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) 2012, 2014 by Delphix. All rights reserved.
24 * Copyright (c) 2014 Integros [integros.com]
25 */
26
27 /* Portions Copyright 2007 Jeremy Teo */
28 /* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */
29
30 #ifdef _KERNEL
31 #include <sys/types.h>
32 #include <sys/param.h>
33 #include <sys/time.h>
34 #include <sys/systm.h>
35 #include <sys/sysmacros.h>
36 #include <sys/resource.h>
37 #include <sys/mntent.h>
38 #include <sys/u8_textprep.h>
39 #include <sys/dsl_dataset.h>
40 #include <sys/vfs.h>
41 #include <sys/vnode.h>
42 #include <sys/file.h>
43 #include <sys/kmem.h>
44 #include <sys/errno.h>
45 #include <sys/unistd.h>
46 #include <sys/atomic.h>
47 #include <sys/zfs_dir.h>
48 #include <sys/zfs_acl.h>
49 #include <sys/zfs_ioctl.h>
50 #include <sys/zfs_rlock.h>
51 #include <sys/zfs_fuid.h>
52 #include <sys/dnode.h>
53 #include <sys/fs/zfs.h>
54 #include <sys/kidmap.h>
55
56 #ifdef __NetBSD__
57 #include <sys/zfs_ctldir.h>
58 #include <miscfs/specfs/specdev.h>
59
60 extern int (**zfs_vnodeop_p)(void *);
61 extern int (**zfs_fifoop_p)(void *);
62 extern int (**zfs_specop_p)(void *);
63
64 #endif
65 #endif /* _KERNEL */
66
67 #include <sys/dmu.h>
68 #include <sys/dmu_objset.h>
69 #include <sys/refcount.h>
70 #include <sys/stat.h>
71 #include <sys/zap.h>
72 #include <sys/zfs_znode.h>
73 #include <sys/sa.h>
74 #include <sys/zfs_sa.h>
75 #include <sys/zfs_stat.h>
76 #include <sys/refcount.h>
77
78 #include "zfs_prop.h"
79 #include "zfs_comutil.h"
80
81 /* Used by fstat(1). */
82 SYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD,
83 SYSCTL_NULL_INT_PTR, sizeof(znode_t), "sizeof(znode_t)");
84
85 /*
86 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
87 * turned on when DEBUG is also defined.
88 */
89 #ifdef DEBUG
90 #define ZNODE_STATS
91 #endif /* DEBUG */
92
93 #ifdef ZNODE_STATS
94 #define ZNODE_STAT_ADD(stat) ((stat)++)
95 #else
96 #define ZNODE_STAT_ADD(stat) /* nothing */
97 #endif /* ZNODE_STATS */
98
99 /*
100 * Functions needed for userland (ie: libzpool) are not put under
101 * #ifdef_KERNEL; the rest of the functions have dependencies
102 * (such as VFS logic) that will not compile easily in userland.
103 */
104 #ifdef _KERNEL
105 /*
106 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
107 * be freed before it can be safely accessed.
108 */
109 krwlock_t zfsvfs_lock;
110
111 static kmem_cache_t *znode_cache = NULL;
112
113 /*ARGSUSED*/
114 static void
znode_evict_error(dmu_buf_t * dbuf,void * user_ptr)115 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
116 {
117 /*
118 * We should never drop all dbuf refs without first clearing
119 * the eviction callback.
120 */
121 panic("evicting znode %p\n", user_ptr);
122 }
123
124 extern struct vop_vector zfs_vnodeops;
125 extern struct vop_vector zfs_fifoops;
126 extern struct vop_vector zfs_shareops;
127
128 static int
zfs_znode_cache_constructor(void * buf,void * arg,int kmflags)129 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
130 {
131 znode_t *zp = buf;
132
133 POINTER_INVALIDATE(&zp->z_zfsvfs);
134
135 list_link_init(&zp->z_link_node);
136
137 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
138
139 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
140 avl_create(&zp->z_range_avl, zfs_range_compare,
141 sizeof (rl_t), offsetof(rl_t, r_node));
142
143 zp->z_acl_cached = NULL;
144 zp->z_vnode = NULL;
145 zp->z_moved = 0;
146 return (0);
147 }
148
149 /*ARGSUSED*/
150 static void
zfs_znode_cache_destructor(void * buf,void * arg)151 zfs_znode_cache_destructor(void *buf, void *arg)
152 {
153 znode_t *zp = buf;
154
155 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
156 ASSERT(ZTOV(zp) == NULL);
157 #ifndef __NetBSD__
158 vn_free(ZTOV(zp));
159 #endif
160 ASSERT(!list_link_active(&zp->z_link_node));
161 mutex_destroy(&zp->z_acl_lock);
162 avl_destroy(&zp->z_range_avl);
163 mutex_destroy(&zp->z_range_lock);
164
165 ASSERT(zp->z_acl_cached == NULL);
166 }
167
168 #ifdef ZNODE_STATS
169 static struct {
170 uint64_t zms_zfsvfs_invalid;
171 uint64_t zms_zfsvfs_recheck1;
172 uint64_t zms_zfsvfs_unmounted;
173 uint64_t zms_zfsvfs_recheck2;
174 uint64_t zms_obj_held;
175 uint64_t zms_vnode_locked;
176 uint64_t zms_not_only_dnlc;
177 } znode_move_stats;
178 #endif /* ZNODE_STATS */
179
180 #ifdef illumos
181 static void
zfs_znode_move_impl(znode_t * ozp,znode_t * nzp)182 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
183 {
184 vnode_t *vp;
185
186 /* Copy fields. */
187 nzp->z_zfsvfs = ozp->z_zfsvfs;
188
189 /* Swap vnodes. */
190 vp = nzp->z_vnode;
191 nzp->z_vnode = ozp->z_vnode;
192 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
193 ZTOV(ozp)->v_data = ozp;
194 ZTOV(nzp)->v_data = nzp;
195
196 nzp->z_id = ozp->z_id;
197 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
198 ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
199 nzp->z_unlinked = ozp->z_unlinked;
200 nzp->z_atime_dirty = ozp->z_atime_dirty;
201 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
202 nzp->z_blksz = ozp->z_blksz;
203 nzp->z_seq = ozp->z_seq;
204 nzp->z_mapcnt = ozp->z_mapcnt;
205 nzp->z_gen = ozp->z_gen;
206 nzp->z_sync_cnt = ozp->z_sync_cnt;
207 nzp->z_is_sa = ozp->z_is_sa;
208 nzp->z_sa_hdl = ozp->z_sa_hdl;
209 bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2);
210 nzp->z_links = ozp->z_links;
211 nzp->z_size = ozp->z_size;
212 nzp->z_pflags = ozp->z_pflags;
213 nzp->z_uid = ozp->z_uid;
214 nzp->z_gid = ozp->z_gid;
215 nzp->z_mode = ozp->z_mode;
216
217 /*
218 * Since this is just an idle znode and kmem is already dealing with
219 * memory pressure, release any cached ACL.
220 */
221 if (ozp->z_acl_cached) {
222 zfs_acl_free(ozp->z_acl_cached);
223 ozp->z_acl_cached = NULL;
224 }
225
226 sa_set_userp(nzp->z_sa_hdl, nzp);
227
228 /*
229 * Invalidate the original znode by clearing fields that provide a
230 * pointer back to the znode. Set the low bit of the vfs pointer to
231 * ensure that zfs_znode_move() recognizes the znode as invalid in any
232 * subsequent callback.
233 */
234 ozp->z_sa_hdl = NULL;
235 POINTER_INVALIDATE(&ozp->z_zfsvfs);
236
237 /*
238 * Mark the znode.
239 */
240 nzp->z_moved = 1;
241 ozp->z_moved = (uint8_t)-1;
242 }
243
244 /*ARGSUSED*/
245 static kmem_cbrc_t
zfs_znode_move(void * buf,void * newbuf,size_t size,void * arg)246 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
247 {
248 znode_t *ozp = buf, *nzp = newbuf;
249 zfsvfs_t *zfsvfs;
250 vnode_t *vp;
251
252 /*
253 * The znode is on the file system's list of known znodes if the vfs
254 * pointer is valid. We set the low bit of the vfs pointer when freeing
255 * the znode to invalidate it, and the memory patterns written by kmem
256 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
257 * created znode sets the vfs pointer last of all to indicate that the
258 * znode is known and in a valid state to be moved by this function.
259 */
260 zfsvfs = ozp->z_zfsvfs;
261 if (!POINTER_IS_VALID(zfsvfs)) {
262 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
263 return (KMEM_CBRC_DONT_KNOW);
264 }
265
266 /*
267 * Close a small window in which it's possible that the filesystem could
268 * be unmounted and freed, and zfsvfs, though valid in the previous
269 * statement, could point to unrelated memory by the time we try to
270 * prevent the filesystem from being unmounted.
271 */
272 rw_enter(&zfsvfs_lock, RW_WRITER);
273 if (zfsvfs != ozp->z_zfsvfs) {
274 rw_exit(&zfsvfs_lock);
275 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
276 return (KMEM_CBRC_DONT_KNOW);
277 }
278
279 /*
280 * If the znode is still valid, then so is the file system. We know that
281 * no valid file system can be freed while we hold zfsvfs_lock, so we
282 * can safely ensure that the filesystem is not and will not be
283 * unmounted. The next statement is equivalent to ZFS_ENTER().
284 */
285 rrm_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
286 if (zfsvfs->z_unmounted) {
287 ZFS_EXIT(zfsvfs);
288 rw_exit(&zfsvfs_lock);
289 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
290 return (KMEM_CBRC_DONT_KNOW);
291 }
292 rw_exit(&zfsvfs_lock);
293
294 mutex_enter(&zfsvfs->z_znodes_lock);
295 /*
296 * Recheck the vfs pointer in case the znode was removed just before
297 * acquiring the lock.
298 */
299 if (zfsvfs != ozp->z_zfsvfs) {
300 mutex_exit(&zfsvfs->z_znodes_lock);
301 ZFS_EXIT(zfsvfs);
302 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
303 return (KMEM_CBRC_DONT_KNOW);
304 }
305
306 /*
307 * At this point we know that as long as we hold z_znodes_lock, the
308 * znode cannot be freed and fields within the znode can be safely
309 * accessed. Now, prevent a race with zfs_zget().
310 */
311 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
312 mutex_exit(&zfsvfs->z_znodes_lock);
313 ZFS_EXIT(zfsvfs);
314 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
315 return (KMEM_CBRC_LATER);
316 }
317
318 vp = ZTOV(ozp);
319 if (mutex_tryenter(&vp->v_lock) == 0) {
320 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
321 mutex_exit(&zfsvfs->z_znodes_lock);
322 ZFS_EXIT(zfsvfs);
323 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
324 return (KMEM_CBRC_LATER);
325 }
326
327 /* Only move znodes that are referenced _only_ by the DNLC. */
328 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
329 mutex_exit(&vp->v_lock);
330 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
331 mutex_exit(&zfsvfs->z_znodes_lock);
332 ZFS_EXIT(zfsvfs);
333 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
334 return (KMEM_CBRC_LATER);
335 }
336
337 /*
338 * The znode is known and in a valid state to move. We're holding the
339 * locks needed to execute the critical section.
340 */
341 zfs_znode_move_impl(ozp, nzp);
342 mutex_exit(&vp->v_lock);
343 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
344
345 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
346 mutex_exit(&zfsvfs->z_znodes_lock);
347 ZFS_EXIT(zfsvfs);
348
349 return (KMEM_CBRC_YES);
350 }
351 #endif /* illumos */
352
353 void
zfs_znode_init(void)354 zfs_znode_init(void)
355 {
356 /*
357 * Initialize zcache
358 */
359 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
360 ASSERT(znode_cache == NULL);
361 znode_cache = kmem_cache_create("zfs_znode_cache",
362 sizeof (znode_t), 0, zfs_znode_cache_constructor,
363 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
364 kmem_cache_set_move(znode_cache, zfs_znode_move);
365 }
366
367 void
zfs_znode_fini(void)368 zfs_znode_fini(void)
369 {
370 #ifdef illumos
371 /*
372 * Cleanup vfs & vnode ops
373 */
374 zfs_remove_op_tables();
375 #endif
376
377 /*
378 * Cleanup zcache
379 */
380 if (znode_cache)
381 kmem_cache_destroy(znode_cache);
382 znode_cache = NULL;
383 rw_destroy(&zfsvfs_lock);
384 }
385
386 #ifdef illumos
387 struct vnodeops *zfs_dvnodeops;
388 struct vnodeops *zfs_fvnodeops;
389 struct vnodeops *zfs_symvnodeops;
390 struct vnodeops *zfs_xdvnodeops;
391 struct vnodeops *zfs_evnodeops;
392 struct vnodeops *zfs_sharevnodeops;
393
394 void
zfs_remove_op_tables()395 zfs_remove_op_tables()
396 {
397 /*
398 * Remove vfs ops
399 */
400 ASSERT(zfsfstype);
401 (void) vfs_freevfsops_by_type(zfsfstype);
402 zfsfstype = 0;
403
404 /*
405 * Remove vnode ops
406 */
407 if (zfs_dvnodeops)
408 vn_freevnodeops(zfs_dvnodeops);
409 if (zfs_fvnodeops)
410 vn_freevnodeops(zfs_fvnodeops);
411 if (zfs_symvnodeops)
412 vn_freevnodeops(zfs_symvnodeops);
413 if (zfs_xdvnodeops)
414 vn_freevnodeops(zfs_xdvnodeops);
415 if (zfs_evnodeops)
416 vn_freevnodeops(zfs_evnodeops);
417 if (zfs_sharevnodeops)
418 vn_freevnodeops(zfs_sharevnodeops);
419
420 zfs_dvnodeops = NULL;
421 zfs_fvnodeops = NULL;
422 zfs_symvnodeops = NULL;
423 zfs_xdvnodeops = NULL;
424 zfs_evnodeops = NULL;
425 zfs_sharevnodeops = NULL;
426 }
427
428 extern const fs_operation_def_t zfs_dvnodeops_template[];
429 extern const fs_operation_def_t zfs_fvnodeops_template[];
430 extern const fs_operation_def_t zfs_xdvnodeops_template[];
431 extern const fs_operation_def_t zfs_symvnodeops_template[];
432 extern const fs_operation_def_t zfs_evnodeops_template[];
433 extern const fs_operation_def_t zfs_sharevnodeops_template[];
434
435 int
zfs_create_op_tables()436 zfs_create_op_tables()
437 {
438 int error;
439
440 /*
441 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
442 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
443 * In this case we just return as the ops vectors are already set up.
444 */
445 if (zfs_dvnodeops)
446 return (0);
447
448 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
449 &zfs_dvnodeops);
450 if (error)
451 return (error);
452
453 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
454 &zfs_fvnodeops);
455 if (error)
456 return (error);
457
458 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
459 &zfs_symvnodeops);
460 if (error)
461 return (error);
462
463 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
464 &zfs_xdvnodeops);
465 if (error)
466 return (error);
467
468 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
469 &zfs_evnodeops);
470 if (error)
471 return (error);
472
473 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
474 &zfs_sharevnodeops);
475
476 return (error);
477 }
478 #endif /* illumos */
479
480 int
zfs_create_share_dir(zfsvfs_t * zfsvfs,dmu_tx_t * tx)481 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
482 {
483 zfs_acl_ids_t acl_ids;
484 vattr_t vattr;
485 znode_t *sharezp;
486 znode_t *zp;
487 int error;
488
489 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
490 vattr.va_type = VDIR;
491 vattr.va_mode = S_IFDIR|0555;
492 vattr.va_uid = crgetuid(kcred);
493 vattr.va_gid = crgetgid(kcred);
494
495 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
496 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
497 sharezp->z_moved = 0;
498 sharezp->z_unlinked = 0;
499 sharezp->z_atime_dirty = 0;
500 sharezp->z_zfsvfs = zfsvfs;
501 sharezp->z_is_sa = zfsvfs->z_use_sa;
502
503 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
504 kcred, NULL, &acl_ids));
505 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
506 ASSERT3P(zp, ==, sharezp);
507 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
508 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
509 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
510 zfsvfs->z_shares_dir = sharezp->z_id;
511
512 zfs_acl_ids_free(&acl_ids);
513 sa_handle_destroy(sharezp->z_sa_hdl);
514 kmem_cache_free(znode_cache, sharezp);
515
516 return (error);
517 }
518
519 /*
520 * define a couple of values we need available
521 * for both 64 and 32 bit environments.
522 */
523 #ifndef NBITSMINOR64
524 #define NBITSMINOR64 32
525 #endif
526 #ifndef MAXMAJ64
527 #define MAXMAJ64 0xffffffffUL
528 #endif
529 #ifndef MAXMIN64
530 #define MAXMIN64 0xffffffffUL
531 #endif
532
533 /*
534 * Create special expldev for ZFS private use.
535 * Can't use standard expldev since it doesn't do
536 * what we want. The standard expldev() takes a
537 * dev32_t in LP64 and expands it to a long dev_t.
538 * We need an interface that takes a dev32_t in ILP32
539 * and expands it to a long dev_t.
540 */
541 static uint64_t
zfs_expldev(dev_t dev)542 zfs_expldev(dev_t dev)
543 {
544 return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev));
545 }
546 /*
547 * Special cmpldev for ZFS private use.
548 * Can't use standard cmpldev since it takes
549 * a long dev_t and compresses it to dev32_t in
550 * LP64. We need to do a compaction of a long dev_t
551 * to a dev32_t in ILP32.
552 */
553 dev_t
zfs_cmpldev(uint64_t dev)554 zfs_cmpldev(uint64_t dev)
555 {
556 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64)));
557 }
558
559 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)560 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
561 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
562 {
563 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
564 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
565
566 ASSERT(zp->z_sa_hdl == NULL);
567 ASSERT(zp->z_acl_cached == NULL);
568 if (sa_hdl == NULL) {
569 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
570 SA_HDL_SHARED, &zp->z_sa_hdl));
571 } else {
572 zp->z_sa_hdl = sa_hdl;
573 sa_set_userp(sa_hdl, zp);
574 }
575
576 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
577
578 /*
579 * Slap on VROOT if we are the root znode unless we are the root
580 * node of a snapshot mounted under .zfs.
581 */
582 if (zp->z_id == zfsvfs->z_root && zfsvfs->z_parent == zfsvfs)
583 ZTOV(zp)->v_flag |= VROOT;
584
585 vn_exists(ZTOV(zp));
586 }
587
588 void
zfs_znode_dmu_fini(znode_t * zp)589 zfs_znode_dmu_fini(znode_t *zp)
590 {
591 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
592 zp->z_unlinked ||
593 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
594
595 sa_handle_destroy(zp->z_sa_hdl);
596 zp->z_sa_hdl = NULL;
597 }
598
599 #ifdef __FreeBSD__
600 static void
zfs_vnode_forget(vnode_t * vp)601 zfs_vnode_forget(vnode_t *vp)
602 {
603
604 /* copied from insmntque_stddtr */
605 vp->v_data = NULL;
606 vp->v_op = &dead_vnodeops;
607 vgone(vp);
608 vput(vp);
609 }
610 #endif /* __FreeBSD__ */
611
612 /*
613 * Construct a new znode/vnode and intialize.
614 *
615 * This does not do a call to dmu_set_user() that is
616 * up to the caller to do, in case you don't want to
617 * return the znode
618 */
619 static znode_t *
620 #ifdef __NetBSD__
zfs_znode_alloc(zfsvfs_t * zfsvfs,dmu_buf_t * db,int blksz,dmu_object_type_t obj_type,sa_handle_t * hdl,vnode_t * vp)621 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
622 dmu_object_type_t obj_type, sa_handle_t *hdl, vnode_t *vp)
623 #else
624 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
625 dmu_object_type_t obj_type, sa_handle_t *hdl)
626 #endif
627 {
628 znode_t *zp;
629 #ifndef __NetBSD__
630 vnode_t *vp;
631 #endif
632 uint64_t mode;
633 uint64_t parent;
634 sa_bulk_attr_t bulk[9];
635 int count = 0;
636 int error;
637
638 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
639
640 #ifndef __NetBSD__
641 KASSERT(curthread->td_vp_reserv > 0,
642 ("zfs_znode_alloc: getnewvnode without any vnodes reserved"));
643 error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp);
644 if (error != 0) {
645 kmem_cache_free(znode_cache, zp);
646 return (NULL);
647 }
648 #endif
649 zp->z_vnode = vp;
650 vp->v_data = zp;
651
652 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
653 zp->z_moved = 0;
654
655 /*
656 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
657 * the zfs_znode_move() callback.
658 */
659 zp->z_sa_hdl = NULL;
660 zp->z_unlinked = 0;
661 zp->z_atime_dirty = 0;
662 zp->z_mapcnt = 0;
663 zp->z_id = db->db_object;
664 zp->z_blksz = blksz;
665 zp->z_seq = 0x7A4653;
666 zp->z_sync_cnt = 0;
667
668 #ifdef __NetBSD__
669 vp->v_op = zfs_vnodeop_p;
670 vp->v_tag = VT_ZFS;
671 zp->z_lockf = NULL;
672 #endif
673
674 vp = ZTOV(zp);
675
676 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
677
678 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
679 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
680 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
681 &zp->z_size, 8);
682 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
683 &zp->z_links, 8);
684 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
685 &zp->z_pflags, 8);
686 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
687 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
688 &zp->z_atime, 16);
689 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
690 &zp->z_uid, 8);
691 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
692 &zp->z_gid, 8);
693
694 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
695 if (hdl == NULL)
696 sa_handle_destroy(zp->z_sa_hdl);
697 #ifndef __NetBSD__
698 zfs_vnode_forget(vp);
699 #endif
700 zp->z_vnode = NULL;
701 kmem_cache_free(znode_cache, zp);
702 return (NULL);
703 }
704
705 zp->z_mode = mode;
706
707 vp->v_type = IFTOVT((mode_t)mode);
708
709 switch (vp->v_type) {
710 case VDIR:
711 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
712 break;
713 #if defined(illumos) || defined(__NetBSD__)
714 case VBLK:
715 case VCHR:
716 {
717 uint64_t rdev;
718 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
719 &rdev, sizeof (rdev)) == 0);
720
721 #ifdef illumos
722 vp->v_rdev = zfs_cmpldev(rdev);
723 #else
724 vp->v_op = zfs_specop_p;
725 spec_node_init(vp, zfs_cmpldev(rdev));
726 #endif
727 }
728 break;
729 #endif
730 case VFIFO:
731 #ifdef __NetBSD__
732 vp->v_op = zfs_fifoop_p;
733 break;
734 #else /* __NetBSD__ */
735 #ifdef illumos
736 case VSOCK:
737 case VDOOR:
738 #endif
739 vp->v_op = &zfs_fifoops;
740 break;
741 case VREG:
742 if (parent == zfsvfs->z_shares_dir) {
743 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
744 vp->v_op = &zfs_shareops;
745 }
746 break;
747 #ifdef illumos
748 case VLNK:
749 vn_setops(vp, zfs_symvnodeops);
750 break;
751 default:
752 vn_setops(vp, zfs_evnodeops);
753 break;
754 #endif
755 #endif /* __NetBSD__ */
756 }
757
758 #ifdef __NetBSD__
759 extern const struct genfs_ops zfs_genfsops;
760 genfs_node_init(vp, &zfs_genfsops);
761 uvm_vnp_setsize(vp, zp->z_size);
762 #endif
763
764 mutex_enter(&zfsvfs->z_znodes_lock);
765 list_insert_tail(&zfsvfs->z_all_znodes, zp);
766 membar_producer();
767 /*
768 * Everything else must be valid before assigning z_zfsvfs makes the
769 * znode eligible for zfs_znode_move().
770 */
771 zp->z_zfsvfs = zfsvfs;
772 mutex_exit(&zfsvfs->z_znodes_lock);
773
774 #ifndef __NetBSD__
775 /*
776 * Acquire vnode lock before making it available to the world.
777 */
778 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
779 VN_LOCK_AREC(vp);
780 if (vp->v_type != VFIFO)
781 VN_LOCK_ASHARE(vp);
782 #endif
783
784 #if defined(illumos) || defined(__NetBSD__)
785 VFS_HOLD(zfsvfs->z_vfs);
786 #endif
787 return (zp);
788 }
789
790 static uint64_t empty_xattr;
791 static uint64_t pad[4];
792 static zfs_acl_phys_t acl_phys;
793 /*
794 * Create a new DMU object to hold a zfs znode.
795 *
796 * IN: dzp - parent directory for new znode
797 * vap - file attributes for new znode
798 * tx - dmu transaction id for zap operations
799 * cr - credentials of caller
800 * flag - flags:
801 * IS_ROOT_NODE - new object will be root
802 * IS_XATTR - new object is an attribute
803 * bonuslen - length of bonus buffer
804 * setaclp - File/Dir initial ACL
805 * fuidp - Tracks fuid allocation.
806 *
807 * OUT: zpp - allocated znode
808 *
809 */
810 #ifdef __NetBSD__
811 struct zfs_newvnode_args {
812 dmu_tx_t *tx;
813 uint_t flag;
814 zfs_acl_ids_t *acl_ids;
815 };
816
817 static void
818 zfs_mknode1(znode_t *, vattr_t *, dmu_tx_t *, cred_t *,
819 uint_t, znode_t **, zfs_acl_ids_t *, vnode_t *);
820
821 int
zfs_loadvnode(struct mount * mp,struct vnode * vp,const void * key,size_t key_len,const void ** new_key)822 zfs_loadvnode(struct mount *mp, struct vnode *vp,
823 const void *key, size_t key_len, const void **new_key)
824 {
825 int err, blksz;
826 uint64_t obj_num;
827 zfsvfs_t *zfsvfs;
828 dmu_buf_t *db;
829 dmu_object_info_t doi;
830 dmu_object_type_t obj_type;
831 sa_handle_t *hdl;
832 znode_t *zp;
833
834 if (key_len != sizeof(obj_num))
835 return zfsctl_loadvnode(mp, vp, key, key_len, new_key);
836
837 memcpy(&obj_num, key, key_len);
838
839 zfsvfs = mp->mnt_data;
840
841 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
842
843 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
844 if (err) {
845 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
846 return (SET_ERROR(err));
847 }
848
849 dmu_object_info_from_db(db, &doi);
850 if (doi.doi_bonus_type != DMU_OT_SA &&
851 (doi.doi_bonus_type != DMU_OT_ZNODE ||
852 (doi.doi_bonus_type == DMU_OT_ZNODE &&
853 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
854 sa_buf_rele(db, NULL);
855 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
856 return (SET_ERROR(EINVAL));
857 }
858 blksz = doi.doi_data_block_size;
859 obj_type = doi.doi_bonus_type;
860 hdl = dmu_buf_get_user(db);
861
862 if (hdl != NULL) {
863 sa_buf_rele(db, NULL);
864 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
865 return (SET_ERROR(ENOENT));
866 }
867
868 zp = zfs_znode_alloc(zfsvfs, db, blksz, obj_type, hdl, vp);
869 if (zp == NULL) {
870 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
871 return (SET_ERROR(ENOENT));
872 }
873 ASSERT(zp == VTOZ(vp));
874 cache_enter_id(vp, zp->z_mode, zp->z_uid, zp->z_gid, true);
875
876 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
877
878 *new_key = &zp->z_id;
879
880 return 0;
881 }
882
883 int
zfs_newvnode(struct mount * mp,vnode_t * dvp,vnode_t * vp,vattr_t * vap,cred_t * cr,void * extra,size_t * key_len,const void ** new_key)884 zfs_newvnode(struct mount *mp, vnode_t *dvp, vnode_t *vp, vattr_t *vap,
885 cred_t *cr, void *extra, size_t *key_len, const void **new_key)
886 {
887 struct zfs_newvnode_args *args = extra;
888 znode_t *zp, *dzp = VTOZ(dvp);
889 dmu_tx_t *tx = args->tx;
890 uint_t flag = args->flag;
891 zfs_acl_ids_t *acl_ids = args->acl_ids;
892
893 zfs_mknode1(dzp, vap, tx, cr, flag, &zp, acl_ids, vp);
894 ASSERT(zp == VTOZ(vp));
895 cache_enter_id(vp, zp->z_mode, zp->z_uid, zp->z_gid, true);
896
897 *key_len = sizeof(zp->z_id);
898 *new_key = &zp->z_id;
899
900 return 0;
901 }
902
903 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)904 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
905 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
906 {
907 vnode_t *vp, *dvp = ZTOV(dzp);
908 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
909 struct zfs_newvnode_args args = { tx, flag, acl_ids };
910
911 if (flag & IS_ROOT_NODE)
912 return zfs_mknode1(dzp, vap, tx, cr, flag, zpp, acl_ids, NULL);
913
914 VERIFY(vcache_new(zfsvfs->z_vfs, dvp, vap, cr, &args, &vp) == 0);
915 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
916 *zpp = VTOZ(vp);
917 }
918
919 static void
zfs_mknode1(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,vnode_t * vp)920 zfs_mknode1(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
921 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids, vnode_t *vp)
922 #else
923 void
924 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
925 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
926 #endif
927 {
928 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
929 uint64_t mode, size, links, parent, pflags;
930 uint64_t dzp_pflags = 0;
931 uint64_t rdev = 0;
932 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
933 dmu_buf_t *db;
934 timestruc_t now;
935 uint64_t gen, obj;
936 int err;
937 int bonuslen;
938 sa_handle_t *sa_hdl;
939 dmu_object_type_t obj_type;
940 sa_bulk_attr_t sa_attrs[ZPL_END];
941 int cnt = 0;
942 zfs_acl_locator_cb_t locate = { 0 };
943
944 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
945
946 if (zfsvfs->z_replay) {
947 obj = vap->va_nodeid;
948 now = vap->va_ctime; /* see zfs_replay_create() */
949 gen = vap->va_nblocks; /* ditto */
950 } else {
951 obj = 0;
952 vfs_timestamp(&now);
953 gen = dmu_tx_get_txg(tx);
954 }
955
956 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
957 bonuslen = (obj_type == DMU_OT_SA) ?
958 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
959
960 /*
961 * Create a new DMU object.
962 */
963 /*
964 * There's currently no mechanism for pre-reading the blocks that will
965 * be needed to allocate a new object, so we accept the small chance
966 * that there will be an i/o error and we will fail one of the
967 * assertions below.
968 */
969 if (vap->va_type == VDIR) {
970 if (zfsvfs->z_replay) {
971 VERIFY0(zap_create_claim_norm(zfsvfs->z_os, obj,
972 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
973 obj_type, bonuslen, tx));
974 } else {
975 obj = zap_create_norm(zfsvfs->z_os,
976 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
977 obj_type, bonuslen, tx);
978 }
979 } else {
980 if (zfsvfs->z_replay) {
981 VERIFY0(dmu_object_claim(zfsvfs->z_os, obj,
982 DMU_OT_PLAIN_FILE_CONTENTS, 0,
983 obj_type, bonuslen, tx));
984 } else {
985 obj = dmu_object_alloc(zfsvfs->z_os,
986 DMU_OT_PLAIN_FILE_CONTENTS, 0,
987 obj_type, bonuslen, tx);
988 }
989 }
990
991 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
992 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
993
994 /*
995 * If this is the root, fix up the half-initialized parent pointer
996 * to reference the just-allocated physical data area.
997 */
998 if (flag & IS_ROOT_NODE) {
999 dzp->z_id = obj;
1000 } else {
1001 dzp_pflags = dzp->z_pflags;
1002 }
1003
1004 /*
1005 * If parent is an xattr, so am I.
1006 */
1007 if (dzp_pflags & ZFS_XATTR) {
1008 flag |= IS_XATTR;
1009 }
1010
1011 if (zfsvfs->z_use_fuids)
1012 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1013 else
1014 pflags = 0;
1015
1016 if (vap->va_type == VDIR) {
1017 size = 2; /* contents ("." and "..") */
1018 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
1019 } else {
1020 size = links = 0;
1021 }
1022
1023 if (vap->va_type == VBLK || vap->va_type == VCHR) {
1024 rdev = zfs_expldev(vap->va_rdev);
1025 }
1026
1027 parent = dzp->z_id;
1028 mode = acl_ids->z_mode;
1029 if (flag & IS_XATTR)
1030 pflags |= ZFS_XATTR;
1031
1032 /*
1033 * No execs denied will be deterimed when zfs_mode_compute() is called.
1034 */
1035 pflags |= acl_ids->z_aclp->z_hints &
1036 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
1037 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
1038
1039 ZFS_TIME_ENCODE(&now, crtime);
1040 ZFS_TIME_ENCODE(&now, ctime);
1041
1042 if (vap->va_mask & AT_ATIME) {
1043 ZFS_TIME_ENCODE(&vap->va_atime, atime);
1044 } else {
1045 ZFS_TIME_ENCODE(&now, atime);
1046 }
1047
1048 if (vap->va_mask & AT_MTIME) {
1049 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
1050 } else {
1051 ZFS_TIME_ENCODE(&now, mtime);
1052 }
1053
1054 /* Now add in all of the "SA" attributes */
1055 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
1056 &sa_hdl));
1057
1058 /*
1059 * Setup the array of attributes to be replaced/set on the new file
1060 *
1061 * order for DMU_OT_ZNODE is critical since it needs to be constructed
1062 * in the old znode_phys_t format. Don't change this ordering
1063 */
1064
1065 if (obj_type == DMU_OT_ZNODE) {
1066 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
1067 NULL, &atime, 16);
1068 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
1069 NULL, &mtime, 16);
1070 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
1071 NULL, &ctime, 16);
1072 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
1073 NULL, &crtime, 16);
1074 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
1075 NULL, &gen, 8);
1076 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
1077 NULL, &mode, 8);
1078 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
1079 NULL, &size, 8);
1080 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
1081 NULL, &parent, 8);
1082 } else {
1083 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
1084 NULL, &mode, 8);
1085 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
1086 NULL, &size, 8);
1087 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
1088 NULL, &gen, 8);
1089 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
1090 &acl_ids->z_fuid, 8);
1091 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
1092 &acl_ids->z_fgid, 8);
1093 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
1094 NULL, &parent, 8);
1095 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
1096 NULL, &pflags, 8);
1097 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
1098 NULL, &atime, 16);
1099 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
1100 NULL, &mtime, 16);
1101 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
1102 NULL, &ctime, 16);
1103 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
1104 NULL, &crtime, 16);
1105 }
1106
1107 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
1108
1109 if (obj_type == DMU_OT_ZNODE) {
1110 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
1111 &empty_xattr, 8);
1112 }
1113 if (obj_type == DMU_OT_ZNODE ||
1114 (vap->va_type == VBLK || vap->va_type == VCHR)) {
1115 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
1116 NULL, &rdev, 8);
1117
1118 }
1119 if (obj_type == DMU_OT_ZNODE) {
1120 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
1121 NULL, &pflags, 8);
1122 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
1123 &acl_ids->z_fuid, 8);
1124 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
1125 &acl_ids->z_fgid, 8);
1126 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
1127 sizeof (uint64_t) * 4);
1128 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
1129 &acl_phys, sizeof (zfs_acl_phys_t));
1130 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
1131 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
1132 &acl_ids->z_aclp->z_acl_count, 8);
1133 locate.cb_aclp = acl_ids->z_aclp;
1134 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
1135 zfs_acl_data_locator, &locate,
1136 acl_ids->z_aclp->z_acl_bytes);
1137 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
1138 acl_ids->z_fuid, acl_ids->z_fgid);
1139 }
1140
1141 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
1142
1143 if (!(flag & IS_ROOT_NODE)) {
1144 #ifdef __NetBSD__
1145 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl, vp);
1146 #else
1147 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
1148 #endif
1149 ASSERT(*zpp != NULL);
1150 } else {
1151 /*
1152 * If we are creating the root node, the "parent" we
1153 * passed in is the znode for the root.
1154 */
1155 *zpp = dzp;
1156
1157 (*zpp)->z_sa_hdl = sa_hdl;
1158 }
1159
1160 (*zpp)->z_pflags = pflags;
1161 (*zpp)->z_mode = mode;
1162
1163 if (vap->va_mask & AT_XVATTR)
1164 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
1165
1166 if (obj_type == DMU_OT_ZNODE ||
1167 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
1168 VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx));
1169 }
1170 #ifndef __NetBSD__
1171 if (!(flag & IS_ROOT_NODE)) {
1172 vnode_t *vp;
1173
1174 vp = ZTOV(*zpp);
1175 vp->v_vflag |= VV_FORCEINSMQ;
1176 err = insmntque(vp, zfsvfs->z_vfs);
1177 vp->v_vflag &= ~VV_FORCEINSMQ;
1178 KASSERT(err == 0, ("insmntque() failed: error %d", err));
1179 }
1180 #endif
1181 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1182 }
1183
1184 /*
1185 * Update in-core attributes. It is assumed the caller will be doing an
1186 * sa_bulk_update to push the changes out.
1187 */
1188 void
zfs_xvattr_set(znode_t * zp,xvattr_t * xvap,dmu_tx_t * tx)1189 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
1190 {
1191 xoptattr_t *xoap;
1192
1193 xoap = xva_getxoptattr(xvap);
1194 ASSERT(xoap);
1195
1196 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1197 uint64_t times[2];
1198 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
1199 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
1200 ×, sizeof (times), tx);
1201 XVA_SET_RTN(xvap, XAT_CREATETIME);
1202 }
1203 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
1204 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
1205 zp->z_pflags, tx);
1206 XVA_SET_RTN(xvap, XAT_READONLY);
1207 }
1208 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
1209 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
1210 zp->z_pflags, tx);
1211 XVA_SET_RTN(xvap, XAT_HIDDEN);
1212 }
1213 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
1214 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
1215 zp->z_pflags, tx);
1216 XVA_SET_RTN(xvap, XAT_SYSTEM);
1217 }
1218 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
1219 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
1220 zp->z_pflags, tx);
1221 XVA_SET_RTN(xvap, XAT_ARCHIVE);
1222 }
1223 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
1224 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
1225 zp->z_pflags, tx);
1226 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
1227 }
1228 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
1229 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
1230 zp->z_pflags, tx);
1231 XVA_SET_RTN(xvap, XAT_NOUNLINK);
1232 }
1233 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
1234 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
1235 zp->z_pflags, tx);
1236 XVA_SET_RTN(xvap, XAT_APPENDONLY);
1237 }
1238 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
1239 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
1240 zp->z_pflags, tx);
1241 XVA_SET_RTN(xvap, XAT_NODUMP);
1242 }
1243 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
1244 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
1245 zp->z_pflags, tx);
1246 XVA_SET_RTN(xvap, XAT_OPAQUE);
1247 }
1248 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1249 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
1250 xoap->xoa_av_quarantined, zp->z_pflags, tx);
1251 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
1252 }
1253 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
1254 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
1255 zp->z_pflags, tx);
1256 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
1257 }
1258 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1259 zfs_sa_set_scanstamp(zp, xvap, tx);
1260 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
1261 }
1262 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
1263 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
1264 zp->z_pflags, tx);
1265 XVA_SET_RTN(xvap, XAT_REPARSE);
1266 }
1267 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
1268 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
1269 zp->z_pflags, tx);
1270 XVA_SET_RTN(xvap, XAT_OFFLINE);
1271 }
1272 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
1273 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
1274 zp->z_pflags, tx);
1275 XVA_SET_RTN(xvap, XAT_SPARSE);
1276 }
1277 }
1278
1279 #ifdef __NetBSD__
1280
1281 int
zfs_zget(zfsvfs_t * zfsvfs,uint64_t obj_num,znode_t ** zpp)1282 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1283 {
1284 int error;
1285 vnode_t *vp;
1286
1287 error = vcache_get(zfsvfs->z_vfs, &obj_num, sizeof(obj_num), &vp);
1288 if (error == 0)
1289 *zpp = VTOZ(vp);
1290
1291 return error;
1292 }
1293
1294 /*
1295 * Get a known cached znode, to be used from zil_commit()->zfs_get_data()
1296 * to resolve log entries. Doesn't take a reference, will never fail and
1297 * depends on zfs_vnops.c::zfs_netbsd_reclaim() running a zil_commit()
1298 * before the znode gets freed.
1299 */
1300 int
zfs_zget_cleaner(zfsvfs_t * zfsvfs,uint64_t obj_num,znode_t ** zpp)1301 zfs_zget_cleaner(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1302 {
1303 dmu_buf_t *db;
1304 sa_handle_t *hdl;
1305 dmu_object_info_t doi;
1306 znode_t *zp;
1307
1308 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1309
1310 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db));
1311
1312 dmu_object_info_from_db(db, &doi);
1313 ASSERT(doi.doi_bonus_type == DMU_OT_SA ||
1314 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1315 doi.doi_bonus_size >= sizeof (znode_phys_t)));
1316
1317 hdl = dmu_buf_get_user(db);
1318 ASSERT3P(hdl, !=, NULL);
1319
1320 zp = sa_get_userdata(hdl);
1321 ASSERT3U(zp->z_id, ==, obj_num);
1322
1323 sa_buf_rele(db, NULL);
1324
1325 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1326
1327 *zpp = zp;
1328 return (0);
1329 }
1330
1331 #else /* __NetBSD__ */
1332
1333 int
zfs_zget(zfsvfs_t * zfsvfs,uint64_t obj_num,znode_t ** zpp)1334 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1335 {
1336 dmu_object_info_t doi;
1337 dmu_buf_t *db;
1338 znode_t *zp;
1339 vnode_t *vp;
1340 sa_handle_t *hdl;
1341 struct thread *td;
1342 int locked;
1343 int err;
1344
1345 td = curthread;
1346 getnewvnode_reserve(1);
1347 again:
1348 *zpp = NULL;
1349 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1350
1351 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1352 if (err) {
1353 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1354 getnewvnode_drop_reserve();
1355 return (err);
1356 }
1357
1358 dmu_object_info_from_db(db, &doi);
1359 if (doi.doi_bonus_type != DMU_OT_SA &&
1360 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1361 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1362 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1363 sa_buf_rele(db, NULL);
1364 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1365 #ifdef __FreeBSD__
1366 getnewvnode_drop_reserve();
1367 #endif
1368 return (SET_ERROR(EINVAL));
1369 }
1370
1371 hdl = dmu_buf_get_user(db);
1372 if (hdl != NULL) {
1373 zp = sa_get_userdata(hdl);
1374
1375 /*
1376 * Since "SA" does immediate eviction we
1377 * should never find a sa handle that doesn't
1378 * know about the znode.
1379 */
1380 ASSERT3P(zp, !=, NULL);
1381 ASSERT3U(zp->z_id, ==, obj_num);
1382 *zpp = zp;
1383 vp = ZTOV(zp);
1384
1385 /* Don't let the vnode disappear after ZFS_OBJ_HOLD_EXIT. */
1386 VN_HOLD(vp);
1387
1388 sa_buf_rele(db, NULL);
1389 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1390
1391 locked = VOP_ISLOCKED(vp);
1392 VI_LOCK(vp);
1393 if ((vp->v_iflag & VI_DOOMED) != 0 &&
1394 locked != LK_EXCLUSIVE) {
1395 /*
1396 * The vnode is doomed and this thread doesn't
1397 * hold the exclusive lock on it, so the vnode
1398 * must be being reclaimed by another thread.
1399 * Otherwise the doomed vnode is being reclaimed
1400 * by this thread and zfs_zget is called from
1401 * ZIL internals.
1402 */
1403 VI_UNLOCK(vp);
1404
1405 /*
1406 * XXX vrele() locks the vnode when the last reference
1407 * is dropped. Although in this case the vnode is
1408 * doomed / dead and so no inactivation is required,
1409 * the vnode lock is still acquired. That could result
1410 * in a LOR with z_teardown_lock if another thread holds
1411 * the vnode's lock and tries to take z_teardown_lock.
1412 * But that is only possible if the other thread peforms
1413 * a ZFS vnode operation on the vnode. That either
1414 * should not happen if the vnode is dead or the thread
1415 * should also have a refrence to the vnode and thus
1416 * our reference is not last.
1417 */
1418 VN_RELE(vp);
1419 goto again;
1420 }
1421 VI_UNLOCK(vp);
1422 getnewvnode_drop_reserve();
1423 return (0);
1424 }
1425
1426 /*
1427 * Not found create new znode/vnode
1428 * but only if file exists.
1429 *
1430 * There is a small window where zfs_vget() could
1431 * find this object while a file create is still in
1432 * progress. This is checked for in zfs_znode_alloc()
1433 *
1434 * if zfs_znode_alloc() fails it will drop the hold on the
1435 * bonus buffer.
1436 */
1437 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1438 doi.doi_bonus_type, NULL);
1439 if (zp == NULL) {
1440 err = SET_ERROR(ENOENT);
1441 } else {
1442 *zpp = zp;
1443 }
1444 if (err == 0) {
1445 vnode_t *vp = ZTOV(zp);
1446
1447 err = insmntque(vp, zfsvfs->z_vfs);
1448 if (err == 0) {
1449 vp->v_hash = obj_num;
1450 VOP_UNLOCK(vp, 0);
1451 } else {
1452 zp->z_vnode = NULL;
1453 zfs_znode_dmu_fini(zp);
1454 zfs_znode_free(zp);
1455 *zpp = NULL;
1456 }
1457 }
1458 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1459 getnewvnode_drop_reserve();
1460 return (err);
1461 }
1462
1463 #endif /* __NetBSD__ */
1464
1465 int
zfs_rezget(znode_t * zp)1466 zfs_rezget(znode_t *zp)
1467 {
1468 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1469 dmu_object_info_t doi;
1470 dmu_buf_t *db;
1471 vnode_t *vp;
1472 uint64_t obj_num = zp->z_id;
1473 uint64_t mode, size;
1474 sa_bulk_attr_t bulk[8];
1475 int err;
1476 int count = 0;
1477 uint64_t gen;
1478
1479 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1480
1481 mutex_enter(&zp->z_acl_lock);
1482 if (zp->z_acl_cached) {
1483 zfs_acl_free(zp->z_acl_cached);
1484 zp->z_acl_cached = NULL;
1485 }
1486
1487 mutex_exit(&zp->z_acl_lock);
1488 ASSERT(zp->z_sa_hdl == NULL);
1489 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1490 if (err) {
1491 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1492 return (err);
1493 }
1494
1495 dmu_object_info_from_db(db, &doi);
1496 if (doi.doi_bonus_type != DMU_OT_SA &&
1497 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1498 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1499 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1500 sa_buf_rele(db, NULL);
1501 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1502 return (SET_ERROR(EINVAL));
1503 }
1504
1505 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1506 size = zp->z_size;
1507
1508 /* reload cached values */
1509 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1510 &gen, sizeof (gen));
1511 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1512 &zp->z_size, sizeof (zp->z_size));
1513 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1514 &zp->z_links, sizeof (zp->z_links));
1515 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1516 &zp->z_pflags, sizeof (zp->z_pflags));
1517 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1518 &zp->z_atime, sizeof (zp->z_atime));
1519 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1520 &zp->z_uid, sizeof (zp->z_uid));
1521 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1522 &zp->z_gid, sizeof (zp->z_gid));
1523 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1524 &mode, sizeof (mode));
1525
1526 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1527 zfs_znode_dmu_fini(zp);
1528 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1529 return (SET_ERROR(EIO));
1530 }
1531
1532 zp->z_mode = mode;
1533
1534 if (gen != zp->z_gen) {
1535 zfs_znode_dmu_fini(zp);
1536 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1537 return (SET_ERROR(EIO));
1538 }
1539
1540 /*
1541 * It is highly improbable but still quite possible that two
1542 * objects in different datasets are created with the same
1543 * object numbers and in transaction groups with the same
1544 * numbers. znodes corresponding to those objects would
1545 * have the same z_id and z_gen, but their other attributes
1546 * may be different.
1547 * zfs recv -F may replace one of such objects with the other.
1548 * As a result file properties recorded in the replaced
1549 * object's vnode may no longer match the received object's
1550 * properties. At present the only cached property is the
1551 * files type recorded in v_type.
1552 * So, handle this case by leaving the old vnode and znode
1553 * disassociated from the actual object. A new vnode and a
1554 * znode will be created if the object is accessed
1555 * (e.g. via a look-up). The old vnode and znode will be
1556 * recycled when the last vnode reference is dropped.
1557 */
1558 vp = ZTOV(zp);
1559 if (vp->v_type != IFTOVT((mode_t)zp->z_mode)) {
1560 zfs_znode_dmu_fini(zp);
1561 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1562 return (EIO);
1563 }
1564
1565 zp->z_unlinked = (zp->z_links == 0);
1566 zp->z_blksz = doi.doi_data_block_size;
1567 #ifdef __NetBSD__
1568 rw_enter(vp->v_uobj.vmobjlock, RW_WRITER);
1569 (void)VOP_PUTPAGES(vp, 0, 0, PGO_ALLPAGES|PGO_FREE|PGO_SYNCIO);
1570 #else
1571 vn_pages_remove(vp, 0, 0);
1572 #endif
1573 if (zp->z_size != size)
1574 vnode_pager_setsize(vp, zp->z_size);
1575 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1576
1577 return (0);
1578 }
1579
1580 void
zfs_znode_delete(znode_t * zp,dmu_tx_t * tx)1581 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1582 {
1583 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1584 objset_t *os = zfsvfs->z_os;
1585 uint64_t obj = zp->z_id;
1586 uint64_t acl_obj = zfs_external_acl(zp);
1587
1588 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1589 if (acl_obj) {
1590 VERIFY(!zp->z_is_sa);
1591 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1592 }
1593 VERIFY(0 == dmu_object_free(os, obj, tx));
1594 zfs_znode_dmu_fini(zp);
1595 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1596 zfs_znode_free(zp);
1597 }
1598
1599 void
zfs_zinactive(znode_t * zp)1600 zfs_zinactive(znode_t *zp)
1601 {
1602 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1603 uint64_t z_id = zp->z_id;
1604
1605 ASSERT(zp->z_sa_hdl);
1606
1607 /*
1608 * Don't allow a zfs_zget() while were trying to release this znode
1609 */
1610 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1611
1612 /*
1613 * If this was the last reference to a file with no links,
1614 * remove the file from the file system.
1615 */
1616 if (zp->z_unlinked) {
1617 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1618 zfs_rmnode(zp);
1619 return;
1620 }
1621
1622 zfs_znode_dmu_fini(zp);
1623 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1624 zfs_znode_free(zp);
1625 }
1626
1627 void
zfs_znode_free(znode_t * zp)1628 zfs_znode_free(znode_t *zp)
1629 {
1630 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1631
1632 #ifdef __NetBSD__
1633 struct vnode *vp = ZTOV(zp);
1634
1635 genfs_node_destroy(vp);
1636
1637 /*
1638 * Interlock with zfs_sync().
1639 */
1640 mutex_enter(vp->v_interlock);
1641 vp->v_data = NULL;
1642 mutex_exit(vp->v_interlock);
1643 #endif
1644
1645 ASSERT(zp->z_sa_hdl == NULL);
1646 zp->z_vnode = NULL;
1647 mutex_enter(&zfsvfs->z_znodes_lock);
1648 POINTER_INVALIDATE(&zp->z_zfsvfs);
1649 list_remove(&zfsvfs->z_all_znodes, zp);
1650 mutex_exit(&zfsvfs->z_znodes_lock);
1651
1652 if (zp->z_acl_cached) {
1653 zfs_acl_free(zp->z_acl_cached);
1654 zp->z_acl_cached = NULL;
1655 }
1656
1657 kmem_cache_free(znode_cache, zp);
1658
1659 #ifdef illumos
1660 VFS_RELE(zfsvfs->z_vfs);
1661 #endif
1662 }
1663
1664 void
zfs_tstamp_update_setup(znode_t * zp,uint_t flag,uint64_t mtime[2],uint64_t ctime[2],boolean_t have_tx)1665 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1666 uint64_t ctime[2], boolean_t have_tx)
1667 {
1668 timestruc_t now;
1669
1670 vfs_timestamp(&now);
1671
1672 if (have_tx) { /* will sa_bulk_update happen really soon? */
1673 zp->z_atime_dirty = 0;
1674 zp->z_seq++;
1675 } else {
1676 zp->z_atime_dirty = 1;
1677 }
1678
1679 if (flag & AT_ATIME) {
1680 ZFS_TIME_ENCODE(&now, zp->z_atime);
1681 }
1682
1683 if (flag & AT_MTIME) {
1684 ZFS_TIME_ENCODE(&now, mtime);
1685 if (zp->z_zfsvfs->z_use_fuids) {
1686 zp->z_pflags |= (ZFS_ARCHIVE |
1687 ZFS_AV_MODIFIED);
1688 }
1689 }
1690
1691 if (flag & AT_CTIME) {
1692 ZFS_TIME_ENCODE(&now, ctime);
1693 if (zp->z_zfsvfs->z_use_fuids)
1694 zp->z_pflags |= ZFS_ARCHIVE;
1695 }
1696 }
1697
1698 /*
1699 * Grow the block size for a file.
1700 *
1701 * IN: zp - znode of file to free data in.
1702 * size - requested block size
1703 * tx - open transaction.
1704 *
1705 * NOTE: this function assumes that the znode is write locked.
1706 */
1707 void
zfs_grow_blocksize(znode_t * zp,uint64_t size,dmu_tx_t * tx)1708 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1709 {
1710 int error;
1711 u_longlong_t dummy;
1712
1713 if (size <= zp->z_blksz)
1714 return;
1715 /*
1716 * If the file size is already greater than the current blocksize,
1717 * we will not grow. If there is more than one block in a file,
1718 * the blocksize cannot change.
1719 */
1720 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1721 return;
1722
1723 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1724 size, 0, tx);
1725
1726 if (error == ENOTSUP)
1727 return;
1728 ASSERT0(error);
1729
1730 /* What blocksize did we actually get? */
1731 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1732 }
1733
1734 #ifdef illumos
1735 /*
1736 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1737 * be calling back into the fs for a putpage(). E.g.: when truncating
1738 * a file, the pages being "thrown away* don't need to be written out.
1739 */
1740 /* ARGSUSED */
1741 static int
zfs_no_putpage(vnode_t * vp,page_t * pp,u_offset_t * offp,size_t * lenp,int flags,cred_t * cr)1742 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1743 int flags, cred_t *cr)
1744 {
1745 ASSERT(0);
1746 return (0);
1747 }
1748 #endif
1749
1750 /*
1751 * Increase the file length
1752 *
1753 * IN: zp - znode of file to free data in.
1754 * end - new end-of-file
1755 *
1756 * RETURN: 0 on success, error code on failure
1757 */
1758 static int
zfs_extend(znode_t * zp,uint64_t end)1759 zfs_extend(znode_t *zp, uint64_t end)
1760 {
1761 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1762 dmu_tx_t *tx;
1763 rl_t *rl;
1764 uint64_t newblksz;
1765 int error;
1766
1767 /*
1768 * We will change zp_size, lock the whole file.
1769 */
1770 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1771
1772 /*
1773 * Nothing to do if file already at desired length.
1774 */
1775 if (end <= zp->z_size) {
1776 zfs_range_unlock(rl);
1777 return (0);
1778 }
1779 tx = dmu_tx_create(zfsvfs->z_os);
1780 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1781 zfs_sa_upgrade_txholds(tx, zp);
1782 if (end > zp->z_blksz &&
1783 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1784 /*
1785 * We are growing the file past the current block size.
1786 */
1787 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1788 /*
1789 * File's blocksize is already larger than the
1790 * "recordsize" property. Only let it grow to
1791 * the next power of 2.
1792 */
1793 ASSERT(!ISP2(zp->z_blksz));
1794 newblksz = MIN(end, 1 << highbit64(zp->z_blksz));
1795 } else {
1796 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1797 }
1798 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1799 } else {
1800 newblksz = 0;
1801 }
1802
1803 error = dmu_tx_assign(tx, TXG_WAIT);
1804 if (error) {
1805 dmu_tx_abort(tx);
1806 zfs_range_unlock(rl);
1807 return (error);
1808 }
1809
1810 if (newblksz)
1811 zfs_grow_blocksize(zp, newblksz, tx);
1812
1813 zp->z_size = end;
1814
1815 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1816 &zp->z_size, sizeof (zp->z_size), tx));
1817
1818 vnode_pager_setsize(ZTOV(zp), end);
1819
1820 zfs_range_unlock(rl);
1821
1822 dmu_tx_commit(tx);
1823
1824 return (0);
1825 }
1826
1827 /*
1828 * Free space in a file.
1829 *
1830 * IN: zp - znode of file to free data in.
1831 * off - start of section to free.
1832 * len - length of section to free.
1833 *
1834 * RETURN: 0 on success, error code on failure
1835 */
1836 static int
zfs_free_range(znode_t * zp,uint64_t off,uint64_t len)1837 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1838 {
1839 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1840 rl_t *rl;
1841 int error;
1842
1843 /*
1844 * Lock the range being freed.
1845 */
1846 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1847
1848 /*
1849 * Nothing to do if file already at desired length.
1850 */
1851 if (off >= zp->z_size) {
1852 zfs_range_unlock(rl);
1853 return (0);
1854 }
1855
1856 if (off + len > zp->z_size)
1857 len = zp->z_size - off;
1858
1859 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1860
1861 if (error == 0) {
1862 /*
1863 * In FreeBSD we cannot free block in the middle of a file,
1864 * but only at the end of a file, so this code path should
1865 * never happen.
1866 */
1867 vnode_pager_setsize(ZTOV(zp), off);
1868 }
1869
1870 zfs_range_unlock(rl);
1871
1872 return (error);
1873 }
1874
1875 /*
1876 * Truncate a file
1877 *
1878 * IN: zp - znode of file to free data in.
1879 * end - new end-of-file.
1880 *
1881 * RETURN: 0 on success, error code on failure
1882 */
1883 static int
zfs_trunc(znode_t * zp,uint64_t end)1884 zfs_trunc(znode_t *zp, uint64_t end)
1885 {
1886 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1887 vnode_t *vp = ZTOV(zp);
1888 dmu_tx_t *tx;
1889 rl_t *rl;
1890 int error;
1891 sa_bulk_attr_t bulk[2];
1892 int count = 0;
1893
1894 /*
1895 * We will change zp_size, lock the whole file.
1896 */
1897 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1898
1899 /*
1900 * Nothing to do if file already at desired length.
1901 */
1902 if (end >= zp->z_size) {
1903 zfs_range_unlock(rl);
1904 return (0);
1905 }
1906
1907 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1908 if (error) {
1909 zfs_range_unlock(rl);
1910 return (error);
1911 }
1912 tx = dmu_tx_create(zfsvfs->z_os);
1913 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1914 zfs_sa_upgrade_txholds(tx, zp);
1915 dmu_tx_mark_netfree(tx);
1916 error = dmu_tx_assign(tx, TXG_WAIT);
1917 if (error) {
1918 dmu_tx_abort(tx);
1919 zfs_range_unlock(rl);
1920 return (error);
1921 }
1922
1923 zp->z_size = end;
1924 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1925 NULL, &zp->z_size, sizeof (zp->z_size));
1926
1927 if (end == 0) {
1928 zp->z_pflags &= ~ZFS_SPARSE;
1929 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1930 NULL, &zp->z_pflags, 8);
1931 }
1932 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1933
1934 dmu_tx_commit(tx);
1935
1936 /*
1937 * Clear any mapped pages in the truncated region. This has to
1938 * happen outside of the transaction to avoid the possibility of
1939 * a deadlock with someone trying to push a page that we are
1940 * about to invalidate.
1941 */
1942 vnode_pager_setsize(vp, end);
1943
1944 zfs_range_unlock(rl);
1945
1946 return (0);
1947 }
1948
1949 /*
1950 * Free space in a file
1951 *
1952 * IN: zp - znode of file to free data in.
1953 * off - start of range
1954 * len - end of range (0 => EOF)
1955 * flag - current file open mode flags.
1956 * log - TRUE if this action should be logged
1957 *
1958 * RETURN: 0 on success, error code on failure
1959 */
1960 int
zfs_freesp(znode_t * zp,uint64_t off,uint64_t len,int flag,boolean_t log)1961 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1962 {
1963 vnode_t *vp = ZTOV(zp);
1964 dmu_tx_t *tx;
1965 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1966 zilog_t *zilog = zfsvfs->z_log;
1967 uint64_t mode;
1968 uint64_t mtime[2], ctime[2];
1969 sa_bulk_attr_t bulk[3];
1970 int count = 0;
1971 int error;
1972
1973 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1974 sizeof (mode))) != 0)
1975 return (error);
1976
1977 if (off > zp->z_size) {
1978 error = zfs_extend(zp, off+len);
1979 if (error == 0 && log)
1980 goto log;
1981 else
1982 return (error);
1983 }
1984
1985 /*
1986 * Check for any locks in the region to be freed.
1987 */
1988
1989 if (MANDLOCK(vp, (mode_t)mode)) {
1990 uint64_t length = (len ? len : zp->z_size - off);
1991 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1992 return (error);
1993 }
1994
1995 if (len == 0) {
1996 error = zfs_trunc(zp, off);
1997 } else {
1998 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1999 off + len > zp->z_size)
2000 error = zfs_extend(zp, off+len);
2001 }
2002 if (error || !log)
2003 return (error);
2004 log:
2005 tx = dmu_tx_create(zfsvfs->z_os);
2006 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
2007 zfs_sa_upgrade_txholds(tx, zp);
2008 error = dmu_tx_assign(tx, TXG_WAIT);
2009 if (error) {
2010 dmu_tx_abort(tx);
2011 return (error);
2012 }
2013
2014 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
2015 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
2016 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
2017 NULL, &zp->z_pflags, 8);
2018 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
2019 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
2020 ASSERT(error == 0);
2021
2022 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
2023
2024 dmu_tx_commit(tx);
2025 return (0);
2026 }
2027
2028 void
zfs_create_fs(objset_t * os,cred_t * cr,nvlist_t * zplprops,dmu_tx_t * tx)2029 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
2030 {
2031 uint64_t moid, obj, sa_obj, version;
2032 uint64_t sense = ZFS_CASE_SENSITIVE;
2033 uint64_t norm = 0;
2034 nvpair_t *elem;
2035 int error;
2036 int i;
2037 znode_t *rootzp = NULL;
2038 zfsvfs_t *zfsvfs;
2039 vattr_t vattr;
2040 znode_t *zp;
2041 zfs_acl_ids_t acl_ids;
2042
2043 /*
2044 * First attempt to create master node.
2045 */
2046 /*
2047 * In an empty objset, there are no blocks to read and thus
2048 * there can be no i/o errors (which we assert below).
2049 */
2050 moid = MASTER_NODE_OBJ;
2051 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
2052 DMU_OT_NONE, 0, tx);
2053 ASSERT(error == 0);
2054
2055 /*
2056 * Set starting attributes.
2057 */
2058 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
2059 elem = NULL;
2060 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
2061 /* For the moment we expect all zpl props to be uint64_ts */
2062 uint64_t val;
2063 char *name;
2064
2065 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
2066 VERIFY(nvpair_value_uint64(elem, &val) == 0);
2067 name = nvpair_name(elem);
2068 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
2069 if (val < version)
2070 version = val;
2071 } else {
2072 error = zap_update(os, moid, name, 8, 1, &val, tx);
2073 }
2074 ASSERT(error == 0);
2075 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
2076 norm = val;
2077 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
2078 sense = val;
2079 }
2080 ASSERT(version != 0);
2081 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
2082
2083 /*
2084 * Create zap object used for SA attribute registration
2085 */
2086
2087 if (version >= ZPL_VERSION_SA) {
2088 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
2089 DMU_OT_NONE, 0, tx);
2090 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
2091 ASSERT(error == 0);
2092 } else {
2093 sa_obj = 0;
2094 }
2095 /*
2096 * Create a delete queue.
2097 */
2098 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
2099
2100 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
2101 ASSERT(error == 0);
2102
2103 /*
2104 * Create root znode. Create minimal znode/vnode/zfsvfs
2105 * to allow zfs_mknode to work.
2106 */
2107 VATTR_NULL(&vattr);
2108 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
2109 vattr.va_type = VDIR;
2110 vattr.va_mode = S_IFDIR|0755;
2111 vattr.va_uid = crgetuid(cr);
2112 vattr.va_gid = crgetgid(cr);
2113
2114 zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP);
2115
2116 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
2117 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
2118 rootzp->z_moved = 0;
2119 rootzp->z_unlinked = 0;
2120 rootzp->z_atime_dirty = 0;
2121 rootzp->z_is_sa = USE_SA(version, os);
2122
2123 zfsvfs->z_os = os;
2124 zfsvfs->z_parent = zfsvfs;
2125 zfsvfs->z_version = version;
2126 zfsvfs->z_use_fuids = USE_FUIDS(version, os);
2127 zfsvfs->z_use_sa = USE_SA(version, os);
2128 zfsvfs->z_norm = norm;
2129
2130 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
2131 &zfsvfs->z_attr_table);
2132
2133 ASSERT(error == 0);
2134
2135 /*
2136 * Fold case on file systems that are always or sometimes case
2137 * insensitive.
2138 */
2139 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
2140 zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
2141
2142 mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
2143 list_create(&zfsvfs->z_all_znodes, sizeof (znode_t),
2144 offsetof(znode_t, z_link_node));
2145
2146 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
2147 mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
2148
2149 rootzp->z_zfsvfs = zfsvfs;
2150 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
2151 cr, NULL, &acl_ids));
2152 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
2153 ASSERT3P(zp, ==, rootzp);
2154 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
2155 ASSERT(error == 0);
2156 zfs_acl_ids_free(&acl_ids);
2157 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
2158
2159 sa_handle_destroy(rootzp->z_sa_hdl);
2160 kmem_cache_free(znode_cache, rootzp);
2161
2162 /*
2163 * Create shares directory
2164 */
2165
2166 error = zfs_create_share_dir(zfsvfs, tx);
2167
2168 ASSERT(error == 0);
2169
2170 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
2171 mutex_destroy(&zfsvfs->z_hold_mtx[i]);
2172 mutex_destroy(&zfsvfs->z_znodes_lock);
2173 kmem_free(zfsvfs, sizeof (zfsvfs_t));
2174 }
2175 #endif /* _KERNEL */
2176
2177 static int
zfs_sa_setup(objset_t * osp,sa_attr_type_t ** sa_table)2178 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
2179 {
2180 uint64_t sa_obj = 0;
2181 int error;
2182
2183 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
2184 if (error != 0 && error != ENOENT)
2185 return (error);
2186
2187 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
2188 return (error);
2189 }
2190
2191 static int
zfs_grab_sa_handle(objset_t * osp,uint64_t obj,sa_handle_t ** hdlp,dmu_buf_t ** db,void * tag)2192 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
2193 dmu_buf_t **db, void *tag)
2194 {
2195 dmu_object_info_t doi;
2196 int error;
2197
2198 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
2199 return (error);
2200
2201 dmu_object_info_from_db(*db, &doi);
2202 if ((doi.doi_bonus_type != DMU_OT_SA &&
2203 doi.doi_bonus_type != DMU_OT_ZNODE) ||
2204 doi.doi_bonus_type == DMU_OT_ZNODE &&
2205 doi.doi_bonus_size < sizeof (znode_phys_t)) {
2206 sa_buf_rele(*db, tag);
2207 return (SET_ERROR(ENOTSUP));
2208 }
2209
2210 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
2211 if (error != 0) {
2212 sa_buf_rele(*db, tag);
2213 return (error);
2214 }
2215
2216 return (0);
2217 }
2218
2219 void
zfs_release_sa_handle(sa_handle_t * hdl,dmu_buf_t * db,void * tag)2220 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
2221 {
2222 sa_handle_destroy(hdl);
2223 sa_buf_rele(db, tag);
2224 }
2225
2226 /*
2227 * Given an object number, return its parent object number and whether
2228 * or not the object is an extended attribute directory.
2229 */
2230 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)2231 zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table,
2232 uint64_t *pobjp, int *is_xattrdir)
2233 {
2234 uint64_t parent;
2235 uint64_t pflags;
2236 uint64_t mode;
2237 uint64_t parent_mode;
2238 sa_bulk_attr_t bulk[3];
2239 sa_handle_t *sa_hdl;
2240 dmu_buf_t *sa_db;
2241 int count = 0;
2242 int error;
2243
2244 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
2245 &parent, sizeof (parent));
2246 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
2247 &pflags, sizeof (pflags));
2248 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2249 &mode, sizeof (mode));
2250
2251 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
2252 return (error);
2253
2254 /*
2255 * When a link is removed its parent pointer is not changed and will
2256 * be invalid. There are two cases where a link is removed but the
2257 * file stays around, when it goes to the delete queue and when there
2258 * are additional links.
2259 */
2260 error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG);
2261 if (error != 0)
2262 return (error);
2263
2264 error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode));
2265 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2266 if (error != 0)
2267 return (error);
2268
2269 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
2270
2271 /*
2272 * Extended attributes can be applied to files, directories, etc.
2273 * Otherwise the parent must be a directory.
2274 */
2275 if (!*is_xattrdir && !S_ISDIR(parent_mode))
2276 return (SET_ERROR(EINVAL));
2277
2278 *pobjp = parent;
2279
2280 return (0);
2281 }
2282
2283 /*
2284 * Given an object number, return some zpl level statistics
2285 */
2286 static int
zfs_obj_to_stats_impl(sa_handle_t * hdl,sa_attr_type_t * sa_table,zfs_stat_t * sb)2287 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
2288 zfs_stat_t *sb)
2289 {
2290 sa_bulk_attr_t bulk[4];
2291 int count = 0;
2292
2293 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
2294 &sb->zs_mode, sizeof (sb->zs_mode));
2295 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
2296 &sb->zs_gen, sizeof (sb->zs_gen));
2297 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
2298 &sb->zs_links, sizeof (sb->zs_links));
2299 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
2300 &sb->zs_ctime, sizeof (sb->zs_ctime));
2301
2302 return (sa_bulk_lookup(hdl, bulk, count));
2303 }
2304
2305 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)2306 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
2307 sa_attr_type_t *sa_table, char *buf, int len)
2308 {
2309 sa_handle_t *sa_hdl;
2310 sa_handle_t *prevhdl = NULL;
2311 dmu_buf_t *prevdb = NULL;
2312 dmu_buf_t *sa_db = NULL;
2313 char *path = buf + len - 1;
2314 int error;
2315
2316 *path = '\0';
2317 sa_hdl = hdl;
2318
2319 for (;;) {
2320 uint64_t pobj;
2321 char component[MAXNAMELEN + 2];
2322 size_t complen;
2323 int is_xattrdir;
2324
2325 if (prevdb)
2326 zfs_release_sa_handle(prevhdl, prevdb, FTAG);
2327
2328 if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj,
2329 &is_xattrdir)) != 0)
2330 break;
2331
2332 if (pobj == obj) {
2333 if (path[0] != '/')
2334 *--path = '/';
2335 break;
2336 }
2337
2338 component[0] = '/';
2339 if (is_xattrdir) {
2340 (void) sprintf(component + 1, "<xattrdir>");
2341 } else {
2342 error = zap_value_search(osp, pobj, obj,
2343 ZFS_DIRENT_OBJ(-1ULL), component + 1);
2344 if (error != 0)
2345 break;
2346 }
2347
2348 complen = strlen(component);
2349 path -= complen;
2350 ASSERT(path >= buf);
2351 bcopy(component, path, complen);
2352 obj = pobj;
2353
2354 if (sa_hdl != hdl) {
2355 prevhdl = sa_hdl;
2356 prevdb = sa_db;
2357 }
2358 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
2359 if (error != 0) {
2360 sa_hdl = prevhdl;
2361 sa_db = prevdb;
2362 break;
2363 }
2364 }
2365
2366 if (sa_hdl != NULL && sa_hdl != hdl) {
2367 ASSERT(sa_db != NULL);
2368 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2369 }
2370
2371 if (error == 0)
2372 (void) memmove(buf, path, buf + len - path);
2373
2374 return (error);
2375 }
2376
2377 int
zfs_obj_to_path(objset_t * osp,uint64_t obj,char * buf,int len)2378 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
2379 {
2380 sa_attr_type_t *sa_table;
2381 sa_handle_t *hdl;
2382 dmu_buf_t *db;
2383 int error;
2384
2385 error = zfs_sa_setup(osp, &sa_table);
2386 if (error != 0)
2387 return (error);
2388
2389 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2390 if (error != 0)
2391 return (error);
2392
2393 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2394
2395 zfs_release_sa_handle(hdl, db, FTAG);
2396 return (error);
2397 }
2398
2399 int
zfs_obj_to_stats(objset_t * osp,uint64_t obj,zfs_stat_t * sb,char * buf,int len)2400 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
2401 char *buf, int len)
2402 {
2403 char *path = buf + len - 1;
2404 sa_attr_type_t *sa_table;
2405 sa_handle_t *hdl;
2406 dmu_buf_t *db;
2407 int error;
2408
2409 *path = '\0';
2410
2411 error = zfs_sa_setup(osp, &sa_table);
2412 if (error != 0)
2413 return (error);
2414
2415 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2416 if (error != 0)
2417 return (error);
2418
2419 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
2420 if (error != 0) {
2421 zfs_release_sa_handle(hdl, db, FTAG);
2422 return (error);
2423 }
2424
2425 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2426
2427 zfs_release_sa_handle(hdl, db, FTAG);
2428 return (error);
2429 }
2430
2431 #ifdef _KERNEL
2432 int
zfs_znode_parent_and_name(znode_t * zp,znode_t ** dzpp,char * buf)2433 zfs_znode_parent_and_name(znode_t *zp, znode_t **dzpp, char *buf)
2434 {
2435 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
2436 uint64_t parent;
2437 int is_xattrdir;
2438 int err;
2439
2440 /* Extended attributes should not be visible as regular files. */
2441 if ((zp->z_pflags & ZFS_XATTR) != 0)
2442 return (SET_ERROR(EINVAL));
2443
2444 err = zfs_obj_to_pobj(zfsvfs->z_os, zp->z_sa_hdl, zfsvfs->z_attr_table,
2445 &parent, &is_xattrdir);
2446 if (err != 0)
2447 return (err);
2448 ASSERT0(is_xattrdir);
2449
2450 /* No name as this is a root object. */
2451 if (parent == zp->z_id)
2452 return (SET_ERROR(EINVAL));
2453
2454 err = zap_value_search(zfsvfs->z_os, parent, zp->z_id,
2455 ZFS_DIRENT_OBJ(-1ULL), buf);
2456 if (err != 0)
2457 return (err);
2458 err = zfs_zget(zfsvfs, parent, dzpp);
2459 return (err);
2460 }
2461 #endif /* _KERNEL */
2462