]> 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
Drop HAVE_XVATTR macros
[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/zfs_vnops.h>
55 #include <sys/dnode.h>
56 #include <sys/fs/zfs.h>
57 #include <sys/kidmap.h>
58 #include <sys/zpl.h>
59 #endif /* _KERNEL */
60
61 #include <sys/dmu.h>
62 #include <sys/refcount.h>
63 #include <sys/stat.h>
64 #include <sys/zap.h>
65 #include <sys/zfs_znode.h>
66 #include <sys/sa.h>
67 #include <sys/zfs_sa.h>
68 #include <sys/zfs_stat.h>
69
70 #include "zfs_prop.h"
71 #include "zfs_comutil.h"
72
73 /*
74 * Define ZNODE_STATS to turn on statistic gathering. By default, it is only
75 * turned on when DEBUG is also defined.
76 */
77 #ifdef DEBUG
78 #define ZNODE_STATS
79 #endif /* DEBUG */
80
81 #ifdef ZNODE_STATS
82 #define ZNODE_STAT_ADD(stat) ((stat)++)
83 #else
84 #define ZNODE_STAT_ADD(stat) /* nothing */
85 #endif /* ZNODE_STATS */
86
87 /*
88 * Functions needed for userland (ie: libzpool) are not put under
89 * #ifdef_KERNEL; the rest of the functions have dependencies
90 * (such as VFS logic) that will not compile easily in userland.
91 */
92 #ifdef _KERNEL
93
94 static kmem_cache_t *znode_cache = NULL;
95
96 /*ARGSUSED*/
97 static int
98 zfs_znode_cache_constructor(void *buf, void *arg, int kmflags)
99 {
100 znode_t *zp = buf;
101
102 inode_init_once(ZTOI(zp));
103 list_link_init(&zp->z_link_node);
104
105 mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL);
106 rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL);
107 rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL);
108 mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL);
109
110 mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL);
111 avl_create(&zp->z_range_avl, zfs_range_compare,
112 sizeof (rl_t), offsetof(rl_t, r_node));
113
114 zp->z_dirlocks = NULL;
115 zp->z_acl_cached = NULL;
116 zp->z_moved = 0;
117 return (0);
118 }
119
120 /*ARGSUSED*/
121 static void
122 zfs_znode_cache_destructor(void *buf, void *arg)
123 {
124 znode_t *zp = buf;
125
126 ASSERT(!list_link_active(&zp->z_link_node));
127 mutex_destroy(&zp->z_lock);
128 rw_destroy(&zp->z_parent_lock);
129 rw_destroy(&zp->z_name_lock);
130 mutex_destroy(&zp->z_acl_lock);
131 avl_destroy(&zp->z_range_avl);
132 mutex_destroy(&zp->z_range_lock);
133
134 ASSERT(zp->z_dirlocks == NULL);
135 ASSERT(zp->z_acl_cached == NULL);
136 }
137
138 void
139 zfs_znode_init(void)
140 {
141 /*
142 * Initialize zcache
143 */
144 ASSERT(znode_cache == NULL);
145 znode_cache = kmem_cache_create("zfs_znode_cache",
146 sizeof (znode_t), 0, zfs_znode_cache_constructor,
147 zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_KMEM);
148 }
149
150 void
151 zfs_znode_fini(void)
152 {
153 /*
154 * Cleanup zcache
155 */
156 if (znode_cache)
157 kmem_cache_destroy(znode_cache);
158 znode_cache = NULL;
159 }
160
161 int
162 zfs_create_share_dir(zfs_sb_t *zsb, dmu_tx_t *tx)
163 {
164 #ifdef HAVE_SHARE
165 zfs_acl_ids_t acl_ids;
166 vattr_t vattr;
167 znode_t *sharezp;
168 vnode_t *vp;
169 znode_t *zp;
170 int error;
171
172 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE;
173 vattr.va_mode = S_IFDIR | 0555;
174 vattr.va_uid = crgetuid(kcred);
175 vattr.va_gid = crgetgid(kcred);
176
177 sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP);
178 sharezp->z_moved = 0;
179 sharezp->z_unlinked = 0;
180 sharezp->z_atime_dirty = 0;
181 sharezp->z_zfsvfs = zfsvfs;
182 sharezp->z_is_sa = zfsvfs->z_use_sa;
183
184 vp = ZTOV(sharezp);
185 vn_reinit(vp);
186 vp->v_type = VDIR;
187
188 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr,
189 kcred, NULL, &acl_ids));
190 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids);
191 ASSERT3P(zp, ==, sharezp);
192 ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */
193 POINTER_INVALIDATE(&sharezp->z_zfsvfs);
194 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
195 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx);
196 zfsvfs->z_shares_dir = sharezp->z_id;
197
198 zfs_acl_ids_free(&acl_ids);
199 // ZTOV(sharezp)->v_count = 0;
200 sa_handle_destroy(sharezp->z_sa_hdl);
201 kmem_cache_free(znode_cache, sharezp);
202
203 return (error);
204 #else
205 return (0);
206 #endif /* HAVE_SHARE */
207 }
208
209 static void
210 zfs_znode_sa_init(zfs_sb_t *zsb, znode_t *zp,
211 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
212 {
213 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zsb, zp->z_id)));
214
215 mutex_enter(&zp->z_lock);
216
217 ASSERT(zp->z_sa_hdl == NULL);
218 ASSERT(zp->z_acl_cached == NULL);
219 if (sa_hdl == NULL) {
220 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, zp,
221 SA_HDL_SHARED, &zp->z_sa_hdl));
222 } else {
223 zp->z_sa_hdl = sa_hdl;
224 sa_set_userp(sa_hdl, zp);
225 }
226
227 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
228
229 mutex_exit(&zp->z_lock);
230 }
231
232 void
233 zfs_znode_dmu_fini(znode_t *zp)
234 {
235 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(ZTOZSB(zp), zp->z_id)) ||
236 zp->z_unlinked ||
237 RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock));
238
239 sa_handle_destroy(zp->z_sa_hdl);
240 zp->z_sa_hdl = NULL;
241 }
242
243 /*
244 * Called by new_inode() to allocate a new inode.
245 */
246 int
247 zfs_inode_alloc(struct super_block *sb, struct inode **ip)
248 {
249 znode_t *zp;
250
251 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
252 *ip = ZTOI(zp);
253
254 return (0);
255 }
256
257 /*
258 * Called in multiple places when an inode should be destroyed.
259 */
260 void
261 zfs_inode_destroy(struct inode *ip)
262 {
263 znode_t *zp = ITOZ(ip);
264 zfs_sb_t *zsb = ZTOZSB(zp);
265
266 mutex_enter(&zsb->z_znodes_lock);
267 list_remove(&zsb->z_all_znodes, zp);
268 mutex_exit(&zsb->z_znodes_lock);
269
270 if (zp->z_acl_cached) {
271 zfs_acl_free(zp->z_acl_cached);
272 zp->z_acl_cached = NULL;
273 }
274
275 kmem_cache_free(znode_cache, zp);
276 }
277
278 static void
279 zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip)
280 {
281 uint64_t rdev;
282
283 switch (ip->i_mode & S_IFMT) {
284 case S_IFREG:
285 ip->i_op = &zpl_inode_operations;
286 ip->i_fop = &zpl_file_operations;
287 ip->i_mapping->a_ops = &zpl_address_space_operations;
288 break;
289
290 case S_IFDIR:
291 ip->i_op = &zpl_dir_inode_operations;
292 ip->i_fop = &zpl_dir_file_operations;
293 ITOZ(ip)->z_zn_prefetch = B_TRUE;
294 break;
295
296 case S_IFLNK:
297 ip->i_op = &zpl_symlink_inode_operations;
298 break;
299
300 case S_IFCHR:
301 case S_IFBLK:
302 case S_IFIFO:
303 case S_IFSOCK:
304 VERIFY(sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb),
305 &rdev, sizeof (rdev)) == 0);
306 init_special_inode(ip, ip->i_mode, rdev);
307 ip->i_op = &zpl_special_inode_operations;
308 break;
309
310 default:
311 printk("ZFS: Invalid mode: 0x%x\n", ip->i_mode);
312 VERIFY(0);
313 }
314 }
315
316 /*
317 * Construct a znode+inode and initialize.
318 *
319 * This does not do a call to dmu_set_user() that is
320 * up to the caller to do, in case you don't want to
321 * return the znode
322 */
323 static znode_t *
324 zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
325 dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl)
326 {
327 znode_t *zp;
328 struct inode *ip;
329 uint64_t parent;
330 sa_bulk_attr_t bulk[9];
331 int count = 0;
332
333 ASSERT(zsb != NULL);
334
335 ip = new_inode(zsb->z_sb);
336 if (ip == NULL)
337 return (NULL);
338
339 zp = ITOZ(ip);
340 ASSERT(zp->z_dirlocks == NULL);
341 zp->z_moved = 0;
342 zp->z_sa_hdl = NULL;
343 zp->z_unlinked = 0;
344 zp->z_atime_dirty = 0;
345 zp->z_mapcnt = 0;
346 zp->z_id = db->db_object;
347 zp->z_blksz = blksz;
348 zp->z_seq = 0x7A4653;
349 zp->z_sync_cnt = 0;
350
351 zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
352
353 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8);
354 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
355 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
356 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
357 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
358 &zp->z_pflags, 8);
359 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
360 &parent, 8);
361 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
362 &zp->z_atime, 16);
363 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
364 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
365
366 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
367 if (hdl == NULL)
368 sa_handle_destroy(zp->z_sa_hdl);
369
370 goto error;
371 }
372
373 ip->i_ino = obj;
374 ip->i_mode = zp->z_mode;
375 ip->i_mtime = ip->i_atime = ip->i_ctime = CURRENT_TIME_SEC;
376 zfs_inode_set_ops(zsb, ip);
377
378 if (insert_inode_locked(ip))
379 goto error;
380
381 mutex_enter(&zsb->z_znodes_lock);
382 list_insert_tail(&zsb->z_all_znodes, zp);
383 membar_producer();
384 mutex_exit(&zsb->z_znodes_lock);
385
386 unlock_new_inode(ip);
387 return (zp);
388
389 error:
390 unlock_new_inode(ip);
391 iput(ip);
392 return NULL;
393 }
394
395 /*
396 * Update the embedded inode given the znode. We should work toward
397 * eliminating this function as soon as possible by removing values
398 * which are duplicated between the znode and inode. If the generic
399 * inode has the correct field it should be used, and the ZFS code
400 * updated to access the inode. This can be done incrementally.
401 */
402 void
403 zfs_inode_update(znode_t *zp)
404 {
405 zfs_sb_t *zsb;
406 struct inode *ip;
407 uint32_t blksize;
408 uint64_t atime[2], mtime[2], ctime[2];
409
410 ASSERT(zp != NULL);
411 zsb = ZTOZSB(zp);
412 ip = ZTOI(zp);
413
414 sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16);
415 sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16);
416 sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16);
417
418 spin_lock(&ip->i_lock);
419 ip->i_generation = zp->z_gen;
420 ip->i_uid = zp->z_uid;
421 ip->i_gid = zp->z_gid;
422 ip->i_nlink = zp->z_links;
423 ip->i_mode = zp->z_mode;
424 ip->i_blkbits = SPA_MINBLOCKSHIFT;
425 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
426 (u_longlong_t *)&ip->i_blocks);
427
428 ZFS_TIME_DECODE(&ip->i_atime, atime);
429 ZFS_TIME_DECODE(&ip->i_mtime, mtime);
430 ZFS_TIME_DECODE(&ip->i_ctime, ctime);
431
432 i_size_write(ip, zp->z_size);
433 spin_unlock(&ip->i_lock);
434 }
435
436 static uint64_t empty_xattr;
437 static uint64_t pad[4];
438 static zfs_acl_phys_t acl_phys;
439 /*
440 * Create a new DMU object to hold a zfs znode.
441 *
442 * IN: dzp - parent directory for new znode
443 * vap - file attributes for new znode
444 * tx - dmu transaction id for zap operations
445 * cr - credentials of caller
446 * flag - flags:
447 * IS_ROOT_NODE - new object will be root
448 * IS_XATTR - new object is an attribute
449 * bonuslen - length of bonus buffer
450 * setaclp - File/Dir initial ACL
451 * fuidp - Tracks fuid allocation.
452 *
453 * OUT: zpp - allocated znode
454 *
455 */
456 void
457 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
458 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
459 {
460 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
461 uint64_t mode, size, links, parent, pflags;
462 uint64_t dzp_pflags = 0;
463 uint64_t rdev = 0;
464 zfs_sb_t *zsb = ZTOZSB(dzp);
465 dmu_buf_t *db;
466 timestruc_t now;
467 uint64_t gen, obj;
468 int err;
469 int bonuslen;
470 sa_handle_t *sa_hdl;
471 dmu_object_type_t obj_type;
472 sa_bulk_attr_t *sa_attrs;
473 int cnt = 0;
474 zfs_acl_locator_cb_t locate = { 0 };
475
476 if (zsb->z_replay) {
477 obj = vap->va_nodeid;
478 now = vap->va_ctime; /* see zfs_replay_create() */
479 gen = vap->va_nblocks; /* ditto */
480 } else {
481 obj = 0;
482 gethrestime(&now);
483 gen = dmu_tx_get_txg(tx);
484 }
485
486 obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
487 bonuslen = (obj_type == DMU_OT_SA) ?
488 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
489
490 /*
491 * Create a new DMU object.
492 */
493 /*
494 * There's currently no mechanism for pre-reading the blocks that will
495 * be needed to allocate a new object, so we accept the small chance
496 * that there will be an i/o error and we will fail one of the
497 * assertions below.
498 */
499 if (S_ISDIR(vap->va_mode)) {
500 if (zsb->z_replay) {
501 err = zap_create_claim_norm(zsb->z_os, obj,
502 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
503 obj_type, bonuslen, tx);
504 ASSERT3U(err, ==, 0);
505 } else {
506 obj = zap_create_norm(zsb->z_os,
507 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
508 obj_type, bonuslen, tx);
509 }
510 } else {
511 if (zsb->z_replay) {
512 err = dmu_object_claim(zsb->z_os, obj,
513 DMU_OT_PLAIN_FILE_CONTENTS, 0,
514 obj_type, bonuslen, tx);
515 ASSERT3U(err, ==, 0);
516 } else {
517 obj = dmu_object_alloc(zsb->z_os,
518 DMU_OT_PLAIN_FILE_CONTENTS, 0,
519 obj_type, bonuslen, tx);
520 }
521 }
522
523 ZFS_OBJ_HOLD_ENTER(zsb, obj);
524 VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db));
525
526 /*
527 * If this is the root, fix up the half-initialized parent pointer
528 * to reference the just-allocated physical data area.
529 */
530 if (flag & IS_ROOT_NODE) {
531 dzp->z_id = obj;
532 } else {
533 dzp_pflags = dzp->z_pflags;
534 }
535
536 /*
537 * If parent is an xattr, so am I.
538 */
539 if (dzp_pflags & ZFS_XATTR) {
540 flag |= IS_XATTR;
541 }
542
543 if (zsb->z_use_fuids)
544 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
545 else
546 pflags = 0;
547
548 if (S_ISDIR(vap->va_mode)) {
549 size = 2; /* contents ("." and "..") */
550 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
551 } else {
552 size = links = 0;
553 }
554
555 if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
556 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))
557 rdev = vap->va_rdev;
558
559 parent = dzp->z_id;
560 mode = acl_ids->z_mode;
561 if (flag & IS_XATTR)
562 pflags |= ZFS_XATTR;
563
564 /*
565 * No execs denied will be deterimed when zfs_mode_compute() is called.
566 */
567 pflags |= acl_ids->z_aclp->z_hints &
568 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
569 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
570
571 ZFS_TIME_ENCODE(&now, crtime);
572 ZFS_TIME_ENCODE(&now, ctime);
573
574 if (vap->va_mask & ATTR_ATIME) {
575 ZFS_TIME_ENCODE(&vap->va_atime, atime);
576 } else {
577 ZFS_TIME_ENCODE(&now, atime);
578 }
579
580 if (vap->va_mask & ATTR_MTIME) {
581 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
582 } else {
583 ZFS_TIME_ENCODE(&now, mtime);
584 }
585
586 /* Now add in all of the "SA" attributes */
587 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED,
588 &sa_hdl));
589
590 /*
591 * Setup the array of attributes to be replaced/set on the new file
592 *
593 * order for DMU_OT_ZNODE is critical since it needs to be constructed
594 * in the old znode_phys_t format. Don't change this ordering
595 */
596 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
597
598 if (obj_type == DMU_OT_ZNODE) {
599 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
600 NULL, &atime, 16);
601 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
602 NULL, &mtime, 16);
603 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
604 NULL, &ctime, 16);
605 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
606 NULL, &crtime, 16);
607 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
608 NULL, &gen, 8);
609 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
610 NULL, &mode, 8);
611 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
612 NULL, &size, 8);
613 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
614 NULL, &parent, 8);
615 } else {
616 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
617 NULL, &mode, 8);
618 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
619 NULL, &size, 8);
620 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
621 NULL, &gen, 8);
622 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb),
623 NULL, &acl_ids->z_fuid, 8);
624 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb),
625 NULL, &acl_ids->z_fgid, 8);
626 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
627 NULL, &parent, 8);
628 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
629 NULL, &pflags, 8);
630 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
631 NULL, &atime, 16);
632 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
633 NULL, &mtime, 16);
634 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
635 NULL, &ctime, 16);
636 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
637 NULL, &crtime, 16);
638 }
639
640 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8);
641
642 if (obj_type == DMU_OT_ZNODE) {
643 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL,
644 &empty_xattr, 8);
645 }
646 if (obj_type == DMU_OT_ZNODE ||
647 (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
648 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))) {
649 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb),
650 NULL, &rdev, 8);
651 }
652 if (obj_type == DMU_OT_ZNODE) {
653 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
654 NULL, &pflags, 8);
655 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL,
656 &acl_ids->z_fuid, 8);
657 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL,
658 &acl_ids->z_fgid, 8);
659 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad,
660 sizeof (uint64_t) * 4);
661 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL,
662 &acl_phys, sizeof (zfs_acl_phys_t));
663 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
664 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL,
665 &acl_ids->z_aclp->z_acl_count, 8);
666 locate.cb_aclp = acl_ids->z_aclp;
667 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb),
668 zfs_acl_data_locator, &locate,
669 acl_ids->z_aclp->z_acl_bytes);
670 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
671 acl_ids->z_fuid, acl_ids->z_fgid);
672 }
673
674 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
675
676 if (!(flag & IS_ROOT_NODE)) {
677 *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl);
678 ASSERT(*zpp != NULL);
679 ASSERT(dzp != NULL);
680 err = zpl_xattr_security_init(ZTOI(*zpp), ZTOI(dzp));
681 ASSERT3S(err, ==, 0);
682 } else {
683 /*
684 * If we are creating the root node, the "parent" we
685 * passed in is the znode for the root.
686 */
687 *zpp = dzp;
688
689 (*zpp)->z_sa_hdl = sa_hdl;
690 }
691
692 (*zpp)->z_pflags = pflags;
693 (*zpp)->z_mode = mode;
694
695 if (obj_type == DMU_OT_ZNODE ||
696 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
697 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
698 ASSERT3S(err, ==, 0);
699 }
700 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
701 ZFS_OBJ_HOLD_EXIT(zsb, obj);
702 }
703
704 /*
705 * zfs_xvattr_set only updates the in-core attributes
706 * it is assumed the caller will be doing an sa_bulk_update
707 * to push the changes out
708 */
709 void
710 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
711 {
712 xoptattr_t *xoap;
713
714 xoap = xva_getxoptattr(xvap);
715 ASSERT(xoap);
716
717 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
718 uint64_t times[2];
719 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
720 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
721 &times, sizeof (times), tx);
722 XVA_SET_RTN(xvap, XAT_CREATETIME);
723 }
724 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
725 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
726 zp->z_pflags, tx);
727 XVA_SET_RTN(xvap, XAT_READONLY);
728 }
729 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
730 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
731 zp->z_pflags, tx);
732 XVA_SET_RTN(xvap, XAT_HIDDEN);
733 }
734 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
735 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
736 zp->z_pflags, tx);
737 XVA_SET_RTN(xvap, XAT_SYSTEM);
738 }
739 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
740 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
741 zp->z_pflags, tx);
742 XVA_SET_RTN(xvap, XAT_ARCHIVE);
743 }
744 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
745 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
746 zp->z_pflags, tx);
747 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
748 }
749 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
750 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
751 zp->z_pflags, tx);
752 XVA_SET_RTN(xvap, XAT_NOUNLINK);
753 }
754 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
755 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
756 zp->z_pflags, tx);
757 XVA_SET_RTN(xvap, XAT_APPENDONLY);
758 }
759 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
760 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
761 zp->z_pflags, tx);
762 XVA_SET_RTN(xvap, XAT_NODUMP);
763 }
764 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
765 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
766 zp->z_pflags, tx);
767 XVA_SET_RTN(xvap, XAT_OPAQUE);
768 }
769 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
770 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
771 xoap->xoa_av_quarantined, zp->z_pflags, tx);
772 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
773 }
774 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
775 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
776 zp->z_pflags, tx);
777 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
778 }
779 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
780 zfs_sa_set_scanstamp(zp, xvap, tx);
781 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
782 }
783 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
784 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
785 zp->z_pflags, tx);
786 XVA_SET_RTN(xvap, XAT_REPARSE);
787 }
788 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
789 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
790 zp->z_pflags, tx);
791 XVA_SET_RTN(xvap, XAT_OFFLINE);
792 }
793 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
794 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
795 zp->z_pflags, tx);
796 XVA_SET_RTN(xvap, XAT_SPARSE);
797 }
798 }
799
800 int
801 zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp)
802 {
803 dmu_object_info_t doi;
804 dmu_buf_t *db;
805 znode_t *zp;
806 int err;
807 sa_handle_t *hdl;
808
809 *zpp = NULL;
810
811 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
812
813 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
814 if (err) {
815 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
816 return (err);
817 }
818
819 dmu_object_info_from_db(db, &doi);
820 if (doi.doi_bonus_type != DMU_OT_SA &&
821 (doi.doi_bonus_type != DMU_OT_ZNODE ||
822 (doi.doi_bonus_type == DMU_OT_ZNODE &&
823 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
824 sa_buf_rele(db, NULL);
825 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
826 return (EINVAL);
827 }
828
829 hdl = dmu_buf_get_user(db);
830 if (hdl != NULL) {
831 zp = sa_get_userdata(hdl);
832
833
834 /*
835 * Since "SA" does immediate eviction we
836 * should never find a sa handle that doesn't
837 * know about the znode.
838 */
839
840 ASSERT3P(zp, !=, NULL);
841
842 mutex_enter(&zp->z_lock);
843 ASSERT3U(zp->z_id, ==, obj_num);
844 if (zp->z_unlinked) {
845 err = ENOENT;
846 } else {
847 igrab(ZTOI(zp));
848 *zpp = zp;
849 err = 0;
850 }
851 sa_buf_rele(db, NULL);
852 mutex_exit(&zp->z_lock);
853 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
854 return (err);
855 }
856
857 /*
858 * Not found create new znode/vnode but only if file exists.
859 *
860 * There is a small window where zfs_vget() could
861 * find this object while a file create is still in
862 * progress. This is checked for in zfs_znode_alloc()
863 *
864 * if zfs_znode_alloc() fails it will drop the hold on the
865 * bonus buffer.
866 */
867 zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size,
868 doi.doi_bonus_type, obj_num, NULL);
869 if (zp == NULL) {
870 err = ENOENT;
871 } else {
872 *zpp = zp;
873 }
874 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
875 return (err);
876 }
877
878 int
879 zfs_rezget(znode_t *zp)
880 {
881 zfs_sb_t *zsb = ZTOZSB(zp);
882 dmu_object_info_t doi;
883 dmu_buf_t *db;
884 uint64_t obj_num = zp->z_id;
885 uint64_t mode;
886 sa_bulk_attr_t bulk[8];
887 int err;
888 int count = 0;
889 uint64_t gen;
890
891 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
892
893 mutex_enter(&zp->z_acl_lock);
894 if (zp->z_acl_cached) {
895 zfs_acl_free(zp->z_acl_cached);
896 zp->z_acl_cached = NULL;
897 }
898
899 mutex_exit(&zp->z_acl_lock);
900 ASSERT(zp->z_sa_hdl == NULL);
901 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
902 if (err) {
903 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
904 return (err);
905 }
906
907 dmu_object_info_from_db(db, &doi);
908 if (doi.doi_bonus_type != DMU_OT_SA &&
909 (doi.doi_bonus_type != DMU_OT_ZNODE ||
910 (doi.doi_bonus_type == DMU_OT_ZNODE &&
911 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
912 sa_buf_rele(db, NULL);
913 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
914 return (EINVAL);
915 }
916
917 zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL);
918
919 /* reload cached values */
920 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL,
921 &gen, sizeof (gen));
922 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL,
923 &zp->z_size, sizeof (zp->z_size));
924 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL,
925 &zp->z_links, sizeof (zp->z_links));
926 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
927 &zp->z_pflags, sizeof (zp->z_pflags));
928 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
929 &zp->z_atime, sizeof (zp->z_atime));
930 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL,
931 &zp->z_uid, sizeof (zp->z_uid));
932 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL,
933 &zp->z_gid, sizeof (zp->z_gid));
934 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
935 &mode, sizeof (mode));
936
937 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
938 zfs_znode_dmu_fini(zp);
939 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
940 return (EIO);
941 }
942
943 zp->z_mode = mode;
944
945 if (gen != zp->z_gen) {
946 zfs_znode_dmu_fini(zp);
947 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
948 return (EIO);
949 }
950
951 zp->z_unlinked = (zp->z_links == 0);
952 zp->z_blksz = doi.doi_data_block_size;
953
954 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
955
956 return (0);
957 }
958
959 void
960 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
961 {
962 zfs_sb_t *zsb = ZTOZSB(zp);
963 objset_t *os = zsb->z_os;
964 uint64_t obj = zp->z_id;
965 uint64_t acl_obj = zfs_external_acl(zp);
966
967 ZFS_OBJ_HOLD_ENTER(zsb, obj);
968 if (acl_obj) {
969 VERIFY(!zp->z_is_sa);
970 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
971 }
972 VERIFY(0 == dmu_object_free(os, obj, tx));
973 zfs_znode_dmu_fini(zp);
974 ZFS_OBJ_HOLD_EXIT(zsb, obj);
975 }
976
977 void
978 zfs_zinactive(znode_t *zp)
979 {
980 zfs_sb_t *zsb = ZTOZSB(zp);
981 uint64_t z_id = zp->z_id;
982
983 ASSERT(zp->z_sa_hdl);
984
985 /*
986 * Don't allow a zfs_zget() while were trying to release this znode
987 */
988 ZFS_OBJ_HOLD_ENTER(zsb, z_id);
989 mutex_enter(&zp->z_lock);
990
991 /*
992 * If this was the last reference to a file with no links,
993 * remove the file from the file system.
994 */
995 if (zp->z_unlinked) {
996 mutex_exit(&zp->z_lock);
997 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
998 zfs_rmnode(zp);
999 return;
1000 }
1001
1002 mutex_exit(&zp->z_lock);
1003 zfs_znode_dmu_fini(zp);
1004 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
1005 }
1006
1007 void
1008 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1009 uint64_t ctime[2], boolean_t have_tx)
1010 {
1011 timestruc_t now;
1012
1013 gethrestime(&now);
1014
1015 if (have_tx) { /* will sa_bulk_update happen really soon? */
1016 zp->z_atime_dirty = 0;
1017 zp->z_seq++;
1018 } else {
1019 zp->z_atime_dirty = 1;
1020 }
1021
1022 if (flag & ATTR_ATIME) {
1023 ZFS_TIME_ENCODE(&now, zp->z_atime);
1024 }
1025
1026 if (flag & ATTR_MTIME) {
1027 ZFS_TIME_ENCODE(&now, mtime);
1028 if (ZTOZSB(zp)->z_use_fuids) {
1029 zp->z_pflags |= (ZFS_ARCHIVE |
1030 ZFS_AV_MODIFIED);
1031 }
1032 }
1033
1034 if (flag & ATTR_CTIME) {
1035 ZFS_TIME_ENCODE(&now, ctime);
1036 if (ZTOZSB(zp)->z_use_fuids)
1037 zp->z_pflags |= ZFS_ARCHIVE;
1038 }
1039 }
1040
1041 /*
1042 * Grow the block size for a file.
1043 *
1044 * IN: zp - znode of file to free data in.
1045 * size - requested block size
1046 * tx - open transaction.
1047 *
1048 * NOTE: this function assumes that the znode is write locked.
1049 */
1050 void
1051 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1052 {
1053 int error;
1054 u_longlong_t dummy;
1055
1056 if (size <= zp->z_blksz)
1057 return;
1058 /*
1059 * If the file size is already greater than the current blocksize,
1060 * we will not grow. If there is more than one block in a file,
1061 * the blocksize cannot change.
1062 */
1063 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1064 return;
1065
1066 error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id,
1067 size, 0, tx);
1068
1069 if (error == ENOTSUP)
1070 return;
1071 ASSERT3U(error, ==, 0);
1072
1073 /* What blocksize did we actually get? */
1074 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1075 }
1076
1077 #ifdef HAVE_MMAP
1078 /*
1079 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1080 * be calling back into the fs for a putpage(). E.g.: when truncating
1081 * a file, the pages being "thrown away* don't need to be written out.
1082 */
1083 /* ARGSUSED */
1084 static int
1085 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1086 int flags, cred_t *cr)
1087 {
1088 ASSERT(0);
1089 return (0);
1090 }
1091 #endif /* HAVE_MMAP */
1092
1093 /*
1094 * Increase the file length
1095 *
1096 * IN: zp - znode of file to free data in.
1097 * end - new end-of-file
1098 *
1099 * RETURN: 0 if success
1100 * error code if failure
1101 */
1102 static int
1103 zfs_extend(znode_t *zp, uint64_t end)
1104 {
1105 zfs_sb_t *zsb = ZTOZSB(zp);
1106 dmu_tx_t *tx;
1107 rl_t *rl;
1108 uint64_t newblksz;
1109 int error;
1110
1111 /*
1112 * We will change zp_size, lock the whole file.
1113 */
1114 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1115
1116 /*
1117 * Nothing to do if file already at desired length.
1118 */
1119 if (end <= zp->z_size) {
1120 zfs_range_unlock(rl);
1121 return (0);
1122 }
1123 top:
1124 tx = dmu_tx_create(zsb->z_os);
1125 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1126 zfs_sa_upgrade_txholds(tx, zp);
1127 if (end > zp->z_blksz &&
1128 (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) {
1129 /*
1130 * We are growing the file past the current block size.
1131 */
1132 if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
1133 ASSERT(!ISP2(zp->z_blksz));
1134 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1135 } else {
1136 newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
1137 }
1138 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1139 } else {
1140 newblksz = 0;
1141 }
1142
1143 error = dmu_tx_assign(tx, TXG_NOWAIT);
1144 if (error) {
1145 if (error == ERESTART) {
1146 dmu_tx_wait(tx);
1147 dmu_tx_abort(tx);
1148 goto top;
1149 }
1150 dmu_tx_abort(tx);
1151 zfs_range_unlock(rl);
1152 return (error);
1153 }
1154
1155 if (newblksz)
1156 zfs_grow_blocksize(zp, newblksz, tx);
1157
1158 zp->z_size = end;
1159
1160 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)),
1161 &zp->z_size, sizeof (zp->z_size), tx));
1162
1163 zfs_range_unlock(rl);
1164
1165 dmu_tx_commit(tx);
1166
1167 return (0);
1168 }
1169
1170 /*
1171 * Free space in a file.
1172 *
1173 * IN: zp - znode of file to free data in.
1174 * off - start of section to free.
1175 * len - length of section to free.
1176 *
1177 * RETURN: 0 if success
1178 * error code if failure
1179 */
1180 static int
1181 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1182 {
1183 zfs_sb_t *zsb = ZTOZSB(zp);
1184 rl_t *rl;
1185 int error;
1186
1187 /*
1188 * Lock the range being freed.
1189 */
1190 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1191
1192 /*
1193 * Nothing to do if file already at desired length.
1194 */
1195 if (off >= zp->z_size) {
1196 zfs_range_unlock(rl);
1197 return (0);
1198 }
1199
1200 if (off + len > zp->z_size)
1201 len = zp->z_size - off;
1202
1203 error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len);
1204
1205 zfs_range_unlock(rl);
1206
1207 return (error);
1208 }
1209
1210 /*
1211 * Truncate a file
1212 *
1213 * IN: zp - znode of file to free data in.
1214 * end - new end-of-file.
1215 *
1216 * RETURN: 0 if success
1217 * error code if failure
1218 */
1219 static int
1220 zfs_trunc(znode_t *zp, uint64_t end)
1221 {
1222 zfs_sb_t *zsb = ZTOZSB(zp);
1223 dmu_tx_t *tx;
1224 rl_t *rl;
1225 int error;
1226 sa_bulk_attr_t bulk[2];
1227 int count = 0;
1228
1229 /*
1230 * We will change zp_size, lock the whole file.
1231 */
1232 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1233
1234 /*
1235 * Nothing to do if file already at desired length.
1236 */
1237 if (end >= zp->z_size) {
1238 zfs_range_unlock(rl);
1239 return (0);
1240 }
1241
1242 error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1);
1243 if (error) {
1244 zfs_range_unlock(rl);
1245 return (error);
1246 }
1247 top:
1248 tx = dmu_tx_create(zsb->z_os);
1249 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1250 zfs_sa_upgrade_txholds(tx, zp);
1251 error = dmu_tx_assign(tx, TXG_NOWAIT);
1252 if (error) {
1253 if (error == ERESTART) {
1254 dmu_tx_wait(tx);
1255 dmu_tx_abort(tx);
1256 goto top;
1257 }
1258 dmu_tx_abort(tx);
1259 zfs_range_unlock(rl);
1260 return (error);
1261 }
1262
1263 zp->z_size = end;
1264 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb),
1265 NULL, &zp->z_size, sizeof (zp->z_size));
1266
1267 if (end == 0) {
1268 zp->z_pflags &= ~ZFS_SPARSE;
1269 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1270 NULL, &zp->z_pflags, 8);
1271 }
1272 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1273
1274 dmu_tx_commit(tx);
1275
1276 zfs_range_unlock(rl);
1277
1278 return (0);
1279 }
1280
1281 /*
1282 * Free space in a file
1283 *
1284 * IN: zp - znode of file to free data in.
1285 * off - start of range
1286 * len - end of range (0 => EOF)
1287 * flag - current file open mode flags.
1288 * log - TRUE if this action should be logged
1289 *
1290 * RETURN: 0 if success
1291 * error code if failure
1292 */
1293 int
1294 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1295 {
1296 #ifdef HAVE_MANDLOCKS
1297 struct inode *ip = ZTOI(zp);
1298 #endif /* HAVE_MANDLOCKS */
1299 dmu_tx_t *tx;
1300 zfs_sb_t *zsb = ZTOZSB(zp);
1301 zilog_t *zilog = zsb->z_log;
1302 uint64_t mode;
1303 uint64_t mtime[2], ctime[2];
1304 sa_bulk_attr_t bulk[3];
1305 int count = 0;
1306 int error;
1307
1308 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode,
1309 sizeof (mode))) != 0)
1310 return (error);
1311
1312 if (off > zp->z_size) {
1313 error = zfs_extend(zp, off+len);
1314 if (error == 0 && log)
1315 goto log;
1316 else
1317 return (error);
1318 }
1319
1320 #ifdef HAVE_MANDLOCKS
1321 /*
1322 * Check for any locks in the region to be freed.
1323 */
1324
1325 if (MANDLOCK(ip, (mode_t)mode)) {
1326 uint64_t length = (len ? len : zp->z_size - off);
1327 if (error = chklock(ip, FWRITE, off, length, flag, NULL))
1328 return (error);
1329 }
1330 #endif /* HAVE_MANDLOCKS */
1331
1332 if (len == 0) {
1333 error = zfs_trunc(zp, off);
1334 } else {
1335 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1336 off + len > zp->z_size)
1337 error = zfs_extend(zp, off+len);
1338 }
1339 if (error || !log)
1340 return (error);
1341 log:
1342 tx = dmu_tx_create(zsb->z_os);
1343 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1344 zfs_sa_upgrade_txholds(tx, zp);
1345 error = dmu_tx_assign(tx, TXG_NOWAIT);
1346 if (error) {
1347 if (error == ERESTART) {
1348 dmu_tx_wait(tx);
1349 dmu_tx_abort(tx);
1350 goto log;
1351 }
1352 dmu_tx_abort(tx);
1353 return (error);
1354 }
1355
1356 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16);
1357 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16);
1358 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1359 NULL, &zp->z_pflags, 8);
1360 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1361 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1362 ASSERT(error == 0);
1363
1364 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1365
1366 dmu_tx_commit(tx);
1367 zfs_inode_update(zp);
1368 return (0);
1369 }
1370
1371 void
1372 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1373 {
1374 uint64_t moid, obj, sa_obj, version;
1375 uint64_t norm = 0;
1376 nvpair_t *elem;
1377 int error;
1378 timestruc_t now;
1379 dmu_buf_t *db;
1380 znode_phys_t *pzp;
1381
1382 /*
1383 * First attempt to create master node.
1384 */
1385 /*
1386 * In an empty objset, there are no blocks to read and thus
1387 * there can be no i/o errors (which we assert below).
1388 */
1389 moid = MASTER_NODE_OBJ;
1390 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1391 DMU_OT_NONE, 0, tx);
1392 ASSERT(error == 0);
1393
1394 /*
1395 * Set starting attributes.
1396 */
1397 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1398 elem = NULL;
1399 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1400 /* For the moment we expect all zpl props to be uint64_ts */
1401 uint64_t val;
1402 char *name;
1403
1404 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1405 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1406 name = nvpair_name(elem);
1407 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1408 if (val < version)
1409 version = val;
1410 } else {
1411 error = zap_update(os, moid, name, 8, 1, &val, tx);
1412 }
1413 ASSERT(error == 0);
1414 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1415 norm = val;
1416 }
1417 ASSERT(version != 0);
1418 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1419
1420 /*
1421 * Create zap object used for SA attribute registration
1422 */
1423
1424 if (version >= ZPL_VERSION_SA) {
1425 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1426 DMU_OT_NONE, 0, tx);
1427 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1428 ASSERT(error == 0);
1429 } else {
1430 sa_obj = 0;
1431 }
1432 /*
1433 * Create a delete queue.
1434 */
1435 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1436
1437 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1438 ASSERT(error == 0);
1439
1440 /*
1441 * Create root znode with code free of VFS dependencies. This
1442 * is important because without a registered filesystem and super
1443 * block all the required VFS hooks will be missing. The critical
1444 * thing is to just crete the required root znode.
1445 */
1446 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1447 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1448
1449 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1450 dmu_buf_will_dirty(db, tx);
1451
1452 /*
1453 * Initialize the znode physical data to zero.
1454 */
1455 ASSERT(db->db_size >= sizeof (znode_phys_t));
1456 bzero(db->db_data, db->db_size);
1457 pzp = db->db_data;
1458
1459 if (USE_FUIDS(version, os))
1460 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1461
1462 pzp->zp_size = 2; /* "." and ".." */
1463 pzp->zp_links = 2;
1464 pzp->zp_parent = obj;
1465 pzp->zp_gen = dmu_tx_get_txg(tx);
1466 pzp->zp_mode = S_IFDIR | 0755;
1467 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1468
1469 gethrestime(&now);
1470
1471 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1472 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1473 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1474 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1475
1476 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1477 ASSERT(error == 0);
1478
1479 dmu_buf_rele(db, FTAG);
1480 }
1481
1482 #endif /* _KERNEL */
1483
1484 static int
1485 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1486 {
1487 uint64_t sa_obj = 0;
1488 int error;
1489
1490 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1491 if (error != 0 && error != ENOENT)
1492 return (error);
1493
1494 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1495 return (error);
1496 }
1497
1498 static int
1499 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1500 dmu_buf_t **db)
1501 {
1502 dmu_object_info_t doi;
1503 int error;
1504
1505 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1506 return (error);
1507
1508 dmu_object_info_from_db(*db, &doi);
1509 if ((doi.doi_bonus_type != DMU_OT_SA &&
1510 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1511 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1512 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1513 sa_buf_rele(*db, FTAG);
1514 return (ENOTSUP);
1515 }
1516
1517 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1518 if (error != 0) {
1519 sa_buf_rele(*db, FTAG);
1520 return (error);
1521 }
1522
1523 return (0);
1524 }
1525
1526 void
1527 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1528 {
1529 sa_handle_destroy(hdl);
1530 sa_buf_rele(db, FTAG);
1531 }
1532
1533 /*
1534 * Given an object number, return its parent object number and whether
1535 * or not the object is an extended attribute directory.
1536 */
1537 static int
1538 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1539 int *is_xattrdir)
1540 {
1541 uint64_t parent;
1542 uint64_t pflags;
1543 uint64_t mode;
1544 sa_bulk_attr_t bulk[3];
1545 int count = 0;
1546 int error;
1547
1548 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1549 &parent, sizeof (parent));
1550 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1551 &pflags, sizeof (pflags));
1552 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1553 &mode, sizeof (mode));
1554
1555 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1556 return (error);
1557
1558 *pobjp = parent;
1559 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1560
1561 return (0);
1562 }
1563
1564 /*
1565 * Given an object number, return some zpl level statistics
1566 */
1567 static int
1568 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1569 zfs_stat_t *sb)
1570 {
1571 sa_bulk_attr_t bulk[4];
1572 int count = 0;
1573
1574 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1575 &sb->zs_mode, sizeof (sb->zs_mode));
1576 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1577 &sb->zs_gen, sizeof (sb->zs_gen));
1578 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1579 &sb->zs_links, sizeof (sb->zs_links));
1580 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1581 &sb->zs_ctime, sizeof (sb->zs_ctime));
1582
1583 return (sa_bulk_lookup(hdl, bulk, count));
1584 }
1585
1586 static int
1587 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1588 sa_attr_type_t *sa_table, char *buf, int len)
1589 {
1590 sa_handle_t *sa_hdl;
1591 sa_handle_t *prevhdl = NULL;
1592 dmu_buf_t *prevdb = NULL;
1593 dmu_buf_t *sa_db = NULL;
1594 char *path = buf + len - 1;
1595 int error;
1596
1597 *path = '\0';
1598 sa_hdl = hdl;
1599
1600 for (;;) {
1601 uint64_t pobj;
1602 char component[MAXNAMELEN + 2];
1603 size_t complen;
1604 int is_xattrdir;
1605
1606 if (prevdb)
1607 zfs_release_sa_handle(prevhdl, prevdb);
1608
1609 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1610 &is_xattrdir)) != 0)
1611 break;
1612
1613 if (pobj == obj) {
1614 if (path[0] != '/')
1615 *--path = '/';
1616 break;
1617 }
1618
1619 component[0] = '/';
1620 if (is_xattrdir) {
1621 (void) sprintf(component + 1, "<xattrdir>");
1622 } else {
1623 error = zap_value_search(osp, pobj, obj,
1624 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1625 if (error != 0)
1626 break;
1627 }
1628
1629 complen = strlen(component);
1630 path -= complen;
1631 ASSERT(path >= buf);
1632 bcopy(component, path, complen);
1633 obj = pobj;
1634
1635 if (sa_hdl != hdl) {
1636 prevhdl = sa_hdl;
1637 prevdb = sa_db;
1638 }
1639 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1640 if (error != 0) {
1641 sa_hdl = prevhdl;
1642 sa_db = prevdb;
1643 break;
1644 }
1645 }
1646
1647 if (sa_hdl != NULL && sa_hdl != hdl) {
1648 ASSERT(sa_db != NULL);
1649 zfs_release_sa_handle(sa_hdl, sa_db);
1650 }
1651
1652 if (error == 0)
1653 (void) memmove(buf, path, buf + len - path);
1654
1655 return (error);
1656 }
1657
1658 int
1659 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1660 {
1661 sa_attr_type_t *sa_table;
1662 sa_handle_t *hdl;
1663 dmu_buf_t *db;
1664 int error;
1665
1666 error = zfs_sa_setup(osp, &sa_table);
1667 if (error != 0)
1668 return (error);
1669
1670 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1671 if (error != 0)
1672 return (error);
1673
1674 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1675
1676 zfs_release_sa_handle(hdl, db);
1677 return (error);
1678 }
1679
1680 int
1681 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1682 char *buf, int len)
1683 {
1684 char *path = buf + len - 1;
1685 sa_attr_type_t *sa_table;
1686 sa_handle_t *hdl;
1687 dmu_buf_t *db;
1688 int error;
1689
1690 *path = '\0';
1691
1692 error = zfs_sa_setup(osp, &sa_table);
1693 if (error != 0)
1694 return (error);
1695
1696 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1697 if (error != 0)
1698 return (error);
1699
1700 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1701 if (error != 0) {
1702 zfs_release_sa_handle(hdl, db);
1703 return (error);
1704 }
1705
1706 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1707
1708 zfs_release_sa_handle(hdl, db);
1709 return (error);
1710 }
1711
1712 #if defined(_KERNEL) && defined(HAVE_SPL)
1713 EXPORT_SYMBOL(zfs_create_fs);
1714 EXPORT_SYMBOL(zfs_obj_to_path);
1715 #endif
ZFS On Linux mirror for PVE