]> 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
Use KM_PUSHPAGE instead of KM_SLEEP
[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_PUSHPAGE);
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_PUSHPAGE);
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 zp->z_is_zvol = 0;
351
352 zfs_znode_sa_init(zsb, zp, db, obj_type, hdl);
353
354 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &zp->z_mode, 8);
355 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8);
356 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8);
357 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8);
358 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
359 &zp->z_pflags, 8);
360 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL,
361 &parent, 8);
362 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
363 &zp->z_atime, 16);
364 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8);
365 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8);
366
367 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) {
368 if (hdl == NULL)
369 sa_handle_destroy(zp->z_sa_hdl);
370
371 goto error;
372 }
373
374 ip->i_ino = obj;
375 ip->i_mode = zp->z_mode;
376 ip->i_mtime = ip->i_atime = ip->i_ctime = CURRENT_TIME_SEC;
377 zfs_inode_set_ops(zsb, ip);
378
379 if (insert_inode_locked(ip))
380 goto error;
381
382 mutex_enter(&zsb->z_znodes_lock);
383 list_insert_tail(&zsb->z_all_znodes, zp);
384 membar_producer();
385 mutex_exit(&zsb->z_znodes_lock);
386
387 unlock_new_inode(ip);
388 return (zp);
389
390 error:
391 unlock_new_inode(ip);
392 iput(ip);
393 return NULL;
394 }
395
396 /*
397 * Update the embedded inode given the znode. We should work toward
398 * eliminating this function as soon as possible by removing values
399 * which are duplicated between the znode and inode. If the generic
400 * inode has the correct field it should be used, and the ZFS code
401 * updated to access the inode. This can be done incrementally.
402 */
403 void
404 zfs_inode_update(znode_t *zp)
405 {
406 zfs_sb_t *zsb;
407 struct inode *ip;
408 uint32_t blksize;
409 uint64_t atime[2], mtime[2], ctime[2];
410
411 ASSERT(zp != NULL);
412 zsb = ZTOZSB(zp);
413 ip = ZTOI(zp);
414
415 sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16);
416 sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16);
417 sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16);
418
419 spin_lock(&ip->i_lock);
420 ip->i_generation = zp->z_gen;
421 ip->i_uid = zp->z_uid;
422 ip->i_gid = zp->z_gid;
423 ip->i_nlink = zp->z_links;
424 ip->i_mode = zp->z_mode;
425 ip->i_blkbits = SPA_MINBLOCKSHIFT;
426 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize,
427 (u_longlong_t *)&ip->i_blocks);
428
429 ZFS_TIME_DECODE(&ip->i_atime, atime);
430 ZFS_TIME_DECODE(&ip->i_mtime, mtime);
431 ZFS_TIME_DECODE(&ip->i_ctime, ctime);
432
433 i_size_write(ip, zp->z_size);
434 spin_unlock(&ip->i_lock);
435 }
436
437 static uint64_t empty_xattr;
438 static uint64_t pad[4];
439 static zfs_acl_phys_t acl_phys;
440 /*
441 * Create a new DMU object to hold a zfs znode.
442 *
443 * IN: dzp - parent directory for new znode
444 * vap - file attributes for new znode
445 * tx - dmu transaction id for zap operations
446 * cr - credentials of caller
447 * flag - flags:
448 * IS_ROOT_NODE - new object will be root
449 * IS_XATTR - new object is an attribute
450 * bonuslen - length of bonus buffer
451 * setaclp - File/Dir initial ACL
452 * fuidp - Tracks fuid allocation.
453 *
454 * OUT: zpp - allocated znode
455 *
456 */
457 void
458 zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr,
459 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids)
460 {
461 uint64_t crtime[2], atime[2], mtime[2], ctime[2];
462 uint64_t mode, size, links, parent, pflags;
463 uint64_t dzp_pflags = 0;
464 uint64_t rdev = 0;
465 zfs_sb_t *zsb = ZTOZSB(dzp);
466 dmu_buf_t *db;
467 timestruc_t now;
468 uint64_t gen, obj;
469 int err;
470 int bonuslen;
471 sa_handle_t *sa_hdl;
472 dmu_object_type_t obj_type;
473 sa_bulk_attr_t *sa_attrs;
474 int cnt = 0;
475 zfs_acl_locator_cb_t locate = { 0 };
476
477 if (zsb->z_replay) {
478 obj = vap->va_nodeid;
479 now = vap->va_ctime; /* see zfs_replay_create() */
480 gen = vap->va_nblocks; /* ditto */
481 } else {
482 obj = 0;
483 gethrestime(&now);
484 gen = dmu_tx_get_txg(tx);
485 }
486
487 obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE;
488 bonuslen = (obj_type == DMU_OT_SA) ?
489 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE;
490
491 /*
492 * Create a new DMU object.
493 */
494 /*
495 * There's currently no mechanism for pre-reading the blocks that will
496 * be needed to allocate a new object, so we accept the small chance
497 * that there will be an i/o error and we will fail one of the
498 * assertions below.
499 */
500 if (S_ISDIR(vap->va_mode)) {
501 if (zsb->z_replay) {
502 err = zap_create_claim_norm(zsb->z_os, obj,
503 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
504 obj_type, bonuslen, tx);
505 ASSERT3U(err, ==, 0);
506 } else {
507 obj = zap_create_norm(zsb->z_os,
508 zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS,
509 obj_type, bonuslen, tx);
510 }
511 } else {
512 if (zsb->z_replay) {
513 err = dmu_object_claim(zsb->z_os, obj,
514 DMU_OT_PLAIN_FILE_CONTENTS, 0,
515 obj_type, bonuslen, tx);
516 ASSERT3U(err, ==, 0);
517 } else {
518 obj = dmu_object_alloc(zsb->z_os,
519 DMU_OT_PLAIN_FILE_CONTENTS, 0,
520 obj_type, bonuslen, tx);
521 }
522 }
523
524 ZFS_OBJ_HOLD_ENTER(zsb, obj);
525 VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db));
526
527 /*
528 * If this is the root, fix up the half-initialized parent pointer
529 * to reference the just-allocated physical data area.
530 */
531 if (flag & IS_ROOT_NODE) {
532 dzp->z_id = obj;
533 } else {
534 dzp_pflags = dzp->z_pflags;
535 }
536
537 /*
538 * If parent is an xattr, so am I.
539 */
540 if (dzp_pflags & ZFS_XATTR) {
541 flag |= IS_XATTR;
542 }
543
544 if (zsb->z_use_fuids)
545 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
546 else
547 pflags = 0;
548
549 if (S_ISDIR(vap->va_mode)) {
550 size = 2; /* contents ("." and "..") */
551 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1;
552 } else {
553 size = links = 0;
554 }
555
556 if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
557 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))
558 rdev = vap->va_rdev;
559
560 parent = dzp->z_id;
561 mode = acl_ids->z_mode;
562 if (flag & IS_XATTR)
563 pflags |= ZFS_XATTR;
564
565 /*
566 * No execs denied will be deterimed when zfs_mode_compute() is called.
567 */
568 pflags |= acl_ids->z_aclp->z_hints &
569 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT|
570 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED);
571
572 ZFS_TIME_ENCODE(&now, crtime);
573 ZFS_TIME_ENCODE(&now, ctime);
574
575 if (vap->va_mask & ATTR_ATIME) {
576 ZFS_TIME_ENCODE(&vap->va_atime, atime);
577 } else {
578 ZFS_TIME_ENCODE(&now, atime);
579 }
580
581 if (vap->va_mask & ATTR_MTIME) {
582 ZFS_TIME_ENCODE(&vap->va_mtime, mtime);
583 } else {
584 ZFS_TIME_ENCODE(&now, mtime);
585 }
586
587 /* Now add in all of the "SA" attributes */
588 VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED,
589 &sa_hdl));
590
591 /*
592 * Setup the array of attributes to be replaced/set on the new file
593 *
594 * order for DMU_OT_ZNODE is critical since it needs to be constructed
595 * in the old znode_phys_t format. Don't change this ordering
596 */
597 sa_attrs = kmem_alloc(sizeof(sa_bulk_attr_t) * ZPL_END, KM_SLEEP);
598
599 if (obj_type == DMU_OT_ZNODE) {
600 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
601 NULL, &atime, 16);
602 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
603 NULL, &mtime, 16);
604 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
605 NULL, &ctime, 16);
606 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
607 NULL, &crtime, 16);
608 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
609 NULL, &gen, 8);
610 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
611 NULL, &mode, 8);
612 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
613 NULL, &size, 8);
614 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
615 NULL, &parent, 8);
616 } else {
617 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb),
618 NULL, &mode, 8);
619 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb),
620 NULL, &size, 8);
621 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb),
622 NULL, &gen, 8);
623 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb),
624 NULL, &acl_ids->z_fuid, 8);
625 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb),
626 NULL, &acl_ids->z_fgid, 8);
627 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb),
628 NULL, &parent, 8);
629 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
630 NULL, &pflags, 8);
631 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb),
632 NULL, &atime, 16);
633 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb),
634 NULL, &mtime, 16);
635 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb),
636 NULL, &ctime, 16);
637 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb),
638 NULL, &crtime, 16);
639 }
640
641 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8);
642
643 if (obj_type == DMU_OT_ZNODE) {
644 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL,
645 &empty_xattr, 8);
646 }
647 if (obj_type == DMU_OT_ZNODE ||
648 (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode) ||
649 S_ISFIFO(vap->va_mode) || S_ISSOCK(vap->va_mode))) {
650 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb),
651 NULL, &rdev, 8);
652 }
653 if (obj_type == DMU_OT_ZNODE) {
654 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb),
655 NULL, &pflags, 8);
656 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL,
657 &acl_ids->z_fuid, 8);
658 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL,
659 &acl_ids->z_fgid, 8);
660 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad,
661 sizeof (uint64_t) * 4);
662 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL,
663 &acl_phys, sizeof (zfs_acl_phys_t));
664 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) {
665 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL,
666 &acl_ids->z_aclp->z_acl_count, 8);
667 locate.cb_aclp = acl_ids->z_aclp;
668 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb),
669 zfs_acl_data_locator, &locate,
670 acl_ids->z_aclp->z_acl_bytes);
671 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags,
672 acl_ids->z_fuid, acl_ids->z_fgid);
673 }
674
675 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0);
676
677 if (!(flag & IS_ROOT_NODE)) {
678 *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl);
679 ASSERT(*zpp != NULL);
680 ASSERT(dzp != NULL);
681 err = zpl_xattr_security_init(ZTOI(*zpp), ZTOI(dzp));
682 ASSERT3S(err, ==, 0);
683 } else {
684 /*
685 * If we are creating the root node, the "parent" we
686 * passed in is the znode for the root.
687 */
688 *zpp = dzp;
689
690 (*zpp)->z_sa_hdl = sa_hdl;
691 }
692
693 (*zpp)->z_pflags = pflags;
694 (*zpp)->z_mode = mode;
695
696 if (obj_type == DMU_OT_ZNODE ||
697 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) {
698 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx);
699 ASSERT3S(err, ==, 0);
700 }
701 kmem_free(sa_attrs, sizeof(sa_bulk_attr_t) * ZPL_END);
702 ZFS_OBJ_HOLD_EXIT(zsb, obj);
703 }
704
705 /*
706 * zfs_xvattr_set only updates the in-core attributes
707 * it is assumed the caller will be doing an sa_bulk_update
708 * to push the changes out
709 */
710 void
711 zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx)
712 {
713 xoptattr_t *xoap;
714
715 xoap = xva_getxoptattr(xvap);
716 ASSERT(xoap);
717
718 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) {
719 uint64_t times[2];
720 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times);
721 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)),
722 &times, sizeof (times), tx);
723 XVA_SET_RTN(xvap, XAT_CREATETIME);
724 }
725 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) {
726 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly,
727 zp->z_pflags, tx);
728 XVA_SET_RTN(xvap, XAT_READONLY);
729 }
730 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) {
731 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden,
732 zp->z_pflags, tx);
733 XVA_SET_RTN(xvap, XAT_HIDDEN);
734 }
735 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) {
736 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system,
737 zp->z_pflags, tx);
738 XVA_SET_RTN(xvap, XAT_SYSTEM);
739 }
740 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) {
741 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive,
742 zp->z_pflags, tx);
743 XVA_SET_RTN(xvap, XAT_ARCHIVE);
744 }
745 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) {
746 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable,
747 zp->z_pflags, tx);
748 XVA_SET_RTN(xvap, XAT_IMMUTABLE);
749 }
750 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) {
751 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink,
752 zp->z_pflags, tx);
753 XVA_SET_RTN(xvap, XAT_NOUNLINK);
754 }
755 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) {
756 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly,
757 zp->z_pflags, tx);
758 XVA_SET_RTN(xvap, XAT_APPENDONLY);
759 }
760 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) {
761 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump,
762 zp->z_pflags, tx);
763 XVA_SET_RTN(xvap, XAT_NODUMP);
764 }
765 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) {
766 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque,
767 zp->z_pflags, tx);
768 XVA_SET_RTN(xvap, XAT_OPAQUE);
769 }
770 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) {
771 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED,
772 xoap->xoa_av_quarantined, zp->z_pflags, tx);
773 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED);
774 }
775 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) {
776 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified,
777 zp->z_pflags, tx);
778 XVA_SET_RTN(xvap, XAT_AV_MODIFIED);
779 }
780 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) {
781 zfs_sa_set_scanstamp(zp, xvap, tx);
782 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP);
783 }
784 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) {
785 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse,
786 zp->z_pflags, tx);
787 XVA_SET_RTN(xvap, XAT_REPARSE);
788 }
789 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) {
790 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline,
791 zp->z_pflags, tx);
792 XVA_SET_RTN(xvap, XAT_OFFLINE);
793 }
794 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) {
795 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse,
796 zp->z_pflags, tx);
797 XVA_SET_RTN(xvap, XAT_SPARSE);
798 }
799 }
800
801 int
802 zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp)
803 {
804 dmu_object_info_t doi;
805 dmu_buf_t *db;
806 znode_t *zp;
807 int err;
808 sa_handle_t *hdl;
809
810 *zpp = NULL;
811
812 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
813
814 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
815 if (err) {
816 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
817 return (err);
818 }
819
820 dmu_object_info_from_db(db, &doi);
821 if (doi.doi_bonus_type != DMU_OT_SA &&
822 (doi.doi_bonus_type != DMU_OT_ZNODE ||
823 (doi.doi_bonus_type == DMU_OT_ZNODE &&
824 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
825 sa_buf_rele(db, NULL);
826 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
827 return (EINVAL);
828 }
829
830 hdl = dmu_buf_get_user(db);
831 if (hdl != NULL) {
832 zp = sa_get_userdata(hdl);
833
834
835 /*
836 * Since "SA" does immediate eviction we
837 * should never find a sa handle that doesn't
838 * know about the znode.
839 */
840
841 ASSERT3P(zp, !=, NULL);
842
843 mutex_enter(&zp->z_lock);
844 ASSERT3U(zp->z_id, ==, obj_num);
845 if (zp->z_unlinked) {
846 err = ENOENT;
847 } else {
848 igrab(ZTOI(zp));
849 *zpp = zp;
850 err = 0;
851 }
852 sa_buf_rele(db, NULL);
853 mutex_exit(&zp->z_lock);
854 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
855 return (err);
856 }
857
858 /*
859 * Not found create new znode/vnode but only if file exists.
860 *
861 * There is a small window where zfs_vget() could
862 * find this object while a file create is still in
863 * progress. This is checked for in zfs_znode_alloc()
864 *
865 * if zfs_znode_alloc() fails it will drop the hold on the
866 * bonus buffer.
867 */
868 zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size,
869 doi.doi_bonus_type, obj_num, NULL);
870 if (zp == NULL) {
871 err = ENOENT;
872 } else {
873 *zpp = zp;
874 }
875 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
876 return (err);
877 }
878
879 int
880 zfs_rezget(znode_t *zp)
881 {
882 zfs_sb_t *zsb = ZTOZSB(zp);
883 dmu_object_info_t doi;
884 dmu_buf_t *db;
885 uint64_t obj_num = zp->z_id;
886 uint64_t mode;
887 sa_bulk_attr_t bulk[8];
888 int err;
889 int count = 0;
890 uint64_t gen;
891
892 ZFS_OBJ_HOLD_ENTER(zsb, obj_num);
893
894 mutex_enter(&zp->z_acl_lock);
895 if (zp->z_acl_cached) {
896 zfs_acl_free(zp->z_acl_cached);
897 zp->z_acl_cached = NULL;
898 }
899
900 mutex_exit(&zp->z_acl_lock);
901 ASSERT(zp->z_sa_hdl == NULL);
902 err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db);
903 if (err) {
904 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
905 return (err);
906 }
907
908 dmu_object_info_from_db(db, &doi);
909 if (doi.doi_bonus_type != DMU_OT_SA &&
910 (doi.doi_bonus_type != DMU_OT_ZNODE ||
911 (doi.doi_bonus_type == DMU_OT_ZNODE &&
912 doi.doi_bonus_size < sizeof (znode_phys_t)))) {
913 sa_buf_rele(db, NULL);
914 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
915 return (EINVAL);
916 }
917
918 zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL);
919
920 /* reload cached values */
921 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL,
922 &gen, sizeof (gen));
923 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL,
924 &zp->z_size, sizeof (zp->z_size));
925 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL,
926 &zp->z_links, sizeof (zp->z_links));
927 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL,
928 &zp->z_pflags, sizeof (zp->z_pflags));
929 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL,
930 &zp->z_atime, sizeof (zp->z_atime));
931 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL,
932 &zp->z_uid, sizeof (zp->z_uid));
933 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL,
934 &zp->z_gid, sizeof (zp->z_gid));
935 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL,
936 &mode, sizeof (mode));
937
938 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) {
939 zfs_znode_dmu_fini(zp);
940 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
941 return (EIO);
942 }
943
944 zp->z_mode = mode;
945
946 if (gen != zp->z_gen) {
947 zfs_znode_dmu_fini(zp);
948 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
949 return (EIO);
950 }
951
952 zp->z_unlinked = (zp->z_links == 0);
953 zp->z_blksz = doi.doi_data_block_size;
954
955 ZFS_OBJ_HOLD_EXIT(zsb, obj_num);
956
957 return (0);
958 }
959
960 void
961 zfs_znode_delete(znode_t *zp, dmu_tx_t *tx)
962 {
963 zfs_sb_t *zsb = ZTOZSB(zp);
964 objset_t *os = zsb->z_os;
965 uint64_t obj = zp->z_id;
966 uint64_t acl_obj = zfs_external_acl(zp);
967
968 ZFS_OBJ_HOLD_ENTER(zsb, obj);
969 if (acl_obj) {
970 VERIFY(!zp->z_is_sa);
971 VERIFY(0 == dmu_object_free(os, acl_obj, tx));
972 }
973 VERIFY(0 == dmu_object_free(os, obj, tx));
974 zfs_znode_dmu_fini(zp);
975 ZFS_OBJ_HOLD_EXIT(zsb, obj);
976 }
977
978 void
979 zfs_zinactive(znode_t *zp)
980 {
981 zfs_sb_t *zsb = ZTOZSB(zp);
982 uint64_t z_id = zp->z_id;
983
984 ASSERT(zp->z_sa_hdl);
985
986 /*
987 * Don't allow a zfs_zget() while were trying to release this znode
988 */
989 ZFS_OBJ_HOLD_ENTER(zsb, z_id);
990 mutex_enter(&zp->z_lock);
991
992 /*
993 * If this was the last reference to a file with no links,
994 * remove the file from the file system.
995 */
996 if (zp->z_unlinked) {
997 mutex_exit(&zp->z_lock);
998 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
999 zfs_rmnode(zp);
1000 return;
1001 }
1002
1003 mutex_exit(&zp->z_lock);
1004 zfs_znode_dmu_fini(zp);
1005 ZFS_OBJ_HOLD_EXIT(zsb, z_id);
1006 }
1007
1008 void
1009 zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2],
1010 uint64_t ctime[2], boolean_t have_tx)
1011 {
1012 timestruc_t now;
1013
1014 gethrestime(&now);
1015
1016 if (have_tx) { /* will sa_bulk_update happen really soon? */
1017 zp->z_atime_dirty = 0;
1018 zp->z_seq++;
1019 } else {
1020 zp->z_atime_dirty = 1;
1021 }
1022
1023 if (flag & ATTR_ATIME) {
1024 ZFS_TIME_ENCODE(&now, zp->z_atime);
1025 }
1026
1027 if (flag & ATTR_MTIME) {
1028 ZFS_TIME_ENCODE(&now, mtime);
1029 if (ZTOZSB(zp)->z_use_fuids) {
1030 zp->z_pflags |= (ZFS_ARCHIVE |
1031 ZFS_AV_MODIFIED);
1032 }
1033 }
1034
1035 if (flag & ATTR_CTIME) {
1036 ZFS_TIME_ENCODE(&now, ctime);
1037 if (ZTOZSB(zp)->z_use_fuids)
1038 zp->z_pflags |= ZFS_ARCHIVE;
1039 }
1040 }
1041
1042 /*
1043 * Grow the block size for a file.
1044 *
1045 * IN: zp - znode of file to free data in.
1046 * size - requested block size
1047 * tx - open transaction.
1048 *
1049 * NOTE: this function assumes that the znode is write locked.
1050 */
1051 void
1052 zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx)
1053 {
1054 int error;
1055 u_longlong_t dummy;
1056
1057 if (size <= zp->z_blksz)
1058 return;
1059 /*
1060 * If the file size is already greater than the current blocksize,
1061 * we will not grow. If there is more than one block in a file,
1062 * the blocksize cannot change.
1063 */
1064 if (zp->z_blksz && zp->z_size > zp->z_blksz)
1065 return;
1066
1067 error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id,
1068 size, 0, tx);
1069
1070 if (error == ENOTSUP)
1071 return;
1072 ASSERT3U(error, ==, 0);
1073
1074 /* What blocksize did we actually get? */
1075 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy);
1076 }
1077
1078 #ifdef HAVE_MMAP
1079 /*
1080 * This is a dummy interface used when pvn_vplist_dirty() should *not*
1081 * be calling back into the fs for a putpage(). E.g.: when truncating
1082 * a file, the pages being "thrown away* don't need to be written out.
1083 */
1084 /* ARGSUSED */
1085 static int
1086 zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp,
1087 int flags, cred_t *cr)
1088 {
1089 ASSERT(0);
1090 return (0);
1091 }
1092 #endif /* HAVE_MMAP */
1093
1094 /*
1095 * Increase the file length
1096 *
1097 * IN: zp - znode of file to free data in.
1098 * end - new end-of-file
1099 *
1100 * RETURN: 0 if success
1101 * error code if failure
1102 */
1103 static int
1104 zfs_extend(znode_t *zp, uint64_t end)
1105 {
1106 zfs_sb_t *zsb = ZTOZSB(zp);
1107 dmu_tx_t *tx;
1108 rl_t *rl;
1109 uint64_t newblksz;
1110 int error;
1111
1112 /*
1113 * We will change zp_size, lock the whole file.
1114 */
1115 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1116
1117 /*
1118 * Nothing to do if file already at desired length.
1119 */
1120 if (end <= zp->z_size) {
1121 zfs_range_unlock(rl);
1122 return (0);
1123 }
1124 top:
1125 tx = dmu_tx_create(zsb->z_os);
1126 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1127 zfs_sa_upgrade_txholds(tx, zp);
1128 if (end > zp->z_blksz &&
1129 (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) {
1130 /*
1131 * We are growing the file past the current block size.
1132 */
1133 if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
1134 ASSERT(!ISP2(zp->z_blksz));
1135 newblksz = MIN(end, SPA_MAXBLOCKSIZE);
1136 } else {
1137 newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
1138 }
1139 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
1140 } else {
1141 newblksz = 0;
1142 }
1143
1144 error = dmu_tx_assign(tx, TXG_NOWAIT);
1145 if (error) {
1146 if (error == ERESTART) {
1147 dmu_tx_wait(tx);
1148 dmu_tx_abort(tx);
1149 goto top;
1150 }
1151 dmu_tx_abort(tx);
1152 zfs_range_unlock(rl);
1153 return (error);
1154 }
1155
1156 if (newblksz)
1157 zfs_grow_blocksize(zp, newblksz, tx);
1158
1159 zp->z_size = end;
1160
1161 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)),
1162 &zp->z_size, sizeof (zp->z_size), tx));
1163
1164 zfs_range_unlock(rl);
1165
1166 dmu_tx_commit(tx);
1167
1168 return (0);
1169 }
1170
1171 /*
1172 * Free space in a file.
1173 *
1174 * IN: zp - znode of file to free data in.
1175 * off - start of section to free.
1176 * len - length of section to free.
1177 *
1178 * RETURN: 0 if success
1179 * error code if failure
1180 */
1181 static int
1182 zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
1183 {
1184 zfs_sb_t *zsb = ZTOZSB(zp);
1185 rl_t *rl;
1186 int error;
1187
1188 /*
1189 * Lock the range being freed.
1190 */
1191 rl = zfs_range_lock(zp, off, len, RL_WRITER);
1192
1193 /*
1194 * Nothing to do if file already at desired length.
1195 */
1196 if (off >= zp->z_size) {
1197 zfs_range_unlock(rl);
1198 return (0);
1199 }
1200
1201 if (off + len > zp->z_size)
1202 len = zp->z_size - off;
1203
1204 error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len);
1205
1206 zfs_range_unlock(rl);
1207
1208 return (error);
1209 }
1210
1211 /*
1212 * Truncate a file
1213 *
1214 * IN: zp - znode of file to free data in.
1215 * end - new end-of-file.
1216 *
1217 * RETURN: 0 if success
1218 * error code if failure
1219 */
1220 static int
1221 zfs_trunc(znode_t *zp, uint64_t end)
1222 {
1223 zfs_sb_t *zsb = ZTOZSB(zp);
1224 dmu_tx_t *tx;
1225 rl_t *rl;
1226 int error;
1227 sa_bulk_attr_t bulk[2];
1228 int count = 0;
1229
1230 /*
1231 * We will change zp_size, lock the whole file.
1232 */
1233 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
1234
1235 /*
1236 * Nothing to do if file already at desired length.
1237 */
1238 if (end >= zp->z_size) {
1239 zfs_range_unlock(rl);
1240 return (0);
1241 }
1242
1243 error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1);
1244 if (error) {
1245 zfs_range_unlock(rl);
1246 return (error);
1247 }
1248 top:
1249 tx = dmu_tx_create(zsb->z_os);
1250 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1251 zfs_sa_upgrade_txholds(tx, zp);
1252 error = dmu_tx_assign(tx, TXG_NOWAIT);
1253 if (error) {
1254 if (error == ERESTART) {
1255 dmu_tx_wait(tx);
1256 dmu_tx_abort(tx);
1257 goto top;
1258 }
1259 dmu_tx_abort(tx);
1260 zfs_range_unlock(rl);
1261 return (error);
1262 }
1263
1264 zp->z_size = end;
1265 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb),
1266 NULL, &zp->z_size, sizeof (zp->z_size));
1267
1268 if (end == 0) {
1269 zp->z_pflags &= ~ZFS_SPARSE;
1270 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1271 NULL, &zp->z_pflags, 8);
1272 }
1273 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
1274
1275 dmu_tx_commit(tx);
1276
1277 zfs_range_unlock(rl);
1278
1279 return (0);
1280 }
1281
1282 /*
1283 * Free space in a file
1284 *
1285 * IN: zp - znode of file to free data in.
1286 * off - start of range
1287 * len - end of range (0 => EOF)
1288 * flag - current file open mode flags.
1289 * log - TRUE if this action should be logged
1290 *
1291 * RETURN: 0 if success
1292 * error code if failure
1293 */
1294 int
1295 zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log)
1296 {
1297 #ifdef HAVE_MANDLOCKS
1298 struct inode *ip = ZTOI(zp);
1299 #endif /* HAVE_MANDLOCKS */
1300 dmu_tx_t *tx;
1301 zfs_sb_t *zsb = ZTOZSB(zp);
1302 zilog_t *zilog = zsb->z_log;
1303 uint64_t mode;
1304 uint64_t mtime[2], ctime[2];
1305 sa_bulk_attr_t bulk[3];
1306 int count = 0;
1307 int error;
1308
1309 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode,
1310 sizeof (mode))) != 0)
1311 return (error);
1312
1313 if (off > zp->z_size) {
1314 error = zfs_extend(zp, off+len);
1315 if (error == 0 && log)
1316 goto log;
1317 else
1318 return (error);
1319 }
1320
1321 #ifdef HAVE_MANDLOCKS
1322 /*
1323 * Check for any locks in the region to be freed.
1324 */
1325
1326 if (MANDLOCK(ip, (mode_t)mode)) {
1327 uint64_t length = (len ? len : zp->z_size - off);
1328 if (error = chklock(ip, FWRITE, off, length, flag, NULL))
1329 return (error);
1330 }
1331 #endif /* HAVE_MANDLOCKS */
1332
1333 if (len == 0) {
1334 error = zfs_trunc(zp, off);
1335 } else {
1336 if ((error = zfs_free_range(zp, off, len)) == 0 &&
1337 off + len > zp->z_size)
1338 error = zfs_extend(zp, off+len);
1339 }
1340 if (error || !log)
1341 return (error);
1342 log:
1343 tx = dmu_tx_create(zsb->z_os);
1344 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
1345 zfs_sa_upgrade_txholds(tx, zp);
1346 error = dmu_tx_assign(tx, TXG_NOWAIT);
1347 if (error) {
1348 if (error == ERESTART) {
1349 dmu_tx_wait(tx);
1350 dmu_tx_abort(tx);
1351 goto log;
1352 }
1353 dmu_tx_abort(tx);
1354 return (error);
1355 }
1356
1357 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16);
1358 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16);
1359 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
1360 NULL, &zp->z_pflags, 8);
1361 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE);
1362 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx);
1363 ASSERT(error == 0);
1364
1365 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len);
1366
1367 dmu_tx_commit(tx);
1368 zfs_inode_update(zp);
1369 return (0);
1370 }
1371
1372 void
1373 zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
1374 {
1375 uint64_t moid, obj, sa_obj, version;
1376 uint64_t norm = 0;
1377 nvpair_t *elem;
1378 int error;
1379 timestruc_t now;
1380 dmu_buf_t *db;
1381 znode_phys_t *pzp;
1382
1383 /*
1384 * First attempt to create master node.
1385 */
1386 /*
1387 * In an empty objset, there are no blocks to read and thus
1388 * there can be no i/o errors (which we assert below).
1389 */
1390 moid = MASTER_NODE_OBJ;
1391 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
1392 DMU_OT_NONE, 0, tx);
1393 ASSERT(error == 0);
1394
1395 /*
1396 * Set starting attributes.
1397 */
1398 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
1399 elem = NULL;
1400 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
1401 /* For the moment we expect all zpl props to be uint64_ts */
1402 uint64_t val;
1403 char *name;
1404
1405 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
1406 VERIFY(nvpair_value_uint64(elem, &val) == 0);
1407 name = nvpair_name(elem);
1408 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
1409 if (val < version)
1410 version = val;
1411 } else {
1412 error = zap_update(os, moid, name, 8, 1, &val, tx);
1413 }
1414 ASSERT(error == 0);
1415 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
1416 norm = val;
1417 }
1418 ASSERT(version != 0);
1419 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
1420
1421 /*
1422 * Create zap object used for SA attribute registration
1423 */
1424
1425 if (version >= ZPL_VERSION_SA) {
1426 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
1427 DMU_OT_NONE, 0, tx);
1428 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
1429 ASSERT(error == 0);
1430 } else {
1431 sa_obj = 0;
1432 }
1433 /*
1434 * Create a delete queue.
1435 */
1436 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
1437
1438 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
1439 ASSERT(error == 0);
1440
1441 /*
1442 * Create root znode with code free of VFS dependencies. This
1443 * is important because without a registered filesystem and super
1444 * block all the required VFS hooks will be missing. The critical
1445 * thing is to just crete the required root znode.
1446 */
1447 obj = zap_create_norm(os, norm, DMU_OT_DIRECTORY_CONTENTS,
1448 DMU_OT_ZNODE, sizeof (znode_phys_t), tx);
1449
1450 VERIFY(0 == dmu_bonus_hold(os, obj, FTAG, &db));
1451 dmu_buf_will_dirty(db, tx);
1452
1453 /*
1454 * Initialize the znode physical data to zero.
1455 */
1456 ASSERT(db->db_size >= sizeof (znode_phys_t));
1457 bzero(db->db_data, db->db_size);
1458 pzp = db->db_data;
1459
1460 if (USE_FUIDS(version, os))
1461 pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED;
1462
1463 pzp->zp_size = 2; /* "." and ".." */
1464 pzp->zp_links = 2;
1465 pzp->zp_parent = obj;
1466 pzp->zp_gen = dmu_tx_get_txg(tx);
1467 pzp->zp_mode = S_IFDIR | 0755;
1468 pzp->zp_flags = ZFS_ACL_TRIVIAL;
1469
1470 gethrestime(&now);
1471
1472 ZFS_TIME_ENCODE(&now, pzp->zp_crtime);
1473 ZFS_TIME_ENCODE(&now, pzp->zp_ctime);
1474 ZFS_TIME_ENCODE(&now, pzp->zp_atime);
1475 ZFS_TIME_ENCODE(&now, pzp->zp_mtime);
1476
1477 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &obj, tx);
1478 ASSERT(error == 0);
1479
1480 dmu_buf_rele(db, FTAG);
1481 }
1482
1483 #endif /* _KERNEL */
1484
1485 static int
1486 zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table)
1487 {
1488 uint64_t sa_obj = 0;
1489 int error;
1490
1491 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj);
1492 if (error != 0 && error != ENOENT)
1493 return (error);
1494
1495 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table);
1496 return (error);
1497 }
1498
1499 static int
1500 zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp,
1501 dmu_buf_t **db)
1502 {
1503 dmu_object_info_t doi;
1504 int error;
1505
1506 if ((error = sa_buf_hold(osp, obj, FTAG, db)) != 0)
1507 return (error);
1508
1509 dmu_object_info_from_db(*db, &doi);
1510 if ((doi.doi_bonus_type != DMU_OT_SA &&
1511 doi.doi_bonus_type != DMU_OT_ZNODE) ||
1512 (doi.doi_bonus_type == DMU_OT_ZNODE &&
1513 doi.doi_bonus_size < sizeof (znode_phys_t))) {
1514 sa_buf_rele(*db, FTAG);
1515 return (ENOTSUP);
1516 }
1517
1518 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp);
1519 if (error != 0) {
1520 sa_buf_rele(*db, FTAG);
1521 return (error);
1522 }
1523
1524 return (0);
1525 }
1526
1527 void
1528 zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db)
1529 {
1530 sa_handle_destroy(hdl);
1531 sa_buf_rele(db, FTAG);
1532 }
1533
1534 /*
1535 * Given an object number, return its parent object number and whether
1536 * or not the object is an extended attribute directory.
1537 */
1538 static int
1539 zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp,
1540 int *is_xattrdir)
1541 {
1542 uint64_t parent;
1543 uint64_t pflags;
1544 uint64_t mode;
1545 sa_bulk_attr_t bulk[3];
1546 int count = 0;
1547 int error;
1548
1549 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL,
1550 &parent, sizeof (parent));
1551 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL,
1552 &pflags, sizeof (pflags));
1553 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1554 &mode, sizeof (mode));
1555
1556 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0)
1557 return (error);
1558
1559 *pobjp = parent;
1560 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode);
1561
1562 return (0);
1563 }
1564
1565 /*
1566 * Given an object number, return some zpl level statistics
1567 */
1568 static int
1569 zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table,
1570 zfs_stat_t *sb)
1571 {
1572 sa_bulk_attr_t bulk[4];
1573 int count = 0;
1574
1575 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL,
1576 &sb->zs_mode, sizeof (sb->zs_mode));
1577 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL,
1578 &sb->zs_gen, sizeof (sb->zs_gen));
1579 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL,
1580 &sb->zs_links, sizeof (sb->zs_links));
1581 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL,
1582 &sb->zs_ctime, sizeof (sb->zs_ctime));
1583
1584 return (sa_bulk_lookup(hdl, bulk, count));
1585 }
1586
1587 static int
1588 zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl,
1589 sa_attr_type_t *sa_table, char *buf, int len)
1590 {
1591 sa_handle_t *sa_hdl;
1592 sa_handle_t *prevhdl = NULL;
1593 dmu_buf_t *prevdb = NULL;
1594 dmu_buf_t *sa_db = NULL;
1595 char *path = buf + len - 1;
1596 int error;
1597
1598 *path = '\0';
1599 sa_hdl = hdl;
1600
1601 for (;;) {
1602 uint64_t pobj;
1603 char component[MAXNAMELEN + 2];
1604 size_t complen;
1605 int is_xattrdir;
1606
1607 if (prevdb)
1608 zfs_release_sa_handle(prevhdl, prevdb);
1609
1610 if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj,
1611 &is_xattrdir)) != 0)
1612 break;
1613
1614 if (pobj == obj) {
1615 if (path[0] != '/')
1616 *--path = '/';
1617 break;
1618 }
1619
1620 component[0] = '/';
1621 if (is_xattrdir) {
1622 (void) sprintf(component + 1, "<xattrdir>");
1623 } else {
1624 error = zap_value_search(osp, pobj, obj,
1625 ZFS_DIRENT_OBJ(-1ULL), component + 1);
1626 if (error != 0)
1627 break;
1628 }
1629
1630 complen = strlen(component);
1631 path -= complen;
1632 ASSERT(path >= buf);
1633 bcopy(component, path, complen);
1634 obj = pobj;
1635
1636 if (sa_hdl != hdl) {
1637 prevhdl = sa_hdl;
1638 prevdb = sa_db;
1639 }
1640 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db);
1641 if (error != 0) {
1642 sa_hdl = prevhdl;
1643 sa_db = prevdb;
1644 break;
1645 }
1646 }
1647
1648 if (sa_hdl != NULL && sa_hdl != hdl) {
1649 ASSERT(sa_db != NULL);
1650 zfs_release_sa_handle(sa_hdl, sa_db);
1651 }
1652
1653 if (error == 0)
1654 (void) memmove(buf, path, buf + len - path);
1655
1656 return (error);
1657 }
1658
1659 int
1660 zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len)
1661 {
1662 sa_attr_type_t *sa_table;
1663 sa_handle_t *hdl;
1664 dmu_buf_t *db;
1665 int error;
1666
1667 error = zfs_sa_setup(osp, &sa_table);
1668 if (error != 0)
1669 return (error);
1670
1671 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1672 if (error != 0)
1673 return (error);
1674
1675 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1676
1677 zfs_release_sa_handle(hdl, db);
1678 return (error);
1679 }
1680
1681 int
1682 zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb,
1683 char *buf, int len)
1684 {
1685 char *path = buf + len - 1;
1686 sa_attr_type_t *sa_table;
1687 sa_handle_t *hdl;
1688 dmu_buf_t *db;
1689 int error;
1690
1691 *path = '\0';
1692
1693 error = zfs_sa_setup(osp, &sa_table);
1694 if (error != 0)
1695 return (error);
1696
1697 error = zfs_grab_sa_handle(osp, obj, &hdl, &db);
1698 if (error != 0)
1699 return (error);
1700
1701 error = zfs_obj_to_stats_impl(hdl, sa_table, sb);
1702 if (error != 0) {
1703 zfs_release_sa_handle(hdl, db);
1704 return (error);
1705 }
1706
1707 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len);
1708
1709 zfs_release_sa_handle(hdl, db);
1710 return (error);
1711 }
1712
1713 #if defined(_KERNEL) && defined(HAVE_SPL)
1714 EXPORT_SYMBOL(zfs_create_fs);
1715 EXPORT_SYMBOL(zfs_obj_to_path);
1716 #endif
ZFS On Linux mirror for PVE