]> 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
Merge branch 'zpl'
[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 /*
210 * define a couple of values we need available
211 * for both 64 and 32 bit environments.
212 */
213 #ifndef NBITSMINOR64
214 #define NBITSMINOR64 32
215 #endif
216 #ifndef MAXMAJ64
217 #define MAXMAJ64 0xffffffffUL
218 #endif
219 #ifndef MAXMIN64
220 #define MAXMIN64 0xffffffffUL
221 #endif
222
223 /*
224 * Create special expldev for ZFS private use.
225 * Can't use standard expldev since it doesn't do
226 * what we want. The standard expldev() takes a
227 * dev32_t in LP64 and expands it to a long dev_t.
228 * We need an interface that takes a dev32_t in ILP32
229 * and expands it to a long dev_t.
230 */
231 static uint64_t
232 zfs_expldev(dev_t dev)
233 {
234 #ifndef _LP64
235 major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32;
236 return (((uint64_t)major << NBITSMINOR64) |
237 ((minor_t)dev & MAXMIN32));
238 #else
239 return (dev);
240 #endif
241 }
242
243 static void
244 zfs_znode_sa_init(zfs_sb_t *zsb, znode_t *zp,
245 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl)
246 {
247 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zsb, zp->z_id)));
248
249 mutex_enter(&zp->z_lock);
250
251 ASSERT(zp->z_sa_hdl == NULL);
252 ASSERT(zp->z_acl_cached == NULL);
253 if (sa_hdl == NULL) {
254 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, zp,
255 SA_HDL_SHARED, &zp->z_sa_hdl));
256 } else {
257 zp->z_sa_hdl = sa_hdl;
258 sa_set_userp(sa_hdl, zp);
259 }
260
261 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE;
262
263 mutex_exit(&zp->z_lock);
264 }
265
266 void
267 zfs_znode_dmu_fini(znode_t *zp)
268 {
269 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(ZTOZSB(zp), zp->z_id)) ||
270 zp->z_unlinked ||
271 RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock));
272
273 sa_handle_destroy(zp->z_sa_hdl);
274 zp->z_sa_hdl = NULL;
275 }
276
277 /*
278 * Called by new_inode() to allocate a new inode.
279 */
280 int
281 zfs_inode_alloc(struct super_block *sb, struct inode **ip)
282 {
283 znode_t *zp;
284
285 zp = kmem_cache_alloc(znode_cache, KM_SLEEP);
286 *ip = ZTOI(zp);
287
288 return (0);
289 }
290
291 /*
292 * Called in multiple places when an inode should be destroyed.
293 */
294 void
295 zfs_inode_destroy(struct inode *ip)
296 {
297 znode_t *zp = ITOZ(ip);
298 zfs_sb_t *zsb = ZTOZSB(zp);
299
300 mutex_enter(&zsb->z_znodes_lock);
301 list_remove(&zsb->z_all_znodes, zp);
302 mutex_exit(&zsb->z_znodes_lock);
303
304 if (zp->z_acl_cached) {
305 zfs_acl_free(zp->z_acl_cached);
306 zp->z_acl_cached = NULL;
307 }
308
309 kmem_cache_free(znode_cache, zp);
310 }
311
312 static void
313 zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip)
314 {
315 uint64_t rdev;
316
317 switch (ip->i_mode & S_IFMT) {
318 case S_IFREG:
319 ip->i_op = &zpl_inode_operations;
320 ip->i_fop = &zpl_file_operations;
321 ip->i_mapping->a_ops = &zpl_address_space_operations;
322 break;
323
324 case S_IFDIR:
325 ip->i_op = &zpl_dir_inode_operations;
326 ip->i_fop = &zpl_dir_file_operations;
327 ITOZ(ip)->z_zn_prefetch = B_TRUE;
328 break;
329
330 case S_IFLNK:
331 ip->i_op = &zpl_symlink_inode_operations;
332 break;
333
334 case S_IFCHR:
335 case S_IFBLK:
336 case S_IFIFO:
337 case S_IFSOCK:
338 VERIFY(sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb),
339 &rdev, sizeof (rdev)) == 0);
340 init_special_inode(ip, ip->i_mode, rdev);
341 ip->i_op = &zpl_special_inode_operations;
342 break;
343
344 default:
345 printk("ZFS: Invalid mode: 0x%x\n", ip->i_mode);
346 VERIFY(0);
347 }
348 }
349
350 /*
351 * Construct a znode+inode and initialize.
352 *
353 * This does not do a call to dmu_set_user() that is
354 * up to the caller to do, in case you don't want to
355 * return the znode
356 */
357 static znode_t *
358 zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz,
359 dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl)
360 {
361 znode_t *zp;
362 struct inode *ip;
363 uint64_t parent;
364 sa_bulk_attr_t bulk[9];
365 int count = 0;
366
367 ASSERT(zsb != NULL);
368
369 ip = new_inode(zsb->z_sb);
370 if (ip == NULL)
371 return (NULL);
372
373 zp = ITOZ(ip);
374 ASSERT(zp->z_dirlocks == NULL);
375 zp->z_moved = 0;
376 zp->z_sa_hdl = NULL;
377 zp->z_unlinked = 0;
378 zp->z_atime_dirty = 0;
379 zp->z_mapcnt = 0;
380 zp->z_id = db->db_object;
381 zp->z_blksz = blksz;
382 zp->z_seq = 0x7A4653;
383 zp->z_sync_cnt = 0;
384
385 zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
386
387 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8);
388 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
389 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
390 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
391 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
392 &zp->z_pflags, 8);
393 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
394 &parent, 8);
395 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
396 &zp->z_atime, 16);
397 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
398 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
399
400 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
401 if (hdl == NULL)
402 sa_handle_destroy(zp->z_sa_hdl);
403
404 goto error;
405 }
406
407 ip->i_ino = obj;
408 ip->i_mode = zp->z_mode;
409 ip->i_mtime = ip->i_atime = ip->i_ctime = CURRENT_TIME_SEC;
410 zfs_inode_set_ops(zsb, ip);
411
412 if (insert_inode_locked(ip))
413 goto error;
414
415 mutex_enter(&zsb->z_znodes_lock);
416 list_insert_tail(&zsb->z_all_znodes, zp);
417 membar_producer();
418 mutex_exit(&zsb->z_znodes_lock);
419
420 unlock_new_inode(ip);
421 return (zp);
422
423 error:
424 unlock_new_inode(ip);
425 iput(ip);
426 return NULL;
427 }
428
429 /*
430 * Update the embedded inode given the znode. We should work toward
431 * eliminating this function as soon as possible by removing values
432 * which are duplicated between the znode and inode. If the generic
433 * inode has the correct field it should be used, and the ZFS code
434 * updated to access the inode. This can be done incrementally.
435 */
436 void
437 zfs_inode_update(znode_t *zp)
438 {
439 zfs_sb_t *zsb;
440 struct inode *ip;
441 uint32_t blksize;
442 uint64_t atime[2], mtime[2], ctime[2];
443
444 ASSERT(zp != NULL);
445 zsb = ZTOZSB(zp);
446 ip = ZTOI(zp);
447
448 sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16);
449 sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16);
450 sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16);
451
452 spin_lock(&ip->i_lock);
453 ip->i_generation = zp->z_gen;
454 ip->i_uid = zp->z_uid;
455 ip->i_gid = zp->z_gid;
456 ip->i_nlink = zp->z_links;
457 ip->i_mode = zp->z_mode;
458 ip->i_blkbits = SPA_MINBLOCKSHIFT;
459 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
460 (u_longlong_t *)&ip->i_blocks);
461
462 ZFS_TIME_DECODE(&ip->i_atime, atime);
463 ZFS_TIME_DECODE(&ip->i_mtime, mtime);
464 ZFS_TIME_DECODE(&ip->i_ctime, ctime);
465
466 i_size_write(ip, zp->z_size);
467 spin_unlock(&ip->i_lock);
468 }
469
470 static uint64_t empty_xattr;
471 static uint64_t pad[4];
472 static zfs_acl_phys_t acl_phys;
473 /*
474 * Create a new DMU object to hold a zfs znode.
475 *
476 * IN: dzp - parent directory for new znode
477 * vap - file attributes for new znode
478 * tx - dmu transaction id for zap operations
479 * cr - credentials of caller
480 * flag - flags:
481 * IS_ROOT_NODE - new object will be root
482 * IS_XATTR - new object is an attribute
483 * bonuslen - length of bonus buffer
484 * setaclp - File/Dir initial ACL
485 * fuidp - Tracks fuid allocation.
486 *
487 * OUT: zpp - allocated znode
488 *
489 */
490 void
491 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
492 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
493 {
494 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
495 uint64_t mode, size, links, parent, pflags;
496 uint64_t dzp_pflags = 0;
497 uint64_t rdev = 0;
498 zfs_sb_t *zsb = ZTOZSB(dzp);
499 dmu_buf_t *db;
500 timestruc_t now;
501 uint64_t gen, obj;
502 int err;
503 int bonuslen;
504 sa_handle_t *sa_hdl;
505 dmu_object_type_t obj_type;
506 sa_bulk_attr_t *sa_attrs;
507 int cnt = 0;
508 zfs_acl_locator_cb_t locate = { 0 };
509
510 if (zsb->z_replay) {
511 obj = vap->va_nodeid;
512 now = vap->va_ctime; /* see zfs_replay_create() */
513 gen = vap->va_nblocks; /* ditto */
514 } else {
515 obj = 0;
516 gethrestime(&now);
517 gen = dmu_tx_get_txg(tx);
518 }
519
520 obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
521 bonuslen = (obj_type == DMU_OT_SA) ?
522 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
523
524 /*
525 * Create a new DMU object.
526 */
527 /*
528 * There's currently no mechanism for pre-reading the blocks that will
529 * be needed to allocate a new object, so we accept the small chance
530 * that there will be an i/o error and we will fail one of the
531 * assertions below.
532 */
533 if (S_ISDIR(vap->va_mode)) {
534 if (zsb->z_replay) {
535 err = zap_create_claim_norm(zsb->z_os, obj,
536 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
537 obj_type, bonuslen, tx);
538 ASSERT3U(err, ==, 0);
539 } else {
540 obj = zap_create_norm(zsb->z_os,
541 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
542 obj_type, bonuslen, tx);
543 }
544 } else {
545 if (zsb->z_replay) {
546 err = dmu_object_claim(zsb->z_os, obj,
547 DMU_OT_PLAIN_FILE_CONTENTS, 0,
548 obj_type, bonuslen, tx);
549 ASSERT3U(err, ==, 0);
550 } else {
551 obj = dmu_object_alloc(zsb->z_os,
552 DMU_OT_PLAIN_FILE_CONTENTS, 0,
553 obj_type, bonuslen, tx);
554 }
555 }
556
557 ZFS_OBJ_HOLD_ENTER(zsb, obj);
558 VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db));
559
560 /*
561 * If this is the root, fix up the half-initialized parent pointer
562 * to reference the just-allocated physical data area.
563 */
564 if (flag & IS_ROOT_NODE) {
565 dzp->z_id = obj;
566 } else {
567 dzp_pflags = dzp->z_pflags;
568 }
569
570 /*
571 * If parent is an xattr, so am I.
572 */
573 if (dzp_pflags & ZFS_XATTR) {
574 flag |= IS_XATTR;
575 }
576
577 if (zsb->z_use_fuids)
578 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
579 else
580 pflags = 0;
581
582 if (S_ISDIR(vap->va_mode)) {
583 size = 2; /* contents ("." and "..") */
584 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
585 } else {
586 size = links = 0;
587 }
588
589 if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
590 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))
591 rdev = zfs_expldev(vap->va_rdev);
592
593 parent = dzp->z_id;
594 mode = acl_ids->z_mode;
595 if (flag & IS_XATTR)
596 pflags |= ZFS_XATTR;
597
598 /*
599 * No execs denied will be deterimed when zfs_mode_compute() is called.
600 */
601 pflags |= acl_ids->z_aclp->z_hints &
602 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
603 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
604
605 ZFS_TIME_ENCODE(&now, crtime);
606 ZFS_TIME_ENCODE(&now, ctime);
607
608 if (vap->va_mask & ATTR_ATIME) {
609 ZFS_TIME_ENCODE(&vap->va_atime, atime);
610 } else {
611 ZFS_TIME_ENCODE(&now, atime);
612 }
613
614 if (vap->va_mask & ATTR_MTIME) {
615 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
616 } else {
617 ZFS_TIME_ENCODE(&now, mtime);
618 }
619
620 /* Now add in all of the "SA" attributes */
621 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED,
622 &sa_hdl));
623
624 /*
625 * Setup the array of attributes to be replaced/set on the new file
626 *
627 * order for DMU_OT_ZNODE is critical since it needs to be constructed
628 * in the old znode_phys_t format. Don't change this ordering
629 */
630 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
631
632 if (obj_type == DMU_OT_ZNODE) {
633 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
634 NULL, &atime, 16);
635 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
636 NULL, &mtime, 16);
637 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
638 NULL, &ctime, 16);
639 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
640 NULL, &crtime, 16);
641 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
642 NULL, &gen, 8);
643 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
644 NULL, &mode, 8);
645 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
646 NULL, &size, 8);
647 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
648 NULL, &parent, 8);
649 } else {
650 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
651 NULL, &mode, 8);
652 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
653 NULL, &size, 8);
654 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
655 NULL, &gen, 8);
656 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb),
657 NULL, &acl_ids->z_fuid, 8);
658 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb),
659 NULL, &acl_ids->z_fgid, 8);
660 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
661 NULL, &parent, 8);
662 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
663 NULL, &pflags, 8);
664 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
665 NULL, &atime, 16);
666 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
667 NULL, &mtime, 16);
668 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
669 NULL, &ctime, 16);
670 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
671 NULL, &crtime, 16);
672 }
673
674 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8);
675
676 if (obj_type == DMU_OT_ZNODE) {
677 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL,
678 &empty_xattr, 8);
679 }
680 if (obj_type == DMU_OT_ZNODE ||
681 (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
682 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))) {
683 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb),
684 NULL, &rdev, 8);
685 }
686 if (obj_type == DMU_OT_ZNODE) {
687 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
688 NULL, &pflags, 8);
689 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL,
690 &acl_ids->z_fuid, 8);
691 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL,
692 &acl_ids->z_fgid, 8);
693 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad,
694 sizeof (uint64_t) * 4);
695 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL,
696 &acl_phys, sizeof (zfs_acl_phys_t));
697 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
698 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL,
699 &acl_ids->z_aclp->z_acl_count, 8);
700 locate.cb_aclp = acl_ids->z_aclp;
701 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb),
702 zfs_acl_data_locator, &locate,
703 acl_ids->z_aclp->z_acl_bytes);
704 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
705 acl_ids->z_fuid, acl_ids->z_fgid);
706 }
707
708 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
709
710 if (!(flag & IS_ROOT_NODE)) {
711 *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl);
712 ASSERT(*zpp != NULL);
713 ASSERT(dzp != NULL);
714 err = zpl_xattr_security_init(ZTOI(*zpp), ZTOI(dzp));
715 ASSERT3S(err, ==, 0);
716 } else {
717 /*
718 * If we are creating the root node, the "parent" we
719 * passed in is the znode for the root.
720 */
721 *zpp = dzp;
722
723 (*zpp)->z_sa_hdl = sa_hdl;
724 }
725
726 (*zpp)->z_pflags = pflags;
727 (*zpp)->z_mode = mode;
728
729 if (obj_type == DMU_OT_ZNODE ||
730 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
731 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
732 ASSERT3S(err, ==, 0);
733 }
734 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
735 ZFS_OBJ_HOLD_EXIT(zsb, obj);
736 }
737
738 int
739 zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp)
740 {
741 dmu_object_info_t doi;
742 dmu_buf_t *db;
743 znode_t *zp;
744 int err;
745 sa_handle_t *hdl;
746
747 *zpp = NULL;
748
749 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
750
751 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
752 if (err) {
753 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
754 return (err);
755 }
756
757 dmu_object_info_from_db(db, &doi);
758 if (doi.doi_bonus_type != DMU_OT_SA &&
759 (doi.doi_bonus_type != DMU_OT_ZNODE ||
760 (doi.doi_bonus_type == DMU_OT_ZNODE &&
761 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
762 sa_buf_rele(db, NULL);
763 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
764 return (EINVAL);
765 }
766
767 hdl = dmu_buf_get_user(db);
768 if (hdl != NULL) {
769 zp = sa_get_userdata(hdl);
770
771
772 /*
773 * Since "SA" does immediate eviction we
774 * should never find a sa handle that doesn't
775 * know about the znode.
776 */
777
778 ASSERT3P(zp, !=, NULL);
779
780 mutex_enter(&zp->z_lock);
781 ASSERT3U(zp->z_id, ==, obj_num);
782 if (zp->z_unlinked) {
783 err = ENOENT;
784 } else {
785 igrab(ZTOI(zp));
786 *zpp = zp;
787 err = 0;
788 }
789 sa_buf_rele(db, NULL);
790 mutex_exit(&zp->z_lock);
791 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
792 return (err);
793 }
794
795 /*
796 * Not found create new znode/vnode but only if file exists.
797 *
798 * There is a small window where zfs_vget() could
799 * find this object while a file create is still in
800 * progress. This is checked for in zfs_znode_alloc()
801 *
802 * if zfs_znode_alloc() fails it will drop the hold on the
803 * bonus buffer.
804 */
805 zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size,
806 doi.doi_bonus_type, obj_num, NULL);
807 if (zp == NULL) {
808 err = ENOENT;
809 } else {
810 *zpp = zp;
811 }
812 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
813 return (err);
814 }
815
816 int
817 zfs_rezget(znode_t *zp)
818 {
819 zfs_sb_t *zsb = ZTOZSB(zp);
820 dmu_object_info_t doi;
821 dmu_buf_t *db;
822 uint64_t obj_num = zp->z_id;
823 uint64_t mode;
824 sa_bulk_attr_t bulk[8];
825 int err;
826 int count = 0;
827 uint64_t gen;
828
829 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
830
831 mutex_enter(&zp->z_acl_lock);
832 if (zp->z_acl_cached) {
833 zfs_acl_free(zp->z_acl_cached);
834 zp->z_acl_cached = NULL;
835 }
836
837 mutex_exit(&zp->z_acl_lock);
838 ASSERT(zp->z_sa_hdl == NULL);
839 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
840 if (err) {
841 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
842 return (err);
843 }
844
845 dmu_object_info_from_db(db, &doi);
846 if (doi.doi_bonus_type != DMU_OT_SA &&
847 (doi.doi_bonus_type != DMU_OT_ZNODE ||
848 (doi.doi_bonus_type == DMU_OT_ZNODE &&
849 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
850 sa_buf_rele(db, NULL);
851 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
852 return (EINVAL);
853 }
854
855 zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL);
856
857 /* reload cached values */
858 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL,
859 &gen, sizeof (gen));
860 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL,
861 &zp->z_size, sizeof (zp->z_size));
862 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL,
863 &zp->z_links, sizeof (zp->z_links));
864 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
865 &zp->z_pflags, sizeof (zp->z_pflags));
866 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
867 &zp->z_atime, sizeof (zp->z_atime));
868 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL,
869 &zp->z_uid, sizeof (zp->z_uid));
870 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL,
871 &zp->z_gid, sizeof (zp->z_gid));
872 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
873 &mode, sizeof (mode));
874
875 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
876 zfs_znode_dmu_fini(zp);
877 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
878 return (EIO);
879 }
880
881 zp->z_mode = mode;
882
883 if (gen != zp->z_gen) {
884 zfs_znode_dmu_fini(zp);
885 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
886 return (EIO);
887 }
888
889 zp->z_unlinked = (zp->z_links == 0);
890 zp->z_blksz = doi.doi_data_block_size;
891
892 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
893
894 return (0);
895 }
896
897 void
898 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
899 {
900 zfs_sb_t *zsb = ZTOZSB(zp);
901 objset_t *os = zsb->z_os;
902 uint64_t obj = zp->z_id;
903 uint64_t acl_obj = zfs_external_acl(zp);
904
905 ZFS_OBJ_HOLD_ENTER(zsb, obj);
906 if (acl_obj) {
907 VERIFY(!zp->z_is_sa);
908 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
909 }
910 VERIFY(0 == dmu_object_free(os, obj, tx));
911 zfs_znode_dmu_fini(zp);
912 ZFS_OBJ_HOLD_EXIT(zsb, obj);
913 }
914
915 void
916 zfs_zinactive(znode_t *zp)
917 {
918 zfs_sb_t *zsb = ZTOZSB(zp);
919 uint64_t z_id = zp->z_id;
920
921 ASSERT(zp->z_sa_hdl);
922
923 /*
924 * Don't allow a zfs_zget() while were trying to release this znode
925 */
926 ZFS_OBJ_HOLD_ENTER(zsb, z_id);
927 mutex_enter(&zp->z_lock);
928
929 /*
930 * If this was the last reference to a file with no links,
931 * remove the file from the file system.
932 */
933 if (zp->z_unlinked) {
934 mutex_exit(&zp->z_lock);
935 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
936 zfs_rmnode(zp);
937 return;
938 }
939
940 mutex_exit(&zp->z_lock);
941 zfs_znode_dmu_fini(zp);
942 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
943 }
944
945 void
946 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
947 uint64_t ctime[2], boolean_t have_tx)
948 {
949 timestruc_t now;
950
951 gethrestime(&now);
952
953 if (have_tx) { /* will sa_bulk_update happen really soon? */
954 zp->z_atime_dirty = 0;
955 zp->z_seq++;
956 } else {
957 zp->z_atime_dirty = 1;
958 }
959
960 if (flag & ATTR_ATIME) {
961 ZFS_TIME_ENCODE(&now, zp->z_atime);
962 }
963
964 if (flag & ATTR_MTIME) {
965 ZFS_TIME_ENCODE(&now, mtime);
966 if (ZTOZSB(zp)->z_use_fuids) {
967 zp->z_pflags |= (ZFS_ARCHIVE |
968 ZFS_AV_MODIFIED);
969 }
970 }
971
972 if (flag & ATTR_CTIME) {
973 ZFS_TIME_ENCODE(&now, ctime);
974 if (ZTOZSB(zp)->z_use_fuids)
975 zp->z_pflags |= ZFS_ARCHIVE;
976 }
977 }
978
979 /*
980 * Grow the block size for a file.
981 *
982 * IN: zp - znode of file to free data in.
983 * size - requested block size
984 * tx - open transaction.
985 *
986 * NOTE: this function assumes that the znode is write locked.
987 */
988 void
989 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
990 {
991 int error;
992 u_longlong_t dummy;
993
994 if (size <= zp->z_blksz)
995 return;
996 /*
997 * If the file size is already greater than the current blocksize,
998 * we will not grow. If there is more than one block in a file,
999 * the blocksize cannot change.
1000 */
1001 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1002 return;
1003
1004 error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id,
1005 size, 0, tx);
1006
1007 if (error == ENOTSUP)
1008 return;
1009 ASSERT3U(error, ==, 0);
1010
1011 /* What blocksize did we actually get? */
1012 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1013 }
1014
1015 #ifdef HAVE_MMAP
1016 /*
1017 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1018 * be calling back into the fs for a putpage(). E.g.: when truncating
1019 * a file, the pages being "thrown away* don't need to be written out.
1020 */
1021 /* ARGSUSED */
1022 static int
1023 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1024 int flags, cred_t *cr)
1025 {
1026 ASSERT(0);
1027 return (0);
1028 }
1029 #endif /* HAVE_MMAP */
1030
1031 /*
1032 * Increase the file length
1033 *
1034 * IN: zp - znode of file to free data in.
1035 * end - new end-of-file
1036 *
1037 * RETURN: 0 if success
1038 * error code if failure
1039 */
1040 static int
1041 zfs_extend(znode_t *zp, uint64_t end)
1042 {
1043 zfs_sb_t *zsb = ZTOZSB(zp);
1044 dmu_tx_t *tx;
1045 rl_t *rl;
1046 uint64_t newblksz;
1047 int error;
1048
1049 /*
1050 * We will change zp_size, lock the whole file.
1051 */
1052 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1053
1054 /*
1055 * Nothing to do if file already at desired length.
1056 */
1057 if (end <= zp->z_size) {
1058 zfs_range_unlock(rl);
1059 return (0);
1060 }
1061 top:
1062 tx = dmu_tx_create(zsb->z_os);
1063 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1064 zfs_sa_upgrade_txholds(tx, zp);
1065 if (end > zp->z_blksz &&
1066 (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) {
1067 /*
1068 * We are growing the file past the current block size.
1069 */
1070 if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
1071 ASSERT(!ISP2(zp->z_blksz));
1072 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1073 } else {
1074 newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
1075 }
1076 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1077 } else {
1078 newblksz = 0;
1079 }
1080
1081 error = dmu_tx_assign(tx, TXG_NOWAIT);
1082 if (error) {
1083 if (error == ERESTART) {
1084 dmu_tx_wait(tx);
1085 dmu_tx_abort(tx);
1086 goto top;
1087 }
1088 dmu_tx_abort(tx);
1089 zfs_range_unlock(rl);
1090 return (error);
1091 }
1092
1093 if (newblksz)
1094 zfs_grow_blocksize(zp, newblksz, tx);
1095
1096 zp->z_size = end;
1097
1098 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)),
1099 &zp->z_size, sizeof (zp->z_size), tx));
1100
1101 zfs_range_unlock(rl);
1102
1103 dmu_tx_commit(tx);
1104
1105 return (0);
1106 }
1107
1108 /*
1109 * Free space in a file.
1110 *
1111 * IN: zp - znode of file to free data in.
1112 * off - start of section to free.
1113 * len - length of section to free.
1114 *
1115 * RETURN: 0 if success
1116 * error code if failure
1117 */
1118 static int
1119 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1120 {
1121 zfs_sb_t *zsb = ZTOZSB(zp);
1122 rl_t *rl;
1123 int error;
1124
1125 /*
1126 * Lock the range being freed.
1127 */
1128 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1129
1130 /*
1131 * Nothing to do if file already at desired length.
1132 */
1133 if (off >= zp->z_size) {
1134 zfs_range_unlock(rl);
1135 return (0);
1136 }
1137
1138 if (off + len > zp->z_size)
1139 len = zp->z_size - off;
1140
1141 error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len);
1142
1143 zfs_range_unlock(rl);
1144
1145 return (error);
1146 }
1147
1148 /*
1149 * Truncate a file
1150 *
1151 * IN: zp - znode of file to free data in.
1152 * end - new end-of-file.
1153 *
1154 * RETURN: 0 if success
1155 * error code if failure
1156 */
1157 static int
1158 zfs_trunc(znode_t *zp, uint64_t end)
1159 {
1160 zfs_sb_t *zsb = ZTOZSB(zp);
1161 dmu_tx_t *tx;
1162 rl_t *rl;
1163 int error;
1164 sa_bulk_attr_t bulk[2];
1165 int count = 0;
1166
1167 /*
1168 * We will change zp_size, lock the whole file.
1169 */
1170 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1171
1172 /*
1173 * Nothing to do if file already at desired length.
1174 */
1175 if (end >= zp->z_size) {
1176 zfs_range_unlock(rl);
1177 return (0);
1178 }
1179
1180 error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1);
1181 if (error) {
1182 zfs_range_unlock(rl);
1183 return (error);
1184 }
1185 top:
1186 tx = dmu_tx_create(zsb->z_os);
1187 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1188 zfs_sa_upgrade_txholds(tx, zp);
1189 error = dmu_tx_assign(tx, TXG_NOWAIT);
1190 if (error) {
1191 if (error == ERESTART) {
1192 dmu_tx_wait(tx);
1193 dmu_tx_abort(tx);
1194 goto top;
1195 }
1196 dmu_tx_abort(tx);
1197 zfs_range_unlock(rl);
1198 return (error);
1199 }
1200
1201 zp->z_size = end;
1202 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb),
1203 NULL, &zp->z_size, sizeof (zp->z_size));
1204
1205 if (end == 0) {
1206 zp->z_pflags &= ~ZFS_SPARSE;
1207 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1208 NULL, &zp->z_pflags, 8);
1209 }
1210 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1211
1212 dmu_tx_commit(tx);
1213
1214 zfs_range_unlock(rl);
1215
1216 return (0);
1217 }
1218
1219 /*
1220 * Free space in a file
1221 *
1222 * IN: zp - znode of file to free data in.
1223 * off - start of range
1224 * len - end of range (0 => EOF)
1225 * flag - current file open mode flags.
1226 * log - TRUE if this action should be logged
1227 *
1228 * RETURN: 0 if success
1229 * error code if failure
1230 */
1231 int
1232 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1233 {
1234 #ifdef HAVE_MANDLOCKS
1235 struct inode *ip = ZTOI(zp);
1236 #endif /* HAVE_MANDLOCKS */
1237 dmu_tx_t *tx;
1238 zfs_sb_t *zsb = ZTOZSB(zp);
1239 zilog_t *zilog = zsb->z_log;
1240 uint64_t mode;
1241 uint64_t mtime[2], ctime[2];
1242 sa_bulk_attr_t bulk[3];
1243 int count = 0;
1244 int error;
1245
1246 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode,
1247 sizeof (mode))) != 0)
1248 return (error);
1249
1250 if (off > zp->z_size) {
1251 error = zfs_extend(zp, off+len);
1252 if (error == 0 && log)
1253 goto log;
1254 else
1255 return (error);
1256 }
1257
1258 #ifdef HAVE_MANDLOCKS
1259 /*
1260 * Check for any locks in the region to be freed.
1261 */
1262
1263 if (MANDLOCK(ip, (mode_t)mode)) {
1264 uint64_t length = (len ? len : zp->z_size - off);
1265 if (error = chklock(ip, FWRITE, off, length, flag, NULL))
1266 return (error);
1267 }
1268 #endif /* HAVE_MANDLOCKS */
1269
1270 if (len == 0) {
1271 error = zfs_trunc(zp, off);
1272 } else {
1273 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1274 off + len > zp->z_size)
1275 error = zfs_extend(zp, off+len);
1276 }
1277 if (error || !log)
1278 return (error);
1279 log:
1280 tx = dmu_tx_create(zsb->z_os);
1281 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1282 zfs_sa_upgrade_txholds(tx, zp);
1283 error = dmu_tx_assign(tx, TXG_NOWAIT);
1284 if (error) {
1285 if (error == ERESTART) {
1286 dmu_tx_wait(tx);
1287 dmu_tx_abort(tx);
1288 goto log;
1289 }
1290 dmu_tx_abort(tx);
1291 return (error);
1292 }
1293
1294 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16);
1295 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16);
1296 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1297 NULL, &zp->z_pflags, 8);
1298 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1299 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1300 ASSERT(error == 0);
1301
1302 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1303
1304 dmu_tx_commit(tx);
1305 zfs_inode_update(zp);
1306 return (0);
1307 }
1308
1309 void
1310 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1311 {
1312 uint64_t moid, obj, sa_obj, version;
1313 uint64_t norm = 0;
1314 nvpair_t *elem;
1315 int error;
1316 timestruc_t now;
1317 dmu_buf_t *db;
1318 znode_phys_t *pzp;
1319
1320 /*
1321 * First attempt to create master node.
1322 */
1323 /*
1324 * In an empty objset, there are no blocks to read and thus
1325 * there can be no i/o errors (which we assert below).
1326 */
1327 moid = MASTER_NODE_OBJ;
1328 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1329 DMU_OT_NONE, 0, tx);
1330 ASSERT(error == 0);
1331
1332 /*
1333 * Set starting attributes.
1334 */
1335 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1336 elem = NULL;
1337 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1338 /* For the moment we expect all zpl props to be uint64_ts */
1339 uint64_t val;
1340 char *name;
1341
1342 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1343 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1344 name = nvpair_name(elem);
1345 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1346 if (val < version)
1347 version = val;
1348 } else {
1349 error = zap_update(os, moid, name, 8, 1, &val, tx);
1350 }
1351 ASSERT(error == 0);
1352 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1353 norm = val;
1354 }
1355 ASSERT(version != 0);
1356 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1357
1358 /*
1359 * Create zap object used for SA attribute registration
1360 */
1361
1362 if (version >= ZPL_VERSION_SA) {
1363 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1364 DMU_OT_NONE, 0, tx);
1365 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1366 ASSERT(error == 0);
1367 } else {
1368 sa_obj = 0;
1369 }
1370 /*
1371 * Create a delete queue.
1372 */
1373 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1374
1375 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1376 ASSERT(error == 0);
1377
1378 /*
1379 * Create root znode with code free of VFS dependencies. This
1380 * is important because without a registered filesystem and super
1381 * block all the required VFS hooks will be missing. The critical
1382 * thing is to just crete the required root znode.
1383 */
1384 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1385 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1386
1387 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1388 dmu_buf_will_dirty(db, tx);
1389
1390 /*
1391 * Initialize the znode physical data to zero.
1392 */
1393 ASSERT(db->db_size >= sizeof (znode_phys_t));
1394 bzero(db->db_data, db->db_size);
1395 pzp = db->db_data;
1396
1397 if (USE_FUIDS(version, os))
1398 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1399
1400 pzp->zp_size = 2; /* "." and ".." */
1401 pzp->zp_links = 2;
1402 pzp->zp_parent = obj;
1403 pzp->zp_gen = dmu_tx_get_txg(tx);
1404 pzp->zp_mode = S_IFDIR | 0755;
1405 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1406
1407 gethrestime(&now);
1408
1409 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1410 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1411 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1412 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1413
1414 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1415 ASSERT(error == 0);
1416
1417 dmu_buf_rele(db, FTAG);
1418 }
1419
1420 #endif /* _KERNEL */
1421
1422 static int
1423 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1424 {
1425 uint64_t sa_obj = 0;
1426 int error;
1427
1428 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1429 if (error != 0 && error != ENOENT)
1430 return (error);
1431
1432 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1433 return (error);
1434 }
1435
1436 static int
1437 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1438 dmu_buf_t **db)
1439 {
1440 dmu_object_info_t doi;
1441 int error;
1442
1443 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1444 return (error);
1445
1446 dmu_object_info_from_db(*db, &doi);
1447 if ((doi.doi_bonus_type != DMU_OT_SA &&
1448 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1449 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1450 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1451 sa_buf_rele(*db, FTAG);
1452 return (ENOTSUP);
1453 }
1454
1455 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1456 if (error != 0) {
1457 sa_buf_rele(*db, FTAG);
1458 return (error);
1459 }
1460
1461 return (0);
1462 }
1463
1464 void
1465 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1466 {
1467 sa_handle_destroy(hdl);
1468 sa_buf_rele(db, FTAG);
1469 }
1470
1471 /*
1472 * Given an object number, return its parent object number and whether
1473 * or not the object is an extended attribute directory.
1474 */
1475 static int
1476 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1477 int *is_xattrdir)
1478 {
1479 uint64_t parent;
1480 uint64_t pflags;
1481 uint64_t mode;
1482 sa_bulk_attr_t bulk[3];
1483 int count = 0;
1484 int error;
1485
1486 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1487 &parent, sizeof (parent));
1488 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1489 &pflags, sizeof (pflags));
1490 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1491 &mode, sizeof (mode));
1492
1493 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1494 return (error);
1495
1496 *pobjp = parent;
1497 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1498
1499 return (0);
1500 }
1501
1502 /*
1503 * Given an object number, return some zpl level statistics
1504 */
1505 static int
1506 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1507 zfs_stat_t *sb)
1508 {
1509 sa_bulk_attr_t bulk[4];
1510 int count = 0;
1511
1512 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1513 &sb->zs_mode, sizeof (sb->zs_mode));
1514 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1515 &sb->zs_gen, sizeof (sb->zs_gen));
1516 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1517 &sb->zs_links, sizeof (sb->zs_links));
1518 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1519 &sb->zs_ctime, sizeof (sb->zs_ctime));
1520
1521 return (sa_bulk_lookup(hdl, bulk, count));
1522 }
1523
1524 static int
1525 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1526 sa_attr_type_t *sa_table, char *buf, int len)
1527 {
1528 sa_handle_t *sa_hdl;
1529 sa_handle_t *prevhdl = NULL;
1530 dmu_buf_t *prevdb = NULL;
1531 dmu_buf_t *sa_db = NULL;
1532 char *path = buf + len - 1;
1533 int error;
1534
1535 *path = '\0';
1536 sa_hdl = hdl;
1537
1538 for (;;) {
1539 uint64_t pobj;
1540 char component[MAXNAMELEN + 2];
1541 size_t complen;
1542 int is_xattrdir;
1543
1544 if (prevdb)
1545 zfs_release_sa_handle(prevhdl, prevdb);
1546
1547 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1548 &is_xattrdir)) != 0)
1549 break;
1550
1551 if (pobj == obj) {
1552 if (path[0] != '/')
1553 *--path = '/';
1554 break;
1555 }
1556
1557 component[0] = '/';
1558 if (is_xattrdir) {
1559 (void) sprintf(component + 1, "<xattrdir>");
1560 } else {
1561 error = zap_value_search(osp, pobj, obj,
1562 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1563 if (error != 0)
1564 break;
1565 }
1566
1567 complen = strlen(component);
1568 path -= complen;
1569 ASSERT(path >= buf);
1570 bcopy(component, path, complen);
1571 obj = pobj;
1572
1573 if (sa_hdl != hdl) {
1574 prevhdl = sa_hdl;
1575 prevdb = sa_db;
1576 }
1577 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1578 if (error != 0) {
1579 sa_hdl = prevhdl;
1580 sa_db = prevdb;
1581 break;
1582 }
1583 }
1584
1585 if (sa_hdl != NULL && sa_hdl != hdl) {
1586 ASSERT(sa_db != NULL);
1587 zfs_release_sa_handle(sa_hdl, sa_db);
1588 }
1589
1590 if (error == 0)
1591 (void) memmove(buf, path, buf + len - path);
1592
1593 return (error);
1594 }
1595
1596 int
1597 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1598 {
1599 sa_attr_type_t *sa_table;
1600 sa_handle_t *hdl;
1601 dmu_buf_t *db;
1602 int error;
1603
1604 error = zfs_sa_setup(osp, &sa_table);
1605 if (error != 0)
1606 return (error);
1607
1608 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1609 if (error != 0)
1610 return (error);
1611
1612 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1613
1614 zfs_release_sa_handle(hdl, db);
1615 return (error);
1616 }
1617
1618 int
1619 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1620 char *buf, int len)
1621 {
1622 char *path = buf + len - 1;
1623 sa_attr_type_t *sa_table;
1624 sa_handle_t *hdl;
1625 dmu_buf_t *db;
1626 int error;
1627
1628 *path = '\0';
1629
1630 error = zfs_sa_setup(osp, &sa_table);
1631 if (error != 0)
1632 return (error);
1633
1634 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1635 if (error != 0)
1636 return (error);
1637
1638 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1639 if (error != 0) {
1640 zfs_release_sa_handle(hdl, db);
1641 return (error);
1642 }
1643
1644 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1645
1646 zfs_release_sa_handle(hdl, db);
1647 return (error);
1648 }
1649
1650 #if defined(_KERNEL) && defined(HAVE_SPL)
1651 EXPORT_SYMBOL(zfs_create_fs);
1652 EXPORT_SYMBOL(zfs_obj_to_path);
1653 #endif
ZFS On Linux mirror for PVE