]> git.proxmox.com Git - mirror_zfs.git/blob - module/zfs/zfs_znode.c
projects / mirror_zfs.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
VFS: Add zfs_inode_update() helper
[mirror_zfs.git] / module / zfs / zfs_znode.c
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 int
101 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
102 {
103 znode_t *zp = buf;
104
105 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
106
107 zp->z_vnode = vn_alloc(kmflags);
108 if (zp->z_vnode == NULL) {
109 return (-1);
110 }
111 ZTOV(zp)->v_data = zp;
112
113 list_link_init(&zp->z_link_node);
114
115 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
116 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
117 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
118 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
119
120 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
121 avl_create(&zp->z_range_avl, zfs_range_compare,
122 sizeof (rl_t), offsetof(rl_t, r_node));
123
124 zp->z_dirlocks = NULL;
125 zp->z_acl_cached = NULL;
126 zp->z_moved = 0;
127 return (0);
128 }
129
130 /*ARGSUSED*/
131 static void
132 zfs_znode_cache_destructor(void *buf, void *arg)
133 {
134 znode_t *zp = buf;
135
136 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
137 ASSERT(ZTOV(zp)->v_data == zp);
138 vn_free(ZTOV(zp));
139 ASSERT(!list_link_active(&zp->z_link_node));
140 mutex_destroy(&zp->z_lock);
141 rw_destroy(&zp->z_parent_lock);
142 rw_destroy(&zp->z_name_lock);
143 mutex_destroy(&zp->z_acl_lock);
144 avl_destroy(&zp->z_range_avl);
145 mutex_destroy(&zp->z_range_lock);
146
147 ASSERT(zp->z_dirlocks == NULL);
148 ASSERT(zp->z_acl_cached == NULL);
149 }
150
151 void
152 zfs_znode_init(void)
153 {
154 /*
155 * Initialize zcache
156 */
157 rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL);
158 ASSERT(znode_cache == NULL);
159 znode_cache = kmem_cache_create("zfs_znode_cache",
160 sizeof (znode_t), 0, zfs_znode_cache_constructor,
161 zfs_znode_cache_destructor, NULL, NULL, NULL, 0);
162 }
163
164 void
165 zfs_znode_fini(void)
166 {
167 /*
168 * Cleanup vfs & vnode ops
169 */
170 #ifdef HAVE_ZPL
171 zfs_remove_op_tables();
172 #endif /* HAVE_ZPL */
173
174 /*
175 * Cleanup zcache
176 */
177 if (znode_cache)
178 kmem_cache_destroy(znode_cache);
179 znode_cache = NULL;
180 rw_destroy(&zfsvfs_lock);
181 }
182
183 #ifdef HAVE_ZPL
184 struct vnodeops *zfs_dvnodeops;
185 struct vnodeops *zfs_fvnodeops;
186 struct vnodeops *zfs_symvnodeops;
187 struct vnodeops *zfs_xdvnodeops;
188 struct vnodeops *zfs_evnodeops;
189 struct vnodeops *zfs_sharevnodeops;
190
191 void
192 zfs_remove_op_tables()
193 {
194 /*
195 * Remove vfs ops
196 */
197 ASSERT(zfsfstype);
198 (void) vfs_freevfsops_by_type(zfsfstype);
199 zfsfstype = 0;
200
201 /*
202 * Remove vnode ops
203 */
204 if (zfs_dvnodeops)
205 vn_freevnodeops(zfs_dvnodeops);
206 if (zfs_fvnodeops)
207 vn_freevnodeops(zfs_fvnodeops);
208 if (zfs_symvnodeops)
209 vn_freevnodeops(zfs_symvnodeops);
210 if (zfs_xdvnodeops)
211 vn_freevnodeops(zfs_xdvnodeops);
212 if (zfs_evnodeops)
213 vn_freevnodeops(zfs_evnodeops);
214 if (zfs_sharevnodeops)
215 vn_freevnodeops(zfs_sharevnodeops);
216
217 zfs_dvnodeops = NULL;
218 zfs_fvnodeops = NULL;
219 zfs_symvnodeops = NULL;
220 zfs_xdvnodeops = NULL;
221 zfs_evnodeops = NULL;
222 zfs_sharevnodeops = NULL;
223 }
224
225 extern const fs_operation_def_t zfs_dvnodeops_template[];
226 extern const fs_operation_def_t zfs_fvnodeops_template[];
227 extern const fs_operation_def_t zfs_xdvnodeops_template[];
228 extern const fs_operation_def_t zfs_symvnodeops_template[];
229 extern const fs_operation_def_t zfs_evnodeops_template[];
230 extern const fs_operation_def_t zfs_sharevnodeops_template[];
231
232 int
233 zfs_create_op_tables()
234 {
235 int error;
236
237 /*
238 * zfs_dvnodeops can be set if mod_remove() calls mod_installfs()
239 * due to a failure to remove the the 2nd modlinkage (zfs_modldrv).
240 * In this case we just return as the ops vectors are already set up.
241 */
242 if (zfs_dvnodeops)
243 return (0);
244
245 error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template,
246 &zfs_dvnodeops);
247 if (error)
248 return (error);
249
250 error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template,
251 &zfs_fvnodeops);
252 if (error)
253 return (error);
254
255 error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template,
256 &zfs_symvnodeops);
257 if (error)
258 return (error);
259
260 error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template,
261 &zfs_xdvnodeops);
262 if (error)
263 return (error);
264
265 error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template,
266 &zfs_evnodeops);
267 if (error)
268 return (error);
269
270 error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template,
271 &zfs_sharevnodeops);
272
273 return (error);
274 }
275
276 int
277 zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
278 {
279 #ifdef HAVE_SHARE
280 zfs_acl_ids_t acl_ids;
281 vattr_t vattr;
282 znode_t *sharezp;
283 vnode_t *vp;
284 znode_t *zp;
285 int error;
286
287 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
288 vattr.va_type = VDIR;
289 vattr.va_mode = S_IFDIR|0555;
290 vattr.va_uid = crgetuid(kcred);
291 vattr.va_gid = crgetgid(kcred);
292
293 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
294 ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs));
295 sharezp->z_moved = 0;
296 sharezp->z_unlinked = 0;
297 sharezp->z_atime_dirty = 0;
298 sharezp->z_zfsvfs = zfsvfs;
299 sharezp->z_is_sa = zfsvfs->z_use_sa;
300
301 vp = ZTOV(sharezp);
302 vn_reinit(vp);
303 vp->v_type = VDIR;
304
305 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
306 kcred, NULL, &acl_ids));
307 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
308 ASSERT3P(zp, ==, sharezp);
309 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
310 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
311 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
312 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
313 zfsvfs->z_shares_dir = sharezp->z_id;
314
315 zfs_acl_ids_free(&acl_ids);
316 ZTOV(sharezp)->v_count = 0;
317 sa_handle_destroy(sharezp->z_sa_hdl);
318 kmem_cache_free(znode_cache, sharezp);
319
320 return (error);
321 #else
322 return (0);
323 #endif /* HAVE_SHARE */
324 }
325
326 /*
327 * define a couple of values we need available
328 * for both 64 and 32 bit environments.
329 */
330 #ifndef NBITSMINOR64
331 #define NBITSMINOR64 32
332 #endif
333 #ifndef MAXMAJ64
334 #define MAXMAJ64 0xffffffffUL
335 #endif
336 #ifndef MAXMIN64
337 #define MAXMIN64 0xffffffffUL
338 #endif
339
340 #endif /* HAVE_ZPL */
341
342 /*
343 * Create special expldev for ZFS private use.
344 * Can't use standard expldev since it doesn't do
345 * what we want. The standard expldev() takes a
346 * dev32_t in LP64 and expands it to a long dev_t.
347 * We need an interface that takes a dev32_t in ILP32
348 * and expands it to a long dev_t.
349 */
350 static uint64_t
351 zfs_expldev(dev_t dev)
352 {
353 #ifndef _LP64
354 major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
355 return (((uint64_t)major << NBITSMINOR64) |
356 ((minor_t)dev & MAXMIN32));
357 #else
358 return (dev);
359 #endif
360 }
361
362 /*
363 * Special cmpldev for ZFS private use.
364 * Can't use standard cmpldev since it takes
365 * a long dev_t and compresses it to dev32_t in
366 * LP64. We need to do a compaction of a long dev_t
367 * to a dev32_t in ILP32.
368 */
369 dev_t
370 zfs_cmpldev(uint64_t dev)
371 {
372 #ifndef _LP64
373 minor_t minor = (minor_t)dev & MAXMIN64;
374 major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64;
375
376 if (major > MAXMAJ32 || minor > MAXMIN32)
377 return (NODEV32);
378
379 return (((dev32_t)major << NBITSMINOR32) | minor);
380 #else
381 return (dev);
382 #endif
383 }
384
385 static void
386 zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp,
387 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
388 {
389 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs));
390 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)));
391
392 mutex_enter(&zp->z_lock);
393
394 ASSERT(zp->z_sa_hdl == NULL);
395 ASSERT(zp->z_acl_cached == NULL);
396 if (sa_hdl == NULL) {
397 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp,
398 SA_HDL_SHARED, &zp->z_sa_hdl));
399 } else {
400 zp->z_sa_hdl = sa_hdl;
401 sa_set_userp(sa_hdl, zp);
402 }
403
404 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
405
406 /*
407 * Slap on VROOT if we are the root znode
408 */
409 if (zp->z_id == zfsvfs->z_root)
410 ZTOV(zp)->v_flag |= VROOT;
411
412 mutex_exit(&zp->z_lock);
413 vn_exists(ZTOV(zp));
414 }
415
416 void
417 zfs_znode_dmu_fini(znode_t *zp)
418 {
419 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) ||
420 zp->z_unlinked ||
421 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock));
422
423 sa_handle_destroy(zp->z_sa_hdl);
424 zp->z_sa_hdl = NULL;
425 }
426
427 /*
428 * Construct a new znode+inode and initialize.
429 *
430 * This does not do a call to dmu_set_user() that is
431 * up to the caller to do, in case you don't want to
432 * return the znode
433 */
434 static znode_t *
435 zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz,
436 dmu_object_type_t obj_type, sa_handle_t *hdl)
437 {
438 znode_t *zp;
439 struct inode *inode;
440 uint64_t parent;
441 sa_bulk_attr_t bulk[9];
442 int count = 0;
443
444 ASSERT(zfsvfs != NULL);
445 ASSERT(zfsvfs->z_vfs != NULL);
446 ASSERT(zfsvfs->z_vfs->mnt_sb != NULL);
447
448 inode = iget_locked(zfsvfs->z_vfs->mnt_sb, db->db_object);
449 zp = ITOZ(inode);
450
451 ASSERT(inode->i_state & I_NEW);
452 ASSERT(zp->z_dirlocks == NULL);
453 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs));
454 zp->z_moved = 0;
455
456 /*
457 * Defer setting z_zfsvfs until the znode is ready to be a candidate for
458 * the zfs_znode_move() callback.
459 */
460 zp->z_sa_hdl = NULL;
461 zp->z_unlinked = 0;
462 zp->z_atime_dirty = 0;
463 zp->z_mapcnt = 0;
464 zp->z_id = db->db_object;
465 zp->z_blksz = blksz;
466 zp->z_seq = 0x7A4653;
467 zp->z_sync_cnt = 0;
468
469 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl);
470
471 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
472 &zp->z_mode, 8);
473 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
474 &zp->z_gen, 8);
475 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
476 &zp->z_size, 8);
477 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
478 &zp->z_links, 8);
479 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
480 &zp->z_pflags, 8);
481 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL,
482 &parent, 8);
483 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
484 &zp->z_atime, 16);
485 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
486 &zp->z_uid, 8);
487 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
488 &zp->z_gid, 8);
489
490 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
491 if (hdl == NULL)
492 sa_handle_destroy(zp->z_sa_hdl);
493 iput(inode);
494 return (NULL);
495 }
496
497 inode->i_mode = (umode_t)zp->z_mode;
498 if ((S_ISCHR(inode->i_mode)) || (S_ISBLK(inode->i_mode))) {
499 uint64_t rdev;
500 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs),
501 &rdev, sizeof (rdev)) == 0);
502 inode->i_rdev = zfs_cmpldev(rdev);
503 }
504
505 /* zp->z_set_ops_inode() must be set in sb->alloc_inode() */
506 ASSERT(zp->z_set_ops_inode != NULL);
507 zp->z_set_ops_inode(inode);
508 unlock_new_inode(inode);
509
510 mutex_enter(&zfsvfs->z_znodes_lock);
511 list_insert_tail(&zfsvfs->z_all_znodes, zp);
512 membar_producer();
513 /*
514 * Everything else must be valid before assigning z_zfsvfs makes the
515 * znode eligible for zfs_znode_move().
516 */
517 zp->z_zfsvfs = zfsvfs;
518 mutex_exit(&zfsvfs->z_znodes_lock);
519
520 VFS_HOLD(zfsvfs->z_vfs);
521 return (zp);
522 }
523
524 /*
525 * Update the embedded inode given the znode. We should work toward
526 * eliminating this function as soon as possible by removing values
527 * which are duplicated between the znode and inode. If the generic
528 * inode has the correct field it should be used, and the ZFS code
529 * updated to access the inode. This can be done incrementally.
530 */
531 void
532 zfs_inode_update(znode_t *zp)
533 {
534 zfsvfs_t *zfsvfs;
535 struct inode *inode;
536 uint32_t blksize;
537 uint64_t atime[2], mtime[2], ctime[2];
538
539 ASSERT(zp != NULL);
540 zfsvfs = zp->z_zfsvfs;
541 inode = ZTOI(zp);
542
543 sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), &atime, 16);
544 sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zfsvfs), &mtime, 16);
545 sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zfsvfs), &ctime, 16);
546
547 spin_lock(&inode->i_lock);
548 inode->i_generation = zp->z_gen;
549 inode->i_uid = zp->z_uid;
550 inode->i_gid = zp->z_gid;
551 inode->i_nlink = zp->z_links;
552 inode->i_mode = zp->z_mode;
553 inode->i_blkbits = SPA_MINBLOCKSHIFT;
554 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
555 (u_longlong_t *)&inode->i_blocks);
556
557 ZFS_TIME_DECODE(&inode->i_atime, atime);
558 ZFS_TIME_DECODE(&inode->i_mtime, mtime);
559 ZFS_TIME_DECODE(&inode->i_ctime, ctime);
560
561 i_size_write(inode, zp->z_size);
562 spin_unlock(&inode->i_lock);
563 }
564
565 static uint64_t empty_xattr;
566 static uint64_t pad[4];
567 static zfs_acl_phys_t acl_phys;
568 /*
569 * Create a new DMU object to hold a zfs znode.
570 *
571 * IN: dzp - parent directory for new znode
572 * vap - file attributes for new znode
573 * tx - dmu transaction id for zap operations
574 * cr - credentials of caller
575 * flag - flags:
576 * IS_ROOT_NODE - new object will be root
577 * IS_XATTR - new object is an attribute
578 * bonuslen - length of bonus buffer
579 * setaclp - File/Dir initial ACL
580 * fuidp - Tracks fuid allocation.
581 *
582 * OUT: zpp - allocated znode
583 *
584 */
585 void
586 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
587 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
588 {
589 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
590 uint64_t mode, size, links, parent, pflags;
591 uint64_t dzp_pflags = 0;
592 uint64_t rdev = 0;
593 zfsvfs_t *zfsvfs = dzp->z_zfsvfs;
594 dmu_buf_t *db;
595 timestruc_t now;
596 uint64_t gen, obj;
597 int err;
598 int bonuslen;
599 sa_handle_t *sa_hdl;
600 dmu_object_type_t obj_type;
601 sa_bulk_attr_t *sa_attrs;
602 int cnt = 0;
603 zfs_acl_locator_cb_t locate = { 0 };
604
605 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE));
606
607 if (zfsvfs->z_replay) {
608 obj = vap->va_nodeid;
609 now = vap->va_ctime; /* see zfs_replay_create() */
610 gen = vap->va_nblocks; /* ditto */
611 } else {
612 obj = 0;
613 gethrestime(&now);
614 gen = dmu_tx_get_txg(tx);
615 }
616
617 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
618 bonuslen = (obj_type == DMU_OT_SA) ?
619 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
620
621 /*
622 * Create a new DMU object.
623 */
624 /*
625 * There's currently no mechanism for pre-reading the blocks that will
626 * be needed to allocate a new object, so we accept the small chance
627 * that there will be an i/o error and we will fail one of the
628 * assertions below.
629 */
630 if (vap->va_type == VDIR) {
631 if (zfsvfs->z_replay) {
632 err = zap_create_claim_norm(zfsvfs->z_os, obj,
633 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
634 obj_type, bonuslen, tx);
635 ASSERT3U(err, ==, 0);
636 } else {
637 obj = zap_create_norm(zfsvfs->z_os,
638 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS,
639 obj_type, bonuslen, tx);
640 }
641 } else {
642 if (zfsvfs->z_replay) {
643 err = dmu_object_claim(zfsvfs->z_os, obj,
644 DMU_OT_PLAIN_FILE_CONTENTS, 0,
645 obj_type, bonuslen, tx);
646 ASSERT3U(err, ==, 0);
647 } else {
648 obj = dmu_object_alloc(zfsvfs->z_os,
649 DMU_OT_PLAIN_FILE_CONTENTS, 0,
650 obj_type, bonuslen, tx);
651 }
652 }
653
654 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
655 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db));
656
657 /*
658 * If this is the root, fix up the half-initialized parent pointer
659 * to reference the just-allocated physical data area.
660 */
661 if (flag & IS_ROOT_NODE) {
662 dzp->z_id = obj;
663 } else {
664 dzp_pflags = dzp->z_pflags;
665 }
666
667 /*
668 * If parent is an xattr, so am I.
669 */
670 if (dzp_pflags & ZFS_XATTR) {
671 flag |= IS_XATTR;
672 }
673
674 if (zfsvfs->z_use_fuids)
675 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
676 else
677 pflags = 0;
678
679 if (vap->va_type == VDIR) {
680 size = 2; /* contents ("." and "..") */
681 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
682 } else {
683 size = links = 0;
684 }
685
686 if (vap->va_type == VBLK || vap->va_type == VCHR) {
687 rdev = zfs_expldev(vap->va_rdev);
688 }
689
690 parent = dzp->z_id;
691 mode = acl_ids->z_mode;
692 if (flag & IS_XATTR)
693 pflags |= ZFS_XATTR;
694
695 /*
696 * No execs denied will be deterimed when zfs_mode_compute() is called.
697 */
698 pflags |= acl_ids->z_aclp->z_hints &
699 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
700 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
701
702 ZFS_TIME_ENCODE(&now, crtime);
703 ZFS_TIME_ENCODE(&now, ctime);
704
705 if (vap->va_mask & AT_ATIME) {
706 ZFS_TIME_ENCODE(&vap->va_atime, atime);
707 } else {
708 ZFS_TIME_ENCODE(&now, atime);
709 }
710
711 if (vap->va_mask & AT_MTIME) {
712 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
713 } else {
714 ZFS_TIME_ENCODE(&now, mtime);
715 }
716
717 /* Now add in all of the "SA" attributes */
718 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED,
719 &sa_hdl));
720
721 /*
722 * Setup the array of attributes to be replaced/set on the new file
723 *
724 * order for DMU_OT_ZNODE is critical since it needs to be constructed
725 * in the old znode_phys_t format. Don't change this ordering
726 */
727 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
728
729 if (obj_type == DMU_OT_ZNODE) {
730 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
731 NULL, &atime, 16);
732 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
733 NULL, &mtime, 16);
734 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
735 NULL, &ctime, 16);
736 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
737 NULL, &crtime, 16);
738 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
739 NULL, &gen, 8);
740 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
741 NULL, &mode, 8);
742 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
743 NULL, &size, 8);
744 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
745 NULL, &parent, 8);
746 } else {
747 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs),
748 NULL, &mode, 8);
749 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs),
750 NULL, &size, 8);
751 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs),
752 NULL, &gen, 8);
753 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
754 &acl_ids->z_fuid, 8);
755 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
756 &acl_ids->z_fgid, 8);
757 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs),
758 NULL, &parent, 8);
759 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
760 NULL, &pflags, 8);
761 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs),
762 NULL, &atime, 16);
763 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs),
764 NULL, &mtime, 16);
765 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs),
766 NULL, &ctime, 16);
767 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs),
768 NULL, &crtime, 16);
769 }
770
771 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8);
772
773 if (obj_type == DMU_OT_ZNODE) {
774 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL,
775 &empty_xattr, 8);
776 }
777 if (obj_type == DMU_OT_ZNODE ||
778 (vap->va_type == VBLK || vap->va_type == VCHR)) {
779 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs),
780 NULL, &rdev, 8);
781
782 }
783 if (obj_type == DMU_OT_ZNODE) {
784 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs),
785 NULL, &pflags, 8);
786 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL,
787 &acl_ids->z_fuid, 8);
788 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL,
789 &acl_ids->z_fgid, 8);
790 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad,
791 sizeof (uint64_t) * 4);
792 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL,
793 &acl_phys, sizeof (zfs_acl_phys_t));
794 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
795 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL,
796 &acl_ids->z_aclp->z_acl_count, 8);
797 locate.cb_aclp = acl_ids->z_aclp;
798 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs),
799 zfs_acl_data_locator, &locate,
800 acl_ids->z_aclp->z_acl_bytes);
801 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
802 acl_ids->z_fuid, acl_ids->z_fgid);
803 }
804
805 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
806
807 if (!(flag & IS_ROOT_NODE)) {
808 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl);
809 ASSERT(*zpp != NULL);
810 } else {
811 /*
812 * If we are creating the root node, the "parent" we
813 * passed in is the znode for the root.
814 */
815 *zpp = dzp;
816
817 (*zpp)->z_sa_hdl = sa_hdl;
818 }
819
820 (*zpp)->z_pflags = pflags;
821 (*zpp)->z_mode = mode;
822
823 if (vap->va_mask & AT_XVATTR)
824 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx);
825
826 if (obj_type == DMU_OT_ZNODE ||
827 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
828 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
829 ASSERT3S(err, ==, 0);
830 }
831 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
832 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
833 }
834
835 /*
836 * zfs_xvattr_set only updates the in-core attributes
837 * it is assumed the caller will be doing an sa_bulk_update
838 * to push the changes out
839 */
840 void
841 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
842 {
843 #ifdef HAVE_XVATTR
844 xoptattr_t *xoap;
845
846 xoap = xva_getxoptattr(xvap);
847 ASSERT(xoap);
848
849 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
850 uint64_t times[2];
851 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
852 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs),
853 &times, sizeof (times), tx);
854 XVA_SET_RTN(xvap, XAT_CREATETIME);
855 }
856 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
857 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
858 zp->z_pflags, tx);
859 XVA_SET_RTN(xvap, XAT_READONLY);
860 }
861 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
862 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
863 zp->z_pflags, tx);
864 XVA_SET_RTN(xvap, XAT_HIDDEN);
865 }
866 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
867 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
868 zp->z_pflags, tx);
869 XVA_SET_RTN(xvap, XAT_SYSTEM);
870 }
871 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
872 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
873 zp->z_pflags, tx);
874 XVA_SET_RTN(xvap, XAT_ARCHIVE);
875 }
876 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
877 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
878 zp->z_pflags, tx);
879 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
880 }
881 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
882 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
883 zp->z_pflags, tx);
884 XVA_SET_RTN(xvap, XAT_NOUNLINK);
885 }
886 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
887 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
888 zp->z_pflags, tx);
889 XVA_SET_RTN(xvap, XAT_APPENDONLY);
890 }
891 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
892 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
893 zp->z_pflags, tx);
894 XVA_SET_RTN(xvap, XAT_NODUMP);
895 }
896 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
897 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
898 zp->z_pflags, tx);
899 XVA_SET_RTN(xvap, XAT_OPAQUE);
900 }
901 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
902 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
903 xoap->xoa_av_quarantined, zp->z_pflags, tx);
904 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
905 }
906 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
907 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
908 zp->z_pflags, tx);
909 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
910 }
911 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
912 zfs_sa_set_scanstamp(zp, xvap, tx);
913 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
914 }
915 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
916 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
917 zp->z_pflags, tx);
918 XVA_SET_RTN(xvap, XAT_REPARSE);
919 }
920 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
921 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
922 zp->z_pflags, tx);
923 XVA_SET_RTN(xvap, XAT_OFFLINE);
924 }
925 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
926 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
927 zp->z_pflags, tx);
928 XVA_SET_RTN(xvap, XAT_SPARSE);
929 }
930 #endif /* HAVE_XVATTR */
931 }
932
933 int
934 zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp)
935 {
936 dmu_object_info_t doi;
937 dmu_buf_t *db;
938 znode_t *zp;
939 int err;
940 sa_handle_t *hdl;
941
942 *zpp = NULL;
943
944 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
945
946 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
947 if (err) {
948 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
949 return (err);
950 }
951
952 dmu_object_info_from_db(db, &doi);
953 if (doi.doi_bonus_type != DMU_OT_SA &&
954 (doi.doi_bonus_type != DMU_OT_ZNODE ||
955 (doi.doi_bonus_type == DMU_OT_ZNODE &&
956 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
957 sa_buf_rele(db, NULL);
958 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
959 return (EINVAL);
960 }
961
962 hdl = dmu_buf_get_user(db);
963 if (hdl != NULL) {
964 zp = sa_get_userdata(hdl);
965
966
967 /*
968 * Since "SA" does immediate eviction we
969 * should never find a sa handle that doesn't
970 * know about the znode.
971 */
972
973 ASSERT3P(zp, !=, NULL);
974
975 mutex_enter(&zp->z_lock);
976 ASSERT3U(zp->z_id, ==, obj_num);
977 if (zp->z_unlinked) {
978 err = ENOENT;
979 } else {
980 VN_HOLD(ZTOV(zp));
981 *zpp = zp;
982 err = 0;
983 }
984 sa_buf_rele(db, NULL);
985 mutex_exit(&zp->z_lock);
986 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
987 return (err);
988 }
989
990 /*
991 * Not found create new znode/vnode
992 * but only if file exists.
993 *
994 * There is a small window where zfs_vget() could
995 * find this object while a file create is still in
996 * progress. This is checked for in zfs_znode_alloc()
997 *
998 * if zfs_znode_alloc() fails it will drop the hold on the
999 * bonus buffer.
1000 */
1001 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size,
1002 doi.doi_bonus_type, NULL);
1003 if (zp == NULL) {
1004 err = ENOENT;
1005 } else {
1006 *zpp = zp;
1007 }
1008 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1009 return (err);
1010 }
1011
1012 int
1013 zfs_rezget(znode_t *zp)
1014 {
1015 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1016 dmu_object_info_t doi;
1017 dmu_buf_t *db;
1018 uint64_t obj_num = zp->z_id;
1019 uint64_t mode;
1020 sa_bulk_attr_t bulk[8];
1021 int err;
1022 int count = 0;
1023 uint64_t gen;
1024
1025 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num);
1026
1027 mutex_enter(&zp->z_acl_lock);
1028 if (zp->z_acl_cached) {
1029 zfs_acl_free(zp->z_acl_cached);
1030 zp->z_acl_cached = NULL;
1031 }
1032
1033 mutex_exit(&zp->z_acl_lock);
1034 ASSERT(zp->z_sa_hdl == NULL);
1035 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db);
1036 if (err) {
1037 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1038 return (err);
1039 }
1040
1041 dmu_object_info_from_db(db, &doi);
1042 if (doi.doi_bonus_type != DMU_OT_SA &&
1043 (doi.doi_bonus_type != DMU_OT_ZNODE ||
1044 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1045 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
1046 sa_buf_rele(db, NULL);
1047 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1048 return (EINVAL);
1049 }
1050
1051 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL);
1052
1053 /* reload cached values */
1054 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL,
1055 &gen, sizeof (gen));
1056 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL,
1057 &zp->z_size, sizeof (zp->z_size));
1058 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL,
1059 &zp->z_links, sizeof (zp->z_links));
1060 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL,
1061 &zp->z_pflags, sizeof (zp->z_pflags));
1062 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL,
1063 &zp->z_atime, sizeof (zp->z_atime));
1064 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL,
1065 &zp->z_uid, sizeof (zp->z_uid));
1066 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL,
1067 &zp->z_gid, sizeof (zp->z_gid));
1068 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL,
1069 &mode, sizeof (mode));
1070
1071 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
1072 zfs_znode_dmu_fini(zp);
1073 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1074 return (EIO);
1075 }
1076
1077 zp->z_mode = mode;
1078
1079 if (gen != zp->z_gen) {
1080 zfs_znode_dmu_fini(zp);
1081 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1082 return (EIO);
1083 }
1084
1085 zp->z_unlinked = (zp->z_links == 0);
1086 zp->z_blksz = doi.doi_data_block_size;
1087
1088 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num);
1089
1090 return (0);
1091 }
1092
1093 void
1094 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
1095 {
1096 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1097 objset_t *os = zfsvfs->z_os;
1098 uint64_t obj = zp->z_id;
1099 uint64_t acl_obj = zfs_external_acl(zp);
1100
1101 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj);
1102 if (acl_obj) {
1103 VERIFY(!zp->z_is_sa);
1104 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
1105 }
1106 VERIFY(0 == dmu_object_free(os, obj, tx));
1107 zfs_znode_dmu_fini(zp);
1108 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj);
1109 zfs_znode_free(zp);
1110 }
1111
1112 void
1113 zfs_zinactive(znode_t *zp)
1114 {
1115 vnode_t *vp = ZTOV(zp);
1116 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1117 uint64_t z_id = zp->z_id;
1118
1119 ASSERT(zp->z_sa_hdl);
1120
1121 /*
1122 * Don't allow a zfs_zget() while were trying to release this znode
1123 */
1124 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id);
1125
1126 mutex_enter(&zp->z_lock);
1127 mutex_enter(&vp->v_lock);
1128 vp->v_count--;
1129 if (vp->v_count > 0 || vn_has_cached_data(vp)) {
1130 /*
1131 * If the hold count is greater than zero, somebody has
1132 * obtained a new reference on this znode while we were
1133 * processing it here, so we are done. If we still have
1134 * mapped pages then we are also done, since we don't
1135 * want to inactivate the znode until the pages get pushed.
1136 *
1137 * XXX - if vn_has_cached_data(vp) is true, but count == 0,
1138 * this seems like it would leave the znode hanging with
1139 * no chance to go inactive...
1140 */
1141 mutex_exit(&vp->v_lock);
1142 mutex_exit(&zp->z_lock);
1143 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1144 return;
1145 }
1146 mutex_exit(&vp->v_lock);
1147
1148 /*
1149 * If this was the last reference to a file with no links,
1150 * remove the file from the file system.
1151 */
1152 if (zp->z_unlinked) {
1153 mutex_exit(&zp->z_lock);
1154 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1155 zfs_rmnode(zp);
1156 return;
1157 }
1158
1159 mutex_exit(&zp->z_lock);
1160 zfs_znode_dmu_fini(zp);
1161 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id);
1162 zfs_znode_free(zp);
1163 }
1164
1165 void
1166 zfs_znode_free(znode_t *zp)
1167 {
1168 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1169
1170 vn_invalid(ZTOV(zp));
1171
1172 ASSERT(ZTOV(zp)->v_count == 0);
1173
1174 mutex_enter(&zfsvfs->z_znodes_lock);
1175 POINTER_INVALIDATE(&zp->z_zfsvfs);
1176 list_remove(&zfsvfs->z_all_znodes, zp);
1177 mutex_exit(&zfsvfs->z_znodes_lock);
1178
1179 if (zp->z_acl_cached) {
1180 zfs_acl_free(zp->z_acl_cached);
1181 zp->z_acl_cached = NULL;
1182 }
1183
1184 kmem_cache_free(znode_cache, zp);
1185
1186 VFS_RELE(zfsvfs->z_vfs);
1187 }
1188
1189 void
1190 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1191 uint64_t ctime[2], boolean_t have_tx)
1192 {
1193 timestruc_t now;
1194
1195 gethrestime(&now);
1196
1197 if (have_tx) { /* will sa_bulk_update happen really soon? */
1198 zp->z_atime_dirty = 0;
1199 zp->z_seq++;
1200 } else {
1201 zp->z_atime_dirty = 1;
1202 }
1203
1204 if (flag & AT_ATIME) {
1205 ZFS_TIME_ENCODE(&now, zp->z_atime);
1206 }
1207
1208 if (flag & AT_MTIME) {
1209 ZFS_TIME_ENCODE(&now, mtime);
1210 if (zp->z_zfsvfs->z_use_fuids) {
1211 zp->z_pflags |= (ZFS_ARCHIVE |
1212 ZFS_AV_MODIFIED);
1213 }
1214 }
1215
1216 if (flag & AT_CTIME) {
1217 ZFS_TIME_ENCODE(&now, ctime);
1218 if (zp->z_zfsvfs->z_use_fuids)
1219 zp->z_pflags |= ZFS_ARCHIVE;
1220 }
1221 }
1222
1223 /*
1224 * Grow the block size for a file.
1225 *
1226 * IN: zp - znode of file to free data in.
1227 * size - requested block size
1228 * tx - open transaction.
1229 *
1230 * NOTE: this function assumes that the znode is write locked.
1231 */
1232 void
1233 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1234 {
1235 int error;
1236 u_longlong_t dummy;
1237
1238 if (size <= zp->z_blksz)
1239 return;
1240 /*
1241 * If the file size is already greater than the current blocksize,
1242 * we will not grow. If there is more than one block in a file,
1243 * the blocksize cannot change.
1244 */
1245 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1246 return;
1247
1248 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id,
1249 size, 0, tx);
1250
1251 if (error == ENOTSUP)
1252 return;
1253 ASSERT3U(error, ==, 0);
1254
1255 /* What blocksize did we actually get? */
1256 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1257 }
1258
1259 #ifdef HAVE_ZPL
1260 /*
1261 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1262 * be calling back into the fs for a putpage(). E.g.: when truncating
1263 * a file, the pages being "thrown away* don't need to be written out.
1264 */
1265 /* ARGSUSED */
1266 static int
1267 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1268 int flags, cred_t *cr)
1269 {
1270 ASSERT(0);
1271 return (0);
1272 }
1273 #endif /* HAVE_ZPL */
1274
1275 /*
1276 * Increase the file length
1277 *
1278 * IN: zp - znode of file to free data in.
1279 * end - new end-of-file
1280 *
1281 * RETURN: 0 if success
1282 * error code if failure
1283 */
1284 static int
1285 zfs_extend(znode_t *zp, uint64_t end)
1286 {
1287 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1288 dmu_tx_t *tx;
1289 rl_t *rl;
1290 uint64_t newblksz;
1291 int error;
1292
1293 /*
1294 * We will change zp_size, lock the whole file.
1295 */
1296 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1297
1298 /*
1299 * Nothing to do if file already at desired length.
1300 */
1301 if (end <= zp->z_size) {
1302 zfs_range_unlock(rl);
1303 return (0);
1304 }
1305 top:
1306 tx = dmu_tx_create(zfsvfs->z_os);
1307 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1308 zfs_sa_upgrade_txholds(tx, zp);
1309 if (end > zp->z_blksz &&
1310 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) {
1311 /*
1312 * We are growing the file past the current block size.
1313 */
1314 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) {
1315 ASSERT(!ISP2(zp->z_blksz));
1316 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1317 } else {
1318 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz);
1319 }
1320 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1321 } else {
1322 newblksz = 0;
1323 }
1324
1325 error = dmu_tx_assign(tx, TXG_NOWAIT);
1326 if (error) {
1327 if (error == ERESTART) {
1328 dmu_tx_wait(tx);
1329 dmu_tx_abort(tx);
1330 goto top;
1331 }
1332 dmu_tx_abort(tx);
1333 zfs_range_unlock(rl);
1334 return (error);
1335 }
1336
1337 if (newblksz)
1338 zfs_grow_blocksize(zp, newblksz, tx);
1339
1340 zp->z_size = end;
1341
1342 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs),
1343 &zp->z_size, sizeof (zp->z_size), tx));
1344
1345 zfs_range_unlock(rl);
1346
1347 dmu_tx_commit(tx);
1348
1349 return (0);
1350 }
1351
1352 /*
1353 * Free space in a file.
1354 *
1355 * IN: zp - znode of file to free data in.
1356 * off - start of section to free.
1357 * len - length of section to free.
1358 *
1359 * RETURN: 0 if success
1360 * error code if failure
1361 */
1362 static int
1363 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1364 {
1365 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1366 rl_t *rl;
1367 int error;
1368
1369 /*
1370 * Lock the range being freed.
1371 */
1372 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1373
1374 /*
1375 * Nothing to do if file already at desired length.
1376 */
1377 if (off >= zp->z_size) {
1378 zfs_range_unlock(rl);
1379 return (0);
1380 }
1381
1382 if (off + len > zp->z_size)
1383 len = zp->z_size - off;
1384
1385 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len);
1386
1387 zfs_range_unlock(rl);
1388
1389 return (error);
1390 }
1391
1392 /*
1393 * Truncate a file
1394 *
1395 * IN: zp - znode of file to free data in.
1396 * end - new end-of-file.
1397 *
1398 * RETURN: 0 if success
1399 * error code if failure
1400 */
1401 static int
1402 zfs_trunc(znode_t *zp, uint64_t end)
1403 {
1404 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1405 #ifdef HAVE_ZPL
1406 vnode_t *vp = ZTOV(zp);
1407 #endif /* HAVE_ZPL */
1408 dmu_tx_t *tx;
1409 rl_t *rl;
1410 int error;
1411 sa_bulk_attr_t bulk[2];
1412 int count = 0;
1413
1414 /*
1415 * We will change zp_size, lock the whole file.
1416 */
1417 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1418
1419 /*
1420 * Nothing to do if file already at desired length.
1421 */
1422 if (end >= zp->z_size) {
1423 zfs_range_unlock(rl);
1424 return (0);
1425 }
1426
1427 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1);
1428 if (error) {
1429 zfs_range_unlock(rl);
1430 return (error);
1431 }
1432 top:
1433 tx = dmu_tx_create(zfsvfs->z_os);
1434 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1435 zfs_sa_upgrade_txholds(tx, zp);
1436 error = dmu_tx_assign(tx, TXG_NOWAIT);
1437 if (error) {
1438 if (error == ERESTART) {
1439 dmu_tx_wait(tx);
1440 dmu_tx_abort(tx);
1441 goto top;
1442 }
1443 dmu_tx_abort(tx);
1444 zfs_range_unlock(rl);
1445 return (error);
1446 }
1447
1448 zp->z_size = end;
1449 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs),
1450 NULL, &zp->z_size, sizeof (zp->z_size));
1451
1452 if (end == 0) {
1453 zp->z_pflags &= ~ZFS_SPARSE;
1454 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1455 NULL, &zp->z_pflags, 8);
1456 }
1457 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1458
1459 dmu_tx_commit(tx);
1460
1461 #ifdef HAVE_ZPL
1462 /*
1463 * Clear any mapped pages in the truncated region. This has to
1464 * happen outside of the transaction to avoid the possibility of
1465 * a deadlock with someone trying to push a page that we are
1466 * about to invalidate.
1467 */
1468 if (vn_has_cached_data(vp)) {
1469 page_t *pp;
1470 uint64_t start = end & PAGEMASK;
1471 int poff = end & PAGEOFFSET;
1472
1473 if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) {
1474 /*
1475 * We need to zero a partial page.
1476 */
1477 pagezero(pp, poff, PAGESIZE - poff);
1478 start += PAGESIZE;
1479 page_unlock(pp);
1480 }
1481 error = pvn_vplist_dirty(vp, start, zfs_no_putpage,
1482 B_INVAL | B_TRUNC, NULL);
1483 ASSERT(error == 0);
1484 }
1485 #endif /* HAVE_ZPL */
1486
1487 zfs_range_unlock(rl);
1488
1489 return (0);
1490 }
1491
1492 /*
1493 * Free space in a file
1494 *
1495 * IN: zp - znode of file to free data in.
1496 * off - start of range
1497 * len - end of range (0 => EOF)
1498 * flag - current file open mode flags.
1499 * log - TRUE if this action should be logged
1500 *
1501 * RETURN: 0 if success
1502 * error code if failure
1503 */
1504 int
1505 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1506 {
1507 #ifdef HAVE_ZPL
1508 vnode_t *vp = ZTOV(zp);
1509 #endif /* HAVE_ZPL */
1510 dmu_tx_t *tx;
1511 zfsvfs_t *zfsvfs = zp->z_zfsvfs;
1512 zilog_t *zilog = zfsvfs->z_log;
1513 uint64_t mode;
1514 uint64_t mtime[2], ctime[2];
1515 sa_bulk_attr_t bulk[3];
1516 int count = 0;
1517 int error;
1518
1519 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode,
1520 sizeof (mode))) != 0)
1521 return (error);
1522
1523 if (off > zp->z_size) {
1524 error = zfs_extend(zp, off+len);
1525 if (error == 0 && log)
1526 goto log;
1527 else
1528 return (error);
1529 }
1530
1531 #ifdef HAVE_ZPL
1532 /*
1533 * Check for any locks in the region to be freed.
1534 */
1535
1536 if (MANDLOCK(vp, (mode_t)mode)) {
1537 uint64_t length = (len ? len : zp->z_size - off);
1538 if (error = chklock(vp, FWRITE, off, length, flag, NULL))
1539 return (error);
1540 }
1541 #endif /* HAVE_ZPL */
1542
1543 if (len == 0) {
1544 error = zfs_trunc(zp, off);
1545 } else {
1546 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1547 off + len > zp->z_size)
1548 error = zfs_extend(zp, off+len);
1549 }
1550 if (error || !log)
1551 return (error);
1552 log:
1553 tx = dmu_tx_create(zfsvfs->z_os);
1554 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1555 zfs_sa_upgrade_txholds(tx, zp);
1556 error = dmu_tx_assign(tx, TXG_NOWAIT);
1557 if (error) {
1558 if (error == ERESTART) {
1559 dmu_tx_wait(tx);
1560 dmu_tx_abort(tx);
1561 goto log;
1562 }
1563 dmu_tx_abort(tx);
1564 return (error);
1565 }
1566
1567 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16);
1568 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16);
1569 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs),
1570 NULL, &zp->z_pflags, 8);
1571 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1572 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1573 ASSERT(error == 0);
1574
1575 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1576
1577 dmu_tx_commit(tx);
1578 zfs_inode_update(zp);
1579 return (0);
1580 }
1581
1582 void
1583 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1584 {
1585 uint64_t moid, obj, sa_obj, version;
1586 uint64_t sense = ZFS_CASE_SENSITIVE;
1587 uint64_t norm = 0;
1588 nvpair_t *elem;
1589 int error;
1590 #ifdef HAVE_ZPL
1591 zfsvfs_t zfsvfs;
1592 int i;
1593 znode_t *rootzp = NULL;
1594 vnode_t *vp;
1595 vattr_t vattr;
1596 znode_t *zp;
1597 zfs_acl_ids_t acl_ids;
1598 #else
1599 timestruc_t now;
1600 dmu_buf_t *db;
1601 znode_phys_t *pzp;
1602 #endif /* HAVE_ZPL */
1603
1604 /*
1605 * First attempt to create master node.
1606 */
1607 /*
1608 * In an empty objset, there are no blocks to read and thus
1609 * there can be no i/o errors (which we assert below).
1610 */
1611 moid = MASTER_NODE_OBJ;
1612 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1613 DMU_OT_NONE, 0, tx);
1614 ASSERT(error == 0);
1615
1616 /*
1617 * Set starting attributes.
1618 */
1619 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1620 elem = NULL;
1621 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1622 /* For the moment we expect all zpl props to be uint64_ts */
1623 uint64_t val;
1624 char *name;
1625
1626 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1627 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1628 name = nvpair_name(elem);
1629 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1630 if (val < version)
1631 version = val;
1632 } else {
1633 error = zap_update(os, moid, name, 8, 1, &val, tx);
1634 }
1635 ASSERT(error == 0);
1636 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1637 norm = val;
1638 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
1639 sense = val;
1640 }
1641 ASSERT(version != 0);
1642 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1643
1644 /*
1645 * Create zap object used for SA attribute registration
1646 */
1647
1648 if (version >= ZPL_VERSION_SA) {
1649 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1650 DMU_OT_NONE, 0, tx);
1651 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1652 ASSERT(error == 0);
1653 } else {
1654 sa_obj = 0;
1655 }
1656 /*
1657 * Create a delete queue.
1658 */
1659 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1660
1661 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1662 ASSERT(error == 0);
1663
1664 #ifdef HAVE_ZPL
1665 /*
1666 * Create root znode. Create minimal znode/vnode/zfsvfs
1667 * to allow zfs_mknode to work.
1668 */
1669 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
1670 vattr.va_type = VDIR;
1671 vattr.va_mode = S_IFDIR|0755;
1672 vattr.va_uid = crgetuid(cr);
1673 vattr.va_gid = crgetgid(cr);
1674
1675 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP);
1676 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs));
1677 rootzp->z_moved = 0;
1678 rootzp->z_unlinked = 0;
1679 rootzp->z_atime_dirty = 0;
1680 rootzp->z_is_sa = USE_SA(version, os);
1681
1682 vp = ZTOV(rootzp);
1683 vn_reinit(vp);
1684 vp->v_type = VDIR;
1685
1686 bzero(&zfsvfs, sizeof (zfsvfs_t));
1687
1688 zfsvfs.z_os = os;
1689 zfsvfs.z_parent = &zfsvfs;
1690 zfsvfs.z_version = version;
1691 zfsvfs.z_use_fuids = USE_FUIDS(version, os);
1692 zfsvfs.z_use_sa = USE_SA(version, os);
1693 zfsvfs.z_norm = norm;
1694
1695 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
1696 &zfsvfs.z_attr_table);
1697
1698 ASSERT(error == 0);
1699
1700 /*
1701 * Fold case on file systems that are always or sometimes case
1702 * insensitive.
1703 */
1704 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
1705 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER;
1706
1707 /* XXX - This must be destroyed but I'm not quite sure yet so
1708 * I'm just annotating that fact when it's an issue. -Brian */
1709 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
1710 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t),
1711 offsetof(znode_t, z_link_node));
1712
1713 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1714 mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
1715
1716 rootzp->z_zfsvfs = &zfsvfs;
1717 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
1718 cr, NULL, &acl_ids));
1719 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
1720 ASSERT3P(zp, ==, rootzp);
1721 ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */
1722 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
1723 ASSERT(error == 0);
1724 zfs_acl_ids_free(&acl_ids);
1725 POINTER_INVALIDATE(&rootzp->z_zfsvfs);
1726
1727 ZTOV(rootzp)->v_count = 0;
1728 sa_handle_destroy(rootzp->z_sa_hdl);
1729 kmem_cache_free(znode_cache, rootzp);
1730 error = zfs_create_share_dir(&zfsvfs, tx);
1731
1732 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
1733 mutex_destroy(&zfsvfs.z_hold_mtx[i]);
1734 #else
1735 /*
1736 * Create root znode with code free of VFS dependencies
1737 */
1738 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1739 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1740
1741 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1742 dmu_buf_will_dirty(db, tx);
1743
1744 /*
1745 * Initialize the znode physical data to zero.
1746 */
1747 ASSERT(db->db_size >= sizeof (znode_phys_t));
1748 bzero(db->db_data, db->db_size);
1749 pzp = db->db_data;
1750
1751 if (USE_FUIDS(version, os))
1752 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1753
1754 pzp->zp_size = 2; /* "." and ".." */
1755 pzp->zp_links = 2;
1756 pzp->zp_parent = obj;
1757 pzp->zp_gen = dmu_tx_get_txg(tx);
1758 pzp->zp_mode = S_IFDIR | 0755;
1759 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1760
1761 gethrestime(&now);
1762
1763 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1764 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1765 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1766 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1767
1768 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1769 ASSERT(error == 0);
1770
1771 dmu_buf_rele(db, FTAG);
1772 #endif /* HAVE_ZPL */
1773 }
1774
1775 #endif /* _KERNEL */
1776
1777 static int
1778 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1779 {
1780 uint64_t sa_obj = 0;
1781 int error;
1782
1783 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1784 if (error != 0 && error != ENOENT)
1785 return (error);
1786
1787 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1788 return (error);
1789 }
1790
1791 static int
1792 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1793 dmu_buf_t **db)
1794 {
1795 dmu_object_info_t doi;
1796 int error;
1797
1798 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1799 return (error);
1800
1801 dmu_object_info_from_db(*db, &doi);
1802 if ((doi.doi_bonus_type != DMU_OT_SA &&
1803 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1804 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1805 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1806 sa_buf_rele(*db, FTAG);
1807 return (ENOTSUP);
1808 }
1809
1810 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1811 if (error != 0) {
1812 sa_buf_rele(*db, FTAG);
1813 return (error);
1814 }
1815
1816 return (0);
1817 }
1818
1819 void
1820 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1821 {
1822 sa_handle_destroy(hdl);
1823 sa_buf_rele(db, FTAG);
1824 }
1825
1826 /*
1827 * Given an object number, return its parent object number and whether
1828 * or not the object is an extended attribute directory.
1829 */
1830 static int
1831 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1832 int *is_xattrdir)
1833 {
1834 uint64_t parent;
1835 uint64_t pflags;
1836 uint64_t mode;
1837 sa_bulk_attr_t bulk[3];
1838 int count = 0;
1839 int error;
1840
1841 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1842 &parent, sizeof (parent));
1843 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1844 &pflags, sizeof (pflags));
1845 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1846 &mode, sizeof (mode));
1847
1848 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1849 return (error);
1850
1851 *pobjp = parent;
1852 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1853
1854 return (0);
1855 }
1856
1857 /*
1858 * Given an object number, return some zpl level statistics
1859 */
1860 static int
1861 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1862 zfs_stat_t *sb)
1863 {
1864 sa_bulk_attr_t bulk[4];
1865 int count = 0;
1866
1867 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1868 &sb->zs_mode, sizeof (sb->zs_mode));
1869 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1870 &sb->zs_gen, sizeof (sb->zs_gen));
1871 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1872 &sb->zs_links, sizeof (sb->zs_links));
1873 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1874 &sb->zs_ctime, sizeof (sb->zs_ctime));
1875
1876 return (sa_bulk_lookup(hdl, bulk, count));
1877 }
1878
1879 static int
1880 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1881 sa_attr_type_t *sa_table, char *buf, int len)
1882 {
1883 sa_handle_t *sa_hdl;
1884 sa_handle_t *prevhdl = NULL;
1885 dmu_buf_t *prevdb = NULL;
1886 dmu_buf_t *sa_db = NULL;
1887 char *path = buf + len - 1;
1888 int error;
1889
1890 *path = '\0';
1891 sa_hdl = hdl;
1892
1893 for (;;) {
1894 uint64_t pobj;
1895 char component[MAXNAMELEN + 2];
1896 size_t complen;
1897 int is_xattrdir;
1898
1899 if (prevdb)
1900 zfs_release_sa_handle(prevhdl, prevdb);
1901
1902 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1903 &is_xattrdir)) != 0)
1904 break;
1905
1906 if (pobj == obj) {
1907 if (path[0] != '/')
1908 *--path = '/';
1909 break;
1910 }
1911
1912 component[0] = '/';
1913 if (is_xattrdir) {
1914 (void) sprintf(component + 1, "<xattrdir>");
1915 } else {
1916 error = zap_value_search(osp, pobj, obj,
1917 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1918 if (error != 0)
1919 break;
1920 }
1921
1922 complen = strlen(component);
1923 path -= complen;
1924 ASSERT(path >= buf);
1925 bcopy(component, path, complen);
1926 obj = pobj;
1927
1928 if (sa_hdl != hdl) {
1929 prevhdl = sa_hdl;
1930 prevdb = sa_db;
1931 }
1932 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1933 if (error != 0) {
1934 sa_hdl = prevhdl;
1935 sa_db = prevdb;
1936 break;
1937 }
1938 }
1939
1940 if (sa_hdl != NULL && sa_hdl != hdl) {
1941 ASSERT(sa_db != NULL);
1942 zfs_release_sa_handle(sa_hdl, sa_db);
1943 }
1944
1945 if (error == 0)
1946 (void) memmove(buf, path, buf + len - path);
1947
1948 return (error);
1949 }
1950
1951 int
1952 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1953 {
1954 sa_attr_type_t *sa_table;
1955 sa_handle_t *hdl;
1956 dmu_buf_t *db;
1957 int error;
1958
1959 error = zfs_sa_setup(osp, &sa_table);
1960 if (error != 0)
1961 return (error);
1962
1963 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1964 if (error != 0)
1965 return (error);
1966
1967 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1968
1969 zfs_release_sa_handle(hdl, db);
1970 return (error);
1971 }
1972
1973 int
1974 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1975 char *buf, int len)
1976 {
1977 char *path = buf + len - 1;
1978 sa_attr_type_t *sa_table;
1979 sa_handle_t *hdl;
1980 dmu_buf_t *db;
1981 int error;
1982
1983 *path = '\0';
1984
1985 error = zfs_sa_setup(osp, &sa_table);
1986 if (error != 0)
1987 return (error);
1988
1989 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1990 if (error != 0)
1991 return (error);
1992
1993 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1994 if (error != 0) {
1995 zfs_release_sa_handle(hdl, db);
1996 return (error);
1997 }
1998
1999 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
2000
2001 zfs_release_sa_handle(hdl, db);
2002 return (error);
2003 }
2004
2005 #if defined(_KERNEL) && defined(HAVE_SPL)
2006 EXPORT_SYMBOL(zfs_create_fs);
2007 EXPORT_SYMBOL(zfs_obj_to_path);
2008 #endif
ZFS On Linux mirror for PVE