]>
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 | |
34dc7c2f BB |
28 | #include <sys/types.h> |
29 | #include <sys/param.h> | |
30 | #include <sys/time.h> | |
31 | #include <sys/systm.h> | |
32 | #include <sys/sysmacros.h> | |
33 | #include <sys/resource.h> | |
34 | #include <sys/vfs.h> | |
35 | #include <sys/vfs_opreg.h> | |
36 | #include <sys/vnode.h> | |
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> | |
44 | #include <sys/vm.h> | |
45 | #include <vm/seg_vn.h> | |
46 | #include <vm/pvn.h> | |
47 | #include <vm/as.h> | |
b128c09f BB |
48 | #include <vm/kpm.h> |
49 | #include <vm/seg_kpm.h> | |
34dc7c2f BB |
50 | #include <sys/mman.h> |
51 | #include <sys/pathname.h> | |
52 | #include <sys/cmn_err.h> | |
53 | #include <sys/errno.h> | |
54 | #include <sys/unistd.h> | |
55 | #include <sys/zfs_dir.h> | |
56 | #include <sys/zfs_acl.h> | |
57 | #include <sys/zfs_ioctl.h> | |
58 | #include <sys/fs/zfs.h> | |
59 | #include <sys/dmu.h> | |
428870ff | 60 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
61 | #include <sys/spa.h> |
62 | #include <sys/txg.h> | |
63 | #include <sys/dbuf.h> | |
64 | #include <sys/zap.h> | |
428870ff | 65 | #include <sys/sa.h> |
34dc7c2f BB |
66 | #include <sys/dirent.h> |
67 | #include <sys/policy.h> | |
68 | #include <sys/sunddi.h> | |
69 | #include <sys/filio.h> | |
b128c09f | 70 | #include <sys/sid.h> |
34dc7c2f BB |
71 | #include "fs/fs_subr.h" |
72 | #include <sys/zfs_ctldir.h> | |
73 | #include <sys/zfs_fuid.h> | |
428870ff | 74 | #include <sys/zfs_sa.h> |
34dc7c2f BB |
75 | #include <sys/dnlc.h> |
76 | #include <sys/zfs_rlock.h> | |
77 | #include <sys/extdirent.h> | |
78 | #include <sys/kidmap.h> | |
428870ff | 79 | #include <sys/cred.h> |
34dc7c2f BB |
80 | #include <sys/attr.h> |
81 | ||
82 | /* | |
83 | * Programming rules. | |
84 | * | |
85 | * Each vnode op performs some logical unit of work. To do this, the ZPL must | |
86 | * properly lock its in-core state, create a DMU transaction, do the work, | |
87 | * record this work in the intent log (ZIL), commit the DMU transaction, | |
88 | * and wait for the intent log to commit if it is a synchronous operation. | |
89 | * Moreover, the vnode ops must work in both normal and log replay context. | |
90 | * The ordering of events is important to avoid deadlocks and references | |
91 | * to freed memory. The example below illustrates the following Big Rules: | |
92 | * | |
93 | * (1) A check must be made in each zfs thread for a mounted file system. | |
94 | * This is done avoiding races using ZFS_ENTER(zfsvfs). | |
95 | * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes | |
96 | * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros | |
97 | * can return EIO from the calling function. | |
98 | * | |
99 | * (2) VN_RELE() should always be the last thing except for zil_commit() | |
100 | * (if necessary) and ZFS_EXIT(). This is for 3 reasons: | |
101 | * First, if it's the last reference, the vnode/znode | |
102 | * can be freed, so the zp may point to freed memory. Second, the last | |
103 | * reference will call zfs_zinactive(), which may induce a lot of work -- | |
104 | * pushing cached pages (which acquires range locks) and syncing out | |
105 | * cached atime changes. Third, zfs_zinactive() may require a new tx, | |
106 | * which could deadlock the system if you were already holding one. | |
9babb374 | 107 | * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC(). |
34dc7c2f BB |
108 | * |
109 | * (3) All range locks must be grabbed before calling dmu_tx_assign(), | |
110 | * as they can span dmu_tx_assign() calls. | |
111 | * | |
fb5f0bc8 | 112 | * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign(). |
34dc7c2f BB |
113 | * This is critical because we don't want to block while holding locks. |
114 | * Note, in particular, that if a lock is sometimes acquired before | |
115 | * the tx assigns, and sometimes after (e.g. z_lock), then failing to | |
116 | * use a non-blocking assign can deadlock the system. The scenario: | |
117 | * | |
118 | * Thread A has grabbed a lock before calling dmu_tx_assign(). | |
119 | * Thread B is in an already-assigned tx, and blocks for this lock. | |
120 | * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() | |
121 | * forever, because the previous txg can't quiesce until B's tx commits. | |
122 | * | |
123 | * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, | |
124 | * then drop all locks, call dmu_tx_wait(), and try again. | |
125 | * | |
126 | * (5) If the operation succeeded, generate the intent log entry for it | |
127 | * before dropping locks. This ensures that the ordering of events | |
128 | * in the intent log matches the order in which they actually occurred. | |
fb5f0bc8 BB |
129 | * During ZIL replay the zfs_log_* functions will update the sequence |
130 | * number to indicate the zil transaction has replayed. | |
34dc7c2f BB |
131 | * |
132 | * (6) At the end of each vnode op, the DMU tx must always commit, | |
133 | * regardless of whether there were any errors. | |
134 | * | |
572e2857 | 135 | * (7) After dropping all locks, invoke zil_commit(zilog, foid) |
34dc7c2f BB |
136 | * to ensure that synchronous semantics are provided when necessary. |
137 | * | |
138 | * In general, this is how things should be ordered in each vnode op: | |
139 | * | |
140 | * ZFS_ENTER(zfsvfs); // exit if unmounted | |
141 | * top: | |
142 | * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD()) | |
143 | * rw_enter(...); // grab any other locks you need | |
144 | * tx = dmu_tx_create(...); // get DMU tx | |
145 | * dmu_tx_hold_*(); // hold each object you might modify | |
fb5f0bc8 | 146 | * error = dmu_tx_assign(tx, TXG_NOWAIT); // try to assign |
34dc7c2f BB |
147 | * if (error) { |
148 | * rw_exit(...); // drop locks | |
149 | * zfs_dirent_unlock(dl); // unlock directory entry | |
150 | * VN_RELE(...); // release held vnodes | |
fb5f0bc8 | 151 | * if (error == ERESTART) { |
34dc7c2f BB |
152 | * dmu_tx_wait(tx); |
153 | * dmu_tx_abort(tx); | |
154 | * goto top; | |
155 | * } | |
156 | * dmu_tx_abort(tx); // abort DMU tx | |
157 | * ZFS_EXIT(zfsvfs); // finished in zfs | |
158 | * return (error); // really out of space | |
159 | * } | |
160 | * error = do_real_work(); // do whatever this VOP does | |
161 | * if (error == 0) | |
162 | * zfs_log_*(...); // on success, make ZIL entry | |
163 | * dmu_tx_commit(tx); // commit DMU tx -- error or not | |
164 | * rw_exit(...); // drop locks | |
165 | * zfs_dirent_unlock(dl); // unlock directory entry | |
166 | * VN_RELE(...); // release held vnodes | |
572e2857 | 167 | * zil_commit(zilog, foid); // synchronous when necessary |
34dc7c2f BB |
168 | * ZFS_EXIT(zfsvfs); // finished in zfs |
169 | * return (error); // done, report error | |
170 | */ | |
171 | ||
172 | /* ARGSUSED */ | |
173 | static int | |
174 | zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct) | |
175 | { | |
176 | znode_t *zp = VTOZ(*vpp); | |
b128c09f BB |
177 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; |
178 | ||
179 | ZFS_ENTER(zfsvfs); | |
180 | ZFS_VERIFY_ZP(zp); | |
34dc7c2f | 181 | |
428870ff | 182 | if ((flag & FWRITE) && (zp->z_pflags & ZFS_APPENDONLY) && |
34dc7c2f | 183 | ((flag & FAPPEND) == 0)) { |
b128c09f | 184 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
185 | return (EPERM); |
186 | } | |
187 | ||
188 | if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && | |
189 | ZTOV(zp)->v_type == VREG && | |
428870ff | 190 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { |
b128c09f BB |
191 | if (fs_vscan(*vpp, cr, 0) != 0) { |
192 | ZFS_EXIT(zfsvfs); | |
34dc7c2f | 193 | return (EACCES); |
b128c09f BB |
194 | } |
195 | } | |
34dc7c2f BB |
196 | |
197 | /* Keep a count of the synchronous opens in the znode */ | |
198 | if (flag & (FSYNC | FDSYNC)) | |
199 | atomic_inc_32(&zp->z_sync_cnt); | |
200 | ||
b128c09f | 201 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
202 | return (0); |
203 | } | |
204 | ||
205 | /* ARGSUSED */ | |
206 | static int | |
207 | zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr, | |
208 | caller_context_t *ct) | |
209 | { | |
210 | znode_t *zp = VTOZ(vp); | |
b128c09f BB |
211 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; |
212 | ||
45d1cae3 BB |
213 | /* |
214 | * Clean up any locks held by this process on the vp. | |
215 | */ | |
216 | cleanlocks(vp, ddi_get_pid(), 0); | |
217 | cleanshares(vp, ddi_get_pid()); | |
218 | ||
b128c09f BB |
219 | ZFS_ENTER(zfsvfs); |
220 | ZFS_VERIFY_ZP(zp); | |
34dc7c2f BB |
221 | |
222 | /* Decrement the synchronous opens in the znode */ | |
223 | if ((flag & (FSYNC | FDSYNC)) && (count == 1)) | |
224 | atomic_dec_32(&zp->z_sync_cnt); | |
225 | ||
34dc7c2f BB |
226 | if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && |
227 | ZTOV(zp)->v_type == VREG && | |
428870ff | 228 | !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) |
34dc7c2f BB |
229 | VERIFY(fs_vscan(vp, cr, 1) == 0); |
230 | ||
b128c09f | 231 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
232 | return (0); |
233 | } | |
234 | ||
235 | /* | |
236 | * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and | |
237 | * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter. | |
238 | */ | |
239 | static int | |
240 | zfs_holey(vnode_t *vp, int cmd, offset_t *off) | |
241 | { | |
242 | znode_t *zp = VTOZ(vp); | |
243 | uint64_t noff = (uint64_t)*off; /* new offset */ | |
244 | uint64_t file_sz; | |
245 | int error; | |
246 | boolean_t hole; | |
247 | ||
428870ff | 248 | file_sz = zp->z_size; |
34dc7c2f BB |
249 | if (noff >= file_sz) { |
250 | return (ENXIO); | |
251 | } | |
252 | ||
253 | if (cmd == _FIO_SEEK_HOLE) | |
254 | hole = B_TRUE; | |
255 | else | |
256 | hole = B_FALSE; | |
257 | ||
258 | error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff); | |
259 | ||
260 | /* end of file? */ | |
261 | if ((error == ESRCH) || (noff > file_sz)) { | |
262 | /* | |
263 | * Handle the virtual hole at the end of file. | |
264 | */ | |
265 | if (hole) { | |
266 | *off = file_sz; | |
267 | return (0); | |
268 | } | |
269 | return (ENXIO); | |
270 | } | |
271 | ||
272 | if (noff < *off) | |
273 | return (error); | |
274 | *off = noff; | |
275 | return (error); | |
276 | } | |
277 | ||
278 | /* ARGSUSED */ | |
279 | static int | |
280 | zfs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred, | |
281 | int *rvalp, caller_context_t *ct) | |
282 | { | |
283 | offset_t off; | |
284 | int error; | |
285 | zfsvfs_t *zfsvfs; | |
286 | znode_t *zp; | |
287 | ||
288 | switch (com) { | |
289 | case _FIOFFS: | |
290 | return (zfs_sync(vp->v_vfsp, 0, cred)); | |
291 | ||
292 | /* | |
293 | * The following two ioctls are used by bfu. Faking out, | |
294 | * necessary to avoid bfu errors. | |
295 | */ | |
296 | case _FIOGDIO: | |
297 | case _FIOSDIO: | |
298 | return (0); | |
299 | ||
300 | case _FIO_SEEK_DATA: | |
301 | case _FIO_SEEK_HOLE: | |
302 | if (ddi_copyin((void *)data, &off, sizeof (off), flag)) | |
303 | return (EFAULT); | |
304 | ||
305 | zp = VTOZ(vp); | |
306 | zfsvfs = zp->z_zfsvfs; | |
307 | ZFS_ENTER(zfsvfs); | |
308 | ZFS_VERIFY_ZP(zp); | |
309 | ||
310 | /* offset parameter is in/out */ | |
311 | error = zfs_holey(vp, com, &off); | |
312 | ZFS_EXIT(zfsvfs); | |
313 | if (error) | |
314 | return (error); | |
315 | if (ddi_copyout(&off, (void *)data, sizeof (off), flag)) | |
316 | return (EFAULT); | |
317 | return (0); | |
318 | } | |
319 | return (ENOTTY); | |
320 | } | |
321 | ||
b128c09f BB |
322 | /* |
323 | * Utility functions to map and unmap a single physical page. These | |
324 | * are used to manage the mappable copies of ZFS file data, and therefore | |
325 | * do not update ref/mod bits. | |
326 | */ | |
327 | caddr_t | |
328 | zfs_map_page(page_t *pp, enum seg_rw rw) | |
329 | { | |
330 | if (kpm_enable) | |
331 | return (hat_kpm_mapin(pp, 0)); | |
332 | ASSERT(rw == S_READ || rw == S_WRITE); | |
333 | return (ppmapin(pp, PROT_READ | ((rw == S_WRITE) ? PROT_WRITE : 0), | |
334 | (caddr_t)-1)); | |
335 | } | |
336 | ||
337 | void | |
338 | zfs_unmap_page(page_t *pp, caddr_t addr) | |
339 | { | |
340 | if (kpm_enable) { | |
341 | hat_kpm_mapout(pp, 0, addr); | |
342 | } else { | |
343 | ppmapout(addr); | |
344 | } | |
345 | } | |
346 | ||
34dc7c2f BB |
347 | /* |
348 | * When a file is memory mapped, we must keep the IO data synchronized | |
349 | * between the DMU cache and the memory mapped pages. What this means: | |
350 | * | |
351 | * On Write: If we find a memory mapped page, we write to *both* | |
352 | * the page and the dmu buffer. | |
34dc7c2f | 353 | */ |
d164b209 BB |
354 | static void |
355 | update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid) | |
34dc7c2f | 356 | { |
d164b209 | 357 | int64_t off; |
34dc7c2f | 358 | |
34dc7c2f BB |
359 | off = start & PAGEOFFSET; |
360 | for (start &= PAGEMASK; len > 0; start += PAGESIZE) { | |
361 | page_t *pp; | |
d164b209 | 362 | uint64_t nbytes = MIN(PAGESIZE - off, len); |
34dc7c2f | 363 | |
34dc7c2f BB |
364 | if (pp = page_lookup(vp, start, SE_SHARED)) { |
365 | caddr_t va; | |
366 | ||
b128c09f | 367 | va = zfs_map_page(pp, S_WRITE); |
9babb374 BB |
368 | (void) dmu_read(os, oid, start+off, nbytes, va+off, |
369 | DMU_READ_PREFETCH); | |
b128c09f | 370 | zfs_unmap_page(pp, va); |
34dc7c2f | 371 | page_unlock(pp); |
34dc7c2f | 372 | } |
d164b209 | 373 | len -= nbytes; |
34dc7c2f | 374 | off = 0; |
34dc7c2f | 375 | } |
34dc7c2f BB |
376 | } |
377 | ||
378 | /* | |
379 | * When a file is memory mapped, we must keep the IO data synchronized | |
380 | * between the DMU cache and the memory mapped pages. What this means: | |
381 | * | |
382 | * On Read: We "read" preferentially from memory mapped pages, | |
383 | * else we default from the dmu buffer. | |
384 | * | |
385 | * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when | |
386 | * the file is memory mapped. | |
387 | */ | |
388 | static int | |
389 | mappedread(vnode_t *vp, int nbytes, uio_t *uio) | |
390 | { | |
391 | znode_t *zp = VTOZ(vp); | |
392 | objset_t *os = zp->z_zfsvfs->z_os; | |
393 | int64_t start, off; | |
394 | int len = nbytes; | |
395 | int error = 0; | |
396 | ||
397 | start = uio->uio_loffset; | |
398 | off = start & PAGEOFFSET; | |
399 | for (start &= PAGEMASK; len > 0; start += PAGESIZE) { | |
400 | page_t *pp; | |
401 | uint64_t bytes = MIN(PAGESIZE - off, len); | |
402 | ||
403 | if (pp = page_lookup(vp, start, SE_SHARED)) { | |
404 | caddr_t va; | |
405 | ||
b128c09f | 406 | va = zfs_map_page(pp, S_READ); |
34dc7c2f | 407 | error = uiomove(va + off, bytes, UIO_READ, uio); |
b128c09f | 408 | zfs_unmap_page(pp, va); |
34dc7c2f BB |
409 | page_unlock(pp); |
410 | } else { | |
411 | error = dmu_read_uio(os, zp->z_id, uio, bytes); | |
412 | } | |
413 | len -= bytes; | |
414 | off = 0; | |
415 | if (error) | |
416 | break; | |
417 | } | |
418 | return (error); | |
419 | } | |
420 | ||
421 | offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */ | |
422 | ||
423 | /* | |
424 | * Read bytes from specified file into supplied buffer. | |
425 | * | |
426 | * IN: vp - vnode of file to be read from. | |
427 | * uio - structure supplying read location, range info, | |
428 | * and return buffer. | |
429 | * ioflag - SYNC flags; used to provide FRSYNC semantics. | |
430 | * cr - credentials of caller. | |
431 | * ct - caller context | |
432 | * | |
433 | * OUT: uio - updated offset and range, buffer filled. | |
434 | * | |
435 | * RETURN: 0 if success | |
436 | * error code if failure | |
437 | * | |
438 | * Side Effects: | |
439 | * vp - atime updated if byte count > 0 | |
440 | */ | |
441 | /* ARGSUSED */ | |
442 | static int | |
443 | zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) | |
444 | { | |
445 | znode_t *zp = VTOZ(vp); | |
446 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
447 | objset_t *os; | |
448 | ssize_t n, nbytes; | |
449 | int error; | |
450 | rl_t *rl; | |
428870ff | 451 | xuio_t *xuio = NULL; |
34dc7c2f BB |
452 | |
453 | ZFS_ENTER(zfsvfs); | |
454 | ZFS_VERIFY_ZP(zp); | |
455 | os = zfsvfs->z_os; | |
456 | ||
428870ff | 457 | if (zp->z_pflags & ZFS_AV_QUARANTINED) { |
34dc7c2f BB |
458 | ZFS_EXIT(zfsvfs); |
459 | return (EACCES); | |
460 | } | |
461 | ||
462 | /* | |
463 | * Validate file offset | |
464 | */ | |
465 | if (uio->uio_loffset < (offset_t)0) { | |
466 | ZFS_EXIT(zfsvfs); | |
467 | return (EINVAL); | |
468 | } | |
469 | ||
470 | /* | |
471 | * Fasttrack empty reads | |
472 | */ | |
473 | if (uio->uio_resid == 0) { | |
474 | ZFS_EXIT(zfsvfs); | |
475 | return (0); | |
476 | } | |
477 | ||
478 | /* | |
479 | * Check for mandatory locks | |
480 | */ | |
428870ff | 481 | if (MANDMODE(zp->z_mode)) { |
34dc7c2f BB |
482 | if (error = chklock(vp, FREAD, |
483 | uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) { | |
484 | ZFS_EXIT(zfsvfs); | |
485 | return (error); | |
486 | } | |
487 | } | |
488 | ||
489 | /* | |
490 | * If we're in FRSYNC mode, sync out this znode before reading it. | |
491 | */ | |
428870ff | 492 | if (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 493 | zil_commit(zfsvfs->z_log, zp->z_id); |
34dc7c2f BB |
494 | |
495 | /* | |
496 | * Lock the range against changes. | |
497 | */ | |
498 | rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER); | |
499 | ||
500 | /* | |
501 | * If we are reading past end-of-file we can skip | |
502 | * to the end; but we might still need to set atime. | |
503 | */ | |
428870ff | 504 | if (uio->uio_loffset >= zp->z_size) { |
34dc7c2f BB |
505 | error = 0; |
506 | goto out; | |
507 | } | |
508 | ||
428870ff BB |
509 | ASSERT(uio->uio_loffset < zp->z_size); |
510 | n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); | |
511 | ||
512 | if ((uio->uio_extflg == UIO_XUIO) && | |
513 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { | |
514 | int nblk; | |
515 | int blksz = zp->z_blksz; | |
516 | uint64_t offset = uio->uio_loffset; | |
517 | ||
518 | xuio = (xuio_t *)uio; | |
519 | if ((ISP2(blksz))) { | |
520 | nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, | |
521 | blksz)) / blksz; | |
522 | } else { | |
523 | ASSERT(offset + n <= blksz); | |
524 | nblk = 1; | |
525 | } | |
526 | (void) dmu_xuio_init(xuio, nblk); | |
527 | ||
528 | if (vn_has_cached_data(vp)) { | |
529 | /* | |
530 | * For simplicity, we always allocate a full buffer | |
531 | * even if we only expect to read a portion of a block. | |
532 | */ | |
533 | while (--nblk >= 0) { | |
534 | (void) dmu_xuio_add(xuio, | |
535 | dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
536 | blksz), 0, blksz); | |
537 | } | |
538 | } | |
539 | } | |
34dc7c2f BB |
540 | |
541 | while (n > 0) { | |
542 | nbytes = MIN(n, zfs_read_chunk_size - | |
543 | P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); | |
544 | ||
545 | if (vn_has_cached_data(vp)) | |
546 | error = mappedread(vp, nbytes, uio); | |
547 | else | |
548 | error = dmu_read_uio(os, zp->z_id, uio, nbytes); | |
b128c09f BB |
549 | if (error) { |
550 | /* convert checksum errors into IO errors */ | |
551 | if (error == ECKSUM) | |
552 | error = EIO; | |
34dc7c2f | 553 | break; |
b128c09f | 554 | } |
34dc7c2f BB |
555 | |
556 | n -= nbytes; | |
557 | } | |
34dc7c2f BB |
558 | out: |
559 | zfs_range_unlock(rl); | |
560 | ||
561 | ZFS_ACCESSTIME_STAMP(zfsvfs, zp); | |
562 | ZFS_EXIT(zfsvfs); | |
563 | return (error); | |
564 | } | |
565 | ||
34dc7c2f BB |
566 | /* |
567 | * Write the bytes to a file. | |
568 | * | |
569 | * IN: vp - vnode of file to be written to. | |
570 | * uio - structure supplying write location, range info, | |
571 | * and data buffer. | |
572 | * ioflag - FAPPEND flag set if in append mode. | |
573 | * cr - credentials of caller. | |
574 | * ct - caller context (NFS/CIFS fem monitor only) | |
575 | * | |
576 | * OUT: uio - updated offset and range. | |
577 | * | |
578 | * RETURN: 0 if success | |
579 | * error code if failure | |
580 | * | |
581 | * Timestamps: | |
582 | * vp - ctime|mtime updated if byte count > 0 | |
583 | */ | |
428870ff | 584 | |
34dc7c2f BB |
585 | /* ARGSUSED */ |
586 | static int | |
587 | zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) | |
588 | { | |
589 | znode_t *zp = VTOZ(vp); | |
590 | rlim64_t limit = uio->uio_llimit; | |
591 | ssize_t start_resid = uio->uio_resid; | |
592 | ssize_t tx_bytes; | |
593 | uint64_t end_size; | |
594 | dmu_tx_t *tx; | |
595 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
596 | zilog_t *zilog; | |
597 | offset_t woff; | |
598 | ssize_t n, nbytes; | |
599 | rl_t *rl; | |
600 | int max_blksz = zfsvfs->z_max_blksz; | |
34dc7c2f | 601 | int error; |
9babb374 | 602 | arc_buf_t *abuf; |
428870ff BB |
603 | iovec_t *aiov; |
604 | xuio_t *xuio = NULL; | |
605 | int i_iov = 0; | |
606 | int iovcnt = uio->uio_iovcnt; | |
607 | iovec_t *iovp = uio->uio_iov; | |
608 | int write_eof; | |
609 | int count = 0; | |
610 | sa_bulk_attr_t bulk[4]; | |
611 | uint64_t mtime[2], ctime[2]; | |
34dc7c2f | 612 | |
34dc7c2f BB |
613 | /* |
614 | * Fasttrack empty write | |
615 | */ | |
616 | n = start_resid; | |
617 | if (n == 0) | |
618 | return (0); | |
619 | ||
620 | if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) | |
621 | limit = MAXOFFSET_T; | |
622 | ||
623 | ZFS_ENTER(zfsvfs); | |
624 | ZFS_VERIFY_ZP(zp); | |
b128c09f | 625 | |
428870ff BB |
626 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
627 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
628 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
629 | &zp->z_size, 8); | |
630 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
631 | &zp->z_pflags, 8); | |
632 | ||
b128c09f BB |
633 | /* |
634 | * If immutable or not appending then return EPERM | |
635 | */ | |
428870ff BB |
636 | if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || |
637 | ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && | |
638 | (uio->uio_loffset < zp->z_size))) { | |
b128c09f BB |
639 | ZFS_EXIT(zfsvfs); |
640 | return (EPERM); | |
641 | } | |
642 | ||
34dc7c2f BB |
643 | zilog = zfsvfs->z_log; |
644 | ||
428870ff BB |
645 | /* |
646 | * Validate file offset | |
647 | */ | |
648 | woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; | |
649 | if (woff < 0) { | |
650 | ZFS_EXIT(zfsvfs); | |
651 | return (EINVAL); | |
652 | } | |
653 | ||
654 | /* | |
655 | * Check for mandatory locks before calling zfs_range_lock() | |
656 | * in order to prevent a deadlock with locks set via fcntl(). | |
657 | */ | |
658 | if (MANDMODE((mode_t)zp->z_mode) && | |
659 | (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) { | |
660 | ZFS_EXIT(zfsvfs); | |
661 | return (error); | |
662 | } | |
663 | ||
34dc7c2f BB |
664 | /* |
665 | * Pre-fault the pages to ensure slow (eg NFS) pages | |
666 | * don't hold up txg. | |
428870ff | 667 | * Skip this if uio contains loaned arc_buf. |
34dc7c2f | 668 | */ |
428870ff BB |
669 | if ((uio->uio_extflg == UIO_XUIO) && |
670 | (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) | |
671 | xuio = (xuio_t *)uio; | |
672 | else | |
572e2857 | 673 | uio_prefaultpages(MIN(n, max_blksz), uio); |
34dc7c2f BB |
674 | |
675 | /* | |
676 | * If in append mode, set the io offset pointer to eof. | |
677 | */ | |
678 | if (ioflag & FAPPEND) { | |
679 | /* | |
428870ff BB |
680 | * Obtain an appending range lock to guarantee file append |
681 | * semantics. We reset the write offset once we have the lock. | |
34dc7c2f BB |
682 | */ |
683 | rl = zfs_range_lock(zp, 0, n, RL_APPEND); | |
428870ff | 684 | woff = rl->r_off; |
34dc7c2f | 685 | if (rl->r_len == UINT64_MAX) { |
428870ff BB |
686 | /* |
687 | * We overlocked the file because this write will cause | |
688 | * the file block size to increase. | |
689 | * Note that zp_size cannot change with this lock held. | |
690 | */ | |
691 | woff = zp->z_size; | |
34dc7c2f | 692 | } |
428870ff | 693 | uio->uio_loffset = woff; |
34dc7c2f | 694 | } else { |
34dc7c2f | 695 | /* |
428870ff BB |
696 | * Note that if the file block size will change as a result of |
697 | * this write, then this range lock will lock the entire file | |
698 | * so that we can re-write the block safely. | |
34dc7c2f BB |
699 | */ |
700 | rl = zfs_range_lock(zp, woff, n, RL_WRITER); | |
701 | } | |
702 | ||
703 | if (woff >= limit) { | |
704 | zfs_range_unlock(rl); | |
705 | ZFS_EXIT(zfsvfs); | |
706 | return (EFBIG); | |
707 | } | |
708 | ||
709 | if ((woff + n) > limit || woff > (limit - n)) | |
710 | n = limit - woff; | |
711 | ||
428870ff BB |
712 | /* Will this write extend the file length? */ |
713 | write_eof = (woff + n > zp->z_size); | |
714 | ||
715 | end_size = MAX(zp->z_size, woff + n); | |
34dc7c2f BB |
716 | |
717 | /* | |
718 | * Write the file in reasonable size chunks. Each chunk is written | |
719 | * in a separate transaction; this keeps the intent log records small | |
720 | * and allows us to do more fine-grained space accounting. | |
721 | */ | |
722 | while (n > 0) { | |
9babb374 BB |
723 | abuf = NULL; |
724 | woff = uio->uio_loffset; | |
9babb374 | 725 | again: |
428870ff BB |
726 | if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || |
727 | zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { | |
9babb374 BB |
728 | if (abuf != NULL) |
729 | dmu_return_arcbuf(abuf); | |
730 | error = EDQUOT; | |
731 | break; | |
732 | } | |
733 | ||
428870ff BB |
734 | if (xuio && abuf == NULL) { |
735 | ASSERT(i_iov < iovcnt); | |
736 | aiov = &iovp[i_iov]; | |
737 | abuf = dmu_xuio_arcbuf(xuio, i_iov); | |
738 | dmu_xuio_clear(xuio, i_iov); | |
739 | DTRACE_PROBE3(zfs_cp_write, int, i_iov, | |
740 | iovec_t *, aiov, arc_buf_t *, abuf); | |
741 | ASSERT((aiov->iov_base == abuf->b_data) || | |
742 | ((char *)aiov->iov_base - (char *)abuf->b_data + | |
743 | aiov->iov_len == arc_buf_size(abuf))); | |
744 | i_iov++; | |
745 | } else if (abuf == NULL && n >= max_blksz && | |
746 | woff >= zp->z_size && | |
9babb374 BB |
747 | P2PHASE(woff, max_blksz) == 0 && |
748 | zp->z_blksz == max_blksz) { | |
428870ff BB |
749 | /* |
750 | * This write covers a full block. "Borrow" a buffer | |
751 | * from the dmu so that we can fill it before we enter | |
752 | * a transaction. This avoids the possibility of | |
753 | * holding up the transaction if the data copy hangs | |
754 | * up on a pagefault (e.g., from an NFS server mapping). | |
755 | */ | |
9babb374 BB |
756 | size_t cbytes; |
757 | ||
428870ff BB |
758 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), |
759 | max_blksz); | |
9babb374 BB |
760 | ASSERT(abuf != NULL); |
761 | ASSERT(arc_buf_size(abuf) == max_blksz); | |
762 | if (error = uiocopy(abuf->b_data, max_blksz, | |
763 | UIO_WRITE, uio, &cbytes)) { | |
764 | dmu_return_arcbuf(abuf); | |
765 | break; | |
766 | } | |
767 | ASSERT(cbytes == max_blksz); | |
768 | } | |
769 | ||
34dc7c2f BB |
770 | /* |
771 | * Start a transaction. | |
772 | */ | |
34dc7c2f | 773 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff | 774 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 775 | dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); |
428870ff | 776 | zfs_sa_upgrade_txholds(tx, zp); |
fb5f0bc8 | 777 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 778 | if (error) { |
fb5f0bc8 | 779 | if (error == ERESTART) { |
34dc7c2f BB |
780 | dmu_tx_wait(tx); |
781 | dmu_tx_abort(tx); | |
9babb374 | 782 | goto again; |
34dc7c2f BB |
783 | } |
784 | dmu_tx_abort(tx); | |
9babb374 BB |
785 | if (abuf != NULL) |
786 | dmu_return_arcbuf(abuf); | |
34dc7c2f BB |
787 | break; |
788 | } | |
789 | ||
790 | /* | |
791 | * If zfs_range_lock() over-locked we grow the blocksize | |
792 | * and then reduce the lock range. This will only happen | |
793 | * on the first iteration since zfs_range_reduce() will | |
794 | * shrink down r_len to the appropriate size. | |
795 | */ | |
796 | if (rl->r_len == UINT64_MAX) { | |
797 | uint64_t new_blksz; | |
798 | ||
799 | if (zp->z_blksz > max_blksz) { | |
800 | ASSERT(!ISP2(zp->z_blksz)); | |
801 | new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE); | |
802 | } else { | |
803 | new_blksz = MIN(end_size, max_blksz); | |
804 | } | |
805 | zfs_grow_blocksize(zp, new_blksz, tx); | |
806 | zfs_range_reduce(rl, woff, n); | |
807 | } | |
808 | ||
809 | /* | |
810 | * XXX - should we really limit each write to z_max_blksz? | |
811 | * Perhaps we should use SPA_MAXBLOCKSIZE chunks? | |
812 | */ | |
813 | nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); | |
34dc7c2f | 814 | |
9babb374 BB |
815 | if (abuf == NULL) { |
816 | tx_bytes = uio->uio_resid; | |
428870ff BB |
817 | error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), |
818 | uio, nbytes, tx); | |
9babb374 BB |
819 | tx_bytes -= uio->uio_resid; |
820 | } else { | |
821 | tx_bytes = nbytes; | |
428870ff BB |
822 | ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); |
823 | /* | |
824 | * If this is not a full block write, but we are | |
825 | * extending the file past EOF and this data starts | |
826 | * block-aligned, use assign_arcbuf(). Otherwise, | |
827 | * write via dmu_write(). | |
828 | */ | |
829 | if (tx_bytes < max_blksz && (!write_eof || | |
830 | aiov->iov_base != abuf->b_data)) { | |
831 | ASSERT(xuio); | |
832 | dmu_write(zfsvfs->z_os, zp->z_id, woff, | |
833 | aiov->iov_len, aiov->iov_base, tx); | |
834 | dmu_return_arcbuf(abuf); | |
835 | xuio_stat_wbuf_copied(); | |
836 | } else { | |
837 | ASSERT(xuio || tx_bytes == max_blksz); | |
838 | dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl), | |
839 | woff, abuf, tx); | |
840 | } | |
9babb374 BB |
841 | ASSERT(tx_bytes <= uio->uio_resid); |
842 | uioskip(uio, tx_bytes); | |
843 | } | |
844 | if (tx_bytes && vn_has_cached_data(vp)) { | |
d164b209 BB |
845 | update_pages(vp, woff, |
846 | tx_bytes, zfsvfs->z_os, zp->z_id); | |
9babb374 | 847 | } |
34dc7c2f BB |
848 | |
849 | /* | |
850 | * If we made no progress, we're done. If we made even | |
851 | * partial progress, update the znode and ZIL accordingly. | |
852 | */ | |
853 | if (tx_bytes == 0) { | |
428870ff BB |
854 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), |
855 | (void *)&zp->z_size, sizeof (uint64_t), tx); | |
34dc7c2f BB |
856 | dmu_tx_commit(tx); |
857 | ASSERT(error != 0); | |
858 | break; | |
859 | } | |
860 | ||
861 | /* | |
862 | * Clear Set-UID/Set-GID bits on successful write if not | |
863 | * privileged and at least one of the excute bits is set. | |
864 | * | |
865 | * It would be nice to to this after all writes have | |
866 | * been done, but that would still expose the ISUID/ISGID | |
867 | * to another app after the partial write is committed. | |
868 | * | |
572e2857 BB |
869 | * Note: we don't call zfs_fuid_map_id() here because |
870 | * user 0 is not an ephemeral uid. | |
34dc7c2f BB |
871 | */ |
872 | mutex_enter(&zp->z_acl_lock); | |
428870ff | 873 | if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | |
34dc7c2f | 874 | (S_IXUSR >> 6))) != 0 && |
428870ff | 875 | (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && |
34dc7c2f | 876 | secpolicy_vnode_setid_retain(cr, |
428870ff BB |
877 | (zp->z_mode & S_ISUID) != 0 && zp->z_uid == 0) != 0) { |
878 | uint64_t newmode; | |
879 | zp->z_mode &= ~(S_ISUID | S_ISGID); | |
880 | newmode = zp->z_mode; | |
881 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), | |
882 | (void *)&newmode, sizeof (uint64_t), tx); | |
34dc7c2f BB |
883 | } |
884 | mutex_exit(&zp->z_acl_lock); | |
885 | ||
428870ff BB |
886 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, |
887 | B_TRUE); | |
34dc7c2f BB |
888 | |
889 | /* | |
890 | * Update the file size (zp_size) if it has changed; | |
891 | * account for possible concurrent updates. | |
892 | */ | |
428870ff BB |
893 | while ((end_size = zp->z_size) < uio->uio_loffset) { |
894 | (void) atomic_cas_64(&zp->z_size, end_size, | |
34dc7c2f | 895 | uio->uio_loffset); |
428870ff BB |
896 | ASSERT(error == 0); |
897 | } | |
572e2857 BB |
898 | /* |
899 | * If we are replaying and eof is non zero then force | |
900 | * the file size to the specified eof. Note, there's no | |
901 | * concurrency during replay. | |
902 | */ | |
903 | if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) | |
904 | zp->z_size = zfsvfs->z_replay_eof; | |
905 | ||
428870ff BB |
906 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
907 | ||
34dc7c2f BB |
908 | zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag); |
909 | dmu_tx_commit(tx); | |
910 | ||
911 | if (error != 0) | |
912 | break; | |
913 | ASSERT(tx_bytes == nbytes); | |
914 | n -= nbytes; | |
572e2857 BB |
915 | |
916 | if (!xuio && n > 0) | |
917 | uio_prefaultpages(MIN(n, max_blksz), uio); | |
34dc7c2f BB |
918 | } |
919 | ||
920 | zfs_range_unlock(rl); | |
921 | ||
922 | /* | |
923 | * If we're in replay mode, or we made no progress, return error. | |
924 | * Otherwise, it's at least a partial write, so it's successful. | |
925 | */ | |
fb5f0bc8 | 926 | if (zfsvfs->z_replay || uio->uio_resid == start_resid) { |
34dc7c2f BB |
927 | ZFS_EXIT(zfsvfs); |
928 | return (error); | |
929 | } | |
930 | ||
428870ff BB |
931 | if (ioflag & (FSYNC | FDSYNC) || |
932 | zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) | |
572e2857 | 933 | zil_commit(zilog, zp->z_id); |
34dc7c2f BB |
934 | |
935 | ZFS_EXIT(zfsvfs); | |
936 | return (0); | |
937 | } | |
938 | ||
939 | void | |
428870ff | 940 | zfs_get_done(zgd_t *zgd, int error) |
34dc7c2f | 941 | { |
428870ff BB |
942 | znode_t *zp = zgd->zgd_private; |
943 | objset_t *os = zp->z_zfsvfs->z_os; | |
944 | ||
945 | if (zgd->zgd_db) | |
946 | dmu_buf_rele(zgd->zgd_db, zgd); | |
947 | ||
948 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f | 949 | |
9babb374 BB |
950 | /* |
951 | * Release the vnode asynchronously as we currently have the | |
952 | * txg stopped from syncing. | |
953 | */ | |
428870ff BB |
954 | VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os))); |
955 | ||
956 | if (error == 0 && zgd->zgd_bp) | |
957 | zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); | |
958 | ||
34dc7c2f BB |
959 | kmem_free(zgd, sizeof (zgd_t)); |
960 | } | |
961 | ||
45d1cae3 BB |
962 | #ifdef DEBUG |
963 | static int zil_fault_io = 0; | |
964 | #endif | |
965 | ||
34dc7c2f BB |
966 | /* |
967 | * Get data to generate a TX_WRITE intent log record. | |
968 | */ | |
969 | int | |
970 | zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) | |
971 | { | |
972 | zfsvfs_t *zfsvfs = arg; | |
973 | objset_t *os = zfsvfs->z_os; | |
974 | znode_t *zp; | |
428870ff BB |
975 | uint64_t object = lr->lr_foid; |
976 | uint64_t offset = lr->lr_offset; | |
977 | uint64_t size = lr->lr_length; | |
978 | blkptr_t *bp = &lr->lr_blkptr; | |
34dc7c2f | 979 | dmu_buf_t *db; |
34dc7c2f | 980 | zgd_t *zgd; |
34dc7c2f BB |
981 | int error = 0; |
982 | ||
428870ff BB |
983 | ASSERT(zio != NULL); |
984 | ASSERT(size != 0); | |
34dc7c2f BB |
985 | |
986 | /* | |
987 | * Nothing to do if the file has been removed | |
988 | */ | |
428870ff | 989 | if (zfs_zget(zfsvfs, object, &zp) != 0) |
34dc7c2f BB |
990 | return (ENOENT); |
991 | if (zp->z_unlinked) { | |
9babb374 BB |
992 | /* |
993 | * Release the vnode asynchronously as we currently have the | |
994 | * txg stopped from syncing. | |
995 | */ | |
996 | VN_RELE_ASYNC(ZTOV(zp), | |
997 | dsl_pool_vnrele_taskq(dmu_objset_pool(os))); | |
34dc7c2f BB |
998 | return (ENOENT); |
999 | } | |
1000 | ||
428870ff BB |
1001 | zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); |
1002 | zgd->zgd_zilog = zfsvfs->z_log; | |
1003 | zgd->zgd_private = zp; | |
1004 | ||
34dc7c2f BB |
1005 | /* |
1006 | * Write records come in two flavors: immediate and indirect. | |
1007 | * For small writes it's cheaper to store the data with the | |
1008 | * log record (immediate); for large writes it's cheaper to | |
1009 | * sync the data and get a pointer to it (indirect) so that | |
1010 | * we don't have to write the data twice. | |
1011 | */ | |
1012 | if (buf != NULL) { /* immediate write */ | |
428870ff | 1013 | zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER); |
34dc7c2f | 1014 | /* test for truncation needs to be done while range locked */ |
428870ff | 1015 | if (offset >= zp->z_size) { |
34dc7c2f | 1016 | error = ENOENT; |
428870ff BB |
1017 | } else { |
1018 | error = dmu_read(os, object, offset, size, buf, | |
1019 | DMU_READ_NO_PREFETCH); | |
34dc7c2f | 1020 | } |
428870ff | 1021 | ASSERT(error == 0 || error == ENOENT); |
34dc7c2f | 1022 | } else { /* indirect write */ |
34dc7c2f BB |
1023 | /* |
1024 | * Have to lock the whole block to ensure when it's | |
1025 | * written out and it's checksum is being calculated | |
1026 | * that no one can change the data. We need to re-check | |
1027 | * blocksize after we get the lock in case it's changed! | |
1028 | */ | |
1029 | for (;;) { | |
428870ff BB |
1030 | uint64_t blkoff; |
1031 | size = zp->z_blksz; | |
1032 | blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; | |
1033 | offset -= blkoff; | |
1034 | zgd->zgd_rl = zfs_range_lock(zp, offset, size, | |
1035 | RL_READER); | |
1036 | if (zp->z_blksz == size) | |
34dc7c2f | 1037 | break; |
428870ff BB |
1038 | offset += blkoff; |
1039 | zfs_range_unlock(zgd->zgd_rl); | |
34dc7c2f BB |
1040 | } |
1041 | /* test for truncation needs to be done while range locked */ | |
428870ff | 1042 | if (lr->lr_offset >= zp->z_size) |
34dc7c2f | 1043 | error = ENOENT; |
45d1cae3 BB |
1044 | #ifdef DEBUG |
1045 | if (zil_fault_io) { | |
1046 | error = EIO; | |
1047 | zil_fault_io = 0; | |
45d1cae3 | 1048 | } |
45d1cae3 | 1049 | #endif |
34dc7c2f | 1050 | if (error == 0) |
428870ff BB |
1051 | error = dmu_buf_hold(os, object, offset, zgd, &db, |
1052 | DMU_READ_NO_PREFETCH); | |
1053 | ||
1054 | if (error == 0) { | |
1055 | zgd->zgd_db = db; | |
1056 | zgd->zgd_bp = bp; | |
1057 | ||
1058 | ASSERT(db->db_offset == offset); | |
1059 | ASSERT(db->db_size == size); | |
1060 | ||
1061 | error = dmu_sync(zio, lr->lr_common.lrc_txg, | |
1062 | zfs_get_done, zgd); | |
1063 | ASSERT(error || lr->lr_length <= zp->z_blksz); | |
1064 | ||
1065 | /* | |
1066 | * On success, we need to wait for the write I/O | |
1067 | * initiated by dmu_sync() to complete before we can | |
1068 | * release this dbuf. We will finish everything up | |
1069 | * in the zfs_get_done() callback. | |
1070 | */ | |
1071 | if (error == 0) | |
1072 | return (0); | |
1073 | ||
1074 | if (error == EALREADY) { | |
1075 | lr->lr_common.lrc_txtype = TX_WRITE2; | |
1076 | error = 0; | |
1077 | } | |
1078 | } | |
34dc7c2f | 1079 | } |
428870ff BB |
1080 | |
1081 | zfs_get_done(zgd, error); | |
1082 | ||
34dc7c2f BB |
1083 | return (error); |
1084 | } | |
1085 | ||
1086 | /*ARGSUSED*/ | |
1087 | static int | |
1088 | zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr, | |
1089 | caller_context_t *ct) | |
1090 | { | |
1091 | znode_t *zp = VTOZ(vp); | |
1092 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
1093 | int error; | |
1094 | ||
1095 | ZFS_ENTER(zfsvfs); | |
1096 | ZFS_VERIFY_ZP(zp); | |
1097 | ||
1098 | if (flag & V_ACE_MASK) | |
1099 | error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); | |
1100 | else | |
1101 | error = zfs_zaccess_rwx(zp, mode, flag, cr); | |
1102 | ||
1103 | ZFS_EXIT(zfsvfs); | |
1104 | return (error); | |
1105 | } | |
1106 | ||
45d1cae3 BB |
1107 | /* |
1108 | * If vnode is for a device return a specfs vnode instead. | |
1109 | */ | |
1110 | static int | |
1111 | specvp_check(vnode_t **vpp, cred_t *cr) | |
1112 | { | |
1113 | int error = 0; | |
1114 | ||
1115 | if (IS_DEVVP(*vpp)) { | |
1116 | struct vnode *svp; | |
1117 | ||
1118 | svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); | |
1119 | VN_RELE(*vpp); | |
1120 | if (svp == NULL) | |
1121 | error = ENOSYS; | |
1122 | *vpp = svp; | |
1123 | } | |
1124 | return (error); | |
1125 | } | |
1126 | ||
1127 | ||
34dc7c2f BB |
1128 | /* |
1129 | * Lookup an entry in a directory, or an extended attribute directory. | |
1130 | * If it exists, return a held vnode reference for it. | |
1131 | * | |
1132 | * IN: dvp - vnode of directory to search. | |
1133 | * nm - name of entry to lookup. | |
1134 | * pnp - full pathname to lookup [UNUSED]. | |
1135 | * flags - LOOKUP_XATTR set if looking for an attribute. | |
1136 | * rdir - root directory vnode [UNUSED]. | |
1137 | * cr - credentials of caller. | |
1138 | * ct - caller context | |
1139 | * direntflags - directory lookup flags | |
1140 | * realpnp - returned pathname. | |
1141 | * | |
1142 | * OUT: vpp - vnode of located entry, NULL if not found. | |
1143 | * | |
1144 | * RETURN: 0 if success | |
1145 | * error code if failure | |
1146 | * | |
1147 | * Timestamps: | |
1148 | * NA | |
1149 | */ | |
1150 | /* ARGSUSED */ | |
1151 | static int | |
1152 | zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp, | |
1153 | int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct, | |
1154 | int *direntflags, pathname_t *realpnp) | |
1155 | { | |
1156 | znode_t *zdp = VTOZ(dvp); | |
1157 | zfsvfs_t *zfsvfs = zdp->z_zfsvfs; | |
45d1cae3 BB |
1158 | int error = 0; |
1159 | ||
1160 | /* fast path */ | |
1161 | if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { | |
1162 | ||
1163 | if (dvp->v_type != VDIR) { | |
1164 | return (ENOTDIR); | |
428870ff | 1165 | } else if (zdp->z_sa_hdl == NULL) { |
45d1cae3 BB |
1166 | return (EIO); |
1167 | } | |
1168 | ||
1169 | if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { | |
1170 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1171 | if (!error) { | |
1172 | *vpp = dvp; | |
1173 | VN_HOLD(*vpp); | |
1174 | return (0); | |
1175 | } | |
1176 | return (error); | |
1177 | } else { | |
1178 | vnode_t *tvp = dnlc_lookup(dvp, nm); | |
1179 | ||
1180 | if (tvp) { | |
1181 | error = zfs_fastaccesschk_execute(zdp, cr); | |
1182 | if (error) { | |
1183 | VN_RELE(tvp); | |
1184 | return (error); | |
1185 | } | |
1186 | if (tvp == DNLC_NO_VNODE) { | |
1187 | VN_RELE(tvp); | |
1188 | return (ENOENT); | |
1189 | } else { | |
1190 | *vpp = tvp; | |
1191 | return (specvp_check(vpp, cr)); | |
1192 | } | |
1193 | } | |
1194 | } | |
1195 | } | |
1196 | ||
1197 | DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm); | |
34dc7c2f BB |
1198 | |
1199 | ZFS_ENTER(zfsvfs); | |
1200 | ZFS_VERIFY_ZP(zdp); | |
1201 | ||
1202 | *vpp = NULL; | |
1203 | ||
1204 | if (flags & LOOKUP_XATTR) { | |
1205 | /* | |
1206 | * If the xattr property is off, refuse the lookup request. | |
1207 | */ | |
1208 | if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) { | |
1209 | ZFS_EXIT(zfsvfs); | |
1210 | return (EINVAL); | |
1211 | } | |
1212 | ||
1213 | /* | |
1214 | * We don't allow recursive attributes.. | |
1215 | * Maybe someday we will. | |
1216 | */ | |
428870ff | 1217 | if (zdp->z_pflags & ZFS_XATTR) { |
34dc7c2f BB |
1218 | ZFS_EXIT(zfsvfs); |
1219 | return (EINVAL); | |
1220 | } | |
1221 | ||
1222 | if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) { | |
1223 | ZFS_EXIT(zfsvfs); | |
1224 | return (error); | |
1225 | } | |
1226 | ||
1227 | /* | |
1228 | * Do we have permission to get into attribute directory? | |
1229 | */ | |
1230 | ||
1231 | if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0, | |
1232 | B_FALSE, cr)) { | |
1233 | VN_RELE(*vpp); | |
1234 | *vpp = NULL; | |
1235 | } | |
1236 | ||
1237 | ZFS_EXIT(zfsvfs); | |
1238 | return (error); | |
1239 | } | |
1240 | ||
1241 | if (dvp->v_type != VDIR) { | |
1242 | ZFS_EXIT(zfsvfs); | |
1243 | return (ENOTDIR); | |
1244 | } | |
1245 | ||
1246 | /* | |
1247 | * Check accessibility of directory. | |
1248 | */ | |
1249 | ||
1250 | if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) { | |
1251 | ZFS_EXIT(zfsvfs); | |
1252 | return (error); | |
1253 | } | |
1254 | ||
1255 | if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), | |
1256 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { | |
1257 | ZFS_EXIT(zfsvfs); | |
1258 | return (EILSEQ); | |
1259 | } | |
1260 | ||
1261 | error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp); | |
45d1cae3 BB |
1262 | if (error == 0) |
1263 | error = specvp_check(vpp, cr); | |
34dc7c2f BB |
1264 | |
1265 | ZFS_EXIT(zfsvfs); | |
1266 | return (error); | |
1267 | } | |
1268 | ||
1269 | /* | |
1270 | * Attempt to create a new entry in a directory. If the entry | |
1271 | * already exists, truncate the file if permissible, else return | |
1272 | * an error. Return the vp of the created or trunc'd file. | |
1273 | * | |
1274 | * IN: dvp - vnode of directory to put new file entry in. | |
1275 | * name - name of new file entry. | |
1276 | * vap - attributes of new file. | |
1277 | * excl - flag indicating exclusive or non-exclusive mode. | |
1278 | * mode - mode to open file with. | |
1279 | * cr - credentials of caller. | |
1280 | * flag - large file flag [UNUSED]. | |
1281 | * ct - caller context | |
1282 | * vsecp - ACL to be set | |
1283 | * | |
1284 | * OUT: vpp - vnode of created or trunc'd entry. | |
1285 | * | |
1286 | * RETURN: 0 if success | |
1287 | * error code if failure | |
1288 | * | |
1289 | * Timestamps: | |
1290 | * dvp - ctime|mtime updated if new entry created | |
1291 | * vp - ctime|mtime always, atime if new | |
1292 | */ | |
1293 | ||
1294 | /* ARGSUSED */ | |
1295 | static int | |
1296 | zfs_create(vnode_t *dvp, char *name, vattr_t *vap, vcexcl_t excl, | |
1297 | int mode, vnode_t **vpp, cred_t *cr, int flag, caller_context_t *ct, | |
1298 | vsecattr_t *vsecp) | |
1299 | { | |
1300 | znode_t *zp, *dzp = VTOZ(dvp); | |
1301 | zfsvfs_t *zfsvfs = dzp->z_zfsvfs; | |
1302 | zilog_t *zilog; | |
1303 | objset_t *os; | |
1304 | zfs_dirlock_t *dl; | |
1305 | dmu_tx_t *tx; | |
1306 | int error; | |
b128c09f BB |
1307 | ksid_t *ksid; |
1308 | uid_t uid; | |
1309 | gid_t gid = crgetgid(cr); | |
428870ff | 1310 | zfs_acl_ids_t acl_ids; |
9babb374 | 1311 | boolean_t fuid_dirtied; |
428870ff | 1312 | boolean_t have_acl = B_FALSE; |
34dc7c2f BB |
1313 | |
1314 | /* | |
1315 | * If we have an ephemeral id, ACL, or XVATTR then | |
1316 | * make sure file system is at proper version | |
1317 | */ | |
1318 | ||
b128c09f BB |
1319 | ksid = crgetsid(cr, KSID_OWNER); |
1320 | if (ksid) | |
1321 | uid = ksid_getid(ksid); | |
1322 | else | |
1323 | uid = crgetuid(cr); | |
1324 | ||
34dc7c2f BB |
1325 | if (zfsvfs->z_use_fuids == B_FALSE && |
1326 | (vsecp || (vap->va_mask & AT_XVATTR) || | |
b128c09f | 1327 | IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) |
34dc7c2f BB |
1328 | return (EINVAL); |
1329 | ||
1330 | ZFS_ENTER(zfsvfs); | |
1331 | ZFS_VERIFY_ZP(dzp); | |
1332 | os = zfsvfs->z_os; | |
1333 | zilog = zfsvfs->z_log; | |
1334 | ||
1335 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), | |
1336 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { | |
1337 | ZFS_EXIT(zfsvfs); | |
1338 | return (EILSEQ); | |
1339 | } | |
1340 | ||
1341 | if (vap->va_mask & AT_XVATTR) { | |
1342 | if ((error = secpolicy_xvattr((xvattr_t *)vap, | |
1343 | crgetuid(cr), cr, vap->va_type)) != 0) { | |
1344 | ZFS_EXIT(zfsvfs); | |
1345 | return (error); | |
1346 | } | |
1347 | } | |
1348 | top: | |
1349 | *vpp = NULL; | |
1350 | ||
1351 | if ((vap->va_mode & VSVTX) && secpolicy_vnode_stky_modify(cr)) | |
1352 | vap->va_mode &= ~VSVTX; | |
1353 | ||
1354 | if (*name == '\0') { | |
1355 | /* | |
1356 | * Null component name refers to the directory itself. | |
1357 | */ | |
1358 | VN_HOLD(dvp); | |
1359 | zp = dzp; | |
1360 | dl = NULL; | |
1361 | error = 0; | |
1362 | } else { | |
1363 | /* possible VN_HOLD(zp) */ | |
1364 | int zflg = 0; | |
1365 | ||
1366 | if (flag & FIGNORECASE) | |
1367 | zflg |= ZCILOOK; | |
1368 | ||
1369 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1370 | NULL, NULL); | |
1371 | if (error) { | |
572e2857 BB |
1372 | if (have_acl) |
1373 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1374 | if (strcmp(name, "..") == 0) |
1375 | error = EISDIR; | |
1376 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
1377 | return (error); |
1378 | } | |
1379 | } | |
428870ff | 1380 | |
34dc7c2f BB |
1381 | if (zp == NULL) { |
1382 | uint64_t txtype; | |
1383 | ||
1384 | /* | |
1385 | * Create a new file object and update the directory | |
1386 | * to reference it. | |
1387 | */ | |
1388 | if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { | |
572e2857 BB |
1389 | if (have_acl) |
1390 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1391 | goto out; |
1392 | } | |
1393 | ||
1394 | /* | |
1395 | * We only support the creation of regular files in | |
1396 | * extended attribute directories. | |
1397 | */ | |
428870ff BB |
1398 | |
1399 | if ((dzp->z_pflags & ZFS_XATTR) && | |
34dc7c2f | 1400 | (vap->va_type != VREG)) { |
572e2857 BB |
1401 | if (have_acl) |
1402 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1403 | error = EINVAL; |
1404 | goto out; | |
1405 | } | |
1406 | ||
428870ff BB |
1407 | if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, |
1408 | cr, vsecp, &acl_ids)) != 0) | |
9babb374 | 1409 | goto out; |
428870ff BB |
1410 | have_acl = B_TRUE; |
1411 | ||
9babb374 | 1412 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
45d1cae3 | 1413 | zfs_acl_ids_free(&acl_ids); |
9babb374 BB |
1414 | error = EDQUOT; |
1415 | goto out; | |
1416 | } | |
1417 | ||
34dc7c2f | 1418 | tx = dmu_tx_create(os); |
428870ff BB |
1419 | |
1420 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1421 | ZFS_SA_BASE_ATTR_SIZE); | |
1422 | ||
9babb374 BB |
1423 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
1424 | if (fuid_dirtied) | |
1425 | zfs_fuid_txhold(zfsvfs, tx); | |
34dc7c2f | 1426 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
1427 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); |
1428 | if (!zfsvfs->z_use_sa && | |
1429 | acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
34dc7c2f | 1430 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
428870ff | 1431 | 0, acl_ids.z_aclp->z_acl_bytes); |
34dc7c2f | 1432 | } |
fb5f0bc8 | 1433 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
1434 | if (error) { |
1435 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1436 | if (error == ERESTART) { |
34dc7c2f BB |
1437 | dmu_tx_wait(tx); |
1438 | dmu_tx_abort(tx); | |
1439 | goto top; | |
1440 | } | |
428870ff | 1441 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
1442 | dmu_tx_abort(tx); |
1443 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
1444 | return (error); |
1445 | } | |
428870ff | 1446 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 BB |
1447 | |
1448 | if (fuid_dirtied) | |
1449 | zfs_fuid_sync(zfsvfs, tx); | |
1450 | ||
34dc7c2f BB |
1451 | (void) zfs_link_create(dl, zp, tx, ZNEW); |
1452 | txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); | |
1453 | if (flag & FIGNORECASE) | |
1454 | txtype |= TX_CI; | |
1455 | zfs_log_create(zilog, tx, txtype, dzp, zp, name, | |
9babb374 BB |
1456 | vsecp, acl_ids.z_fuidp, vap); |
1457 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
1458 | dmu_tx_commit(tx); |
1459 | } else { | |
1460 | int aflags = (flag & FAPPEND) ? V_APPEND : 0; | |
1461 | ||
572e2857 BB |
1462 | if (have_acl) |
1463 | zfs_acl_ids_free(&acl_ids); | |
1464 | have_acl = B_FALSE; | |
1465 | ||
34dc7c2f BB |
1466 | /* |
1467 | * A directory entry already exists for this name. | |
1468 | */ | |
1469 | /* | |
1470 | * Can't truncate an existing file if in exclusive mode. | |
1471 | */ | |
1472 | if (excl == EXCL) { | |
1473 | error = EEXIST; | |
1474 | goto out; | |
1475 | } | |
1476 | /* | |
1477 | * Can't open a directory for writing. | |
1478 | */ | |
1479 | if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) { | |
1480 | error = EISDIR; | |
1481 | goto out; | |
1482 | } | |
1483 | /* | |
1484 | * Verify requested access to file. | |
1485 | */ | |
1486 | if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { | |
1487 | goto out; | |
1488 | } | |
1489 | ||
1490 | mutex_enter(&dzp->z_lock); | |
1491 | dzp->z_seq++; | |
1492 | mutex_exit(&dzp->z_lock); | |
1493 | ||
1494 | /* | |
1495 | * Truncate regular files if requested. | |
1496 | */ | |
1497 | if ((ZTOV(zp)->v_type == VREG) && | |
1498 | (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) { | |
b128c09f BB |
1499 | /* we can't hold any locks when calling zfs_freesp() */ |
1500 | zfs_dirent_unlock(dl); | |
1501 | dl = NULL; | |
34dc7c2f | 1502 | error = zfs_freesp(zp, 0, 0, mode, TRUE); |
34dc7c2f BB |
1503 | if (error == 0) { |
1504 | vnevent_create(ZTOV(zp), ct); | |
1505 | } | |
1506 | } | |
1507 | } | |
1508 | out: | |
1509 | ||
1510 | if (dl) | |
1511 | zfs_dirent_unlock(dl); | |
1512 | ||
1513 | if (error) { | |
1514 | if (zp) | |
1515 | VN_RELE(ZTOV(zp)); | |
1516 | } else { | |
1517 | *vpp = ZTOV(zp); | |
45d1cae3 | 1518 | error = specvp_check(vpp, cr); |
34dc7c2f | 1519 | } |
34dc7c2f | 1520 | |
428870ff | 1521 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1522 | zil_commit(zilog, 0); |
428870ff | 1523 | |
34dc7c2f BB |
1524 | ZFS_EXIT(zfsvfs); |
1525 | return (error); | |
1526 | } | |
1527 | ||
1528 | /* | |
1529 | * Remove an entry from a directory. | |
1530 | * | |
1531 | * IN: dvp - vnode of directory to remove entry from. | |
1532 | * name - name of entry to remove. | |
1533 | * cr - credentials of caller. | |
1534 | * ct - caller context | |
1535 | * flags - case flags | |
1536 | * | |
1537 | * RETURN: 0 if success | |
1538 | * error code if failure | |
1539 | * | |
1540 | * Timestamps: | |
1541 | * dvp - ctime|mtime | |
1542 | * vp - ctime (if nlink > 0) | |
1543 | */ | |
428870ff BB |
1544 | |
1545 | uint64_t null_xattr = 0; | |
1546 | ||
34dc7c2f BB |
1547 | /*ARGSUSED*/ |
1548 | static int | |
1549 | zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct, | |
1550 | int flags) | |
1551 | { | |
1552 | znode_t *zp, *dzp = VTOZ(dvp); | |
572e2857 | 1553 | znode_t *xzp; |
34dc7c2f BB |
1554 | vnode_t *vp; |
1555 | zfsvfs_t *zfsvfs = dzp->z_zfsvfs; | |
1556 | zilog_t *zilog; | |
572e2857 | 1557 | uint64_t acl_obj, xattr_obj; |
428870ff | 1558 | uint64_t xattr_obj_unlinked = 0; |
572e2857 | 1559 | uint64_t obj = 0; |
34dc7c2f BB |
1560 | zfs_dirlock_t *dl; |
1561 | dmu_tx_t *tx; | |
1562 | boolean_t may_delete_now, delete_now = FALSE; | |
b128c09f | 1563 | boolean_t unlinked, toobig = FALSE; |
34dc7c2f BB |
1564 | uint64_t txtype; |
1565 | pathname_t *realnmp = NULL; | |
1566 | pathname_t realnm; | |
1567 | int error; | |
1568 | int zflg = ZEXISTS; | |
1569 | ||
1570 | ZFS_ENTER(zfsvfs); | |
1571 | ZFS_VERIFY_ZP(dzp); | |
1572 | zilog = zfsvfs->z_log; | |
1573 | ||
1574 | if (flags & FIGNORECASE) { | |
1575 | zflg |= ZCILOOK; | |
1576 | pn_alloc(&realnm); | |
1577 | realnmp = &realnm; | |
1578 | } | |
1579 | ||
1580 | top: | |
572e2857 BB |
1581 | xattr_obj = 0; |
1582 | xzp = NULL; | |
34dc7c2f BB |
1583 | /* |
1584 | * Attempt to lock directory; fail if entry doesn't exist. | |
1585 | */ | |
1586 | if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1587 | NULL, realnmp)) { | |
1588 | if (realnmp) | |
1589 | pn_free(realnmp); | |
1590 | ZFS_EXIT(zfsvfs); | |
1591 | return (error); | |
1592 | } | |
1593 | ||
1594 | vp = ZTOV(zp); | |
1595 | ||
1596 | if (error = zfs_zaccess_delete(dzp, zp, cr)) { | |
1597 | goto out; | |
1598 | } | |
1599 | ||
1600 | /* | |
1601 | * Need to use rmdir for removing directories. | |
1602 | */ | |
1603 | if (vp->v_type == VDIR) { | |
1604 | error = EPERM; | |
1605 | goto out; | |
1606 | } | |
1607 | ||
1608 | vnevent_remove(vp, dvp, name, ct); | |
1609 | ||
1610 | if (realnmp) | |
1611 | dnlc_remove(dvp, realnmp->pn_buf); | |
1612 | else | |
1613 | dnlc_remove(dvp, name); | |
1614 | ||
1615 | mutex_enter(&vp->v_lock); | |
1616 | may_delete_now = vp->v_count == 1 && !vn_has_cached_data(vp); | |
1617 | mutex_exit(&vp->v_lock); | |
1618 | ||
1619 | /* | |
1620 | * We may delete the znode now, or we may put it in the unlinked set; | |
1621 | * it depends on whether we're the last link, and on whether there are | |
1622 | * other holds on the vnode. So we dmu_tx_hold() the right things to | |
1623 | * allow for either case. | |
1624 | */ | |
572e2857 | 1625 | obj = zp->z_id; |
34dc7c2f BB |
1626 | tx = dmu_tx_create(zfsvfs->z_os); |
1627 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); | |
428870ff BB |
1628 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1629 | zfs_sa_upgrade_txholds(tx, zp); | |
1630 | zfs_sa_upgrade_txholds(tx, dzp); | |
b128c09f BB |
1631 | if (may_delete_now) { |
1632 | toobig = | |
428870ff | 1633 | zp->z_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT; |
b128c09f BB |
1634 | /* if the file is too big, only hold_free a token amount */ |
1635 | dmu_tx_hold_free(tx, zp->z_id, 0, | |
1636 | (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); | |
1637 | } | |
34dc7c2f BB |
1638 | |
1639 | /* are there any extended attributes? */ | |
428870ff BB |
1640 | error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
1641 | &xattr_obj, sizeof (xattr_obj)); | |
572e2857 | 1642 | if (error == 0 && xattr_obj) { |
428870ff BB |
1643 | error = zfs_zget(zfsvfs, xattr_obj, &xzp); |
1644 | ASSERT3U(error, ==, 0); | |
1645 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
1646 | dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
1647 | } |
1648 | ||
572e2857 BB |
1649 | mutex_enter(&zp->z_lock); |
1650 | if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) | |
34dc7c2f | 1651 | dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); |
572e2857 | 1652 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1653 | |
1654 | /* charge as an update -- would be nice not to charge at all */ | |
1655 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); | |
1656 | ||
fb5f0bc8 | 1657 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
1658 | if (error) { |
1659 | zfs_dirent_unlock(dl); | |
1660 | VN_RELE(vp); | |
572e2857 BB |
1661 | if (xzp) |
1662 | VN_RELE(ZTOV(xzp)); | |
fb5f0bc8 | 1663 | if (error == ERESTART) { |
34dc7c2f BB |
1664 | dmu_tx_wait(tx); |
1665 | dmu_tx_abort(tx); | |
1666 | goto top; | |
1667 | } | |
1668 | if (realnmp) | |
1669 | pn_free(realnmp); | |
1670 | dmu_tx_abort(tx); | |
1671 | ZFS_EXIT(zfsvfs); | |
1672 | return (error); | |
1673 | } | |
1674 | ||
1675 | /* | |
1676 | * Remove the directory entry. | |
1677 | */ | |
1678 | error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); | |
1679 | ||
1680 | if (error) { | |
1681 | dmu_tx_commit(tx); | |
1682 | goto out; | |
1683 | } | |
1684 | ||
1685 | if (unlinked) { | |
428870ff | 1686 | |
572e2857 BB |
1687 | /* |
1688 | * Hold z_lock so that we can make sure that the ACL obj | |
1689 | * hasn't changed. Could have been deleted due to | |
1690 | * zfs_sa_upgrade(). | |
1691 | */ | |
1692 | mutex_enter(&zp->z_lock); | |
34dc7c2f | 1693 | mutex_enter(&vp->v_lock); |
428870ff BB |
1694 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
1695 | &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); | |
b128c09f | 1696 | delete_now = may_delete_now && !toobig && |
34dc7c2f | 1697 | vp->v_count == 1 && !vn_has_cached_data(vp) && |
572e2857 | 1698 | xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == |
428870ff | 1699 | acl_obj; |
34dc7c2f BB |
1700 | mutex_exit(&vp->v_lock); |
1701 | } | |
1702 | ||
1703 | if (delete_now) { | |
428870ff BB |
1704 | if (xattr_obj_unlinked) { |
1705 | ASSERT3U(xzp->z_links, ==, 2); | |
34dc7c2f BB |
1706 | mutex_enter(&xzp->z_lock); |
1707 | xzp->z_unlinked = 1; | |
428870ff BB |
1708 | xzp->z_links = 0; |
1709 | error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), | |
1710 | &xzp->z_links, sizeof (xzp->z_links), tx); | |
1711 | ASSERT3U(error, ==, 0); | |
34dc7c2f BB |
1712 | mutex_exit(&xzp->z_lock); |
1713 | zfs_unlinked_add(xzp, tx); | |
572e2857 | 1714 | |
428870ff BB |
1715 | if (zp->z_is_sa) |
1716 | error = sa_remove(zp->z_sa_hdl, | |
1717 | SA_ZPL_XATTR(zfsvfs), tx); | |
1718 | else | |
1719 | error = sa_update(zp->z_sa_hdl, | |
1720 | SA_ZPL_XATTR(zfsvfs), &null_xattr, | |
1721 | sizeof (uint64_t), tx); | |
1722 | ASSERT3U(error, ==, 0); | |
34dc7c2f | 1723 | } |
34dc7c2f BB |
1724 | mutex_enter(&vp->v_lock); |
1725 | vp->v_count--; | |
1726 | ASSERT3U(vp->v_count, ==, 0); | |
1727 | mutex_exit(&vp->v_lock); | |
1728 | mutex_exit(&zp->z_lock); | |
1729 | zfs_znode_delete(zp, tx); | |
1730 | } else if (unlinked) { | |
572e2857 | 1731 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
1732 | zfs_unlinked_add(zp, tx); |
1733 | } | |
1734 | ||
1735 | txtype = TX_REMOVE; | |
1736 | if (flags & FIGNORECASE) | |
1737 | txtype |= TX_CI; | |
572e2857 | 1738 | zfs_log_remove(zilog, tx, txtype, dzp, name, obj); |
34dc7c2f BB |
1739 | |
1740 | dmu_tx_commit(tx); | |
1741 | out: | |
1742 | if (realnmp) | |
1743 | pn_free(realnmp); | |
1744 | ||
1745 | zfs_dirent_unlock(dl); | |
1746 | ||
428870ff | 1747 | if (!delete_now) |
34dc7c2f | 1748 | VN_RELE(vp); |
428870ff | 1749 | if (xzp) |
34dc7c2f | 1750 | VN_RELE(ZTOV(xzp)); |
428870ff BB |
1751 | |
1752 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) | |
572e2857 | 1753 | zil_commit(zilog, 0); |
34dc7c2f BB |
1754 | |
1755 | ZFS_EXIT(zfsvfs); | |
1756 | return (error); | |
1757 | } | |
1758 | ||
1759 | /* | |
1760 | * Create a new directory and insert it into dvp using the name | |
1761 | * provided. Return a pointer to the inserted directory. | |
1762 | * | |
1763 | * IN: dvp - vnode of directory to add subdir to. | |
1764 | * dirname - name of new directory. | |
1765 | * vap - attributes of new directory. | |
1766 | * cr - credentials of caller. | |
1767 | * ct - caller context | |
1768 | * vsecp - ACL to be set | |
1769 | * | |
1770 | * OUT: vpp - vnode of created directory. | |
1771 | * | |
1772 | * RETURN: 0 if success | |
1773 | * error code if failure | |
1774 | * | |
1775 | * Timestamps: | |
1776 | * dvp - ctime|mtime updated | |
1777 | * vp - ctime|mtime|atime updated | |
1778 | */ | |
1779 | /*ARGSUSED*/ | |
1780 | static int | |
1781 | zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr, | |
1782 | caller_context_t *ct, int flags, vsecattr_t *vsecp) | |
1783 | { | |
1784 | znode_t *zp, *dzp = VTOZ(dvp); | |
1785 | zfsvfs_t *zfsvfs = dzp->z_zfsvfs; | |
1786 | zilog_t *zilog; | |
1787 | zfs_dirlock_t *dl; | |
1788 | uint64_t txtype; | |
1789 | dmu_tx_t *tx; | |
1790 | int error; | |
34dc7c2f | 1791 | int zf = ZNEW; |
b128c09f BB |
1792 | ksid_t *ksid; |
1793 | uid_t uid; | |
1794 | gid_t gid = crgetgid(cr); | |
428870ff | 1795 | zfs_acl_ids_t acl_ids; |
9babb374 | 1796 | boolean_t fuid_dirtied; |
34dc7c2f BB |
1797 | |
1798 | ASSERT(vap->va_type == VDIR); | |
1799 | ||
1800 | /* | |
1801 | * If we have an ephemeral id, ACL, or XVATTR then | |
1802 | * make sure file system is at proper version | |
1803 | */ | |
1804 | ||
b128c09f BB |
1805 | ksid = crgetsid(cr, KSID_OWNER); |
1806 | if (ksid) | |
1807 | uid = ksid_getid(ksid); | |
1808 | else | |
1809 | uid = crgetuid(cr); | |
34dc7c2f | 1810 | if (zfsvfs->z_use_fuids == B_FALSE && |
b128c09f BB |
1811 | (vsecp || (vap->va_mask & AT_XVATTR) || |
1812 | IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) | |
34dc7c2f BB |
1813 | return (EINVAL); |
1814 | ||
1815 | ZFS_ENTER(zfsvfs); | |
1816 | ZFS_VERIFY_ZP(dzp); | |
1817 | zilog = zfsvfs->z_log; | |
1818 | ||
428870ff | 1819 | if (dzp->z_pflags & ZFS_XATTR) { |
34dc7c2f BB |
1820 | ZFS_EXIT(zfsvfs); |
1821 | return (EINVAL); | |
1822 | } | |
1823 | ||
1824 | if (zfsvfs->z_utf8 && u8_validate(dirname, | |
1825 | strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { | |
1826 | ZFS_EXIT(zfsvfs); | |
1827 | return (EILSEQ); | |
1828 | } | |
1829 | if (flags & FIGNORECASE) | |
1830 | zf |= ZCILOOK; | |
1831 | ||
428870ff | 1832 | if (vap->va_mask & AT_XVATTR) { |
34dc7c2f BB |
1833 | if ((error = secpolicy_xvattr((xvattr_t *)vap, |
1834 | crgetuid(cr), cr, vap->va_type)) != 0) { | |
1835 | ZFS_EXIT(zfsvfs); | |
1836 | return (error); | |
1837 | } | |
428870ff | 1838 | } |
34dc7c2f | 1839 | |
428870ff BB |
1840 | if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, |
1841 | vsecp, &acl_ids)) != 0) { | |
1842 | ZFS_EXIT(zfsvfs); | |
1843 | return (error); | |
1844 | } | |
34dc7c2f BB |
1845 | /* |
1846 | * First make sure the new directory doesn't exist. | |
428870ff BB |
1847 | * |
1848 | * Existence is checked first to make sure we don't return | |
1849 | * EACCES instead of EEXIST which can cause some applications | |
1850 | * to fail. | |
34dc7c2f BB |
1851 | */ |
1852 | top: | |
1853 | *vpp = NULL; | |
1854 | ||
1855 | if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, | |
1856 | NULL, NULL)) { | |
428870ff | 1857 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
1858 | ZFS_EXIT(zfsvfs); |
1859 | return (error); | |
1860 | } | |
1861 | ||
1862 | if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) { | |
428870ff | 1863 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
1864 | zfs_dirent_unlock(dl); |
1865 | ZFS_EXIT(zfsvfs); | |
1866 | return (error); | |
1867 | } | |
1868 | ||
9babb374 | 1869 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
45d1cae3 | 1870 | zfs_acl_ids_free(&acl_ids); |
9babb374 BB |
1871 | zfs_dirent_unlock(dl); |
1872 | ZFS_EXIT(zfsvfs); | |
1873 | return (EDQUOT); | |
1874 | } | |
1875 | ||
34dc7c2f BB |
1876 | /* |
1877 | * Add a new entry to the directory. | |
1878 | */ | |
1879 | tx = dmu_tx_create(zfsvfs->z_os); | |
1880 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); | |
1881 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
9babb374 BB |
1882 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
1883 | if (fuid_dirtied) | |
1884 | zfs_fuid_txhold(zfsvfs, tx); | |
428870ff BB |
1885 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
1886 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, | |
1887 | acl_ids.z_aclp->z_acl_bytes); | |
1888 | } | |
1889 | ||
1890 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + | |
1891 | ZFS_SA_BASE_ATTR_SIZE); | |
1892 | ||
fb5f0bc8 | 1893 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
1894 | if (error) { |
1895 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 1896 | if (error == ERESTART) { |
34dc7c2f BB |
1897 | dmu_tx_wait(tx); |
1898 | dmu_tx_abort(tx); | |
1899 | goto top; | |
1900 | } | |
428870ff | 1901 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
1902 | dmu_tx_abort(tx); |
1903 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
1904 | return (error); |
1905 | } | |
1906 | ||
1907 | /* | |
1908 | * Create new node. | |
1909 | */ | |
428870ff | 1910 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
34dc7c2f | 1911 | |
9babb374 BB |
1912 | if (fuid_dirtied) |
1913 | zfs_fuid_sync(zfsvfs, tx); | |
428870ff | 1914 | |
34dc7c2f BB |
1915 | /* |
1916 | * Now put new name in parent dir. | |
1917 | */ | |
1918 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
1919 | ||
1920 | *vpp = ZTOV(zp); | |
1921 | ||
1922 | txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); | |
1923 | if (flags & FIGNORECASE) | |
1924 | txtype |= TX_CI; | |
9babb374 BB |
1925 | zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, |
1926 | acl_ids.z_fuidp, vap); | |
34dc7c2f | 1927 | |
9babb374 | 1928 | zfs_acl_ids_free(&acl_ids); |
428870ff | 1929 | |
34dc7c2f BB |
1930 | dmu_tx_commit(tx); |
1931 | ||
1932 | zfs_dirent_unlock(dl); | |
1933 | ||
428870ff | 1934 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 1935 | zil_commit(zilog, 0); |
428870ff | 1936 | |
34dc7c2f BB |
1937 | ZFS_EXIT(zfsvfs); |
1938 | return (0); | |
1939 | } | |
1940 | ||
1941 | /* | |
1942 | * Remove a directory subdir entry. If the current working | |
1943 | * directory is the same as the subdir to be removed, the | |
1944 | * remove will fail. | |
1945 | * | |
1946 | * IN: dvp - vnode of directory to remove from. | |
1947 | * name - name of directory to be removed. | |
1948 | * cwd - vnode of current working directory. | |
1949 | * cr - credentials of caller. | |
1950 | * ct - caller context | |
1951 | * flags - case flags | |
1952 | * | |
1953 | * RETURN: 0 if success | |
1954 | * error code if failure | |
1955 | * | |
1956 | * Timestamps: | |
1957 | * dvp - ctime|mtime updated | |
1958 | */ | |
1959 | /*ARGSUSED*/ | |
1960 | static int | |
1961 | zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr, | |
1962 | caller_context_t *ct, int flags) | |
1963 | { | |
1964 | znode_t *dzp = VTOZ(dvp); | |
1965 | znode_t *zp; | |
1966 | vnode_t *vp; | |
1967 | zfsvfs_t *zfsvfs = dzp->z_zfsvfs; | |
1968 | zilog_t *zilog; | |
1969 | zfs_dirlock_t *dl; | |
1970 | dmu_tx_t *tx; | |
1971 | int error; | |
1972 | int zflg = ZEXISTS; | |
1973 | ||
1974 | ZFS_ENTER(zfsvfs); | |
1975 | ZFS_VERIFY_ZP(dzp); | |
1976 | zilog = zfsvfs->z_log; | |
1977 | ||
1978 | if (flags & FIGNORECASE) | |
1979 | zflg |= ZCILOOK; | |
1980 | top: | |
1981 | zp = NULL; | |
1982 | ||
1983 | /* | |
1984 | * Attempt to lock directory; fail if entry doesn't exist. | |
1985 | */ | |
1986 | if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, | |
1987 | NULL, NULL)) { | |
1988 | ZFS_EXIT(zfsvfs); | |
1989 | return (error); | |
1990 | } | |
1991 | ||
1992 | vp = ZTOV(zp); | |
1993 | ||
1994 | if (error = zfs_zaccess_delete(dzp, zp, cr)) { | |
1995 | goto out; | |
1996 | } | |
1997 | ||
1998 | if (vp->v_type != VDIR) { | |
1999 | error = ENOTDIR; | |
2000 | goto out; | |
2001 | } | |
2002 | ||
2003 | if (vp == cwd) { | |
2004 | error = EINVAL; | |
2005 | goto out; | |
2006 | } | |
2007 | ||
2008 | vnevent_rmdir(vp, dvp, name, ct); | |
2009 | ||
2010 | /* | |
2011 | * Grab a lock on the directory to make sure that noone is | |
2012 | * trying to add (or lookup) entries while we are removing it. | |
2013 | */ | |
2014 | rw_enter(&zp->z_name_lock, RW_WRITER); | |
2015 | ||
2016 | /* | |
2017 | * Grab a lock on the parent pointer to make sure we play well | |
2018 | * with the treewalk and directory rename code. | |
2019 | */ | |
2020 | rw_enter(&zp->z_parent_lock, RW_WRITER); | |
2021 | ||
2022 | tx = dmu_tx_create(zfsvfs->z_os); | |
2023 | dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); | |
428870ff | 2024 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
34dc7c2f | 2025 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
428870ff BB |
2026 | zfs_sa_upgrade_txholds(tx, zp); |
2027 | zfs_sa_upgrade_txholds(tx, dzp); | |
fb5f0bc8 | 2028 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
2029 | if (error) { |
2030 | rw_exit(&zp->z_parent_lock); | |
2031 | rw_exit(&zp->z_name_lock); | |
2032 | zfs_dirent_unlock(dl); | |
2033 | VN_RELE(vp); | |
fb5f0bc8 | 2034 | if (error == ERESTART) { |
34dc7c2f BB |
2035 | dmu_tx_wait(tx); |
2036 | dmu_tx_abort(tx); | |
2037 | goto top; | |
2038 | } | |
2039 | dmu_tx_abort(tx); | |
2040 | ZFS_EXIT(zfsvfs); | |
2041 | return (error); | |
2042 | } | |
2043 | ||
2044 | error = zfs_link_destroy(dl, zp, tx, zflg, NULL); | |
2045 | ||
2046 | if (error == 0) { | |
2047 | uint64_t txtype = TX_RMDIR; | |
2048 | if (flags & FIGNORECASE) | |
2049 | txtype |= TX_CI; | |
572e2857 | 2050 | zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); |
34dc7c2f BB |
2051 | } |
2052 | ||
2053 | dmu_tx_commit(tx); | |
2054 | ||
2055 | rw_exit(&zp->z_parent_lock); | |
2056 | rw_exit(&zp->z_name_lock); | |
2057 | out: | |
2058 | zfs_dirent_unlock(dl); | |
2059 | ||
2060 | VN_RELE(vp); | |
2061 | ||
428870ff | 2062 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 2063 | zil_commit(zilog, 0); |
428870ff | 2064 | |
34dc7c2f BB |
2065 | ZFS_EXIT(zfsvfs); |
2066 | return (error); | |
2067 | } | |
2068 | ||
2069 | /* | |
2070 | * Read as many directory entries as will fit into the provided | |
2071 | * buffer from the given directory cursor position (specified in | |
2072 | * the uio structure. | |
2073 | * | |
2074 | * IN: vp - vnode of directory to read. | |
2075 | * uio - structure supplying read location, range info, | |
2076 | * and return buffer. | |
2077 | * cr - credentials of caller. | |
2078 | * ct - caller context | |
2079 | * flags - case flags | |
2080 | * | |
2081 | * OUT: uio - updated offset and range, buffer filled. | |
2082 | * eofp - set to true if end-of-file detected. | |
2083 | * | |
2084 | * RETURN: 0 if success | |
2085 | * error code if failure | |
2086 | * | |
2087 | * Timestamps: | |
2088 | * vp - atime updated | |
2089 | * | |
2090 | * Note that the low 4 bits of the cookie returned by zap is always zero. | |
2091 | * This allows us to use the low range for "special" directory entries: | |
2092 | * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, | |
2093 | * we use the offset 2 for the '.zfs' directory. | |
2094 | */ | |
2095 | /* ARGSUSED */ | |
2096 | static int | |
2097 | zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, | |
2098 | caller_context_t *ct, int flags) | |
2099 | { | |
2100 | znode_t *zp = VTOZ(vp); | |
2101 | iovec_t *iovp; | |
2102 | edirent_t *eodp; | |
2103 | dirent64_t *odp; | |
2104 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
2105 | objset_t *os; | |
2106 | caddr_t outbuf; | |
2107 | size_t bufsize; | |
2108 | zap_cursor_t zc; | |
2109 | zap_attribute_t zap; | |
2110 | uint_t bytes_wanted; | |
2111 | uint64_t offset; /* must be unsigned; checks for < 1 */ | |
428870ff | 2112 | uint64_t parent; |
34dc7c2f BB |
2113 | int local_eof; |
2114 | int outcount; | |
2115 | int error; | |
2116 | uint8_t prefetch; | |
2117 | boolean_t check_sysattrs; | |
2118 | ||
2119 | ZFS_ENTER(zfsvfs); | |
2120 | ZFS_VERIFY_ZP(zp); | |
2121 | ||
428870ff BB |
2122 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), |
2123 | &parent, sizeof (parent))) != 0) { | |
2124 | ZFS_EXIT(zfsvfs); | |
2125 | return (error); | |
2126 | } | |
2127 | ||
34dc7c2f BB |
2128 | /* |
2129 | * If we are not given an eof variable, | |
2130 | * use a local one. | |
2131 | */ | |
2132 | if (eofp == NULL) | |
2133 | eofp = &local_eof; | |
2134 | ||
2135 | /* | |
2136 | * Check for valid iov_len. | |
2137 | */ | |
2138 | if (uio->uio_iov->iov_len <= 0) { | |
2139 | ZFS_EXIT(zfsvfs); | |
2140 | return (EINVAL); | |
2141 | } | |
2142 | ||
2143 | /* | |
2144 | * Quit if directory has been removed (posix) | |
2145 | */ | |
2146 | if ((*eofp = zp->z_unlinked) != 0) { | |
2147 | ZFS_EXIT(zfsvfs); | |
2148 | return (0); | |
2149 | } | |
2150 | ||
2151 | error = 0; | |
2152 | os = zfsvfs->z_os; | |
2153 | offset = uio->uio_loffset; | |
2154 | prefetch = zp->z_zn_prefetch; | |
2155 | ||
2156 | /* | |
2157 | * Initialize the iterator cursor. | |
2158 | */ | |
2159 | if (offset <= 3) { | |
2160 | /* | |
2161 | * Start iteration from the beginning of the directory. | |
2162 | */ | |
2163 | zap_cursor_init(&zc, os, zp->z_id); | |
2164 | } else { | |
2165 | /* | |
2166 | * The offset is a serialized cursor. | |
2167 | */ | |
2168 | zap_cursor_init_serialized(&zc, os, zp->z_id, offset); | |
2169 | } | |
2170 | ||
2171 | /* | |
2172 | * Get space to change directory entries into fs independent format. | |
2173 | */ | |
2174 | iovp = uio->uio_iov; | |
2175 | bytes_wanted = iovp->iov_len; | |
2176 | if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) { | |
2177 | bufsize = bytes_wanted; | |
2178 | outbuf = kmem_alloc(bufsize, KM_SLEEP); | |
2179 | odp = (struct dirent64 *)outbuf; | |
2180 | } else { | |
2181 | bufsize = bytes_wanted; | |
2182 | odp = (struct dirent64 *)iovp->iov_base; | |
2183 | } | |
2184 | eodp = (struct edirent *)odp; | |
2185 | ||
2186 | /* | |
b128c09f BB |
2187 | * If this VFS supports the system attribute view interface; and |
2188 | * we're looking at an extended attribute directory; and we care | |
2189 | * about normalization conflicts on this vfs; then we must check | |
2190 | * for normalization conflicts with the sysattr name space. | |
34dc7c2f | 2191 | */ |
b128c09f | 2192 | check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) && |
34dc7c2f BB |
2193 | (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm && |
2194 | (flags & V_RDDIR_ENTFLAGS); | |
2195 | ||
2196 | /* | |
2197 | * Transform to file-system independent format | |
2198 | */ | |
2199 | outcount = 0; | |
2200 | while (outcount < bytes_wanted) { | |
2201 | ino64_t objnum; | |
2202 | ushort_t reclen; | |
572e2857 | 2203 | off64_t *next = NULL; |
34dc7c2f BB |
2204 | |
2205 | /* | |
2206 | * Special case `.', `..', and `.zfs'. | |
2207 | */ | |
2208 | if (offset == 0) { | |
2209 | (void) strcpy(zap.za_name, "."); | |
2210 | zap.za_normalization_conflict = 0; | |
2211 | objnum = zp->z_id; | |
2212 | } else if (offset == 1) { | |
2213 | (void) strcpy(zap.za_name, ".."); | |
2214 | zap.za_normalization_conflict = 0; | |
428870ff | 2215 | objnum = parent; |
34dc7c2f BB |
2216 | } else if (offset == 2 && zfs_show_ctldir(zp)) { |
2217 | (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); | |
2218 | zap.za_normalization_conflict = 0; | |
2219 | objnum = ZFSCTL_INO_ROOT; | |
2220 | } else { | |
2221 | /* | |
2222 | * Grab next entry. | |
2223 | */ | |
2224 | if (error = zap_cursor_retrieve(&zc, &zap)) { | |
2225 | if ((*eofp = (error == ENOENT)) != 0) | |
2226 | break; | |
2227 | else | |
2228 | goto update; | |
2229 | } | |
2230 | ||
2231 | if (zap.za_integer_length != 8 || | |
2232 | zap.za_num_integers != 1) { | |
2233 | cmn_err(CE_WARN, "zap_readdir: bad directory " | |
2234 | "entry, obj = %lld, offset = %lld\n", | |
2235 | (u_longlong_t)zp->z_id, | |
2236 | (u_longlong_t)offset); | |
2237 | error = ENXIO; | |
2238 | goto update; | |
2239 | } | |
2240 | ||
2241 | objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); | |
2242 | /* | |
2243 | * MacOS X can extract the object type here such as: | |
2244 | * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer); | |
2245 | */ | |
2246 | ||
2247 | if (check_sysattrs && !zap.za_normalization_conflict) { | |
2248 | zap.za_normalization_conflict = | |
2249 | xattr_sysattr_casechk(zap.za_name); | |
2250 | } | |
2251 | } | |
2252 | ||
9babb374 BB |
2253 | if (flags & V_RDDIR_ACCFILTER) { |
2254 | /* | |
2255 | * If we have no access at all, don't include | |
2256 | * this entry in the returned information | |
2257 | */ | |
2258 | znode_t *ezp; | |
2259 | if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0) | |
2260 | goto skip_entry; | |
2261 | if (!zfs_has_access(ezp, cr)) { | |
2262 | VN_RELE(ZTOV(ezp)); | |
2263 | goto skip_entry; | |
2264 | } | |
2265 | VN_RELE(ZTOV(ezp)); | |
2266 | } | |
2267 | ||
34dc7c2f BB |
2268 | if (flags & V_RDDIR_ENTFLAGS) |
2269 | reclen = EDIRENT_RECLEN(strlen(zap.za_name)); | |
2270 | else | |
2271 | reclen = DIRENT64_RECLEN(strlen(zap.za_name)); | |
2272 | ||
2273 | /* | |
2274 | * Will this entry fit in the buffer? | |
2275 | */ | |
2276 | if (outcount + reclen > bufsize) { | |
2277 | /* | |
2278 | * Did we manage to fit anything in the buffer? | |
2279 | */ | |
2280 | if (!outcount) { | |
2281 | error = EINVAL; | |
2282 | goto update; | |
2283 | } | |
2284 | break; | |
2285 | } | |
2286 | if (flags & V_RDDIR_ENTFLAGS) { | |
2287 | /* | |
2288 | * Add extended flag entry: | |
2289 | */ | |
2290 | eodp->ed_ino = objnum; | |
2291 | eodp->ed_reclen = reclen; | |
2292 | /* NOTE: ed_off is the offset for the *next* entry */ | |
2293 | next = &(eodp->ed_off); | |
2294 | eodp->ed_eflags = zap.za_normalization_conflict ? | |
2295 | ED_CASE_CONFLICT : 0; | |
2296 | (void) strncpy(eodp->ed_name, zap.za_name, | |
2297 | EDIRENT_NAMELEN(reclen)); | |
2298 | eodp = (edirent_t *)((intptr_t)eodp + reclen); | |
2299 | } else { | |
2300 | /* | |
2301 | * Add normal entry: | |
2302 | */ | |
2303 | odp->d_ino = objnum; | |
2304 | odp->d_reclen = reclen; | |
2305 | /* NOTE: d_off is the offset for the *next* entry */ | |
2306 | next = &(odp->d_off); | |
2307 | (void) strncpy(odp->d_name, zap.za_name, | |
2308 | DIRENT64_NAMELEN(reclen)); | |
2309 | odp = (dirent64_t *)((intptr_t)odp + reclen); | |
2310 | } | |
2311 | outcount += reclen; | |
2312 | ||
2313 | ASSERT(outcount <= bufsize); | |
2314 | ||
2315 | /* Prefetch znode */ | |
2316 | if (prefetch) | |
2317 | dmu_prefetch(os, objnum, 0, 0); | |
2318 | ||
9babb374 | 2319 | skip_entry: |
34dc7c2f BB |
2320 | /* |
2321 | * Move to the next entry, fill in the previous offset. | |
2322 | */ | |
2323 | if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { | |
2324 | zap_cursor_advance(&zc); | |
2325 | offset = zap_cursor_serialize(&zc); | |
2326 | } else { | |
2327 | offset += 1; | |
2328 | } | |
572e2857 BB |
2329 | if (next) |
2330 | *next = offset; | |
34dc7c2f BB |
2331 | } |
2332 | zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ | |
2333 | ||
2334 | if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) { | |
2335 | iovp->iov_base += outcount; | |
2336 | iovp->iov_len -= outcount; | |
2337 | uio->uio_resid -= outcount; | |
2338 | } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) { | |
2339 | /* | |
2340 | * Reset the pointer. | |
2341 | */ | |
2342 | offset = uio->uio_loffset; | |
2343 | } | |
2344 | ||
2345 | update: | |
2346 | zap_cursor_fini(&zc); | |
2347 | if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) | |
2348 | kmem_free(outbuf, bufsize); | |
2349 | ||
2350 | if (error == ENOENT) | |
2351 | error = 0; | |
2352 | ||
2353 | ZFS_ACCESSTIME_STAMP(zfsvfs, zp); | |
2354 | ||
2355 | uio->uio_loffset = offset; | |
2356 | ZFS_EXIT(zfsvfs); | |
2357 | return (error); | |
2358 | } | |
2359 | ||
2360 | ulong_t zfs_fsync_sync_cnt = 4; | |
2361 | ||
2362 | static int | |
2363 | zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct) | |
2364 | { | |
2365 | znode_t *zp = VTOZ(vp); | |
2366 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
2367 | ||
2368 | /* | |
2369 | * Regardless of whether this is required for standards conformance, | |
2370 | * this is the logical behavior when fsync() is called on a file with | |
2371 | * dirty pages. We use B_ASYNC since the ZIL transactions are already | |
2372 | * going to be pushed out as part of the zil_commit(). | |
2373 | */ | |
2374 | if (vn_has_cached_data(vp) && !(syncflag & FNODSYNC) && | |
2375 | (vp->v_type == VREG) && !(IS_SWAPVP(vp))) | |
2376 | (void) VOP_PUTPAGE(vp, (offset_t)0, (size_t)0, B_ASYNC, cr, ct); | |
2377 | ||
2378 | (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); | |
2379 | ||
428870ff BB |
2380 | if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { |
2381 | ZFS_ENTER(zfsvfs); | |
2382 | ZFS_VERIFY_ZP(zp); | |
572e2857 | 2383 | zil_commit(zfsvfs->z_log, zp->z_id); |
428870ff BB |
2384 | ZFS_EXIT(zfsvfs); |
2385 | } | |
34dc7c2f BB |
2386 | return (0); |
2387 | } | |
2388 | ||
2389 | ||
2390 | /* | |
2391 | * Get the requested file attributes and place them in the provided | |
2392 | * vattr structure. | |
2393 | * | |
2394 | * IN: vp - vnode of file. | |
2395 | * vap - va_mask identifies requested attributes. | |
2396 | * If AT_XVATTR set, then optional attrs are requested | |
2397 | * flags - ATTR_NOACLCHECK (CIFS server context) | |
2398 | * cr - credentials of caller. | |
2399 | * ct - caller context | |
2400 | * | |
2401 | * OUT: vap - attribute values. | |
2402 | * | |
2403 | * RETURN: 0 (always succeeds) | |
2404 | */ | |
2405 | /* ARGSUSED */ | |
2406 | static int | |
2407 | zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, | |
2408 | caller_context_t *ct) | |
2409 | { | |
2410 | znode_t *zp = VTOZ(vp); | |
2411 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
34dc7c2f BB |
2412 | int error = 0; |
2413 | uint64_t links; | |
428870ff | 2414 | uint64_t mtime[2], ctime[2]; |
34dc7c2f BB |
2415 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ |
2416 | xoptattr_t *xoap = NULL; | |
2417 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
428870ff BB |
2418 | sa_bulk_attr_t bulk[2]; |
2419 | int count = 0; | |
34dc7c2f BB |
2420 | |
2421 | ZFS_ENTER(zfsvfs); | |
2422 | ZFS_VERIFY_ZP(zp); | |
428870ff | 2423 | |
572e2857 BB |
2424 | zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); |
2425 | ||
428870ff BB |
2426 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); |
2427 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
2428 | ||
2429 | if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { | |
2430 | ZFS_EXIT(zfsvfs); | |
2431 | return (error); | |
2432 | } | |
34dc7c2f | 2433 | |
34dc7c2f BB |
2434 | /* |
2435 | * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. | |
2436 | * Also, if we are the owner don't bother, since owner should | |
2437 | * always be allowed to read basic attributes of file. | |
2438 | */ | |
572e2857 BB |
2439 | if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && |
2440 | (vap->va_uid != crgetuid(cr))) { | |
34dc7c2f BB |
2441 | if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, |
2442 | skipaclchk, cr)) { | |
34dc7c2f BB |
2443 | ZFS_EXIT(zfsvfs); |
2444 | return (error); | |
2445 | } | |
2446 | } | |
2447 | ||
2448 | /* | |
2449 | * Return all attributes. It's cheaper to provide the answer | |
2450 | * than to determine whether we were asked the question. | |
2451 | */ | |
2452 | ||
9babb374 | 2453 | mutex_enter(&zp->z_lock); |
34dc7c2f | 2454 | vap->va_type = vp->v_type; |
428870ff | 2455 | vap->va_mode = zp->z_mode & MODEMASK; |
34dc7c2f BB |
2456 | vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev; |
2457 | vap->va_nodeid = zp->z_id; | |
2458 | if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp)) | |
428870ff | 2459 | links = zp->z_links + 1; |
34dc7c2f | 2460 | else |
428870ff | 2461 | links = zp->z_links; |
34dc7c2f | 2462 | vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */ |
428870ff | 2463 | vap->va_size = zp->z_size; |
34dc7c2f BB |
2464 | vap->va_rdev = vp->v_rdev; |
2465 | vap->va_seq = zp->z_seq; | |
2466 | ||
2467 | /* | |
2468 | * Add in any requested optional attributes and the create time. | |
2469 | * Also set the corresponding bits in the returned attribute bitmap. | |
2470 | */ | |
2471 | if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { | |
2472 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
2473 | xoap->xoa_archive = | |
428870ff | 2474 | ((zp->z_pflags & ZFS_ARCHIVE) != 0); |
34dc7c2f BB |
2475 | XVA_SET_RTN(xvap, XAT_ARCHIVE); |
2476 | } | |
2477 | ||
2478 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
2479 | xoap->xoa_readonly = | |
428870ff | 2480 | ((zp->z_pflags & ZFS_READONLY) != 0); |
34dc7c2f BB |
2481 | XVA_SET_RTN(xvap, XAT_READONLY); |
2482 | } | |
2483 | ||
2484 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
2485 | xoap->xoa_system = | |
428870ff | 2486 | ((zp->z_pflags & ZFS_SYSTEM) != 0); |
34dc7c2f BB |
2487 | XVA_SET_RTN(xvap, XAT_SYSTEM); |
2488 | } | |
2489 | ||
2490 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
2491 | xoap->xoa_hidden = | |
428870ff | 2492 | ((zp->z_pflags & ZFS_HIDDEN) != 0); |
34dc7c2f BB |
2493 | XVA_SET_RTN(xvap, XAT_HIDDEN); |
2494 | } | |
2495 | ||
2496 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2497 | xoap->xoa_nounlink = | |
428870ff | 2498 | ((zp->z_pflags & ZFS_NOUNLINK) != 0); |
34dc7c2f BB |
2499 | XVA_SET_RTN(xvap, XAT_NOUNLINK); |
2500 | } | |
2501 | ||
2502 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2503 | xoap->xoa_immutable = | |
428870ff | 2504 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0); |
34dc7c2f BB |
2505 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); |
2506 | } | |
2507 | ||
2508 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2509 | xoap->xoa_appendonly = | |
428870ff | 2510 | ((zp->z_pflags & ZFS_APPENDONLY) != 0); |
34dc7c2f BB |
2511 | XVA_SET_RTN(xvap, XAT_APPENDONLY); |
2512 | } | |
2513 | ||
2514 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2515 | xoap->xoa_nodump = | |
428870ff | 2516 | ((zp->z_pflags & ZFS_NODUMP) != 0); |
34dc7c2f BB |
2517 | XVA_SET_RTN(xvap, XAT_NODUMP); |
2518 | } | |
2519 | ||
2520 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
2521 | xoap->xoa_opaque = | |
428870ff | 2522 | ((zp->z_pflags & ZFS_OPAQUE) != 0); |
34dc7c2f BB |
2523 | XVA_SET_RTN(xvap, XAT_OPAQUE); |
2524 | } | |
2525 | ||
2526 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2527 | xoap->xoa_av_quarantined = | |
428870ff | 2528 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); |
34dc7c2f BB |
2529 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); |
2530 | } | |
2531 | ||
2532 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2533 | xoap->xoa_av_modified = | |
428870ff | 2534 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); |
34dc7c2f BB |
2535 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); |
2536 | } | |
2537 | ||
2538 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && | |
428870ff BB |
2539 | vp->v_type == VREG) { |
2540 | zfs_sa_get_scanstamp(zp, xvap); | |
34dc7c2f BB |
2541 | } |
2542 | ||
2543 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { | |
428870ff BB |
2544 | uint64_t times[2]; |
2545 | ||
2546 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), | |
2547 | times, sizeof (times)); | |
2548 | ZFS_TIME_DECODE(&xoap->xoa_createtime, times); | |
34dc7c2f BB |
2549 | XVA_SET_RTN(xvap, XAT_CREATETIME); |
2550 | } | |
428870ff BB |
2551 | |
2552 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
2553 | xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); | |
2554 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
2555 | } | |
572e2857 BB |
2556 | if (XVA_ISSET_REQ(xvap, XAT_GEN)) { |
2557 | xoap->xoa_generation = zp->z_gen; | |
2558 | XVA_SET_RTN(xvap, XAT_GEN); | |
2559 | } | |
2560 | ||
2561 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
2562 | xoap->xoa_offline = | |
2563 | ((zp->z_pflags & ZFS_OFFLINE) != 0); | |
2564 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
2565 | } | |
2566 | ||
2567 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
2568 | xoap->xoa_sparse = | |
2569 | ((zp->z_pflags & ZFS_SPARSE) != 0); | |
2570 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
2571 | } | |
34dc7c2f BB |
2572 | } |
2573 | ||
428870ff BB |
2574 | ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime); |
2575 | ZFS_TIME_DECODE(&vap->va_mtime, mtime); | |
2576 | ZFS_TIME_DECODE(&vap->va_ctime, ctime); | |
34dc7c2f BB |
2577 | |
2578 | mutex_exit(&zp->z_lock); | |
2579 | ||
428870ff | 2580 | sa_object_size(zp->z_sa_hdl, &vap->va_blksize, &vap->va_nblocks); |
34dc7c2f BB |
2581 | |
2582 | if (zp->z_blksz == 0) { | |
2583 | /* | |
2584 | * Block size hasn't been set; suggest maximal I/O transfers. | |
2585 | */ | |
2586 | vap->va_blksize = zfsvfs->z_max_blksz; | |
2587 | } | |
2588 | ||
2589 | ZFS_EXIT(zfsvfs); | |
2590 | return (0); | |
2591 | } | |
2592 | ||
2593 | /* | |
2594 | * Set the file attributes to the values contained in the | |
2595 | * vattr structure. | |
2596 | * | |
2597 | * IN: vp - vnode of file to be modified. | |
2598 | * vap - new attribute values. | |
2599 | * If AT_XVATTR set, then optional attrs are being set | |
2600 | * flags - ATTR_UTIME set if non-default time values provided. | |
2601 | * - ATTR_NOACLCHECK (CIFS context only). | |
2602 | * cr - credentials of caller. | |
2603 | * ct - caller context | |
2604 | * | |
2605 | * RETURN: 0 if success | |
2606 | * error code if failure | |
2607 | * | |
2608 | * Timestamps: | |
2609 | * vp - ctime updated, mtime updated if size changed. | |
2610 | */ | |
2611 | /* ARGSUSED */ | |
2612 | static int | |
2613 | zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, | |
2614 | caller_context_t *ct) | |
2615 | { | |
2616 | znode_t *zp = VTOZ(vp); | |
34dc7c2f BB |
2617 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; |
2618 | zilog_t *zilog; | |
2619 | dmu_tx_t *tx; | |
2620 | vattr_t oldva; | |
fb5f0bc8 | 2621 | xvattr_t tmpxvattr; |
34dc7c2f BB |
2622 | uint_t mask = vap->va_mask; |
2623 | uint_t saved_mask; | |
2624 | int trim_mask = 0; | |
2625 | uint64_t new_mode; | |
9babb374 | 2626 | uint64_t new_uid, new_gid; |
572e2857 | 2627 | uint64_t xattr_obj; |
428870ff | 2628 | uint64_t mtime[2], ctime[2]; |
34dc7c2f BB |
2629 | znode_t *attrzp; |
2630 | int need_policy = FALSE; | |
428870ff | 2631 | int err, err2; |
34dc7c2f BB |
2632 | zfs_fuid_info_t *fuidp = NULL; |
2633 | xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ | |
2634 | xoptattr_t *xoap; | |
572e2857 | 2635 | zfs_acl_t *aclp; |
34dc7c2f | 2636 | boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; |
428870ff BB |
2637 | boolean_t fuid_dirtied = B_FALSE; |
2638 | sa_bulk_attr_t bulk[7], xattr_bulk[7]; | |
2639 | int count = 0, xattr_count = 0; | |
34dc7c2f BB |
2640 | |
2641 | if (mask == 0) | |
2642 | return (0); | |
2643 | ||
2644 | if (mask & AT_NOSET) | |
2645 | return (EINVAL); | |
2646 | ||
2647 | ZFS_ENTER(zfsvfs); | |
2648 | ZFS_VERIFY_ZP(zp); | |
2649 | ||
34dc7c2f BB |
2650 | zilog = zfsvfs->z_log; |
2651 | ||
2652 | /* | |
2653 | * Make sure that if we have ephemeral uid/gid or xvattr specified | |
2654 | * that file system is at proper version level | |
2655 | */ | |
2656 | ||
2657 | if (zfsvfs->z_use_fuids == B_FALSE && | |
2658 | (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) || | |
2659 | ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) || | |
2660 | (mask & AT_XVATTR))) { | |
2661 | ZFS_EXIT(zfsvfs); | |
2662 | return (EINVAL); | |
2663 | } | |
2664 | ||
2665 | if (mask & AT_SIZE && vp->v_type == VDIR) { | |
2666 | ZFS_EXIT(zfsvfs); | |
2667 | return (EISDIR); | |
2668 | } | |
2669 | ||
2670 | if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) { | |
2671 | ZFS_EXIT(zfsvfs); | |
2672 | return (EINVAL); | |
2673 | } | |
2674 | ||
2675 | /* | |
2676 | * If this is an xvattr_t, then get a pointer to the structure of | |
2677 | * optional attributes. If this is NULL, then we have a vattr_t. | |
2678 | */ | |
2679 | xoap = xva_getxoptattr(xvap); | |
2680 | ||
fb5f0bc8 BB |
2681 | xva_init(&tmpxvattr); |
2682 | ||
34dc7c2f BB |
2683 | /* |
2684 | * Immutable files can only alter immutable bit and atime | |
2685 | */ | |
428870ff | 2686 | if ((zp->z_pflags & ZFS_IMMUTABLE) && |
34dc7c2f BB |
2687 | ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) || |
2688 | ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { | |
2689 | ZFS_EXIT(zfsvfs); | |
2690 | return (EPERM); | |
2691 | } | |
2692 | ||
428870ff | 2693 | if ((mask & AT_SIZE) && (zp->z_pflags & ZFS_READONLY)) { |
34dc7c2f BB |
2694 | ZFS_EXIT(zfsvfs); |
2695 | return (EPERM); | |
2696 | } | |
2697 | ||
2698 | /* | |
2699 | * Verify timestamps doesn't overflow 32 bits. | |
2700 | * ZFS can handle large timestamps, but 32bit syscalls can't | |
2701 | * handle times greater than 2039. This check should be removed | |
2702 | * once large timestamps are fully supported. | |
2703 | */ | |
2704 | if (mask & (AT_ATIME | AT_MTIME)) { | |
2705 | if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || | |
2706 | ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { | |
2707 | ZFS_EXIT(zfsvfs); | |
2708 | return (EOVERFLOW); | |
2709 | } | |
2710 | } | |
2711 | ||
2712 | top: | |
2713 | attrzp = NULL; | |
572e2857 | 2714 | aclp = NULL; |
34dc7c2f | 2715 | |
45d1cae3 | 2716 | /* Can this be moved to before the top label? */ |
34dc7c2f BB |
2717 | if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) { |
2718 | ZFS_EXIT(zfsvfs); | |
2719 | return (EROFS); | |
2720 | } | |
2721 | ||
2722 | /* | |
2723 | * First validate permissions | |
2724 | */ | |
2725 | ||
2726 | if (mask & AT_SIZE) { | |
2727 | err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); | |
2728 | if (err) { | |
2729 | ZFS_EXIT(zfsvfs); | |
2730 | return (err); | |
2731 | } | |
2732 | /* | |
2733 | * XXX - Note, we are not providing any open | |
2734 | * mode flags here (like FNDELAY), so we may | |
2735 | * block if there are locks present... this | |
2736 | * should be addressed in openat(). | |
2737 | */ | |
b128c09f BB |
2738 | /* XXX - would it be OK to generate a log record here? */ |
2739 | err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); | |
34dc7c2f BB |
2740 | if (err) { |
2741 | ZFS_EXIT(zfsvfs); | |
2742 | return (err); | |
2743 | } | |
2744 | } | |
2745 | ||
2746 | if (mask & (AT_ATIME|AT_MTIME) || | |
2747 | ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || | |
2748 | XVA_ISSET_REQ(xvap, XAT_READONLY) || | |
2749 | XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || | |
572e2857 BB |
2750 | XVA_ISSET_REQ(xvap, XAT_OFFLINE) || |
2751 | XVA_ISSET_REQ(xvap, XAT_SPARSE) || | |
34dc7c2f | 2752 | XVA_ISSET_REQ(xvap, XAT_CREATETIME) || |
428870ff | 2753 | XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { |
34dc7c2f BB |
2754 | need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, |
2755 | skipaclchk, cr); | |
428870ff | 2756 | } |
34dc7c2f BB |
2757 | |
2758 | if (mask & (AT_UID|AT_GID)) { | |
2759 | int idmask = (mask & (AT_UID|AT_GID)); | |
2760 | int take_owner; | |
2761 | int take_group; | |
2762 | ||
2763 | /* | |
2764 | * NOTE: even if a new mode is being set, | |
2765 | * we may clear S_ISUID/S_ISGID bits. | |
2766 | */ | |
2767 | ||
2768 | if (!(mask & AT_MODE)) | |
428870ff | 2769 | vap->va_mode = zp->z_mode; |
34dc7c2f BB |
2770 | |
2771 | /* | |
2772 | * Take ownership or chgrp to group we are a member of | |
2773 | */ | |
2774 | ||
2775 | take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr)); | |
2776 | take_group = (mask & AT_GID) && | |
2777 | zfs_groupmember(zfsvfs, vap->va_gid, cr); | |
2778 | ||
2779 | /* | |
2780 | * If both AT_UID and AT_GID are set then take_owner and | |
2781 | * take_group must both be set in order to allow taking | |
2782 | * ownership. | |
2783 | * | |
2784 | * Otherwise, send the check through secpolicy_vnode_setattr() | |
2785 | * | |
2786 | */ | |
2787 | ||
2788 | if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) || | |
2789 | ((idmask == AT_UID) && take_owner) || | |
2790 | ((idmask == AT_GID) && take_group)) { | |
2791 | if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, | |
2792 | skipaclchk, cr) == 0) { | |
2793 | /* | |
2794 | * Remove setuid/setgid for non-privileged users | |
2795 | */ | |
2796 | secpolicy_setid_clear(vap, cr); | |
2797 | trim_mask = (mask & (AT_UID|AT_GID)); | |
2798 | } else { | |
2799 | need_policy = TRUE; | |
2800 | } | |
2801 | } else { | |
2802 | need_policy = TRUE; | |
2803 | } | |
2804 | } | |
2805 | ||
2806 | mutex_enter(&zp->z_lock); | |
428870ff | 2807 | oldva.va_mode = zp->z_mode; |
572e2857 | 2808 | zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); |
34dc7c2f | 2809 | if (mask & AT_XVATTR) { |
fb5f0bc8 BB |
2810 | /* |
2811 | * Update xvattr mask to include only those attributes | |
2812 | * that are actually changing. | |
2813 | * | |
2814 | * the bits will be restored prior to actually setting | |
2815 | * the attributes so the caller thinks they were set. | |
2816 | */ | |
2817 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
2818 | if (xoap->xoa_appendonly != | |
428870ff | 2819 | ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { |
fb5f0bc8 BB |
2820 | need_policy = TRUE; |
2821 | } else { | |
2822 | XVA_CLR_REQ(xvap, XAT_APPENDONLY); | |
2823 | XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY); | |
2824 | } | |
2825 | } | |
2826 | ||
2827 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
2828 | if (xoap->xoa_nounlink != | |
428870ff | 2829 | ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { |
fb5f0bc8 BB |
2830 | need_policy = TRUE; |
2831 | } else { | |
2832 | XVA_CLR_REQ(xvap, XAT_NOUNLINK); | |
2833 | XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK); | |
2834 | } | |
2835 | } | |
2836 | ||
2837 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
2838 | if (xoap->xoa_immutable != | |
428870ff | 2839 | ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { |
fb5f0bc8 BB |
2840 | need_policy = TRUE; |
2841 | } else { | |
2842 | XVA_CLR_REQ(xvap, XAT_IMMUTABLE); | |
2843 | XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE); | |
2844 | } | |
2845 | } | |
2846 | ||
2847 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
2848 | if (xoap->xoa_nodump != | |
428870ff | 2849 | ((zp->z_pflags & ZFS_NODUMP) != 0)) { |
fb5f0bc8 BB |
2850 | need_policy = TRUE; |
2851 | } else { | |
2852 | XVA_CLR_REQ(xvap, XAT_NODUMP); | |
2853 | XVA_SET_REQ(&tmpxvattr, XAT_NODUMP); | |
2854 | } | |
2855 | } | |
2856 | ||
2857 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
2858 | if (xoap->xoa_av_modified != | |
428870ff | 2859 | ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { |
fb5f0bc8 BB |
2860 | need_policy = TRUE; |
2861 | } else { | |
2862 | XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); | |
2863 | XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED); | |
2864 | } | |
2865 | } | |
2866 | ||
2867 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
2868 | if ((vp->v_type != VREG && | |
2869 | xoap->xoa_av_quarantined) || | |
2870 | xoap->xoa_av_quarantined != | |
428870ff | 2871 | ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { |
fb5f0bc8 BB |
2872 | need_policy = TRUE; |
2873 | } else { | |
2874 | XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); | |
2875 | XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED); | |
2876 | } | |
2877 | } | |
2878 | ||
428870ff BB |
2879 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { |
2880 | mutex_exit(&zp->z_lock); | |
2881 | ZFS_EXIT(zfsvfs); | |
2882 | return (EPERM); | |
2883 | } | |
2884 | ||
fb5f0bc8 BB |
2885 | if (need_policy == FALSE && |
2886 | (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || | |
2887 | XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { | |
34dc7c2f BB |
2888 | need_policy = TRUE; |
2889 | } | |
2890 | } | |
2891 | ||
2892 | mutex_exit(&zp->z_lock); | |
2893 | ||
2894 | if (mask & AT_MODE) { | |
2895 | if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { | |
2896 | err = secpolicy_setid_setsticky_clear(vp, vap, | |
2897 | &oldva, cr); | |
2898 | if (err) { | |
2899 | ZFS_EXIT(zfsvfs); | |
2900 | return (err); | |
2901 | } | |
2902 | trim_mask |= AT_MODE; | |
2903 | } else { | |
2904 | need_policy = TRUE; | |
2905 | } | |
2906 | } | |
2907 | ||
2908 | if (need_policy) { | |
2909 | /* | |
2910 | * If trim_mask is set then take ownership | |
2911 | * has been granted or write_acl is present and user | |
2912 | * has the ability to modify mode. In that case remove | |
2913 | * UID|GID and or MODE from mask so that | |
2914 | * secpolicy_vnode_setattr() doesn't revoke it. | |
2915 | */ | |
2916 | ||
2917 | if (trim_mask) { | |
2918 | saved_mask = vap->va_mask; | |
2919 | vap->va_mask &= ~trim_mask; | |
2920 | } | |
2921 | err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags, | |
2922 | (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); | |
2923 | if (err) { | |
2924 | ZFS_EXIT(zfsvfs); | |
2925 | return (err); | |
2926 | } | |
2927 | ||
2928 | if (trim_mask) | |
2929 | vap->va_mask |= saved_mask; | |
2930 | } | |
2931 | ||
2932 | /* | |
2933 | * secpolicy_vnode_setattr, or take ownership may have | |
2934 | * changed va_mask | |
2935 | */ | |
2936 | mask = vap->va_mask; | |
2937 | ||
428870ff | 2938 | if ((mask & (AT_UID | AT_GID))) { |
572e2857 BB |
2939 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), |
2940 | &xattr_obj, sizeof (xattr_obj)); | |
428870ff | 2941 | |
572e2857 | 2942 | if (err == 0 && xattr_obj) { |
428870ff BB |
2943 | err = zfs_zget(zp->z_zfsvfs, xattr_obj, &attrzp); |
2944 | if (err) | |
2945 | goto out2; | |
2946 | } | |
2947 | if (mask & AT_UID) { | |
2948 | new_uid = zfs_fuid_create(zfsvfs, | |
2949 | (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); | |
572e2857 | 2950 | if (new_uid != zp->z_uid && |
428870ff | 2951 | zfs_fuid_overquota(zfsvfs, B_FALSE, new_uid)) { |
572e2857 BB |
2952 | if (attrzp) |
2953 | VN_RELE(ZTOV(attrzp)); | |
428870ff BB |
2954 | err = EDQUOT; |
2955 | goto out2; | |
2956 | } | |
2957 | } | |
2958 | ||
2959 | if (mask & AT_GID) { | |
2960 | new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid, | |
2961 | cr, ZFS_GROUP, &fuidp); | |
2962 | if (new_gid != zp->z_gid && | |
2963 | zfs_fuid_overquota(zfsvfs, B_TRUE, new_gid)) { | |
572e2857 BB |
2964 | if (attrzp) |
2965 | VN_RELE(ZTOV(attrzp)); | |
428870ff BB |
2966 | err = EDQUOT; |
2967 | goto out2; | |
2968 | } | |
2969 | } | |
2970 | } | |
34dc7c2f | 2971 | tx = dmu_tx_create(zfsvfs->z_os); |
34dc7c2f BB |
2972 | |
2973 | if (mask & AT_MODE) { | |
428870ff | 2974 | uint64_t pmode = zp->z_mode; |
572e2857 | 2975 | uint64_t acl_obj; |
34dc7c2f BB |
2976 | new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); |
2977 | ||
572e2857 | 2978 | zfs_acl_chmod_setattr(zp, &aclp, new_mode); |
428870ff | 2979 | |
572e2857 BB |
2980 | mutex_enter(&zp->z_lock); |
2981 | if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { | |
428870ff BB |
2982 | /* |
2983 | * Are we upgrading ACL from old V0 format | |
2984 | * to V1 format? | |
2985 | */ | |
572e2857 BB |
2986 | if (zfsvfs->z_version >= ZPL_VERSION_FUID && |
2987 | zfs_znode_acl_version(zp) == | |
34dc7c2f | 2988 | ZFS_ACL_VERSION_INITIAL) { |
572e2857 | 2989 | dmu_tx_hold_free(tx, acl_obj, 0, |
34dc7c2f BB |
2990 | DMU_OBJECT_END); |
2991 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, | |
2992 | 0, aclp->z_acl_bytes); | |
2993 | } else { | |
572e2857 | 2994 | dmu_tx_hold_write(tx, acl_obj, 0, |
34dc7c2f BB |
2995 | aclp->z_acl_bytes); |
2996 | } | |
428870ff | 2997 | } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { |
34dc7c2f BB |
2998 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, |
2999 | 0, aclp->z_acl_bytes); | |
3000 | } | |
572e2857 | 3001 | mutex_exit(&zp->z_lock); |
428870ff BB |
3002 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); |
3003 | } else { | |
3004 | if ((mask & AT_XVATTR) && | |
3005 | XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) | |
3006 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); | |
3007 | else | |
3008 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3009 | } |
3010 | ||
428870ff BB |
3011 | if (attrzp) { |
3012 | dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3013 | } |
3014 | ||
428870ff BB |
3015 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
3016 | if (fuid_dirtied) | |
3017 | zfs_fuid_txhold(zfsvfs, tx); | |
3018 | ||
3019 | zfs_sa_upgrade_txholds(tx, zp); | |
3020 | ||
fb5f0bc8 | 3021 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 3022 | if (err) { |
9babb374 | 3023 | if (err == ERESTART) |
34dc7c2f | 3024 | dmu_tx_wait(tx); |
9babb374 | 3025 | goto out; |
34dc7c2f BB |
3026 | } |
3027 | ||
428870ff | 3028 | count = 0; |
34dc7c2f BB |
3029 | /* |
3030 | * Set each attribute requested. | |
3031 | * We group settings according to the locks they need to acquire. | |
3032 | * | |
3033 | * Note: you cannot set ctime directly, although it will be | |
3034 | * updated as a side-effect of calling this function. | |
3035 | */ | |
3036 | ||
572e2857 BB |
3037 | |
3038 | if (mask & (AT_UID|AT_GID|AT_MODE)) | |
3039 | mutex_enter(&zp->z_acl_lock); | |
34dc7c2f BB |
3040 | mutex_enter(&zp->z_lock); |
3041 | ||
428870ff BB |
3042 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
3043 | &zp->z_pflags, sizeof (zp->z_pflags)); | |
3044 | ||
3045 | if (attrzp) { | |
572e2857 BB |
3046 | if (mask & (AT_UID|AT_GID|AT_MODE)) |
3047 | mutex_enter(&attrzp->z_acl_lock); | |
428870ff BB |
3048 | mutex_enter(&attrzp->z_lock); |
3049 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3050 | SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, | |
3051 | sizeof (attrzp->z_pflags)); | |
3052 | } | |
3053 | ||
3054 | if (mask & (AT_UID|AT_GID)) { | |
3055 | ||
3056 | if (mask & AT_UID) { | |
3057 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, | |
3058 | &new_uid, sizeof (new_uid)); | |
572e2857 | 3059 | zp->z_uid = new_uid; |
428870ff BB |
3060 | if (attrzp) { |
3061 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3062 | SA_ZPL_UID(zfsvfs), NULL, &new_uid, | |
3063 | sizeof (new_uid)); | |
572e2857 | 3064 | attrzp->z_uid = new_uid; |
428870ff BB |
3065 | } |
3066 | } | |
3067 | ||
3068 | if (mask & AT_GID) { | |
3069 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), | |
3070 | NULL, &new_gid, sizeof (new_gid)); | |
572e2857 | 3071 | zp->z_gid = new_gid; |
428870ff BB |
3072 | if (attrzp) { |
3073 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3074 | SA_ZPL_GID(zfsvfs), NULL, &new_gid, | |
3075 | sizeof (new_gid)); | |
572e2857 | 3076 | attrzp->z_gid = new_gid; |
428870ff BB |
3077 | } |
3078 | } | |
3079 | if (!(mask & AT_MODE)) { | |
3080 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), | |
3081 | NULL, &new_mode, sizeof (new_mode)); | |
3082 | new_mode = zp->z_mode; | |
3083 | } | |
3084 | err = zfs_acl_chown_setattr(zp); | |
3085 | ASSERT(err == 0); | |
3086 | if (attrzp) { | |
3087 | err = zfs_acl_chown_setattr(attrzp); | |
3088 | ASSERT(err == 0); | |
3089 | } | |
3090 | } | |
3091 | ||
34dc7c2f | 3092 | if (mask & AT_MODE) { |
428870ff BB |
3093 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
3094 | &new_mode, sizeof (new_mode)); | |
3095 | zp->z_mode = new_mode; | |
3096 | ASSERT3U((uintptr_t)aclp, !=, NULL); | |
9babb374 | 3097 | err = zfs_aclset_common(zp, aclp, cr, tx); |
34dc7c2f | 3098 | ASSERT3U(err, ==, 0); |
572e2857 BB |
3099 | if (zp->z_acl_cached) |
3100 | zfs_acl_free(zp->z_acl_cached); | |
45d1cae3 BB |
3101 | zp->z_acl_cached = aclp; |
3102 | aclp = NULL; | |
34dc7c2f BB |
3103 | } |
3104 | ||
34dc7c2f | 3105 | |
428870ff BB |
3106 | if (mask & AT_ATIME) { |
3107 | ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime); | |
3108 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, | |
3109 | &zp->z_atime, sizeof (zp->z_atime)); | |
34dc7c2f BB |
3110 | } |
3111 | ||
428870ff BB |
3112 | if (mask & AT_MTIME) { |
3113 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); | |
3114 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, | |
3115 | mtime, sizeof (mtime)); | |
34dc7c2f BB |
3116 | } |
3117 | ||
b128c09f | 3118 | /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */ |
428870ff BB |
3119 | if (mask & AT_SIZE && !(mask & AT_MTIME)) { |
3120 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), | |
3121 | NULL, mtime, sizeof (mtime)); | |
3122 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, | |
3123 | &ctime, sizeof (ctime)); | |
3124 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, | |
3125 | B_TRUE); | |
3126 | } else if (mask != 0) { | |
3127 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, | |
3128 | &ctime, sizeof (ctime)); | |
3129 | zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, | |
3130 | B_TRUE); | |
3131 | if (attrzp) { | |
3132 | SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, | |
3133 | SA_ZPL_CTIME(zfsvfs), NULL, | |
3134 | &ctime, sizeof (ctime)); | |
3135 | zfs_tstamp_update_setup(attrzp, STATE_CHANGED, | |
3136 | mtime, ctime, B_TRUE); | |
3137 | } | |
3138 | } | |
34dc7c2f BB |
3139 | /* |
3140 | * Do this after setting timestamps to prevent timestamp | |
3141 | * update from toggling bit | |
3142 | */ | |
3143 | ||
3144 | if (xoap && (mask & AT_XVATTR)) { | |
fb5f0bc8 BB |
3145 | |
3146 | /* | |
3147 | * restore trimmed off masks | |
3148 | * so that return masks can be set for caller. | |
3149 | */ | |
3150 | ||
3151 | if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) { | |
3152 | XVA_SET_REQ(xvap, XAT_APPENDONLY); | |
3153 | } | |
3154 | if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) { | |
3155 | XVA_SET_REQ(xvap, XAT_NOUNLINK); | |
3156 | } | |
3157 | if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) { | |
3158 | XVA_SET_REQ(xvap, XAT_IMMUTABLE); | |
3159 | } | |
3160 | if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) { | |
3161 | XVA_SET_REQ(xvap, XAT_NODUMP); | |
3162 | } | |
3163 | if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) { | |
3164 | XVA_SET_REQ(xvap, XAT_AV_MODIFIED); | |
3165 | } | |
3166 | if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) { | |
3167 | XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); | |
3168 | } | |
3169 | ||
428870ff | 3170 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) |
34dc7c2f BB |
3171 | ASSERT(vp->v_type == VREG); |
3172 | ||
428870ff | 3173 | zfs_xvattr_set(zp, xvap, tx); |
34dc7c2f BB |
3174 | } |
3175 | ||
9babb374 BB |
3176 | if (fuid_dirtied) |
3177 | zfs_fuid_sync(zfsvfs, tx); | |
3178 | ||
34dc7c2f BB |
3179 | if (mask != 0) |
3180 | zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); | |
3181 | ||
34dc7c2f | 3182 | mutex_exit(&zp->z_lock); |
572e2857 BB |
3183 | if (mask & (AT_UID|AT_GID|AT_MODE)) |
3184 | mutex_exit(&zp->z_acl_lock); | |
34dc7c2f | 3185 | |
572e2857 BB |
3186 | if (attrzp) { |
3187 | if (mask & (AT_UID|AT_GID|AT_MODE)) | |
3188 | mutex_exit(&attrzp->z_acl_lock); | |
3189 | mutex_exit(&attrzp->z_lock); | |
3190 | } | |
9babb374 | 3191 | out: |
428870ff BB |
3192 | if (err == 0 && attrzp) { |
3193 | err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, | |
3194 | xattr_count, tx); | |
3195 | ASSERT(err2 == 0); | |
3196 | } | |
3197 | ||
34dc7c2f BB |
3198 | if (attrzp) |
3199 | VN_RELE(ZTOV(attrzp)); | |
45d1cae3 | 3200 | if (aclp) |
9babb374 | 3201 | zfs_acl_free(aclp); |
9babb374 BB |
3202 | |
3203 | if (fuidp) { | |
3204 | zfs_fuid_info_free(fuidp); | |
3205 | fuidp = NULL; | |
3206 | } | |
3207 | ||
428870ff | 3208 | if (err) { |
9babb374 | 3209 | dmu_tx_abort(tx); |
428870ff BB |
3210 | if (err == ERESTART) |
3211 | goto top; | |
3212 | } else { | |
3213 | err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
9babb374 | 3214 | dmu_tx_commit(tx); |
428870ff BB |
3215 | } |
3216 | ||
428870ff BB |
3217 | out2: |
3218 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) | |
572e2857 | 3219 | zil_commit(zilog, 0); |
34dc7c2f BB |
3220 | |
3221 | ZFS_EXIT(zfsvfs); | |
3222 | return (err); | |
3223 | } | |
3224 | ||
3225 | typedef struct zfs_zlock { | |
3226 | krwlock_t *zl_rwlock; /* lock we acquired */ | |
3227 | znode_t *zl_znode; /* znode we held */ | |
3228 | struct zfs_zlock *zl_next; /* next in list */ | |
3229 | } zfs_zlock_t; | |
3230 | ||
3231 | /* | |
3232 | * Drop locks and release vnodes that were held by zfs_rename_lock(). | |
3233 | */ | |
3234 | static void | |
3235 | zfs_rename_unlock(zfs_zlock_t **zlpp) | |
3236 | { | |
3237 | zfs_zlock_t *zl; | |
3238 | ||
3239 | while ((zl = *zlpp) != NULL) { | |
3240 | if (zl->zl_znode != NULL) | |
3241 | VN_RELE(ZTOV(zl->zl_znode)); | |
3242 | rw_exit(zl->zl_rwlock); | |
3243 | *zlpp = zl->zl_next; | |
3244 | kmem_free(zl, sizeof (*zl)); | |
3245 | } | |
3246 | } | |
3247 | ||
3248 | /* | |
3249 | * Search back through the directory tree, using the ".." entries. | |
3250 | * Lock each directory in the chain to prevent concurrent renames. | |
3251 | * Fail any attempt to move a directory into one of its own descendants. | |
3252 | * XXX - z_parent_lock can overlap with map or grow locks | |
3253 | */ | |
3254 | static int | |
3255 | zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) | |
3256 | { | |
3257 | zfs_zlock_t *zl; | |
3258 | znode_t *zp = tdzp; | |
3259 | uint64_t rootid = zp->z_zfsvfs->z_root; | |
428870ff | 3260 | uint64_t oidp = zp->z_id; |
34dc7c2f BB |
3261 | krwlock_t *rwlp = &szp->z_parent_lock; |
3262 | krw_t rw = RW_WRITER; | |
3263 | ||
3264 | /* | |
3265 | * First pass write-locks szp and compares to zp->z_id. | |
3266 | * Later passes read-lock zp and compare to zp->z_parent. | |
3267 | */ | |
3268 | do { | |
3269 | if (!rw_tryenter(rwlp, rw)) { | |
3270 | /* | |
3271 | * Another thread is renaming in this path. | |
3272 | * Note that if we are a WRITER, we don't have any | |
3273 | * parent_locks held yet. | |
3274 | */ | |
3275 | if (rw == RW_READER && zp->z_id > szp->z_id) { | |
3276 | /* | |
3277 | * Drop our locks and restart | |
3278 | */ | |
3279 | zfs_rename_unlock(&zl); | |
3280 | *zlpp = NULL; | |
3281 | zp = tdzp; | |
428870ff | 3282 | oidp = zp->z_id; |
34dc7c2f BB |
3283 | rwlp = &szp->z_parent_lock; |
3284 | rw = RW_WRITER; | |
3285 | continue; | |
3286 | } else { | |
3287 | /* | |
3288 | * Wait for other thread to drop its locks | |
3289 | */ | |
3290 | rw_enter(rwlp, rw); | |
3291 | } | |
3292 | } | |
3293 | ||
3294 | zl = kmem_alloc(sizeof (*zl), KM_SLEEP); | |
3295 | zl->zl_rwlock = rwlp; | |
3296 | zl->zl_znode = NULL; | |
3297 | zl->zl_next = *zlpp; | |
3298 | *zlpp = zl; | |
3299 | ||
428870ff | 3300 | if (oidp == szp->z_id) /* We're a descendant of szp */ |
34dc7c2f BB |
3301 | return (EINVAL); |
3302 | ||
428870ff | 3303 | if (oidp == rootid) /* We've hit the top */ |
34dc7c2f BB |
3304 | return (0); |
3305 | ||
3306 | if (rw == RW_READER) { /* i.e. not the first pass */ | |
428870ff | 3307 | int error = zfs_zget(zp->z_zfsvfs, oidp, &zp); |
34dc7c2f BB |
3308 | if (error) |
3309 | return (error); | |
3310 | zl->zl_znode = zp; | |
3311 | } | |
428870ff BB |
3312 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zp->z_zfsvfs), |
3313 | &oidp, sizeof (oidp)); | |
34dc7c2f BB |
3314 | rwlp = &zp->z_parent_lock; |
3315 | rw = RW_READER; | |
3316 | ||
3317 | } while (zp->z_id != sdzp->z_id); | |
3318 | ||
3319 | return (0); | |
3320 | } | |
3321 | ||
3322 | /* | |
3323 | * Move an entry from the provided source directory to the target | |
3324 | * directory. Change the entry name as indicated. | |
3325 | * | |
3326 | * IN: sdvp - Source directory containing the "old entry". | |
3327 | * snm - Old entry name. | |
3328 | * tdvp - Target directory to contain the "new entry". | |
3329 | * tnm - New entry name. | |
3330 | * cr - credentials of caller. | |
3331 | * ct - caller context | |
3332 | * flags - case flags | |
3333 | * | |
3334 | * RETURN: 0 if success | |
3335 | * error code if failure | |
3336 | * | |
3337 | * Timestamps: | |
3338 | * sdvp,tdvp - ctime|mtime updated | |
3339 | */ | |
3340 | /*ARGSUSED*/ | |
3341 | static int | |
3342 | zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr, | |
3343 | caller_context_t *ct, int flags) | |
3344 | { | |
3345 | znode_t *tdzp, *szp, *tzp; | |
3346 | znode_t *sdzp = VTOZ(sdvp); | |
3347 | zfsvfs_t *zfsvfs = sdzp->z_zfsvfs; | |
3348 | zilog_t *zilog; | |
3349 | vnode_t *realvp; | |
3350 | zfs_dirlock_t *sdl, *tdl; | |
3351 | dmu_tx_t *tx; | |
3352 | zfs_zlock_t *zl; | |
3353 | int cmp, serr, terr; | |
3354 | int error = 0; | |
3355 | int zflg = 0; | |
3356 | ||
3357 | ZFS_ENTER(zfsvfs); | |
3358 | ZFS_VERIFY_ZP(sdzp); | |
3359 | zilog = zfsvfs->z_log; | |
3360 | ||
3361 | /* | |
3362 | * Make sure we have the real vp for the target directory. | |
3363 | */ | |
3364 | if (VOP_REALVP(tdvp, &realvp, ct) == 0) | |
3365 | tdvp = realvp; | |
3366 | ||
428870ff | 3367 | if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) { |
34dc7c2f BB |
3368 | ZFS_EXIT(zfsvfs); |
3369 | return (EXDEV); | |
3370 | } | |
3371 | ||
3372 | tdzp = VTOZ(tdvp); | |
3373 | ZFS_VERIFY_ZP(tdzp); | |
3374 | if (zfsvfs->z_utf8 && u8_validate(tnm, | |
3375 | strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { | |
3376 | ZFS_EXIT(zfsvfs); | |
3377 | return (EILSEQ); | |
3378 | } | |
3379 | ||
3380 | if (flags & FIGNORECASE) | |
3381 | zflg |= ZCILOOK; | |
3382 | ||
3383 | top: | |
3384 | szp = NULL; | |
3385 | tzp = NULL; | |
3386 | zl = NULL; | |
3387 | ||
3388 | /* | |
3389 | * This is to prevent the creation of links into attribute space | |
3390 | * by renaming a linked file into/outof an attribute directory. | |
3391 | * See the comment in zfs_link() for why this is considered bad. | |
3392 | */ | |
428870ff | 3393 | if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { |
34dc7c2f BB |
3394 | ZFS_EXIT(zfsvfs); |
3395 | return (EINVAL); | |
3396 | } | |
3397 | ||
3398 | /* | |
3399 | * Lock source and target directory entries. To prevent deadlock, | |
3400 | * a lock ordering must be defined. We lock the directory with | |
3401 | * the smallest object id first, or if it's a tie, the one with | |
3402 | * the lexically first name. | |
3403 | */ | |
3404 | if (sdzp->z_id < tdzp->z_id) { | |
3405 | cmp = -1; | |
3406 | } else if (sdzp->z_id > tdzp->z_id) { | |
3407 | cmp = 1; | |
3408 | } else { | |
3409 | /* | |
3410 | * First compare the two name arguments without | |
3411 | * considering any case folding. | |
3412 | */ | |
3413 | int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); | |
3414 | ||
3415 | cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); | |
3416 | ASSERT(error == 0 || !zfsvfs->z_utf8); | |
3417 | if (cmp == 0) { | |
3418 | /* | |
3419 | * POSIX: "If the old argument and the new argument | |
3420 | * both refer to links to the same existing file, | |
3421 | * the rename() function shall return successfully | |
3422 | * and perform no other action." | |
3423 | */ | |
3424 | ZFS_EXIT(zfsvfs); | |
3425 | return (0); | |
3426 | } | |
3427 | /* | |
3428 | * If the file system is case-folding, then we may | |
3429 | * have some more checking to do. A case-folding file | |
3430 | * system is either supporting mixed case sensitivity | |
3431 | * access or is completely case-insensitive. Note | |
3432 | * that the file system is always case preserving. | |
3433 | * | |
3434 | * In mixed sensitivity mode case sensitive behavior | |
3435 | * is the default. FIGNORECASE must be used to | |
3436 | * explicitly request case insensitive behavior. | |
3437 | * | |
3438 | * If the source and target names provided differ only | |
3439 | * by case (e.g., a request to rename 'tim' to 'Tim'), | |
3440 | * we will treat this as a special case in the | |
3441 | * case-insensitive mode: as long as the source name | |
3442 | * is an exact match, we will allow this to proceed as | |
3443 | * a name-change request. | |
3444 | */ | |
3445 | if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || | |
3446 | (zfsvfs->z_case == ZFS_CASE_MIXED && | |
3447 | flags & FIGNORECASE)) && | |
3448 | u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, | |
3449 | &error) == 0) { | |
3450 | /* | |
3451 | * case preserving rename request, require exact | |
3452 | * name matches | |
3453 | */ | |
3454 | zflg |= ZCIEXACT; | |
3455 | zflg &= ~ZCILOOK; | |
3456 | } | |
3457 | } | |
3458 | ||
428870ff BB |
3459 | /* |
3460 | * If the source and destination directories are the same, we should | |
3461 | * grab the z_name_lock of that directory only once. | |
3462 | */ | |
3463 | if (sdzp == tdzp) { | |
3464 | zflg |= ZHAVELOCK; | |
3465 | rw_enter(&sdzp->z_name_lock, RW_READER); | |
3466 | } | |
3467 | ||
34dc7c2f BB |
3468 | if (cmp < 0) { |
3469 | serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, | |
3470 | ZEXISTS | zflg, NULL, NULL); | |
3471 | terr = zfs_dirent_lock(&tdl, | |
3472 | tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); | |
3473 | } else { | |
3474 | terr = zfs_dirent_lock(&tdl, | |
3475 | tdzp, tnm, &tzp, zflg, NULL, NULL); | |
3476 | serr = zfs_dirent_lock(&sdl, | |
3477 | sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, | |
3478 | NULL, NULL); | |
3479 | } | |
3480 | ||
3481 | if (serr) { | |
3482 | /* | |
3483 | * Source entry invalid or not there. | |
3484 | */ | |
3485 | if (!terr) { | |
3486 | zfs_dirent_unlock(tdl); | |
3487 | if (tzp) | |
3488 | VN_RELE(ZTOV(tzp)); | |
3489 | } | |
428870ff BB |
3490 | |
3491 | if (sdzp == tdzp) | |
3492 | rw_exit(&sdzp->z_name_lock); | |
3493 | ||
34dc7c2f BB |
3494 | if (strcmp(snm, "..") == 0) |
3495 | serr = EINVAL; | |
3496 | ZFS_EXIT(zfsvfs); | |
3497 | return (serr); | |
3498 | } | |
3499 | if (terr) { | |
3500 | zfs_dirent_unlock(sdl); | |
3501 | VN_RELE(ZTOV(szp)); | |
428870ff BB |
3502 | |
3503 | if (sdzp == tdzp) | |
3504 | rw_exit(&sdzp->z_name_lock); | |
3505 | ||
34dc7c2f BB |
3506 | if (strcmp(tnm, "..") == 0) |
3507 | terr = EINVAL; | |
3508 | ZFS_EXIT(zfsvfs); | |
3509 | return (terr); | |
3510 | } | |
3511 | ||
3512 | /* | |
3513 | * Must have write access at the source to remove the old entry | |
3514 | * and write access at the target to create the new entry. | |
3515 | * Note that if target and source are the same, this can be | |
3516 | * done in a single check. | |
3517 | */ | |
3518 | ||
3519 | if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)) | |
3520 | goto out; | |
3521 | ||
3522 | if (ZTOV(szp)->v_type == VDIR) { | |
3523 | /* | |
3524 | * Check to make sure rename is valid. | |
3525 | * Can't do a move like this: /usr/a/b to /usr/a/b/c/d | |
3526 | */ | |
3527 | if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl)) | |
3528 | goto out; | |
3529 | } | |
3530 | ||
3531 | /* | |
3532 | * Does target exist? | |
3533 | */ | |
3534 | if (tzp) { | |
3535 | /* | |
3536 | * Source and target must be the same type. | |
3537 | */ | |
3538 | if (ZTOV(szp)->v_type == VDIR) { | |
3539 | if (ZTOV(tzp)->v_type != VDIR) { | |
3540 | error = ENOTDIR; | |
3541 | goto out; | |
3542 | } | |
3543 | } else { | |
3544 | if (ZTOV(tzp)->v_type == VDIR) { | |
3545 | error = EISDIR; | |
3546 | goto out; | |
3547 | } | |
3548 | } | |
3549 | /* | |
3550 | * POSIX dictates that when the source and target | |
3551 | * entries refer to the same file object, rename | |
3552 | * must do nothing and exit without error. | |
3553 | */ | |
3554 | if (szp->z_id == tzp->z_id) { | |
3555 | error = 0; | |
3556 | goto out; | |
3557 | } | |
3558 | } | |
3559 | ||
3560 | vnevent_rename_src(ZTOV(szp), sdvp, snm, ct); | |
3561 | if (tzp) | |
3562 | vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct); | |
3563 | ||
3564 | /* | |
3565 | * notify the target directory if it is not the same | |
3566 | * as source directory. | |
3567 | */ | |
3568 | if (tdvp != sdvp) { | |
3569 | vnevent_rename_dest_dir(tdvp, ct); | |
3570 | } | |
3571 | ||
3572 | tx = dmu_tx_create(zfsvfs->z_os); | |
428870ff BB |
3573 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
3574 | dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); | |
34dc7c2f BB |
3575 | dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); |
3576 | dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); | |
428870ff BB |
3577 | if (sdzp != tdzp) { |
3578 | dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); | |
3579 | zfs_sa_upgrade_txholds(tx, tdzp); | |
3580 | } | |
3581 | if (tzp) { | |
3582 | dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); | |
3583 | zfs_sa_upgrade_txholds(tx, tzp); | |
3584 | } | |
3585 | ||
3586 | zfs_sa_upgrade_txholds(tx, szp); | |
34dc7c2f | 3587 | dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); |
fb5f0bc8 | 3588 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
3589 | if (error) { |
3590 | if (zl != NULL) | |
3591 | zfs_rename_unlock(&zl); | |
3592 | zfs_dirent_unlock(sdl); | |
3593 | zfs_dirent_unlock(tdl); | |
428870ff BB |
3594 | |
3595 | if (sdzp == tdzp) | |
3596 | rw_exit(&sdzp->z_name_lock); | |
3597 | ||
34dc7c2f BB |
3598 | VN_RELE(ZTOV(szp)); |
3599 | if (tzp) | |
3600 | VN_RELE(ZTOV(tzp)); | |
fb5f0bc8 | 3601 | if (error == ERESTART) { |
34dc7c2f BB |
3602 | dmu_tx_wait(tx); |
3603 | dmu_tx_abort(tx); | |
3604 | goto top; | |
3605 | } | |
3606 | dmu_tx_abort(tx); | |
3607 | ZFS_EXIT(zfsvfs); | |
3608 | return (error); | |
3609 | } | |
3610 | ||
3611 | if (tzp) /* Attempt to remove the existing target */ | |
3612 | error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); | |
3613 | ||
3614 | if (error == 0) { | |
3615 | error = zfs_link_create(tdl, szp, tx, ZRENAMING); | |
3616 | if (error == 0) { | |
428870ff | 3617 | szp->z_pflags |= ZFS_AV_MODIFIED; |
34dc7c2f | 3618 | |
428870ff BB |
3619 | error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), |
3620 | (void *)&szp->z_pflags, sizeof (uint64_t), tx); | |
3621 | ASSERT3U(error, ==, 0); | |
34dc7c2f | 3622 | |
428870ff BB |
3623 | error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); |
3624 | if (error == 0) { | |
3625 | zfs_log_rename(zilog, tx, TX_RENAME | | |
572e2857 BB |
3626 | (flags & FIGNORECASE ? TX_CI : 0), sdzp, |
3627 | sdl->dl_name, tdzp, tdl->dl_name, szp); | |
b128c09f | 3628 | |
428870ff BB |
3629 | /* |
3630 | * Update path information for the target vnode | |
3631 | */ | |
3632 | vn_renamepath(tdvp, ZTOV(szp), tnm, | |
3633 | strlen(tnm)); | |
3634 | } else { | |
3635 | /* | |
3636 | * At this point, we have successfully created | |
3637 | * the target name, but have failed to remove | |
3638 | * the source name. Since the create was done | |
3639 | * with the ZRENAMING flag, there are | |
3640 | * complications; for one, the link count is | |
3641 | * wrong. The easiest way to deal with this | |
3642 | * is to remove the newly created target, and | |
3643 | * return the original error. This must | |
3644 | * succeed; fortunately, it is very unlikely to | |
3645 | * fail, since we just created it. | |
3646 | */ | |
3647 | VERIFY3U(zfs_link_destroy(tdl, szp, tx, | |
3648 | ZRENAMING, NULL), ==, 0); | |
3649 | } | |
34dc7c2f BB |
3650 | } |
3651 | } | |
3652 | ||
3653 | dmu_tx_commit(tx); | |
3654 | out: | |
3655 | if (zl != NULL) | |
3656 | zfs_rename_unlock(&zl); | |
3657 | ||
3658 | zfs_dirent_unlock(sdl); | |
3659 | zfs_dirent_unlock(tdl); | |
3660 | ||
428870ff BB |
3661 | if (sdzp == tdzp) |
3662 | rw_exit(&sdzp->z_name_lock); | |
3663 | ||
3664 | ||
34dc7c2f BB |
3665 | VN_RELE(ZTOV(szp)); |
3666 | if (tzp) | |
3667 | VN_RELE(ZTOV(tzp)); | |
3668 | ||
428870ff | 3669 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3670 | zil_commit(zilog, 0); |
428870ff | 3671 | |
34dc7c2f BB |
3672 | ZFS_EXIT(zfsvfs); |
3673 | return (error); | |
3674 | } | |
3675 | ||
3676 | /* | |
3677 | * Insert the indicated symbolic reference entry into the directory. | |
3678 | * | |
3679 | * IN: dvp - Directory to contain new symbolic link. | |
3680 | * link - Name for new symlink entry. | |
3681 | * vap - Attributes of new entry. | |
3682 | * target - Target path of new symlink. | |
3683 | * cr - credentials of caller. | |
3684 | * ct - caller context | |
3685 | * flags - case flags | |
3686 | * | |
3687 | * RETURN: 0 if success | |
3688 | * error code if failure | |
3689 | * | |
3690 | * Timestamps: | |
3691 | * dvp - ctime|mtime updated | |
3692 | */ | |
3693 | /*ARGSUSED*/ | |
3694 | static int | |
3695 | zfs_symlink(vnode_t *dvp, char *name, vattr_t *vap, char *link, cred_t *cr, | |
3696 | caller_context_t *ct, int flags) | |
3697 | { | |
3698 | znode_t *zp, *dzp = VTOZ(dvp); | |
3699 | zfs_dirlock_t *dl; | |
3700 | dmu_tx_t *tx; | |
3701 | zfsvfs_t *zfsvfs = dzp->z_zfsvfs; | |
3702 | zilog_t *zilog; | |
428870ff | 3703 | uint64_t len = strlen(link); |
34dc7c2f BB |
3704 | int error; |
3705 | int zflg = ZNEW; | |
9babb374 BB |
3706 | zfs_acl_ids_t acl_ids; |
3707 | boolean_t fuid_dirtied; | |
428870ff | 3708 | uint64_t txtype = TX_SYMLINK; |
34dc7c2f BB |
3709 | |
3710 | ASSERT(vap->va_type == VLNK); | |
3711 | ||
3712 | ZFS_ENTER(zfsvfs); | |
3713 | ZFS_VERIFY_ZP(dzp); | |
3714 | zilog = zfsvfs->z_log; | |
3715 | ||
3716 | if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), | |
3717 | NULL, U8_VALIDATE_ENTIRE, &error) < 0) { | |
3718 | ZFS_EXIT(zfsvfs); | |
3719 | return (EILSEQ); | |
3720 | } | |
3721 | if (flags & FIGNORECASE) | |
3722 | zflg |= ZCILOOK; | |
34dc7c2f BB |
3723 | |
3724 | if (len > MAXPATHLEN) { | |
3725 | ZFS_EXIT(zfsvfs); | |
3726 | return (ENAMETOOLONG); | |
3727 | } | |
3728 | ||
428870ff BB |
3729 | if ((error = zfs_acl_ids_create(dzp, 0, |
3730 | vap, cr, NULL, &acl_ids)) != 0) { | |
3731 | ZFS_EXIT(zfsvfs); | |
3732 | return (error); | |
3733 | } | |
3734 | top: | |
34dc7c2f BB |
3735 | /* |
3736 | * Attempt to lock directory; fail if entry already exists. | |
3737 | */ | |
3738 | error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); | |
3739 | if (error) { | |
428870ff BB |
3740 | zfs_acl_ids_free(&acl_ids); |
3741 | ZFS_EXIT(zfsvfs); | |
3742 | return (error); | |
3743 | } | |
3744 | ||
3745 | if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { | |
3746 | zfs_acl_ids_free(&acl_ids); | |
3747 | zfs_dirent_unlock(dl); | |
34dc7c2f BB |
3748 | ZFS_EXIT(zfsvfs); |
3749 | return (error); | |
3750 | } | |
3751 | ||
9babb374 BB |
3752 | if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { |
3753 | zfs_acl_ids_free(&acl_ids); | |
3754 | zfs_dirent_unlock(dl); | |
3755 | ZFS_EXIT(zfsvfs); | |
3756 | return (EDQUOT); | |
3757 | } | |
34dc7c2f | 3758 | tx = dmu_tx_create(zfsvfs->z_os); |
9babb374 | 3759 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
34dc7c2f | 3760 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); |
34dc7c2f | 3761 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3762 | dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + |
3763 | ZFS_SA_BASE_ATTR_SIZE + len); | |
3764 | dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); | |
3765 | if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { | |
3766 | dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, | |
3767 | acl_ids.z_aclp->z_acl_bytes); | |
3768 | } | |
9babb374 BB |
3769 | if (fuid_dirtied) |
3770 | zfs_fuid_txhold(zfsvfs, tx); | |
fb5f0bc8 | 3771 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
3772 | if (error) { |
3773 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3774 | if (error == ERESTART) { |
34dc7c2f BB |
3775 | dmu_tx_wait(tx); |
3776 | dmu_tx_abort(tx); | |
3777 | goto top; | |
3778 | } | |
428870ff | 3779 | zfs_acl_ids_free(&acl_ids); |
34dc7c2f BB |
3780 | dmu_tx_abort(tx); |
3781 | ZFS_EXIT(zfsvfs); | |
3782 | return (error); | |
3783 | } | |
3784 | ||
34dc7c2f BB |
3785 | /* |
3786 | * Create a new object for the symlink. | |
428870ff | 3787 | * for version 4 ZPL datsets the symlink will be an SA attribute |
34dc7c2f | 3788 | */ |
428870ff | 3789 | zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); |
9babb374 | 3790 | |
428870ff BB |
3791 | if (fuid_dirtied) |
3792 | zfs_fuid_sync(zfsvfs, tx); | |
34dc7c2f | 3793 | |
572e2857 | 3794 | mutex_enter(&zp->z_lock); |
428870ff BB |
3795 | if (zp->z_is_sa) |
3796 | error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), | |
3797 | link, len, tx); | |
3798 | else | |
3799 | zfs_sa_symlink(zp, link, len, tx); | |
572e2857 | 3800 | mutex_exit(&zp->z_lock); |
34dc7c2f | 3801 | |
428870ff BB |
3802 | zp->z_size = len; |
3803 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), | |
3804 | &zp->z_size, sizeof (zp->z_size), tx); | |
34dc7c2f BB |
3805 | /* |
3806 | * Insert the new object into the directory. | |
3807 | */ | |
3808 | (void) zfs_link_create(dl, zp, tx, ZNEW); | |
428870ff BB |
3809 | |
3810 | if (flags & FIGNORECASE) | |
3811 | txtype |= TX_CI; | |
3812 | zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); | |
9babb374 BB |
3813 | |
3814 | zfs_acl_ids_free(&acl_ids); | |
34dc7c2f BB |
3815 | |
3816 | dmu_tx_commit(tx); | |
3817 | ||
3818 | zfs_dirent_unlock(dl); | |
3819 | ||
3820 | VN_RELE(ZTOV(zp)); | |
3821 | ||
428870ff | 3822 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 3823 | zil_commit(zilog, 0); |
428870ff | 3824 | |
34dc7c2f BB |
3825 | ZFS_EXIT(zfsvfs); |
3826 | return (error); | |
3827 | } | |
3828 | ||
3829 | /* | |
3830 | * Return, in the buffer contained in the provided uio structure, | |
3831 | * the symbolic path referred to by vp. | |
3832 | * | |
3833 | * IN: vp - vnode of symbolic link. | |
3834 | * uoip - structure to contain the link path. | |
3835 | * cr - credentials of caller. | |
3836 | * ct - caller context | |
3837 | * | |
3838 | * OUT: uio - structure to contain the link path. | |
3839 | * | |
3840 | * RETURN: 0 if success | |
3841 | * error code if failure | |
3842 | * | |
3843 | * Timestamps: | |
3844 | * vp - atime updated | |
3845 | */ | |
3846 | /* ARGSUSED */ | |
3847 | static int | |
3848 | zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct) | |
3849 | { | |
3850 | znode_t *zp = VTOZ(vp); | |
3851 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
34dc7c2f BB |
3852 | int error; |
3853 | ||
3854 | ZFS_ENTER(zfsvfs); | |
3855 | ZFS_VERIFY_ZP(zp); | |
3856 | ||
572e2857 | 3857 | mutex_enter(&zp->z_lock); |
428870ff BB |
3858 | if (zp->z_is_sa) |
3859 | error = sa_lookup_uio(zp->z_sa_hdl, | |
3860 | SA_ZPL_SYMLINK(zfsvfs), uio); | |
3861 | else | |
3862 | error = zfs_sa_readlink(zp, uio); | |
572e2857 | 3863 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
3864 | |
3865 | ZFS_ACCESSTIME_STAMP(zfsvfs, zp); | |
428870ff | 3866 | |
34dc7c2f BB |
3867 | ZFS_EXIT(zfsvfs); |
3868 | return (error); | |
3869 | } | |
3870 | ||
3871 | /* | |
3872 | * Insert a new entry into directory tdvp referencing svp. | |
3873 | * | |
3874 | * IN: tdvp - Directory to contain new entry. | |
3875 | * svp - vnode of new entry. | |
3876 | * name - name of new entry. | |
3877 | * cr - credentials of caller. | |
3878 | * ct - caller context | |
3879 | * | |
3880 | * RETURN: 0 if success | |
3881 | * error code if failure | |
3882 | * | |
3883 | * Timestamps: | |
3884 | * tdvp - ctime|mtime updated | |
3885 | * svp - ctime updated | |
3886 | */ | |
3887 | /* ARGSUSED */ | |
3888 | static int | |
3889 | zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr, | |
3890 | caller_context_t *ct, int flags) | |
3891 | { | |
3892 | znode_t *dzp = VTOZ(tdvp); | |
3893 | znode_t *tzp, *szp; | |
3894 | zfsvfs_t *zfsvfs = dzp->z_zfsvfs; | |
3895 | zilog_t *zilog; | |
3896 | zfs_dirlock_t *dl; | |
3897 | dmu_tx_t *tx; | |
3898 | vnode_t *realvp; | |
3899 | int error; | |
3900 | int zf = ZNEW; | |
428870ff | 3901 | uint64_t parent; |
572e2857 | 3902 | uid_t owner; |
34dc7c2f BB |
3903 | |
3904 | ASSERT(tdvp->v_type == VDIR); | |
3905 | ||
3906 | ZFS_ENTER(zfsvfs); | |
3907 | ZFS_VERIFY_ZP(dzp); | |
3908 | zilog = zfsvfs->z_log; | |
3909 | ||
3910 | if (VOP_REALVP(svp, &realvp, ct) == 0) | |
3911 | svp = realvp; | |
3912 | ||
428870ff BB |
3913 | /* |
3914 | * POSIX dictates that we return EPERM here. | |
3915 | * Better choices include ENOTSUP or EISDIR. | |
3916 | */ | |
3917 | if (svp->v_type == VDIR) { | |
3918 | ZFS_EXIT(zfsvfs); | |
3919 | return (EPERM); | |
3920 | } | |
3921 | ||
3922 | if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) { | |
34dc7c2f BB |
3923 | ZFS_EXIT(zfsvfs); |
3924 | return (EXDEV); | |
3925 | } | |
428870ff | 3926 | |
34dc7c2f BB |
3927 | szp = VTOZ(svp); |
3928 | ZFS_VERIFY_ZP(szp); | |
3929 | ||
428870ff BB |
3930 | /* Prevent links to .zfs/shares files */ |
3931 | ||
3932 | if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), | |
3933 | &parent, sizeof (uint64_t))) != 0) { | |
3934 | ZFS_EXIT(zfsvfs); | |
3935 | return (error); | |
3936 | } | |
3937 | if (parent == zfsvfs->z_shares_dir) { | |
3938 | ZFS_EXIT(zfsvfs); | |
3939 | return (EPERM); | |
3940 | } | |
3941 | ||
34dc7c2f BB |
3942 | if (zfsvfs->z_utf8 && u8_validate(name, |
3943 | strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { | |
3944 | ZFS_EXIT(zfsvfs); | |
3945 | return (EILSEQ); | |
3946 | } | |
3947 | if (flags & FIGNORECASE) | |
3948 | zf |= ZCILOOK; | |
3949 | ||
34dc7c2f BB |
3950 | /* |
3951 | * We do not support links between attributes and non-attributes | |
3952 | * because of the potential security risk of creating links | |
3953 | * into "normal" file space in order to circumvent restrictions | |
3954 | * imposed in attribute space. | |
3955 | */ | |
428870ff | 3956 | if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { |
34dc7c2f BB |
3957 | ZFS_EXIT(zfsvfs); |
3958 | return (EINVAL); | |
3959 | } | |
3960 | ||
34dc7c2f | 3961 | |
572e2857 BB |
3962 | owner = zfs_fuid_map_id(zfsvfs, szp->z_uid, cr, ZFS_OWNER); |
3963 | if (owner != crgetuid(cr) && secpolicy_basic_link(cr) != 0) { | |
34dc7c2f BB |
3964 | ZFS_EXIT(zfsvfs); |
3965 | return (EPERM); | |
3966 | } | |
3967 | ||
3968 | if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { | |
3969 | ZFS_EXIT(zfsvfs); | |
3970 | return (error); | |
3971 | } | |
3972 | ||
428870ff | 3973 | top: |
34dc7c2f BB |
3974 | /* |
3975 | * Attempt to lock directory; fail if entry already exists. | |
3976 | */ | |
3977 | error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); | |
3978 | if (error) { | |
3979 | ZFS_EXIT(zfsvfs); | |
3980 | return (error); | |
3981 | } | |
3982 | ||
3983 | tx = dmu_tx_create(zfsvfs->z_os); | |
428870ff | 3984 | dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); |
34dc7c2f | 3985 | dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); |
428870ff BB |
3986 | zfs_sa_upgrade_txholds(tx, szp); |
3987 | zfs_sa_upgrade_txholds(tx, dzp); | |
fb5f0bc8 | 3988 | error = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f BB |
3989 | if (error) { |
3990 | zfs_dirent_unlock(dl); | |
fb5f0bc8 | 3991 | if (error == ERESTART) { |
34dc7c2f BB |
3992 | dmu_tx_wait(tx); |
3993 | dmu_tx_abort(tx); | |
3994 | goto top; | |
3995 | } | |
3996 | dmu_tx_abort(tx); | |
3997 | ZFS_EXIT(zfsvfs); | |
3998 | return (error); | |
3999 | } | |
4000 | ||
4001 | error = zfs_link_create(dl, szp, tx, 0); | |
4002 | ||
4003 | if (error == 0) { | |
4004 | uint64_t txtype = TX_LINK; | |
4005 | if (flags & FIGNORECASE) | |
4006 | txtype |= TX_CI; | |
4007 | zfs_log_link(zilog, tx, txtype, dzp, szp, name); | |
4008 | } | |
4009 | ||
4010 | dmu_tx_commit(tx); | |
4011 | ||
4012 | zfs_dirent_unlock(dl); | |
4013 | ||
4014 | if (error == 0) { | |
4015 | vnevent_link(svp, ct); | |
4016 | } | |
4017 | ||
428870ff | 4018 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4019 | zil_commit(zilog, 0); |
428870ff | 4020 | |
34dc7c2f BB |
4021 | ZFS_EXIT(zfsvfs); |
4022 | return (error); | |
4023 | } | |
4024 | ||
4025 | /* | |
4026 | * zfs_null_putapage() is used when the file system has been force | |
4027 | * unmounted. It just drops the pages. | |
4028 | */ | |
4029 | /* ARGSUSED */ | |
4030 | static int | |
4031 | zfs_null_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, | |
4032 | size_t *lenp, int flags, cred_t *cr) | |
4033 | { | |
4034 | pvn_write_done(pp, B_INVAL|B_FORCE|B_ERROR); | |
4035 | return (0); | |
4036 | } | |
4037 | ||
4038 | /* | |
4039 | * Push a page out to disk, klustering if possible. | |
4040 | * | |
4041 | * IN: vp - file to push page to. | |
4042 | * pp - page to push. | |
4043 | * flags - additional flags. | |
4044 | * cr - credentials of caller. | |
4045 | * | |
4046 | * OUT: offp - start of range pushed. | |
4047 | * lenp - len of range pushed. | |
4048 | * | |
4049 | * RETURN: 0 if success | |
4050 | * error code if failure | |
4051 | * | |
4052 | * NOTE: callers must have locked the page to be pushed. On | |
4053 | * exit, the page (and all other pages in the kluster) must be | |
4054 | * unlocked. | |
4055 | */ | |
4056 | /* ARGSUSED */ | |
4057 | static int | |
4058 | zfs_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, | |
4059 | size_t *lenp, int flags, cred_t *cr) | |
4060 | { | |
4061 | znode_t *zp = VTOZ(vp); | |
4062 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
34dc7c2f | 4063 | dmu_tx_t *tx; |
34dc7c2f BB |
4064 | u_offset_t off, koff; |
4065 | size_t len, klen; | |
34dc7c2f BB |
4066 | int err; |
4067 | ||
34dc7c2f BB |
4068 | off = pp->p_offset; |
4069 | len = PAGESIZE; | |
4070 | /* | |
4071 | * If our blocksize is bigger than the page size, try to kluster | |
fb5f0bc8 | 4072 | * multiple pages so that we write a full block (thus avoiding |
34dc7c2f BB |
4073 | * a read-modify-write). |
4074 | */ | |
428870ff | 4075 | if (off < zp->z_size && zp->z_blksz > PAGESIZE) { |
d164b209 BB |
4076 | klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE); |
4077 | koff = ISP2(klen) ? P2ALIGN(off, (u_offset_t)klen) : 0; | |
428870ff BB |
4078 | ASSERT(koff <= zp->z_size); |
4079 | if (koff + klen > zp->z_size) | |
4080 | klen = P2ROUNDUP(zp->z_size - koff, (uint64_t)PAGESIZE); | |
34dc7c2f BB |
4081 | pp = pvn_write_kluster(vp, pp, &off, &len, koff, klen, flags); |
4082 | } | |
4083 | ASSERT3U(btop(len), ==, btopr(len)); | |
d164b209 | 4084 | |
34dc7c2f BB |
4085 | /* |
4086 | * Can't push pages past end-of-file. | |
4087 | */ | |
428870ff | 4088 | if (off >= zp->z_size) { |
34dc7c2f BB |
4089 | /* ignore all pages */ |
4090 | err = 0; | |
4091 | goto out; | |
428870ff BB |
4092 | } else if (off + len > zp->z_size) { |
4093 | int npages = btopr(zp->z_size - off); | |
34dc7c2f BB |
4094 | page_t *trunc; |
4095 | ||
4096 | page_list_break(&pp, &trunc, npages); | |
4097 | /* ignore pages past end of file */ | |
4098 | if (trunc) | |
4099 | pvn_write_done(trunc, flags); | |
428870ff | 4100 | len = zp->z_size - off; |
34dc7c2f | 4101 | } |
9babb374 | 4102 | |
428870ff BB |
4103 | if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || |
4104 | zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { | |
9babb374 BB |
4105 | err = EDQUOT; |
4106 | goto out; | |
4107 | } | |
d164b209 | 4108 | top: |
34dc7c2f BB |
4109 | tx = dmu_tx_create(zfsvfs->z_os); |
4110 | dmu_tx_hold_write(tx, zp->z_id, off, len); | |
428870ff BB |
4111 | |
4112 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); | |
4113 | zfs_sa_upgrade_txholds(tx, zp); | |
fb5f0bc8 | 4114 | err = dmu_tx_assign(tx, TXG_NOWAIT); |
34dc7c2f | 4115 | if (err != 0) { |
fb5f0bc8 | 4116 | if (err == ERESTART) { |
34dc7c2f BB |
4117 | dmu_tx_wait(tx); |
4118 | dmu_tx_abort(tx); | |
34dc7c2f BB |
4119 | goto top; |
4120 | } | |
4121 | dmu_tx_abort(tx); | |
4122 | goto out; | |
4123 | } | |
4124 | ||
4125 | if (zp->z_blksz <= PAGESIZE) { | |
b128c09f | 4126 | caddr_t va = zfs_map_page(pp, S_READ); |
34dc7c2f BB |
4127 | ASSERT3U(len, <=, PAGESIZE); |
4128 | dmu_write(zfsvfs->z_os, zp->z_id, off, len, va, tx); | |
b128c09f | 4129 | zfs_unmap_page(pp, va); |
34dc7c2f BB |
4130 | } else { |
4131 | err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, pp, tx); | |
4132 | } | |
4133 | ||
4134 | if (err == 0) { | |
428870ff BB |
4135 | uint64_t mtime[2], ctime[2]; |
4136 | sa_bulk_attr_t bulk[3]; | |
4137 | int count = 0; | |
4138 | ||
4139 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, | |
4140 | &mtime, 16); | |
4141 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, | |
4142 | &ctime, 16); | |
4143 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
4144 | &zp->z_pflags, 8); | |
4145 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, | |
4146 | B_TRUE); | |
d164b209 | 4147 | zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off, len, 0); |
34dc7c2f | 4148 | } |
45d1cae3 | 4149 | dmu_tx_commit(tx); |
34dc7c2f BB |
4150 | |
4151 | out: | |
34dc7c2f BB |
4152 | pvn_write_done(pp, (err ? B_ERROR : 0) | flags); |
4153 | if (offp) | |
4154 | *offp = off; | |
4155 | if (lenp) | |
4156 | *lenp = len; | |
4157 | ||
4158 | return (err); | |
4159 | } | |
4160 | ||
4161 | /* | |
4162 | * Copy the portion of the file indicated from pages into the file. | |
4163 | * The pages are stored in a page list attached to the files vnode. | |
4164 | * | |
4165 | * IN: vp - vnode of file to push page data to. | |
4166 | * off - position in file to put data. | |
4167 | * len - amount of data to write. | |
4168 | * flags - flags to control the operation. | |
4169 | * cr - credentials of caller. | |
4170 | * ct - caller context. | |
4171 | * | |
4172 | * RETURN: 0 if success | |
4173 | * error code if failure | |
4174 | * | |
4175 | * Timestamps: | |
4176 | * vp - ctime|mtime updated | |
4177 | */ | |
4178 | /*ARGSUSED*/ | |
4179 | static int | |
4180 | zfs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr, | |
4181 | caller_context_t *ct) | |
4182 | { | |
4183 | znode_t *zp = VTOZ(vp); | |
4184 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
4185 | page_t *pp; | |
4186 | size_t io_len; | |
4187 | u_offset_t io_off; | |
d164b209 BB |
4188 | uint_t blksz; |
4189 | rl_t *rl; | |
34dc7c2f BB |
4190 | int error = 0; |
4191 | ||
4192 | ZFS_ENTER(zfsvfs); | |
4193 | ZFS_VERIFY_ZP(zp); | |
4194 | ||
d164b209 BB |
4195 | /* |
4196 | * Align this request to the file block size in case we kluster. | |
4197 | * XXX - this can result in pretty aggresive locking, which can | |
4198 | * impact simultanious read/write access. One option might be | |
4199 | * to break up long requests (len == 0) into block-by-block | |
4200 | * operations to get narrower locking. | |
4201 | */ | |
4202 | blksz = zp->z_blksz; | |
4203 | if (ISP2(blksz)) | |
4204 | io_off = P2ALIGN_TYPED(off, blksz, u_offset_t); | |
4205 | else | |
4206 | io_off = 0; | |
4207 | if (len > 0 && ISP2(blksz)) | |
9babb374 | 4208 | io_len = P2ROUNDUP_TYPED(len + (off - io_off), blksz, size_t); |
d164b209 BB |
4209 | else |
4210 | io_len = 0; | |
4211 | ||
4212 | if (io_len == 0) { | |
34dc7c2f | 4213 | /* |
d164b209 | 4214 | * Search the entire vp list for pages >= io_off. |
34dc7c2f | 4215 | */ |
d164b209 BB |
4216 | rl = zfs_range_lock(zp, io_off, UINT64_MAX, RL_WRITER); |
4217 | error = pvn_vplist_dirty(vp, io_off, zfs_putapage, flags, cr); | |
34dc7c2f BB |
4218 | goto out; |
4219 | } | |
d164b209 | 4220 | rl = zfs_range_lock(zp, io_off, io_len, RL_WRITER); |
34dc7c2f | 4221 | |
428870ff | 4222 | if (off > zp->z_size) { |
34dc7c2f | 4223 | /* past end of file */ |
d164b209 | 4224 | zfs_range_unlock(rl); |
34dc7c2f BB |
4225 | ZFS_EXIT(zfsvfs); |
4226 | return (0); | |
4227 | } | |
4228 | ||
428870ff | 4229 | len = MIN(io_len, P2ROUNDUP(zp->z_size, PAGESIZE) - io_off); |
34dc7c2f | 4230 | |
d164b209 | 4231 | for (off = io_off; io_off < off + len; io_off += io_len) { |
34dc7c2f BB |
4232 | if ((flags & B_INVAL) || ((flags & B_ASYNC) == 0)) { |
4233 | pp = page_lookup(vp, io_off, | |
4234 | (flags & (B_INVAL | B_FREE)) ? SE_EXCL : SE_SHARED); | |
4235 | } else { | |
4236 | pp = page_lookup_nowait(vp, io_off, | |
4237 | (flags & B_FREE) ? SE_EXCL : SE_SHARED); | |
4238 | } | |
4239 | ||
4240 | if (pp != NULL && pvn_getdirty(pp, flags)) { | |
4241 | int err; | |
4242 | ||
4243 | /* | |
4244 | * Found a dirty page to push | |
4245 | */ | |
4246 | err = zfs_putapage(vp, pp, &io_off, &io_len, flags, cr); | |
4247 | if (err) | |
4248 | error = err; | |
4249 | } else { | |
4250 | io_len = PAGESIZE; | |
4251 | } | |
4252 | } | |
4253 | out: | |
d164b209 | 4254 | zfs_range_unlock(rl); |
428870ff | 4255 | if ((flags & B_ASYNC) == 0 || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) |
572e2857 | 4256 | zil_commit(zfsvfs->z_log, zp->z_id); |
34dc7c2f BB |
4257 | ZFS_EXIT(zfsvfs); |
4258 | return (error); | |
4259 | } | |
4260 | ||
4261 | /*ARGSUSED*/ | |
4262 | void | |
4263 | zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct) | |
4264 | { | |
4265 | znode_t *zp = VTOZ(vp); | |
4266 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
4267 | int error; | |
4268 | ||
4269 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); | |
428870ff | 4270 | if (zp->z_sa_hdl == NULL) { |
34dc7c2f BB |
4271 | /* |
4272 | * The fs has been unmounted, or we did a | |
4273 | * suspend/resume and this file no longer exists. | |
4274 | */ | |
4275 | if (vn_has_cached_data(vp)) { | |
4276 | (void) pvn_vplist_dirty(vp, 0, zfs_null_putapage, | |
4277 | B_INVAL, cr); | |
4278 | } | |
4279 | ||
4280 | mutex_enter(&zp->z_lock); | |
428870ff BB |
4281 | mutex_enter(&vp->v_lock); |
4282 | ASSERT(vp->v_count == 1); | |
4283 | vp->v_count = 0; | |
4284 | mutex_exit(&vp->v_lock); | |
34dc7c2f BB |
4285 | mutex_exit(&zp->z_lock); |
4286 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
4287 | zfs_znode_free(zp); | |
4288 | return; | |
4289 | } | |
4290 | ||
4291 | /* | |
4292 | * Attempt to push any data in the page cache. If this fails | |
4293 | * we will get kicked out later in zfs_zinactive(). | |
4294 | */ | |
4295 | if (vn_has_cached_data(vp)) { | |
4296 | (void) pvn_vplist_dirty(vp, 0, zfs_putapage, B_INVAL|B_ASYNC, | |
4297 | cr); | |
4298 | } | |
4299 | ||
4300 | if (zp->z_atime_dirty && zp->z_unlinked == 0) { | |
4301 | dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); | |
4302 | ||
428870ff BB |
4303 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
4304 | zfs_sa_upgrade_txholds(tx, zp); | |
34dc7c2f BB |
4305 | error = dmu_tx_assign(tx, TXG_WAIT); |
4306 | if (error) { | |
4307 | dmu_tx_abort(tx); | |
4308 | } else { | |
34dc7c2f | 4309 | mutex_enter(&zp->z_lock); |
428870ff BB |
4310 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), |
4311 | (void *)&zp->z_atime, sizeof (zp->z_atime), tx); | |
34dc7c2f BB |
4312 | zp->z_atime_dirty = 0; |
4313 | mutex_exit(&zp->z_lock); | |
4314 | dmu_tx_commit(tx); | |
4315 | } | |
4316 | } | |
4317 | ||
4318 | zfs_zinactive(zp); | |
4319 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
4320 | } | |
4321 | ||
4322 | /* | |
4323 | * Bounds-check the seek operation. | |
4324 | * | |
4325 | * IN: vp - vnode seeking within | |
4326 | * ooff - old file offset | |
4327 | * noffp - pointer to new file offset | |
4328 | * ct - caller context | |
4329 | * | |
4330 | * RETURN: 0 if success | |
4331 | * EINVAL if new offset invalid | |
4332 | */ | |
4333 | /* ARGSUSED */ | |
4334 | static int | |
4335 | zfs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp, | |
4336 | caller_context_t *ct) | |
4337 | { | |
4338 | if (vp->v_type == VDIR) | |
4339 | return (0); | |
4340 | return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); | |
4341 | } | |
4342 | ||
4343 | /* | |
4344 | * Pre-filter the generic locking function to trap attempts to place | |
4345 | * a mandatory lock on a memory mapped file. | |
4346 | */ | |
4347 | static int | |
4348 | zfs_frlock(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset, | |
4349 | flk_callback_t *flk_cbp, cred_t *cr, caller_context_t *ct) | |
4350 | { | |
4351 | znode_t *zp = VTOZ(vp); | |
4352 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
34dc7c2f BB |
4353 | |
4354 | ZFS_ENTER(zfsvfs); | |
4355 | ZFS_VERIFY_ZP(zp); | |
4356 | ||
4357 | /* | |
4358 | * We are following the UFS semantics with respect to mapcnt | |
4359 | * here: If we see that the file is mapped already, then we will | |
4360 | * return an error, but we don't worry about races between this | |
4361 | * function and zfs_map(). | |
4362 | */ | |
428870ff | 4363 | if (zp->z_mapcnt > 0 && MANDMODE(zp->z_mode)) { |
34dc7c2f BB |
4364 | ZFS_EXIT(zfsvfs); |
4365 | return (EAGAIN); | |
4366 | } | |
34dc7c2f | 4367 | ZFS_EXIT(zfsvfs); |
428870ff | 4368 | return (fs_frlock(vp, cmd, bfp, flag, offset, flk_cbp, cr, ct)); |
34dc7c2f BB |
4369 | } |
4370 | ||
4371 | /* | |
4372 | * If we can't find a page in the cache, we will create a new page | |
4373 | * and fill it with file data. For efficiency, we may try to fill | |
d164b209 | 4374 | * multiple pages at once (klustering) to fill up the supplied page |
9babb374 BB |
4375 | * list. Note that the pages to be filled are held with an exclusive |
4376 | * lock to prevent access by other threads while they are being filled. | |
34dc7c2f BB |
4377 | */ |
4378 | static int | |
4379 | zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg, | |
4380 | caddr_t addr, page_t *pl[], size_t plsz, enum seg_rw rw) | |
4381 | { | |
4382 | znode_t *zp = VTOZ(vp); | |
4383 | page_t *pp, *cur_pp; | |
4384 | objset_t *os = zp->z_zfsvfs->z_os; | |
34dc7c2f | 4385 | u_offset_t io_off, total; |
34dc7c2f | 4386 | size_t io_len; |
34dc7c2f BB |
4387 | int err; |
4388 | ||
34dc7c2f | 4389 | if (plsz == PAGESIZE || zp->z_blksz <= PAGESIZE) { |
d164b209 BB |
4390 | /* |
4391 | * We only have a single page, don't bother klustering | |
4392 | */ | |
34dc7c2f BB |
4393 | io_off = off; |
4394 | io_len = PAGESIZE; | |
9babb374 BB |
4395 | pp = page_create_va(vp, io_off, io_len, |
4396 | PG_EXCL | PG_WAIT, seg, addr); | |
34dc7c2f BB |
4397 | } else { |
4398 | /* | |
d164b209 | 4399 | * Try to find enough pages to fill the page list |
34dc7c2f | 4400 | */ |
34dc7c2f | 4401 | pp = pvn_read_kluster(vp, off, seg, addr, &io_off, |
d164b209 | 4402 | &io_len, off, plsz, 0); |
34dc7c2f BB |
4403 | } |
4404 | if (pp == NULL) { | |
4405 | /* | |
d164b209 | 4406 | * The page already exists, nothing to do here. |
34dc7c2f BB |
4407 | */ |
4408 | *pl = NULL; | |
4409 | return (0); | |
4410 | } | |
4411 | ||
4412 | /* | |
4413 | * Fill the pages in the kluster. | |
4414 | */ | |
4415 | cur_pp = pp; | |
4416 | for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { | |
d164b209 BB |
4417 | caddr_t va; |
4418 | ||
34dc7c2f | 4419 | ASSERT3U(io_off, ==, cur_pp->p_offset); |
b128c09f | 4420 | va = zfs_map_page(cur_pp, S_WRITE); |
9babb374 BB |
4421 | err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, |
4422 | DMU_READ_PREFETCH); | |
b128c09f | 4423 | zfs_unmap_page(cur_pp, va); |
34dc7c2f BB |
4424 | if (err) { |
4425 | /* On error, toss the entire kluster */ | |
4426 | pvn_read_done(pp, B_ERROR); | |
b128c09f BB |
4427 | /* convert checksum errors into IO errors */ |
4428 | if (err == ECKSUM) | |
4429 | err = EIO; | |
34dc7c2f BB |
4430 | return (err); |
4431 | } | |
4432 | cur_pp = cur_pp->p_next; | |
4433 | } | |
d164b209 | 4434 | |
34dc7c2f | 4435 | /* |
d164b209 BB |
4436 | * Fill in the page list array from the kluster starting |
4437 | * from the desired offset `off'. | |
34dc7c2f BB |
4438 | * NOTE: the page list will always be null terminated. |
4439 | */ | |
4440 | pvn_plist_init(pp, pl, plsz, off, io_len, rw); | |
d164b209 | 4441 | ASSERT(pl == NULL || (*pl)->p_offset == off); |
34dc7c2f BB |
4442 | |
4443 | return (0); | |
4444 | } | |
4445 | ||
4446 | /* | |
4447 | * Return pointers to the pages for the file region [off, off + len] | |
4448 | * in the pl array. If plsz is greater than len, this function may | |
d164b209 BB |
4449 | * also return page pointers from after the specified region |
4450 | * (i.e. the region [off, off + plsz]). These additional pages are | |
4451 | * only returned if they are already in the cache, or were created as | |
4452 | * part of a klustered read. | |
34dc7c2f BB |
4453 | * |
4454 | * IN: vp - vnode of file to get data from. | |
4455 | * off - position in file to get data from. | |
4456 | * len - amount of data to retrieve. | |
4457 | * plsz - length of provided page list. | |
4458 | * seg - segment to obtain pages for. | |
4459 | * addr - virtual address of fault. | |
4460 | * rw - mode of created pages. | |
4461 | * cr - credentials of caller. | |
4462 | * ct - caller context. | |
4463 | * | |
4464 | * OUT: protp - protection mode of created pages. | |
4465 | * pl - list of pages created. | |
4466 | * | |
4467 | * RETURN: 0 if success | |
4468 | * error code if failure | |
4469 | * | |
4470 | * Timestamps: | |
4471 | * vp - atime updated | |
4472 | */ | |
4473 | /* ARGSUSED */ | |
4474 | static int | |
4475 | zfs_getpage(vnode_t *vp, offset_t off, size_t len, uint_t *protp, | |
4476 | page_t *pl[], size_t plsz, struct seg *seg, caddr_t addr, | |
4477 | enum seg_rw rw, cred_t *cr, caller_context_t *ct) | |
4478 | { | |
4479 | znode_t *zp = VTOZ(vp); | |
4480 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
d164b209 BB |
4481 | page_t **pl0 = pl; |
4482 | int err = 0; | |
4483 | ||
4484 | /* we do our own caching, faultahead is unnecessary */ | |
4485 | if (pl == NULL) | |
4486 | return (0); | |
4487 | else if (len > plsz) | |
4488 | len = plsz; | |
4489 | else | |
4490 | len = P2ROUNDUP(len, PAGESIZE); | |
4491 | ASSERT(plsz >= len); | |
34dc7c2f BB |
4492 | |
4493 | ZFS_ENTER(zfsvfs); | |
4494 | ZFS_VERIFY_ZP(zp); | |
4495 | ||
4496 | if (protp) | |
4497 | *protp = PROT_ALL; | |
4498 | ||
34dc7c2f | 4499 | /* |
9babb374 | 4500 | * Loop through the requested range [off, off + len) looking |
34dc7c2f BB |
4501 | * for pages. If we don't find a page, we will need to create |
4502 | * a new page and fill it with data from the file. | |
4503 | */ | |
4504 | while (len > 0) { | |
d164b209 BB |
4505 | if (*pl = page_lookup(vp, off, SE_SHARED)) |
4506 | *(pl+1) = NULL; | |
4507 | else if (err = zfs_fillpage(vp, off, seg, addr, pl, plsz, rw)) | |
4508 | goto out; | |
4509 | while (*pl) { | |
4510 | ASSERT3U((*pl)->p_offset, ==, off); | |
34dc7c2f BB |
4511 | off += PAGESIZE; |
4512 | addr += PAGESIZE; | |
d164b209 BB |
4513 | if (len > 0) { |
4514 | ASSERT3U(len, >=, PAGESIZE); | |
4515 | len -= PAGESIZE; | |
34dc7c2f | 4516 | } |
d164b209 BB |
4517 | ASSERT3U(plsz, >=, PAGESIZE); |
4518 | plsz -= PAGESIZE; | |
4519 | pl++; | |
34dc7c2f BB |
4520 | } |
4521 | } | |
4522 | ||
4523 | /* | |
4524 | * Fill out the page array with any pages already in the cache. | |
4525 | */ | |
d164b209 BB |
4526 | while (plsz > 0 && |
4527 | (*pl++ = page_lookup_nowait(vp, off, SE_SHARED))) { | |
4528 | off += PAGESIZE; | |
4529 | plsz -= PAGESIZE; | |
34dc7c2f | 4530 | } |
34dc7c2f | 4531 | out: |
34dc7c2f BB |
4532 | if (err) { |
4533 | /* | |
4534 | * Release any pages we have previously locked. | |
4535 | */ | |
4536 | while (pl > pl0) | |
4537 | page_unlock(*--pl); | |
d164b209 BB |
4538 | } else { |
4539 | ZFS_ACCESSTIME_STAMP(zfsvfs, zp); | |
34dc7c2f BB |
4540 | } |
4541 | ||
4542 | *pl = NULL; | |
4543 | ||
34dc7c2f BB |
4544 | ZFS_EXIT(zfsvfs); |
4545 | return (err); | |
4546 | } | |
4547 | ||
4548 | /* | |
4549 | * Request a memory map for a section of a file. This code interacts | |
4550 | * with common code and the VM system as follows: | |
4551 | * | |
4552 | * common code calls mmap(), which ends up in smmap_common() | |
4553 | * | |
4554 | * this calls VOP_MAP(), which takes you into (say) zfs | |
4555 | * | |
4556 | * zfs_map() calls as_map(), passing segvn_create() as the callback | |
4557 | * | |
4558 | * segvn_create() creates the new segment and calls VOP_ADDMAP() | |
4559 | * | |
4560 | * zfs_addmap() updates z_mapcnt | |
4561 | */ | |
4562 | /*ARGSUSED*/ | |
4563 | static int | |
4564 | zfs_map(vnode_t *vp, offset_t off, struct as *as, caddr_t *addrp, | |
4565 | size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr, | |
4566 | caller_context_t *ct) | |
4567 | { | |
4568 | znode_t *zp = VTOZ(vp); | |
4569 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
4570 | segvn_crargs_t vn_a; | |
4571 | int error; | |
4572 | ||
4573 | ZFS_ENTER(zfsvfs); | |
4574 | ZFS_VERIFY_ZP(zp); | |
4575 | ||
428870ff BB |
4576 | if ((prot & PROT_WRITE) && (zp->z_pflags & |
4577 | (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { | |
34dc7c2f BB |
4578 | ZFS_EXIT(zfsvfs); |
4579 | return (EPERM); | |
4580 | } | |
4581 | ||
4582 | if ((prot & (PROT_READ | PROT_EXEC)) && | |
428870ff | 4583 | (zp->z_pflags & ZFS_AV_QUARANTINED)) { |
34dc7c2f BB |
4584 | ZFS_EXIT(zfsvfs); |
4585 | return (EACCES); | |
4586 | } | |
4587 | ||
4588 | if (vp->v_flag & VNOMAP) { | |
4589 | ZFS_EXIT(zfsvfs); | |
4590 | return (ENOSYS); | |
4591 | } | |
4592 | ||
4593 | if (off < 0 || len > MAXOFFSET_T - off) { | |
4594 | ZFS_EXIT(zfsvfs); | |
4595 | return (ENXIO); | |
4596 | } | |
4597 | ||
4598 | if (vp->v_type != VREG) { | |
4599 | ZFS_EXIT(zfsvfs); | |
4600 | return (ENODEV); | |
4601 | } | |
4602 | ||
4603 | /* | |
4604 | * If file is locked, disallow mapping. | |
4605 | */ | |
428870ff | 4606 | if (MANDMODE(zp->z_mode) && vn_has_flocks(vp)) { |
34dc7c2f BB |
4607 | ZFS_EXIT(zfsvfs); |
4608 | return (EAGAIN); | |
4609 | } | |
4610 | ||
4611 | as_rangelock(as); | |
4612 | error = choose_addr(as, addrp, len, off, ADDR_VACALIGN, flags); | |
4613 | if (error != 0) { | |
4614 | as_rangeunlock(as); | |
4615 | ZFS_EXIT(zfsvfs); | |
4616 | return (error); | |
4617 | } | |
4618 | ||
4619 | vn_a.vp = vp; | |
4620 | vn_a.offset = (u_offset_t)off; | |
4621 | vn_a.type = flags & MAP_TYPE; | |
4622 | vn_a.prot = prot; | |
4623 | vn_a.maxprot = maxprot; | |
4624 | vn_a.cred = cr; | |
4625 | vn_a.amp = NULL; | |
4626 | vn_a.flags = flags & ~MAP_TYPE; | |
4627 | vn_a.szc = 0; | |
4628 | vn_a.lgrp_mem_policy_flags = 0; | |
4629 | ||
4630 | error = as_map(as, *addrp, len, segvn_create, &vn_a); | |
4631 | ||
4632 | as_rangeunlock(as); | |
4633 | ZFS_EXIT(zfsvfs); | |
4634 | return (error); | |
4635 | } | |
4636 | ||
4637 | /* ARGSUSED */ | |
4638 | static int | |
4639 | zfs_addmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr, | |
4640 | size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr, | |
4641 | caller_context_t *ct) | |
4642 | { | |
4643 | uint64_t pages = btopr(len); | |
4644 | ||
4645 | atomic_add_64(&VTOZ(vp)->z_mapcnt, pages); | |
4646 | return (0); | |
4647 | } | |
4648 | ||
4649 | /* | |
4650 | * The reason we push dirty pages as part of zfs_delmap() is so that we get a | |
4651 | * more accurate mtime for the associated file. Since we don't have a way of | |
4652 | * detecting when the data was actually modified, we have to resort to | |
4653 | * heuristics. If an explicit msync() is done, then we mark the mtime when the | |
4654 | * last page is pushed. The problem occurs when the msync() call is omitted, | |
4655 | * which by far the most common case: | |
4656 | * | |
4657 | * open() | |
4658 | * mmap() | |
4659 | * <modify memory> | |
4660 | * munmap() | |
4661 | * close() | |
4662 | * <time lapse> | |
4663 | * putpage() via fsflush | |
4664 | * | |
4665 | * If we wait until fsflush to come along, we can have a modification time that | |
4666 | * is some arbitrary point in the future. In order to prevent this in the | |
4667 | * common case, we flush pages whenever a (MAP_SHARED, PROT_WRITE) mapping is | |
4668 | * torn down. | |
4669 | */ | |
4670 | /* ARGSUSED */ | |
4671 | static int | |
4672 | zfs_delmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr, | |
4673 | size_t len, uint_t prot, uint_t maxprot, uint_t flags, cred_t *cr, | |
4674 | caller_context_t *ct) | |
4675 | { | |
4676 | uint64_t pages = btopr(len); | |
4677 | ||
4678 | ASSERT3U(VTOZ(vp)->z_mapcnt, >=, pages); | |
4679 | atomic_add_64(&VTOZ(vp)->z_mapcnt, -pages); | |
4680 | ||
4681 | if ((flags & MAP_SHARED) && (prot & PROT_WRITE) && | |
4682 | vn_has_cached_data(vp)) | |
4683 | (void) VOP_PUTPAGE(vp, off, len, B_ASYNC, cr, ct); | |
4684 | ||
4685 | return (0); | |
4686 | } | |
4687 | ||
4688 | /* | |
4689 | * Free or allocate space in a file. Currently, this function only | |
4690 | * supports the `F_FREESP' command. However, this command is somewhat | |
4691 | * misnamed, as its functionality includes the ability to allocate as | |
4692 | * well as free space. | |
4693 | * | |
4694 | * IN: vp - vnode of file to free data in. | |
4695 | * cmd - action to take (only F_FREESP supported). | |
4696 | * bfp - section of file to free/alloc. | |
4697 | * flag - current file open mode flags. | |
4698 | * offset - current file offset. | |
4699 | * cr - credentials of caller [UNUSED]. | |
4700 | * ct - caller context. | |
4701 | * | |
4702 | * RETURN: 0 if success | |
4703 | * error code if failure | |
4704 | * | |
4705 | * Timestamps: | |
4706 | * vp - ctime|mtime updated | |
4707 | */ | |
4708 | /* ARGSUSED */ | |
4709 | static int | |
4710 | zfs_space(vnode_t *vp, int cmd, flock64_t *bfp, int flag, | |
4711 | offset_t offset, cred_t *cr, caller_context_t *ct) | |
4712 | { | |
4713 | znode_t *zp = VTOZ(vp); | |
4714 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
4715 | uint64_t off, len; | |
4716 | int error; | |
4717 | ||
4718 | ZFS_ENTER(zfsvfs); | |
4719 | ZFS_VERIFY_ZP(zp); | |
4720 | ||
34dc7c2f BB |
4721 | if (cmd != F_FREESP) { |
4722 | ZFS_EXIT(zfsvfs); | |
4723 | return (EINVAL); | |
4724 | } | |
4725 | ||
4726 | if (error = convoff(vp, bfp, 0, offset)) { | |
4727 | ZFS_EXIT(zfsvfs); | |
4728 | return (error); | |
4729 | } | |
4730 | ||
4731 | if (bfp->l_len < 0) { | |
4732 | ZFS_EXIT(zfsvfs); | |
4733 | return (EINVAL); | |
4734 | } | |
4735 | ||
4736 | off = bfp->l_start; | |
4737 | len = bfp->l_len; /* 0 means from off to end of file */ | |
4738 | ||
b128c09f | 4739 | error = zfs_freesp(zp, off, len, flag, TRUE); |
34dc7c2f BB |
4740 | |
4741 | ZFS_EXIT(zfsvfs); | |
4742 | return (error); | |
4743 | } | |
4744 | ||
4745 | /*ARGSUSED*/ | |
4746 | static int | |
4747 | zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct) | |
4748 | { | |
4749 | znode_t *zp = VTOZ(vp); | |
4750 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
4751 | uint32_t gen; | |
428870ff | 4752 | uint64_t gen64; |
34dc7c2f BB |
4753 | uint64_t object = zp->z_id; |
4754 | zfid_short_t *zfid; | |
428870ff | 4755 | int size, i, error; |
34dc7c2f BB |
4756 | |
4757 | ZFS_ENTER(zfsvfs); | |
4758 | ZFS_VERIFY_ZP(zp); | |
428870ff BB |
4759 | |
4760 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), | |
4761 | &gen64, sizeof (uint64_t))) != 0) { | |
4762 | ZFS_EXIT(zfsvfs); | |
4763 | return (error); | |
4764 | } | |
4765 | ||
4766 | gen = (uint32_t)gen64; | |
34dc7c2f BB |
4767 | |
4768 | size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN; | |
4769 | if (fidp->fid_len < size) { | |
4770 | fidp->fid_len = size; | |
4771 | ZFS_EXIT(zfsvfs); | |
4772 | return (ENOSPC); | |
4773 | } | |
4774 | ||
4775 | zfid = (zfid_short_t *)fidp; | |
4776 | ||
4777 | zfid->zf_len = size; | |
4778 | ||
4779 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
4780 | zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); | |
4781 | ||
4782 | /* Must have a non-zero generation number to distinguish from .zfs */ | |
4783 | if (gen == 0) | |
4784 | gen = 1; | |
4785 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
4786 | zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); | |
4787 | ||
4788 | if (size == LONG_FID_LEN) { | |
4789 | uint64_t objsetid = dmu_objset_id(zfsvfs->z_os); | |
4790 | zfid_long_t *zlfid; | |
4791 | ||
4792 | zlfid = (zfid_long_t *)fidp; | |
4793 | ||
4794 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
4795 | zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); | |
4796 | ||
4797 | /* XXX - this should be the generation number for the objset */ | |
4798 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
4799 | zlfid->zf_setgen[i] = 0; | |
4800 | } | |
4801 | ||
4802 | ZFS_EXIT(zfsvfs); | |
4803 | return (0); | |
4804 | } | |
4805 | ||
4806 | static int | |
4807 | zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr, | |
4808 | caller_context_t *ct) | |
4809 | { | |
4810 | znode_t *zp, *xzp; | |
4811 | zfsvfs_t *zfsvfs; | |
4812 | zfs_dirlock_t *dl; | |
4813 | int error; | |
4814 | ||
4815 | switch (cmd) { | |
4816 | case _PC_LINK_MAX: | |
4817 | *valp = ULONG_MAX; | |
4818 | return (0); | |
4819 | ||
4820 | case _PC_FILESIZEBITS: | |
4821 | *valp = 64; | |
4822 | return (0); | |
4823 | ||
4824 | case _PC_XATTR_EXISTS: | |
4825 | zp = VTOZ(vp); | |
4826 | zfsvfs = zp->z_zfsvfs; | |
4827 | ZFS_ENTER(zfsvfs); | |
4828 | ZFS_VERIFY_ZP(zp); | |
4829 | *valp = 0; | |
4830 | error = zfs_dirent_lock(&dl, zp, "", &xzp, | |
4831 | ZXATTR | ZEXISTS | ZSHARED, NULL, NULL); | |
4832 | if (error == 0) { | |
4833 | zfs_dirent_unlock(dl); | |
4834 | if (!zfs_dirempty(xzp)) | |
4835 | *valp = 1; | |
4836 | VN_RELE(ZTOV(xzp)); | |
4837 | } else if (error == ENOENT) { | |
4838 | /* | |
4839 | * If there aren't extended attributes, it's the | |
4840 | * same as having zero of them. | |
4841 | */ | |
4842 | error = 0; | |
4843 | } | |
4844 | ZFS_EXIT(zfsvfs); | |
4845 | return (error); | |
4846 | ||
4847 | case _PC_SATTR_ENABLED: | |
4848 | case _PC_SATTR_EXISTS: | |
b128c09f | 4849 | *valp = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) && |
34dc7c2f BB |
4850 | (vp->v_type == VREG || vp->v_type == VDIR); |
4851 | return (0); | |
4852 | ||
9babb374 BB |
4853 | case _PC_ACCESS_FILTERING: |
4854 | *valp = vfs_has_feature(vp->v_vfsp, VFSFT_ACCESS_FILTER) && | |
4855 | vp->v_type == VDIR; | |
4856 | return (0); | |
4857 | ||
34dc7c2f BB |
4858 | case _PC_ACL_ENABLED: |
4859 | *valp = _ACL_ACE_ENABLED; | |
4860 | return (0); | |
4861 | ||
4862 | case _PC_MIN_HOLE_SIZE: | |
4863 | *valp = (ulong_t)SPA_MINBLOCKSIZE; | |
4864 | return (0); | |
4865 | ||
428870ff BB |
4866 | case _PC_TIMESTAMP_RESOLUTION: |
4867 | /* nanosecond timestamp resolution */ | |
4868 | *valp = 1L; | |
4869 | return (0); | |
4870 | ||
34dc7c2f BB |
4871 | default: |
4872 | return (fs_pathconf(vp, cmd, valp, cr, ct)); | |
4873 | } | |
4874 | } | |
4875 | ||
4876 | /*ARGSUSED*/ | |
4877 | static int | |
4878 | zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr, | |
4879 | caller_context_t *ct) | |
4880 | { | |
4881 | znode_t *zp = VTOZ(vp); | |
4882 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
4883 | int error; | |
4884 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
4885 | ||
4886 | ZFS_ENTER(zfsvfs); | |
4887 | ZFS_VERIFY_ZP(zp); | |
4888 | error = zfs_getacl(zp, vsecp, skipaclchk, cr); | |
4889 | ZFS_EXIT(zfsvfs); | |
4890 | ||
4891 | return (error); | |
4892 | } | |
4893 | ||
4894 | /*ARGSUSED*/ | |
4895 | static int | |
4896 | zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr, | |
4897 | caller_context_t *ct) | |
4898 | { | |
4899 | znode_t *zp = VTOZ(vp); | |
4900 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
4901 | int error; | |
4902 | boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; | |
428870ff | 4903 | zilog_t *zilog = zfsvfs->z_log; |
34dc7c2f BB |
4904 | |
4905 | ZFS_ENTER(zfsvfs); | |
4906 | ZFS_VERIFY_ZP(zp); | |
428870ff | 4907 | |
34dc7c2f | 4908 | error = zfs_setacl(zp, vsecp, skipaclchk, cr); |
428870ff BB |
4909 | |
4910 | if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) | |
572e2857 | 4911 | zil_commit(zilog, 0); |
428870ff | 4912 | |
34dc7c2f BB |
4913 | ZFS_EXIT(zfsvfs); |
4914 | return (error); | |
4915 | } | |
4916 | ||
428870ff BB |
4917 | /* |
4918 | * Tunable, both must be a power of 2. | |
4919 | * | |
4920 | * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf | |
4921 | * zcr_blksz_max: if set to less than the file block size, allow loaning out of | |
4922 | * an arcbuf for a partial block read | |
4923 | */ | |
4924 | int zcr_blksz_min = (1 << 10); /* 1K */ | |
4925 | int zcr_blksz_max = (1 << 17); /* 128K */ | |
4926 | ||
4927 | /*ARGSUSED*/ | |
4928 | static int | |
4929 | zfs_reqzcbuf(vnode_t *vp, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr, | |
4930 | caller_context_t *ct) | |
4931 | { | |
4932 | znode_t *zp = VTOZ(vp); | |
4933 | zfsvfs_t *zfsvfs = zp->z_zfsvfs; | |
4934 | int max_blksz = zfsvfs->z_max_blksz; | |
4935 | uio_t *uio = &xuio->xu_uio; | |
4936 | ssize_t size = uio->uio_resid; | |
4937 | offset_t offset = uio->uio_loffset; | |
4938 | int blksz; | |
4939 | int fullblk, i; | |
4940 | arc_buf_t *abuf; | |
4941 | ssize_t maxsize; | |
4942 | int preamble, postamble; | |
4943 | ||
4944 | if (xuio->xu_type != UIOTYPE_ZEROCOPY) | |
4945 | return (EINVAL); | |
4946 | ||
4947 | ZFS_ENTER(zfsvfs); | |
4948 | ZFS_VERIFY_ZP(zp); | |
4949 | switch (ioflag) { | |
4950 | case UIO_WRITE: | |
4951 | /* | |
4952 | * Loan out an arc_buf for write if write size is bigger than | |
4953 | * max_blksz, and the file's block size is also max_blksz. | |
4954 | */ | |
4955 | blksz = max_blksz; | |
4956 | if (size < blksz || zp->z_blksz != blksz) { | |
4957 | ZFS_EXIT(zfsvfs); | |
4958 | return (EINVAL); | |
4959 | } | |
4960 | /* | |
4961 | * Caller requests buffers for write before knowing where the | |
4962 | * write offset might be (e.g. NFS TCP write). | |
4963 | */ | |
4964 | if (offset == -1) { | |
4965 | preamble = 0; | |
4966 | } else { | |
4967 | preamble = P2PHASE(offset, blksz); | |
4968 | if (preamble) { | |
4969 | preamble = blksz - preamble; | |
4970 | size -= preamble; | |
4971 | } | |
4972 | } | |
4973 | ||
4974 | postamble = P2PHASE(size, blksz); | |
4975 | size -= postamble; | |
4976 | ||
4977 | fullblk = size / blksz; | |
4978 | (void) dmu_xuio_init(xuio, | |
4979 | (preamble != 0) + fullblk + (postamble != 0)); | |
4980 | DTRACE_PROBE3(zfs_reqzcbuf_align, int, preamble, | |
4981 | int, postamble, int, | |
4982 | (preamble != 0) + fullblk + (postamble != 0)); | |
4983 | ||
4984 | /* | |
4985 | * Have to fix iov base/len for partial buffers. They | |
4986 | * currently represent full arc_buf's. | |
4987 | */ | |
4988 | if (preamble) { | |
4989 | /* data begins in the middle of the arc_buf */ | |
4990 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4991 | blksz); | |
4992 | ASSERT(abuf); | |
4993 | (void) dmu_xuio_add(xuio, abuf, | |
4994 | blksz - preamble, preamble); | |
4995 | } | |
4996 | ||
4997 | for (i = 0; i < fullblk; i++) { | |
4998 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
4999 | blksz); | |
5000 | ASSERT(abuf); | |
5001 | (void) dmu_xuio_add(xuio, abuf, 0, blksz); | |
5002 | } | |
5003 | ||
5004 | if (postamble) { | |
5005 | /* data ends in the middle of the arc_buf */ | |
5006 | abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), | |
5007 | blksz); | |
5008 | ASSERT(abuf); | |
5009 | (void) dmu_xuio_add(xuio, abuf, 0, postamble); | |
5010 | } | |
5011 | break; | |
5012 | case UIO_READ: | |
5013 | /* | |
5014 | * Loan out an arc_buf for read if the read size is larger than | |
5015 | * the current file block size. Block alignment is not | |
5016 | * considered. Partial arc_buf will be loaned out for read. | |
5017 | */ | |
5018 | blksz = zp->z_blksz; | |
5019 | if (blksz < zcr_blksz_min) | |
5020 | blksz = zcr_blksz_min; | |
5021 | if (blksz > zcr_blksz_max) | |
5022 | blksz = zcr_blksz_max; | |
5023 | /* avoid potential complexity of dealing with it */ | |
5024 | if (blksz > max_blksz) { | |
5025 | ZFS_EXIT(zfsvfs); | |
5026 | return (EINVAL); | |
5027 | } | |
5028 | ||
5029 | maxsize = zp->z_size - uio->uio_loffset; | |
5030 | if (size > maxsize) | |
5031 | size = maxsize; | |
5032 | ||
5033 | if (size < blksz || vn_has_cached_data(vp)) { | |
5034 | ZFS_EXIT(zfsvfs); | |
5035 | return (EINVAL); | |
5036 | } | |
5037 | break; | |
5038 | default: | |
5039 | ZFS_EXIT(zfsvfs); | |
5040 | return (EINVAL); | |
5041 | } | |
5042 | ||
5043 | uio->uio_extflg = UIO_XUIO; | |
5044 | XUIO_XUZC_RW(xuio) = ioflag; | |
5045 | ZFS_EXIT(zfsvfs); | |
5046 | return (0); | |
5047 | } | |
5048 | ||
5049 | /*ARGSUSED*/ | |
5050 | static int | |
5051 | zfs_retzcbuf(vnode_t *vp, xuio_t *xuio, cred_t *cr, caller_context_t *ct) | |
5052 | { | |
5053 | int i; | |
5054 | arc_buf_t *abuf; | |
5055 | int ioflag = XUIO_XUZC_RW(xuio); | |
5056 | ||
5057 | ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); | |
5058 | ||
5059 | i = dmu_xuio_cnt(xuio); | |
5060 | while (i-- > 0) { | |
5061 | abuf = dmu_xuio_arcbuf(xuio, i); | |
5062 | /* | |
5063 | * if abuf == NULL, it must be a write buffer | |
5064 | * that has been returned in zfs_write(). | |
5065 | */ | |
5066 | if (abuf) | |
5067 | dmu_return_arcbuf(abuf); | |
5068 | ASSERT(abuf || ioflag == UIO_WRITE); | |
5069 | } | |
5070 | ||
5071 | dmu_xuio_fini(xuio); | |
5072 | return (0); | |
5073 | } | |
5074 | ||
34dc7c2f BB |
5075 | /* |
5076 | * Predeclare these here so that the compiler assumes that | |
5077 | * this is an "old style" function declaration that does | |
5078 | * not include arguments => we won't get type mismatch errors | |
5079 | * in the initializations that follow. | |
5080 | */ | |
5081 | static int zfs_inval(); | |
5082 | static int zfs_isdir(); | |
5083 | ||
5084 | static int | |
5085 | zfs_inval() | |
5086 | { | |
5087 | return (EINVAL); | |
5088 | } | |
5089 | ||
5090 | static int | |
5091 | zfs_isdir() | |
5092 | { | |
5093 | return (EISDIR); | |
5094 | } | |
5095 | /* | |
5096 | * Directory vnode operations template | |
5097 | */ | |
5098 | vnodeops_t *zfs_dvnodeops; | |
5099 | const fs_operation_def_t zfs_dvnodeops_template[] = { | |
5100 | VOPNAME_OPEN, { .vop_open = zfs_open }, | |
5101 | VOPNAME_CLOSE, { .vop_close = zfs_close }, | |
5102 | VOPNAME_READ, { .error = zfs_isdir }, | |
5103 | VOPNAME_WRITE, { .error = zfs_isdir }, | |
5104 | VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl }, | |
5105 | VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, | |
5106 | VOPNAME_SETATTR, { .vop_setattr = zfs_setattr }, | |
5107 | VOPNAME_ACCESS, { .vop_access = zfs_access }, | |
5108 | VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup }, | |
5109 | VOPNAME_CREATE, { .vop_create = zfs_create }, | |
5110 | VOPNAME_REMOVE, { .vop_remove = zfs_remove }, | |
5111 | VOPNAME_LINK, { .vop_link = zfs_link }, | |
5112 | VOPNAME_RENAME, { .vop_rename = zfs_rename }, | |
5113 | VOPNAME_MKDIR, { .vop_mkdir = zfs_mkdir }, | |
5114 | VOPNAME_RMDIR, { .vop_rmdir = zfs_rmdir }, | |
5115 | VOPNAME_READDIR, { .vop_readdir = zfs_readdir }, | |
5116 | VOPNAME_SYMLINK, { .vop_symlink = zfs_symlink }, | |
5117 | VOPNAME_FSYNC, { .vop_fsync = zfs_fsync }, | |
5118 | VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, | |
5119 | VOPNAME_FID, { .vop_fid = zfs_fid }, | |
5120 | VOPNAME_SEEK, { .vop_seek = zfs_seek }, | |
5121 | VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, | |
5122 | VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr }, | |
5123 | VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr }, | |
5124 | VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, | |
5125 | NULL, NULL | |
5126 | }; | |
5127 | ||
5128 | /* | |
5129 | * Regular file vnode operations template | |
5130 | */ | |
5131 | vnodeops_t *zfs_fvnodeops; | |
5132 | const fs_operation_def_t zfs_fvnodeops_template[] = { | |
5133 | VOPNAME_OPEN, { .vop_open = zfs_open }, | |
5134 | VOPNAME_CLOSE, { .vop_close = zfs_close }, | |
5135 | VOPNAME_READ, { .vop_read = zfs_read }, | |
5136 | VOPNAME_WRITE, { .vop_write = zfs_write }, | |
5137 | VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl }, | |
5138 | VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, | |
5139 | VOPNAME_SETATTR, { .vop_setattr = zfs_setattr }, | |
5140 | VOPNAME_ACCESS, { .vop_access = zfs_access }, | |
5141 | VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup }, | |
5142 | VOPNAME_RENAME, { .vop_rename = zfs_rename }, | |
5143 | VOPNAME_FSYNC, { .vop_fsync = zfs_fsync }, | |
5144 | VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, | |
5145 | VOPNAME_FID, { .vop_fid = zfs_fid }, | |
5146 | VOPNAME_SEEK, { .vop_seek = zfs_seek }, | |
5147 | VOPNAME_FRLOCK, { .vop_frlock = zfs_frlock }, | |
5148 | VOPNAME_SPACE, { .vop_space = zfs_space }, | |
5149 | VOPNAME_GETPAGE, { .vop_getpage = zfs_getpage }, | |
5150 | VOPNAME_PUTPAGE, { .vop_putpage = zfs_putpage }, | |
5151 | VOPNAME_MAP, { .vop_map = zfs_map }, | |
5152 | VOPNAME_ADDMAP, { .vop_addmap = zfs_addmap }, | |
5153 | VOPNAME_DELMAP, { .vop_delmap = zfs_delmap }, | |
5154 | VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, | |
5155 | VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr }, | |
5156 | VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr }, | |
5157 | VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, | |
428870ff BB |
5158 | VOPNAME_REQZCBUF, { .vop_reqzcbuf = zfs_reqzcbuf }, |
5159 | VOPNAME_RETZCBUF, { .vop_retzcbuf = zfs_retzcbuf }, | |
34dc7c2f BB |
5160 | NULL, NULL |
5161 | }; | |
5162 | ||
5163 | /* | |
5164 | * Symbolic link vnode operations template | |
5165 | */ | |
5166 | vnodeops_t *zfs_symvnodeops; | |
5167 | const fs_operation_def_t zfs_symvnodeops_template[] = { | |
5168 | VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, | |
5169 | VOPNAME_SETATTR, { .vop_setattr = zfs_setattr }, | |
5170 | VOPNAME_ACCESS, { .vop_access = zfs_access }, | |
5171 | VOPNAME_RENAME, { .vop_rename = zfs_rename }, | |
5172 | VOPNAME_READLINK, { .vop_readlink = zfs_readlink }, | |
5173 | VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, | |
5174 | VOPNAME_FID, { .vop_fid = zfs_fid }, | |
5175 | VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, | |
5176 | VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, | |
5177 | NULL, NULL | |
5178 | }; | |
5179 | ||
9babb374 BB |
5180 | /* |
5181 | * special share hidden files vnode operations template | |
5182 | */ | |
5183 | vnodeops_t *zfs_sharevnodeops; | |
5184 | const fs_operation_def_t zfs_sharevnodeops_template[] = { | |
5185 | VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, | |
5186 | VOPNAME_ACCESS, { .vop_access = zfs_access }, | |
5187 | VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, | |
5188 | VOPNAME_FID, { .vop_fid = zfs_fid }, | |
5189 | VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, | |
5190 | VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr }, | |
5191 | VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr }, | |
5192 | VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, | |
5193 | NULL, NULL | |
5194 | }; | |
5195 | ||
34dc7c2f BB |
5196 | /* |
5197 | * Extended attribute directory vnode operations template | |
5198 | * This template is identical to the directory vnodes | |
5199 | * operation template except for restricted operations: | |
5200 | * VOP_MKDIR() | |
5201 | * VOP_SYMLINK() | |
5202 | * Note that there are other restrictions embedded in: | |
5203 | * zfs_create() - restrict type to VREG | |
5204 | * zfs_link() - no links into/out of attribute space | |
5205 | * zfs_rename() - no moves into/out of attribute space | |
5206 | */ | |
5207 | vnodeops_t *zfs_xdvnodeops; | |
5208 | const fs_operation_def_t zfs_xdvnodeops_template[] = { | |
5209 | VOPNAME_OPEN, { .vop_open = zfs_open }, | |
5210 | VOPNAME_CLOSE, { .vop_close = zfs_close }, | |
5211 | VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl }, | |
5212 | VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, | |
5213 | VOPNAME_SETATTR, { .vop_setattr = zfs_setattr }, | |
5214 | VOPNAME_ACCESS, { .vop_access = zfs_access }, | |
5215 | VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup }, | |
5216 | VOPNAME_CREATE, { .vop_create = zfs_create }, | |
5217 | VOPNAME_REMOVE, { .vop_remove = zfs_remove }, | |
5218 | VOPNAME_LINK, { .vop_link = zfs_link }, | |
5219 | VOPNAME_RENAME, { .vop_rename = zfs_rename }, | |
5220 | VOPNAME_MKDIR, { .error = zfs_inval }, | |
5221 | VOPNAME_RMDIR, { .vop_rmdir = zfs_rmdir }, | |
5222 | VOPNAME_READDIR, { .vop_readdir = zfs_readdir }, | |
5223 | VOPNAME_SYMLINK, { .error = zfs_inval }, | |
5224 | VOPNAME_FSYNC, { .vop_fsync = zfs_fsync }, | |
5225 | VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, | |
5226 | VOPNAME_FID, { .vop_fid = zfs_fid }, | |
5227 | VOPNAME_SEEK, { .vop_seek = zfs_seek }, | |
5228 | VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, | |
5229 | VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr }, | |
5230 | VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr }, | |
5231 | VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, | |
5232 | NULL, NULL | |
5233 | }; | |
5234 | ||
5235 | /* | |
5236 | * Error vnode operations template | |
5237 | */ | |
5238 | vnodeops_t *zfs_evnodeops; | |
5239 | const fs_operation_def_t zfs_evnodeops_template[] = { | |
5240 | VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, | |
5241 | VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, | |
5242 | NULL, NULL | |
5243 | }; |