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 */
24
25 /* Portions Copyright 2007 Jeremy Teo */
26
27 #ifdef _KERNEL
28 #include <sys/types.h>
29 #include <sys/param.h>
30 #include <sys/time.h>
31 #include <sys/systm.h>
32 #include <sys/sysmacros.h>
33 #include <sys/resource.h>
34 #include <sys/mntent.h>
35 #include <sys/mkdev.h>
36 #include <sys/u8_textprep.h>
37 #include <sys/dsl_dataset.h>
38 #include <sys/vfs.h>
39 #include <sys/vfs_opreg.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/mode.h>
46 #include <sys/atomic.h>
47 #include <vm/pvn.h>
48 #include "fs/fs_subr.h"
49 #include <sys/zfs_dir.h>
50 #include <sys/zfs_acl.h>
51 #include <sys/zfs_ioctl.h>
52 #include <sys/zfs_rlock.h>
53 #include <sys/zfs_fuid.h>
54 #include <sys/dnode.h>
55 #include <sys/fs/zfs.h>
56 #include <sys/kidmap.h>
57 #endif /* _KERNEL */
58
59 #include <sys/dmu.h>
60 #include <sys/refcount.h>
61 #include <sys/stat.h>
62 #include <sys/zap.h>
63 #include <sys/zfs_znode.h>
64 #include <sys/sa.h>
65 #include <sys/zfs_sa.h>
66 #include <sys/zfs_stat.h>
67
68 #include "zfs_prop.h"
69 #include "zfs_comutil.h"
70
71 /*
72 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
73 * turned on when DEBUG is also defined.
74 */
75 #ifdef DEBUG
76 #define ZNODE_STATS
77 #endif /* DEBUG */
78
79 #ifdef ZNODE_STATS
80 #define ZNODE_STAT_ADD(stat) ((stat)++)
81 #else
82 #define ZNODE_STAT_ADD(stat) /* nothing */
83 #endif /* ZNODE_STATS */
84
85 /*
86 * Functions needed for userland (ie: libzpool) are not put under
87 * #ifdef_KERNEL; the rest of the functions have dependencies
88 * (such as VFS logic) that will not compile easily in userland.
89 */
90 #ifdef _KERNEL
91 /*
92 * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to
93 * be freed before it can be safely accessed.
94 */
95 krwlock_t zfsvfs_lock;
96
97 static kmem_cache_t *znode_cache = NULL;
98
99 /*ARGSUSED*/
100 static void
101 znode_evict_error(dmu_buf_t *dbuf, void *user_ptr)
102 {
103 /*
104 * We should never drop all dbuf refs without first clearing
105 * the eviction callback.
106 */
107 panic("evicting znode %p\n", user_ptr);
108 }
109
110 /*ARGSUSED*/
111 static int
112 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
113 {
114 znode_t *zp = buf;
115
116 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
117
118 zp->z_vnode = vn_alloc(kmflags);
119 if (zp->z_vnode == NULL) {
120 return (-1);
121 }
122 ZTOV(zp)->v_data = zp;
123
124 list_link_init(&zp->z_link_node);
125
126 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
127 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
128 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
129 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
130
131 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
132 avl_create(&zp->z_range_avl, zfs_range_compare,
133 sizeof (rl_t), offsetof(rl_t, r_node));
134
135 zp->z_dirlocks = NULL;
136 zp->z_acl_cached = NULL;
137 zp->z_moved = 0;
138 return (0);
139 }
140
141 /*ARGSUSED*/
142 static void
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)->v_data == zp);
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 static void
175 zfs_znode_move_impl(znode_t *ozp, znode_t *nzp)
176 {
177 vnode_t *vp;
178
179 /* Copy fields. */
180 nzp->z_zfsvfs = ozp->z_zfsvfs;
181
182 /* Swap vnodes. */
183 vp = nzp->z_vnode;
184 nzp->z_vnode = ozp->z_vnode;
185 ozp->z_vnode = vp; /* let destructor free the overwritten vnode */
186 ZTOV(ozp)->v_data = ozp;
187 ZTOV(nzp)->v_data = nzp;
188
189 nzp->z_id = ozp->z_id;
190 ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */
191 ASSERT(avl_numnodes(&ozp->z_range_avl) == 0);
192 nzp->z_unlinked = ozp->z_unlinked;
193 nzp->z_atime_dirty = ozp->z_atime_dirty;
194 nzp->z_zn_prefetch = ozp->z_zn_prefetch;
195 nzp->z_blksz = ozp->z_blksz;
196 nzp->z_seq = ozp->z_seq;
197 nzp->z_mapcnt = ozp->z_mapcnt;
198 nzp->z_gen = ozp->z_gen;
199 nzp->z_sync_cnt = ozp->z_sync_cnt;
200 nzp->z_is_sa = ozp->z_is_sa;
201 nzp->z_sa_hdl = ozp->z_sa_hdl;
202 bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2);
203 nzp->z_links = ozp->z_links;
204 nzp->z_size = ozp->z_size;
205 nzp->z_pflags = ozp->z_pflags;
206 nzp->z_uid = ozp->z_uid;
207 nzp->z_gid = ozp->z_gid;
208 nzp->z_mode = ozp->z_mode;
209
210 /*
211 * Since this is just an idle znode and kmem is already dealing with
212 * memory pressure, release any cached ACL.
213 */
214 if (ozp->z_acl_cached) {
215 zfs_acl_free(ozp->z_acl_cached);
216 ozp->z_acl_cached = NULL;
217 }
218
219 sa_set_userp(nzp->z_sa_hdl, nzp);
220
221 /*
222 * Invalidate the original znode by clearing fields that provide a
223 * pointer back to the znode. Set the low bit of the vfs pointer to
224 * ensure that zfs_znode_move() recognizes the znode as invalid in any
225 * subsequent callback.
226 */
227 ozp->z_sa_hdl = NULL;
228 POINTER_INVALIDATE(&ozp->z_zfsvfs);
229
230 /*
231 * Mark the znode.
232 */
233 nzp->z_moved = 1;
234 ozp->z_moved = (uint8_t)-1;
235 }
236
237 /*ARGSUSED*/
238 static kmem_cbrc_t
239 zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg)
240 {
241 znode_t *ozp = buf, *nzp = newbuf;
242 zfsvfs_t *zfsvfs;
243 vnode_t *vp;
244
245 /*
246 * The znode is on the file system's list of known znodes if the vfs
247 * pointer is valid. We set the low bit of the vfs pointer when freeing
248 * the znode to invalidate it, and the memory patterns written by kmem
249 * (baddcafe and deadbeef) set at least one of the two low bits. A newly
250 * created znode sets the vfs pointer last of all to indicate that the
251 * znode is known and in a valid state to be moved by this function.
252 */
253 zfsvfs = ozp->z_zfsvfs;
254 if (!POINTER_IS_VALID(zfsvfs)) {
255 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid);
256 return (KMEM_CBRC_DONT_KNOW);
257 }
258
259 /*
260 * Close a small window in which it's possible that the filesystem could
261 * be unmounted and freed, and zfsvfs, though valid in the previous
262 * statement, could point to unrelated memory by the time we try to
263 * prevent the filesystem from being unmounted.
264 */
265 rw_enter(&zfsvfs_lock, RW_WRITER);
266 if (zfsvfs != ozp->z_zfsvfs) {
267 rw_exit(&zfsvfs_lock);
268 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1);
269 return (KMEM_CBRC_DONT_KNOW);
270 }
271
272 /*
273 * If the znode is still valid, then so is the file system. We know that
274 * no valid file system can be freed while we hold zfsvfs_lock, so we
275 * can safely ensure that the filesystem is not and will not be
276 * unmounted. The next statement is equivalent to ZFS_ENTER().
277 */
278 rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG);
279 if (zfsvfs->z_unmounted) {
280 ZFS_EXIT(zfsvfs);
281 rw_exit(&zfsvfs_lock);
282 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted);
283 return (KMEM_CBRC_DONT_KNOW);
284 }
285 rw_exit(&zfsvfs_lock);
286
287 mutex_enter(&zfsvfs->z_znodes_lock);
288 /*
289 * Recheck the vfs pointer in case the znode was removed just before
290 * acquiring the lock.
291 */
292 if (zfsvfs != ozp->z_zfsvfs) {
293 mutex_exit(&zfsvfs->z_znodes_lock);
294 ZFS_EXIT(zfsvfs);
295 ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2);
296 return (KMEM_CBRC_DONT_KNOW);
297 }
298
299 /*
300 * At this point we know that as long as we hold z_znodes_lock, the
301 * znode cannot be freed and fields within the znode can be safely
302 * accessed. Now, prevent a race with zfs_zget().
303 */
304 if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) {
305 mutex_exit(&zfsvfs->z_znodes_lock);
306 ZFS_EXIT(zfsvfs);
307 ZNODE_STAT_ADD(znode_move_stats.zms_obj_held);
308 return (KMEM_CBRC_LATER);
309 }
310
311 vp = ZTOV(ozp);
312 if (mutex_tryenter(&vp->v_lock) == 0) {
313 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
314 mutex_exit(&zfsvfs->z_znodes_lock);
315 ZFS_EXIT(zfsvfs);
316 ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked);
317 return (KMEM_CBRC_LATER);
318 }
319
320 /* Only move znodes that are referenced _only_ by the DNLC. */
321 if (vp->v_count != 1 || !vn_in_dnlc(vp)) {
322 mutex_exit(&vp->v_lock);
323 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
324 mutex_exit(&zfsvfs->z_znodes_lock);
325 ZFS_EXIT(zfsvfs);
326 ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc);
327 return (KMEM_CBRC_LATER);
328 }
329
330 /*
331 * The znode is known and in a valid state to move. We're holding the
332 * locks needed to execute the critical section.
333 */
334 zfs_znode_move_impl(ozp, nzp);
335 mutex_exit(&vp->v_lock);
336 ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id);
337
338 list_link_replace(&ozp->z_link_node, &nzp->z_link_node);
339 mutex_exit(&zfsvfs->z_znodes_lock);
340 ZFS_EXIT(zfsvfs);
341
342 return (KMEM_CBRC_YES);
343 }
344
345 void
346 zfs_znode_init(void)
347 {
348 /*
349 * Initialize zcache
350 */
351 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
352 ASSERT(znode_cache == NULL);
353 znode_cache = kmem_cache_create("zfs_znode_cache",
354 sizeof (znode_t), 0, zfs_znode_cache_constructor,
355 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
356 kmem_cache_set_move(znode_cache, zfs_znode_move);
357 }
358
359 void
360 zfs_znode_fini(void)
361 {
362 /*
363 * Cleanup vfs & vnode ops
364 */
365 zfs_remove_op_tables();
366
367 /*
368 * Cleanup zcache
369 */
370 if (znode_cache)
371 kmem_cache_destroy(znode_cache);
372 znode_cache = NULL;
373 rw_destroy(&zfsvfs_lock);
374 }
375
376 struct vnodeops *zfs_dvnodeops;
377 struct vnodeops *zfs_fvnodeops;
378 struct vnodeops *zfs_symvnodeops;
379 struct vnodeops *zfs_xdvnodeops;
380 struct vnodeops *zfs_evnodeops;
381 struct vnodeops *zfs_sharevnodeops;
382
383 void
384 zfs_remove_op_tables()
385 {
386 /*
387 * Remove vfs ops
388 */
389 ASSERT(zfsfstype);
390 (void) vfs_freevfsops_by_type(zfsfstype);
391 zfsfstype = 0;
392
393 /*
394 * Remove vnode ops
395 */
396 if (zfs_dvnodeops)
397 vn_freevnodeops(zfs_dvnodeops);
398 if (zfs_fvnodeops)
399 vn_freevnodeops(zfs_fvnodeops);
400 if (zfs_symvnodeops)
401 vn_freevnodeops(zfs_symvnodeops);
402 if (zfs_xdvnodeops)
403 vn_freevnodeops(zfs_xdvnodeops);
404 if (zfs_evnodeops)
405 vn_freevnodeops(zfs_evnodeops);
406 if (zfs_sharevnodeops)
407 vn_freevnodeops(zfs_sharevnodeops);
408
409 zfs_dvnodeops = NULL;
410 zfs_fvnodeops = NULL;
411 zfs_symvnodeops = NULL;
412 zfs_xdvnodeops = NULL;
413 zfs_evnodeops = NULL;
414 zfs_sharevnodeops = NULL;
415 }
416
417 extern const fs_operation_def_t zfs_dvnodeops_template[];
418 extern const fs_operation_def_t zfs_fvnodeops_template[];
419 extern const fs_operation_def_t zfs_xdvnodeops_template[];
420 extern const fs_operation_def_t zfs_symvnodeops_template[];
421 extern const fs_operation_def_t zfs_evnodeops_template[];
422 extern const fs_operation_def_t zfs_sharevnodeops_template[];
423
424 int
425 zfs_create_op_tables()
426 {
427 int error;
428
429 /*
430 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
431 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
432 * In this case we just return as the ops vectors are already set up.
433 */
434 if (zfs_dvnodeops)
435 return (0);
436
437 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
438 &zfs_dvnodeops);
439 if (error)
440 return (error);
441
442 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
443 &zfs_fvnodeops);
444 if (error)
445 return (error);
446
447 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
448 &zfs_symvnodeops);
449 if (error)
450 return (error);
451
452 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
453 &zfs_xdvnodeops);
454 if (error)
455 return (error);
456
457 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
458 &zfs_evnodeops);
459 if (error)
460 return (error);
461
462 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
463 &zfs_sharevnodeops);
464
465 return (error);
466 }
467
468 int
469 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
470 {
471 zfs_acl_ids_t acl_ids;
472 vattr_t vattr;
473 znode_t *sharezp;
474 vnode_t *vp;
475 znode_t *zp;
476 int error;
477
478 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
479 vattr.va_type = VDIR;
480 vattr.va_mode = S_IFDIR|0555;
481 vattr.va_uid = crgetuid(kcred);
482 vattr.va_gid = crgetgid(kcred);
483
484 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
485 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
486 sharezp->z_moved = 0;
487 sharezp->z_unlinked = 0;
488 sharezp->z_atime_dirty = 0;
489 sharezp->z_zfsvfs = zfsvfs;
490 sharezp->z_is_sa = zfsvfs->z_use_sa;
491
492 vp = ZTOV(sharezp);
493 vn_reinit(vp);
494 vp->v_type = VDIR;
495
496 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
497 kcred, NULL, &acl_ids));
498 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
499 ASSERT3P(zp, ==, sharezp);
500 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
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 ZTOV(sharezp)->v_count = 0;
508 sa_handle_destroy(sharezp->z_sa_hdl);
509 kmem_cache_free(znode_cache, sharezp);
510
511 return (error);
512 }
513
514 /*
515 * define a couple of values we need available
516 * for both 64 and 32 bit environments.
517 */
518 #ifndef NBITSMINOR64
519 #define NBITSMINOR64 32
520 #endif
521 #ifndef MAXMAJ64
522 #define MAXMAJ64 0xffffffffUL
523 #endif
524 #ifndef MAXMIN64
525 #define MAXMIN64 0xffffffffUL
526 #endif
527
528 /*
529 * Create special expldev for ZFS private use.
530 * Can't use standard expldev since it doesn't do
531 * what we want. The standard expldev() takes a
532 * dev32_t in LP64 and expands it to a long dev_t.
533 * We need an interface that takes a dev32_t in ILP32
534 * and expands it to a long dev_t.
535 */
536 static uint64_t
537 zfs_expldev(dev_t dev)
538 {
539 #ifndef _LP64
540 major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
541 return (((uint64_t)major << NBITSMINOR64) |
542 ((minor_t)dev & MAXMIN32));
543 #else
544 return (dev);
545 #endif
546 }
547
548 /*
549 * Special cmpldev for ZFS private use.
550 * Can't use standard cmpldev since it takes
551 * a long dev_t and compresses it to dev32_t in
552 * LP64. We need to do a compaction of a long dev_t
553 * to a dev32_t in ILP32.
554 */
555 dev_t
556 zfs_cmpldev(uint64_t dev)
557 {
558 #ifndef _LP64
559 minor_t minor = (minor_t)dev & MAXMIN64;
560 major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64;
561
562 if (major > MAXMAJ32 || minor > MAXMIN32)
563 return (NODEV32);
564
565 return (((dev32_t)major << NBITSMINOR32) | minor);
566 #else
567 return (dev);
568 #endif
569 }
570
571 static void
572 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
573 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
574 {
575 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
576 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
577
578 mutex_enter(&zp->z_lock);
579
580 ASSERT(zp->z_sa_hdl == NULL);
581 ASSERT(zp->z_acl_cached == NULL);
582 if (sa_hdl == NULL) {
583 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
584 SA_HDL_SHARED, &zp->z_sa_hdl));
585 } else {
586 zp->z_sa_hdl = sa_hdl;
587 sa_set_userp(sa_hdl, zp);
588 }
589
590 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
591
592 /*
593 * Slap on VROOT if we are the root znode
594 */
595 if (zp->z_id == zfsvfs->z_root)
596 ZTOV(zp)->v_flag |= VROOT;
597
598 mutex_exit(&zp->z_lock);
599 vn_exists(ZTOV(zp));
600 }
601
602 void
603 zfs_znode_dmu_fini(znode_t *zp)
604 {
605 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
606 zp->z_unlinked ||
607 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
608
609 sa_handle_destroy(zp->z_sa_hdl);
610 zp->z_sa_hdl = NULL;
611 }
612
613 /*
614 * Construct a new znode/vnode and intialize.
615 *
616 * This does not do a call to dmu_set_user() that is
617 * up to the caller to do, in case you don't want to
618 * return the znode
619 */
620 static znode_t *
621 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
622 dmu_object_type_t obj_type, sa_handle_t *hdl)
623 {
624 znode_t *zp;
625 vnode_t *vp;
626 uint64_t mode;
627 uint64_t parent;
628 sa_bulk_attr_t bulk[9];
629 int count = 0;
630
631 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
632
633 ASSERT(zp->z_dirlocks == NULL);
634 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
635 zp->z_moved = 0;
636
637 /*
638 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
639 * the zfs_znode_move() callback.
640 */
641 zp->z_sa_hdl = NULL;
642 zp->z_unlinked = 0;
643 zp->z_atime_dirty = 0;
644 zp->z_mapcnt = 0;
645 zp->z_id = db->db_object;
646 zp->z_blksz = blksz;
647 zp->z_seq = 0x7A4653;
648 zp->z_sync_cnt = 0;
649
650 vp = ZTOV(zp);
651 vn_reinit(vp);
652
653 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
654
655 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8);
656 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8);
657 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
658 &zp->z_size, 8);
659 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
660 &zp->z_links, 8);
661 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
662 &zp->z_pflags, 8);
663 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8);
664 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
665 &zp->z_atime, 16);
666 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
667 &zp->z_uid, 8);
668 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
669 &zp->z_gid, 8);
670
671 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
672 if (hdl == NULL)
673 sa_handle_destroy(zp->z_sa_hdl);
674 kmem_cache_free(znode_cache, zp);
675 return (NULL);
676 }
677
678 zp->z_mode = mode;
679 vp->v_vfsp = zfsvfs->z_parent->z_vfs;
680
681 vp->v_type = IFTOVT((mode_t)mode);
682
683 switch (vp->v_type) {
684 case VDIR:
685 if (zp->z_pflags & ZFS_XATTR) {
686 vn_setops(vp, zfs_xdvnodeops);
687 vp->v_flag |= V_XATTRDIR;
688 } else {
689 vn_setops(vp, zfs_dvnodeops);
690 }
691 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */
692 break;
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 /*FALLTHROUGH*/
703 case VFIFO:
704 case VSOCK:
705 case VDOOR:
706 vn_setops(vp, zfs_fvnodeops);
707 break;
708 case VREG:
709 vp->v_flag |= VMODSORT;
710 if (parent == zfsvfs->z_shares_dir) {
711 ASSERT(zp->z_uid == 0 && zp->z_gid == 0);
712 vn_setops(vp, zfs_sharevnodeops);
713 } else {
714 vn_setops(vp, zfs_fvnodeops);
715 }
716 break;
717 case VLNK:
718 vn_setops(vp, zfs_symvnodeops);
719 break;
720 default:
721 vn_setops(vp, zfs_evnodeops);
722 break;
723 }
724
725 mutex_enter(&zfsvfs->z_znodes_lock);
726 list_insert_tail(&zfsvfs->z_all_znodes, zp);
727 membar_producer();
728 /*
729 * Everything else must be valid before assigning z_zfsvfs makes the
730 * znode eligible for zfs_znode_move().
731 */
732 zp->z_zfsvfs = zfsvfs;
733 mutex_exit(&zfsvfs->z_znodes_lock);
734
735 VFS_HOLD(zfsvfs->z_vfs);
736 return (zp);
737 }
738
739 static uint64_t empty_xattr;
740 static uint64_t pad[4];
741 static zfs_acl_phys_t acl_phys;
742 /*
743 * Create a new DMU object to hold a zfs znode.
744 *
745 * IN: dzp - parent directory for new znode
746 * vap - file attributes for new znode
747 * tx - dmu transaction id for zap operations
748 * cr - credentials of caller
749 * flag - flags:
750 * IS_ROOT_NODE - new object will be root
751 * IS_XATTR - new object is an attribute
752 * bonuslen - length of bonus buffer
753 * setaclp - File/Dir initial ACL
754 * fuidp - Tracks fuid allocation.
755 *
756 * OUT: zpp - allocated znode
757 *
758 */
759 void
760 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
761 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
762 {
763 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
764 uint64_t mode, size, links, parent, pflags;
765 uint64_t dzp_pflags = 0;
766 uint64_t rdev = 0;
767 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
768 dmu_buf_t *db;
769 timestruc_t now;
770 uint64_t gen, obj;
771 int err;
772 int bonuslen;
773 sa_handle_t *sa_hdl;
774 dmu_object_type_t obj_type;
775 sa_bulk_attr_t sa_attrs[ZPL_END];
776 int cnt = 0;
777 zfs_acl_locator_cb_t locate = { 0 };
778
779 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
780
781 if (zfsvfs->z_replay) {
782 obj = vap->va_nodeid;
783 now = vap->va_ctime; /* see zfs_replay_create() */
784 gen = vap->va_nblocks; /* ditto */
785 } else {
786 obj = 0;
787 gethrestime(&now);
788 gen = dmu_tx_get_txg(tx);
789 }
790
791 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
792 bonuslen = (obj_type == DMU_OT_SA) ?
793 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
794
795 /*
796 * Create a new DMU object.
797 */
798 /*
799 * There's currently no mechanism for pre-reading the blocks that will
800 * be needed to allocate a new object, so we accept the small chance
801 * that there will be an i/o error and we will fail one of the
802 * assertions below.
803 */
804 if (vap->va_type == VDIR) {
805 if (zfsvfs->z_replay) {
806 err = zap_create_claim_norm(zfsvfs->z_os, obj,
807 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
808 obj_type, bonuslen, tx);
809 ASSERT3U(err, ==, 0);
810 } else {
811 obj = zap_create_norm(zfsvfs->z_os,
812 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
813 obj_type, bonuslen, tx);
814 }
815 } else {
816 if (zfsvfs->z_replay) {
817 err = dmu_object_claim(zfsvfs->z_os, obj,
818 DMU_OT_PLAIN_FILE_CONTENTS, 0,
819 obj_type, bonuslen, tx);
820 ASSERT3U(err, ==, 0);
821 } else {
822 obj = dmu_object_alloc(zfsvfs->z_os,
823 DMU_OT_PLAIN_FILE_CONTENTS, 0,
824 obj_type, bonuslen, tx);
825 }
826 }
827
828 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
829 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
830
831 /*
832 * If this is the root, fix up the half-initialized parent pointer
833 * to reference the just-allocated physical data area.
834 */
835 if (flag & IS_ROOT_NODE) {
836 dzp->z_id = obj;
837 } else {
838 dzp_pflags = dzp->z_pflags;
839 }
840
841 /*
842 * If parent is an xattr, so am I.
843 */
844 if (dzp_pflags & ZFS_XATTR) {
845 flag |= IS_XATTR;
846 }
847
848 if (zfsvfs->z_use_fuids)
849 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
850 else
851 pflags = 0;
852
853 if (vap->va_type == VDIR) {
854 size = 2; /* contents ("." and "..") */
855 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
856 } else {
857 size = links = 0;
858 }
859
860 if (vap->va_type == VBLK || vap->va_type == VCHR) {
861 rdev = zfs_expldev(vap->va_rdev);
862 }
863
864 parent = dzp->z_id;
865 mode = acl_ids->z_mode;
866 if (flag & IS_XATTR)
867 pflags |= ZFS_XATTR;
868
869 /*
870 * No execs denied will be deterimed when zfs_mode_compute() is called.
871 */
872 pflags |= acl_ids->z_aclp->z_hints &
873 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
874 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
875
876 ZFS_TIME_ENCODE(&now, crtime);
877 ZFS_TIME_ENCODE(&now, ctime);
878
879 if (vap->va_mask & AT_ATIME) {
880 ZFS_TIME_ENCODE(&vap->va_atime, atime);
881 } else {
882 ZFS_TIME_ENCODE(&now, atime);
883 }
884
885 if (vap->va_mask & AT_MTIME) {
886 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
887 } else {
888 ZFS_TIME_ENCODE(&now, mtime);
889 }
890
891 /* Now add in all of the "SA" attributes */
892 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
893 &sa_hdl));
894
895 /*
896 * Setup the array of attributes to be replaced/set on the new file
897 *
898 * order for DMU_OT_ZNODE is critical since it needs to be constructed
899 * in the old znode_phys_t format. Don't change this ordering
900 */
901
902 if (obj_type == DMU_OT_ZNODE) {
903 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
904 NULL, &atime, 16);
905 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
906 NULL, &mtime, 16);
907 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
908 NULL, &ctime, 16);
909 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
910 NULL, &crtime, 16);
911 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
912 NULL, &gen, 8);
913 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
914 NULL, &mode, 8);
915 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
916 NULL, &size, 8);
917 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
918 NULL, &parent, 8);
919 } else {
920 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
921 NULL, &mode, 8);
922 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
923 NULL, &size, 8);
924 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
925 NULL, &gen, 8);
926 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
927 &acl_ids->z_fuid, 8);
928 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
929 &acl_ids->z_fgid, 8);
930 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
931 NULL, &parent, 8);
932 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
933 NULL, &pflags, 8);
934 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
935 NULL, &atime, 16);
936 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
937 NULL, &mtime, 16);
938 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
939 NULL, &ctime, 16);
940 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
941 NULL, &crtime, 16);
942 }
943
944 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
945
946 if (obj_type == DMU_OT_ZNODE) {
947 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
948 &empty_xattr, 8);
949 }
950 if (obj_type == DMU_OT_ZNODE ||
951 (vap->va_type == VBLK || vap->va_type == VCHR)) {
952 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
953 NULL, &rdev, 8);
954
955 }
956 if (obj_type == DMU_OT_ZNODE) {
957 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
958 NULL, &pflags, 8);
959 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
960 &acl_ids->z_fuid, 8);
961 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
962 &acl_ids->z_fgid, 8);
963 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
964 sizeof (uint64_t) * 4);
965 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
966 &acl_phys, sizeof (zfs_acl_phys_t));
967 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
968 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
969 &acl_ids->z_aclp->z_acl_count, 8);
970 locate.cb_aclp = acl_ids->z_aclp;
971 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
972 zfs_acl_data_locator, &locate,
973 acl_ids->z_aclp->z_acl_bytes);
974 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
975 acl_ids->z_fuid, acl_ids->z_fgid);
976 }
977
978 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
979
980 if (!(flag & IS_ROOT_NODE)) {
981 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
982 ASSERT(*zpp != NULL);
983 } else {
984 /*
985 * If we are creating the root node, the "parent" we
986 * passed in is the znode for the root.
987 */
988 *zpp = dzp;
989
990 (*zpp)->z_sa_hdl = sa_hdl;
991 }
992
993 (*zpp)->z_pflags = pflags;
994 (*zpp)->z_mode = mode;
995
996 if (vap->va_mask & AT_XVATTR)
997 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
998
999 if (obj_type == DMU_OT_ZNODE ||
1000 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
1001 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
1002 ASSERT3P(err, ==, 0);
1003 }
1004 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1005 }
1006
1007 /*
1008 * zfs_xvattr_set only updates the in-core attributes
1009 * it is assumed the caller will be doing an sa_bulk_update
1010 * to push the changes out
1011 */
1012 void
1013 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
1014 {
1015 xoptattr_t *xoap;
1016
1017 xoap = xva_getxoptattr(xvap);
1018 ASSERT(xoap);
1019
1020 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
1021 uint64_t times[2];
1022 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
1023 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
1024 ×, sizeof (times), tx);
1025 XVA_SET_RTN(xvap, XAT_CREATETIME);
1026 }
1027 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
1028 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
1029 zp->z_pflags, tx);
1030 XVA_SET_RTN(xvap, XAT_READONLY);
1031 }
1032 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
1033 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
1034 zp->z_pflags, tx);
1035 XVA_SET_RTN(xvap, XAT_HIDDEN);
1036 }
1037 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
1038 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
1039 zp->z_pflags, tx);
1040 XVA_SET_RTN(xvap, XAT_SYSTEM);
1041 }
1042 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
1043 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
1044 zp->z_pflags, tx);
1045 XVA_SET_RTN(xvap, XAT_ARCHIVE);
1046 }
1047 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
1048 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
1049 zp->z_pflags, tx);
1050 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
1051 }
1052 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
1053 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
1054 zp->z_pflags, tx);
1055 XVA_SET_RTN(xvap, XAT_NOUNLINK);
1056 }
1057 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
1058 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
1059 zp->z_pflags, tx);
1060 XVA_SET_RTN(xvap, XAT_APPENDONLY);
1061 }
1062 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
1063 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
1064 zp->z_pflags, tx);
1065 XVA_SET_RTN(xvap, XAT_NODUMP);
1066 }
1067 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
1068 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
1069 zp->z_pflags, tx);
1070 XVA_SET_RTN(xvap, XAT_OPAQUE);
1071 }
1072 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
1073 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
1074 xoap->xoa_av_quarantined, zp->z_pflags, tx);
1075 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
1076 }
1077 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
1078 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
1079 zp->z_pflags, tx);
1080 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
1081 }
1082 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
1083 zfs_sa_set_scanstamp(zp, xvap, tx);
1084 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
1085 }
1086 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
1087 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
1088 zp->z_pflags, tx);
1089 XVA_SET_RTN(xvap, XAT_REPARSE);
1090 }
1091 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
1092 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
1093 zp->z_pflags, tx);
1094 XVA_SET_RTN(xvap, XAT_OFFLINE);
1095 }
1096 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
1097 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
1098 zp->z_pflags, tx);
1099 XVA_SET_RTN(xvap, XAT_SPARSE);
1100 }
1101 }
1102
1103 int
1104 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
1105 {
1106 dmu_object_info_t doi;
1107 dmu_buf_t *db;
1108 znode_t *zp;
1109 int err;
1110 sa_handle_t *hdl;
1111
1112 *zpp = NULL;
1113
1114 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1115
1116 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1117 if (err) {
1118 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1119 return (err);
1120 }
1121
1122 dmu_object_info_from_db(db, &doi);
1123 if (doi.doi_bonus_type != DMU_OT_SA &&
1124 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1125 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1126 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1127 sa_buf_rele(db, NULL);
1128 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1129 return (EINVAL);
1130 }
1131
1132 hdl = dmu_buf_get_user(db);
1133 if (hdl != NULL) {
1134 zp = sa_get_userdata(hdl);
1135
1136
1137 /*
1138 * Since "SA" does immediate eviction we
1139 * should never find a sa handle that doesn't
1140 * know about the znode.
1141 */
1142
1143 ASSERT3P(zp, !=, NULL);
1144
1145 mutex_enter(&zp->z_lock);
1146 ASSERT3U(zp->z_id, ==, obj_num);
1147 if (zp->z_unlinked) {
1148 err = ENOENT;
1149 } else {
1150 VN_HOLD(ZTOV(zp));
1151 *zpp = zp;
1152 err = 0;
1153 }
1154 sa_buf_rele(db, NULL);
1155 mutex_exit(&zp->z_lock);
1156 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1157 return (err);
1158 }
1159
1160 /*
1161 * Not found create new znode/vnode
1162 * but only if file exists.
1163 *
1164 * There is a small window where zfs_vget() could
1165 * find this object while a file create is still in
1166 * progress. This is checked for in zfs_znode_alloc()
1167 *
1168 * if zfs_znode_alloc() fails it will drop the hold on the
1169 * bonus buffer.
1170 */
1171 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1172 doi.doi_bonus_type, NULL);
1173 if (zp == NULL) {
1174 err = ENOENT;
1175 } else {
1176 *zpp = zp;
1177 }
1178 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1179 return (err);
1180 }
1181
1182 int
1183 zfs_rezget(znode_t *zp)
1184 {
1185 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1186 dmu_object_info_t doi;
1187 dmu_buf_t *db;
1188 uint64_t obj_num = zp->z_id;
1189 uint64_t mode;
1190 sa_bulk_attr_t bulk[8];
1191 int err;
1192 int count = 0;
1193 uint64_t gen;
1194
1195 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1196
1197 mutex_enter(&zp->z_acl_lock);
1198 if (zp->z_acl_cached) {
1199 zfs_acl_free(zp->z_acl_cached);
1200 zp->z_acl_cached = NULL;
1201 }
1202
1203 mutex_exit(&zp->z_acl_lock);
1204 ASSERT(zp->z_sa_hdl == NULL);
1205 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1206 if (err) {
1207 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1208 return (err);
1209 }
1210
1211 dmu_object_info_from_db(db, &doi);
1212 if (doi.doi_bonus_type != DMU_OT_SA &&
1213 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1214 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1215 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1216 sa_buf_rele(db, NULL);
1217 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1218 return (EINVAL);
1219 }
1220
1221 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1222
1223 /* reload cached values */
1224 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1225 &gen, sizeof (gen));
1226 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1227 &zp->z_size, sizeof (zp->z_size));
1228 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1229 &zp->z_links, sizeof (zp->z_links));
1230 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1231 &zp->z_pflags, sizeof (zp->z_pflags));
1232 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1233 &zp->z_atime, sizeof (zp->z_atime));
1234 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1235 &zp->z_uid, sizeof (zp->z_uid));
1236 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1237 &zp->z_gid, sizeof (zp->z_gid));
1238 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1239 &mode, sizeof (mode));
1240
1241 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1242 zfs_znode_dmu_fini(zp);
1243 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1244 return (EIO);
1245 }
1246
1247 zp->z_mode = mode;
1248
1249 if (gen != zp->z_gen) {
1250 zfs_znode_dmu_fini(zp);
1251 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1252 return (EIO);
1253 }
1254
1255 zp->z_unlinked = (zp->z_links == 0);
1256 zp->z_blksz = doi.doi_data_block_size;
1257
1258 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1259
1260 return (0);
1261 }
1262
1263 void
1264 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1265 {
1266 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1267 objset_t *os = zfsvfs->z_os;
1268 uint64_t obj = zp->z_id;
1269 uint64_t acl_obj = zfs_external_acl(zp);
1270
1271 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1272 if (acl_obj) {
1273 VERIFY(!zp->z_is_sa);
1274 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1275 }
1276 VERIFY(0 == dmu_object_free(os, obj, tx));
1277 zfs_znode_dmu_fini(zp);
1278 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1279 zfs_znode_free(zp);
1280 }
1281
1282 void
1283 zfs_zinactive(znode_t *zp)
1284 {
1285 vnode_t *vp = ZTOV(zp);
1286 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1287 uint64_t z_id = zp->z_id;
1288
1289 ASSERT(zp->z_sa_hdl);
1290
1291 /*
1292 * Don't allow a zfs_zget() while were trying to release this znode
1293 */
1294 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1295
1296 mutex_enter(&zp->z_lock);
1297 mutex_enter(&vp->v_lock);
1298 vp->v_count--;
1299 if (vp->v_count > 0 || vn_has_cached_data(vp)) {
1300 /*
1301 * If the hold count is greater than zero, somebody has
1302 * obtained a new reference on this znode while we were
1303 * processing it here, so we are done. If we still have
1304 * mapped pages then we are also done, since we don't
1305 * want to inactivate the znode until the pages get pushed.
1306 *
1307 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1308 * this seems like it would leave the znode hanging with
1309 * no chance to go inactive...
1310 */
1311 mutex_exit(&vp->v_lock);
1312 mutex_exit(&zp->z_lock);
1313 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1314 return;
1315 }
1316 mutex_exit(&vp->v_lock);
1317
1318 /*
1319 * If this was the last reference to a file with no links,
1320 * remove the file from the file system.
1321 */
1322 if (zp->z_unlinked) {
1323 mutex_exit(&zp->z_lock);
1324 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1325 zfs_rmnode(zp);
1326 return;
1327 }
1328
1329 mutex_exit(&zp->z_lock);
1330 zfs_znode_dmu_fini(zp);
1331 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1332 zfs_znode_free(zp);
1333 }
1334
1335 void
1336 zfs_znode_free(znode_t *zp)
1337 {
1338 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1339
1340 vn_invalid(ZTOV(zp));
1341
1342 ASSERT(ZTOV(zp)->v_count == 0);
1343
1344 mutex_enter(&zfsvfs->z_znodes_lock);
1345 POINTER_INVALIDATE(&zp->z_zfsvfs);
1346 list_remove(&zfsvfs->z_all_znodes, zp);
1347 mutex_exit(&zfsvfs->z_znodes_lock);
1348
1349 if (zp->z_acl_cached) {
1350 zfs_acl_free(zp->z_acl_cached);
1351 zp->z_acl_cached = NULL;
1352 }
1353
1354 kmem_cache_free(znode_cache, zp);
1355
1356 VFS_RELE(zfsvfs->z_vfs);
1357 }
1358
1359 void
1360 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1361 uint64_t ctime[2], boolean_t have_tx)
1362 {
1363 timestruc_t now;
1364
1365 gethrestime(&now);
1366
1367 if (have_tx) { /* will sa_bulk_update happen really soon? */
1368 zp->z_atime_dirty = 0;
1369 zp->z_seq++;
1370 } else {
1371 zp->z_atime_dirty = 1;
1372 }
1373
1374 if (flag & AT_ATIME) {
1375 ZFS_TIME_ENCODE(&now, zp->z_atime);
1376 }
1377
1378 if (flag & AT_MTIME) {
1379 ZFS_TIME_ENCODE(&now, mtime);
1380 if (zp->z_zfsvfs->z_use_fuids) {
1381 zp->z_pflags |= (ZFS_ARCHIVE |
1382 ZFS_AV_MODIFIED);
1383 }
1384 }
1385
1386 if (flag & AT_CTIME) {
1387 ZFS_TIME_ENCODE(&now, ctime);
1388 if (zp->z_zfsvfs->z_use_fuids)
1389 zp->z_pflags |= ZFS_ARCHIVE;
1390 }
1391 }
1392
1393 /*
1394 * Grow the block size for a file.
1395 *
1396 * IN: zp - znode of file to free data in.
1397 * size - requested block size
1398 * tx - open transaction.
1399 *
1400 * NOTE: this function assumes that the znode is write locked.
1401 */
1402 void
1403 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1404 {
1405 int error;
1406 u_longlong_t dummy;
1407
1408 if (size <= zp->z_blksz)
1409 return;
1410 /*
1411 * If the file size is already greater than the current blocksize,
1412 * we will not grow. If there is more than one block in a file,
1413 * the blocksize cannot change.
1414 */
1415 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1416 return;
1417
1418 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1419 size, 0, tx);
1420
1421 if (error == ENOTSUP)
1422 return;
1423 ASSERT3U(error, ==, 0);
1424
1425 /* What blocksize did we actually get? */
1426 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1427 }
1428
1429 /*
1430 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1431 * be calling back into the fs for a putpage(). E.g.: when truncating
1432 * a file, the pages being "thrown away* don't need to be written out.
1433 */
1434 /* ARGSUSED */
1435 static int
1436 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1437 int flags, cred_t *cr)
1438 {
1439 ASSERT(0);
1440 return (0);
1441 }
1442
1443 /*
1444 * Increase the file length
1445 *
1446 * IN: zp - znode of file to free data in.
1447 * end - new end-of-file
1448 *
1449 * RETURN: 0 if success
1450 * error code if failure
1451 */
1452 static int
1453 zfs_extend(znode_t *zp, uint64_t end)
1454 {
1455 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1456 dmu_tx_t *tx;
1457 rl_t *rl;
1458 uint64_t newblksz;
1459 int error;
1460
1461 /*
1462 * We will change zp_size, lock the whole file.
1463 */
1464 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1465
1466 /*
1467 * Nothing to do if file already at desired length.
1468 */
1469 if (end <= zp->z_size) {
1470 zfs_range_unlock(rl);
1471 return (0);
1472 }
1473 top:
1474 tx = dmu_tx_create(zfsvfs->z_os);
1475 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1476 zfs_sa_upgrade_txholds(tx, zp);
1477 if (end > zp->z_blksz &&
1478 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1479 /*
1480 * We are growing the file past the current block size.
1481 */
1482 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1483 ASSERT(!ISP2(zp->z_blksz));
1484 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1485 } else {
1486 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1487 }
1488 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1489 } else {
1490 newblksz = 0;
1491 }
1492
1493 error = dmu_tx_assign(tx, TXG_NOWAIT);
1494 if (error) {
1495 if (error == ERESTART) {
1496 dmu_tx_wait(tx);
1497 dmu_tx_abort(tx);
1498 goto top;
1499 }
1500 dmu_tx_abort(tx);
1501 zfs_range_unlock(rl);
1502 return (error);
1503 }
1504
1505 if (newblksz)
1506 zfs_grow_blocksize(zp, newblksz, tx);
1507
1508 zp->z_size = end;
1509
1510 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1511 &zp->z_size, sizeof (zp->z_size), tx));
1512
1513 zfs_range_unlock(rl);
1514
1515 dmu_tx_commit(tx);
1516
1517 return (0);
1518 }
1519
1520 /*
1521 * Free space in a file.
1522 *
1523 * IN: zp - znode of file to free data in.
1524 * off - start of section to free.
1525 * len - length of section to free.
1526 *
1527 * RETURN: 0 if success
1528 * error code if failure
1529 */
1530 static int
1531 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1532 {
1533 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1534 rl_t *rl;
1535 int error;
1536
1537 /*
1538 * Lock the range being freed.
1539 */
1540 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1541
1542 /*
1543 * Nothing to do if file already at desired length.
1544 */
1545 if (off >= zp->z_size) {
1546 zfs_range_unlock(rl);
1547 return (0);
1548 }
1549
1550 if (off + len > zp->z_size)
1551 len = zp->z_size - off;
1552
1553 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1554
1555 zfs_range_unlock(rl);
1556
1557 return (error);
1558 }
1559
1560 /*
1561 * Truncate a file
1562 *
1563 * IN: zp - znode of file to free data in.
1564 * end - new end-of-file.
1565 *
1566 * RETURN: 0 if success
1567 * error code if failure
1568 */
1569 static int
1570 zfs_trunc(znode_t *zp, uint64_t end)
1571 {
1572 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1573 vnode_t *vp = ZTOV(zp);
1574 dmu_tx_t *tx;
1575 rl_t *rl;
1576 int error;
1577 sa_bulk_attr_t bulk[2];
1578 int count = 0;
1579
1580 /*
1581 * We will change zp_size, lock the whole file.
1582 */
1583 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1584
1585 /*
1586 * Nothing to do if file already at desired length.
1587 */
1588 if (end >= zp->z_size) {
1589 zfs_range_unlock(rl);
1590 return (0);
1591 }
1592
1593 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1594 if (error) {
1595 zfs_range_unlock(rl);
1596 return (error);
1597 }
1598 top:
1599 tx = dmu_tx_create(zfsvfs->z_os);
1600 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1601 zfs_sa_upgrade_txholds(tx, zp);
1602 error = dmu_tx_assign(tx, TXG_NOWAIT);
1603 if (error) {
1604 if (error == ERESTART) {
1605 dmu_tx_wait(tx);
1606 dmu_tx_abort(tx);
1607 goto top;
1608 }
1609 dmu_tx_abort(tx);
1610 zfs_range_unlock(rl);
1611 return (error);
1612 }
1613
1614 zp->z_size = end;
1615 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1616 NULL, &zp->z_size, sizeof (zp->z_size));
1617
1618 if (end == 0) {
1619 zp->z_pflags &= ~ZFS_SPARSE;
1620 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1621 NULL, &zp->z_pflags, 8);
1622 }
1623 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1624
1625 dmu_tx_commit(tx);
1626
1627 /*
1628 * Clear any mapped pages in the truncated region. This has to
1629 * happen outside of the transaction to avoid the possibility of
1630 * a deadlock with someone trying to push a page that we are
1631 * about to invalidate.
1632 */
1633 if (vn_has_cached_data(vp)) {
1634 page_t *pp;
1635 uint64_t start = end & PAGEMASK;
1636 int poff = end & PAGEOFFSET;
1637
1638 if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
1639 /*
1640 * We need to zero a partial page.
1641 */
1642 pagezero(pp, poff, PAGESIZE - poff);
1643 start += PAGESIZE;
1644 page_unlock(pp);
1645 }
1646 error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
1647 B_INVAL | B_TRUNC, NULL);
1648 ASSERT(error == 0);
1649 }
1650
1651 zfs_range_unlock(rl);
1652
1653 return (0);
1654 }
1655
1656 /*
1657 * Free space in a file
1658 *
1659 * IN: zp - znode of file to free data in.
1660 * off - start of range
1661 * len - end of range (0 => EOF)
1662 * flag - current file open mode flags.
1663 * log - TRUE if this action should be logged
1664 *
1665 * RETURN: 0 if success
1666 * error code if failure
1667 */
1668 int
1669 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1670 {
1671 vnode_t *vp = ZTOV(zp);
1672 dmu_tx_t *tx;
1673 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1674 zilog_t *zilog = zfsvfs->z_log;
1675 uint64_t mode;
1676 uint64_t mtime[2], ctime[2];
1677 sa_bulk_attr_t bulk[3];
1678 int count = 0;
1679 int error;
1680
1681 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1682 sizeof (mode))) != 0)
1683 return (error);
1684
1685 if (off > zp->z_size) {
1686 error = zfs_extend(zp, off+len);
1687 if (error == 0 && log)
1688 goto log;
1689 else
1690 return (error);
1691 }
1692
1693 /*
1694 * Check for any locks in the region to be freed.
1695 */
1696
1697 if (MANDLOCK(vp, (mode_t)mode)) {
1698 uint64_t length = (len ? len : zp->z_size - off);
1699 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1700 return (error);
1701 }
1702
1703 if (len == 0) {
1704 error = zfs_trunc(zp, off);
1705 } else {
1706 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1707 off + len > zp->z_size)
1708 error = zfs_extend(zp, off+len);
1709 }
1710 if (error || !log)
1711 return (error);
1712 log:
1713 tx = dmu_tx_create(zfsvfs->z_os);
1714 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1715 zfs_sa_upgrade_txholds(tx, zp);
1716 error = dmu_tx_assign(tx, TXG_NOWAIT);
1717 if (error) {
1718 if (error == ERESTART) {
1719 dmu_tx_wait(tx);
1720 dmu_tx_abort(tx);
1721 goto log;
1722 }
1723 dmu_tx_abort(tx);
1724 return (error);
1725 }
1726
1727 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1728 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1729 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1730 NULL, &zp->z_pflags, 8);
1731 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1732 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1733 ASSERT(error == 0);
1734
1735 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1736
1737 dmu_tx_commit(tx);
1738 return (0);
1739 }
1740
1741 void
1742 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1743 {
1744 zfsvfs_t zfsvfs;
1745 uint64_t moid, obj, sa_obj, version;
1746 uint64_t sense = ZFS_CASE_SENSITIVE;
1747 uint64_t norm = 0;
1748 nvpair_t *elem;
1749 int error;
1750 int i;
1751 znode_t *rootzp = NULL;
1752 vnode_t *vp;
1753 vattr_t vattr;
1754 znode_t *zp;
1755 zfs_acl_ids_t acl_ids;
1756
1757 /*
1758 * First attempt to create master node.
1759 */
1760 /*
1761 * In an empty objset, there are no blocks to read and thus
1762 * there can be no i/o errors (which we assert below).
1763 */
1764 moid = MASTER_NODE_OBJ;
1765 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1766 DMU_OT_NONE, 0, tx);
1767 ASSERT(error == 0);
1768
1769 /*
1770 * Set starting attributes.
1771 */
1772 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1773 elem = NULL;
1774 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1775 /* For the moment we expect all zpl props to be uint64_ts */
1776 uint64_t val;
1777 char *name;
1778
1779 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1780 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1781 name = nvpair_name(elem);
1782 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1783 if (val < version)
1784 version = val;
1785 } else {
1786 error = zap_update(os, moid, name, 8, 1, &val, tx);
1787 }
1788 ASSERT(error == 0);
1789 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1790 norm = val;
1791 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1792 sense = val;
1793 }
1794 ASSERT(version != 0);
1795 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1796
1797 /*
1798 * Create zap object used for SA attribute registration
1799 */
1800
1801 if (version >= ZPL_VERSION_SA) {
1802 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1803 DMU_OT_NONE, 0, tx);
1804 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1805 ASSERT(error == 0);
1806 } else {
1807 sa_obj = 0;
1808 }
1809 /*
1810 * Create a delete queue.
1811 */
1812 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1813
1814 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1815 ASSERT(error == 0);
1816
1817 /*
1818 * Create root znode. Create minimal znode/vnode/zfsvfs
1819 * to allow zfs_mknode to work.
1820 */
1821 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1822 vattr.va_type = VDIR;
1823 vattr.va_mode = S_IFDIR|0755;
1824 vattr.va_uid = crgetuid(cr);
1825 vattr.va_gid = crgetgid(cr);
1826
1827 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1828 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1829 rootzp->z_moved = 0;
1830 rootzp->z_unlinked = 0;
1831 rootzp->z_atime_dirty = 0;
1832 rootzp->z_is_sa = USE_SA(version, os);
1833
1834 vp = ZTOV(rootzp);
1835 vn_reinit(vp);
1836 vp->v_type = VDIR;
1837
1838 bzero(&zfsvfs, sizeof (zfsvfs_t));
1839
1840 zfsvfs.z_os = os;
1841 zfsvfs.z_parent = &zfsvfs;
1842 zfsvfs.z_version = version;
1843 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1844 zfsvfs.z_use_sa = USE_SA(version, os);
1845 zfsvfs.z_norm = norm;
1846
1847 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1848 &zfsvfs.z_attr_table);
1849
1850 ASSERT(error == 0);
1851
1852 /*
1853 * Fold case on file systems that are always or sometimes case
1854 * insensitive.
1855 */
1856 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1857 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1858
1859 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1860 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1861 offsetof(znode_t, z_link_node));
1862
1863 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1864 mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1865
1866 rootzp->z_zfsvfs = &zfsvfs;
1867 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1868 cr, NULL, &acl_ids));
1869 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1870 ASSERT3P(zp, ==, rootzp);
1871 ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
1872 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1873 ASSERT(error == 0);
1874 zfs_acl_ids_free(&acl_ids);
1875 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1876
1877 ZTOV(rootzp)->v_count = 0;
1878 sa_handle_destroy(rootzp->z_sa_hdl);
1879 kmem_cache_free(znode_cache, rootzp);
1880
1881 /*
1882 * Create shares directory
1883 */
1884
1885 error = zfs_create_share_dir(&zfsvfs, tx);
1886
1887 ASSERT(error == 0);
1888
1889 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1890 mutex_destroy(&zfsvfs.z_hold_mtx[i]);
1891 }
1892
1893 #endif /* _KERNEL */
1894
1895 static int
1896 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1897 {
1898 uint64_t sa_obj = 0;
1899 int error;
1900
1901 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1902 if (error != 0 && error != ENOENT)
1903 return (error);
1904
1905 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1906 return (error);
1907 }
1908
1909 static int
1910 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1911 dmu_buf_t **db, void *tag)
1912 {
1913 dmu_object_info_t doi;
1914 int error;
1915
1916 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0)
1917 return (error);
1918
1919 dmu_object_info_from_db(*db, &doi);
1920 if ((doi.doi_bonus_type != DMU_OT_SA &&
1921 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1922 doi.doi_bonus_type == DMU_OT_ZNODE &&
1923 doi.doi_bonus_size < sizeof (znode_phys_t)) {
1924 sa_buf_rele(*db, tag);
1925 return (ENOTSUP);
1926 }
1927
1928 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1929 if (error != 0) {
1930 sa_buf_rele(*db, tag);
1931 return (error);
1932 }
1933
1934 return (0);
1935 }
1936
1937 void
1938 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag)
1939 {
1940 sa_handle_destroy(hdl);
1941 sa_buf_rele(db, tag);
1942 }
1943
1944 /*
1945 * Given an object number, return its parent object number and whether
1946 * or not the object is an extended attribute directory.
1947 */
1948 static int
1949 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1950 int *is_xattrdir)
1951 {
1952 uint64_t parent;
1953 uint64_t pflags;
1954 uint64_t mode;
1955 sa_bulk_attr_t bulk[3];
1956 int count = 0;
1957 int error;
1958
1959 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1960 &parent, sizeof (parent));
1961 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1962 &pflags, sizeof (pflags));
1963 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1964 &mode, sizeof (mode));
1965
1966 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1967 return (error);
1968
1969 *pobjp = parent;
1970 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1971
1972 return (0);
1973 }
1974
1975 /*
1976 * Given an object number, return some zpl level statistics
1977 */
1978 static int
1979 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1980 zfs_stat_t *sb)
1981 {
1982 sa_bulk_attr_t bulk[4];
1983 int count = 0;
1984
1985 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1986 &sb->zs_mode, sizeof (sb->zs_mode));
1987 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1988 &sb->zs_gen, sizeof (sb->zs_gen));
1989 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1990 &sb->zs_links, sizeof (sb->zs_links));
1991 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1992 &sb->zs_ctime, sizeof (sb->zs_ctime));
1993
1994 return (sa_bulk_lookup(hdl, bulk, count));
1995 }
1996
1997 static int
1998 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1999 sa_attr_type_t *sa_table, char *buf, int len)
2000 {
2001 sa_handle_t *sa_hdl;
2002 sa_handle_t *prevhdl = NULL;
2003 dmu_buf_t *prevdb = NULL;
2004 dmu_buf_t *sa_db = NULL;
2005 char *path = buf + len - 1;
2006 int error;
2007
2008 *path = '\0';
2009 sa_hdl = hdl;
2010
2011 for (;;) {
2012 uint64_t pobj;
2013 char component[MAXNAMELEN + 2];
2014 size_t complen;
2015 int is_xattrdir;
2016
2017 if (prevdb)
2018 zfs_release_sa_handle(prevhdl, prevdb, FTAG);
2019
2020 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
2021 &is_xattrdir)) != 0)
2022 break;
2023
2024 if (pobj == obj) {
2025 if (path[0] != '/')
2026 *--path = '/';
2027 break;
2028 }
2029
2030 component[0] = '/';
2031 if (is_xattrdir) {
2032 (void) sprintf(component + 1, "<xattrdir>");
2033 } else {
2034 error = zap_value_search(osp, pobj, obj,
2035 ZFS_DIRENT_OBJ(-1ULL), component + 1);
2036 if (error != 0)
2037 break;
2038 }
2039
2040 complen = strlen(component);
2041 path -= complen;
2042 ASSERT(path >= buf);
2043 bcopy(component, path, complen);
2044 obj = pobj;
2045
2046 if (sa_hdl != hdl) {
2047 prevhdl = sa_hdl;
2048 prevdb = sa_db;
2049 }
2050 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG);
2051 if (error != 0) {
2052 sa_hdl = prevhdl;
2053 sa_db = prevdb;
2054 break;
2055 }
2056 }
2057
2058 if (sa_hdl != NULL && sa_hdl != hdl) {
2059 ASSERT(sa_db != NULL);
2060 zfs_release_sa_handle(sa_hdl, sa_db, FTAG);
2061 }
2062
2063 if (error == 0)
2064 (void) memmove(buf, path, buf + len - path);
2065
2066 return (error);
2067 }
2068
2069 int
2070 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
2071 {
2072 sa_attr_type_t *sa_table;
2073 sa_handle_t *hdl;
2074 dmu_buf_t *db;
2075 int error;
2076
2077 error = zfs_sa_setup(osp, &sa_table);
2078 if (error != 0)
2079 return (error);
2080
2081 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2082 if (error != 0)
2083 return (error);
2084
2085 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2086
2087 zfs_release_sa_handle(hdl, db, FTAG);
2088 return (error);
2089 }
2090
2091 int
2092 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
2093 char *buf, int len)
2094 {
2095 char *path = buf + len - 1;
2096 sa_attr_type_t *sa_table;
2097 sa_handle_t *hdl;
2098 dmu_buf_t *db;
2099 int error;
2100
2101 *path = '\0';
2102
2103 error = zfs_sa_setup(osp, &sa_table);
2104 if (error != 0)
2105 return (error);
2106
2107 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG);
2108 if (error != 0)
2109 return (error);
2110
2111 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
2112 if (error != 0) {
2113 zfs_release_sa_handle(hdl, db, FTAG);
2114 return (error);
2115 }
2116
2117 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2118
2119 zfs_release_sa_handle(hdl, db, FTAG);
2120 return (error);
2121 }