_uio(os, zp->z_id, uio, bytes); | |
332 | } | |
c0d35759 | 333 | |
34dc7c2f BB |
334 | len -= bytes; |
335 | off = 0; | |
336 | if (error) | |
337 | break; | |
338 | } | |
339 | return (error); | |
340 | } | |
c0d35759 | 341 | #endif /* _KERNEL */ |
34dc7c2f | 342 | |
c409e464 | 343 | unsigned long zfs_read_chunk_size = 1024 * 1024; /* Tunable */ |
34dc7c2f BB |
344 | |
345 | /* | |
346 | * Read bytes from specified file into supplied buffer. | |
347 | * | |
3558fd73 | 348 | * IN: ip - inode of file to be read from. |
34dc7c2f BB |
349 | * uio - structure supplying read location, range info, |
350 | * and return buffer. | |
c0d35759 BB |
351 | * ioflag - FSYNC flags; used to provide FRSYNC semantics. |
352 | * O_DIRECT flag; used to bypass page cache. | |
34dc7c2f | 353 | * cr - credentials of caller. |
34dc7c2f BB |
354 | * |
355 | * OUT: uio - updated offset and range, buffer filled. | |
356 | * | |
357 | * RETURN: 0 if success | |
358 | * error code if failure | |
359 | * | |
360 | * Side Effects: | |
3558fd73 | 361 | * inode - atime updated if byte count > 0 |
34dc7c2f BB |
362 | */ |
363 | /* ARGSUSED */ | |
e5c39b95 | 364 | int |
3558fd73 | 365 | zfs_read(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 366 | { |
3558fd73 BB |
367 | znode_t *zp = ITOZ(ip); |
368 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
369 | objset_t *os; |
370 | ssize_t n, nbytes; | |
149e873a | 371 | int error = 0; |
34dc7c2f | 372 | rl_t *rl; |
3558fd73 | 373 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff | 374 | xuio_t *xuio = NULL; |
3558fd73 | 375 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f | 376 | |
3558fd73 | 377 | ZFS_ENTER(zsb); |
34dc7c2f | 378 | ZFS_VERIFY_ZP(zp); |
3558fd73 | 379 | os = zsb->z_os; |
34dc7c2f | 380 | |
428870ff | 381 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
3558fd73 | 382 | ZFS_EXIT(zsb); |
34dc7c2f BB |
383 | return (EACCES); |
384 | } | |
385 | ||
386 | /* | |
387 | * Validate file offset | |
388 | */ | |
389 | if (uio->uio_loffset < (offset_t)0) { | |
3558fd73 | 390 | ZFS_EXIT(zsb); |
34dc7c2f BB |
391 | return (EINVAL); |
392 | } | |
393 | ||
394 | /* | |
395 | * Fasttrack empty reads | |
396 | */ | |
397 | if (uio->uio_resid == 0) { | |
3558fd73 | 398 | ZFS_EXIT(zsb); |
34dc7c2f BB |
399 | return (0); |
400 | } | |
401 | ||
402 | /* | |
403 | * Check for mandatory locks | |
404 | */ | |
a89c3e0b RP |
405 | if (mandatory_lock(ip) && |
406 | !lock_may_read(ip, uio->uio_loffset, uio->uio_resid)) { | |
407 | ZFS_EXIT(zsb); | |
408 | return (EAGAIN); | |
34dc7c2f BB |
409 | } |
410 | ||
411 | /* | |
412 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
413 | */ | |
3558fd73 BB |
414 | if (ioflag & FRSYNC || zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
415 | zil_commit(zsb->z_log, zp->z_id); | |
34dc7c2f BB |
416 | |
417 | /* | |
418 | * Lock the range against changes. | |
419 | */ | |
420 | rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER); | |
421 | ||
422 | /* | |
423 | * If we are reading past end-of-file we can skip | |
424 | * to the end; but we might still need to set atime. | |
425 | */ | |
428870ff | 426 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
427 | error = 0; |
428 | goto out; | |
429 | } | |
430 | ||
428870ff BB |
431 | ASSERT(uio->uio_loffset < zp->z_size); |
432 | n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); | |
433 | ||
3558fd73 | 434 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
435 | if ((uio->uio_extflg == UIO_XUIO) && |
436 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
437 | int nblk; | |
438 | int blksz = zp->z_blksz; | |
439 | uint64_t offset = uio->uio_loffset; | |
440 | ||
441 | xuio = (xuio_t *)uio; | |
442 | if ((ISP2(blksz))) { | |
443 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
444 | blksz)) / blksz; | |
445 | } else { | |
446 | ASSERT(offset + n <= blksz); | |
447 | nblk = 1; | |
448 | } | |
449 | (void) dmu_xuio_init(xuio, nblk); | |
450 | ||
3558fd73 | 451 | if (vn_has_cached_data(ip)) { |
428870ff BB |
452 | /* |
453 | * For simplicity, we always allocate a full buffer | |
454 | * even if we only expect to read a portion of a block. | |
455 | */ | |
456 | while (--nblk >= 0) { | |
457 | (void) dmu_xuio_add(xuio, | |
458 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
459 | blksz), 0, blksz); | |
460 | } | |
461 | } | |
462 | } | |
3558fd73 | 463 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
464 | |
465 | while (n > 0) { | |
466 | nbytes = MIN(n, zfs_read_chunk_size - | |
467 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); | |
468 | ||
c0d35759 | 469 | if (zp->z_is_mapped && !(ioflag & O_DIRECT)) |
3558fd73 | 470 | error = mappedread(ip, nbytes, uio); |
34dc7c2f BB |
471 | else |
472 | error = dmu_read_uio(os, zp->z_id, uio, nbytes); | |
c0d35759 | 473 | |
b128c09f BB |
474 | if (error) { |
475 | /* convert checksum errors into IO errors */ | |
476 | if (error == ECKSUM) | |
477 | error = EIO; | |
34dc7c2f | 478 | break; |
b128c09f | 479 | } |
34dc7c2f BB |
480 | |
481 | n -= nbytes; | |
482 | } | |
34dc7c2f BB |
483 | out: |
484 | zfs_range_unlock(rl); | |
485 | ||
3558fd73 | 486 | ZFS_ACCESSTIME_STAMP(zsb, zp); |
960e08fe | 487 | zfs_inode_update(zp); |
3558fd73 | 488 | ZFS_EXIT(zsb); |
34dc7c2f BB |
489 | return (error); |
490 | } | |
e5c39b95 | 491 | EXPORT_SYMBOL(zfs_read); |
34dc7c2f | 492 | |
34dc7c2f BB |
493 | /* |
494 | * Write the bytes to a file. | |
495 | * | |
3558fd73 | 496 | * IN: ip - inode of file to be written to. |
34dc7c2f BB |
497 | * uio - structure supplying write location, range info, |
498 | * and data buffer. | |
499 | * ioflag - FAPPEND flag set if in append mode. | |
c0d35759 | 500 | * O_DIRECT flag; used to bypass page cache. |
34dc7c2f | 501 | * cr - credentials of caller. |
34dc7c2f BB |
502 | * |
503 | * OUT: uio - updated offset and range. | |
504 | * | |
505 | * RETURN: 0 if success | |
506 | * error code if failure | |
507 | * | |
508 | * Timestamps: | |
3558fd73 | 509 | * ip - ctime|mtime updated if byte count > 0 |
34dc7c2f | 510 | */ |
428870ff | 511 | |
34dc7c2f | 512 | /* ARGSUSED */ |
e5c39b95 | 513 | int |
3558fd73 | 514 | zfs_write(struct inode *ip, uio_t *uio, int ioflag, cred_t *cr) |
34dc7c2f | 515 | { |
3558fd73 BB |
516 | znode_t *zp = ITOZ(ip); |
517 | rlim64_t limit = uio->uio_limit; | |
34dc7c2f BB |
518 | ssize_t start_resid = uio->uio_resid; |
519 | ssize_t tx_bytes; | |
520 | uint64_t end_size; | |
521 | dmu_tx_t *tx; | |
3558fd73 | 522 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
523 | zilog_t *zilog; |
524 | offset_t woff; | |
525 | ssize_t n, nbytes; | |
526 | rl_t *rl; | |
3558fd73 BB |
527 | int max_blksz = zsb->z_max_blksz; |
528 | int error = 0; | |
9babb374 | 529 | arc_buf_t *abuf; |
3558fd73 | 530 | iovec_t *aiov = NULL; |
428870ff BB |
531 | xuio_t *xuio = NULL; |
532 | int i_iov = 0; | |
428870ff BB |
533 | iovec_t *iovp = uio->uio_iov; |
534 | int write_eof; | |
535 | int count = 0; | |
536 | sa_bulk_attr_t bulk[4]; | |
537 | uint64_t mtime[2], ctime[2]; | |
3558fd73 | 538 | ASSERTV(int iovcnt = uio->uio_iovcnt); |
34dc7c2f | 539 | |
34dc7c2f BB |
540 | /* |
541 | * Fasttrack empty write | |
542 | */ | |
543 | n = start_resid; | |
544 | if (n == 0) | |
545 | return (0); | |
546 | ||
547 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) | |
548 | limit = MAXOFFSET_T; | |
549 | ||
3558fd73 | 550 | ZFS_ENTER(zsb); |
34dc7c2f | 551 | ZFS_VERIFY_ZP(zp); |
b128c09f | 552 | |
3558fd73 BB |
553 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
554 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
555 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); | |
556 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, | |
428870ff BB |
557 | &zp->z_pflags, 8); |
558 | ||
b128c09f BB |
559 | /* |
560 | * If immutable or not appending then return EPERM | |
561 | */ | |
428870ff BB |
562 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
563 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
564 | (uio->uio_loffset < zp->z_size))) { | |
3558fd73 | 565 | ZFS_EXIT(zsb); |
b128c09f BB |
566 | return (EPERM); |
567 | } | |
568 | ||
3558fd73 | 569 | zilog = zsb->z_log; |
34dc7c2f | 570 | |
428870ff BB |
571 | /* |
572 | * Validate file offset | |
573 | */ | |
574 | woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; | |
575 | if (woff < 0) { | |
3558fd73 | 576 | ZFS_EXIT(zsb); |
428870ff BB |
577 | return (EINVAL); |
578 | } | |
579 | ||
580 | /* | |
581 | * Check for mandatory locks before calling zfs_range_lock() | |
582 | * in order to prevent a deadlock with locks set via fcntl(). | |
583 | */ | |
a89c3e0b | 584 | if (mandatory_lock(ip) && !lock_may_write(ip, woff, n)) { |
3558fd73 | 585 | ZFS_EXIT(zsb); |
a89c3e0b | 586 | return (EAGAIN); |
428870ff BB |
587 | } |
588 | ||
3558fd73 | 589 | #ifdef HAVE_UIO_ZEROCOPY |
34dc7c2f BB |
590 | /* |
591 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
592 | * don't hold up txg. | |
428870ff | 593 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 594 | */ |
428870ff BB |
595 | if ((uio->uio_extflg == UIO_XUIO) && |
596 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
597 | xuio = (xuio_t *)uio; | |
598 | else | |
572e2857 | 599 | uio_prefaultpages(MIN(n, max_blksz), uio); |
3558fd73 | 600 | #endif /* HAVE_UIO_ZEROCOPY */ |
34dc7c2f BB |
601 | |
602 | /* | |
603 | * If in append mode, set the io offset pointer to eof. | |
604 | */ | |
605 | if (ioflag & FAPPEND) { | |
606 | /* | |
428870ff BB |
607 | * Obtain an appending range lock to guarantee file append |
608 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f BB |
609 | */ |
610 | rl = zfs_range_lock(zp, 0, n, RL_APPEND); | |
428870ff | 611 | woff = rl->r_off; |
34dc7c2f | 612 | if (rl->r_len == UINT64_MAX) { |
428870ff BB |
613 | /* |
614 | * We overlocked the file because this write will cause | |
615 | * the file block size to increase. | |
616 | * Note that zp_size cannot change with this lock held. | |
617 | */ | |
618 | woff = zp->z_size; | |
34dc7c2f | 619 | } |
428870ff | 620 | uio->uio_loffset = woff; |
34dc7c2f | 621 | } else { |
34dc7c2f | 622 | /* |
428870ff BB |
623 | * Note that if the file block size will change as a result of |
624 | * this write, then this range lock will lock the entire file | |
625 | * so that we can re-write the block safely. | |
34dc7c2f BB |
626 | */ |
627 | rl = zfs_range_lock(zp, woff, n, RL_WRITER); | |
628 | } | |
629 | ||
630 | if (woff >= limit) { | |
631 | zfs_range_unlock(rl); | |
3558fd73 | 632 | ZFS_EXIT(zsb); |
34dc7c2f BB |
633 | return (EFBIG); |
634 | } | |
635 | ||
636 | if ((woff + n) > limit || woff > (limit - n)) | |
637 | n = limit - woff; | |
638 | ||
428870ff BB |
639 | /* Will this write extend the file length? */ |
640 | write_eof = (woff + n > zp->z_size); | |
641 | ||
642 | end_size = MAX(zp->z_size, woff + n); | |
34dc7c2f BB |
643 | |
644 | /* | |
645 | * Write the file in reasonable size chunks. Each chunk is written | |
646 | * in a separate transaction; this keeps the intent log records small | |
647 | * and allows us to do more fine-grained space accounting. | |
648 | */ | |
649 | while (n > 0) { | |
9babb374 BB |
650 | abuf = NULL; |
651 | woff = uio->uio_loffset; | |
9babb374 | 652 | again: |
3558fd73 BB |
653 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
654 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 BB |
655 | if (abuf != NULL) |
656 | dmu_return_arcbuf(abuf); | |
657 | error = EDQUOT; | |
658 | break; | |
659 | } | |
660 | ||
428870ff BB |
661 | if (xuio && abuf == NULL) { |
662 | ASSERT(i_iov < iovcnt); | |
663 | aiov = &iovp[i_iov]; | |
664 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
665 | dmu_xuio_clear(xuio, i_iov); | |
428870ff BB |
666 | ASSERT((aiov->iov_base == abuf->b_data) || |
667 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
668 | aiov->iov_len == arc_buf_size(abuf))); | |
669 | i_iov++; | |
670 | } else if (abuf == NULL && n >= max_blksz && | |
671 | woff >= zp->z_size && | |
9babb374 BB |
672 | P2PHASE(woff, max_blksz) == 0 && |
673 | zp->z_blksz == max_blksz) { | |
428870ff BB |
674 | /* |
675 | * This write covers a full block. "Borrow" a buffer | |
676 | * from the dmu so that we can fill it before we enter | |
677 | * a transaction. This avoids the possibility of | |
678 | * holding up the transaction if the data copy hangs | |
679 | * up on a pagefault (e.g., from an NFS server mapping). | |
680 | */ | |
9babb374 BB |
681 | size_t cbytes; |
682 | ||
428870ff BB |
683 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
684 | max_blksz); | |
9babb374 BB |
685 | ASSERT(abuf != NULL); |
686 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
149e873a BB |
687 | if ((error = uiocopy(abuf->b_data, max_blksz, |
688 | UIO_WRITE, uio, &cbytes))) { | |
9babb374 BB |
689 | dmu_return_arcbuf(abuf); |
690 | break; | |
691 | } | |
692 | ASSERT(cbytes == max_blksz); | |
693 | } | |
694 | ||
34dc7c2f BB |
695 | /* |
696 | * Start a transaction. | |
697 | */ | |
3558fd73 | 698 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 699 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 700 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 701 | zfs_sa_upgrade_txholds(tx, zp); |
fb5f0bc8 | 702 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 703 | if (error) { |
fb5f0bc8 | 704 | if (error == ERESTART) { |
34dc7c2f BB |
705 | dmu_tx_wait(tx); |
706 | dmu_tx_abort(tx); | |
9babb374 | 707 | goto again; |
34dc7c2f BB |
708 | } |
709 | dmu_tx_abort(tx); | |
9babb374 BB |
710 | if (abuf != NULL) |
711 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
712 | break; |
713 | } | |
714 | ||
715 | /* | |
716 | * If zfs_range_lock() over-locked we grow the blocksize | |
717 | * and then reduce the lock range. This will only happen | |
718 | * on the first iteration since zfs_range_reduce() will | |
719 | * shrink down r_len to the appropriate size. | |
720 | */ | |
721 | if (rl->r_len == UINT64_MAX) { | |
722 | uint64_t new_blksz; | |
723 | ||
724 | if (zp->z_blksz > max_blksz) { | |
725 | ASSERT(!ISP2(zp->z_blksz)); | |
726 | new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE); | |
727 | } else { | |
728 | new_blksz = MIN(end_size, max_blksz); | |
729 | } | |
730 | zfs_grow_blocksize(zp, new_blksz, tx); | |
731 | zfs_range_reduce(rl, woff, n); | |
732 | } | |
733 | ||
734 | /* | |
735 | * XXX - should we really limit each write to z_max_blksz? | |
736 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
737 | */ | |
738 | nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); | |
34dc7c2f | 739 | |
9babb374 BB |
740 | if (abuf == NULL) { |
741 | tx_bytes = uio->uio_resid; | |
428870ff BB |
742 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
743 | uio, nbytes, tx); | |
9babb374 BB |
744 | tx_bytes -= uio->uio_resid; |
745 | } else { | |
746 | tx_bytes = nbytes; | |
428870ff BB |
747 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
748 | /* | |
749 | * If this is not a full block write, but we are | |
750 | * extending the file past EOF and this data starts | |
751 | * block-aligned, use assign_arcbuf(). Otherwise, | |
752 | * write via dmu_write(). | |
753 | */ | |
754 | if (tx_bytes < max_blksz && (!write_eof || | |
755 | aiov->iov_base != abuf->b_data)) { | |
756 | ASSERT(xuio); | |
3558fd73 | 757 | dmu_write(zsb->z_os, zp->z_id, woff, |
428870ff BB |
758 | aiov->iov_len, aiov->iov_base, tx); |
759 | dmu_return_arcbuf(abuf); | |
760 | xuio_stat_wbuf_copied(); | |
761 | } else { | |
762 | ASSERT(xuio || tx_bytes == max_blksz); | |
763 | dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl), | |
764 | woff, abuf, tx); | |
765 | } | |
9babb374 BB |
766 | ASSERT(tx_bytes <= uio->uio_resid); |
767 | uioskip(uio, tx_bytes); | |
768 | } | |
c0d35759 BB |
769 | |
770 | if (tx_bytes && zp->z_is_mapped && !(ioflag & O_DIRECT)) | |
771 | update_pages(ip, woff, tx_bytes, zsb->z_os, zp->z_id); | |
34dc7c2f BB |
772 | |
773 | /* | |
774 | * If we made no progress, we're done. If we made even | |
775 | * partial progress, update the znode and ZIL accordingly. | |
776 | */ | |
777 | if (tx_bytes == 0) { | |
3558fd73 | 778 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 779 | (void *)&zp->z_size, sizeof (uint64_t), tx); |
34dc7c2f BB |
780 | dmu_tx_commit(tx); |
781 | ASSERT(error != 0); | |
782 | break; | |
783 | } | |
784 | ||
785 | /* | |
786 | * Clear Set-UID/Set-GID bits on successful write if not | |
787 | * privileged and at least one of the excute bits is set. | |
788 | * | |
789 | * It would be nice to to this after all writes have | |
790 | * been done, but that would still expose the ISUID/ISGID | |
791 | * to another app after the partial write is committed. | |
792 | * | |
572e2857 BB |
793 | * Note: we don't call zfs_fuid_map_id() here because |
794 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
795 | */ |
796 | mutex_enter(&zp->z_acl_lock); | |
428870ff | 797 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 798 | (S_IXUSR >> 6))) != 0 && |
428870ff | 799 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 800 | secpolicy_vnode_setid_retain(cr, |
428870ff BB |
801 | (zp->z_mode & S_ISUID) != 0 && zp->z_uid == 0) != 0) { |
802 | uint64_t newmode; | |
803 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
804 | newmode = zp->z_mode; | |
3558fd73 | 805 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zsb), |
428870ff | 806 | (void *)&newmode, sizeof (uint64_t), tx); |
34dc7c2f BB |
807 | } |
808 | mutex_exit(&zp->z_acl_lock); | |
809 | ||
428870ff BB |
810 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, |
811 | B_TRUE); | |
34dc7c2f BB |
812 | |
813 | /* | |
814 | * Update the file size (zp_size) if it has changed; | |
815 | * account for possible concurrent updates. | |
816 | */ | |
428870ff BB |
817 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
818 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 819 | uio->uio_loffset); |
428870ff BB |
820 | ASSERT(error == 0); |
821 | } | |
572e2857 BB |
822 | /* |
823 | * If we are replaying and eof is non zero then force | |
824 | * the file size to the specified eof. Note, there's no | |
825 | * concurrency during replay. | |
826 | */ | |
3558fd73 BB |
827 | if (zsb->z_replay && zsb->z_replay_eof != 0) |
828 | zp->z_size = zsb->z_replay_eof; | |
572e2857 | 829 | |
428870ff BB |
830 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
831 | ||
34dc7c2f BB |
832 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag); |
833 | dmu_tx_commit(tx); | |
834 | ||
835 | if (error != 0) | |
836 | break; | |
837 | ASSERT(tx_bytes == nbytes); | |
838 | n -= nbytes; | |
572e2857 BB |
839 | |
840 | if (!xuio && n > 0) | |
841 | uio_prefaultpages(MIN(n, max_blksz), uio); | |
34dc7c2f BB |
842 | } |
843 | ||
844 | zfs_range_unlock(rl); | |
845 | ||
846 | /* | |
847 | * If we're in replay mode, or we made no progress, return error. | |
848 | * Otherwise, it's at least a partial write, so it's successful. | |
849 | */ | |
3558fd73 BB |
850 | if (zsb->z_replay || uio->uio_resid == start_resid) { |
851 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
852 | return (error); |
853 | } | |
854 | ||
428870ff | 855 | if (ioflag & (FSYNC | FDSYNC) || |
3558fd73 | 856 | zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 857 | zil_commit(zilog, zp->z_id); |
34dc7c2f | 858 | |
960e08fe | 859 | zfs_inode_update(zp); |
3558fd73 | 860 | ZFS_EXIT(zsb); |
34dc7c2f BB |
861 | return (0); |
862 | } | |
e5c39b95 | 863 | EXPORT_SYMBOL(zfs_write); |
34dc7c2f | 864 | |
3558fd73 BB |
865 | static void |
866 | iput_async(struct inode *ip, taskq_t *taskq) | |
867 | { | |
868 | ASSERT(atomic_read(&ip->i_count) > 0); | |
869 | if (atomic_read(&ip->i_count) == 1) | |
870 | taskq_dispatch(taskq, (task_func_t *)iput, ip, TQ_SLEEP); | |
871 | else | |
872 | iput(ip); | |
873 | } | |
874 | ||
34dc7c2f | 875 | void |
428870ff | 876 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 877 | { |
428870ff | 878 | znode_t *zp = zgd->zgd_private; |
3558fd73 | 879 | objset_t *os = ZTOZSB(zp)->z_os; |
428870ff BB |
880 | |
881 | if (zgd->zgd_db) | |
882 | dmu_buf_rele(zgd->zgd_db, zgd); | |
883 | ||
884 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f | 885 | |
9babb374 BB |
886 | /* |
887 | * Release the vnode asynchronously as we currently have the | |
888 | * txg stopped from syncing. | |
889 | */ | |
3558fd73 | 890 | iput_async(ZTOI(zp), dsl_pool_iput_taskq(dmu_objset_pool(os))); |
428870ff BB |
891 | |
892 | if (error == 0 && zgd->zgd_bp) | |
893 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
894 | ||
34dc7c2f BB |
895 | kmem_free(zgd, sizeof (zgd_t)); |
896 | } | |
897 | ||
45d1cae3 BB |
898 | #ifdef DEBUG |
899 | static int zil_fault_io = 0; | |
900 | #endif | |
901 | ||
34dc7c2f BB |
902 | /* |
903 | * Get data to generate a TX_WRITE intent log record. | |
904 | */ | |
905 | int | |
906 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
907 | { | |
3558fd73 BB |
908 | zfs_sb_t *zsb = arg; |
909 | objset_t *os = zsb->z_os; | |
34dc7c2f | 910 | znode_t *zp; |
428870ff BB |
911 | uint64_t object = lr->lr_foid; |
912 | uint64_t offset = lr->lr_offset; | |
913 | uint64_t size = lr->lr_length; | |
914 | blkptr_t *bp = &lr->lr_blkptr; | |
34dc7c2f | 915 | dmu_buf_t *db; |
34dc7c2f | 916 | zgd_t *zgd; |
34dc7c2f BB |
917 | int error = 0; |
918 | ||
428870ff BB |
919 | ASSERT(zio != NULL); |
920 | ASSERT(size != 0); | |
34dc7c2f BB |
921 | |
922 | /* | |
923 | * Nothing to do if the file has been removed | |
924 | */ | |
3558fd73 | 925 | if (zfs_zget(zsb, object, &zp) != 0) |
34dc7c2f BB |
926 | return (ENOENT); |
927 | if (zp->z_unlinked) { | |
9babb374 BB |
928 | /* |
929 | * Release the vnode asynchronously as we currently have the | |
930 | * txg stopped from syncing. | |
931 | */ | |
3558fd73 | 932 | iput_async(ZTOI(zp), dsl_pool_iput_taskq(dmu_objset_pool(os))); |
34dc7c2f BB |
933 | return (ENOENT); |
934 | } | |
935 | ||
428870ff | 936 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
3558fd73 | 937 | zgd->zgd_zilog = zsb->z_log; |
428870ff BB |
938 | zgd->zgd_private = zp; |
939 | ||
34dc7c2f BB |
940 | /* |
941 | * Write records come in two flavors: immediate and indirect. | |
942 | * For small writes it's cheaper to store the data with the | |
943 | * log record (immediate); for large writes it's cheaper to | |
944 | * sync the data and get a pointer to it (indirect) so that | |
945 | * we don't have to write the data twice. | |
946 | */ | |
947 | if (buf != NULL) { /* immediate write */ | |
428870ff | 948 | zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER); |
34dc7c2f | 949 | /* test for truncation needs to be done while range locked */ |
428870ff | 950 | if (offset >= zp->z_size) { |
34dc7c2f | 951 | error = ENOENT; |
428870ff BB |
952 | } else { |
953 | error = dmu_read(os, object, offset, size, buf, | |
954 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 955 | } |
428870ff | 956 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 957 | } else { /* indirect write */ |
34dc7c2f BB |
958 | /* |
959 | * Have to lock the whole block to ensure when it's | |
960 | * written out and it's checksum is being calculated | |
961 | * that no one can change the data. We need to re-check | |
962 | * blocksize after we get the lock in case it's changed! | |
963 | */ | |
964 | for (;;) { | |
428870ff BB |
965 | uint64_t blkoff; |
966 | size = zp->z_blksz; | |
967 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
968 | offset -= blkoff; | |
969 | zgd->zgd_rl = zfs_range_lock(zp, offset, size, | |
970 | RL_READER); | |
971 | if (zp->z_blksz == size) | |
34dc7c2f | 972 | break; |
428870ff BB |
973 | offset += blkoff; |
974 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f BB |
975 | } |
976 | /* test for truncation needs to be done while range locked */ | |
428870ff | 977 | if (lr->lr_offset >= zp->z_size) |
34dc7c2f | 978 | error = ENOENT; |
45d1cae3 BB |
979 | #ifdef DEBUG |
980 | if (zil_fault_io) { | |
981 | error = EIO; | |
982 | zil_fault_io = 0; | |
45d1cae3 | 983 | } |
45d1cae3 | 984 | #endif |
34dc7c2f | 985 | if (error == 0) |
428870ff BB |
986 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
987 | DMU_READ_NO_PREFETCH); | |
988 | ||
989 | if (error == 0) { | |
990 | zgd->zgd_db = db; | |
991 | zgd->zgd_bp = bp; | |
992 | ||
993 | ASSERT(db->db_offset == offset); | |
994 | ASSERT(db->db_size == size); | |
995 | ||
996 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
997 | zfs_get_done, zgd); | |
998 | ASSERT(error || lr->lr_length <= zp->z_blksz); | |
999 | ||
1000 | /* | |
1001 | * On success, we need to wait for the write I/O | |
1002 | * initiated by dmu_sync() to complete before we can | |
1003 | * release this dbuf. We will finish everything up | |
1004 | * in the zfs_get_done() callback. | |
1005 | */ | |
1006 | if (error == 0) | |
1007 | return (0); | |
1008 | ||
1009 | if (error == EALREADY) { | |
1010 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
1011 | error = 0; | |
1012 | } | |
1013 | } | |
34dc7c2f | 1014 | } |
428870ff BB |
1015 | |
1016 | zfs_get_done(zgd, error); | |
1017 | ||
34dc7c2f BB |
1018 | return (error); |
1019 | } | |
1020 | ||
1021 | /*ARGSUSED*/ | |
3558fd73 BB |
1022 | int |
1023 | zfs_access(struct inode *ip, int mode, int flag, cred_t *cr) | |
34dc7c2f | 1024 | { |
3558fd73 BB |
1025 | znode_t *zp = ITOZ(ip); |
1026 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
1027 | int error; |
1028 | ||
3558fd73 | 1029 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1030 | ZFS_VERIFY_ZP(zp); |
1031 | ||
1032 | if (flag & V_ACE_MASK) | |
1033 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1034 | else | |
1035 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1036 | ||
3558fd73 | 1037 | ZFS_EXIT(zsb); |
45d1cae3 BB |
1038 | return (error); |
1039 | } | |
3558fd73 | 1040 | EXPORT_SYMBOL(zfs_access); |
45d1cae3 | 1041 | |
34dc7c2f BB |
1042 | /* |
1043 | * Lookup an entry in a directory, or an extended attribute directory. | |
3558fd73 | 1044 | * If it exists, return a held inode reference for it. |
34dc7c2f | 1045 | * |
3558fd73 | 1046 | * IN: dip - inode of directory to search. |
34dc7c2f | 1047 | * nm - name of entry to lookup. |
34dc7c2f | 1048 | * flags - LOOKUP_XATTR set if looking for an attribute. |
34dc7c2f | 1049 | * cr - credentials of caller. |
34dc7c2f BB |
1050 | * direntflags - directory lookup flags |
1051 | * realpnp - returned pathname. | |
1052 | * | |
3558fd73 | 1053 | * OUT: ipp - inode of located entry, NULL if not found. |
34dc7c2f BB |
1054 | * |
1055 | * RETURN: 0 if success | |
1056 | * error code if failure | |
1057 | * | |
1058 | * Timestamps: | |
1059 | * NA | |
1060 | */ | |
1061 | /* ARGSUSED */ | |
e5c39b95 | 1062 | int |
3558fd73 BB |
1063 | zfs_lookup(struct inode *dip, char *nm, struct inode **ipp, int flags, |
1064 | cred_t *cr, int *direntflags, pathname_t *realpnp) | |
34dc7c2f | 1065 | { |
3558fd73 BB |
1066 | znode_t *zdp = ITOZ(dip); |
1067 | zfs_sb_t *zsb = ITOZSB(dip); | |
1068 | int error = 0; | |
45d1cae3 BB |
1069 | |
1070 | /* fast path */ | |
1071 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { | |
1072 | ||
3558fd73 | 1073 | if (!S_ISDIR(dip->i_mode)) { |
45d1cae3 | 1074 | return (ENOTDIR); |
428870ff | 1075 | } else if (zdp->z_sa_hdl == NULL) { |
45d1cae3 BB |
1076 | return (EIO); |
1077 | } | |
1078 | ||
1079 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1080 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1081 | if (!error) { | |
3558fd73 BB |
1082 | *ipp = dip; |
1083 | igrab(*ipp); | |
45d1cae3 BB |
1084 | return (0); |
1085 | } | |
1086 | return (error); | |
3558fd73 | 1087 | #ifdef HAVE_DNLC |
45d1cae3 BB |
1088 | } else { |
1089 | vnode_t *tvp = dnlc_lookup(dvp, nm); | |
1090 | ||
1091 | if (tvp) { | |
1092 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1093 | if (error) { | |
3558fd73 | 1094 | iput(tvp); |
45d1cae3 BB |
1095 | return (error); |
1096 | } | |
1097 | if (tvp == DNLC_NO_VNODE) { | |
3558fd73 | 1098 | iput(tvp); |
45d1cae3 BB |
1099 | return (ENOENT); |
1100 | } else { | |
1101 | *vpp = tvp; | |
1102 | return (specvp_check(vpp, cr)); | |
1103 | } | |
1104 | } | |
3558fd73 | 1105 | #endif /* HAVE_DNLC */ |
45d1cae3 BB |
1106 | } |
1107 | } | |
1108 | ||
3558fd73 | 1109 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1110 | ZFS_VERIFY_ZP(zdp); |
1111 | ||
3558fd73 | 1112 | *ipp = NULL; |
34dc7c2f BB |
1113 | |
1114 | if (flags & LOOKUP_XATTR) { | |
34dc7c2f BB |
1115 | /* |
1116 | * We don't allow recursive attributes.. | |
1117 | * Maybe someday we will. | |
1118 | */ | |
428870ff | 1119 | if (zdp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1120 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1121 | return (EINVAL); |
1122 | } | |
1123 | ||
3558fd73 BB |
1124 | if ((error = zfs_get_xattrdir(zdp, ipp, cr, flags))) { |
1125 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1126 | return (error); |
1127 | } | |
1128 | ||
1129 | /* | |
1130 | * Do we have permission to get into attribute directory? | |
1131 | */ | |
1132 | ||
3558fd73 | 1133 | if ((error = zfs_zaccess(ITOZ(*ipp), ACE_EXECUTE, 0, |
149e873a | 1134 | B_FALSE, cr))) { |
3558fd73 BB |
1135 | iput(*ipp); |
1136 | *ipp = NULL; | |
34dc7c2f BB |
1137 | } |
1138 | ||
3558fd73 | 1139 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1140 | return (error); |
1141 | } | |
1142 | ||
3558fd73 BB |
1143 | if (!S_ISDIR(dip->i_mode)) { |
1144 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1145 | return (ENOTDIR); |
1146 | } | |
1147 | ||
1148 | /* | |
1149 | * Check accessibility of directory. | |
1150 | */ | |
1151 | ||
149e873a | 1152 | if ((error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr))) { |
3558fd73 | 1153 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1154 | return (error); |
1155 | } | |
1156 | ||
3558fd73 | 1157 | if (zsb->z_utf8 && u8_validate(nm, strlen(nm), |
34dc7c2f | 1158 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1159 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1160 | return (EILSEQ); |
1161 | } | |
1162 | ||
3558fd73 BB |
1163 | error = zfs_dirlook(zdp, nm, ipp, flags, direntflags, realpnp); |
1164 | if ((error == 0) && (*ipp)) | |
1165 | zfs_inode_update(ITOZ(*ipp)); | |
34dc7c2f | 1166 | |
3558fd73 | 1167 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1168 | return (error); |
1169 | } | |
e5c39b95 | 1170 | EXPORT_SYMBOL(zfs_lookup); |
34dc7c2f BB |
1171 | |
1172 | /* | |
1173 | * Attempt to create a new entry in a directory. If the entry | |
1174 | * already exists, truncate the file if permissible, else return | |
3558fd73 | 1175 | * an error. Return the ip of the created or trunc'd file. |
34dc7c2f | 1176 | * |
3558fd73 | 1177 | * IN: dip - inode of directory to put new file entry in. |
34dc7c2f BB |
1178 | * name - name of new file entry. |
1179 | * vap - attributes of new file. | |
1180 | * excl - flag indicating exclusive or non-exclusive mode. | |
1181 | * mode - mode to open file with. | |
1182 | * cr - credentials of caller. | |
1183 | * flag - large file flag [UNUSED]. | |
3558fd73 | 1184 | * vsecp - ACL to be set |
34dc7c2f | 1185 | * |
3558fd73 | 1186 | * OUT: ipp - inode of created or trunc'd entry. |
34dc7c2f BB |
1187 | * |
1188 | * RETURN: 0 if success | |
1189 | * error code if failure | |
1190 | * | |
1191 | * Timestamps: | |
3558fd73 BB |
1192 | * dip - ctime|mtime updated if new entry created |
1193 | * ip - ctime|mtime always, atime if new | |
34dc7c2f BB |
1194 | */ |
1195 | ||
1196 | /* ARGSUSED */ | |
e5c39b95 | 1197 | int |
3558fd73 BB |
1198 | zfs_create(struct inode *dip, char *name, vattr_t *vap, int excl, |
1199 | int mode, struct inode **ipp, cred_t *cr, int flag, vsecattr_t *vsecp) | |
34dc7c2f | 1200 | { |
3558fd73 BB |
1201 | znode_t *zp, *dzp = ITOZ(dip); |
1202 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1203 | zilog_t *zilog; |
1204 | objset_t *os; | |
1205 | zfs_dirlock_t *dl; | |
1206 | dmu_tx_t *tx; | |
1207 | int error; | |
b128c09f | 1208 | uid_t uid; |
149e873a | 1209 | gid_t gid; |
428870ff | 1210 | zfs_acl_ids_t acl_ids; |
9babb374 | 1211 | boolean_t fuid_dirtied; |
428870ff | 1212 | boolean_t have_acl = B_FALSE; |
34dc7c2f BB |
1213 | |
1214 | /* | |
1215 | * If we have an ephemeral id, ACL, or XVATTR then | |
1216 | * make sure file system is at proper version | |
1217 | */ | |
1218 | ||
149e873a | 1219 | gid = crgetgid(cr); |
3558fd73 | 1220 | uid = crgetuid(cr); |
b128c09f | 1221 | |
3558fd73 BB |
1222 | if (zsb->z_use_fuids == B_FALSE && |
1223 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
34dc7c2f BB |
1224 | return (EINVAL); |
1225 | ||
3558fd73 | 1226 | ZFS_ENTER(zsb); |
34dc7c2f | 1227 | ZFS_VERIFY_ZP(dzp); |
3558fd73 BB |
1228 | os = zsb->z_os; |
1229 | zilog = zsb->z_log; | |
34dc7c2f | 1230 | |
3558fd73 | 1231 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 1232 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1233 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1234 | return (EILSEQ); |
1235 | } | |
1236 | ||
5484965a | 1237 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1238 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1239 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1240 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1241 | return (error); |
1242 | } | |
1243 | } | |
34dc7c2f | 1244 | |
3558fd73 BB |
1245 | top: |
1246 | *ipp = NULL; | |
34dc7c2f BB |
1247 | if (*name == '\0') { |
1248 | /* | |
1249 | * Null component name refers to the directory itself. | |
1250 | */ | |
3558fd73 | 1251 | igrab(dip); |
34dc7c2f BB |
1252 | zp = dzp; |
1253 | dl = NULL; | |
1254 | error = 0; | |
1255 | } else { | |
3558fd73 | 1256 | /* possible igrab(zp) */ |
34dc7c2f BB |
1257 | int zflg = 0; |
1258 | ||
1259 | if (flag & FIGNORECASE) | |
1260 | zflg |= ZCILOOK; | |
1261 | ||
1262 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1263 | NULL, NULL); | |
1264 | if (error) { | |
572e2857 BB |
1265 | if (have_acl) |
1266 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1267 | if (strcmp(name, "..") == 0) |
1268 | error = EISDIR; | |
3558fd73 | 1269 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1270 | return (error); |
1271 | } | |
1272 | } | |
428870ff | 1273 | |
34dc7c2f BB |
1274 | if (zp == NULL) { |
1275 | uint64_t txtype; | |
1276 | ||
1277 | /* | |
1278 | * Create a new file object and update the directory | |
1279 | * to reference it. | |
1280 | */ | |
149e873a | 1281 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
572e2857 BB |
1282 | if (have_acl) |
1283 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1284 | goto out; |
1285 | } | |
1286 | ||
1287 | /* | |
1288 | * We only support the creation of regular files in | |
1289 | * extended attribute directories. | |
1290 | */ | |
428870ff | 1291 | |
3558fd73 | 1292 | if ((dzp->z_pflags & ZFS_XATTR) && !S_ISREG(vap->va_mode)) { |
572e2857 BB |
1293 | if (have_acl) |
1294 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1295 | error = EINVAL; |
1296 | goto out; | |
1297 | } | |
1298 | ||
428870ff BB |
1299 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1300 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1301 | goto out; |
428870ff BB |
1302 | have_acl = B_TRUE; |
1303 | ||
3558fd73 | 1304 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1305 | zfs_acl_ids_free(&acl_ids); |
9babb374 BB |
1306 | error = EDQUOT; |
1307 | goto out; | |
1308 | } | |
1309 | ||
34dc7c2f | 1310 | tx = dmu_tx_create(os); |
428870ff BB |
1311 | |
1312 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1313 | ZFS_SA_BASE_ATTR_SIZE); | |
1314 | ||
3558fd73 | 1315 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1316 | if (fuid_dirtied) |
3558fd73 | 1317 | zfs_fuid_txhold(zsb, tx); |
34dc7c2f | 1318 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff | 1319 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
3558fd73 | 1320 | if (!zsb->z_use_sa && |
428870ff | 1321 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f | 1322 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1323 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1324 | } |
fb5f0bc8 | 1325 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
1326 | if (error) { |
1327 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1328 | if (error == ERESTART) { |
34dc7c2f BB |
1329 | dmu_tx_wait(tx); |
1330 | dmu_tx_abort(tx); | |
1331 | goto top; | |
1332 | } | |
428870ff | 1333 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1334 | dmu_tx_abort(tx); |
3558fd73 | 1335 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1336 | return (error); |
1337 | } | |
428870ff | 1338 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 BB |
1339 | |
1340 | if (fuid_dirtied) | |
3558fd73 | 1341 | zfs_fuid_sync(zsb, tx); |
9babb374 | 1342 | |
34dc7c2f BB |
1343 | (void) zfs_link_create(dl, zp, tx, ZNEW); |
1344 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); | |
1345 | if (flag & FIGNORECASE) | |
1346 | txtype |= TX_CI; | |
1347 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1348 | vsecp, acl_ids.z_fuidp, vap); |
1349 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1350 | dmu_tx_commit(tx); |
1351 | } else { | |
1352 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1353 | ||
572e2857 BB |
1354 | if (have_acl) |
1355 | zfs_acl_ids_free(&acl_ids); | |
1356 | have_acl = B_FALSE; | |
1357 | ||
34dc7c2f BB |
1358 | /* |
1359 | * A directory entry already exists for this name. | |
1360 | */ | |
1361 | /* | |
1362 | * Can't truncate an existing file if in exclusive mode. | |
1363 | */ | |
3558fd73 | 1364 | if (excl) { |
34dc7c2f BB |
1365 | error = EEXIST; |
1366 | goto out; | |
1367 | } | |
1368 | /* | |
1369 | * Can't open a directory for writing. | |
1370 | */ | |
3558fd73 | 1371 | if (S_ISDIR(ZTOI(zp)->i_mode)) { |
34dc7c2f BB |
1372 | error = EISDIR; |
1373 | goto out; | |
1374 | } | |
1375 | /* | |
1376 | * Verify requested access to file. | |
1377 | */ | |
1378 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1379 | goto out; | |
1380 | } | |
1381 | ||
1382 | mutex_enter(&dzp->z_lock); | |
1383 | dzp->z_seq++; | |
1384 | mutex_exit(&dzp->z_lock); | |
1385 | ||
1386 | /* | |
1387 | * Truncate regular files if requested. | |
1388 | */ | |
3558fd73 BB |
1389 | if (S_ISREG(ZTOI(zp)->i_mode) && |
1390 | (vap->va_mask & ATTR_SIZE) && (vap->va_size == 0)) { | |
b128c09f BB |
1391 | /* we can't hold any locks when calling zfs_freesp() */ |
1392 | zfs_dirent_unlock(dl); | |
1393 | dl = NULL; | |
34dc7c2f | 1394 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1395 | } |
1396 | } | |
1397 | out: | |
1398 | ||
1399 | if (dl) | |
1400 | zfs_dirent_unlock(dl); | |
1401 | ||
1402 | if (error) { | |
1403 | if (zp) | |
3558fd73 | 1404 | iput(ZTOI(zp)); |
34dc7c2f | 1405 | } else { |
960e08fe BB |
1406 | zfs_inode_update(dzp); |
1407 | zfs_inode_update(zp); | |
3558fd73 | 1408 | *ipp = ZTOI(zp); |
34dc7c2f | 1409 | } |
34dc7c2f | 1410 | |
3558fd73 | 1411 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1412 | zil_commit(zilog, 0); |
428870ff | 1413 | |
3558fd73 | 1414 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1415 | return (error); |
1416 | } | |
e5c39b95 | 1417 | EXPORT_SYMBOL(zfs_create); |
34dc7c2f BB |
1418 | |
1419 | /* | |
1420 | * Remove an entry from a directory. | |
1421 | * | |
3558fd73 | 1422 | * IN: dip - inode of directory to remove entry from. |
34dc7c2f BB |
1423 | * name - name of entry to remove. |
1424 | * cr - credentials of caller. | |
34dc7c2f BB |
1425 | * |
1426 | * RETURN: 0 if success | |
1427 | * error code if failure | |
1428 | * | |
1429 | * Timestamps: | |
3558fd73 BB |
1430 | * dip - ctime|mtime |
1431 | * ip - ctime (if nlink > 0) | |
34dc7c2f | 1432 | */ |
428870ff BB |
1433 | |
1434 | uint64_t null_xattr = 0; | |
1435 | ||
34dc7c2f | 1436 | /*ARGSUSED*/ |
e5c39b95 | 1437 | int |
3558fd73 | 1438 | zfs_remove(struct inode *dip, char *name, cred_t *cr) |
34dc7c2f | 1439 | { |
3558fd73 | 1440 | znode_t *zp, *dzp = ITOZ(dip); |
572e2857 | 1441 | znode_t *xzp; |
3558fd73 BB |
1442 | struct inode *ip; |
1443 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f | 1444 | zilog_t *zilog; |
3558fd73 BB |
1445 | uint64_t xattr_obj; |
1446 | uint64_t xattr_obj_unlinked = 0; | |
572e2857 | 1447 | uint64_t obj = 0; |
34dc7c2f BB |
1448 | zfs_dirlock_t *dl; |
1449 | dmu_tx_t *tx; | |
3558fd73 | 1450 | boolean_t unlinked; |
34dc7c2f BB |
1451 | uint64_t txtype; |
1452 | pathname_t *realnmp = NULL; | |
3558fd73 | 1453 | #ifdef HAVE_PN_UTILS |
34dc7c2f | 1454 | pathname_t realnm; |
3558fd73 | 1455 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
1456 | int error; |
1457 | int zflg = ZEXISTS; | |
1458 | ||
3558fd73 | 1459 | ZFS_ENTER(zsb); |
34dc7c2f | 1460 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1461 | zilog = zsb->z_log; |
34dc7c2f | 1462 | |
3558fd73 | 1463 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1464 | if (flags & FIGNORECASE) { |
1465 | zflg |= ZCILOOK; | |
1466 | pn_alloc(&realnm); | |
1467 | realnmp = &realnm; | |
1468 | } | |
3558fd73 | 1469 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
1470 | |
1471 | top: | |
572e2857 BB |
1472 | xattr_obj = 0; |
1473 | xzp = NULL; | |
34dc7c2f BB |
1474 | /* |
1475 | * Attempt to lock directory; fail if entry doesn't exist. | |
1476 | */ | |
149e873a BB |
1477 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1478 | NULL, realnmp))) { | |
3558fd73 | 1479 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1480 | if (realnmp) |
1481 | pn_free(realnmp); | |
3558fd73 BB |
1482 | #endif /* HAVE_PN_UTILS */ |
1483 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1484 | return (error); |
1485 | } | |
1486 | ||
3558fd73 | 1487 | ip = ZTOI(zp); |
34dc7c2f | 1488 | |
149e873a | 1489 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1490 | goto out; |
1491 | } | |
1492 | ||
1493 | /* | |
1494 | * Need to use rmdir for removing directories. | |
1495 | */ | |
3558fd73 | 1496 | if (S_ISDIR(ip->i_mode)) { |
34dc7c2f BB |
1497 | error = EPERM; |
1498 | goto out; | |
1499 | } | |
1500 | ||
3558fd73 | 1501 | #ifdef HAVE_DNLC |
34dc7c2f BB |
1502 | if (realnmp) |
1503 | dnlc_remove(dvp, realnmp->pn_buf); | |
1504 | else | |
1505 | dnlc_remove(dvp, name); | |
3558fd73 | 1506 | #endif /* HAVE_DNLC */ |
34dc7c2f BB |
1507 | |
1508 | /* | |
3558fd73 BB |
1509 | * We never delete the znode and always place it in the unlinked |
1510 | * set. The dentry cache will always hold the last reference and | |
1511 | * is responsible for safely freeing the znode. | |
34dc7c2f | 1512 | */ |
572e2857 | 1513 | obj = zp->z_id; |
3558fd73 | 1514 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 1515 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff BB |
1516 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1517 | zfs_sa_upgrade_txholds(tx, zp); | |
1518 | zfs_sa_upgrade_txholds(tx, dzp); | |
34dc7c2f BB |
1519 | |
1520 | /* are there any extended attributes? */ | |
3558fd73 | 1521 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1522 | &xattr_obj, sizeof (xattr_obj)); |
572e2857 | 1523 | if (error == 0 && xattr_obj) { |
3558fd73 | 1524 | error = zfs_zget(zsb, xattr_obj, &xzp); |
428870ff BB |
1525 | ASSERT3U(error, ==, 0); |
1526 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
1527 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1528 | } |
1529 | ||
34dc7c2f | 1530 | /* charge as an update -- would be nice not to charge at all */ |
3558fd73 | 1531 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
34dc7c2f | 1532 | |
fb5f0bc8 | 1533 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
1534 | if (error) { |
1535 | zfs_dirent_unlock(dl); | |
3558fd73 | 1536 | iput(ip); |
572e2857 | 1537 | if (xzp) |
3558fd73 | 1538 | iput(ZTOI(xzp)); |
fb5f0bc8 | 1539 | if (error == ERESTART) { |
34dc7c2f BB |
1540 | dmu_tx_wait(tx); |
1541 | dmu_tx_abort(tx); | |
1542 | goto top; | |
1543 | } | |
3558fd73 | 1544 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1545 | if (realnmp) |
1546 | pn_free(realnmp); | |
3558fd73 | 1547 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f | 1548 | dmu_tx_abort(tx); |
3558fd73 | 1549 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1550 | return (error); |
1551 | } | |
1552 | ||
1553 | /* | |
1554 | * Remove the directory entry. | |
1555 | */ | |
1556 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1557 | ||
1558 | if (error) { | |
1559 | dmu_tx_commit(tx); | |
1560 | goto out; | |
1561 | } | |
1562 | ||
1563 | if (unlinked) { | |
572e2857 BB |
1564 | /* |
1565 | * Hold z_lock so that we can make sure that the ACL obj | |
1566 | * hasn't changed. Could have been deleted due to | |
1567 | * zfs_sa_upgrade(). | |
1568 | */ | |
1569 | mutex_enter(&zp->z_lock); | |
3558fd73 | 1570 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), |
428870ff | 1571 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); |
572e2857 | 1572 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1573 | zfs_unlinked_add(zp, tx); |
1574 | } | |
1575 | ||
1576 | txtype = TX_REMOVE; | |
3558fd73 | 1577 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1578 | if (flags & FIGNORECASE) |
1579 | txtype |= TX_CI; | |
3558fd73 | 1580 | #endif /* HAVE_PN_UTILS */ |
572e2857 | 1581 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1582 | |
1583 | dmu_tx_commit(tx); | |
1584 | out: | |
3558fd73 | 1585 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
1586 | if (realnmp) |
1587 | pn_free(realnmp); | |
3558fd73 | 1588 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
1589 | |
1590 | zfs_dirent_unlock(dl); | |
960e08fe BB |
1591 | zfs_inode_update(dzp); |
1592 | zfs_inode_update(zp); | |
1593 | if (xzp) | |
1594 | zfs_inode_update(xzp); | |
34dc7c2f | 1595 | |
3558fd73 | 1596 | iput(ip); |
428870ff | 1597 | if (xzp) |
3558fd73 | 1598 | iput(ZTOI(xzp)); |
428870ff | 1599 | |
3558fd73 | 1600 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1601 | zil_commit(zilog, 0); |
34dc7c2f | 1602 | |
3558fd73 | 1603 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1604 | return (error); |
1605 | } | |
e5c39b95 | 1606 | EXPORT_SYMBOL(zfs_remove); |
34dc7c2f BB |
1607 | |
1608 | /* | |
3558fd73 | 1609 | * Create a new directory and insert it into dip using the name |
34dc7c2f BB |
1610 | * provided. Return a pointer to the inserted directory. |
1611 | * | |
3558fd73 | 1612 | * IN: dip - inode of directory to add subdir to. |
34dc7c2f BB |
1613 | * dirname - name of new directory. |
1614 | * vap - attributes of new directory. | |
1615 | * cr - credentials of caller. | |
34dc7c2f BB |
1616 | * vsecp - ACL to be set |
1617 | * | |
3558fd73 | 1618 | * OUT: ipp - inode of created directory. |
34dc7c2f BB |
1619 | * |
1620 | * RETURN: 0 if success | |
1621 | * error code if failure | |
1622 | * | |
1623 | * Timestamps: | |
3558fd73 BB |
1624 | * dip - ctime|mtime updated |
1625 | * ipp - ctime|mtime|atime updated | |
34dc7c2f BB |
1626 | */ |
1627 | /*ARGSUSED*/ | |
e5c39b95 | 1628 | int |
3558fd73 BB |
1629 | zfs_mkdir(struct inode *dip, char *dirname, vattr_t *vap, struct inode **ipp, |
1630 | cred_t *cr, int flags, vsecattr_t *vsecp) | |
34dc7c2f | 1631 | { |
3558fd73 BB |
1632 | znode_t *zp, *dzp = ITOZ(dip); |
1633 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1634 | zilog_t *zilog; |
1635 | zfs_dirlock_t *dl; | |
1636 | uint64_t txtype; | |
1637 | dmu_tx_t *tx; | |
1638 | int error; | |
34dc7c2f | 1639 | int zf = ZNEW; |
b128c09f BB |
1640 | uid_t uid; |
1641 | gid_t gid = crgetgid(cr); | |
428870ff | 1642 | zfs_acl_ids_t acl_ids; |
9babb374 | 1643 | boolean_t fuid_dirtied; |
34dc7c2f | 1644 | |
3558fd73 | 1645 | ASSERT(S_ISDIR(vap->va_mode)); |
34dc7c2f BB |
1646 | |
1647 | /* | |
1648 | * If we have an ephemeral id, ACL, or XVATTR then | |
1649 | * make sure file system is at proper version | |
1650 | */ | |
1651 | ||
3558fd73 BB |
1652 | uid = crgetuid(cr); |
1653 | if (zsb->z_use_fuids == B_FALSE && | |
1654 | (vsecp || IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
34dc7c2f BB |
1655 | return (EINVAL); |
1656 | ||
3558fd73 | 1657 | ZFS_ENTER(zsb); |
34dc7c2f | 1658 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1659 | zilog = zsb->z_log; |
34dc7c2f | 1660 | |
428870ff | 1661 | if (dzp->z_pflags & ZFS_XATTR) { |
3558fd73 | 1662 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1663 | return (EINVAL); |
1664 | } | |
1665 | ||
3558fd73 | 1666 | if (zsb->z_utf8 && u8_validate(dirname, |
34dc7c2f | 1667 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 1668 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1669 | return (EILSEQ); |
1670 | } | |
1671 | if (flags & FIGNORECASE) | |
1672 | zf |= ZCILOOK; | |
1673 | ||
5484965a | 1674 | if (vap->va_mask & ATTR_XVATTR) { |
34dc7c2f | 1675 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
3558fd73 BB |
1676 | crgetuid(cr), cr, vap->va_mode)) != 0) { |
1677 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1678 | return (error); |
1679 | } | |
428870ff | 1680 | } |
34dc7c2f | 1681 | |
428870ff BB |
1682 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1683 | vsecp, &acl_ids)) != 0) { | |
3558fd73 | 1684 | ZFS_EXIT(zsb); |
428870ff BB |
1685 | return (error); |
1686 | } | |
34dc7c2f BB |
1687 | /* |
1688 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1689 | * |
1690 | * Existence is checked first to make sure we don't return | |
1691 | * EACCES instead of EEXIST which can cause some applications | |
1692 | * to fail. | |
34dc7c2f BB |
1693 | */ |
1694 | top: | |
3558fd73 | 1695 | *ipp = NULL; |
34dc7c2f | 1696 | |
149e873a BB |
1697 | if ((error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, |
1698 | NULL, NULL))) { | |
428870ff | 1699 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 1700 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1701 | return (error); |
1702 | } | |
1703 | ||
149e873a | 1704 | if ((error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr))) { |
428870ff | 1705 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1706 | zfs_dirent_unlock(dl); |
3558fd73 | 1707 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1708 | return (error); |
1709 | } | |
1710 | ||
3558fd73 | 1711 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
45d1cae3 | 1712 | zfs_acl_ids_free(&acl_ids); |
9babb374 | 1713 | zfs_dirent_unlock(dl); |
3558fd73 | 1714 | ZFS_EXIT(zsb); |
9babb374 BB |
1715 | return (EDQUOT); |
1716 | } | |
1717 | ||
34dc7c2f BB |
1718 | /* |
1719 | * Add a new entry to the directory. | |
1720 | */ | |
3558fd73 | 1721 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f BB |
1722 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); |
1723 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
3558fd73 | 1724 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 1725 | if (fuid_dirtied) |
3558fd73 BB |
1726 | zfs_fuid_txhold(zsb, tx); |
1727 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
428870ff BB |
1728 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
1729 | acl_ids.z_aclp->z_acl_bytes); | |
1730 | } | |
1731 | ||
1732 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1733 | ZFS_SA_BASE_ATTR_SIZE); | |
1734 | ||
fb5f0bc8 | 1735 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
1736 | if (error) { |
1737 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1738 | if (error == ERESTART) { |
34dc7c2f BB |
1739 | dmu_tx_wait(tx); |
1740 | dmu_tx_abort(tx); | |
1741 | goto top; | |
1742 | } | |
428870ff | 1743 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 1744 | dmu_tx_abort(tx); |
3558fd73 | 1745 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1746 | return (error); |
1747 | } | |
1748 | ||
1749 | /* | |
1750 | * Create new node. | |
1751 | */ | |
428870ff | 1752 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f | 1753 | |
9babb374 | 1754 | if (fuid_dirtied) |
3558fd73 | 1755 | zfs_fuid_sync(zsb, tx); |
428870ff | 1756 | |
34dc7c2f BB |
1757 | /* |
1758 | * Now put new name in parent dir. | |
1759 | */ | |
1760 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
1761 | ||
3558fd73 | 1762 | *ipp = ZTOI(zp); |
34dc7c2f BB |
1763 | |
1764 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
1765 | if (flags & FIGNORECASE) | |
1766 | txtype |= TX_CI; | |
9babb374 BB |
1767 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
1768 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 1769 | |
9babb374 | 1770 | zfs_acl_ids_free(&acl_ids); |
428870ff | 1771 | |
34dc7c2f BB |
1772 | dmu_tx_commit(tx); |
1773 | ||
1774 | zfs_dirent_unlock(dl); | |
1775 | ||
3558fd73 | 1776 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1777 | zil_commit(zilog, 0); |
428870ff | 1778 | |
960e08fe BB |
1779 | zfs_inode_update(dzp); |
1780 | zfs_inode_update(zp); | |
3558fd73 | 1781 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1782 | return (0); |
1783 | } | |
e5c39b95 | 1784 | EXPORT_SYMBOL(zfs_mkdir); |
34dc7c2f BB |
1785 | |
1786 | /* | |
1787 | * Remove a directory subdir entry. If the current working | |
1788 | * directory is the same as the subdir to be removed, the | |
1789 | * remove will fail. | |
1790 | * | |
3558fd73 | 1791 | * IN: dip - inode of directory to remove from. |
34dc7c2f | 1792 | * name - name of directory to be removed. |
3558fd73 | 1793 | * cwd - inode of current working directory. |
34dc7c2f | 1794 | * cr - credentials of caller. |
34dc7c2f BB |
1795 | * flags - case flags |
1796 | * | |
1797 | * RETURN: 0 if success | |
1798 | * error code if failure | |
1799 | * | |
1800 | * Timestamps: | |
3558fd73 | 1801 | * dip - ctime|mtime updated |
34dc7c2f BB |
1802 | */ |
1803 | /*ARGSUSED*/ | |
e5c39b95 | 1804 | int |
3558fd73 BB |
1805 | zfs_rmdir(struct inode *dip, char *name, struct inode *cwd, cred_t *cr, |
1806 | int flags) | |
34dc7c2f | 1807 | { |
3558fd73 | 1808 | znode_t *dzp = ITOZ(dip); |
34dc7c2f | 1809 | znode_t *zp; |
3558fd73 BB |
1810 | struct inode *ip; |
1811 | zfs_sb_t *zsb = ITOZSB(dip); | |
34dc7c2f BB |
1812 | zilog_t *zilog; |
1813 | zfs_dirlock_t *dl; | |
1814 | dmu_tx_t *tx; | |
1815 | int error; | |
1816 | int zflg = ZEXISTS; | |
1817 | ||
3558fd73 | 1818 | ZFS_ENTER(zsb); |
34dc7c2f | 1819 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 1820 | zilog = zsb->z_log; |
34dc7c2f BB |
1821 | |
1822 | if (flags & FIGNORECASE) | |
1823 | zflg |= ZCILOOK; | |
1824 | top: | |
1825 | zp = NULL; | |
1826 | ||
1827 | /* | |
1828 | * Attempt to lock directory; fail if entry doesn't exist. | |
1829 | */ | |
149e873a BB |
1830 | if ((error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, |
1831 | NULL, NULL))) { | |
3558fd73 | 1832 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1833 | return (error); |
1834 | } | |
1835 | ||
3558fd73 | 1836 | ip = ZTOI(zp); |
34dc7c2f | 1837 | |
149e873a | 1838 | if ((error = zfs_zaccess_delete(dzp, zp, cr))) { |
34dc7c2f BB |
1839 | goto out; |
1840 | } | |
1841 | ||
3558fd73 | 1842 | if (!S_ISDIR(ip->i_mode)) { |
34dc7c2f BB |
1843 | error = ENOTDIR; |
1844 | goto out; | |
1845 | } | |
1846 | ||
3558fd73 | 1847 | if (ip == cwd) { |
34dc7c2f BB |
1848 | error = EINVAL; |
1849 | goto out; | |
1850 | } | |
1851 | ||
34dc7c2f BB |
1852 | /* |
1853 | * Grab a lock on the directory to make sure that noone is | |
1854 | * trying to add (or lookup) entries while we are removing it. | |
1855 | */ | |
1856 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
1857 | ||
1858 | /* | |
1859 | * Grab a lock on the parent pointer to make sure we play well | |
1860 | * with the treewalk and directory rename code. | |
1861 | */ | |
1862 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
1863 | ||
3558fd73 | 1864 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 1865 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); |
428870ff | 1866 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3558fd73 | 1867 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
1868 | zfs_sa_upgrade_txholds(tx, zp); |
1869 | zfs_sa_upgrade_txholds(tx, dzp); | |
fb5f0bc8 | 1870 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
1871 | if (error) { |
1872 | rw_exit(&zp->z_parent_lock); | |
1873 | rw_exit(&zp->z_name_lock); | |
1874 | zfs_dirent_unlock(dl); | |
3558fd73 | 1875 | iput(ip); |
fb5f0bc8 | 1876 | if (error == ERESTART) { |
34dc7c2f BB |
1877 | dmu_tx_wait(tx); |
1878 | dmu_tx_abort(tx); | |
1879 | goto top; | |
1880 | } | |
1881 | dmu_tx_abort(tx); | |
3558fd73 | 1882 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1883 | return (error); |
1884 | } | |
1885 | ||
1886 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
1887 | ||
1888 | if (error == 0) { | |
1889 | uint64_t txtype = TX_RMDIR; | |
1890 | if (flags & FIGNORECASE) | |
1891 | txtype |= TX_CI; | |
572e2857 | 1892 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
1893 | } |
1894 | ||
1895 | dmu_tx_commit(tx); | |
1896 | ||
1897 | rw_exit(&zp->z_parent_lock); | |
1898 | rw_exit(&zp->z_name_lock); | |
1899 | out: | |
1900 | zfs_dirent_unlock(dl); | |
1901 | ||
3558fd73 | 1902 | iput(ip); |
34dc7c2f | 1903 | |
3558fd73 | 1904 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1905 | zil_commit(zilog, 0); |
428870ff | 1906 | |
960e08fe BB |
1907 | zfs_inode_update(dzp); |
1908 | zfs_inode_update(zp); | |
3558fd73 | 1909 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1910 | return (error); |
1911 | } | |
e5c39b95 | 1912 | EXPORT_SYMBOL(zfs_rmdir); |
34dc7c2f BB |
1913 | |
1914 | /* | |
1915 | * Read as many directory entries as will fit into the provided | |
3558fd73 | 1916 | * dirent buffer from the given directory cursor position. |
34dc7c2f | 1917 | * |
3558fd73 BB |
1918 | * IN: ip - inode of directory to read. |
1919 | * dirent - buffer for directory entries. | |
34dc7c2f | 1920 | * |
3558fd73 | 1921 | * OUT: dirent - filler buffer of directory entries. |
34dc7c2f BB |
1922 | * |
1923 | * RETURN: 0 if success | |
1924 | * error code if failure | |
1925 | * | |
1926 | * Timestamps: | |
3558fd73 | 1927 | * ip - atime updated |
34dc7c2f BB |
1928 | * |
1929 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
1930 | * This allows us to use the low range for "special" directory entries: | |
1931 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
1932 | * we use the offset 2 for the '.zfs' directory. | |
1933 | */ | |
1934 | /* ARGSUSED */ | |
3558fd73 BB |
1935 | int |
1936 | zfs_readdir(struct inode *ip, void *dirent, filldir_t filldir, | |
1937 | loff_t *pos, cred_t *cr) | |
34dc7c2f | 1938 | { |
3558fd73 BB |
1939 | znode_t *zp = ITOZ(ip); |
1940 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 1941 | objset_t *os; |
34dc7c2f BB |
1942 | zap_cursor_t zc; |
1943 | zap_attribute_t zap; | |
34dc7c2f BB |
1944 | int outcount; |
1945 | int error; | |
1946 | uint8_t prefetch; | |
3558fd73 BB |
1947 | int done = 0; |
1948 | uint64_t parent; | |
34dc7c2f | 1949 | |
3558fd73 | 1950 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1951 | ZFS_VERIFY_ZP(zp); |
1952 | ||
3558fd73 BB |
1953 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
1954 | &parent, sizeof (parent))) != 0) | |
1955 | goto out; | |
34dc7c2f BB |
1956 | |
1957 | /* | |
1958 | * Quit if directory has been removed (posix) | |
1959 | */ | |
34dc7c2f | 1960 | error = 0; |
3558fd73 BB |
1961 | if (zp->z_unlinked) |
1962 | goto out; | |
1963 | ||
1964 | os = zsb->z_os; | |
34dc7c2f BB |
1965 | prefetch = zp->z_zn_prefetch; |
1966 | ||
1967 | /* | |
1968 | * Initialize the iterator cursor. | |
1969 | */ | |
3558fd73 | 1970 | if (*pos <= 3) { |
34dc7c2f BB |
1971 | /* |
1972 | * Start iteration from the beginning of the directory. | |
1973 | */ | |
1974 | zap_cursor_init(&zc, os, zp->z_id); | |
1975 | } else { | |
1976 | /* | |
1977 | * The offset is a serialized cursor. | |
1978 | */ | |
3558fd73 | 1979 | zap_cursor_init_serialized(&zc, os, zp->z_id, *pos); |
34dc7c2f BB |
1980 | } |
1981 | ||
34dc7c2f BB |
1982 | /* |
1983 | * Transform to file-system independent format | |
1984 | */ | |
1985 | outcount = 0; | |
34dc7c2f | 1986 | |
3558fd73 BB |
1987 | while (!done) { |
1988 | uint64_t objnum; | |
34dc7c2f BB |
1989 | /* |
1990 | * Special case `.', `..', and `.zfs'. | |
1991 | */ | |
3558fd73 | 1992 | if (*pos == 0) { |
34dc7c2f BB |
1993 | (void) strcpy(zap.za_name, "."); |
1994 | zap.za_normalization_conflict = 0; | |
1995 | objnum = zp->z_id; | |
3558fd73 | 1996 | } else if (*pos == 1) { |
34dc7c2f BB |
1997 | (void) strcpy(zap.za_name, ".."); |
1998 | zap.za_normalization_conflict = 0; | |
428870ff | 1999 | objnum = parent; |
3558fd73 | 2000 | } else if (*pos == 2 && zfs_show_ctldir(zp)) { |
34dc7c2f BB |
2001 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); |
2002 | zap.za_normalization_conflict = 0; | |
2003 | objnum = ZFSCTL_INO_ROOT; | |
2004 | } else { | |
2005 | /* | |
2006 | * Grab next entry. | |
2007 | */ | |
3558fd73 BB |
2008 | if ((error = zap_cursor_retrieve(&zc, &zap))) { |
2009 | if (error == ENOENT) | |
34dc7c2f BB |
2010 | break; |
2011 | else | |
2012 | goto update; | |
2013 | } | |
2014 | ||
2015 | if (zap.za_integer_length != 8 || | |
2016 | zap.za_num_integers != 1) { | |
2017 | cmn_err(CE_WARN, "zap_readdir: bad directory " | |
2018 | "entry, obj = %lld, offset = %lld\n", | |
2019 | (u_longlong_t)zp->z_id, | |
3558fd73 | 2020 | (u_longlong_t)*pos); |
34dc7c2f BB |
2021 | error = ENXIO; |
2022 | goto update; | |
2023 | } | |
2024 | ||
2025 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
34dc7c2f | 2026 | } |
3558fd73 BB |
2027 | done = filldir(dirent, zap.za_name, strlen(zap.za_name), |
2028 | zap_cursor_serialize(&zc), objnum, 0); | |
2029 | if (done) { | |
34dc7c2f BB |
2030 | break; |
2031 | } | |
34dc7c2f BB |
2032 | |
2033 | /* Prefetch znode */ | |
3558fd73 | 2034 | if (prefetch) { |
34dc7c2f | 2035 | dmu_prefetch(os, objnum, 0, 0); |
3558fd73 | 2036 | } |
34dc7c2f | 2037 | |
3558fd73 | 2038 | if (*pos >= 2) { |
34dc7c2f | 2039 | zap_cursor_advance(&zc); |
3558fd73 | 2040 | *pos = zap_cursor_serialize(&zc); |
34dc7c2f | 2041 | } else { |
3558fd73 | 2042 | (*pos)++; |
34dc7c2f | 2043 | } |
34dc7c2f BB |
2044 | } |
2045 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2046 | ||
34dc7c2f BB |
2047 | update: |
2048 | zap_cursor_fini(&zc); | |
34dc7c2f BB |
2049 | if (error == ENOENT) |
2050 | error = 0; | |
2051 | ||
3558fd73 BB |
2052 | ZFS_ACCESSTIME_STAMP(zsb, zp); |
2053 | zfs_inode_update(zp); | |
2054 | ||
2055 | out: | |
2056 | ZFS_EXIT(zsb); | |
34dc7c2f | 2057 | |
34dc7c2f BB |
2058 | return (error); |
2059 | } | |
3558fd73 | 2060 | EXPORT_SYMBOL(zfs_readdir); |
34dc7c2f BB |
2061 | |
2062 | ulong_t zfs_fsync_sync_cnt = 4; | |
2063 | ||
e5c39b95 | 2064 | int |
3558fd73 | 2065 | zfs_fsync(struct inode *ip, int syncflag, cred_t *cr) |
34dc7c2f | 2066 | { |
3558fd73 BB |
2067 | znode_t *zp = ITOZ(ip); |
2068 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2069 | |
2070 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); | |
2071 | ||
3558fd73 BB |
2072 | if (zsb->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2073 | ZFS_ENTER(zsb); | |
428870ff | 2074 | ZFS_VERIFY_ZP(zp); |
3558fd73 BB |
2075 | zil_commit(zsb->z_log, zp->z_id); |
2076 | ZFS_EXIT(zsb); | |
428870ff | 2077 | } |
34dc7c2f BB |
2078 | return (0); |
2079 | } | |
e5c39b95 | 2080 | EXPORT_SYMBOL(zfs_fsync); |
34dc7c2f BB |
2081 | |
2082 | ||
2083 | /* | |
2084 | * Get the requested file attributes and place them in the provided | |
2085 | * vattr structure. | |
2086 | * | |
3558fd73 | 2087 | * IN: ip - inode of file. |
5484965a BB |
2088 | * vap - va_mask identifies requested attributes. |
2089 | * If ATTR_XVATTR set, then optional attrs are requested | |
34dc7c2f BB |
2090 | * flags - ATTR_NOACLCHECK (CIFS server context) |
2091 | * cr - credentials of caller. | |
34dc7c2f | 2092 | * |
5484965a BB |
2093 | * OUT: vap - attribute values. |
2094 | * | |
2095 | * RETURN: 0 (always succeeds) | |
34dc7c2f BB |
2096 | */ |
2097 | /* ARGSUSED */ | |
e5c39b95 | 2098 | int |
5484965a | 2099 | zfs_getattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2100 | { |
3558fd73 BB |
2101 | znode_t *zp = ITOZ(ip); |
2102 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2103 | int error = 0; |
2104 | uint64_t links; | |
428870ff | 2105 | uint64_t mtime[2], ctime[2]; |
5484965a BB |
2106 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2107 | xoptattr_t *xoap = NULL; | |
34dc7c2f | 2108 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff BB |
2109 | sa_bulk_attr_t bulk[2]; |
2110 | int count = 0; | |
34dc7c2f | 2111 | |
3558fd73 | 2112 | ZFS_ENTER(zsb); |
34dc7c2f | 2113 | ZFS_VERIFY_ZP(zp); |
428870ff | 2114 | |
5484965a | 2115 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
572e2857 | 2116 | |
3558fd73 BB |
2117 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, &mtime, 16); |
2118 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, &ctime, 16); | |
428870ff BB |
2119 | |
2120 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
3558fd73 | 2121 | ZFS_EXIT(zsb); |
428870ff BB |
2122 | return (error); |
2123 | } | |
34dc7c2f | 2124 | |
34dc7c2f BB |
2125 | /* |
2126 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2127 | * Also, if we are the owner don't bother, since owner should | |
2128 | * always be allowed to read basic attributes of file. | |
2129 | */ | |
572e2857 | 2130 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
5484965a | 2131 | (vap->va_uid != crgetuid(cr))) { |
149e873a BB |
2132 | if ((error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2133 | skipaclchk, cr))) { | |
3558fd73 | 2134 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2135 | return (error); |
2136 | } | |
2137 | } | |
2138 | ||
2139 | /* | |
2140 | * Return all attributes. It's cheaper to provide the answer | |
2141 | * than to determine whether we were asked the question. | |
2142 | */ | |
2143 | ||
9babb374 | 2144 | mutex_enter(&zp->z_lock); |
5484965a BB |
2145 | vap->va_type = vn_mode_to_vtype(zp->z_mode); |
2146 | vap->va_mode = zp->z_mode; | |
53cf50e0 | 2147 | vap->va_fsid = ZTOI(zp)->i_sb->s_dev; |
5484965a | 2148 | vap->va_nodeid = zp->z_id; |
3558fd73 | 2149 | if ((zp->z_id == zsb->z_root) && zfs_show_ctldir(zp)) |
428870ff | 2150 | links = zp->z_links + 1; |
34dc7c2f | 2151 | else |
428870ff | 2152 | links = zp->z_links; |
5484965a BB |
2153 | vap->va_nlink = MIN(links, ZFS_LINK_MAX); |
2154 | vap->va_size = i_size_read(ip); | |
2155 | vap->va_rdev = ip->i_rdev; | |
2156 | vap->va_seq = ip->i_generation; | |
2157 | ||
2158 | /* | |
2159 | * Add in any requested optional attributes and the create time. | |
2160 | * Also set the corresponding bits in the returned attribute bitmap. | |
2161 | */ | |
2162 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zsb->z_use_fuids) { | |
2163 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
2164 | xoap->xoa_archive = | |
2165 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); | |
2166 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
2167 | } | |
2168 | ||
2169 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2170 | xoap->xoa_readonly = | |
2171 | ((zp->z_pflags & ZFS_READONLY) != 0); | |
2172 | XVA_SET_RTN(xvap, XAT_READONLY); | |
2173 | } | |
2174 | ||
2175 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2176 | xoap->xoa_system = | |
2177 | ((zp->z_pflags & ZFS_SYSTEM) != 0); | |
2178 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
2179 | } | |
2180 | ||
2181 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2182 | xoap->xoa_hidden = | |
2183 | ((zp->z_pflags & ZFS_HIDDEN) != 0); | |
2184 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
2185 | } | |
2186 | ||
2187 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2188 | xoap->xoa_nounlink = | |
2189 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); | |
2190 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
2191 | } | |
2192 | ||
2193 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2194 | xoap->xoa_immutable = | |
2195 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); | |
2196 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
2197 | } | |
2198 | ||
2199 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2200 | xoap->xoa_appendonly = | |
2201 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); | |
2202 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
2203 | } | |
2204 | ||
2205 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2206 | xoap->xoa_nodump = | |
2207 | ((zp->z_pflags & ZFS_NODUMP) != 0); | |
2208 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
2209 | } | |
2210 | ||
2211 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2212 | xoap->xoa_opaque = | |
2213 | ((zp->z_pflags & ZFS_OPAQUE) != 0); | |
2214 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
2215 | } | |
2216 | ||
2217 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2218 | xoap->xoa_av_quarantined = | |
2219 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); | |
2220 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
2221 | } | |
2222 | ||
2223 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2224 | xoap->xoa_av_modified = | |
2225 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); | |
2226 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
2227 | } | |
2228 | ||
2229 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
2230 | S_ISREG(ip->i_mode)) { | |
2231 | zfs_sa_get_scanstamp(zp, xvap); | |
2232 | } | |
34dc7c2f | 2233 | |
5484965a BB |
2234 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { |
2235 | uint64_t times[2]; | |
2236 | ||
2237 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zsb), | |
2238 | times, sizeof (times)); | |
2239 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
2240 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
2241 | } | |
2242 | ||
2243 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2244 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2245 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2246 | } | |
2247 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { | |
2248 | xoap->xoa_generation = zp->z_gen; | |
2249 | XVA_SET_RTN(xvap, XAT_GEN); | |
2250 | } | |
2251 | ||
2252 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2253 | xoap->xoa_offline = | |
2254 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2255 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2256 | } | |
2257 | ||
2258 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2259 | xoap->xoa_sparse = | |
2260 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2261 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2262 | } | |
2263 | } | |
2264 | ||
2265 | ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime); | |
2266 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); | |
2267 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2268 | |
2269 | mutex_exit(&zp->z_lock); | |
2270 | ||
5484965a | 2271 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2272 | |
2273 | if (zp->z_blksz == 0) { | |
2274 | /* | |
2275 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2276 | */ | |
5484965a | 2277 | vap->va_blksize = zsb->z_max_blksz; |
34dc7c2f BB |
2278 | } |
2279 | ||
3558fd73 | 2280 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2281 | return (0); |
2282 | } | |
e5c39b95 | 2283 | EXPORT_SYMBOL(zfs_getattr); |
34dc7c2f BB |
2284 | |
2285 | /* | |
2286 | * Set the file attributes to the values contained in the | |
2287 | * vattr structure. | |
2288 | * | |
3558fd73 | 2289 | * IN: ip - inode of file to be modified. |
34dc7c2f | 2290 | * vap - new attribute values. |
5484965a | 2291 | * If ATTR_XVATTR set, then optional attrs are being set |
34dc7c2f BB |
2292 | * flags - ATTR_UTIME set if non-default time values provided. |
2293 | * - ATTR_NOACLCHECK (CIFS context only). | |
2294 | * cr - credentials of caller. | |
34dc7c2f BB |
2295 | * |
2296 | * RETURN: 0 if success | |
2297 | * error code if failure | |
2298 | * | |
2299 | * Timestamps: | |
3558fd73 | 2300 | * ip - ctime updated, mtime updated if size changed. |
34dc7c2f BB |
2301 | */ |
2302 | /* ARGSUSED */ | |
e5c39b95 | 2303 | int |
5484965a | 2304 | zfs_setattr(struct inode *ip, vattr_t *vap, int flags, cred_t *cr) |
34dc7c2f | 2305 | { |
3558fd73 BB |
2306 | znode_t *zp = ITOZ(ip); |
2307 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
2308 | zilog_t *zilog; |
2309 | dmu_tx_t *tx; | |
2310 | vattr_t oldva; | |
f4ea75d4 | 2311 | xvattr_t *tmpxvattr; |
5484965a | 2312 | uint_t mask = vap->va_mask; |
34dc7c2f BB |
2313 | uint_t saved_mask; |
2314 | int trim_mask = 0; | |
2315 | uint64_t new_mode; | |
9babb374 | 2316 | uint64_t new_uid, new_gid; |
572e2857 | 2317 | uint64_t xattr_obj; |
428870ff | 2318 | uint64_t mtime[2], ctime[2]; |
34dc7c2f BB |
2319 | znode_t *attrzp; |
2320 | int need_policy = FALSE; | |
428870ff | 2321 | int err, err2; |
34dc7c2f | 2322 | zfs_fuid_info_t *fuidp = NULL; |
5484965a BB |
2323 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2324 | xoptattr_t *xoap; | |
2325 | zfs_acl_t *aclp; | |
34dc7c2f | 2326 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff | 2327 | boolean_t fuid_dirtied = B_FALSE; |
17c37660 | 2328 | sa_bulk_attr_t *bulk, *xattr_bulk; |
428870ff | 2329 | int count = 0, xattr_count = 0; |
34dc7c2f BB |
2330 | |
2331 | if (mask == 0) | |
2332 | return (0); | |
2333 | ||
3558fd73 | 2334 | ZFS_ENTER(zsb); |
34dc7c2f BB |
2335 | ZFS_VERIFY_ZP(zp); |
2336 | ||
3558fd73 | 2337 | zilog = zsb->z_log; |
34dc7c2f BB |
2338 | |
2339 | /* | |
2340 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2341 | * that file system is at proper version level | |
2342 | */ | |
5484965a | 2343 | |
3558fd73 | 2344 | if (zsb->z_use_fuids == B_FALSE && |
5484965a BB |
2345 | (((mask & ATTR_UID) && IS_EPHEMERAL(vap->va_uid)) || |
2346 | ((mask & ATTR_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2347 | (mask & ATTR_XVATTR))) { | |
3558fd73 | 2348 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2349 | return (EINVAL); |
2350 | } | |
2351 | ||
3558fd73 BB |
2352 | if (mask & ATTR_SIZE && S_ISDIR(ip->i_mode)) { |
2353 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
2354 | return (EISDIR); |
2355 | } | |
2356 | ||
3558fd73 BB |
2357 | if (mask & ATTR_SIZE && !S_ISREG(ip->i_mode) && !S_ISFIFO(ip->i_mode)) { |
2358 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
2359 | return (EINVAL); |
2360 | } | |
2361 | ||
5484965a BB |
2362 | /* |
2363 | * If this is an xvattr_t, then get a pointer to the structure of | |
2364 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2365 | */ | |
2366 | xoap = xva_getxoptattr(xvap); | |
2367 | ||
f4ea75d4 BB |
2368 | tmpxvattr = kmem_alloc(sizeof(xvattr_t), KM_SLEEP); |
2369 | xva_init(tmpxvattr); | |
5484965a | 2370 | |
17c37660 BB |
2371 | bulk = kmem_alloc(sizeof(sa_bulk_attr_t) * 7, KM_SLEEP); |
2372 | xattr_bulk = kmem_alloc(sizeof(sa_bulk_attr_t) * 7, KM_SLEEP); | |
2373 | ||
5484965a BB |
2374 | /* |
2375 | * Immutable files can only alter immutable bit and atime | |
2376 | */ | |
2377 | if ((zp->z_pflags & ZFS_IMMUTABLE) && | |
2378 | ((mask & (ATTR_SIZE|ATTR_UID|ATTR_GID|ATTR_MTIME|ATTR_MODE)) || | |
2379 | ((mask & ATTR_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
f4ea75d4 BB |
2380 | err = EPERM; |
2381 | goto out3; | |
5484965a BB |
2382 | } |
2383 | ||
3558fd73 | 2384 | if ((mask & ATTR_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
f4ea75d4 BB |
2385 | err = EPERM; |
2386 | goto out3; | |
34dc7c2f BB |
2387 | } |
2388 | ||
5484965a BB |
2389 | /* |
2390 | * Verify timestamps doesn't overflow 32 bits. | |
2391 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2392 | * handle times greater than 2039. This check should be removed | |
2393 | * once large timestamps are fully supported. | |
2394 | */ | |
2395 | if (mask & (ATTR_ATIME | ATTR_MTIME)) { | |
2396 | if (((mask & ATTR_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2397 | ((mask & ATTR_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
f4ea75d4 BB |
2398 | err = EOVERFLOW; |
2399 | goto out3; | |
5484965a BB |
2400 | } |
2401 | } | |
2402 | ||
34dc7c2f BB |
2403 | top: |
2404 | attrzp = NULL; | |
572e2857 | 2405 | aclp = NULL; |
34dc7c2f | 2406 | |
45d1cae3 | 2407 | /* Can this be moved to before the top label? */ |
2cf7f52b | 2408 | if (zfs_is_readonly(zsb)) { |
f4ea75d4 BB |
2409 | err = EROFS; |
2410 | goto out3; | |
34dc7c2f BB |
2411 | } |
2412 | ||
2413 | /* | |
2414 | * First validate permissions | |
2415 | */ | |
2416 | ||
3558fd73 | 2417 | if (mask & ATTR_SIZE) { |
34dc7c2f | 2418 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); |
f4ea75d4 BB |
2419 | if (err) |
2420 | goto out3; | |
2421 | ||
218b8eaf PJ |
2422 | truncate_setsize(ip, vap->va_size); |
2423 | ||
34dc7c2f BB |
2424 | /* |
2425 | * XXX - Note, we are not providing any open | |
2426 | * mode flags here (like FNDELAY), so we may | |
2427 | * block if there are locks present... this | |
2428 | * should be addressed in openat(). | |
2429 | */ | |
b128c09f | 2430 | /* XXX - would it be OK to generate a log record here? */ |
5484965a | 2431 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); |
f4ea75d4 BB |
2432 | if (err) |
2433 | goto out3; | |
428870ff | 2434 | } |
34dc7c2f | 2435 | |
5484965a BB |
2436 | if (mask & (ATTR_ATIME|ATTR_MTIME) || |
2437 | ((mask & ATTR_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2438 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2439 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
2440 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || | |
2441 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
2442 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || | |
2443 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { | |
2444 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, | |
2445 | skipaclchk, cr); | |
2446 | } | |
2447 | ||
3558fd73 BB |
2448 | if (mask & (ATTR_UID|ATTR_GID)) { |
2449 | int idmask = (mask & (ATTR_UID|ATTR_GID)); | |
34dc7c2f BB |
2450 | int take_owner; |
2451 | int take_group; | |
2452 | ||
2453 | /* | |
2454 | * NOTE: even if a new mode is being set, | |
2455 | * we may clear S_ISUID/S_ISGID bits. | |
2456 | */ | |
2457 | ||
3558fd73 | 2458 | if (!(mask & ATTR_MODE)) |
5484965a | 2459 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2460 | |
2461 | /* | |
2462 | * Take ownership or chgrp to group we are a member of | |
2463 | */ | |
2464 | ||
5484965a | 2465 | take_owner = (mask & ATTR_UID) && (vap->va_uid == crgetuid(cr)); |
3558fd73 | 2466 | take_group = (mask & ATTR_GID) && |
5484965a | 2467 | zfs_groupmember(zsb, vap->va_gid, cr); |
34dc7c2f BB |
2468 | |
2469 | /* | |
5484965a | 2470 | * If both ATTR_UID and ATTR_GID are set then take_owner and |
34dc7c2f BB |
2471 | * take_group must both be set in order to allow taking |
2472 | * ownership. | |
2473 | * | |
2474 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2475 | * | |
2476 | */ | |
2477 | ||
3558fd73 BB |
2478 | if (((idmask == (ATTR_UID|ATTR_GID)) && |
2479 | take_owner && take_group) || | |
2480 | ((idmask == ATTR_UID) && take_owner) || | |
2481 | ((idmask == ATTR_GID) && take_group)) { | |
34dc7c2f BB |
2482 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, |
2483 | skipaclchk, cr) == 0) { | |
2484 | /* | |
2485 | * Remove setuid/setgid for non-privileged users | |
2486 | */ | |
5484965a | 2487 | (void) secpolicy_setid_clear(vap, cr); |
3558fd73 | 2488 | trim_mask = (mask & (ATTR_UID|ATTR_GID)); |
34dc7c2f BB |
2489 | } else { |
2490 | need_policy = TRUE; | |
2491 | } | |
2492 | } else { | |
2493 | need_policy = TRUE; | |
2494 | } | |
2495 | } | |
2496 | ||
2497 | mutex_enter(&zp->z_lock); | |
428870ff | 2498 | oldva.va_mode = zp->z_mode; |
572e2857 | 2499 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
5484965a BB |
2500 | if (mask & ATTR_XVATTR) { |
2501 | /* | |
2502 | * Update xvattr mask to include only those attributes | |
2503 | * that are actually changing. | |
2504 | * | |
2505 | * the bits will be restored prior to actually setting | |
2506 | * the attributes so the caller thinks they were set. | |
2507 | */ | |
2508 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2509 | if (xoap->xoa_appendonly != | |
2510 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { | |
2511 | need_policy = TRUE; | |
2512 | } else { | |
2513 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
f4ea75d4 | 2514 | XVA_SET_REQ(tmpxvattr, XAT_APPENDONLY); |
5484965a BB |
2515 | } |
2516 | } | |
2517 | ||
2518 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2519 | if (xoap->xoa_nounlink != | |
2520 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { | |
2521 | need_policy = TRUE; | |
2522 | } else { | |
2523 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
f4ea75d4 | 2524 | XVA_SET_REQ(tmpxvattr, XAT_NOUNLINK); |
5484965a BB |
2525 | } |
2526 | } | |
2527 | ||
2528 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2529 | if (xoap->xoa_immutable != | |
2530 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { | |
2531 | need_policy = TRUE; | |
2532 | } else { | |
2533 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
f4ea75d4 | 2534 | XVA_SET_REQ(tmpxvattr, XAT_IMMUTABLE); |
5484965a BB |
2535 | } |
2536 | } | |
2537 | ||
2538 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2539 | if (xoap->xoa_nodump != | |
2540 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { | |
2541 | need_policy = TRUE; | |
2542 | } else { | |
2543 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
f4ea75d4 | 2544 | XVA_SET_REQ(tmpxvattr, XAT_NODUMP); |
5484965a BB |
2545 | } |
2546 | } | |
2547 | ||
2548 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2549 | if (xoap->xoa_av_modified != | |
2550 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { | |
2551 | need_policy = TRUE; | |
2552 | } else { | |
2553 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
f4ea75d4 | 2554 | XVA_SET_REQ(tmpxvattr, XAT_AV_MODIFIED); |
5484965a BB |
2555 | } |
2556 | } | |
2557 | ||
2558 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2559 | if ((!S_ISREG(ip->i_mode) && | |
2560 | xoap->xoa_av_quarantined) || | |
2561 | xoap->xoa_av_quarantined != | |
2562 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { | |
2563 | need_policy = TRUE; | |
2564 | } else { | |
2565 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
f4ea75d4 | 2566 | XVA_SET_REQ(tmpxvattr, XAT_AV_QUARANTINED); |
5484965a BB |
2567 | } |
2568 | } | |
2569 | ||
2570 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2571 | mutex_exit(&zp->z_lock); | |
f4ea75d4 BB |
2572 | err = EPERM; |
2573 | goto out3; | |
5484965a BB |
2574 | } |
2575 | ||
2576 | if (need_policy == FALSE && | |
2577 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2578 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
2579 | need_policy = TRUE; | |
2580 | } | |
2581 | } | |
34dc7c2f BB |
2582 | |
2583 | mutex_exit(&zp->z_lock); | |
2584 | ||
3558fd73 | 2585 | if (mask & ATTR_MODE) { |
34dc7c2f | 2586 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { |
5484965a | 2587 | err = secpolicy_setid_setsticky_clear(ip, vap, |
34dc7c2f | 2588 | &oldva, cr); |
f4ea75d4 BB |
2589 | if (err) |
2590 | goto out3; | |
2591 | ||
3558fd73 | 2592 | trim_mask |= ATTR_MODE; |
34dc7c2f BB |
2593 | } else { |
2594 | need_policy = TRUE; | |
2595 | } | |
2596 | } | |
2597 | ||
2598 | if (need_policy) { | |
2599 | /* | |
2600 | * If trim_mask is set then take ownership | |
2601 | * has been granted or write_acl is present and user | |
2602 | * has the ability to modify mode. In that case remove | |
2603 | * UID|GID and or MODE from mask so that | |
2604 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2605 | */ | |
2606 | ||
2607 | if (trim_mask) { | |
5484965a BB |
2608 | saved_mask = vap->va_mask; |
2609 | vap->va_mask &= ~trim_mask; | |
34dc7c2f | 2610 | } |
5484965a | 2611 | err = secpolicy_vnode_setattr(cr, ip, vap, &oldva, flags, |
34dc7c2f | 2612 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); |
f4ea75d4 BB |
2613 | if (err) |
2614 | goto out3; | |
34dc7c2f BB |
2615 | |
2616 | if (trim_mask) | |
5484965a | 2617 | vap->va_mask |= saved_mask; |
34dc7c2f BB |
2618 | } |
2619 | ||
2620 | /* | |
2621 | * secpolicy_vnode_setattr, or take ownership may have | |
2622 | * changed va_mask | |
2623 | */ | |
5484965a | 2624 | mask = vap->va_mask; |
34dc7c2f | 2625 | |
3558fd73 BB |
2626 | if ((mask & (ATTR_UID | ATTR_GID))) { |
2627 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zsb), | |
572e2857 | 2628 | &xattr_obj, sizeof (xattr_obj)); |
428870ff | 2629 | |
572e2857 | 2630 | if (err == 0 && xattr_obj) { |
3558fd73 | 2631 | err = zfs_zget(ZTOZSB(zp), xattr_obj, &attrzp); |
428870ff BB |
2632 | if (err) |
2633 | goto out2; | |
2634 | } | |
3558fd73 BB |
2635 | if (mask & ATTR_UID) { |
2636 | new_uid = zfs_fuid_create(zsb, | |
5484965a | 2637 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); |
572e2857 | 2638 | if (new_uid != zp->z_uid && |
3558fd73 | 2639 | zfs_fuid_overquota(zsb, B_FALSE, new_uid)) { |
572e2857 | 2640 | if (attrzp) |
3558fd73 | 2641 | iput(ZTOI(attrzp)); |
428870ff BB |
2642 | err = EDQUOT; |
2643 | goto out2; | |
2644 | } | |
2645 | } | |
2646 | ||
3558fd73 | 2647 | if (mask & ATTR_GID) { |
5484965a | 2648 | new_gid = zfs_fuid_create(zsb, (uint64_t)vap->va_gid, |
428870ff BB |
2649 | cr, ZFS_GROUP, &fuidp); |
2650 | if (new_gid != zp->z_gid && | |
3558fd73 | 2651 | zfs_fuid_overquota(zsb, B_TRUE, new_gid)) { |
572e2857 | 2652 | if (attrzp) |
3558fd73 | 2653 | iput(ZTOI(attrzp)); |
428870ff BB |
2654 | err = EDQUOT; |
2655 | goto out2; | |
2656 | } | |
2657 | } | |
2658 | } | |
3558fd73 | 2659 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 2660 | |
3558fd73 | 2661 | if (mask & ATTR_MODE) { |
428870ff | 2662 | uint64_t pmode = zp->z_mode; |
572e2857 | 2663 | uint64_t acl_obj; |
5484965a | 2664 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
34dc7c2f | 2665 | |
572e2857 | 2666 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 2667 | |
572e2857 BB |
2668 | mutex_enter(&zp->z_lock); |
2669 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
2670 | /* |
2671 | * Are we upgrading ACL from old V0 format | |
2672 | * to V1 format? | |
2673 | */ | |
3558fd73 | 2674 | if (zsb->z_version >= ZPL_VERSION_FUID && |
572e2857 | 2675 | zfs_znode_acl_version(zp) == |
34dc7c2f | 2676 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 2677 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
2678 | DMU_OBJECT_END); |
2679 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
2680 | 0, aclp->z_acl_bytes); | |
2681 | } else { | |
572e2857 | 2682 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
2683 | aclp->z_acl_bytes); |
2684 | } | |
428870ff | 2685 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
2686 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
2687 | 0, aclp->z_acl_bytes); | |
2688 | } | |
572e2857 | 2689 | mutex_exit(&zp->z_lock); |
428870ff BB |
2690 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
2691 | } else { | |
5484965a BB |
2692 | if ((mask & ATTR_XVATTR) && |
2693 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
2694 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
2695 | else | |
2696 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2697 | } |
2698 | ||
428870ff BB |
2699 | if (attrzp) { |
2700 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
2701 | } |
2702 | ||
3558fd73 | 2703 | fuid_dirtied = zsb->z_fuid_dirty; |
428870ff | 2704 | if (fuid_dirtied) |
3558fd73 | 2705 | zfs_fuid_txhold(zsb, tx); |
428870ff BB |
2706 | |
2707 | zfs_sa_upgrade_txholds(tx, zp); | |
2708 | ||
fb5f0bc8 | 2709 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 2710 | if (err) { |
9babb374 | 2711 | if (err == ERESTART) |
34dc7c2f | 2712 | dmu_tx_wait(tx); |
9babb374 | 2713 | goto out; |
34dc7c2f BB |
2714 | } |
2715 | ||
428870ff | 2716 | count = 0; |
34dc7c2f BB |
2717 | /* |
2718 | * Set each attribute requested. | |
2719 | * We group settings according to the locks they need to acquire. | |
2720 | * | |
2721 | * Note: you cannot set ctime directly, although it will be | |
2722 | * updated as a side-effect of calling this function. | |
2723 | */ | |
2724 | ||
572e2857 | 2725 | |
3558fd73 | 2726 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2727 | mutex_enter(&zp->z_acl_lock); |
34dc7c2f BB |
2728 | mutex_enter(&zp->z_lock); |
2729 | ||
3558fd73 | 2730 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff BB |
2731 | &zp->z_pflags, sizeof (zp->z_pflags)); |
2732 | ||
2733 | if (attrzp) { | |
3558fd73 | 2734 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2735 | mutex_enter(&attrzp->z_acl_lock); |
428870ff BB |
2736 | mutex_enter(&attrzp->z_lock); |
2737 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2738 | SA_ZPL_FLAGS(zsb), NULL, &attrzp->z_pflags, |
428870ff BB |
2739 | sizeof (attrzp->z_pflags)); |
2740 | } | |
2741 | ||
3558fd73 | 2742 | if (mask & (ATTR_UID|ATTR_GID)) { |
428870ff | 2743 | |
3558fd73 BB |
2744 | if (mask & ATTR_UID) { |
2745 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, | |
428870ff | 2746 | &new_uid, sizeof (new_uid)); |
572e2857 | 2747 | zp->z_uid = new_uid; |
428870ff BB |
2748 | if (attrzp) { |
2749 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2750 | SA_ZPL_UID(zsb), NULL, &new_uid, |
428870ff | 2751 | sizeof (new_uid)); |
572e2857 | 2752 | attrzp->z_uid = new_uid; |
428870ff BB |
2753 | } |
2754 | } | |
2755 | ||
3558fd73 BB |
2756 | if (mask & ATTR_GID) { |
2757 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), | |
428870ff | 2758 | NULL, &new_gid, sizeof (new_gid)); |
572e2857 | 2759 | zp->z_gid = new_gid; |
428870ff BB |
2760 | if (attrzp) { |
2761 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2762 | SA_ZPL_GID(zsb), NULL, &new_gid, |
428870ff | 2763 | sizeof (new_gid)); |
572e2857 | 2764 | attrzp->z_gid = new_gid; |
428870ff BB |
2765 | } |
2766 | } | |
3558fd73 BB |
2767 | if (!(mask & ATTR_MODE)) { |
2768 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), | |
428870ff BB |
2769 | NULL, &new_mode, sizeof (new_mode)); |
2770 | new_mode = zp->z_mode; | |
2771 | } | |
2772 | err = zfs_acl_chown_setattr(zp); | |
2773 | ASSERT(err == 0); | |
2774 | if (attrzp) { | |
2775 | err = zfs_acl_chown_setattr(attrzp); | |
2776 | ASSERT(err == 0); | |
2777 | } | |
2778 | } | |
2779 | ||
3558fd73 BB |
2780 | if (mask & ATTR_MODE) { |
2781 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, | |
428870ff BB |
2782 | &new_mode, sizeof (new_mode)); |
2783 | zp->z_mode = new_mode; | |
99c564bc | 2784 | ASSERT3P(aclp, !=, NULL); |
9babb374 | 2785 | err = zfs_aclset_common(zp, aclp, cr, tx); |
34dc7c2f | 2786 | ASSERT3U(err, ==, 0); |
572e2857 BB |
2787 | if (zp->z_acl_cached) |
2788 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
2789 | zp->z_acl_cached = aclp; |
2790 | aclp = NULL; | |
34dc7c2f BB |
2791 | } |
2792 | ||
34dc7c2f | 2793 | |
3558fd73 | 2794 | if (mask & ATTR_ATIME) { |
5484965a | 2795 | ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime); |
3558fd73 | 2796 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, |
428870ff | 2797 | &zp->z_atime, sizeof (zp->z_atime)); |
34dc7c2f BB |
2798 | } |
2799 | ||
3558fd73 | 2800 | if (mask & ATTR_MTIME) { |
5484965a | 2801 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
3558fd73 | 2802 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, |
428870ff | 2803 | mtime, sizeof (mtime)); |
34dc7c2f BB |
2804 | } |
2805 | ||
b128c09f | 2806 | /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */ |
3558fd73 BB |
2807 | if (mask & ATTR_SIZE && !(mask & ATTR_MTIME)) { |
2808 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), | |
428870ff | 2809 | NULL, mtime, sizeof (mtime)); |
3558fd73 | 2810 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
2811 | &ctime, sizeof (ctime)); |
2812 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, | |
2813 | B_TRUE); | |
2814 | } else if (mask != 0) { | |
3558fd73 | 2815 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
2816 | &ctime, sizeof (ctime)); |
2817 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, | |
2818 | B_TRUE); | |
2819 | if (attrzp) { | |
2820 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3558fd73 | 2821 | SA_ZPL_CTIME(zsb), NULL, |
428870ff BB |
2822 | &ctime, sizeof (ctime)); |
2823 | zfs_tstamp_update_setup(attrzp, STATE_CHANGED, | |
2824 | mtime, ctime, B_TRUE); | |
2825 | } | |
2826 | } | |
34dc7c2f BB |
2827 | /* |
2828 | * Do this after setting timestamps to prevent timestamp | |
2829 | * update from toggling bit | |
2830 | */ | |
2831 | ||
5484965a BB |
2832 | if (xoap && (mask & ATTR_XVATTR)) { |
2833 | ||
2834 | /* | |
2835 | * restore trimmed off masks | |
2836 | * so that return masks can be set for caller. | |
2837 | */ | |
2838 | ||
f4ea75d4 | 2839 | if (XVA_ISSET_REQ(tmpxvattr, XAT_APPENDONLY)) { |
5484965a BB |
2840 | XVA_SET_REQ(xvap, XAT_APPENDONLY); |
2841 | } | |
f4ea75d4 | 2842 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NOUNLINK)) { |
5484965a BB |
2843 | XVA_SET_REQ(xvap, XAT_NOUNLINK); |
2844 | } | |
f4ea75d4 | 2845 | if (XVA_ISSET_REQ(tmpxvattr, XAT_IMMUTABLE)) { |
5484965a BB |
2846 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); |
2847 | } | |
f4ea75d4 | 2848 | if (XVA_ISSET_REQ(tmpxvattr, XAT_NODUMP)) { |
5484965a BB |
2849 | XVA_SET_REQ(xvap, XAT_NODUMP); |
2850 | } | |
f4ea75d4 | 2851 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_MODIFIED)) { |
5484965a BB |
2852 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); |
2853 | } | |
f4ea75d4 | 2854 | if (XVA_ISSET_REQ(tmpxvattr, XAT_AV_QUARANTINED)) { |
5484965a BB |
2855 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); |
2856 | } | |
2857 | ||
2858 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
2859 | ASSERT(S_ISREG(ip->i_mode)); | |
2860 | ||
2861 | zfs_xvattr_set(zp, xvap, tx); | |
2862 | } | |
2863 | ||
9babb374 | 2864 | if (fuid_dirtied) |
3558fd73 | 2865 | zfs_fuid_sync(zsb, tx); |
9babb374 | 2866 | |
34dc7c2f | 2867 | if (mask != 0) |
5484965a | 2868 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); |
34dc7c2f | 2869 | |
34dc7c2f | 2870 | mutex_exit(&zp->z_lock); |
3558fd73 | 2871 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 | 2872 | mutex_exit(&zp->z_acl_lock); |
34dc7c2f | 2873 | |
572e2857 | 2874 | if (attrzp) { |
3558fd73 | 2875 | if (mask & (ATTR_UID|ATTR_GID|ATTR_MODE)) |
572e2857 BB |
2876 | mutex_exit(&attrzp->z_acl_lock); |
2877 | mutex_exit(&attrzp->z_lock); | |
2878 | } | |
9babb374 | 2879 | out: |
428870ff BB |
2880 | if (err == 0 && attrzp) { |
2881 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, | |
2882 | xattr_count, tx); | |
2883 | ASSERT(err2 == 0); | |
2884 | } | |
2885 | ||
34dc7c2f | 2886 | if (attrzp) |
3558fd73 | 2887 | iput(ZTOI(attrzp)); |
45d1cae3 | 2888 | if (aclp) |
9babb374 | 2889 | zfs_acl_free(aclp); |
9babb374 BB |
2890 | |
2891 | if (fuidp) { | |
2892 | zfs_fuid_info_free(fuidp); | |
2893 | fuidp = NULL; | |
2894 | } | |
2895 | ||
428870ff | 2896 | if (err) { |
9babb374 | 2897 | dmu_tx_abort(tx); |
428870ff BB |
2898 | if (err == ERESTART) |
2899 | goto top; | |
2900 | } else { | |
2901 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 2902 | dmu_tx_commit(tx); |
037849f8 | 2903 | zfs_inode_update(zp); |
428870ff BB |
2904 | } |
2905 | ||
428870ff | 2906 | out2: |
3558fd73 | 2907 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2908 | zil_commit(zilog, 0); |
34dc7c2f | 2909 | |
f4ea75d4 | 2910 | out3: |
17c37660 BB |
2911 | kmem_free(xattr_bulk, sizeof(sa_bulk_attr_t) * 7); |
2912 | kmem_free(bulk, sizeof(sa_bulk_attr_t) * 7); | |
f4ea75d4 | 2913 | kmem_free(tmpxvattr, sizeof(xvattr_t)); |
3558fd73 | 2914 | ZFS_EXIT(zsb); |
34dc7c2f BB |
2915 | return (err); |
2916 | } | |
e5c39b95 | 2917 | EXPORT_SYMBOL(zfs_setattr); |
34dc7c2f BB |
2918 | |
2919 | typedef struct zfs_zlock { | |
2920 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
2921 | znode_t *zl_znode; /* znode we held */ | |
2922 | struct zfs_zlock *zl_next; /* next in list */ | |
2923 | } zfs_zlock_t; | |
2924 | ||
2925 | /* | |
2926 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
2927 | */ | |
2928 | static void | |
2929 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
2930 | { | |
2931 | zfs_zlock_t *zl; | |
2932 | ||
2933 | while ((zl = *zlpp) != NULL) { | |
2934 | if (zl->zl_znode != NULL) | |
3558fd73 | 2935 | iput(ZTOI(zl->zl_znode)); |
34dc7c2f BB |
2936 | rw_exit(zl->zl_rwlock); |
2937 | *zlpp = zl->zl_next; | |
2938 | kmem_free(zl, sizeof (*zl)); | |
2939 | } | |
2940 | } | |
2941 | ||
2942 | /* | |
2943 | * Search back through the directory tree, using the ".." entries. | |
2944 | * Lock each directory in the chain to prevent concurrent renames. | |
2945 | * Fail any attempt to move a directory into one of its own descendants. | |
2946 | * XXX - z_parent_lock can overlap with map or grow locks | |
2947 | */ | |
2948 | static int | |
2949 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
2950 | { | |
2951 | zfs_zlock_t *zl; | |
2952 | znode_t *zp = tdzp; | |
3558fd73 | 2953 | uint64_t rootid = ZTOZSB(zp)->z_root; |
428870ff | 2954 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
2955 | krwlock_t *rwlp = &szp->z_parent_lock; |
2956 | krw_t rw = RW_WRITER; | |
2957 | ||
2958 | /* | |
2959 | * First pass write-locks szp and compares to zp->z_id. | |
2960 | * Later passes read-lock zp and compare to zp->z_parent. | |
2961 | */ | |
2962 | do { | |
2963 | if (!rw_tryenter(rwlp, rw)) { | |
2964 | /* | |
2965 | * Another thread is renaming in this path. | |
2966 | * Note that if we are a WRITER, we don't have any | |
2967 | * parent_locks held yet. | |
2968 | */ | |
2969 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
2970 | /* | |
2971 | * Drop our locks and restart | |
2972 | */ | |
2973 | zfs_rename_unlock(&zl); | |
2974 | *zlpp = NULL; | |
2975 | zp = tdzp; | |
428870ff | 2976 | oidp = zp->z_id; |
34dc7c2f BB |
2977 | rwlp = &szp->z_parent_lock; |
2978 | rw = RW_WRITER; | |
2979 | continue; | |
2980 | } else { | |
2981 | /* | |
2982 | * Wait for other thread to drop its locks | |
2983 | */ | |
2984 | rw_enter(rwlp, rw); | |
2985 | } | |
2986 | } | |
2987 | ||
2988 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
2989 | zl->zl_rwlock = rwlp; | |
2990 | zl->zl_znode = NULL; | |
2991 | zl->zl_next = *zlpp; | |
2992 | *zlpp = zl; | |
2993 | ||
428870ff | 2994 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
34dc7c2f BB |
2995 | return (EINVAL); |
2996 | ||
428870ff | 2997 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
2998 | return (0); |
2999 | ||
3000 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
3558fd73 | 3001 | int error = zfs_zget(ZTOZSB(zp), oidp, &zp); |
34dc7c2f BB |
3002 | if (error) |
3003 | return (error); | |
3004 | zl->zl_znode = zp; | |
3005 | } | |
3558fd73 | 3006 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(ZTOZSB(zp)), |
428870ff | 3007 | &oidp, sizeof (oidp)); |
34dc7c2f BB |
3008 | rwlp = &zp->z_parent_lock; |
3009 | rw = RW_READER; | |
3010 | ||
3011 | } while (zp->z_id != sdzp->z_id); | |
3012 | ||
3013 | return (0); | |
3014 | } | |
3015 | ||
3016 | /* | |
3017 | * Move an entry from the provided source directory to the target | |
3018 | * directory. Change the entry name as indicated. | |
3019 | * | |
3558fd73 | 3020 | * IN: sdip - Source directory containing the "old entry". |
34dc7c2f | 3021 | * snm - Old entry name. |
3558fd73 | 3022 | * tdip - Target directory to contain the "new entry". |
34dc7c2f BB |
3023 | * tnm - New entry name. |
3024 | * cr - credentials of caller. | |
34dc7c2f BB |
3025 | * flags - case flags |
3026 | * | |
3027 | * RETURN: 0 if success | |
3028 | * error code if failure | |
3029 | * | |
3030 | * Timestamps: | |
3558fd73 | 3031 | * sdip,tdip - ctime|mtime updated |
34dc7c2f BB |
3032 | */ |
3033 | /*ARGSUSED*/ | |
e5c39b95 | 3034 | int |
3558fd73 BB |
3035 | zfs_rename(struct inode *sdip, char *snm, struct inode *tdip, char *tnm, |
3036 | cred_t *cr, int flags) | |
34dc7c2f BB |
3037 | { |
3038 | znode_t *tdzp, *szp, *tzp; | |
3558fd73 BB |
3039 | znode_t *sdzp = ITOZ(sdip); |
3040 | zfs_sb_t *zsb = ITOZSB(sdip); | |
34dc7c2f | 3041 | zilog_t *zilog; |
34dc7c2f BB |
3042 | zfs_dirlock_t *sdl, *tdl; |
3043 | dmu_tx_t *tx; | |
3044 | zfs_zlock_t *zl; | |
3045 | int cmp, serr, terr; | |
3046 | int error = 0; | |
3047 | int zflg = 0; | |
3048 | ||
3558fd73 | 3049 | ZFS_ENTER(zsb); |
34dc7c2f | 3050 | ZFS_VERIFY_ZP(sdzp); |
3558fd73 | 3051 | zilog = zsb->z_log; |
34dc7c2f | 3052 | |
3558fd73 BB |
3053 | if (tdip->i_sb != sdip->i_sb) { |
3054 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
3055 | return (EXDEV); |
3056 | } | |
3057 | ||
3558fd73 | 3058 | tdzp = ITOZ(tdip); |
34dc7c2f | 3059 | ZFS_VERIFY_ZP(tdzp); |
3558fd73 | 3060 | if (zsb->z_utf8 && u8_validate(tnm, |
34dc7c2f | 3061 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3062 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3063 | return (EILSEQ); |
3064 | } | |
3065 | ||
3066 | if (flags & FIGNORECASE) | |
3067 | zflg |= ZCILOOK; | |
3068 | ||
3069 | top: | |
3070 | szp = NULL; | |
3071 | tzp = NULL; | |
3072 | zl = NULL; | |
3073 | ||
3074 | /* | |
3075 | * This is to prevent the creation of links into attribute space | |
3076 | * by renaming a linked file into/outof an attribute directory. | |
3077 | * See the comment in zfs_link() for why this is considered bad. | |
3078 | */ | |
428870ff | 3079 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 3080 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3081 | return (EINVAL); |
3082 | } | |
3083 | ||
3084 | /* | |
3085 | * Lock source and target directory entries. To prevent deadlock, | |
3086 | * a lock ordering must be defined. We lock the directory with | |
3087 | * the smallest object id first, or if it's a tie, the one with | |
3088 | * the lexically first name. | |
3089 | */ | |
3090 | if (sdzp->z_id < tdzp->z_id) { | |
3091 | cmp = -1; | |
3092 | } else if (sdzp->z_id > tdzp->z_id) { | |
3093 | cmp = 1; | |
3094 | } else { | |
3095 | /* | |
3096 | * First compare the two name arguments without | |
3097 | * considering any case folding. | |
3098 | */ | |
3558fd73 | 3099 | int nofold = (zsb->z_norm & ~U8_TEXTPREP_TOUPPER); |
34dc7c2f BB |
3100 | |
3101 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
3558fd73 | 3102 | ASSERT(error == 0 || !zsb->z_utf8); |
34dc7c2f BB |
3103 | if (cmp == 0) { |
3104 | /* | |
3105 | * POSIX: "If the old argument and the new argument | |
3106 | * both refer to links to the same existing file, | |
3107 | * the rename() function shall return successfully | |
3108 | * and perform no other action." | |
3109 | */ | |
3558fd73 | 3110 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3111 | return (0); |
3112 | } | |
3113 | /* | |
3114 | * If the file system is case-folding, then we may | |
3115 | * have some more checking to do. A case-folding file | |
3116 | * system is either supporting mixed case sensitivity | |
3117 | * access or is completely case-insensitive. Note | |
3118 | * that the file system is always case preserving. | |
3119 | * | |
3120 | * In mixed sensitivity mode case sensitive behavior | |
3121 | * is the default. FIGNORECASE must be used to | |
3122 | * explicitly request case insensitive behavior. | |
3123 | * | |
3124 | * If the source and target names provided differ only | |
3125 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3126 | * we will treat this as a special case in the | |
3127 | * case-insensitive mode: as long as the source name | |
3128 | * is an exact match, we will allow this to proceed as | |
3129 | * a name-change request. | |
3130 | */ | |
3558fd73 BB |
3131 | if ((zsb->z_case == ZFS_CASE_INSENSITIVE || |
3132 | (zsb->z_case == ZFS_CASE_MIXED && | |
34dc7c2f | 3133 | flags & FIGNORECASE)) && |
3558fd73 | 3134 | u8_strcmp(snm, tnm, 0, zsb->z_norm, U8_UNICODE_LATEST, |
34dc7c2f BB |
3135 | &error) == 0) { |
3136 | /* | |
3137 | * case preserving rename request, require exact | |
3138 | * name matches | |
3139 | */ | |
3140 | zflg |= ZCIEXACT; | |
3141 | zflg &= ~ZCILOOK; | |
3142 | } | |
3143 | } | |
3144 | ||
428870ff BB |
3145 | /* |
3146 | * If the source and destination directories are the same, we should | |
3147 | * grab the z_name_lock of that directory only once. | |
3148 | */ | |
3149 | if (sdzp == tdzp) { | |
3150 | zflg |= ZHAVELOCK; | |
3151 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3152 | } | |
3153 | ||
34dc7c2f BB |
3154 | if (cmp < 0) { |
3155 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3156 | ZEXISTS | zflg, NULL, NULL); | |
3157 | terr = zfs_dirent_lock(&tdl, | |
3158 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3159 | } else { | |
3160 | terr = zfs_dirent_lock(&tdl, | |
3161 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3162 | serr = zfs_dirent_lock(&sdl, | |
3163 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3164 | NULL, NULL); | |
3165 | } | |
3166 | ||
3167 | if (serr) { | |
3168 | /* | |
3169 | * Source entry invalid or not there. | |
3170 | */ | |
3171 | if (!terr) { | |
3172 | zfs_dirent_unlock(tdl); | |
3173 | if (tzp) | |
3558fd73 | 3174 | iput(ZTOI(tzp)); |
34dc7c2f | 3175 | } |
428870ff BB |
3176 | |
3177 | if (sdzp == tdzp) | |
3178 | rw_exit(&sdzp->z_name_lock); | |
3179 | ||
34dc7c2f BB |
3180 | if (strcmp(snm, "..") == 0) |
3181 | serr = EINVAL; | |
3558fd73 | 3182 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3183 | return (serr); |
3184 | } | |
3185 | if (terr) { | |
3186 | zfs_dirent_unlock(sdl); | |
3558fd73 | 3187 | iput(ZTOI(szp)); |
428870ff BB |
3188 | |
3189 | if (sdzp == tdzp) | |
3190 | rw_exit(&sdzp->z_name_lock); | |
3191 | ||
34dc7c2f BB |
3192 | if (strcmp(tnm, "..") == 0) |
3193 | terr = EINVAL; | |
3558fd73 | 3194 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3195 | return (terr); |
3196 | } | |
3197 | ||
3198 | /* | |
3199 | * Must have write access at the source to remove the old entry | |
3200 | * and write access at the target to create the new entry. | |
3201 | * Note that if target and source are the same, this can be | |
3202 | * done in a single check. | |
3203 | */ | |
3204 | ||
149e873a | 3205 | if ((error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr))) |
34dc7c2f BB |
3206 | goto out; |
3207 | ||
3558fd73 | 3208 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
34dc7c2f BB |
3209 | /* |
3210 | * Check to make sure rename is valid. | |
3211 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3212 | */ | |
149e873a | 3213 | if ((error = zfs_rename_lock(szp, tdzp, sdzp, &zl))) |
34dc7c2f BB |
3214 | goto out; |
3215 | } | |
3216 | ||
3217 | /* | |
3218 | * Does target exist? | |
3219 | */ | |
3220 | if (tzp) { | |
3221 | /* | |
3222 | * Source and target must be the same type. | |
3223 | */ | |
3558fd73 BB |
3224 | if (S_ISDIR(ZTOI(szp)->i_mode)) { |
3225 | if (!S_ISDIR(ZTOI(tzp)->i_mode)) { | |
34dc7c2f BB |
3226 | error = ENOTDIR; |
3227 | goto out; | |
3228 | } | |
3229 | } else { | |
3558fd73 | 3230 | if (S_ISDIR(ZTOI(tzp)->i_mode)) { |
34dc7c2f BB |
3231 | error = EISDIR; |
3232 | goto out; | |
3233 | } | |
3234 | } | |
3235 | /* | |
3236 | * POSIX dictates that when the source and target | |
3237 | * entries refer to the same file object, rename | |
3238 | * must do nothing and exit without error. | |
3239 | */ | |
3240 | if (szp->z_id == tzp->z_id) { | |
3241 | error = 0; | |
3242 | goto out; | |
3243 | } | |
3244 | } | |
3245 | ||
3558fd73 | 3246 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
3247 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3248 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3249 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3250 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3251 | if (sdzp != tdzp) { |
3252 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3253 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3254 | } | |
3255 | if (tzp) { | |
3256 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3257 | zfs_sa_upgrade_txholds(tx, tzp); | |
3258 | } | |
3259 | ||
3260 | zfs_sa_upgrade_txholds(tx, szp); | |
3558fd73 | 3261 | dmu_tx_hold_zap(tx, zsb->z_unlinkedobj, FALSE, NULL); |
fb5f0bc8 | 3262 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
3263 | if (error) { |
3264 | if (zl != NULL) | |
3265 | zfs_rename_unlock(&zl); | |
3266 | zfs_dirent_unlock(sdl); | |
3267 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3268 | |
3269 | if (sdzp == tdzp) | |
3270 | rw_exit(&sdzp->z_name_lock); | |
3271 | ||
3558fd73 | 3272 | iput(ZTOI(szp)); |
34dc7c2f | 3273 | if (tzp) |
3558fd73 | 3274 | iput(ZTOI(tzp)); |
fb5f0bc8 | 3275 | if (error == ERESTART) { |
34dc7c2f BB |
3276 | dmu_tx_wait(tx); |
3277 | dmu_tx_abort(tx); | |
3278 | goto top; | |
3279 | } | |
3280 | dmu_tx_abort(tx); | |
3558fd73 | 3281 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3282 | return (error); |
3283 | } | |
3284 | ||
3285 | if (tzp) /* Attempt to remove the existing target */ | |
3286 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3287 | ||
3288 | if (error == 0) { | |
3289 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3290 | if (error == 0) { | |
428870ff | 3291 | szp->z_pflags |= ZFS_AV_MODIFIED; |
34dc7c2f | 3292 | |
3558fd73 | 3293 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zsb), |
428870ff BB |
3294 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); |
3295 | ASSERT3U(error, ==, 0); | |
34dc7c2f | 3296 | |
428870ff BB |
3297 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3298 | if (error == 0) { | |
3299 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3300 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3301 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
428870ff BB |
3302 | } else { |
3303 | /* | |
3304 | * At this point, we have successfully created | |
3305 | * the target name, but have failed to remove | |
3306 | * the source name. Since the create was done | |
3307 | * with the ZRENAMING flag, there are | |
3308 | * complications; for one, the link count is | |
3309 | * wrong. The easiest way to deal with this | |
3310 | * is to remove the newly created target, and | |
3311 | * return the original error. This must | |
3312 | * succeed; fortunately, it is very unlikely to | |
3313 | * fail, since we just created it. | |
3314 | */ | |
3315 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3316 | ZRENAMING, NULL), ==, 0); | |
3317 | } | |
34dc7c2f BB |
3318 | } |
3319 | } | |
3320 | ||
3321 | dmu_tx_commit(tx); | |
3322 | out: | |
3323 | if (zl != NULL) | |
3324 | zfs_rename_unlock(&zl); | |
3325 | ||
3326 | zfs_dirent_unlock(sdl); | |
3327 | zfs_dirent_unlock(tdl); | |
3328 | ||
960e08fe | 3329 | zfs_inode_update(sdzp); |
428870ff BB |
3330 | if (sdzp == tdzp) |
3331 | rw_exit(&sdzp->z_name_lock); | |
3332 | ||
960e08fe BB |
3333 | if (sdzp != tdzp) |
3334 | zfs_inode_update(tdzp); | |
428870ff | 3335 | |
960e08fe | 3336 | zfs_inode_update(szp); |
3558fd73 | 3337 | iput(ZTOI(szp)); |
960e08fe BB |
3338 | if (tzp) { |
3339 | zfs_inode_update(tzp); | |
3558fd73 | 3340 | iput(ZTOI(tzp)); |
960e08fe | 3341 | } |
34dc7c2f | 3342 | |
3558fd73 | 3343 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3344 | zil_commit(zilog, 0); |
428870ff | 3345 | |
3558fd73 | 3346 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3347 | return (error); |
3348 | } | |
e5c39b95 | 3349 | EXPORT_SYMBOL(zfs_rename); |
34dc7c2f BB |
3350 | |
3351 | /* | |
3352 | * Insert the indicated symbolic reference entry into the directory. | |
3353 | * | |
3558fd73 | 3354 | * IN: dip - Directory to contain new symbolic link. |
34dc7c2f BB |
3355 | * link - Name for new symlink entry. |
3356 | * vap - Attributes of new entry. | |
3357 | * target - Target path of new symlink. | |
3558fd73 | 3358 | * |
34dc7c2f | 3359 | * cr - credentials of caller. |
34dc7c2f BB |
3360 | * flags - case flags |
3361 | * | |
3362 | * RETURN: 0 if success | |
3363 | * error code if failure | |
3364 | * | |
3365 | * Timestamps: | |
3558fd73 | 3366 | * dip - ctime|mtime updated |
34dc7c2f BB |
3367 | */ |
3368 | /*ARGSUSED*/ | |
e5c39b95 | 3369 | int |
3558fd73 BB |
3370 | zfs_symlink(struct inode *dip, char *name, vattr_t *vap, char *link, |
3371 | struct inode **ipp, cred_t *cr, int flags) | |
34dc7c2f | 3372 | { |
3558fd73 | 3373 | znode_t *zp, *dzp = ITOZ(dip); |
34dc7c2f BB |
3374 | zfs_dirlock_t *dl; |
3375 | dmu_tx_t *tx; | |
3558fd73 | 3376 | zfs_sb_t *zsb = ITOZSB(dip); |
34dc7c2f | 3377 | zilog_t *zilog; |
428870ff | 3378 | uint64_t len = strlen(link); |
34dc7c2f BB |
3379 | int error; |
3380 | int zflg = ZNEW; | |
9babb374 BB |
3381 | zfs_acl_ids_t acl_ids; |
3382 | boolean_t fuid_dirtied; | |
428870ff | 3383 | uint64_t txtype = TX_SYMLINK; |
34dc7c2f | 3384 | |
3558fd73 | 3385 | ASSERT(S_ISLNK(vap->va_mode)); |
34dc7c2f | 3386 | |
3558fd73 | 3387 | ZFS_ENTER(zsb); |
34dc7c2f | 3388 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3389 | zilog = zsb->z_log; |
34dc7c2f | 3390 | |
3558fd73 | 3391 | if (zsb->z_utf8 && u8_validate(name, strlen(name), |
34dc7c2f | 3392 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3393 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3394 | return (EILSEQ); |
3395 | } | |
3396 | if (flags & FIGNORECASE) | |
3397 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3398 | |
3399 | if (len > MAXPATHLEN) { | |
3558fd73 | 3400 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3401 | return (ENAMETOOLONG); |
3402 | } | |
3403 | ||
428870ff BB |
3404 | if ((error = zfs_acl_ids_create(dzp, 0, |
3405 | vap, cr, NULL, &acl_ids)) != 0) { | |
3558fd73 | 3406 | ZFS_EXIT(zsb); |
428870ff BB |
3407 | return (error); |
3408 | } | |
3409 | top: | |
3558fd73 BB |
3410 | *ipp = NULL; |
3411 | ||
34dc7c2f BB |
3412 | /* |
3413 | * Attempt to lock directory; fail if entry already exists. | |
3414 | */ | |
3415 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3416 | if (error) { | |
428870ff | 3417 | zfs_acl_ids_free(&acl_ids); |
3558fd73 | 3418 | ZFS_EXIT(zsb); |
428870ff BB |
3419 | return (error); |
3420 | } | |
3421 | ||
149e873a | 3422 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
428870ff BB |
3423 | zfs_acl_ids_free(&acl_ids); |
3424 | zfs_dirent_unlock(dl); | |
3558fd73 | 3425 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3426 | return (error); |
3427 | } | |
3428 | ||
3558fd73 | 3429 | if (zfs_acl_ids_overquota(zsb, &acl_ids)) { |
9babb374 BB |
3430 | zfs_acl_ids_free(&acl_ids); |
3431 | zfs_dirent_unlock(dl); | |
3558fd73 | 3432 | ZFS_EXIT(zsb); |
9babb374 BB |
3433 | return (EDQUOT); |
3434 | } | |
3558fd73 BB |
3435 | tx = dmu_tx_create(zsb->z_os); |
3436 | fuid_dirtied = zsb->z_fuid_dirty; | |
34dc7c2f | 3437 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3438 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3439 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3440 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3441 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
3558fd73 | 3442 | if (!zsb->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
428870ff BB |
3443 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, |
3444 | acl_ids.z_aclp->z_acl_bytes); | |
3445 | } | |
9babb374 | 3446 | if (fuid_dirtied) |
3558fd73 | 3447 | zfs_fuid_txhold(zsb, tx); |
fb5f0bc8 | 3448 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
3449 | if (error) { |
3450 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3451 | if (error == ERESTART) { |
34dc7c2f BB |
3452 | dmu_tx_wait(tx); |
3453 | dmu_tx_abort(tx); | |
3454 | goto top; | |
3455 | } | |
428870ff | 3456 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f | 3457 | dmu_tx_abort(tx); |
3558fd73 | 3458 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3459 | return (error); |
3460 | } | |
3461 | ||
34dc7c2f BB |
3462 | /* |
3463 | * Create a new object for the symlink. | |
428870ff | 3464 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 3465 | */ |
428870ff | 3466 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3467 | |
428870ff | 3468 | if (fuid_dirtied) |
3558fd73 | 3469 | zfs_fuid_sync(zsb, tx); |
34dc7c2f | 3470 | |
572e2857 | 3471 | mutex_enter(&zp->z_lock); |
428870ff | 3472 | if (zp->z_is_sa) |
3558fd73 | 3473 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zsb), |
428870ff BB |
3474 | link, len, tx); |
3475 | else | |
3476 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3477 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3478 | |
428870ff | 3479 | zp->z_size = len; |
3558fd73 | 3480 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zsb), |
428870ff | 3481 | &zp->z_size, sizeof (zp->z_size), tx); |
34dc7c2f BB |
3482 | /* |
3483 | * Insert the new object into the directory. | |
3484 | */ | |
3485 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
428870ff BB |
3486 | |
3487 | if (flags & FIGNORECASE) | |
3488 | txtype |= TX_CI; | |
3489 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
9babb374 | 3490 | |
960e08fe BB |
3491 | zfs_inode_update(dzp); |
3492 | zfs_inode_update(zp); | |
3493 | ||
9babb374 | 3494 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3495 | |
3496 | dmu_tx_commit(tx); | |
3497 | ||
3498 | zfs_dirent_unlock(dl); | |
3499 | ||
3558fd73 | 3500 | *ipp = ZTOI(zp); |
34dc7c2f | 3501 | |
3558fd73 | 3502 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3503 | zil_commit(zilog, 0); |
428870ff | 3504 | |
3558fd73 | 3505 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3506 | return (error); |
3507 | } | |
e5c39b95 | 3508 | EXPORT_SYMBOL(zfs_symlink); |
34dc7c2f BB |
3509 | |
3510 | /* | |
3511 | * Return, in the buffer contained in the provided uio structure, | |
3558fd73 | 3512 | * the symbolic path referred to by ip. |
34dc7c2f | 3513 | * |
8b4f9a2d BB |
3514 | * IN: ip - inode of symbolic link |
3515 | * uio - structure to contain the link path. | |
3516 | * cr - credentials of caller. | |
34dc7c2f BB |
3517 | * |
3518 | * RETURN: 0 if success | |
3519 | * error code if failure | |
3520 | * | |
3521 | * Timestamps: | |
3558fd73 | 3522 | * ip - atime updated |
34dc7c2f BB |
3523 | */ |
3524 | /* ARGSUSED */ | |
e5c39b95 | 3525 | int |
8b4f9a2d | 3526 | zfs_readlink(struct inode *ip, uio_t *uio, cred_t *cr) |
34dc7c2f | 3527 | { |
3558fd73 BB |
3528 | znode_t *zp = ITOZ(ip); |
3529 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
3530 | int error; |
3531 | ||
3558fd73 | 3532 | ZFS_ENTER(zsb); |
34dc7c2f BB |
3533 | ZFS_VERIFY_ZP(zp); |
3534 | ||
572e2857 | 3535 | mutex_enter(&zp->z_lock); |
428870ff | 3536 | if (zp->z_is_sa) |
8b4f9a2d BB |
3537 | error = sa_lookup_uio(zp->z_sa_hdl, |
3538 | SA_ZPL_SYMLINK(zsb), uio); | |
428870ff | 3539 | else |
8b4f9a2d | 3540 | error = zfs_sa_readlink(zp, uio); |
572e2857 | 3541 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3542 | |
3558fd73 | 3543 | ZFS_ACCESSTIME_STAMP(zsb, zp); |
960e08fe | 3544 | zfs_inode_update(zp); |
3558fd73 | 3545 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3546 | return (error); |
3547 | } | |
8b4f9a2d | 3548 | EXPORT_SYMBOL(zfs_readlink); |
34dc7c2f BB |
3549 | |
3550 | /* | |
3558fd73 | 3551 | * Insert a new entry into directory tdip referencing sip. |
34dc7c2f | 3552 | * |
3558fd73 BB |
3553 | * IN: tdip - Directory to contain new entry. |
3554 | * sip - inode of new entry. | |
34dc7c2f BB |
3555 | * name - name of new entry. |
3556 | * cr - credentials of caller. | |
34dc7c2f BB |
3557 | * |
3558 | * RETURN: 0 if success | |
3559 | * error code if failure | |
3560 | * | |
3561 | * Timestamps: | |
3558fd73 BB |
3562 | * tdip - ctime|mtime updated |
3563 | * sip - ctime updated | |
34dc7c2f BB |
3564 | */ |
3565 | /* ARGSUSED */ | |
e5c39b95 | 3566 | int |
3558fd73 | 3567 | zfs_link(struct inode *tdip, struct inode *sip, char *name, cred_t *cr) |
34dc7c2f | 3568 | { |
3558fd73 | 3569 | znode_t *dzp = ITOZ(tdip); |
34dc7c2f | 3570 | znode_t *tzp, *szp; |
3558fd73 | 3571 | zfs_sb_t *zsb = ITOZSB(tdip); |
34dc7c2f BB |
3572 | zilog_t *zilog; |
3573 | zfs_dirlock_t *dl; | |
3574 | dmu_tx_t *tx; | |
34dc7c2f BB |
3575 | int error; |
3576 | int zf = ZNEW; | |
428870ff | 3577 | uint64_t parent; |
572e2857 | 3578 | uid_t owner; |
34dc7c2f | 3579 | |
3558fd73 | 3580 | ASSERT(S_ISDIR(tdip->i_mode)); |
34dc7c2f | 3581 | |
3558fd73 | 3582 | ZFS_ENTER(zsb); |
34dc7c2f | 3583 | ZFS_VERIFY_ZP(dzp); |
3558fd73 | 3584 | zilog = zsb->z_log; |
34dc7c2f | 3585 | |
428870ff BB |
3586 | /* |
3587 | * POSIX dictates that we return EPERM here. | |
3588 | * Better choices include ENOTSUP or EISDIR. | |
3589 | */ | |
3558fd73 BB |
3590 | if (S_ISDIR(sip->i_mode)) { |
3591 | ZFS_EXIT(zsb); | |
428870ff BB |
3592 | return (EPERM); |
3593 | } | |
3594 | ||
3558fd73 BB |
3595 | if (sip->i_sb != tdip->i_sb) { |
3596 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
3597 | return (EXDEV); |
3598 | } | |
428870ff | 3599 | |
3558fd73 | 3600 | szp = ITOZ(sip); |
34dc7c2f BB |
3601 | ZFS_VERIFY_ZP(szp); |
3602 | ||
428870ff BB |
3603 | /* Prevent links to .zfs/shares files */ |
3604 | ||
3558fd73 | 3605 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zsb), |
428870ff | 3606 | &parent, sizeof (uint64_t))) != 0) { |
3558fd73 | 3607 | ZFS_EXIT(zsb); |
428870ff BB |
3608 | return (error); |
3609 | } | |
3558fd73 BB |
3610 | if (parent == zsb->z_shares_dir) { |
3611 | ZFS_EXIT(zsb); | |
428870ff BB |
3612 | return (EPERM); |
3613 | } | |
3614 | ||
3558fd73 | 3615 | if (zsb->z_utf8 && u8_validate(name, |
34dc7c2f | 3616 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { |
3558fd73 | 3617 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3618 | return (EILSEQ); |
3619 | } | |
3558fd73 | 3620 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
3621 | if (flags & FIGNORECASE) |
3622 | zf |= ZCILOOK; | |
3558fd73 | 3623 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f | 3624 | |
34dc7c2f BB |
3625 | /* |
3626 | * We do not support links between attributes and non-attributes | |
3627 | * because of the potential security risk of creating links | |
3628 | * into "normal" file space in order to circumvent restrictions | |
3629 | * imposed in attribute space. | |
3630 | */ | |
428870ff | 3631 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
3558fd73 | 3632 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3633 | return (EINVAL); |
3634 | } | |
3635 | ||
3558fd73 | 3636 | owner = zfs_fuid_map_id(zsb, szp->z_uid, cr, ZFS_OWNER); |
572e2857 | 3637 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { |
3558fd73 | 3638 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3639 | return (EPERM); |
3640 | } | |
3641 | ||
149e873a | 3642 | if ((error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr))) { |
3558fd73 | 3643 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3644 | return (error); |
3645 | } | |
3646 | ||
428870ff | 3647 | top: |
34dc7c2f BB |
3648 | /* |
3649 | * Attempt to lock directory; fail if entry already exists. | |
3650 | */ | |
3651 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
3652 | if (error) { | |
3558fd73 | 3653 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3654 | return (error); |
3655 | } | |
3656 | ||
3558fd73 | 3657 | tx = dmu_tx_create(zsb->z_os); |
428870ff | 3658 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 3659 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3660 | zfs_sa_upgrade_txholds(tx, szp); |
3661 | zfs_sa_upgrade_txholds(tx, dzp); | |
fb5f0bc8 | 3662 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
3663 | if (error) { |
3664 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3665 | if (error == ERESTART) { |
34dc7c2f BB |
3666 | dmu_tx_wait(tx); |
3667 | dmu_tx_abort(tx); | |
3668 | goto top; | |
3669 | } | |
3670 | dmu_tx_abort(tx); | |
3558fd73 | 3671 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3672 | return (error); |
3673 | } | |
3674 | ||
3675 | error = zfs_link_create(dl, szp, tx, 0); | |
3676 | ||
3677 | if (error == 0) { | |
3678 | uint64_t txtype = TX_LINK; | |
3558fd73 | 3679 | #ifdef HAVE_PN_UTILS |
34dc7c2f BB |
3680 | if (flags & FIGNORECASE) |
3681 | txtype |= TX_CI; | |
3558fd73 | 3682 | #endif /* HAVE_PN_UTILS */ |
34dc7c2f BB |
3683 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); |
3684 | } | |
3685 | ||
3686 | dmu_tx_commit(tx); | |
3687 | ||
3688 | zfs_dirent_unlock(dl); | |
3689 | ||
3558fd73 | 3690 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3691 | zil_commit(zilog, 0); |
428870ff | 3692 | |
960e08fe BB |
3693 | zfs_inode_update(dzp); |
3694 | zfs_inode_update(szp); | |
3558fd73 | 3695 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3696 | return (error); |
3697 | } | |
e5c39b95 | 3698 | EXPORT_SYMBOL(zfs_link); |
34dc7c2f | 3699 | |
34dc7c2f | 3700 | /* |
dde471ef | 3701 | * Push a page out to disk |
34dc7c2f BB |
3702 | * |
3703 | * IN: vp - file to push page to. | |
3704 | * pp - page to push. | |
dde471ef PJ |
3705 | * off - start of range pushed. |
3706 | * len - len of range pushed. | |
34dc7c2f | 3707 | * |
34dc7c2f BB |
3708 | * |
3709 | * RETURN: 0 if success | |
3710 | * error code if failure | |
3711 | * | |
dde471ef | 3712 | * NOTE: callers must have locked the page to be pushed. |
34dc7c2f BB |
3713 | */ |
3714 | /* ARGSUSED */ | |
3715 | static int | |
dde471ef | 3716 | zfs_putapage(struct inode *ip, struct page *pp, u_offset_t off, size_t len) |
34dc7c2f | 3717 | { |
dde471ef PJ |
3718 | znode_t *zp = ITOZ(ip); |
3719 | zfs_sb_t *zsb = ITOZSB(ip); | |
3720 | dmu_tx_t *tx; | |
3721 | caddr_t va; | |
3722 | int err; | |
d164b209 | 3723 | |
34dc7c2f BB |
3724 | /* |
3725 | * Can't push pages past end-of-file. | |
3726 | */ | |
428870ff | 3727 | if (off >= zp->z_size) { |
34dc7c2f BB |
3728 | /* ignore all pages */ |
3729 | err = 0; | |
3730 | goto out; | |
dde471ef | 3731 | } else if (off + len > zp->z_size) |
428870ff | 3732 | len = zp->z_size - off; |
9babb374 | 3733 | |
dde471ef PJ |
3734 | if (zfs_owner_overquota(zsb, zp, B_FALSE) || |
3735 | zfs_owner_overquota(zsb, zp, B_TRUE)) { | |
9babb374 BB |
3736 | err = EDQUOT; |
3737 | goto out; | |
3738 | } | |
d164b209 | 3739 | top: |
dde471ef | 3740 | tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 3741 | dmu_tx_hold_write(tx, zp->z_id, off, len); |
428870ff BB |
3742 | |
3743 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
3744 | zfs_sa_upgrade_txholds(tx, zp); | |
fb5f0bc8 | 3745 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 3746 | if (err != 0) { |
fb5f0bc8 | 3747 | if (err == ERESTART) { |
34dc7c2f BB |
3748 | dmu_tx_wait(tx); |
3749 | dmu_tx_abort(tx); | |
34dc7c2f BB |
3750 | goto top; |
3751 | } | |
3752 | dmu_tx_abort(tx); | |
3753 | goto out; | |
3754 | } | |
3755 | ||
dde471ef PJ |
3756 | va = kmap(pp); |
3757 | ASSERT3U(len, <=, PAGESIZE); | |
3758 | dmu_write(zsb->z_os, zp->z_id, off, len, va, tx); | |
3759 | kunmap(pp); | |
34dc7c2f BB |
3760 | |
3761 | if (err == 0) { | |
428870ff BB |
3762 | uint64_t mtime[2], ctime[2]; |
3763 | sa_bulk_attr_t bulk[3]; | |
3764 | int count = 0; | |
3765 | ||
dde471ef | 3766 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, |
428870ff | 3767 | &mtime, 16); |
dde471ef | 3768 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, |
428870ff | 3769 | &ctime, 16); |
dde471ef | 3770 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff BB |
3771 | &zp->z_pflags, 8); |
3772 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, | |
3773 | B_TRUE); | |
dde471ef | 3774 | zfs_log_write(zsb->z_log, tx, TX_WRITE, zp, off, len, 0); |
34dc7c2f | 3775 | } |
45d1cae3 | 3776 | dmu_tx_commit(tx); |
34dc7c2f BB |
3777 | |
3778 | out: | |
34dc7c2f BB |
3779 | return (err); |
3780 | } | |
3781 | ||
3782 | /* | |
dde471ef | 3783 | * Copy the portion of the file indicated from page into the file. |
34dc7c2f | 3784 | * |
dde471ef PJ |
3785 | * IN: ip - inode of file to push page data to. |
3786 | * wbc - Unused parameter | |
3787 | * data - pointer to address_space | |
34dc7c2f BB |
3788 | * |
3789 | * RETURN: 0 if success | |
3790 | * error code if failure | |
3791 | * | |
3792 | * Timestamps: | |
3793 | * vp - ctime|mtime updated | |
3794 | */ | |
3795 | /*ARGSUSED*/ | |
dde471ef PJ |
3796 | int |
3797 | zfs_putpage(struct page *page, struct writeback_control *wbc, void *data) | |
34dc7c2f | 3798 | { |
dde471ef PJ |
3799 | struct address_space *mapping = data; |
3800 | struct inode *ip = mapping->host; | |
3801 | znode_t *zp = ITOZ(ip); | |
3802 | zfs_sb_t *zsb = ITOZSB(ip); | |
3803 | rl_t *rl; | |
3804 | u_offset_t io_off; | |
3805 | size_t io_len; | |
3806 | size_t len; | |
3807 | int error; | |
3808 | ||
3809 | io_off = page_offset(page); | |
3810 | io_len = PAGESIZE; | |
34dc7c2f | 3811 | |
dde471ef | 3812 | ZFS_ENTER(zsb); |
34dc7c2f BB |
3813 | ZFS_VERIFY_ZP(zp); |
3814 | ||
d164b209 | 3815 | rl = zfs_range_lock(zp, io_off, io_len, RL_WRITER); |
34dc7c2f | 3816 | |
dde471ef | 3817 | if (io_off > zp->z_size) { |
34dc7c2f | 3818 | /* past end of file */ |
d164b209 | 3819 | zfs_range_unlock(rl); |
dde471ef | 3820 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3821 | return (0); |
3822 | } | |
3823 | ||
428870ff | 3824 | len = MIN(io_len, P2ROUNDUP(zp->z_size, PAGESIZE) - io_off); |
34dc7c2f | 3825 | |
dde471ef | 3826 | error = zfs_putapage(ip, page, io_off, len); |
d164b209 | 3827 | zfs_range_unlock(rl); |
dde471ef PJ |
3828 | |
3829 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) | |
3830 | zil_commit(zsb->z_log, zp->z_id); | |
3831 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
3832 | return (error); |
3833 | } | |
dde471ef | 3834 | EXPORT_SYMBOL(zfs_putpage); |
34dc7c2f BB |
3835 | |
3836 | /*ARGSUSED*/ | |
3837 | void | |
c0d35759 | 3838 | zfs_inactive(struct inode *ip) |
34dc7c2f | 3839 | { |
c0d35759 BB |
3840 | znode_t *zp = ITOZ(ip); |
3841 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
3842 | int error; |
3843 | ||
c0d35759 BB |
3844 | #ifdef HAVE_SNAPSHOT |
3845 | /* Early return for snapshot inode? */ | |
3846 | #endif /* HAVE_SNAPSHOT */ | |
34dc7c2f | 3847 | |
c0d35759 BB |
3848 | rw_enter(&zsb->z_teardown_inactive_lock, RW_READER); |
3849 | if (zp->z_sa_hdl == NULL) { | |
3850 | rw_exit(&zsb->z_teardown_inactive_lock); | |
3851 | return; | |
34dc7c2f BB |
3852 | } |
3853 | ||
3854 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
c0d35759 | 3855 | dmu_tx_t *tx = dmu_tx_create(zsb->z_os); |
34dc7c2f | 3856 | |
428870ff BB |
3857 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
3858 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
3859 | error = dmu_tx_assign(tx, TXG_WAIT); |
3860 | if (error) { | |
3861 | dmu_tx_abort(tx); | |
3862 | } else { | |
34dc7c2f | 3863 | mutex_enter(&zp->z_lock); |
3558fd73 | 3864 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), |
428870ff | 3865 | (void *)&zp->z_atime, sizeof (zp->z_atime), tx); |
34dc7c2f BB |
3866 | zp->z_atime_dirty = 0; |
3867 | mutex_exit(&zp->z_lock); | |
3868 | dmu_tx_commit(tx); | |
3869 | } | |
3870 | } | |
3871 | ||
3872 | zfs_zinactive(zp); | |
3558fd73 | 3873 | rw_exit(&zsb->z_teardown_inactive_lock); |
34dc7c2f | 3874 | } |
e5c39b95 | 3875 | EXPORT_SYMBOL(zfs_inactive); |
34dc7c2f BB |
3876 | |
3877 | /* | |
3878 | * Bounds-check the seek operation. | |
3879 | * | |
3558fd73 | 3880 | * IN: ip - inode seeking within |
34dc7c2f BB |
3881 | * ooff - old file offset |
3882 | * noffp - pointer to new file offset | |
3883 | * ct - caller context | |
3884 | * | |
3885 | * RETURN: 0 if success | |
3886 | * EINVAL if new offset invalid | |
3887 | */ | |
3888 | /* ARGSUSED */ | |
3558fd73 | 3889 | int |
9623f736 | 3890 | zfs_seek(struct inode *ip, offset_t ooff, offset_t *noffp) |
34dc7c2f | 3891 | { |
3558fd73 | 3892 | if (S_ISDIR(ip->i_mode)) |
34dc7c2f BB |
3893 | return (0); |
3894 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
3895 | } | |
3558fd73 | 3896 | EXPORT_SYMBOL(zfs_seek); |
34dc7c2f | 3897 | |
34dc7c2f | 3898 | /* |
dde471ef | 3899 | * Fill pages with data from the disk. |
34dc7c2f BB |
3900 | */ |
3901 | static int | |
dde471ef | 3902 | zfs_fillpage(struct inode *ip, struct page *pl[], int nr_pages) |
34dc7c2f | 3903 | { |
dde471ef PJ |
3904 | znode_t *zp = ITOZ(ip); |
3905 | zfs_sb_t *zsb = ITOZSB(ip); | |
3906 | objset_t *os; | |
3907 | struct page *cur_pp; | |
3908 | u_offset_t io_off, total; | |
3909 | size_t io_len; | |
3910 | loff_t i_size; | |
3911 | unsigned page_idx; | |
3912 | int err; | |
34dc7c2f | 3913 | |
dde471ef PJ |
3914 | os = zsb->z_os; |
3915 | io_len = nr_pages << PAGE_CACHE_SHIFT; | |
3916 | i_size = i_size_read(ip); | |
3917 | io_off = page_offset(pl[0]); | |
3918 | ||
3919 | if (io_off + io_len > i_size) | |
3920 | io_len = i_size - io_off; | |
34dc7c2f BB |
3921 | |
3922 | /* | |
dde471ef | 3923 | * Iterate over list of pages and read each page individually. |
34dc7c2f | 3924 | */ |
dde471ef PJ |
3925 | page_idx = 0; |
3926 | cur_pp = pl[0]; | |
34dc7c2f | 3927 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { |
d164b209 BB |
3928 | caddr_t va; |
3929 | ||
dde471ef | 3930 | va = kmap(cur_pp); |
9babb374 BB |
3931 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
3932 | DMU_READ_PREFETCH); | |
dde471ef | 3933 | kunmap(cur_pp); |
34dc7c2f | 3934 | if (err) { |
b128c09f BB |
3935 | /* convert checksum errors into IO errors */ |
3936 | if (err == ECKSUM) | |
3937 | err = EIO; | |
34dc7c2f BB |
3938 | return (err); |
3939 | } | |
dde471ef | 3940 | cur_pp = pl[++page_idx]; |
34dc7c2f | 3941 | } |
d164b209 | 3942 | |
34dc7c2f BB |
3943 | return (0); |
3944 | } | |
3945 | ||
3946 | /* | |
dde471ef | 3947 | * Uses zfs_fillpage to read data from the file and fill the pages. |
34dc7c2f | 3948 | * |
dde471ef PJ |
3949 | * IN: ip - inode of file to get data from. |
3950 | * pl - list of pages to read | |
3951 | * nr_pages - number of pages to read | |
34dc7c2f BB |
3952 | * |
3953 | * RETURN: 0 if success | |
3954 | * error code if failure | |
3955 | * | |
3956 | * Timestamps: | |
3957 | * vp - atime updated | |
3958 | */ | |
3959 | /* ARGSUSED */ | |
dde471ef PJ |
3960 | int |
3961 | zfs_getpage(struct inode *ip, struct page *pl[], int nr_pages) | |
34dc7c2f | 3962 | { |
dde471ef PJ |
3963 | znode_t *zp = ITOZ(ip); |
3964 | zfs_sb_t *zsb = ITOZSB(ip); | |
3965 | int err; | |
d164b209 | 3966 | |
d164b209 BB |
3967 | if (pl == NULL) |
3968 | return (0); | |
34dc7c2f | 3969 | |
dde471ef | 3970 | ZFS_ENTER(zsb); |
34dc7c2f BB |
3971 | ZFS_VERIFY_ZP(zp); |
3972 | ||
dde471ef | 3973 | err = zfs_fillpage(ip, pl, nr_pages); |
34dc7c2f | 3974 | |
dde471ef PJ |
3975 | if (!err) |
3976 | ZFS_ACCESSTIME_STAMP(zsb, zp); | |
34dc7c2f | 3977 | |
dde471ef | 3978 | ZFS_EXIT(zsb); |
34dc7c2f BB |
3979 | return (err); |
3980 | } | |
dde471ef | 3981 | EXPORT_SYMBOL(zfs_getpage); |
34dc7c2f BB |
3982 | |
3983 | /* | |
e2e7aa2d | 3984 | * Check ZFS specific permissions to memory map a section of a file. |
34dc7c2f | 3985 | * |
e2e7aa2d BB |
3986 | * IN: ip - inode of the file to mmap |
3987 | * off - file offset | |
3988 | * addrp - start address in memory region | |
3989 | * len - length of memory region | |
3990 | * vm_flags- address flags | |
34dc7c2f | 3991 | * |
e2e7aa2d BB |
3992 | * RETURN: 0 if success |
3993 | * error code if failure | |
34dc7c2f BB |
3994 | */ |
3995 | /*ARGSUSED*/ | |
e2e7aa2d BB |
3996 | int |
3997 | zfs_map(struct inode *ip, offset_t off, caddr_t *addrp, size_t len, | |
3998 | unsigned long vm_flags) | |
34dc7c2f | 3999 | { |
e2e7aa2d BB |
4000 | znode_t *zp = ITOZ(ip); |
4001 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4002 | |
e2e7aa2d | 4003 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4004 | ZFS_VERIFY_ZP(zp); |
4005 | ||
e2e7aa2d | 4006 | if ((vm_flags & VM_WRITE) && (zp->z_pflags & |
428870ff | 4007 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { |
e2e7aa2d | 4008 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4009 | return (EPERM); |
4010 | } | |
4011 | ||
e2e7aa2d | 4012 | if ((vm_flags & (VM_READ | VM_EXEC)) && |
428870ff | 4013 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
e2e7aa2d | 4014 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4015 | return (EACCES); |
4016 | } | |
4017 | ||
34dc7c2f | 4018 | if (off < 0 || len > MAXOFFSET_T - off) { |
e2e7aa2d | 4019 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4020 | return (ENXIO); |
4021 | } | |
4022 | ||
e2e7aa2d | 4023 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4024 | return (0); |
4025 | } | |
e2e7aa2d | 4026 | EXPORT_SYMBOL(zfs_map); |
34dc7c2f | 4027 | |
3558fd73 BB |
4028 | /* |
4029 | * convoff - converts the given data (start, whence) to the | |
4030 | * given whence. | |
4031 | */ | |
4032 | int | |
4033 | convoff(struct inode *ip, flock64_t *lckdat, int whence, offset_t offset) | |
4034 | { | |
5484965a | 4035 | vattr_t vap; |
3558fd73 BB |
4036 | int error; |
4037 | ||
4038 | if ((lckdat->l_whence == 2) || (whence == 2)) { | |
5484965a | 4039 | if ((error = zfs_getattr(ip, &vap, 0, CRED()) != 0)) |
3558fd73 BB |
4040 | return (error); |
4041 | } | |
4042 | ||
4043 | switch (lckdat->l_whence) { | |
4044 | case 1: | |
4045 | lckdat->l_start += offset; | |
4046 | break; | |
4047 | case 2: | |
5484965a | 4048 | lckdat->l_start += vap.va_size; |
3558fd73 BB |
4049 | /* FALLTHRU */ |
4050 | case 0: | |
4051 | break; | |
4052 | default: | |
4053 | return (EINVAL); | |
4054 | } | |
4055 | ||
4056 | if (lckdat->l_start < 0) | |
4057 | return (EINVAL); | |
4058 | ||
4059 | switch (whence) { | |
4060 | case 1: | |
4061 | lckdat->l_start -= offset; | |
4062 | break; | |
4063 | case 2: | |
5484965a | 4064 | lckdat->l_start -= vap.va_size; |
3558fd73 BB |
4065 | /* FALLTHRU */ |
4066 | case 0: | |
4067 | break; | |
4068 | default: | |
4069 | return (EINVAL); | |
4070 | } | |
4071 | ||
4072 | lckdat->l_whence = (short)whence; | |
4073 | return (0); | |
4074 | } | |
4075 | ||
34dc7c2f BB |
4076 | /* |
4077 | * Free or allocate space in a file. Currently, this function only | |
4078 | * supports the `F_FREESP' command. However, this command is somewhat | |
4079 | * misnamed, as its functionality includes the ability to allocate as | |
4080 | * well as free space. | |
4081 | * | |
3558fd73 | 4082 | * IN: ip - inode of file to free data in. |
34dc7c2f BB |
4083 | * cmd - action to take (only F_FREESP supported). |
4084 | * bfp - section of file to free/alloc. | |
4085 | * flag - current file open mode flags. | |
4086 | * offset - current file offset. | |
4087 | * cr - credentials of caller [UNUSED]. | |
34dc7c2f BB |
4088 | * |
4089 | * RETURN: 0 if success | |
4090 | * error code if failure | |
4091 | * | |
4092 | * Timestamps: | |
3558fd73 | 4093 | * ip - ctime|mtime updated |
34dc7c2f BB |
4094 | */ |
4095 | /* ARGSUSED */ | |
e5c39b95 | 4096 | int |
3558fd73 BB |
4097 | zfs_space(struct inode *ip, int cmd, flock64_t *bfp, int flag, |
4098 | offset_t offset, cred_t *cr) | |
34dc7c2f | 4099 | { |
3558fd73 BB |
4100 | znode_t *zp = ITOZ(ip); |
4101 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4102 | uint64_t off, len; |
4103 | int error; | |
4104 | ||
3558fd73 | 4105 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4106 | ZFS_VERIFY_ZP(zp); |
4107 | ||
34dc7c2f | 4108 | if (cmd != F_FREESP) { |
3558fd73 | 4109 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4110 | return (EINVAL); |
4111 | } | |
4112 | ||
3558fd73 BB |
4113 | if ((error = convoff(ip, bfp, 0, offset))) { |
4114 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
4115 | return (error); |
4116 | } | |
4117 | ||
4118 | if (bfp->l_len < 0) { | |
3558fd73 | 4119 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4120 | return (EINVAL); |
4121 | } | |
4122 | ||
4123 | off = bfp->l_start; | |
4124 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4125 | ||
b128c09f | 4126 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f | 4127 | |
3558fd73 | 4128 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4129 | return (error); |
4130 | } | |
e5c39b95 | 4131 | EXPORT_SYMBOL(zfs_space); |
34dc7c2f BB |
4132 | |
4133 | /*ARGSUSED*/ | |
e5c39b95 | 4134 | int |
3558fd73 | 4135 | zfs_fid(struct inode *ip, fid_t *fidp) |
34dc7c2f | 4136 | { |
3558fd73 BB |
4137 | znode_t *zp = ITOZ(ip); |
4138 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f | 4139 | uint32_t gen; |
428870ff | 4140 | uint64_t gen64; |
34dc7c2f BB |
4141 | uint64_t object = zp->z_id; |
4142 | zfid_short_t *zfid; | |
428870ff | 4143 | int size, i, error; |
34dc7c2f | 4144 | |
3558fd73 | 4145 | ZFS_ENTER(zsb); |
34dc7c2f | 4146 | ZFS_VERIFY_ZP(zp); |
428870ff | 4147 | |
3558fd73 | 4148 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), |
428870ff | 4149 | &gen64, sizeof (uint64_t))) != 0) { |
3558fd73 | 4150 | ZFS_EXIT(zsb); |
428870ff BB |
4151 | return (error); |
4152 | } | |
4153 | ||
4154 | gen = (uint32_t)gen64; | |
34dc7c2f | 4155 | |
3558fd73 | 4156 | size = (zsb->z_parent != zsb) ? LONG_FID_LEN : SHORT_FID_LEN; |
34dc7c2f BB |
4157 | if (fidp->fid_len < size) { |
4158 | fidp->fid_len = size; | |
3558fd73 | 4159 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4160 | return (ENOSPC); |
4161 | } | |
4162 | ||
4163 | zfid = (zfid_short_t *)fidp; | |
4164 | ||
4165 | zfid->zf_len = size; | |
4166 | ||
4167 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4168 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4169 | ||
4170 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4171 | if (gen == 0) | |
4172 | gen = 1; | |
4173 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4174 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4175 | ||
4176 | if (size == LONG_FID_LEN) { | |
3558fd73 | 4177 | uint64_t objsetid = dmu_objset_id(zsb->z_os); |
34dc7c2f BB |
4178 | zfid_long_t *zlfid; |
4179 | ||
4180 | zlfid = (zfid_long_t *)fidp; | |
4181 | ||
4182 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
4183 | zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); | |
4184 | ||
4185 | /* XXX - this should be the generation number for the objset */ | |
4186 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
4187 | zlfid->zf_setgen[i] = 0; | |
4188 | } | |
4189 | ||
3558fd73 | 4190 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4191 | return (0); |
4192 | } | |
e5c39b95 | 4193 | EXPORT_SYMBOL(zfs_fid); |
34dc7c2f | 4194 | |
34dc7c2f | 4195 | /*ARGSUSED*/ |
e5c39b95 | 4196 | int |
3558fd73 | 4197 | zfs_getsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4198 | { |
3558fd73 BB |
4199 | znode_t *zp = ITOZ(ip); |
4200 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4201 | int error; |
4202 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
4203 | ||
3558fd73 | 4204 | ZFS_ENTER(zsb); |
34dc7c2f BB |
4205 | ZFS_VERIFY_ZP(zp); |
4206 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
3558fd73 | 4207 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4208 | |
4209 | return (error); | |
4210 | } | |
e5c39b95 | 4211 | EXPORT_SYMBOL(zfs_getsecattr); |
34dc7c2f BB |
4212 | |
4213 | /*ARGSUSED*/ | |
e5c39b95 | 4214 | int |
3558fd73 | 4215 | zfs_setsecattr(struct inode *ip, vsecattr_t *vsecp, int flag, cred_t *cr) |
34dc7c2f | 4216 | { |
3558fd73 BB |
4217 | znode_t *zp = ITOZ(ip); |
4218 | zfs_sb_t *zsb = ITOZSB(ip); | |
34dc7c2f BB |
4219 | int error; |
4220 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
3558fd73 | 4221 | zilog_t *zilog = zsb->z_log; |
34dc7c2f | 4222 | |
3558fd73 | 4223 | ZFS_ENTER(zsb); |
34dc7c2f | 4224 | ZFS_VERIFY_ZP(zp); |
428870ff | 4225 | |
34dc7c2f | 4226 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff | 4227 | |
3558fd73 | 4228 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4229 | zil_commit(zilog, 0); |
428870ff | 4230 | |
3558fd73 | 4231 | ZFS_EXIT(zsb); |
34dc7c2f BB |
4232 | return (error); |
4233 | } | |
e5c39b95 | 4234 | EXPORT_SYMBOL(zfs_setsecattr); |
34dc7c2f | 4235 | |
3558fd73 | 4236 | #ifdef HAVE_UIO_ZEROCOPY |
428870ff BB |
4237 | /* |
4238 | * Tunable, both must be a power of 2. | |
4239 | * | |
4240 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
4241 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
3558fd73 | 4242 | * an arcbuf for a partial block read |
428870ff BB |
4243 | */ |
4244 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
4245 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
4246 | ||
4247 | /*ARGSUSED*/ | |
4248 | static int | |
3558fd73 | 4249 | zfs_reqzcbuf(struct inode *ip, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr) |
428870ff | 4250 | { |
3558fd73 BB |
4251 | znode_t *zp = ITOZ(ip); |
4252 | zfs_sb_t *zsb = ITOZSB(ip); | |
4253 | int max_blksz = zsb->z_max_blksz; | |
428870ff BB |
4254 | uio_t *uio = &xuio->xu_uio; |
4255 | ssize_t size = uio->uio_resid; | |
4256 | offset_t offset = uio->uio_loffset; | |
4257 | int blksz; | |
4258 | int fullblk, i; | |
4259 | arc_buf_t *abuf; | |
4260 | ssize_t maxsize; | |
4261 | int preamble, postamble; | |
4262 | ||
4263 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
4264 | return (EINVAL); | |
4265 | ||
3558fd73 | 4266 | ZFS_ENTER(zsb); |
428870ff BB |
4267 | ZFS_VERIFY_ZP(zp); |
4268 | switch (ioflag) { | |
4269 | case UIO_WRITE: | |
4270 | /* | |
4271 | * Loan out an arc_buf for write if write size is bigger than | |
4272 | * max_blksz, and the file's block size is also max_blksz. | |
4273 | */ | |
4274 | blksz = max_blksz; | |
4275 | if (size < blksz || zp->z_blksz != blksz) { | |
3558fd73 | 4276 | ZFS_EXIT(zsb); |
428870ff BB |
4277 | return (EINVAL); |
4278 | } | |
4279 | /* | |
4280 | * Caller requests buffers for write before knowing where the | |
4281 | * write offset might be (e.g. NFS TCP write). | |
4282 | */ | |
4283 | if (offset == -1) { | |
4284 | preamble = 0; | |
4285 | } else { | |
4286 | preamble = P2PHASE(offset, blksz); | |
4287 | if (preamble) { | |
4288 | preamble = blksz - preamble; | |
4289 | size -= preamble; | |
4290 | } | |
4291 | } | |
4292 | ||
4293 | postamble = P2PHASE(size, blksz); | |
4294 | size -= postamble; | |
4295 | ||
4296 | fullblk = size / blksz; | |
4297 | (void) dmu_xuio_init(xuio, | |
4298 | (preamble != 0) + fullblk + (postamble != 0)); | |
428870ff BB |
4299 | |
4300 | /* | |
4301 | * Have to fix iov base/len for partial buffers. They | |
4302 | * currently represent full arc_buf's. | |
4303 | */ | |
4304 | if (preamble) { | |
4305 | /* data begins in the middle of the arc_buf */ | |
4306 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4307 | blksz); | |
4308 | ASSERT(abuf); | |
4309 | (void) dmu_xuio_add(xuio, abuf, | |
4310 | blksz - preamble, preamble); | |
4311 | } | |
4312 | ||
4313 | for (i = 0; i < fullblk; i++) { | |
4314 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4315 | blksz); | |
4316 | ASSERT(abuf); | |
4317 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
4318 | } | |
4319 | ||
4320 | if (postamble) { | |
4321 | /* data ends in the middle of the arc_buf */ | |
4322 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4323 | blksz); | |
4324 | ASSERT(abuf); | |
4325 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
4326 | } | |
4327 | break; | |
4328 | case UIO_READ: | |
4329 | /* | |
4330 | * Loan out an arc_buf for read if the read size is larger than | |
4331 | * the current file block size. Block alignment is not | |
4332 | * considered. Partial arc_buf will be loaned out for read. | |
4333 | */ | |
4334 | blksz = zp->z_blksz; | |
4335 | if (blksz < zcr_blksz_min) | |
4336 | blksz = zcr_blksz_min; | |
4337 | if (blksz > zcr_blksz_max) | |
4338 | blksz = zcr_blksz_max; | |
4339 | /* avoid potential complexity of dealing with it */ | |
4340 | if (blksz > max_blksz) { | |
3558fd73 | 4341 | ZFS_EXIT(zsb); |
428870ff BB |
4342 | return (EINVAL); |
4343 | } | |
4344 | ||
4345 | maxsize = zp->z_size - uio->uio_loffset; | |
4346 | if (size > maxsize) | |
4347 | size = maxsize; | |
4348 | ||
3558fd73 BB |
4349 | if (size < blksz) { |
4350 | ZFS_EXIT(zsb); | |
428870ff BB |
4351 | return (EINVAL); |
4352 | } | |
4353 | break; | |
4354 | default: | |
3558fd73 | 4355 | ZFS_EXIT(zsb); |
428870ff BB |
4356 | return (EINVAL); |
4357 | } | |
4358 | ||
4359 | uio->uio_extflg = UIO_XUIO; | |
4360 | XUIO_XUZC_RW(xuio) = ioflag; | |
3558fd73 | 4361 | ZFS_EXIT(zsb); |
428870ff BB |
4362 | return (0); |
4363 | } | |
4364 | ||
4365 | /*ARGSUSED*/ | |
4366 | static int | |
3558fd73 | 4367 | zfs_retzcbuf(struct inode *ip, xuio_t *xuio, cred_t *cr) |
428870ff BB |
4368 | { |
4369 | int i; | |
4370 | arc_buf_t *abuf; | |
4371 | int ioflag = XUIO_XUZC_RW(xuio); | |
4372 | ||
4373 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
4374 | ||
4375 | i = dmu_xuio_cnt(xuio); | |
4376 | while (i-- > 0) { | |
4377 | abuf = dmu_xuio_arcbuf(xuio, i); | |
4378 | /* | |
4379 | * if abuf == NULL, it must be a write buffer | |
4380 | * that has been returned in zfs_write(). | |
4381 | */ | |
4382 | if (abuf) | |
4383 | dmu_return_arcbuf(abuf); | |
4384 | ASSERT(abuf || ioflag == UIO_WRITE); | |
4385 | } | |
4386 | ||
4387 | dmu_xuio_fini(xuio); | |
4388 | return (0); | |
4389 | } | |
3558fd73 | 4390 | #endif /* HAVE_UIO_ZEROCOPY */ |
c409e464 BB |
4391 | |
4392 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
4393 | module_param(zfs_read_chunk_size, long, 0644); | |
4394 | MODULE_PARM_DESC(zfs_read_chunk_size, "Bytes to read per chunk"); | |
4395 | #endif |