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 /*
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2013 by Delphix. All rights reserved.
24 */
25
26 /* Portions Copyright 2007 Jeremy Teo */
27 /* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */
28
29 #ifdef _KERNEL
30 #include <sys/types.h>
31 #include <sys/param.h>
32 #include <sys/time.h>
33 #include <sys/systm.h>
34 #include <sys/sysmacros.h>
35 #include <sys/resource.h>
36 #include <sys/mntent.h>
37 #include <sys/u8_textprep.h>
38 #include <sys/dsl_dataset.h>
39 #include <sys/vfs.h>
40 #include <sys/vnode.h>
41 #include <sys/file.h>
42 #include <sys/kmem.h>
43 #include <sys/errno.h>
44 #include <sys/unistd.h>
45 #include <sys/atomic.h>
46 #include <sys/zfs_dir.h>
47 #include <sys/zfs_acl.h>
48 #include <sys/zfs_ioctl.h>
49 #include <sys/zfs_rlock.h>
50 #include <sys/zfs_fuid.h>
51 #include <sys/dnode.h>
52 #include <sys/fs/zfs.h>
53 #include <sys/kidmap.h>
54 #endif /* _KERNEL */
55
56 #include <sys/dmu.h>
57 #include <sys/refcount.h>
58 #include <sys/stat.h>
59 #include <sys/zap.h>
60 #include <sys/zfs_znode.h>
61 #include <sys/sa.h>
62 #include <sys/zfs_sa.h>
63 #include <sys/zfs_stat.h>
64 #include <sys/refcount.h>
65
66 #include "zfs_prop.h"
67 #include "zfs_comutil.h"
68
69 /* Used by fstat(1). */
70 SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD,
71 SYSCTL_NULL_INT_PTR, sizeof(znode_t), "sizeof(znode_t)");
72
73 /*
74 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
75 * turned on when DEBUG is also defined.
76 */
77 #ifdef DEBUG
78 #define ZNODE_STATS
79 #endif /* DEBUG */
80
81 #ifdef ZNODE_STATS
82 #define ZNODE_STAT_ADD(stat) ((stat)++)
83 #else
84 #define ZNODE_STAT_ADD(stat) /* nothing */
85 #endif /* ZNODE_STATS */
86
87 /*
88 * Functions needed for userland (ie: libzpool) are not put under
89 * #ifdef_KERNEL; the rest of the functions have dependencies
90 * (such as VFS logic) that will not compile easily in userland.
91 */
92 #ifdef _KERNEL
93 /*
94 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
95 * be freed before it can be safely accessed.
96 */
97 krwlock_t zfsvfs_lock;
98
99 static kmem_cache_t *znode_cache = NULL;
100
101 /*ARGSUSED*/
102 static void
znode_evict_error(dmu_buf_t * dbuf,void * user_ptr)103 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
104 {
105 /*
106 * We should never drop all dbuf refs without first clearing
107 * the eviction callback.
108 */
109 panic("evicting znode %p\n", user_ptr);
110 }
111
112 extern struct vop_vector zfs_vnodeops;
113 extern struct vop_vector zfs_fifoops;
114 extern struct vop_vector zfs_shareops;
115
116 static int
zfs_znode_cache_constructor(void * buf,void * arg,int kmflags)117 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
118 {
119 znode_t *zp = buf;
120
121 POINTER_INVALIDATE(&zp->z_zfsvfs);
122
123 list_link_init(&zp->z_link_node);
124
125 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
126 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
127 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
128 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
129
130 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
131 avl_create(&zp->z_range_avl, zfs_range_compare,
132 sizeof (rl_t), offsetof(rl_t, r_node));
133
134 zp->z_dirlocks = NULL;
135 zp->z_acl_cached = NULL;
136 zp->z_vnode = NULL;
137 zp->z_moved = 0;
138 return (0);
139 }
140
141 /*ARGSUSED*/
142 static void
zfs_znode_cache_destructor(void * buf,void * arg)143 zfs_znode_cache_destructor(void *buf, void *arg)
144 {
145 znode_t *zp = buf;
146
147 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
148 ASSERT(ZTOV(zp) == NULL);
149 vn_free(ZTOV(zp));
150 ASSERT(!list_link_active(&zp->z_link_node));
151 mutex_destroy(&zp->z_lock);
152 rw_destroy(&zp->z_parent_lock);
153 rw_destroy(&zp->z_name_lock);
154 mutex_destroy(&zp->z_acl_lock);
155 avl_destroy(&zp->z_range_avl);
156 mutex_destroy(&zp->z_range_lock);
157
158 ASSERT(zp->z_dirlocks == NULL);
159 ASSERT(zp->z_acl_cached == NULL);
160 }
161
162 #ifdef ZNODE_STATS
163 static struct {
164 uint64_t zms_zfsvfs_invalid;
165 uint64_t zms_zfsvfs_recheck1;
166 uint64_t zms_zfsvfs_unmounted;
167 uint64_t zms_zfsvfs_recheck2;
168 uint64_t zms_obj_held;
169 uint64_t zms_vnode_locked;
170 uint64_t zms_not_only_dnlc;
171 } znode_move_stats;
172 #endif /* ZNODE_STATS */
173
174 #ifdef sun
175 static void
zfs_znode_move_impl(znode_t * ozp,znode_t * nzp)176 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
177 {
178 vnode_t *vp;
179
180 /* Copy fields. */
181 nzp->z_zfsvfs = ozp->z_zfsvfs;
182
183 /* Swap vnodes. */
184 vp = nzp->z_vnode;
185 nzp->z_vnode = ozp->z_vnode;
186 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
187 ZTOV(ozp)->v_data = ozp;
188 ZTOV(nzp)->v_data = nzp;
189
190 nzp->z_id = ozp->z_id;
191 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
192 ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
193 nzp->z_unlinked = ozp->z_unlinked;
194 nzp->z_atime_dirty = ozp->z_atime_dirty;
195 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
196 nzp->z_blksz = ozp->z_blksz;
197 nzp->z_seq = ozp->z_seq;
198 nzp->z_mapcnt = ozp->z_mapcnt;
199 nzp->z_gen = ozp->z_gen;
200 nzp->z_sync_cnt = ozp->z_sync_cnt;
201 nzp->z_is_sa = ozp->z_is_sa;
202 nzp->z_sa_hdl = ozp->z_sa_hdl;
203 bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2);
204 nzp->z_links = ozp->z_links;
205 nzp->z_size = ozp->z_size;
206 nzp->z_pflags = ozp->z_pflags;
207 nzp->z_uid = ozp->z_uid;
208 nzp->z_gid = ozp->z_gid;
209 nzp->z_mode = ozp->z_mode;
210
211 /*
212 * Since this is just an idle znode and kmem is already dealing with
213 * memory pressure, release any cached ACL.
214 */
215 if (ozp->z_acl_cached) {
216 zfs_acl_free(ozp->z_acl_cached);
217 ozp->z_acl_cached = NULL;
218 }
219
220 sa_set_userp(nzp->z_sa_hdl, nzp);
221
222 /*
223 * Invalidate the original znode by clearing fields that provide a
224 * pointer back to the znode. Set the low bit of the vfs pointer to
225 * ensure that zfs_znode_move() recognizes the znode as invalid in any
226 * subsequent callback.
227 */
228 ozp->z_sa_hdl = NULL;
229 POINTER_INVALIDATE(&ozp->z_zfsvfs);
230
231 /*
232 * Mark the znode.
233 */
234 nzp->z_moved = 1;
235 ozp->z_moved = (uint8_t)-1;
236 }
237
238 /*ARGSUSED*/
239 static kmem_cbrc_t
zfs_znode_move(void * buf,void * newbuf,size_t size,void * arg)240 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
241 {
242 znode_t *ozp = buf, *nzp = newbuf;
243 zfsvfs_t *zfsvfs;
244 vnode_t *vp;
245
246 /*
247 * The znode is on the file system's list of known znodes if the vfs
248 * pointer is valid. We set the low bit of the vfs pointer when freeing
249 * the znode to invalidate it, and the memory patterns written by kmem
250 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
251 * created znode sets the vfs pointer last of all to indicate that the
252 * znode is known and in a valid state to be moved by this function.
253 */
254 zfsvfs = ozp->z_zfsvfs;
255 if (!POINTER_IS_VALID(zfsvfs)) {
256 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
257 return (KMEM_CBRC_DONT_KNOW);
258 }
259
260 /*
261 * Close a small window in which it's possible that the filesystem could
262 * be unmounted and freed, and zfsvfs, though valid in the previous
263 * statement, could point to unrelated memory by the time we try to
264 * prevent the filesystem from being unmounted.
265 */
266 rw_enter(&zfsvfs_lock, RW_WRITER);
267 if (zfsvfs != ozp->z_zfsvfs) {
268 rw_exit(&zfsvfs_lock);
269 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
270 return (KMEM_CBRC_DONT_KNOW);
271 }
272
273 /*
274 * If the znode is still valid, then so is the file system. We know that
275 * no valid file system can be freed while we hold zfsvfs_lock, so we
276 * can safely ensure that the filesystem is not and will not be
277 * unmounted. The next statement is equivalent to ZFS_ENTER().
278 */
279 rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
280 if (zfsvfs->z_unmounted) {
281 ZFS_EXIT(zfsvfs);
282 rw_exit(&zfsvfs_lock);
283 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
284 return (KMEM_CBRC_DONT_KNOW);
285 }
286 rw_exit(&zfsvfs_lock);
287
288 mutex_enter(&zfsvfs->z_znodes_lock);
289 /*
290 * Recheck the vfs pointer in case the znode was removed just before
291 * acquiring the lock.
292 */
293 if (zfsvfs != ozp->z_zfsvfs) {
294 mutex_exit(&zfsvfs->z_znodes_lock);
295 ZFS_EXIT(zfsvfs);
296 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
297 return (KMEM_CBRC_DONT_KNOW);
298 }
299
300 /*
301 * At this point we know that as long as we hold z_znodes_lock, the
302 * znode cannot be freed and fields within the znode can be safely
303 * accessed. Now, prevent a race with zfs_zget().
304 */
305 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
306 mutex_exit(&zfsvfs->z_znodes_lock);
307 ZFS_EXIT(zfsvfs);
308 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
309 return (KMEM_CBRC_LATER);
310 }
311
312 vp = ZTOV(ozp);
313 if (mutex_tryenter(&vp->v_lock) == 0) {
314 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
315 mutex_exit(&zfsvfs->z_znodes_lock);
316 ZFS_EXIT(zfsvfs);
317 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
318 return (KMEM_CBRC_LATER);
319 }
320
321 /* Only move znodes that are referenced _only_ by the DNLC. */
322 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
323 mutex_exit(&vp->v_lock);
324 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
325 mutex_exit(&zfsvfs->z_znodes_lock);
326 ZFS_EXIT(zfsvfs);
327 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
328 return (KMEM_CBRC_LATER);
329 }
330
331 /*
332 * The znode is known and in a valid state to move. We're holding the
333 * locks needed to execute the critical section.
334 */
335 zfs_znode_move_impl(ozp, nzp);
336 mutex_exit(&vp->v_lock);
337 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
338
339 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
340 mutex_exit(&zfsvfs->z_znodes_lock);
341 ZFS_EXIT(zfsvfs);
342
343 return (KMEM_CBRC_YES);
344 }
345 #endif /* sun */
346
347 void
zfs_znode_init(void)348 zfs_znode_init(void)
349 {
350 /*
351 * Initialize zcache
352 */
353 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
354 ASSERT(znode_cache == NULL);
355 znode_cache = kmem_cache_create("zfs_znode_cache",
356 sizeof (znode_t), 0, zfs_znode_cache_constructor,
357 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
358 kmem_cache_set_move(znode_cache, zfs_znode_move);
359 }
360
361 void
zfs_znode_fini(void)362 zfs_znode_fini(void)
363 {
364 #ifdef sun
365 /*
366 * Cleanup vfs & vnode ops
367 */
368 zfs_remove_op_tables();
369 #endif /* sun */
370
371 /*
372 * Cleanup zcache
373 */
374 if (znode_cache)
375 kmem_cache_destroy(znode_cache);
376 znode_cache = NULL;
377 rw_destroy(&zfsvfs_lock);
378 }
379
380 #ifdef sun
381 struct vnodeops *zfs_dvnodeops;
382 struct vnodeops *zfs_fvnodeops;
383 struct vnodeops *zfs_symvnodeops;
384 struct vnodeops *zfs_xdvnodeops;
385 struct vnodeops *zfs_evnodeops;
386 struct vnodeops *zfs_sharevnodeops;
387
388 void
zfs_remove_op_tables()389 zfs_remove_op_tables()
390 {
391 /*
392 * Remove vfs ops
393 */
394 ASSERT(zfsfstype);
395 (void) vfs_freevfsops_by_type(zfsfstype);
396 zfsfstype = 0;
397
398 /*
399 * Remove vnode ops
400 */
401 if (zfs_dvnodeops)
402 vn_freevnodeops(zfs_dvnodeops);
403 if (zfs_fvnodeops)
404 vn_freevnodeops(zfs_fvnodeops);
405 if (zfs_symvnodeops)
406 vn_freevnodeops(zfs_symvnodeops);
407 if (zfs_xdvnodeops)
408 vn_freevnodeops(zfs_xdvnodeops);
409 if (zfs_evnodeops)
410 vn_freevnodeops(zfs_evnodeops);
411 if (zfs_sharevnodeops)
412 vn_freevnodeops(zfs_sharevnodeops);
413
414 zfs_dvnodeops = NULL;
415 zfs_fvnodeops = NULL;
416 zfs_symvnodeops = NULL;
417 zfs_xdvnodeops = NULL;
418 zfs_evnodeops = NULL;
419 zfs_sharevnodeops = NULL;
420 }
421
422 extern const fs_operation_def_t zfs_dvnodeops_template[];
423 extern const fs_operation_def_t zfs_fvnodeops_template[];
424 extern const fs_operation_def_t zfs_xdvnodeops_template[];
425 extern const fs_operation_def_t zfs_symvnodeops_template[];
426 extern const fs_operation_def_t zfs_evnodeops_template[];
427 extern const fs_operation_def_t zfs_sharevnodeops_template[];
428
429 int
zfs_create_op_tables()430 zfs_create_op_tables()
431 {
432 int error;
433
434 /*
435 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
436 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
437 * In this case we just return as the ops vectors are already set up.
438 */
439 if (zfs_dvnodeops)
440 return (0);
441
442 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
443 &zfs_dvnodeops);
444 if (error)
445 return (error);
446
447 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
448 &zfs_fvnodeops);
449 if (error)
450 return (error);
451
452 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
453 &zfs_symvnodeops);
454 if (error)
455 return (error);
456
457 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
458 &zfs_xdvnodeops);
459 if (error)
460 return (error);
461
462 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
463 &zfs_evnodeops);
464 if (error)
465 return (error);
466
467 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
468 &zfs_sharevnodeops);
469
470 return (error);
471 }
472 #endif /* sun */
473
474 int
zfs_create_share_dir(zfsvfs_t * zfsvfs,dmu_tx_t * tx)475 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
476 {
477 zfs_acl_ids_t acl_ids;
478 vattr_t vattr;
479 znode_t *sharezp;
480 znode_t *zp;
481 int error;
482
483 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
484 vattr.va_type = VDIR;
485 vattr.va_mode = S_IFDIR|0555;
486 vattr.va_uid = crgetuid(kcred);
487 vattr.va_gid = crgetgid(kcred);
488
489 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
490 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
491 sharezp->z_moved = 0;
492 sharezp->z_unlinked = 0;
493 sharezp->z_atime_dirty = 0;
494 sharezp->z_zfsvfs = zfsvfs;
495 sharezp->z_is_sa = zfsvfs->z_use_sa;
496
497 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
498 kcred, NULL, &acl_ids));
499 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
500 ASSERT3P(zp, ==, sharezp);
501 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
502 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
503 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
504 zfsvfs->z_shares_dir = sharezp->z_id;
505
506 zfs_acl_ids_free(&acl_ids);
507 sa_handle_destroy(sharezp->z_sa_hdl);
508 kmem_cache_free(znode_cache, sharezp);
509
510 return (error);
511 }
512
513 /*
514 * define a couple of values we need available
515 * for both 64 and 32 bit environments.
516 */
517 #ifndef NBITSMINOR64
518 #define NBITSMINOR64 32
519 #endif
520 #ifndef MAXMAJ64
521 #define MAXMAJ64 0xffffffffUL
522 #endif
523 #ifndef MAXMIN64
524 #define MAXMIN64 0xffffffffUL
525 #endif
526
527 /*
528 * Create special expldev for ZFS private use.
529 * Can't use standard expldev since it doesn't do
530 * what we want. The standard expldev() takes a
531 * dev32_t in LP64 and expands it to a long dev_t.
532 * We need an interface that takes a dev32_t in ILP32
533 * and expands it to a long dev_t.
534 */
535 static uint64_t
zfs_expldev(dev_t dev)536 zfs_expldev(dev_t dev)
537 {
538 return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev));
539 }
540 /*
541 * Special cmpldev for ZFS private use.
542 * Can't use standard cmpldev since it takes
543 * a long dev_t and compresses it to dev32_t in
544 * LP64. We need to do a compaction of a long dev_t
545 * to a dev32_t in ILP32.
546 */
547 dev_t
zfs_cmpldev(uint64_t dev)548 zfs_cmpldev(uint64_t dev)
549 {
550 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64)));
551 }
552
553 static void
zfs_znode_sa_init(zfsvfs_t * zfsvfs,znode_t * zp,dmu_buf_t * db,dmu_object_type_t obj_type,sa_handle_t * sa_hdl)554 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
555 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
556 {
557 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
558 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
559
560 mutex_enter(&zp->z_lock);
561
562 ASSERT(zp->z_sa_hdl == NULL);
563 ASSERT(zp->z_acl_cached == NULL);
564 if (sa_hdl == NULL) {
565 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
566 SA_HDL_SHARED, &zp->z_sa_hdl));
567 } else {
568 zp->z_sa_hdl = sa_hdl;
569 sa_set_userp(sa_hdl, zp);
570 }
571
572 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
573
574 /*
575 * Slap on VROOT if we are the root znode unless we are the root
576 * node of a snapshot mounted under .zfs.
577 */
578 if (zp->z_id == zfsvfs->z_root && zfsvfs->z_parent == zfsvfs)
579 ZTOV(zp)->v_flag |= VROOT;
580
581 mutex_exit(&zp->z_lock);
582 vn_exists(ZTOV(zp));
583 }
584
585 void
zfs_znode_dmu_fini(znode_t * zp)586 zfs_znode_dmu_fini(znode_t *zp)
587 {
588 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
589 zp->z_unlinked ||
590 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
591
592 sa_handle_destroy(zp->z_sa_hdl);
593 zp->z_sa_hdl = NULL;
594 }
595
596 static void
zfs_vnode_forget(vnode_t * vp)597 zfs_vnode_forget(vnode_t *vp)
598 {
599
600 /* copied from insmntque_stddtr */
601 vp->v_data = NULL;
602 vp->v_op = &dead_vnodeops;
603 vgone(vp);
604 vput(vp);
605 }
606
607 /*
608 * Construct a new znode/vnode and intialize.
609 *
610 * This does not do a call to dmu_set_user() that is
611 * up to the caller to do, in case you don't want to
612 * return the znode
613 */
614 static znode_t *
zfs_znode_alloc(zfsvfs_t * zfsvfs,dmu_buf_t * db,int blksz,dmu_object_type_t obj_type,sa_handle_t * hdl)615 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
616 dmu_object_type_t obj_type, sa_handle_t *hdl)
617 {
618 znode_t *zp;
619 vnode_t *vp;
620 uint64_t mode;
621 uint64_t parent;
622 sa_bulk_attr_t bulk[9];
623 int count = 0;
624 int error;
625
626 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
627
628 KASSERT(curthread->td_vp_reserv > 0,
629 ("zfs_znode_alloc: getnewvnode without any vnodes reserved"));
630 error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp);
631 if (error != 0) {
632 kmem_cache_free(znode_cache, zp);
633 return (NULL);
634 }
635 zp->z_vnode = vp;
636 vp->v_data = zp;
637
638 ASSERT(zp->z_dirlocks == NULL);
639 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
640 zp->z_moved = 0;
641
642 /*
643 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
644 * the zfs_znode_move() callback.
645 */
646 zp->z_sa_hdl = NULL;
647 zp->z_unlinked = 0;
648 zp->z_atime_dirty = 0;
649 zp->z_mapcnt = 0;
650 zp->z_id = db->db_object;
651 zp->z_blksz = blksz;
652 zp->z_seq = 0x7A4653;
653 zp->z_sync_cnt = 0;
654
655 vp = ZTOV(zp);
656
657 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
658
659 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
660 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
661 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
662 &zp->z_size, 8);
663 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
664 &zp->z_links, 8);
665 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
666 &zp->z_pflags, 8);
667 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
668 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
669 &zp->z_atime, 16);
670 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
671 &zp->z_uid, 8);
672 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
673 &zp->z_gid, 8);
674
675 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
676 if (hdl == NULL)
677 sa_handle_destroy(zp->z_sa_hdl);
678 zfs_vnode_forget(vp);
679 zp->z_vnode = NULL;
680 kmem_cache_free(znode_cache, zp);
681 return (NULL);
682 }
683
684 zp->z_mode = mode;
685
686 vp->v_type = IFTOVT((mode_t)mode);
687
688 switch (vp->v_type) {
689 case VDIR:
690 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
691 break;
692 #ifdef sun
693 case VBLK:
694 case VCHR:
695 {
696 uint64_t rdev;
697 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
698 &rdev, sizeof (rdev)) == 0);
699
700 vp->v_rdev = zfs_cmpldev(rdev);
701 }
702 break;
703 #endif /* sun */
704 case VFIFO:
705 #ifdef sun
706 case VSOCK:
707 case VDOOR:
708 #endif /* sun */
709 vp->v_op = &zfs_fifoops;
710 break;
711 case VREG:
712 if (parent == zfsvfs->z_shares_dir) {
713 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
714 vp->v_op = &zfs_shareops;
715 }
716 break;
717 #ifdef sun
718 case VLNK:
719 vn_setops(vp, zfs_symvnodeops);
720 break;
721 default:
722 vn_setops(vp, zfs_evnodeops);
723 break;
724 #endif /* sun */
725 }
726
727 mutex_enter(&zfsvfs->z_znodes_lock);
728 list_insert_tail(&zfsvfs->z_all_znodes, zp);
729 membar_producer();
730 /*
731 * Everything else must be valid before assigning z_zfsvfs makes the
732 * znode eligible for zfs_znode_move().
733 */
734 zp->z_zfsvfs = zfsvfs;
735 mutex_exit(&zfsvfs->z_znodes_lock);
736
737 /*
738 * Acquire vnode lock before making it available to the world.
739 */
740 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
741 VN_LOCK_AREC(vp);
742 if (vp->v_type != VFIFO)
743 VN_LOCK_ASHARE(vp);
744
745 #ifdef illumos
746 VFS_HOLD(zfsvfs->z_vfs);
747 #endif
748 return (zp);
749 }
750
751 static uint64_t empty_xattr;
752 static uint64_t pad[4];
753 static zfs_acl_phys_t acl_phys;
754 /*
755 * Create a new DMU object to hold a zfs znode.
756 *
757 * IN: dzp - parent directory for new znode
758 * vap - file attributes for new znode
759 * tx - dmu transaction id for zap operations
760 * cr - credentials of caller
761 * flag - flags:
762 * IS_ROOT_NODE - new object will be root
763 * IS_XATTR - new object is an attribute
764 * bonuslen - length of bonus buffer
765 * setaclp - File/Dir initial ACL
766 * fuidp - Tracks fuid allocation.
767 *
768 * OUT: zpp - allocated znode
769 *
770 */
771 void
zfs_mknode(znode_t * dzp,vattr_t * vap,dmu_tx_t * tx,cred_t * cr,uint_t flag,znode_t ** zpp,zfs_acl_ids_t * acl_ids)772 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
773 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
774 {
775 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
776 uint64_t mode, size, links, parent, pflags;
777 uint64_t dzp_pflags = 0;
778 uint64_t rdev = 0;
779 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
780 dmu_buf_t *db;
781 timestruc_t now;
782 uint64_t gen, obj;
783 int err;
784 int bonuslen;
785 sa_handle_t *sa_hdl;
786 dmu_object_type_t obj_type;
787 sa_bulk_attr_t sa_attrs[ZPL_END];
788 int cnt = 0;
789 zfs_acl_locator_cb_t locate = { 0 };
790
791 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
792
793 if (zfsvfs->z_replay) {
794 obj = vap->va_nodeid;
795 now = vap->va_ctime; /* see zfs_replay_create() */
796 gen = vap->va_nblocks; /* ditto */
797 } else {
798 obj = 0;
799 gethrestime(&now);
800 gen = dmu_tx_get_txg(tx);
801 }
802
803 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
804 bonuslen = (obj_type == DMU_OT_SA) ?
805 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
806
807 /*
808 * Create a new DMU object.
809 */
810 /*
811 * There's currently no mechanism for pre-reading the blocks that will
812 * be needed to allocate a new object, so we accept the small chance
813 * that there will be an i/o error and we will fail one of the
814 * assertions below.
815 */
816 if (vap->va_type == VDIR) {
817 if (zfsvfs->z_replay) {
818 VERIFY0(zap_create_claim_norm(zfsvfs->z_os, obj,
819 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
820 obj_type, bonuslen, tx));
821 } else {
822 obj = zap_create_norm(zfsvfs->z_os,
823 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
824 obj_type, bonuslen, tx);
825 }
826 } else {
827 if (zfsvfs->z_replay) {
828 VERIFY0(dmu_object_claim(zfsvfs->z_os, obj,
829 DMU_OT_PLAIN_FILE_CONTENTS, 0,
830 obj_type, bonuslen, tx));
831 } else {
832 obj = dmu_object_alloc(zfsvfs->z_os,
833 DMU_OT_PLAIN_FILE_CONTENTS, 0,
834 obj_type, bonuslen, tx);
835 }
836 }
837
838 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
839 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
840
841 /*
842 * If this is the root, fix up the half-initialized parent pointer
843 * to reference the just-allocated physical data area.
844 */
845 if (flag & IS_ROOT_NODE) {
846 dzp->z_id = obj;
847 } else {
848 dzp_pflags = dzp->z_pflags;
849 }
850
851 /*
852 * If parent is an xattr, so am I.
853 */
854 if (dzp_pflags & ZFS_XATTR) {
855 flag |= IS_XATTR;
856 }
857
858 if (zfsvfs->z_use_fuids)
859 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
860 else
861 pflags = 0;
862
863 if (vap->va_type == VDIR) {
864 size = 2; /* contents ("." and "..") */
865 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
866 } else {
867 size = links = 0;
868 }
869
870 if (vap->va_type == VBLK || vap->va_type == VCHR) {
871 rdev = zfs_expldev(vap->va_rdev);
872 }
873
874 parent = dzp->z_id;
875 mode = acl_ids->z_mode;
876 if (flag & IS_XATTR)
877 pflags |= ZFS_XATTR;
878
879 /*
880 * No execs denied will be deterimed when zfs_mode_compute() is called.
881 */
882 pflags |= acl_ids->z_aclp->z_hints &
883 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
884 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
885
886 ZFS_TIME_ENCODE(&now, crtime);
887 ZFS_TIME_ENCODE(&now, ctime);
888
889 if (vap->va_mask & AT_ATIME) {
890 ZFS_TIME_ENCODE(&vap->va_atime, atime);
891 } else {
892 ZFS_TIME_ENCODE(&now, atime);
893 }
894
895 if (vap->va_mask & AT_MTIME) {
896 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
897 } else {
898 ZFS_TIME_ENCODE(&now, mtime);
899 }
900
901 /* Now add in all of the "SA" attributes */
902 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
903 &sa_hdl));
904
905 /*
906 * Setup the array of attributes to be replaced/set on the new file
907 *
908 * order for DMU_OT_ZNODE is critical since it needs to be constructed
909 * in the old znode_phys_t format. Don't change this ordering
910 */
911
912 if (obj_type == DMU_OT_ZNODE) {
913 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
914 NULL, &atime, 16);
915 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
916 NULL, &mtime, 16);
917 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
918 NULL, &ctime, 16);
919 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
920 NULL, &crtime, 16);
921 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
922 NULL, &gen, 8);
923 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
924 NULL, &mode, 8);
925 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
926 NULL, &size, 8);
927 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
928 NULL, &parent, 8);
929 } else {
930 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
931 NULL, &mode, 8);
932 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
933 NULL, &size, 8);
934 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
935 NULL, &gen, 8);
936 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
937 &acl_ids->z_fuid, 8);
938 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
939 &acl_ids->z_fgid, 8);
940 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
941 NULL, &parent, 8);
942 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
943 NULL, &pflags, 8);
944 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
945 NULL, &atime, 16);
946 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
947 NULL, &mtime, 16);
948 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
949 NULL, &ctime, 16);
950 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
951 NULL, &crtime, 16);
952 }
953
954 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
955
956 if (obj_type == DMU_OT_ZNODE) {
957 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
958 &empty_xattr, 8);
959 }
960 if (obj_type == DMU_OT_ZNODE ||
961 (vap->va_type == VBLK || vap->va_type == VCHR)) {
962 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
963 NULL, &rdev, 8);
964
965 }
966 if (obj_type == DMU_OT_ZNODE) {
967 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
968 NULL, &pflags, 8);
969 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
970 &acl_ids->z_fuid, 8);
971 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
972 &acl_ids->z_fgid, 8);
973 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
974 sizeof (uint64_t) * 4);
975 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
976 &acl_phys, sizeof (zfs_acl_phys_t));
977 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
978 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
979 &acl_ids->z_aclp->z_acl_count, 8);
980 locate.cb_aclp = acl_ids->z_aclp;
981 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
982 zfs_acl_data_locator, &locate,
983 acl_ids->z_aclp->z_acl_bytes);
984 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
985 acl_ids->z_fuid, acl_ids->z_fgid);
986 }
987
988 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
989
990 if (!(flag & IS_ROOT_NODE)) {
991 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
992 ASSERT(*zpp != NULL);
993 } else {
994 /*
995 * If we are creating the root node, the "parent" we
996 * passed in is the znode for the root.
997 */
998 *zpp = dzp;
999
1000 (*zpp)->z_sa_hdl = sa_hdl;
1001 }
1002
1003 (*zpp)->z_pflags = pflags;
1004 (*zpp)->z_mode = mode;
1005
1006 if (vap->va_mask & AT_XVATTR)
1007 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
1008
1009 if (obj_type == DMU_OT_ZNODE ||
1010 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
1011 VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
1012 }
1013 if (!(flag & IS_ROOT_NODE)) {
1014 vnode_t *vp;
1015
1016 vp = ZTOV(*zpp);
1017 vp->v_vflag |= VV_FORCEINSMQ;
1018 err = insmntque(vp, zfsvfs->z_vfs);
1019 vp->v_vflag &= ~VV_FORCEINSMQ;
1020 KASSERT(err == 0, ("insmntque() failed: error %d", err));
1021 }
1022 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1023 }
1024
1025 /*
1026 * Update in-core attributes. It is assumed the caller will be doing an
1027 * sa_bulk_update to push the changes out.
1028 */
1029 void
zfs_xvattr_set(znode_t * zp,xvattr_t * xvap,dmu_tx_t * tx)1030 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
1031 {
1032 xoptattr_t *xoap;
1033
1034 xoap = xva_getxoptattr(xvap);
1035 ASSERT(xoap);
1036
1037 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1038 uint64_t times[2];
1039 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
1040 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
1041 ×, sizeof (times), tx);
1042 XVA_SET_RTN(xvap, XAT_CREATETIME);
1043 }
1044 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
1045 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
1046 zp->z_pflags, tx);
1047 XVA_SET_RTN(xvap, XAT_READONLY);
1048 }
1049 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
1050 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
1051 zp->z_pflags, tx);
1052 XVA_SET_RTN(xvap, XAT_HIDDEN);
1053 }
1054 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
1055 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
1056 zp->z_pflags, tx);
1057 XVA_SET_RTN(xvap, XAT_SYSTEM);
1058 }
1059 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
1060 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
1061 zp->z_pflags, tx);
1062 XVA_SET_RTN(xvap, XAT_ARCHIVE);
1063 }
1064 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
1065 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
1066 zp->z_pflags, tx);
1067 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
1068 }
1069 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
1070 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
1071 zp->z_pflags, tx);
1072 XVA_SET_RTN(xvap, XAT_NOUNLINK);
1073 }
1074 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
1075 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
1076 zp->z_pflags, tx);
1077 XVA_SET_RTN(xvap, XAT_APPENDONLY);
1078 }
1079 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
1080 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
1081 zp->z_pflags, tx);
1082 XVA_SET_RTN(xvap, XAT_NODUMP);
1083 }
1084 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
1085 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
1086 zp->z_pflags, tx);
1087 XVA_SET_RTN(xvap, XAT_OPAQUE);
1088 }
1089 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1090 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
1091 xoap->xoa_av_quarantined, zp->z_pflags, tx);
1092 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
1093 }
1094 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
1095 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
1096 zp->z_pflags, tx);
1097 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
1098 }
1099 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1100 zfs_sa_set_scanstamp(zp, xvap, tx);
1101 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
1102 }
1103 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
1104 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
1105 zp->z_pflags, tx);
1106 XVA_SET_RTN(xvap, XAT_REPARSE);
1107 }
1108 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
1109 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
1110 zp->z_pflags, tx);
1111 XVA_SET_RTN(xvap, XAT_OFFLINE);
1112 }
1113 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
1114 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
1115 zp->z_pflags, tx);
1116 XVA_SET_RTN(xvap, XAT_SPARSE);
1117 }
1118 }
1119
1120 int
zfs_zget(zfsvfs_t * zfsvfs,uint64_t obj_num,znode_t ** zpp)1121 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1122 {
1123 dmu_object_info_t doi;
1124 dmu_buf_t *db;
1125 znode_t *zp;
1126 vnode_t *vp;
1127 sa_handle_t *hdl;
1128 struct thread *td;
1129 int locked;
1130 int err;
1131
1132 td = curthread;
1133 getnewvnode_reserve(1);
1134 again:
1135 *zpp = NULL;
1136 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1137
1138 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1139 if (err) {
1140 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1141 getnewvnode_drop_reserve();
1142 return (err);
1143 }
1144
1145 dmu_object_info_from_db(db, &doi);
1146 if (doi.doi_bonus_type != DMU_OT_SA &&
1147 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1148 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1149 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1150 sa_buf_rele(db, NULL);
1151 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1152 #ifdef __FreeBSD__
1153 getnewvnode_drop_reserve();
1154 #endif
1155 return (SET_ERROR(EINVAL));
1156 }
1157
1158 hdl = dmu_buf_get_user(db);
1159 if (hdl != NULL) {
1160 zp = sa_get_userdata(hdl);
1161
1162
1163 /*
1164 * Since "SA" does immediate eviction we
1165 * should never find a sa handle that doesn't
1166 * know about the znode.
1167 */
1168
1169 ASSERT3P(zp, !=, NULL);
1170
1171 mutex_enter(&zp->z_lock);
1172 ASSERT3U(zp->z_id, ==, obj_num);
1173 if (zp->z_unlinked) {
1174 err = SET_ERROR(ENOENT);
1175 } else {
1176 vp = ZTOV(zp);
1177 *zpp = zp;
1178 err = 0;
1179 }
1180 sa_buf_rele(db, NULL);
1181
1182 /* Don't let the vnode disappear after ZFS_OBJ_HOLD_EXIT. */
1183 if (err == 0)
1184 VN_HOLD(vp);
1185
1186 mutex_exit(&zp->z_lock);
1187 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1188
1189 if (err == 0) {
1190 locked = VOP_ISLOCKED(vp);
1191 VI_LOCK(vp);
1192 if ((vp->v_iflag & VI_DOOMED) != 0 &&
1193 locked != LK_EXCLUSIVE) {
1194 /*
1195 * The vnode is doomed and this thread doesn't
1196 * hold the exclusive lock on it, so the vnode
1197 * must be being reclaimed by another thread.
1198 * Otherwise the doomed vnode is being reclaimed
1199 * by this thread and zfs_zget is called from
1200 * ZIL internals.
1201 */
1202 VI_UNLOCK(vp);
1203 VN_RELE(vp);
1204 goto again;
1205 }
1206 VI_UNLOCK(vp);
1207 }
1208 getnewvnode_drop_reserve();
1209 return (err);
1210 }
1211
1212 /*
1213 * Not found create new znode/vnode
1214 * but only if file exists.
1215 *
1216 * There is a small window where zfs_vget() could
1217 * find this object while a file create is still in
1218 * progress. This is checked for in zfs_znode_alloc()
1219 *
1220 * if zfs_znode_alloc() fails it will drop the hold on the
1221 * bonus buffer.
1222 */
1223 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1224 doi.doi_bonus_type, NULL);
1225 if (zp == NULL) {
1226 err = SET_ERROR(ENOENT);
1227 } else {
1228 *zpp = zp;
1229 }
1230 if (err == 0) {
1231 vnode_t *vp = ZTOV(zp);
1232
1233 err = insmntque(vp, zfsvfs->z_vfs);
1234 if (err == 0) {
1235 vp->v_hash = obj_num;
1236 VOP_UNLOCK(vp, 0);
1237 } else {
1238 zp->z_vnode = NULL;
1239 zfs_znode_dmu_fini(zp);
1240 zfs_znode_free(zp);
1241 *zpp = NULL;
1242 }
1243 }
1244 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1245 getnewvnode_drop_reserve();
1246 return (err);
1247 }
1248
1249 int
zfs_rezget(znode_t * zp)1250 zfs_rezget(znode_t *zp)
1251 {
1252 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1253 dmu_object_info_t doi;
1254 dmu_buf_t *db;
1255 vnode_t *vp;
1256 uint64_t obj_num = zp->z_id;
1257 uint64_t mode, size;
1258 sa_bulk_attr_t bulk[8];
1259 int err;
1260 int count = 0;
1261 uint64_t gen;
1262
1263 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1264
1265 mutex_enter(&zp->z_acl_lock);
1266 if (zp->z_acl_cached) {
1267 zfs_acl_free(zp->z_acl_cached);
1268 zp->z_acl_cached = NULL;
1269 }
1270
1271 mutex_exit(&zp->z_acl_lock);
1272 ASSERT(zp->z_sa_hdl == NULL);
1273 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1274 if (err) {
1275 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1276 return (err);
1277 }
1278
1279 dmu_object_info_from_db(db, &doi);
1280 if (doi.doi_bonus_type != DMU_OT_SA &&
1281 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1282 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1283 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1284 sa_buf_rele(db, NULL);
1285 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1286 return (SET_ERROR(EINVAL));
1287 }
1288
1289 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1290 size = zp->z_size;
1291
1292 /* reload cached values */
1293 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1294 &gen, sizeof (gen));
1295 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1296 &zp->z_size, sizeof (zp->z_size));
1297 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1298 &zp->z_links, sizeof (zp->z_links));
1299 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1300 &zp->z_pflags, sizeof (zp->z_pflags));
1301 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1302 &zp->z_atime, sizeof (zp->z_atime));
1303 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1304 &zp->z_uid, sizeof (zp->z_uid));
1305 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1306 &zp->z_gid, sizeof (zp->z_gid));
1307 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1308 &mode, sizeof (mode));
1309
1310 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1311 zfs_znode_dmu_fini(zp);
1312 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1313 return (SET_ERROR(EIO));
1314 }
1315
1316 zp->z_mode = mode;
1317
1318 if (gen != zp->z_gen) {
1319 zfs_znode_dmu_fini(zp);
1320 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1321 return (SET_ERROR(EIO));
1322 }
1323
1324 /*
1325 * It is highly improbable but still quite possible that two
1326 * objects in different datasets are created with the same
1327 * object numbers and in transaction groups with the same
1328 * numbers. znodes corresponding to those objects would
1329 * have the same z_id and z_gen, but their other attributes
1330 * may be different.
1331 * zfs recv -F may replace one of such objects with the other.
1332 * As a result file properties recorded in the replaced
1333 * object's vnode may no longer match the received object's
1334 * properties. At present the only cached property is the
1335 * files type recorded in v_type.
1336 * So, handle this case by leaving the old vnode and znode
1337 * disassociated from the actual object. A new vnode and a
1338 * znode will be created if the object is accessed
1339 * (e.g. via a look-up). The old vnode and znode will be
1340 * recycled when the last vnode reference is dropped.
1341 */
1342 vp = ZTOV(zp);
1343 if (vp->v_type != IFTOVT((mode_t)zp->z_mode)) {
1344 zfs_znode_dmu_fini(zp);
1345 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1346 return (EIO);
1347 }
1348
1349 zp->z_unlinked = (zp->z_links == 0);
1350 zp->z_blksz = doi.doi_data_block_size;
1351 vn_pages_remove(vp, 0, 0);
1352 if (zp->z_size != size)
1353 vnode_pager_setsize(vp, zp->z_size);
1354
1355 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1356
1357 return (0);
1358 }
1359
1360 void
zfs_znode_delete(znode_t * zp,dmu_tx_t * tx)1361 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1362 {
1363 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1364 objset_t *os = zfsvfs->z_os;
1365 uint64_t obj = zp->z_id;
1366 uint64_t acl_obj = zfs_external_acl(zp);
1367
1368 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1369 if (acl_obj) {
1370 VERIFY(!zp->z_is_sa);
1371 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1372 }
1373 VERIFY(0 == dmu_object_free(os, obj, tx));
1374 zfs_znode_dmu_fini(zp);
1375 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1376 zfs_znode_free(zp);
1377 }
1378
1379 void
zfs_zinactive(znode_t * zp)1380 zfs_zinactive(znode_t *zp)
1381 {
1382 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1383 uint64_t z_id = zp->z_id;
1384
1385 ASSERT(zp->z_sa_hdl);
1386
1387 /*
1388 * Don't allow a zfs_zget() while were trying to release this znode
1389 */
1390 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1391
1392 mutex_enter(&zp->z_lock);
1393
1394 /*
1395 * If this was the last reference to a file with no links,
1396 * remove the file from the file system.
1397 */
1398 if (zp->z_unlinked) {
1399 mutex_exit(&zp->z_lock);
1400 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1401 zfs_rmnode(zp);
1402 return;
1403 }
1404
1405 mutex_exit(&zp->z_lock);
1406 zfs_znode_dmu_fini(zp);
1407 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1408 zfs_znode_free(zp);
1409 }
1410
1411 void
zfs_znode_free(znode_t * zp)1412 zfs_znode_free(znode_t *zp)
1413 {
1414 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1415
1416 ASSERT(zp->z_sa_hdl == NULL);
1417 zp->z_vnode = NULL;
1418 mutex_enter(&zfsvfs->z_znodes_lock);
1419 POINTER_INVALIDATE(&zp->z_zfsvfs);
1420 list_remove(&zfsvfs->z_all_znodes, zp);
1421 mutex_exit(&zfsvfs->z_znodes_lock);
1422
1423 if (zp->z_acl_cached) {
1424 zfs_acl_free(zp->z_acl_cached);
1425 zp->z_acl_cached = NULL;
1426 }
1427
1428 kmem_cache_free(znode_cache, zp);
1429
1430 #ifdef illumos
1431 VFS_RELE(zfsvfs->z_vfs);
1432 #endif
1433 }
1434
1435 void
zfs_tstamp_update_setup(znode_t * zp,uint_t flag,uint64_t mtime[2],uint64_t ctime[2],boolean_t have_tx)1436 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1437 uint64_t ctime[2], boolean_t have_tx)
1438 {
1439 timestruc_t now;
1440
1441 gethrestime(&now);
1442
1443 if (have_tx) { /* will sa_bulk_update happen really soon? */
1444 zp->z_atime_dirty = 0;
1445 zp->z_seq++;
1446 } else {
1447 zp->z_atime_dirty = 1;
1448 }
1449
1450 if (flag & AT_ATIME) {
1451 ZFS_TIME_ENCODE(&now, zp->z_atime);
1452 }
1453
1454 if (flag & AT_MTIME) {
1455 ZFS_TIME_ENCODE(&now, mtime);
1456 if (zp->z_zfsvfs->z_use_fuids) {
1457 zp->z_pflags |= (ZFS_ARCHIVE |
1458 ZFS_AV_MODIFIED);
1459 }
1460 }
1461
1462 if (flag & AT_CTIME) {
1463 ZFS_TIME_ENCODE(&now, ctime);
1464 if (zp->z_zfsvfs->z_use_fuids)
1465 zp->z_pflags |= ZFS_ARCHIVE;
1466 }
1467 }
1468
1469 /*
1470 * Grow the block size for a file.
1471 *
1472 * IN: zp - znode of file to free data in.
1473 * size - requested block size
1474 * tx - open transaction.
1475 *
1476 * NOTE: this function assumes that the znode is write locked.
1477 */
1478 void
zfs_grow_blocksize(znode_t * zp,uint64_t size,dmu_tx_t * tx)1479 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1480 {
1481 int error;
1482 u_longlong_t dummy;
1483
1484 if (size <= zp->z_blksz)
1485 return;
1486 /*
1487 * If the file size is already greater than the current blocksize,
1488 * we will not grow. If there is more than one block in a file,
1489 * the blocksize cannot change.
1490 */
1491 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1492 return;
1493
1494 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1495 size, 0, tx);
1496
1497 if (error == ENOTSUP)
1498 return;
1499 ASSERT0(error);
1500
1501 /* What blocksize did we actually get? */
1502 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1503 }
1504
1505 #ifdef sun
1506 /*
1507 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1508 * be calling back into the fs for a putpage(). E.g.: when truncating
1509 * a file, the pages being "thrown away* don't need to be written out.
1510 */
1511 /* ARGSUSED */
1512 static int
zfs_no_putpage(vnode_t * vp,page_t * pp,u_offset_t * offp,size_t * lenp,int flags,cred_t * cr)1513 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1514 int flags, cred_t *cr)
1515 {
1516 ASSERT(0);
1517 return (0);
1518 }
1519 #endif /* sun */
1520
1521 /*
1522 * Increase the file length
1523 *
1524 * IN: zp - znode of file to free data in.
1525 * end - new end-of-file
1526 *
1527 * RETURN: 0 on success, error code on failure
1528 */
1529 static int
zfs_extend(znode_t * zp,uint64_t end)1530 zfs_extend(znode_t *zp, uint64_t end)
1531 {
1532 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1533 dmu_tx_t *tx;
1534 rl_t *rl;
1535 uint64_t newblksz;
1536 int error;
1537
1538 /*
1539 * We will change zp_size, lock the whole file.
1540 */
1541 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1542
1543 /*
1544 * Nothing to do if file already at desired length.
1545 */
1546 if (end <= zp->z_size) {
1547 zfs_range_unlock(rl);
1548 return (0);
1549 }
1550 tx = dmu_tx_create(zfsvfs->z_os);
1551 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1552 zfs_sa_upgrade_txholds(tx, zp);
1553 if (end > zp->z_blksz &&
1554 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1555 /*
1556 * We are growing the file past the current block size.
1557 */
1558 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1559 ASSERT(!ISP2(zp->z_blksz));
1560 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1561 } else {
1562 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1563 }
1564 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1565 } else {
1566 newblksz = 0;
1567 }
1568
1569 error = dmu_tx_assign(tx, TXG_WAIT);
1570 if (error) {
1571 dmu_tx_abort(tx);
1572 zfs_range_unlock(rl);
1573 return (error);
1574 }
1575
1576 if (newblksz)
1577 zfs_grow_blocksize(zp, newblksz, tx);
1578
1579 zp->z_size = end;
1580
1581 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1582 &zp->z_size, sizeof (zp->z_size), tx));
1583
1584 vnode_pager_setsize(ZTOV(zp), end);
1585
1586 zfs_range_unlock(rl);
1587
1588 dmu_tx_commit(tx);
1589
1590 return (0);
1591 }
1592
1593 /*
1594 * Free space in a file.
1595 *
1596 * IN: zp - znode of file to free data in.
1597 * off - start of section to free.
1598 * len - length of section to free.
1599 *
1600 * RETURN: 0 on success, error code on failure
1601 */
1602 static int
zfs_free_range(znode_t * zp,uint64_t off,uint64_t len)1603 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1604 {
1605 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1606 rl_t *rl;
1607 int error;
1608
1609 /*
1610 * Lock the range being freed.
1611 */
1612 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1613
1614 /*
1615 * Nothing to do if file already at desired length.
1616 */
1617 if (off >= zp->z_size) {
1618 zfs_range_unlock(rl);
1619 return (0);
1620 }
1621
1622 if (off + len > zp->z_size)
1623 len = zp->z_size - off;
1624
1625 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1626
1627 if (error == 0) {
1628 /*
1629 * In FreeBSD we cannot free block in the middle of a file,
1630 * but only at the end of a file, so this code path should
1631 * never happen.
1632 */
1633 vnode_pager_setsize(ZTOV(zp), off);
1634 }
1635
1636 zfs_range_unlock(rl);
1637
1638 return (error);
1639 }
1640
1641 /*
1642 * Truncate a file
1643 *
1644 * IN: zp - znode of file to free data in.
1645 * end - new end-of-file.
1646 *
1647 * RETURN: 0 on success, error code on failure
1648 */
1649 static int
zfs_trunc(znode_t * zp,uint64_t end)1650 zfs_trunc(znode_t *zp, uint64_t end)
1651 {
1652 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1653 vnode_t *vp = ZTOV(zp);
1654 dmu_tx_t *tx;
1655 rl_t *rl;
1656 int error;
1657 sa_bulk_attr_t bulk[2];
1658 int count = 0;
1659
1660 /*
1661 * We will change zp_size, lock the whole file.
1662 */
1663 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1664
1665 /*
1666 * Nothing to do if file already at desired length.
1667 */
1668 if (end >= zp->z_size) {
1669 zfs_range_unlock(rl);
1670 return (0);
1671 }
1672
1673 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1674 if (error) {
1675 zfs_range_unlock(rl);
1676 return (error);
1677 }
1678 tx = dmu_tx_create(zfsvfs->z_os);
1679 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1680 zfs_sa_upgrade_txholds(tx, zp);
1681 error = dmu_tx_assign(tx, TXG_WAIT);
1682 if (error) {
1683 dmu_tx_abort(tx);
1684 zfs_range_unlock(rl);
1685 return (error);
1686 }
1687
1688 zp->z_size = end;
1689 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1690 NULL, &zp->z_size, sizeof (zp->z_size));
1691
1692 if (end == 0) {
1693 zp->z_pflags &= ~ZFS_SPARSE;
1694 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1695 NULL, &zp->z_pflags, 8);
1696 }
1697 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1698
1699 dmu_tx_commit(tx);
1700
1701 /*
1702 * Clear any mapped pages in the truncated region. This has to
1703 * happen outside of the transaction to avoid the possibility of
1704 * a deadlock with someone trying to push a page that we are
1705 * about to invalidate.
1706 */
1707 vnode_pager_setsize(vp, end);
1708
1709 zfs_range_unlock(rl);
1710
1711 return (0);
1712 }
1713
1714 /*
1715 * Free space in a file
1716 *
1717 * IN: zp - znode of file to free data in.
1718 * off - start of range
1719 * len - end of range (0 => EOF)
1720 * flag - current file open mode flags.
1721 * log - TRUE if this action should be logged
1722 *
1723 * RETURN: 0 on success, error code on failure
1724 */
1725 int
zfs_freesp(znode_t * zp,uint64_t off,uint64_t len,int flag,boolean_t log)1726 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1727 {
1728 vnode_t *vp = ZTOV(zp);
1729 dmu_tx_t *tx;
1730 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1731 zilog_t *zilog = zfsvfs->z_log;
1732 uint64_t mode;
1733 uint64_t mtime[2], ctime[2];
1734 sa_bulk_attr_t bulk[3];
1735 int count = 0;
1736 int error;
1737
1738 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1739 sizeof (mode))) != 0)
1740 return (error);
1741
1742 if (off > zp->z_size) {
1743 error = zfs_extend(zp, off+len);
1744 if (error == 0 && log)
1745 goto log;
1746 else
1747 return (error);
1748 }
1749
1750 /*
1751 * Check for any locks in the region to be freed.
1752 */
1753
1754 if (MANDLOCK(vp, (mode_t)mode)) {
1755 uint64_t length = (len ? len : zp->z_size - off);
1756 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1757 return (error);
1758 }
1759
1760 if (len == 0) {
1761 error = zfs_trunc(zp, off);
1762 } else {
1763 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1764 off + len > zp->z_size)
1765 error = zfs_extend(zp, off+len);
1766 }
1767 if (error || !log)
1768 return (error);
1769 log:
1770 tx = dmu_tx_create(zfsvfs->z_os);
1771 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1772 zfs_sa_upgrade_txholds(tx, zp);
1773 error = dmu_tx_assign(tx, TXG_WAIT);
1774 if (error) {
1775 dmu_tx_abort(tx);
1776 return (error);
1777 }
1778
1779 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1780 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1781 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1782 NULL, &zp->z_pflags, 8);
1783 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1784 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1785 ASSERT(error == 0);
1786
1787 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1788
1789 dmu_tx_commit(tx);
1790 return (0);
1791 }
1792
1793 void
zfs_create_fs(objset_t * os,cred_t * cr,nvlist_t * zplprops,dmu_tx_t * tx)1794 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1795 {
1796 uint64_t moid, obj, sa_obj, version;
1797 uint64_t sense = ZFS_CASE_SENSITIVE;
1798 uint64_t norm = 0;
1799 nvpair_t *elem;
1800 int error;
1801 int i;
1802 znode_t *rootzp = NULL;
1803 zfsvfs_t *zfsvfs;
1804 vattr_t vattr;
1805 znode_t *zp;
1806 zfs_acl_ids_t acl_ids;
1807
1808 /*
1809 * First attempt to create master node.
1810 */
1811 /*
1812 * In an empty objset, there are no blocks to read and thus
1813 * there can be no i/o errors (which we assert below).
1814 */
1815 moid = MASTER_NODE_OBJ;
1816 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1817 DMU_OT_NONE, 0, tx);
1818 ASSERT(error == 0);
1819
1820 /*
1821 * Set starting attributes.
1822 */
1823 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1824 elem = NULL;
1825 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1826 /* For the moment we expect all zpl props to be uint64_ts */
1827 uint64_t val;
1828 char *name;
1829
1830 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1831 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1832 name = nvpair_name(elem);
1833 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1834 if (val < version)
1835 version = val;
1836 } else {
1837 error = zap_update(os, moid, name, 8, 1, &val, tx);
1838 }
1839 ASSERT(error == 0);
1840 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1841 norm = val;
1842 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1843 sense = val;
1844 }
1845 ASSERT(version != 0);
1846 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1847
1848 /*
1849 * Create zap object used for SA attribute registration
1850 */
1851
1852 if (version >= ZPL_VERSION_SA) {
1853 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1854 DMU_OT_NONE, 0, tx);
1855 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1856 ASSERT(error == 0);
1857 } else {
1858 sa_obj = 0;
1859 }
1860 /*
1861 * Create a delete queue.
1862 */
1863 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1864
1865 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1866 ASSERT(error == 0);
1867
1868 /*
1869 * Create root znode. Create minimal znode/vnode/zfsvfs
1870 * to allow zfs_mknode to work.
1871 */
1872 VATTR_NULL(&vattr);
1873 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1874 vattr.va_type = VDIR;
1875 vattr.va_mode = S_IFDIR|0755;
1876 vattr.va_uid = crgetuid(cr);
1877 vattr.va_gid = crgetgid(cr);
1878
1879 zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
1880
1881 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1882 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1883 rootzp->z_moved = 0;
1884 rootzp->z_unlinked = 0;
1885 rootzp->z_atime_dirty = 0;
1886 rootzp->z_is_sa = USE_SA(version, os);
1887
1888 zfsvfs->z_os = os;
1889 zfsvfs->z_parent = zfsvfs;
1890 zfsvfs->z_version = version;
1891 zfsvfs->z_use_fuids = USE_FUIDS(version, os);
1892 zfsvfs->z_use_sa = USE_SA(version, os);
1893 zfsvfs->z_norm = norm;
1894
1895 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1896 &zfsvfs->z_attr_table);
1897
1898 ASSERT(error == 0);
1899
1900 /*
1901 * Fold case on file systems that are always or sometimes case
1902 * insensitive.
1903 */
1904 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1905 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
1906
1907 mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1908 list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
1909 offsetof(znode_t, z_link_node));
1910
1911 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1912 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1913
1914 rootzp->z_zfsvfs = zfsvfs;
1915 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1916 cr, NULL, &acl_ids));
1917 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1918 ASSERT3P(zp, ==, rootzp);
1919 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1920 ASSERT(error == 0);
1921 zfs_acl_ids_free(&acl_ids);
1922 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1923
1924 sa_handle_destroy(rootzp->z_sa_hdl);
1925 kmem_cache_free(znode_cache, rootzp);
1926
1927 /*
1928 * Create shares directory
1929 */
1930
1931 error = zfs_create_share_dir(zfsvfs, tx);
1932
1933 ASSERT(error == 0);
1934
1935 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1936 mutex_destroy(&zfsvfs->z_hold_mtx[i]);
1937 kmem_free(zfsvfs, sizeof (zfsvfs_t));
1938 }
1939
1940 #endif /* _KERNEL */
1941
1942 static int
zfs_sa_setup(objset_t * osp,sa_attr_type_t ** sa_table)1943 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1944 {
1945 uint64_t sa_obj = 0;
1946 int error;
1947
1948 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1949 if (error != 0 && error != ENOENT)
1950 return (error);
1951
1952 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1953 return (error);
1954 }
1955
1956 static int
zfs_grab_sa_handle(objset_t * osp,uint64_t obj,sa_handle_t ** hdlp,dmu_buf_t ** db,void * tag)1957 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1958 dmu_buf_t **db, void *tag)
1959 {
1960 dmu_object_info_t doi;
1961 int error;
1962
1963 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
1964 return (error);
1965
1966 dmu_object_info_from_db(*db, &doi);
1967 if ((doi.doi_bonus_type != DMU_OT_SA &&
1968 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1969 doi.doi_bonus_type == DMU_OT_ZNODE &&
1970 doi.doi_bonus_size < sizeof (znode_phys_t)) {
1971 sa_buf_rele(*db, tag);
1972 return (SET_ERROR(ENOTSUP));
1973 }
1974
1975 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1976 if (error != 0) {
1977 sa_buf_rele(*db, tag);
1978 return (error);
1979 }
1980
1981 return (0);
1982 }
1983
1984 void
zfs_release_sa_handle(sa_handle_t * hdl,dmu_buf_t * db,void * tag)1985 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
1986 {
1987 sa_handle_destroy(hdl);
1988 sa_buf_rele(db, tag);
1989 }
1990
1991 /*
1992 * Given an object number, return its parent object number and whether
1993 * or not the object is an extended attribute directory.
1994 */
1995 static int
zfs_obj_to_pobj(objset_t * osp,sa_handle_t * hdl,sa_attr_type_t * sa_table,uint64_t * pobjp,int * is_xattrdir)1996 zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table,
1997 uint64_t *pobjp, int *is_xattrdir)
1998 {
1999 uint64_t parent;
2000 uint64_t pflags;
2001 uint64_t mode;
2002 uint64_t parent_mode;
2003 sa_bulk_attr_t bulk[3];
2004 sa_handle_t *sa_hdl;
2005 dmu_buf_t *sa_db;
2006 int count = 0;
2007 int error;
2008
2009 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
2010 &parent, sizeof (parent));
2011 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
2012 &pflags, sizeof (pflags));
2013 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2014 &mode, sizeof (mode));
2015
2016 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
2017 return (error);
2018
2019 /*
2020 * When a link is removed its parent pointer is not changed and will
2021 * be invalid. There are two cases where a link is removed but the
2022 * file stays around, when it goes to the delete queue and when there
2023 * are additional links.
2024 */
2025 error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG);
2026 if (error != 0)
2027 return (error);
2028
2029 error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode));
2030 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2031 if (error != 0)
2032 return (error);
2033
2034 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
2035
2036 /*
2037 * Extended attributes can be applied to files, directories, etc.
2038 * Otherwise the parent must be a directory.
2039 */
2040 if (!*is_xattrdir && !S_ISDIR(parent_mode))
2041 return (SET_ERROR(EINVAL));
2042
2043 *pobjp = parent;
2044
2045 return (0);
2046 }
2047
2048 /*
2049 * Given an object number, return some zpl level statistics
2050 */
2051 static int
zfs_obj_to_stats_impl(sa_handle_t * hdl,sa_attr_type_t * sa_table,zfs_stat_t * sb)2052 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
2053 zfs_stat_t *sb)
2054 {
2055 sa_bulk_attr_t bulk[4];
2056 int count = 0;
2057
2058 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2059 &sb->zs_mode, sizeof (sb->zs_mode));
2060 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
2061 &sb->zs_gen, sizeof (sb->zs_gen));
2062 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
2063 &sb->zs_links, sizeof (sb->zs_links));
2064 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
2065 &sb->zs_ctime, sizeof (sb->zs_ctime));
2066
2067 return (sa_bulk_lookup(hdl, bulk, count));
2068 }
2069
2070 static int
zfs_obj_to_path_impl(objset_t * osp,uint64_t obj,sa_handle_t * hdl,sa_attr_type_t * sa_table,char * buf,int len)2071 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
2072 sa_attr_type_t *sa_table, char *buf, int len)
2073 {
2074 sa_handle_t *sa_hdl;
2075 sa_handle_t *prevhdl = NULL;
2076 dmu_buf_t *prevdb = NULL;
2077 dmu_buf_t *sa_db = NULL;
2078 char *path = buf + len - 1;
2079 int error;
2080
2081 *path = '\0';
2082 sa_hdl = hdl;
2083
2084 for (;;) {
2085 uint64_t pobj;
2086 char component[MAXNAMELEN + 2];
2087 size_t complen;
2088 int is_xattrdir;
2089
2090 if (prevdb)
2091 zfs_release_sa_handle(prevhdl, prevdb, FTAG);
2092
2093 if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj,
2094 &is_xattrdir)) != 0)
2095 break;
2096
2097 if (pobj == obj) {
2098 if (path[0] != '/')
2099 *--path = '/';
2100 break;
2101 }
2102
2103 component[0] = '/';
2104 if (is_xattrdir) {
2105 (void) sprintf(component + 1, "<xattrdir>");
2106 } else {
2107 error = zap_value_search(osp, pobj, obj,
2108 ZFS_DIRENT_OBJ(-1ULL), component + 1);
2109 if (error != 0)
2110 break;
2111 }
2112
2113 complen = strlen(component);
2114 path -= complen;
2115 ASSERT(path >= buf);
2116 bcopy(component, path, complen);
2117 obj = pobj;
2118
2119 if (sa_hdl != hdl) {
2120 prevhdl = sa_hdl;
2121 prevdb = sa_db;
2122 }
2123 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
2124 if (error != 0) {
2125 sa_hdl = prevhdl;
2126 sa_db = prevdb;
2127 break;
2128 }
2129 }
2130
2131 if (sa_hdl != NULL && sa_hdl != hdl) {
2132 ASSERT(sa_db != NULL);
2133 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2134 }
2135
2136 if (error == 0)
2137 (void) memmove(buf, path, buf + len - path);
2138
2139 return (error);
2140 }
2141
2142 int
zfs_obj_to_path(objset_t * osp,uint64_t obj,char * buf,int len)2143 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
2144 {
2145 sa_attr_type_t *sa_table;
2146 sa_handle_t *hdl;
2147 dmu_buf_t *db;
2148 int error;
2149
2150 error = zfs_sa_setup(osp, &sa_table);
2151 if (error != 0)
2152 return (error);
2153
2154 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2155 if (error != 0)
2156 return (error);
2157
2158 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2159
2160 zfs_release_sa_handle(hdl, db, FTAG);
2161 return (error);
2162 }
2163
2164 int
zfs_obj_to_stats(objset_t * osp,uint64_t obj,zfs_stat_t * sb,char * buf,int len)2165 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
2166 char *buf, int len)
2167 {
2168 char *path = buf + len - 1;
2169 sa_attr_type_t *sa_table;
2170 sa_handle_t *hdl;
2171 dmu_buf_t *db;
2172 int error;
2173
2174 *path = '\0';
2175
2176 error = zfs_sa_setup(osp, &sa_table);
2177 if (error != 0)
2178 return (error);
2179
2180 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2181 if (error != 0)
2182 return (error);
2183
2184 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
2185 if (error != 0) {
2186 zfs_release_sa_handle(hdl, db, FTAG);
2187 return (error);
2188 }
2189
2190 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2191
2192 zfs_release_sa_handle(hdl, db, FTAG);
2193 return (error);
2194 }
2195