]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
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 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
5d43cc9a | 23 | * Copyright (c) 2012, 2018 by Delphix. All rights reserved. |
34dc7c2f BB |
24 | */ |
25 | ||
26 | /* Portions Copyright 2007 Jeremy Teo */ | |
27 | ||
34dc7c2f BB |
28 | #ifdef _KERNEL |
29 | #include <sys/types.h> | |
30 | #include <sys/param.h> | |
31 | #include <sys/time.h> | |
34dc7c2f | 32 | #include <sys/sysmacros.h> |
34dc7c2f | 33 | #include <sys/mntent.h> |
34dc7c2f BB |
34 | #include <sys/u8_textprep.h> |
35 | #include <sys/dsl_dataset.h> | |
36 | #include <sys/vfs.h> | |
34dc7c2f BB |
37 | #include <sys/vnode.h> |
38 | #include <sys/file.h> | |
39 | #include <sys/kmem.h> | |
40 | #include <sys/errno.h> | |
34dc7c2f BB |
41 | #include <sys/mode.h> |
42 | #include <sys/atomic.h> | |
34dc7c2f BB |
43 | #include <sys/zfs_dir.h> |
44 | #include <sys/zfs_acl.h> | |
45 | #include <sys/zfs_ioctl.h> | |
46 | #include <sys/zfs_rlock.h> | |
47 | #include <sys/zfs_fuid.h> | |
3558fd73 | 48 | #include <sys/zfs_vnops.h> |
ebe7e575 | 49 | #include <sys/zfs_ctldir.h> |
428870ff | 50 | #include <sys/dnode.h> |
34dc7c2f | 51 | #include <sys/fs/zfs.h> |
3558fd73 | 52 | #include <sys/zpl.h> |
34dc7c2f BB |
53 | #endif /* _KERNEL */ |
54 | ||
55 | #include <sys/dmu.h> | |
f1512ee6 | 56 | #include <sys/dmu_objset.h> |
50c957f7 | 57 | #include <sys/dmu_tx.h> |
34dc7c2f BB |
58 | #include <sys/refcount.h> |
59 | #include <sys/stat.h> | |
60 | #include <sys/zap.h> | |
61 | #include <sys/zfs_znode.h> | |
428870ff BB |
62 | #include <sys/sa.h> |
63 | #include <sys/zfs_sa.h> | |
572e2857 | 64 | #include <sys/zfs_stat.h> |
34dc7c2f BB |
65 | |
66 | #include "zfs_prop.h" | |
428870ff | 67 | #include "zfs_comutil.h" |
34dc7c2f | 68 | |
b128c09f BB |
69 | /* |
70 | * Define ZNODE_STATS to turn on statistic gathering. By default, it is only | |
71 | * turned on when DEBUG is also defined. | |
72 | */ | |
73 | #ifdef DEBUG | |
74 | #define ZNODE_STATS | |
75 | #endif /* DEBUG */ | |
76 | ||
77 | #ifdef ZNODE_STATS | |
78 | #define ZNODE_STAT_ADD(stat) ((stat)++) | |
79 | #else | |
80 | #define ZNODE_STAT_ADD(stat) /* nothing */ | |
81 | #endif /* ZNODE_STATS */ | |
82 | ||
34dc7c2f BB |
83 | /* |
84 | * Functions needed for userland (ie: libzpool) are not put under | |
85 | * #ifdef_KERNEL; the rest of the functions have dependencies | |
86 | * (such as VFS logic) that will not compile easily in userland. | |
87 | */ | |
88 | #ifdef _KERNEL | |
9babb374 | 89 | |
b128c09f | 90 | static kmem_cache_t *znode_cache = NULL; |
c96c36fa | 91 | static kmem_cache_t *znode_hold_cache = NULL; |
0720116d | 92 | unsigned int zfs_object_mutex_size = ZFS_OBJ_MTX_SZ; |
34dc7c2f | 93 | |
dcec0a12 AP |
94 | /* |
95 | * This is used by the test suite so that it can delay znodes from being | |
96 | * freed in order to inspect the unlinked set. | |
97 | */ | |
98 | int zfs_unlink_suspend_progress = 0; | |
99 | ||
5d43cc9a MA |
100 | /* |
101 | * This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on | |
102 | * z_rangelock. It will modify the offset and length of the lock to reflect | |
103 | * znode-specific information, and convert RL_APPEND to RL_WRITER. This is | |
104 | * called with the rangelock_t's rl_lock held, which avoids races. | |
105 | */ | |
106 | static void | |
107 | zfs_rangelock_cb(locked_range_t *new, void *arg) | |
108 | { | |
109 | znode_t *zp = arg; | |
110 | ||
111 | /* | |
112 | * If in append mode, convert to writer and lock starting at the | |
113 | * current end of file. | |
114 | */ | |
115 | if (new->lr_type == RL_APPEND) { | |
116 | new->lr_offset = zp->z_size; | |
117 | new->lr_type = RL_WRITER; | |
118 | } | |
119 | ||
120 | /* | |
121 | * If we need to grow the block size then lock the whole file range. | |
122 | */ | |
123 | uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length); | |
124 | if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) || | |
125 | zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) { | |
126 | new->lr_offset = 0; | |
127 | new->lr_length = UINT64_MAX; | |
128 | } | |
129 | } | |
130 | ||
34dc7c2f BB |
131 | /*ARGSUSED*/ |
132 | static int | |
b128c09f | 133 | zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) |
34dc7c2f BB |
134 | { |
135 | znode_t *zp = buf; | |
136 | ||
3558fd73 | 137 | inode_init_once(ZTOI(zp)); |
b128c09f BB |
138 | list_link_init(&zp->z_link_node); |
139 | ||
34dc7c2f | 140 | mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 141 | rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); |
448d7aaa | 142 | rw_init(&zp->z_name_lock, NULL, RW_NOLOCKDEP, NULL); |
34dc7c2f | 143 | mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); |
82a37189 | 144 | rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); |
34dc7c2f | 145 | |
5d43cc9a | 146 | rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp); |
34dc7c2f | 147 | |
b128c09f | 148 | zp->z_dirlocks = NULL; |
45d1cae3 | 149 | zp->z_acl_cached = NULL; |
82a37189 | 150 | zp->z_xattr_cached = NULL; |
98701490 | 151 | zp->z_xattr_parent = 0; |
572e2857 | 152 | zp->z_moved = 0; |
34dc7c2f BB |
153 | return (0); |
154 | } | |
155 | ||
156 | /*ARGSUSED*/ | |
157 | static void | |
b128c09f | 158 | zfs_znode_cache_destructor(void *buf, void *arg) |
34dc7c2f BB |
159 | { |
160 | znode_t *zp = buf; | |
161 | ||
b128c09f | 162 | ASSERT(!list_link_active(&zp->z_link_node)); |
34dc7c2f | 163 | mutex_destroy(&zp->z_lock); |
34dc7c2f BB |
164 | rw_destroy(&zp->z_parent_lock); |
165 | rw_destroy(&zp->z_name_lock); | |
166 | mutex_destroy(&zp->z_acl_lock); | |
82a37189 | 167 | rw_destroy(&zp->z_xattr_lock); |
5d43cc9a | 168 | rangelock_fini(&zp->z_rangelock); |
34dc7c2f | 169 | |
b128c09f | 170 | ASSERT(zp->z_dirlocks == NULL); |
45d1cae3 | 171 | ASSERT(zp->z_acl_cached == NULL); |
82a37189 | 172 | ASSERT(zp->z_xattr_cached == NULL); |
b128c09f BB |
173 | } |
174 | ||
c96c36fa BB |
175 | static int |
176 | zfs_znode_hold_cache_constructor(void *buf, void *arg, int kmflags) | |
177 | { | |
178 | znode_hold_t *zh = buf; | |
179 | ||
180 | mutex_init(&zh->zh_lock, NULL, MUTEX_DEFAULT, NULL); | |
424fd7c3 | 181 | zfs_refcount_create(&zh->zh_refcount); |
c96c36fa BB |
182 | zh->zh_obj = ZFS_NO_OBJECT; |
183 | ||
184 | return (0); | |
185 | } | |
186 | ||
187 | static void | |
188 | zfs_znode_hold_cache_destructor(void *buf, void *arg) | |
189 | { | |
190 | znode_hold_t *zh = buf; | |
191 | ||
192 | mutex_destroy(&zh->zh_lock); | |
424fd7c3 | 193 | zfs_refcount_destroy(&zh->zh_refcount); |
c96c36fa BB |
194 | } |
195 | ||
34dc7c2f BB |
196 | void |
197 | zfs_znode_init(void) | |
198 | { | |
199 | /* | |
5074bfe8 TC |
200 | * Initialize zcache. The KMC_SLAB hint is used in order that it be |
201 | * backed by kmalloc() when on the Linux slab in order that any | |
202 | * wait_on_bit() operations on the related inode operate properly. | |
34dc7c2f BB |
203 | */ |
204 | ASSERT(znode_cache == NULL); | |
205 | znode_cache = kmem_cache_create("zfs_znode_cache", | |
206 | sizeof (znode_t), 0, zfs_znode_cache_constructor, | |
5074bfe8 | 207 | zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_SLAB); |
c96c36fa BB |
208 | |
209 | ASSERT(znode_hold_cache == NULL); | |
210 | znode_hold_cache = kmem_cache_create("zfs_znode_hold_cache", | |
211 | sizeof (znode_hold_t), 0, zfs_znode_hold_cache_constructor, | |
212 | zfs_znode_hold_cache_destructor, NULL, NULL, NULL, 0); | |
34dc7c2f BB |
213 | } |
214 | ||
215 | void | |
216 | zfs_znode_fini(void) | |
217 | { | |
34dc7c2f BB |
218 | /* |
219 | * Cleanup zcache | |
220 | */ | |
221 | if (znode_cache) | |
222 | kmem_cache_destroy(znode_cache); | |
223 | znode_cache = NULL; | |
c96c36fa BB |
224 | |
225 | if (znode_hold_cache) | |
226 | kmem_cache_destroy(znode_hold_cache); | |
227 | znode_hold_cache = NULL; | |
228 | } | |
229 | ||
230 | /* | |
231 | * The zfs_znode_hold_enter() / zfs_znode_hold_exit() functions are used to | |
232 | * serialize access to a znode and its SA buffer while the object is being | |
233 | * created or destroyed. This kind of locking would normally reside in the | |
234 | * znode itself but in this case that's impossible because the znode and SA | |
235 | * buffer may not yet exist. Therefore the locking is handled externally | |
236 | * with an array of mutexs and AVLs trees which contain per-object locks. | |
237 | * | |
238 | * In zfs_znode_hold_enter() a per-object lock is created as needed, inserted | |
239 | * in to the correct AVL tree and finally the per-object lock is held. In | |
240 | * zfs_znode_hold_exit() the process is reversed. The per-object lock is | |
241 | * released, removed from the AVL tree and destroyed if there are no waiters. | |
242 | * | |
243 | * This scheme has two important properties: | |
244 | * | |
245 | * 1) No memory allocations are performed while holding one of the z_hold_locks. | |
246 | * This ensures evict(), which can be called from direct memory reclaim, will | |
247 | * never block waiting on a z_hold_locks which just happens to have hashed | |
248 | * to the same index. | |
249 | * | |
250 | * 2) All locks used to serialize access to an object are per-object and never | |
251 | * shared. This minimizes lock contention without creating a large number | |
252 | * of dedicated locks. | |
253 | * | |
254 | * On the downside it does require znode_lock_t structures to be frequently | |
255 | * allocated and freed. However, because these are backed by a kmem cache | |
256 | * and very short lived this cost is minimal. | |
257 | */ | |
258 | int | |
259 | zfs_znode_hold_compare(const void *a, const void *b) | |
260 | { | |
ee36c709 GN |
261 | const znode_hold_t *zh_a = (const znode_hold_t *)a; |
262 | const znode_hold_t *zh_b = (const znode_hold_t *)b; | |
263 | ||
264 | return (AVL_CMP(zh_a->zh_obj, zh_b->zh_obj)); | |
c96c36fa BB |
265 | } |
266 | ||
267 | boolean_t | |
0037b49e | 268 | zfs_znode_held(zfsvfs_t *zfsvfs, uint64_t obj) |
c96c36fa BB |
269 | { |
270 | znode_hold_t *zh, search; | |
0037b49e | 271 | int i = ZFS_OBJ_HASH(zfsvfs, obj); |
37c56346 | 272 | boolean_t held; |
c96c36fa BB |
273 | |
274 | search.zh_obj = obj; | |
275 | ||
0037b49e BB |
276 | mutex_enter(&zfsvfs->z_hold_locks[i]); |
277 | zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL); | |
37c56346 | 278 | held = (zh && MUTEX_HELD(&zh->zh_lock)) ? B_TRUE : B_FALSE; |
0037b49e | 279 | mutex_exit(&zfsvfs->z_hold_locks[i]); |
c96c36fa | 280 | |
37c56346 | 281 | return (held); |
c96c36fa BB |
282 | } |
283 | ||
284 | static znode_hold_t * | |
0037b49e | 285 | zfs_znode_hold_enter(zfsvfs_t *zfsvfs, uint64_t obj) |
c96c36fa BB |
286 | { |
287 | znode_hold_t *zh, *zh_new, search; | |
0037b49e | 288 | int i = ZFS_OBJ_HASH(zfsvfs, obj); |
c96c36fa BB |
289 | boolean_t found = B_FALSE; |
290 | ||
291 | zh_new = kmem_cache_alloc(znode_hold_cache, KM_SLEEP); | |
292 | zh_new->zh_obj = obj; | |
293 | search.zh_obj = obj; | |
294 | ||
0037b49e BB |
295 | mutex_enter(&zfsvfs->z_hold_locks[i]); |
296 | zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL); | |
c96c36fa BB |
297 | if (likely(zh == NULL)) { |
298 | zh = zh_new; | |
0037b49e | 299 | avl_add(&zfsvfs->z_hold_trees[i], zh); |
c96c36fa BB |
300 | } else { |
301 | ASSERT3U(zh->zh_obj, ==, obj); | |
302 | found = B_TRUE; | |
303 | } | |
c13060e4 | 304 | zfs_refcount_add(&zh->zh_refcount, NULL); |
0037b49e | 305 | mutex_exit(&zfsvfs->z_hold_locks[i]); |
c96c36fa BB |
306 | |
307 | if (found == B_TRUE) | |
308 | kmem_cache_free(znode_hold_cache, zh_new); | |
309 | ||
310 | ASSERT(MUTEX_NOT_HELD(&zh->zh_lock)); | |
424fd7c3 | 311 | ASSERT3S(zfs_refcount_count(&zh->zh_refcount), >, 0); |
c96c36fa BB |
312 | mutex_enter(&zh->zh_lock); |
313 | ||
314 | return (zh); | |
315 | } | |
316 | ||
317 | static void | |
0037b49e | 318 | zfs_znode_hold_exit(zfsvfs_t *zfsvfs, znode_hold_t *zh) |
c96c36fa | 319 | { |
0037b49e | 320 | int i = ZFS_OBJ_HASH(zfsvfs, zh->zh_obj); |
c96c36fa BB |
321 | boolean_t remove = B_FALSE; |
322 | ||
0037b49e | 323 | ASSERT(zfs_znode_held(zfsvfs, zh->zh_obj)); |
424fd7c3 | 324 | ASSERT3S(zfs_refcount_count(&zh->zh_refcount), >, 0); |
c96c36fa BB |
325 | mutex_exit(&zh->zh_lock); |
326 | ||
0037b49e | 327 | mutex_enter(&zfsvfs->z_hold_locks[i]); |
424fd7c3 | 328 | if (zfs_refcount_remove(&zh->zh_refcount, NULL) == 0) { |
0037b49e | 329 | avl_remove(&zfsvfs->z_hold_trees[i], zh); |
c96c36fa BB |
330 | remove = B_TRUE; |
331 | } | |
0037b49e | 332 | mutex_exit(&zfsvfs->z_hold_locks[i]); |
c96c36fa BB |
333 | |
334 | if (remove == B_TRUE) | |
335 | kmem_cache_free(znode_hold_cache, zh); | |
34dc7c2f BB |
336 | } |
337 | ||
34dc7c2f | 338 | static void |
0037b49e | 339 | zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, |
428870ff | 340 | dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) |
34dc7c2f | 341 | { |
0037b49e | 342 | ASSERT(zfs_znode_held(zfsvfs, zp->z_id)); |
34dc7c2f BB |
343 | |
344 | mutex_enter(&zp->z_lock); | |
345 | ||
428870ff BB |
346 | ASSERT(zp->z_sa_hdl == NULL); |
347 | ASSERT(zp->z_acl_cached == NULL); | |
348 | if (sa_hdl == NULL) { | |
0037b49e | 349 | VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp, |
428870ff BB |
350 | SA_HDL_SHARED, &zp->z_sa_hdl)); |
351 | } else { | |
352 | zp->z_sa_hdl = sa_hdl; | |
353 | sa_set_userp(sa_hdl, zp); | |
354 | } | |
34dc7c2f | 355 | |
428870ff | 356 | zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; |
34dc7c2f | 357 | |
34dc7c2f | 358 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
359 | } |
360 | ||
361 | void | |
362 | zfs_znode_dmu_fini(znode_t *zp) | |
363 | { | |
c96c36fa | 364 | ASSERT(zfs_znode_held(ZTOZSB(zp), zp->z_id) || zp->z_unlinked || |
3558fd73 | 365 | RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock)); |
428870ff BB |
366 | |
367 | sa_handle_destroy(zp->z_sa_hdl); | |
368 | zp->z_sa_hdl = NULL; | |
34dc7c2f BB |
369 | } |
370 | ||
371 | /* | |
3558fd73 BB |
372 | * Called by new_inode() to allocate a new inode. |
373 | */ | |
374 | int | |
375 | zfs_inode_alloc(struct super_block *sb, struct inode **ip) | |
376 | { | |
377 | znode_t *zp; | |
378 | ||
79c76d5b | 379 | zp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
3558fd73 BB |
380 | *ip = ZTOI(zp); |
381 | ||
382 | return (0); | |
383 | } | |
384 | ||
385 | /* | |
386 | * Called in multiple places when an inode should be destroyed. | |
387 | */ | |
388 | void | |
389 | zfs_inode_destroy(struct inode *ip) | |
390 | { | |
391 | znode_t *zp = ITOZ(ip); | |
0037b49e | 392 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
3558fd73 | 393 | |
0037b49e | 394 | mutex_enter(&zfsvfs->z_znodes_lock); |
7b3e34ba | 395 | if (list_link_active(&zp->z_link_node)) { |
0037b49e BB |
396 | list_remove(&zfsvfs->z_all_znodes, zp); |
397 | zfsvfs->z_nr_znodes--; | |
7b3e34ba | 398 | } |
0037b49e | 399 | mutex_exit(&zfsvfs->z_znodes_lock); |
3558fd73 BB |
400 | |
401 | if (zp->z_acl_cached) { | |
402 | zfs_acl_free(zp->z_acl_cached); | |
403 | zp->z_acl_cached = NULL; | |
404 | } | |
405 | ||
82a37189 BB |
406 | if (zp->z_xattr_cached) { |
407 | nvlist_free(zp->z_xattr_cached); | |
408 | zp->z_xattr_cached = NULL; | |
409 | } | |
410 | ||
3558fd73 BB |
411 | kmem_cache_free(znode_cache, zp); |
412 | } | |
413 | ||
414 | static void | |
0037b49e | 415 | zfs_inode_set_ops(zfsvfs_t *zfsvfs, struct inode *ip) |
3558fd73 | 416 | { |
aa6d8c10 | 417 | uint64_t rdev = 0; |
3558fd73 BB |
418 | |
419 | switch (ip->i_mode & S_IFMT) { | |
420 | case S_IFREG: | |
421 | ip->i_op = &zpl_inode_operations; | |
422 | ip->i_fop = &zpl_file_operations; | |
423 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
424 | break; | |
425 | ||
426 | case S_IFDIR: | |
427 | ip->i_op = &zpl_dir_inode_operations; | |
428 | ip->i_fop = &zpl_dir_file_operations; | |
429 | ITOZ(ip)->z_zn_prefetch = B_TRUE; | |
430 | break; | |
431 | ||
432 | case S_IFLNK: | |
433 | ip->i_op = &zpl_symlink_inode_operations; | |
434 | break; | |
435 | ||
aa6d8c10 NB |
436 | /* |
437 | * rdev is only stored in a SA only for device files. | |
438 | */ | |
3558fd73 BB |
439 | case S_IFCHR: |
440 | case S_IFBLK: | |
0037b49e | 441 | (void) sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zfsvfs), &rdev, |
53b1d979 | 442 | sizeof (rdev)); |
aa6d8c10 NB |
443 | /*FALLTHROUGH*/ |
444 | case S_IFIFO: | |
445 | case S_IFSOCK: | |
3558fd73 BB |
446 | init_special_inode(ip, ip->i_mode, rdev); |
447 | ip->i_op = &zpl_special_inode_operations; | |
448 | break; | |
449 | ||
450 | default: | |
53b1d979 BB |
451 | zfs_panic_recover("inode %llu has invalid mode: 0x%x\n", |
452 | (u_longlong_t)ip->i_ino, ip->i_mode); | |
453 | ||
454 | /* Assume the inode is a file and attempt to continue */ | |
455 | ip->i_mode = S_IFREG | 0644; | |
456 | ip->i_op = &zpl_inode_operations; | |
457 | ip->i_fop = &zpl_file_operations; | |
458 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
459 | break; | |
3558fd73 BB |
460 | } |
461 | } | |
462 | ||
7bb1325f CC |
463 | void |
464 | zfs_set_inode_flags(znode_t *zp, struct inode *ip) | |
465 | { | |
466 | /* | |
467 | * Linux and Solaris have different sets of file attributes, so we | |
468 | * restrict this conversion to the intersection of the two. | |
469 | */ | |
a5248129 CC |
470 | #ifdef HAVE_INODE_SET_FLAGS |
471 | unsigned int flags = 0; | |
472 | if (zp->z_pflags & ZFS_IMMUTABLE) | |
473 | flags |= S_IMMUTABLE; | |
474 | if (zp->z_pflags & ZFS_APPENDONLY) | |
475 | flags |= S_APPEND; | |
7bb1325f | 476 | |
a5248129 CC |
477 | inode_set_flags(ip, flags, S_IMMUTABLE|S_APPEND); |
478 | #else | |
7bb1325f CC |
479 | if (zp->z_pflags & ZFS_IMMUTABLE) |
480 | ip->i_flags |= S_IMMUTABLE; | |
481 | else | |
482 | ip->i_flags &= ~S_IMMUTABLE; | |
483 | ||
484 | if (zp->z_pflags & ZFS_APPENDONLY) | |
485 | ip->i_flags |= S_APPEND; | |
486 | else | |
487 | ip->i_flags &= ~S_APPEND; | |
a5248129 | 488 | #endif |
7bb1325f CC |
489 | } |
490 | ||
704cd075 CC |
491 | /* |
492 | * Update the embedded inode given the znode. We should work toward | |
493 | * eliminating this function as soon as possible by removing values | |
494 | * which are duplicated between the znode and inode. If the generic | |
495 | * inode has the correct field it should be used, and the ZFS code | |
496 | * updated to access the inode. This can be done incrementally. | |
497 | */ | |
9f5f0019 NB |
498 | void |
499 | zfs_inode_update(znode_t *zp) | |
704cd075 | 500 | { |
0037b49e | 501 | zfsvfs_t *zfsvfs; |
704cd075 CC |
502 | struct inode *ip; |
503 | uint32_t blksize; | |
504 | u_longlong_t i_blocks; | |
704cd075 CC |
505 | |
506 | ASSERT(zp != NULL); | |
0037b49e | 507 | zfsvfs = ZTOZSB(zp); |
704cd075 CC |
508 | ip = ZTOI(zp); |
509 | ||
510 | /* Skip .zfs control nodes which do not exist on disk. */ | |
511 | if (zfsctl_is_node(ip)) | |
512 | return; | |
513 | ||
704cd075 CC |
514 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, &i_blocks); |
515 | ||
516 | spin_lock(&ip->i_lock); | |
704cd075 | 517 | ip->i_blocks = i_blocks; |
704cd075 CC |
518 | i_size_write(ip, zp->z_size); |
519 | spin_unlock(&ip->i_lock); | |
520 | } | |
521 | ||
704cd075 | 522 | |
3558fd73 BB |
523 | /* |
524 | * Construct a znode+inode and initialize. | |
34dc7c2f BB |
525 | * |
526 | * This does not do a call to dmu_set_user() that is | |
527 | * up to the caller to do, in case you don't want to | |
528 | * return the znode | |
529 | */ | |
530 | static znode_t * | |
0037b49e | 531 | zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, |
31b6111f | 532 | dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl) |
34dc7c2f BB |
533 | { |
534 | znode_t *zp; | |
3558fd73 | 535 | struct inode *ip; |
7f89ae6b | 536 | uint64_t mode; |
428870ff | 537 | uint64_t parent; |
278f2236 | 538 | uint64_t tmp_gen; |
dfbc8630 | 539 | uint64_t links; |
2c6abf15 | 540 | uint64_t z_uid, z_gid; |
9f5f0019 | 541 | uint64_t atime[2], mtime[2], ctime[2]; |
9c5167d1 | 542 | uint64_t projid = ZFS_DEFAULT_PROJID; |
9f5f0019 | 543 | sa_bulk_attr_t bulk[11]; |
428870ff | 544 | int count = 0; |
34dc7c2f | 545 | |
0037b49e | 546 | ASSERT(zfsvfs != NULL); |
34dc7c2f | 547 | |
0037b49e | 548 | ip = new_inode(zfsvfs->z_sb); |
3558fd73 BB |
549 | if (ip == NULL) |
550 | return (NULL); | |
7304b6e5 | 551 | |
3558fd73 | 552 | zp = ITOZ(ip); |
34dc7c2f | 553 | ASSERT(zp->z_dirlocks == NULL); |
ebe7e575 BB |
554 | ASSERT3P(zp->z_acl_cached, ==, NULL); |
555 | ASSERT3P(zp->z_xattr_cached, ==, NULL); | |
572e2857 | 556 | zp->z_moved = 0; |
428870ff | 557 | zp->z_sa_hdl = NULL; |
34dc7c2f BB |
558 | zp->z_unlinked = 0; |
559 | zp->z_atime_dirty = 0; | |
560 | zp->z_mapcnt = 0; | |
34dc7c2f BB |
561 | zp->z_id = db->db_object; |
562 | zp->z_blksz = blksz; | |
563 | zp->z_seq = 0x7A4653; | |
564 | zp->z_sync_cnt = 0; | |
ebe7e575 BB |
565 | zp->z_is_mapped = B_FALSE; |
566 | zp->z_is_ctldir = B_FALSE; | |
7b3e34ba | 567 | zp->z_is_stale = B_FALSE; |
34dc7c2f | 568 | |
0037b49e | 569 | zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); |
3558fd73 | 570 | |
0037b49e BB |
571 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
572 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8); | |
573 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
574 | &zp->z_size, 8); | |
575 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); | |
576 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
428870ff | 577 | &zp->z_pflags, 8); |
0037b49e | 578 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, |
7304b6e5 | 579 | &parent, 8); |
0037b49e BB |
580 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &z_uid, 8); |
581 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &z_gid, 8); | |
582 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
583 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
584 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
428870ff | 585 | |
9c5167d1 NF |
586 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || tmp_gen == 0 || |
587 | (dmu_objset_projectquota_enabled(zfsvfs->z_os) && | |
588 | (zp->z_pflags & ZFS_PROJID) && | |
589 | sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) { | |
428870ff BB |
590 | if (hdl == NULL) |
591 | sa_handle_destroy(zp->z_sa_hdl); | |
07d63f0c | 592 | zp->z_sa_hdl = NULL; |
3558fd73 | 593 | goto error; |
34dc7c2f | 594 | } |
7304b6e5 | 595 | |
9c5167d1 | 596 | zp->z_projid = projid; |
12fa7f34 | 597 | zp->z_mode = ip->i_mode = mode; |
278f2236 | 598 | ip->i_generation = (uint32_t)tmp_gen; |
ba2fe6af | 599 | ip->i_blkbits = SPA_MINBLOCKSHIFT; |
dfbc8630 | 600 | set_nlink(ip, (uint32_t)links); |
2c6abf15 NB |
601 | zfs_uid_write(ip, z_uid); |
602 | zfs_gid_write(ip, z_gid); | |
7bb1325f | 603 | zfs_set_inode_flags(zp, ip); |
7f89ae6b | 604 | |
98701490 CC |
605 | /* Cache the xattr parent id */ |
606 | if (zp->z_pflags & ZFS_XATTR) | |
607 | zp->z_xattr_parent = parent; | |
608 | ||
9f5f0019 NB |
609 | ZFS_TIME_DECODE(&ip->i_atime, atime); |
610 | ZFS_TIME_DECODE(&ip->i_mtime, mtime); | |
611 | ZFS_TIME_DECODE(&ip->i_ctime, ctime); | |
612 | ||
3558fd73 | 613 | ip->i_ino = obj; |
9f5f0019 | 614 | zfs_inode_update(zp); |
0037b49e | 615 | zfs_inode_set_ops(zfsvfs, ip); |
3558fd73 | 616 | |
7b3e34ba BB |
617 | /* |
618 | * The only way insert_inode_locked() can fail is if the ip->i_ino | |
619 | * number is already hashed for this super block. This can never | |
620 | * happen because the inode numbers map 1:1 with the object numbers. | |
621 | * | |
622 | * The one exception is rolling back a mounted file system, but in | |
623 | * this case all the active inode are unhashed during the rollback. | |
624 | */ | |
625 | VERIFY3S(insert_inode_locked(ip), ==, 0); | |
c85b224f | 626 | |
0037b49e BB |
627 | mutex_enter(&zfsvfs->z_znodes_lock); |
628 | list_insert_tail(&zfsvfs->z_all_znodes, zp); | |
629 | zfsvfs->z_nr_znodes++; | |
b128c09f | 630 | membar_producer(); |
0037b49e | 631 | mutex_exit(&zfsvfs->z_znodes_lock); |
b128c09f | 632 | |
3558fd73 | 633 | unlock_new_inode(ip); |
34dc7c2f | 634 | return (zp); |
3558fd73 BB |
635 | |
636 | error: | |
3558fd73 | 637 | iput(ip); |
d1d7e268 | 638 | return (NULL); |
34dc7c2f BB |
639 | } |
640 | ||
1e8db771 BB |
641 | /* |
642 | * Safely mark an inode dirty. Inodes which are part of a read-only | |
643 | * file system or snapshot may not be dirtied. | |
644 | */ | |
645 | void | |
646 | zfs_mark_inode_dirty(struct inode *ip) | |
647 | { | |
0037b49e | 648 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
1e8db771 | 649 | |
0037b49e | 650 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
1e8db771 BB |
651 | return; |
652 | ||
653 | mark_inode_dirty(ip); | |
654 | } | |
655 | ||
428870ff BB |
656 | static uint64_t empty_xattr; |
657 | static uint64_t pad[4]; | |
658 | static zfs_acl_phys_t acl_phys; | |
34dc7c2f BB |
659 | /* |
660 | * Create a new DMU object to hold a zfs znode. | |
661 | * | |
662 | * IN: dzp - parent directory for new znode | |
663 | * vap - file attributes for new znode | |
664 | * tx - dmu transaction id for zap operations | |
665 | * cr - credentials of caller | |
666 | * flag - flags: | |
667 | * IS_ROOT_NODE - new object will be root | |
668 | * IS_XATTR - new object is an attribute | |
34dc7c2f BB |
669 | * bonuslen - length of bonus buffer |
670 | * setaclp - File/Dir initial ACL | |
671 | * fuidp - Tracks fuid allocation. | |
672 | * | |
673 | * OUT: zpp - allocated znode | |
674 | * | |
675 | */ | |
676 | void | |
677 | zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, | |
428870ff | 678 | uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) |
34dc7c2f | 679 | { |
428870ff BB |
680 | uint64_t crtime[2], atime[2], mtime[2], ctime[2]; |
681 | uint64_t mode, size, links, parent, pflags; | |
9c5167d1 | 682 | uint64_t projid = ZFS_DEFAULT_PROJID; |
428870ff | 683 | uint64_t rdev = 0; |
0037b49e | 684 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
428870ff | 685 | dmu_buf_t *db; |
6413c95f | 686 | inode_timespec_t now; |
34dc7c2f | 687 | uint64_t gen, obj; |
428870ff | 688 | int bonuslen; |
50c957f7 | 689 | int dnodesize; |
428870ff BB |
690 | sa_handle_t *sa_hdl; |
691 | dmu_object_type_t obj_type; | |
f30484af | 692 | sa_bulk_attr_t *sa_attrs; |
428870ff BB |
693 | int cnt = 0; |
694 | zfs_acl_locator_cb_t locate = { 0 }; | |
c96c36fa | 695 | znode_hold_t *zh; |
34dc7c2f | 696 | |
0037b49e | 697 | if (zfsvfs->z_replay) { |
34dc7c2f | 698 | obj = vap->va_nodeid; |
34dc7c2f BB |
699 | now = vap->va_ctime; /* see zfs_replay_create() */ |
700 | gen = vap->va_nblocks; /* ditto */ | |
50c957f7 | 701 | dnodesize = vap->va_fsid; /* ditto */ |
34dc7c2f BB |
702 | } else { |
703 | obj = 0; | |
704 | gethrestime(&now); | |
705 | gen = dmu_tx_get_txg(tx); | |
0037b49e | 706 | dnodesize = dmu_objset_dnodesize(zfsvfs->z_os); |
34dc7c2f BB |
707 | } |
708 | ||
50c957f7 NB |
709 | if (dnodesize == 0) |
710 | dnodesize = DNODE_MIN_SIZE; | |
711 | ||
0037b49e | 712 | obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; |
50c957f7 | 713 | |
428870ff | 714 | bonuslen = (obj_type == DMU_OT_SA) ? |
50c957f7 | 715 | DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE; |
428870ff | 716 | |
34dc7c2f BB |
717 | /* |
718 | * Create a new DMU object. | |
719 | */ | |
720 | /* | |
721 | * There's currently no mechanism for pre-reading the blocks that will | |
572e2857 | 722 | * be needed to allocate a new object, so we accept the small chance |
34dc7c2f BB |
723 | * that there will be an i/o error and we will fail one of the |
724 | * assertions below. | |
725 | */ | |
3558fd73 | 726 | if (S_ISDIR(vap->va_mode)) { |
0037b49e BB |
727 | if (zfsvfs->z_replay) { |
728 | VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj, | |
729 | zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
50c957f7 | 730 | obj_type, bonuslen, dnodesize, tx)); |
34dc7c2f | 731 | } else { |
0037b49e BB |
732 | obj = zap_create_norm_dnsize(zfsvfs->z_os, |
733 | zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
50c957f7 | 734 | obj_type, bonuslen, dnodesize, tx); |
34dc7c2f BB |
735 | } |
736 | } else { | |
0037b49e BB |
737 | if (zfsvfs->z_replay) { |
738 | VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj, | |
34dc7c2f | 739 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
50c957f7 | 740 | obj_type, bonuslen, dnodesize, tx)); |
34dc7c2f | 741 | } else { |
0037b49e | 742 | obj = dmu_object_alloc_dnsize(zfsvfs->z_os, |
34dc7c2f | 743 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
50c957f7 | 744 | obj_type, bonuslen, dnodesize, tx); |
34dc7c2f BB |
745 | } |
746 | } | |
34dc7c2f | 747 | |
0037b49e | 748 | zh = zfs_znode_hold_enter(zfsvfs, obj); |
9631681b | 749 | VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db)); |
34dc7c2f BB |
750 | |
751 | /* | |
752 | * If this is the root, fix up the half-initialized parent pointer | |
753 | * to reference the just-allocated physical data area. | |
754 | */ | |
755 | if (flag & IS_ROOT_NODE) { | |
34dc7c2f BB |
756 | dzp->z_id = obj; |
757 | } | |
758 | ||
759 | /* | |
760 | * If parent is an xattr, so am I. | |
761 | */ | |
9c5167d1 | 762 | if (dzp->z_pflags & ZFS_XATTR) { |
34dc7c2f | 763 | flag |= IS_XATTR; |
34dc7c2f BB |
764 | } |
765 | ||
0037b49e | 766 | if (zfsvfs->z_use_fuids) |
428870ff BB |
767 | pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; |
768 | else | |
769 | pflags = 0; | |
34dc7c2f | 770 | |
3558fd73 | 771 | if (S_ISDIR(vap->va_mode)) { |
428870ff | 772 | size = 2; /* contents ("." and "..") */ |
dfbc8630 | 773 | links = 2; |
428870ff | 774 | } else { |
dfbc8630 | 775 | size = 0; |
ace1eae8 | 776 | links = (flag & IS_TMPFILE) ? 0 : 1; |
34dc7c2f BB |
777 | } |
778 | ||
aa6d8c10 | 779 | if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode)) |
dc1d7665 | 780 | rdev = vap->va_rdev; |
428870ff BB |
781 | |
782 | parent = dzp->z_id; | |
783 | mode = acl_ids->z_mode; | |
34dc7c2f | 784 | if (flag & IS_XATTR) |
428870ff | 785 | pflags |= ZFS_XATTR; |
34dc7c2f | 786 | |
9c5167d1 NF |
787 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) { |
788 | /* | |
789 | * With ZFS_PROJID flag, we can easily know whether there is | |
790 | * project ID stored on disk or not. See zfs_space_delta_cb(). | |
791 | */ | |
792 | if (obj_type != DMU_OT_ZNODE && | |
793 | dmu_objset_projectquota_enabled(zfsvfs->z_os)) | |
794 | pflags |= ZFS_PROJID; | |
795 | ||
796 | /* | |
797 | * Inherit project ID from parent if required. | |
798 | */ | |
799 | projid = zfs_inherit_projid(dzp); | |
800 | if (dzp->z_pflags & ZFS_PROJINHERIT) | |
801 | pflags |= ZFS_PROJINHERIT; | |
802 | } | |
803 | ||
428870ff BB |
804 | /* |
805 | * No execs denied will be deterimed when zfs_mode_compute() is called. | |
806 | */ | |
807 | pflags |= acl_ids->z_aclp->z_hints & | |
808 | (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| | |
809 | ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); | |
34dc7c2f | 810 | |
428870ff BB |
811 | ZFS_TIME_ENCODE(&now, crtime); |
812 | ZFS_TIME_ENCODE(&now, ctime); | |
34dc7c2f | 813 | |
3558fd73 | 814 | if (vap->va_mask & ATTR_ATIME) { |
428870ff | 815 | ZFS_TIME_ENCODE(&vap->va_atime, atime); |
34dc7c2f | 816 | } else { |
428870ff | 817 | ZFS_TIME_ENCODE(&now, atime); |
34dc7c2f BB |
818 | } |
819 | ||
3558fd73 | 820 | if (vap->va_mask & ATTR_MTIME) { |
428870ff BB |
821 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
822 | } else { | |
823 | ZFS_TIME_ENCODE(&now, mtime); | |
824 | } | |
825 | ||
826 | /* Now add in all of the "SA" attributes */ | |
0037b49e | 827 | VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, |
428870ff BB |
828 | &sa_hdl)); |
829 | ||
830 | /* | |
831 | * Setup the array of attributes to be replaced/set on the new file | |
832 | * | |
833 | * order for DMU_OT_ZNODE is critical since it needs to be constructed | |
834 | * in the old znode_phys_t format. Don't change this ordering | |
835 | */ | |
79c76d5b | 836 | sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); |
428870ff BB |
837 | |
838 | if (obj_type == DMU_OT_ZNODE) { | |
0037b49e | 839 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), |
428870ff | 840 | NULL, &atime, 16); |
0037b49e | 841 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), |
428870ff | 842 | NULL, &mtime, 16); |
0037b49e | 843 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), |
428870ff | 844 | NULL, &ctime, 16); |
0037b49e | 845 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), |
428870ff | 846 | NULL, &crtime, 16); |
0037b49e | 847 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), |
428870ff | 848 | NULL, &gen, 8); |
0037b49e | 849 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), |
428870ff | 850 | NULL, &mode, 8); |
0037b49e | 851 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), |
428870ff | 852 | NULL, &size, 8); |
0037b49e | 853 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), |
428870ff | 854 | NULL, &parent, 8); |
34dc7c2f | 855 | } else { |
0037b49e | 856 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), |
428870ff | 857 | NULL, &mode, 8); |
0037b49e | 858 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), |
428870ff | 859 | NULL, &size, 8); |
0037b49e | 860 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), |
428870ff | 861 | NULL, &gen, 8); |
0037b49e | 862 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), |
3558fd73 | 863 | NULL, &acl_ids->z_fuid, 8); |
0037b49e | 864 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), |
3558fd73 | 865 | NULL, &acl_ids->z_fgid, 8); |
0037b49e | 866 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), |
428870ff | 867 | NULL, &parent, 8); |
0037b49e | 868 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 869 | NULL, &pflags, 8); |
0037b49e | 870 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), |
428870ff | 871 | NULL, &atime, 16); |
0037b49e | 872 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), |
428870ff | 873 | NULL, &mtime, 16); |
0037b49e | 874 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), |
428870ff | 875 | NULL, &ctime, 16); |
0037b49e | 876 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), |
428870ff BB |
877 | NULL, &crtime, 16); |
878 | } | |
879 | ||
0037b49e | 880 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); |
428870ff BB |
881 | |
882 | if (obj_type == DMU_OT_ZNODE) { | |
0037b49e | 883 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, |
428870ff | 884 | &empty_xattr, 8); |
9c5167d1 NF |
885 | } else if (dmu_objset_projectquota_enabled(zfsvfs->z_os) && |
886 | pflags & ZFS_PROJID) { | |
887 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PROJID(zfsvfs), | |
888 | NULL, &projid, 8); | |
34dc7c2f | 889 | } |
428870ff | 890 | if (obj_type == DMU_OT_ZNODE || |
aa6d8c10 | 891 | (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) { |
0037b49e | 892 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs), |
428870ff | 893 | NULL, &rdev, 8); |
428870ff BB |
894 | } |
895 | if (obj_type == DMU_OT_ZNODE) { | |
0037b49e | 896 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 897 | NULL, &pflags, 8); |
0037b49e | 898 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 899 | &acl_ids->z_fuid, 8); |
0037b49e | 900 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, |
428870ff | 901 | &acl_ids->z_fgid, 8); |
0037b49e | 902 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad, |
428870ff | 903 | sizeof (uint64_t) * 4); |
0037b49e | 904 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, |
428870ff BB |
905 | &acl_phys, sizeof (zfs_acl_phys_t)); |
906 | } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { | |
0037b49e | 907 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL, |
428870ff BB |
908 | &acl_ids->z_aclp->z_acl_count, 8); |
909 | locate.cb_aclp = acl_ids->z_aclp; | |
0037b49e | 910 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs), |
428870ff BB |
911 | zfs_acl_data_locator, &locate, |
912 | acl_ids->z_aclp->z_acl_bytes); | |
913 | mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, | |
914 | acl_ids->z_fuid, acl_ids->z_fgid); | |
915 | } | |
916 | ||
917 | VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); | |
34dc7c2f | 918 | |
34dc7c2f | 919 | if (!(flag & IS_ROOT_NODE)) { |
8d703987 BB |
920 | /* |
921 | * The call to zfs_znode_alloc() may fail if memory is low | |
922 | * via the call path: alloc_inode() -> inode_init_always() -> | |
923 | * security_inode_alloc() -> inode_alloc_security(). Since | |
924 | * the existing code is written such that zfs_mknode() can | |
925 | * not fail retry until sufficient memory has been reclaimed. | |
926 | */ | |
927 | do { | |
928 | *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, obj, | |
929 | sa_hdl); | |
930 | } while (*zpp == NULL); | |
931 | ||
7b3e34ba BB |
932 | VERIFY(*zpp != NULL); |
933 | VERIFY(dzp != NULL); | |
34dc7c2f BB |
934 | } else { |
935 | /* | |
936 | * If we are creating the root node, the "parent" we | |
937 | * passed in is the znode for the root. | |
938 | */ | |
939 | *zpp = dzp; | |
428870ff BB |
940 | |
941 | (*zpp)->z_sa_hdl = sa_hdl; | |
34dc7c2f | 942 | } |
428870ff BB |
943 | |
944 | (*zpp)->z_pflags = pflags; | |
12fa7f34 | 945 | (*zpp)->z_mode = ZTOI(*zpp)->i_mode = mode; |
50c957f7 | 946 | (*zpp)->z_dnodesize = dnodesize; |
9c5167d1 | 947 | (*zpp)->z_projid = projid; |
428870ff | 948 | |
428870ff BB |
949 | if (obj_type == DMU_OT_ZNODE || |
950 | acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { | |
b0bc7a84 | 951 | VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); |
428870ff | 952 | } |
d1d7e268 | 953 | kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); |
0037b49e | 954 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
955 | } |
956 | ||
5484965a | 957 | /* |
d3cc8b15 WA |
958 | * Update in-core attributes. It is assumed the caller will be doing an |
959 | * sa_bulk_update to push the changes out. | |
5484965a BB |
960 | */ |
961 | void | |
962 | zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) | |
963 | { | |
964 | xoptattr_t *xoap; | |
7bb1325f | 965 | boolean_t update_inode = B_FALSE; |
5484965a BB |
966 | |
967 | xoap = xva_getxoptattr(xvap); | |
968 | ASSERT(xoap); | |
969 | ||
970 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { | |
971 | uint64_t times[2]; | |
972 | ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); | |
973 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)), | |
974 | ×, sizeof (times), tx); | |
975 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
976 | } | |
977 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
978 | ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, | |
979 | zp->z_pflags, tx); | |
980 | XVA_SET_RTN(xvap, XAT_READONLY); | |
981 | } | |
982 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
983 | ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, | |
984 | zp->z_pflags, tx); | |
985 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
986 | } | |
987 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
988 | ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, | |
989 | zp->z_pflags, tx); | |
990 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
991 | } | |
992 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
993 | ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, | |
994 | zp->z_pflags, tx); | |
995 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
996 | } | |
997 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
998 | ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, | |
999 | zp->z_pflags, tx); | |
1000 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
64c688d7 | 1001 | |
7bb1325f | 1002 | update_inode = B_TRUE; |
5484965a BB |
1003 | } |
1004 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
1005 | ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, | |
1006 | zp->z_pflags, tx); | |
1007 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
1008 | } | |
1009 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
1010 | ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, | |
1011 | zp->z_pflags, tx); | |
1012 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
64c688d7 | 1013 | |
7bb1325f | 1014 | update_inode = B_TRUE; |
5484965a BB |
1015 | } |
1016 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
1017 | ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, | |
1018 | zp->z_pflags, tx); | |
1019 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
1020 | } | |
1021 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
1022 | ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, | |
1023 | zp->z_pflags, tx); | |
1024 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
1025 | } | |
1026 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
1027 | ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, | |
1028 | xoap->xoa_av_quarantined, zp->z_pflags, tx); | |
1029 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
1030 | } | |
1031 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
1032 | ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, | |
1033 | zp->z_pflags, tx); | |
1034 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
1035 | } | |
1036 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { | |
1037 | zfs_sa_set_scanstamp(zp, xvap, tx); | |
1038 | XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); | |
1039 | } | |
1040 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
1041 | ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, | |
1042 | zp->z_pflags, tx); | |
1043 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
1044 | } | |
1045 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
1046 | ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, | |
1047 | zp->z_pflags, tx); | |
1048 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
1049 | } | |
1050 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
1051 | ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, | |
1052 | zp->z_pflags, tx); | |
1053 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
1054 | } | |
9c5167d1 NF |
1055 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { |
1056 | ZFS_ATTR_SET(zp, ZFS_PROJINHERIT, xoap->xoa_projinherit, | |
1057 | zp->z_pflags, tx); | |
1058 | XVA_SET_RTN(xvap, XAT_PROJINHERIT); | |
1059 | } | |
7bb1325f CC |
1060 | |
1061 | if (update_inode) | |
1062 | zfs_set_inode_flags(zp, ZTOI(zp)); | |
5484965a BB |
1063 | } |
1064 | ||
34dc7c2f | 1065 | int |
0037b49e | 1066 | zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) |
34dc7c2f BB |
1067 | { |
1068 | dmu_object_info_t doi; | |
1069 | dmu_buf_t *db; | |
1070 | znode_t *zp; | |
c96c36fa | 1071 | znode_hold_t *zh; |
34dc7c2f | 1072 | int err; |
428870ff | 1073 | sa_handle_t *hdl; |
34dc7c2f BB |
1074 | |
1075 | *zpp = NULL; | |
1076 | ||
6f9548c4 | 1077 | again: |
0037b49e | 1078 | zh = zfs_znode_hold_enter(zfsvfs, obj_num); |
34dc7c2f | 1079 | |
0037b49e | 1080 | err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); |
34dc7c2f | 1081 | if (err) { |
0037b49e | 1082 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1083 | return (err); |
1084 | } | |
1085 | ||
1086 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1087 | if (doi.doi_bonus_type != DMU_OT_SA && |
1088 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1089 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1090 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1091 | sa_buf_rele(db, NULL); | |
0037b49e | 1092 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1093 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1094 | } |
1095 | ||
428870ff BB |
1096 | hdl = dmu_buf_get_user(db); |
1097 | if (hdl != NULL) { | |
36df2843 | 1098 | zp = sa_get_userdata(hdl); |
34dc7c2f | 1099 | |
8ac67298 | 1100 | |
34dc7c2f | 1101 | /* |
428870ff BB |
1102 | * Since "SA" does immediate eviction we |
1103 | * should never find a sa handle that doesn't | |
1104 | * know about the znode. | |
34dc7c2f | 1105 | */ |
428870ff BB |
1106 | |
1107 | ASSERT3P(zp, !=, NULL); | |
1108 | ||
1109 | mutex_enter(&zp->z_lock); | |
34dc7c2f | 1110 | ASSERT3U(zp->z_id, ==, obj_num); |
98701490 CC |
1111 | /* |
1112 | * If igrab() returns NULL the VFS has independently | |
1113 | * determined the inode should be evicted and has | |
1114 | * called iput_final() to start the eviction process. | |
1115 | * The SA handle is still valid but because the VFS | |
1116 | * requires that the eviction succeed we must drop | |
1117 | * our locks and references to allow the eviction to | |
1118 | * complete. The zfs_zget() may then be retried. | |
1119 | * | |
1120 | * This unlikely case could be optimized by registering | |
1121 | * a sops->drop_inode() callback. The callback would | |
1122 | * need to detect the active SA hold thereby informing | |
1123 | * the VFS that this inode should not be evicted. | |
1124 | */ | |
1125 | if (igrab(ZTOI(zp)) == NULL) { | |
1126 | mutex_exit(&zp->z_lock); | |
1127 | sa_buf_rele(db, NULL); | |
0037b49e | 1128 | zfs_znode_hold_exit(zfsvfs, zh); |
98701490 CC |
1129 | /* inode might need this to finish evict */ |
1130 | cond_resched(); | |
1131 | goto again; | |
34dc7c2f | 1132 | } |
98701490 CC |
1133 | *zpp = zp; |
1134 | err = 0; | |
34dc7c2f | 1135 | mutex_exit(&zp->z_lock); |
f3ad9cd6 | 1136 | sa_buf_rele(db, NULL); |
0037b49e | 1137 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1138 | return (err); |
1139 | } | |
1140 | ||
1141 | /* | |
3558fd73 | 1142 | * Not found create new znode/vnode but only if file exists. |
428870ff BB |
1143 | * |
1144 | * There is a small window where zfs_vget() could | |
1145 | * find this object while a file create is still in | |
1146 | * progress. This is checked for in zfs_znode_alloc() | |
1147 | * | |
1148 | * if zfs_znode_alloc() fails it will drop the hold on the | |
1149 | * bonus buffer. | |
34dc7c2f | 1150 | */ |
0037b49e | 1151 | zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size, |
31b6111f | 1152 | doi.doi_bonus_type, obj_num, NULL); |
428870ff | 1153 | if (zp == NULL) { |
2e528b49 | 1154 | err = SET_ERROR(ENOENT); |
428870ff BB |
1155 | } else { |
1156 | *zpp = zp; | |
1157 | } | |
0037b49e | 1158 | zfs_znode_hold_exit(zfsvfs, zh); |
428870ff | 1159 | return (err); |
34dc7c2f BB |
1160 | } |
1161 | ||
1162 | int | |
1163 | zfs_rezget(znode_t *zp) | |
1164 | { | |
0037b49e | 1165 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
1166 | dmu_object_info_t doi; |
1167 | dmu_buf_t *db; | |
1168 | uint64_t obj_num = zp->z_id; | |
428870ff | 1169 | uint64_t mode; |
dfbc8630 | 1170 | uint64_t links; |
9f5f0019 | 1171 | sa_bulk_attr_t bulk[10]; |
34dc7c2f | 1172 | int err; |
428870ff BB |
1173 | int count = 0; |
1174 | uint64_t gen; | |
2c6abf15 | 1175 | uint64_t z_uid, z_gid; |
9f5f0019 | 1176 | uint64_t atime[2], mtime[2], ctime[2]; |
9c5167d1 | 1177 | uint64_t projid = ZFS_DEFAULT_PROJID; |
c96c36fa | 1178 | znode_hold_t *zh; |
34dc7c2f | 1179 | |
cbecb4fb CC |
1180 | /* |
1181 | * skip ctldir, otherwise they will always get invalidated. This will | |
1182 | * cause funny behaviour for the mounted snapdirs. Especially for | |
1183 | * Linux >= 3.18, d_invalidate will detach the mountpoint and prevent | |
1184 | * anyone automount it again as long as someone is still using the | |
1185 | * detached mount. | |
1186 | */ | |
1187 | if (zp->z_is_ctldir) | |
1188 | return (0); | |
1189 | ||
0037b49e | 1190 | zh = zfs_znode_hold_enter(zfsvfs, obj_num); |
34dc7c2f | 1191 | |
428870ff BB |
1192 | mutex_enter(&zp->z_acl_lock); |
1193 | if (zp->z_acl_cached) { | |
1194 | zfs_acl_free(zp->z_acl_cached); | |
1195 | zp->z_acl_cached = NULL; | |
1196 | } | |
428870ff | 1197 | mutex_exit(&zp->z_acl_lock); |
7b3e34ba | 1198 | |
228b461b | 1199 | rw_enter(&zp->z_xattr_lock, RW_WRITER); |
7b3e34ba BB |
1200 | if (zp->z_xattr_cached) { |
1201 | nvlist_free(zp->z_xattr_cached); | |
1202 | zp->z_xattr_cached = NULL; | |
1203 | } | |
7b3e34ba BB |
1204 | rw_exit(&zp->z_xattr_lock); |
1205 | ||
428870ff | 1206 | ASSERT(zp->z_sa_hdl == NULL); |
0037b49e | 1207 | err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); |
34dc7c2f | 1208 | if (err) { |
0037b49e | 1209 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1210 | return (err); |
1211 | } | |
1212 | ||
1213 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1214 | if (doi.doi_bonus_type != DMU_OT_SA && |
1215 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1216 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1217 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1218 | sa_buf_rele(db, NULL); | |
0037b49e | 1219 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1220 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1221 | } |
1222 | ||
0037b49e | 1223 | zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL); |
428870ff BB |
1224 | |
1225 | /* reload cached values */ | |
0037b49e | 1226 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, |
428870ff | 1227 | &gen, sizeof (gen)); |
0037b49e | 1228 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, |
428870ff | 1229 | &zp->z_size, sizeof (zp->z_size)); |
0037b49e | 1230 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, |
dfbc8630 | 1231 | &links, sizeof (links)); |
0037b49e | 1232 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff | 1233 | &zp->z_pflags, sizeof (zp->z_pflags)); |
0037b49e | 1234 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
2c6abf15 | 1235 | &z_uid, sizeof (z_uid)); |
0037b49e | 1236 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, |
2c6abf15 | 1237 | &z_gid, sizeof (z_gid)); |
0037b49e | 1238 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 1239 | &mode, sizeof (mode)); |
0037b49e | 1240 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
9f5f0019 | 1241 | &atime, 16); |
0037b49e | 1242 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
9f5f0019 | 1243 | &mtime, 16); |
0037b49e | 1244 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
9f5f0019 | 1245 | &ctime, 16); |
428870ff | 1246 | |
428870ff BB |
1247 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { |
1248 | zfs_znode_dmu_fini(zp); | |
0037b49e | 1249 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1250 | return (SET_ERROR(EIO)); |
428870ff BB |
1251 | } |
1252 | ||
9c5167d1 NF |
1253 | if (dmu_objset_projectquota_enabled(zfsvfs->z_os)) { |
1254 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), | |
1255 | &projid, 8); | |
1256 | if (err != 0 && err != ENOENT) { | |
1257 | zfs_znode_dmu_fini(zp); | |
1258 | zfs_znode_hold_exit(zfsvfs, zh); | |
1259 | return (SET_ERROR(err)); | |
1260 | } | |
1261 | } | |
1262 | ||
1263 | zp->z_projid = projid; | |
12fa7f34 | 1264 | zp->z_mode = ZTOI(zp)->i_mode = mode; |
2c6abf15 NB |
1265 | zfs_uid_write(ZTOI(zp), z_uid); |
1266 | zfs_gid_write(ZTOI(zp), z_gid); | |
572e2857 | 1267 | |
9f5f0019 NB |
1268 | ZFS_TIME_DECODE(&ZTOI(zp)->i_atime, atime); |
1269 | ZFS_TIME_DECODE(&ZTOI(zp)->i_mtime, mtime); | |
1270 | ZFS_TIME_DECODE(&ZTOI(zp)->i_ctime, ctime); | |
1271 | ||
278f2236 | 1272 | if (gen != ZTOI(zp)->i_generation) { |
428870ff | 1273 | zfs_znode_dmu_fini(zp); |
0037b49e | 1274 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1275 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1276 | } |
1277 | ||
dfbc8630 | 1278 | set_nlink(ZTOI(zp), (uint32_t)links); |
7bb1325f | 1279 | zfs_set_inode_flags(zp, ZTOI(zp)); |
dfbc8630 | 1280 | |
34dc7c2f | 1281 | zp->z_blksz = doi.doi_data_block_size; |
704cd075 | 1282 | zp->z_atime_dirty = 0; |
9f5f0019 | 1283 | zfs_inode_update(zp); |
34dc7c2f | 1284 | |
6a218566 AG |
1285 | /* |
1286 | * If the file has zero links, then it has been unlinked on the send | |
1287 | * side and it must be in the received unlinked set. | |
1288 | * We call zfs_znode_dmu_fini() now to prevent any accesses to the | |
1289 | * stale data and to prevent automatical removal of the file in | |
1290 | * zfs_zinactive(). The file will be removed either when it is removed | |
1291 | * on the send side and the next incremental stream is received or | |
1292 | * when the unlinked set gets processed. | |
1293 | */ | |
1294 | zp->z_unlinked = (ZTOI(zp)->i_nlink == 0); | |
1295 | if (zp->z_unlinked) | |
1296 | zfs_znode_dmu_fini(zp); | |
1297 | ||
0037b49e | 1298 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1299 | |
1300 | return (0); | |
1301 | } | |
1302 | ||
1303 | void | |
1304 | zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) | |
1305 | { | |
0037b49e BB |
1306 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
1307 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 1308 | uint64_t obj = zp->z_id; |
572e2857 | 1309 | uint64_t acl_obj = zfs_external_acl(zp); |
c96c36fa | 1310 | znode_hold_t *zh; |
34dc7c2f | 1311 | |
0037b49e | 1312 | zh = zfs_znode_hold_enter(zfsvfs, obj); |
572e2857 BB |
1313 | if (acl_obj) { |
1314 | VERIFY(!zp->z_is_sa); | |
b128c09f | 1315 | VERIFY(0 == dmu_object_free(os, acl_obj, tx)); |
572e2857 | 1316 | } |
b128c09f | 1317 | VERIFY(0 == dmu_object_free(os, obj, tx)); |
34dc7c2f | 1318 | zfs_znode_dmu_fini(zp); |
0037b49e | 1319 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1320 | } |
1321 | ||
1322 | void | |
1323 | zfs_zinactive(znode_t *zp) | |
1324 | { | |
0037b49e | 1325 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f | 1326 | uint64_t z_id = zp->z_id; |
c96c36fa | 1327 | znode_hold_t *zh; |
34dc7c2f | 1328 | |
428870ff | 1329 | ASSERT(zp->z_sa_hdl); |
34dc7c2f BB |
1330 | |
1331 | /* | |
d6bd8eaa | 1332 | * Don't allow a zfs_zget() while were trying to release this znode. |
34dc7c2f | 1333 | */ |
0037b49e | 1334 | zh = zfs_znode_hold_enter(zfsvfs, z_id); |
d6bd8eaa | 1335 | |
34dc7c2f | 1336 | mutex_enter(&zp->z_lock); |
34dc7c2f BB |
1337 | |
1338 | /* | |
6a218566 AG |
1339 | * If this was the last reference to a file with no links, remove |
1340 | * the file from the file system unless the file system is mounted | |
1341 | * read-only. That can happen, for example, if the file system was | |
1342 | * originally read-write, the file was opened, then unlinked and | |
1343 | * the file system was made read-only before the file was finally | |
1344 | * closed. The file will remain in the unlinked set. | |
34dc7c2f BB |
1345 | */ |
1346 | if (zp->z_unlinked) { | |
6a218566 | 1347 | ASSERT(!zfsvfs->z_issnap); |
dcec0a12 | 1348 | if (!zfs_is_readonly(zfsvfs) && !zfs_unlink_suspend_progress) { |
6a218566 AG |
1349 | mutex_exit(&zp->z_lock); |
1350 | zfs_znode_hold_exit(zfsvfs, zh); | |
1351 | zfs_rmnode(zp); | |
1352 | return; | |
1353 | } | |
34dc7c2f | 1354 | } |
428870ff | 1355 | |
34dc7c2f BB |
1356 | mutex_exit(&zp->z_lock); |
1357 | zfs_znode_dmu_fini(zp); | |
d6bd8eaa | 1358 | |
0037b49e | 1359 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1360 | } |
1361 | ||
6d111134 TC |
1362 | static inline int |
1363 | zfs_compare_timespec(struct timespec *t1, struct timespec *t2) | |
1364 | { | |
1365 | if (t1->tv_sec < t2->tv_sec) | |
1366 | return (-1); | |
1367 | ||
1368 | if (t1->tv_sec > t2->tv_sec) | |
1369 | return (1); | |
1370 | ||
1371 | return (t1->tv_nsec - t2->tv_nsec); | |
1372 | } | |
1373 | ||
6d111134 TC |
1374 | /* |
1375 | * Prepare to update znode time stamps. | |
1376 | * | |
1377 | * IN: zp - znode requiring timestamp update | |
0df9673f | 1378 | * flag - ATTR_MTIME, ATTR_CTIME flags |
6d111134 | 1379 | * |
0df9673f | 1380 | * OUT: zp - z_seq |
6d111134 TC |
1381 | * mtime - new mtime |
1382 | * ctime - new ctime | |
1383 | * | |
0df9673f CC |
1384 | * Note: We don't update atime here, because we rely on Linux VFS to do |
1385 | * atime updating. | |
6d111134 | 1386 | */ |
34dc7c2f | 1387 | void |
428870ff | 1388 | zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], |
0df9673f | 1389 | uint64_t ctime[2]) |
34dc7c2f | 1390 | { |
6413c95f | 1391 | inode_timespec_t now; |
34dc7c2f | 1392 | |
34dc7c2f BB |
1393 | gethrestime(&now); |
1394 | ||
0df9673f | 1395 | zp->z_seq++; |
34dc7c2f | 1396 | |
3558fd73 | 1397 | if (flag & ATTR_MTIME) { |
428870ff | 1398 | ZFS_TIME_ENCODE(&now, mtime); |
9f5f0019 | 1399 | ZFS_TIME_DECODE(&(ZTOI(zp)->i_mtime), mtime); |
3558fd73 | 1400 | if (ZTOZSB(zp)->z_use_fuids) { |
428870ff BB |
1401 | zp->z_pflags |= (ZFS_ARCHIVE | |
1402 | ZFS_AV_MODIFIED); | |
1403 | } | |
34dc7c2f BB |
1404 | } |
1405 | ||
3558fd73 | 1406 | if (flag & ATTR_CTIME) { |
428870ff | 1407 | ZFS_TIME_ENCODE(&now, ctime); |
9f5f0019 | 1408 | ZFS_TIME_DECODE(&(ZTOI(zp)->i_ctime), ctime); |
3558fd73 | 1409 | if (ZTOZSB(zp)->z_use_fuids) |
428870ff | 1410 | zp->z_pflags |= ZFS_ARCHIVE; |
34dc7c2f BB |
1411 | } |
1412 | } | |
1413 | ||
34dc7c2f BB |
1414 | /* |
1415 | * Grow the block size for a file. | |
1416 | * | |
1417 | * IN: zp - znode of file to free data in. | |
1418 | * size - requested block size | |
1419 | * tx - open transaction. | |
1420 | * | |
1421 | * NOTE: this function assumes that the znode is write locked. | |
1422 | */ | |
1423 | void | |
1424 | zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) | |
1425 | { | |
1426 | int error; | |
1427 | u_longlong_t dummy; | |
1428 | ||
1429 | if (size <= zp->z_blksz) | |
1430 | return; | |
1431 | /* | |
1432 | * If the file size is already greater than the current blocksize, | |
1433 | * we will not grow. If there is more than one block in a file, | |
1434 | * the blocksize cannot change. | |
1435 | */ | |
428870ff | 1436 | if (zp->z_blksz && zp->z_size > zp->z_blksz) |
34dc7c2f BB |
1437 | return; |
1438 | ||
3558fd73 | 1439 | error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id, |
34dc7c2f | 1440 | size, 0, tx); |
428870ff | 1441 | |
34dc7c2f BB |
1442 | if (error == ENOTSUP) |
1443 | return; | |
c99c9001 | 1444 | ASSERT0(error); |
34dc7c2f BB |
1445 | |
1446 | /* What blocksize did we actually get? */ | |
428870ff | 1447 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); |
34dc7c2f BB |
1448 | } |
1449 | ||
34dc7c2f | 1450 | /* |
b128c09f | 1451 | * Increase the file length |
34dc7c2f BB |
1452 | * |
1453 | * IN: zp - znode of file to free data in. | |
b128c09f | 1454 | * end - new end-of-file |
34dc7c2f | 1455 | * |
19d55079 | 1456 | * RETURN: 0 on success, error code on failure |
34dc7c2f | 1457 | */ |
b128c09f BB |
1458 | static int |
1459 | zfs_extend(znode_t *zp, uint64_t end) | |
34dc7c2f | 1460 | { |
0037b49e | 1461 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
b128c09f | 1462 | dmu_tx_t *tx; |
5d43cc9a | 1463 | locked_range_t *lr; |
b128c09f | 1464 | uint64_t newblksz; |
34dc7c2f BB |
1465 | int error; |
1466 | ||
34dc7c2f | 1467 | /* |
b128c09f | 1468 | * We will change zp_size, lock the whole file. |
34dc7c2f | 1469 | */ |
5d43cc9a | 1470 | lr = rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); |
34dc7c2f BB |
1471 | |
1472 | /* | |
1473 | * Nothing to do if file already at desired length. | |
1474 | */ | |
428870ff | 1475 | if (end <= zp->z_size) { |
5d43cc9a | 1476 | rangelock_exit(lr); |
34dc7c2f BB |
1477 | return (0); |
1478 | } | |
0037b49e | 1479 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1480 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1481 | zfs_sa_upgrade_txholds(tx, zp); | |
b128c09f | 1482 | if (end > zp->z_blksz && |
0037b49e | 1483 | (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { |
34dc7c2f BB |
1484 | /* |
1485 | * We are growing the file past the current block size. | |
1486 | */ | |
3558fd73 | 1487 | if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) { |
f1512ee6 MA |
1488 | /* |
1489 | * File's blocksize is already larger than the | |
1490 | * "recordsize" property. Only let it grow to | |
1491 | * the next power of 2. | |
1492 | */ | |
34dc7c2f | 1493 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 | 1494 | newblksz = MIN(end, 1 << highbit64(zp->z_blksz)); |
34dc7c2f | 1495 | } else { |
3558fd73 | 1496 | newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz); |
34dc7c2f | 1497 | } |
b128c09f BB |
1498 | dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); |
1499 | } else { | |
1500 | newblksz = 0; | |
34dc7c2f BB |
1501 | } |
1502 | ||
384f8a09 | 1503 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 1504 | if (error) { |
34dc7c2f | 1505 | dmu_tx_abort(tx); |
5d43cc9a | 1506 | rangelock_exit(lr); |
34dc7c2f BB |
1507 | return (error); |
1508 | } | |
1509 | ||
b128c09f BB |
1510 | if (newblksz) |
1511 | zfs_grow_blocksize(zp, newblksz, tx); | |
34dc7c2f | 1512 | |
428870ff BB |
1513 | zp->z_size = end; |
1514 | ||
3558fd73 | 1515 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), |
428870ff | 1516 | &zp->z_size, sizeof (zp->z_size), tx)); |
34dc7c2f | 1517 | |
5d43cc9a | 1518 | rangelock_exit(lr); |
34dc7c2f | 1519 | |
b128c09f | 1520 | dmu_tx_commit(tx); |
34dc7c2f | 1521 | |
b128c09f BB |
1522 | return (0); |
1523 | } | |
1524 | ||
223df016 TC |
1525 | /* |
1526 | * zfs_zero_partial_page - Modeled after update_pages() but | |
1527 | * with different arguments and semantics for use by zfs_freesp(). | |
1528 | * | |
1529 | * Zeroes a piece of a single page cache entry for zp at offset | |
1530 | * start and length len. | |
1531 | * | |
1532 | * Caller must acquire a range lock on the file for the region | |
1533 | * being zeroed in order that the ARC and page cache stay in sync. | |
1534 | */ | |
1535 | static void | |
1536 | zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len) | |
1537 | { | |
1538 | struct address_space *mp = ZTOI(zp)->i_mapping; | |
1539 | struct page *pp; | |
1540 | int64_t off; | |
1541 | void *pb; | |
1542 | ||
8b1899d3 | 1543 | ASSERT((start & PAGE_MASK) == ((start + len - 1) & PAGE_MASK)); |
223df016 | 1544 | |
8b1899d3 BB |
1545 | off = start & (PAGE_SIZE - 1); |
1546 | start &= PAGE_MASK; | |
223df016 | 1547 | |
8b1899d3 | 1548 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
223df016 TC |
1549 | if (pp) { |
1550 | if (mapping_writably_mapped(mp)) | |
1551 | flush_dcache_page(pp); | |
1552 | ||
1553 | pb = kmap(pp); | |
1554 | bzero(pb + off, len); | |
1555 | kunmap(pp); | |
1556 | ||
1557 | if (mapping_writably_mapped(mp)) | |
1558 | flush_dcache_page(pp); | |
1559 | ||
1560 | mark_page_accessed(pp); | |
1561 | SetPageUptodate(pp); | |
1562 | ClearPageError(pp); | |
1563 | unlock_page(pp); | |
8b1899d3 | 1564 | put_page(pp); |
223df016 TC |
1565 | } |
1566 | } | |
1567 | ||
b128c09f BB |
1568 | /* |
1569 | * Free space in a file. | |
1570 | * | |
1571 | * IN: zp - znode of file to free data in. | |
1572 | * off - start of section to free. | |
1573 | * len - length of section to free. | |
1574 | * | |
19d55079 | 1575 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1576 | */ |
1577 | static int | |
1578 | zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) | |
1579 | { | |
0037b49e | 1580 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
5d43cc9a | 1581 | locked_range_t *lr; |
b128c09f BB |
1582 | int error; |
1583 | ||
1584 | /* | |
1585 | * Lock the range being freed. | |
1586 | */ | |
5d43cc9a | 1587 | lr = rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER); |
b128c09f BB |
1588 | |
1589 | /* | |
1590 | * Nothing to do if file already at desired length. | |
1591 | */ | |
428870ff | 1592 | if (off >= zp->z_size) { |
5d43cc9a | 1593 | rangelock_exit(lr); |
b128c09f | 1594 | return (0); |
34dc7c2f BB |
1595 | } |
1596 | ||
428870ff BB |
1597 | if (off + len > zp->z_size) |
1598 | len = zp->z_size - off; | |
b128c09f | 1599 | |
0037b49e | 1600 | error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); |
b128c09f | 1601 | |
223df016 TC |
1602 | /* |
1603 | * Zero partial page cache entries. This must be done under a | |
1604 | * range lock in order to keep the ARC and page cache in sync. | |
1605 | */ | |
1606 | if (zp->z_is_mapped) { | |
1607 | loff_t first_page, last_page, page_len; | |
1608 | loff_t first_page_offset, last_page_offset; | |
1609 | ||
1610 | /* first possible full page in hole */ | |
8b1899d3 | 1611 | first_page = (off + PAGE_SIZE - 1) >> PAGE_SHIFT; |
223df016 | 1612 | /* last page of hole */ |
8b1899d3 | 1613 | last_page = (off + len) >> PAGE_SHIFT; |
223df016 TC |
1614 | |
1615 | /* offset of first_page */ | |
8b1899d3 | 1616 | first_page_offset = first_page << PAGE_SHIFT; |
223df016 | 1617 | /* offset of last_page */ |
8b1899d3 | 1618 | last_page_offset = last_page << PAGE_SHIFT; |
223df016 | 1619 | |
cb08f063 TC |
1620 | /* truncate whole pages */ |
1621 | if (last_page_offset > first_page_offset) { | |
1622 | truncate_inode_pages_range(ZTOI(zp)->i_mapping, | |
1623 | first_page_offset, last_page_offset - 1); | |
1624 | } | |
1625 | ||
1626 | /* truncate sub-page ranges */ | |
223df016 TC |
1627 | if (first_page > last_page) { |
1628 | /* entire punched area within a single page */ | |
1629 | zfs_zero_partial_page(zp, off, len); | |
1630 | } else { | |
1631 | /* beginning of punched area at the end of a page */ | |
1632 | page_len = first_page_offset - off; | |
1633 | if (page_len > 0) | |
1634 | zfs_zero_partial_page(zp, off, page_len); | |
1635 | ||
1636 | /* end of punched area at the beginning of a page */ | |
1637 | page_len = off + len - last_page_offset; | |
1638 | if (page_len > 0) | |
1639 | zfs_zero_partial_page(zp, last_page_offset, | |
1640 | page_len); | |
1641 | } | |
1642 | } | |
5d43cc9a | 1643 | rangelock_exit(lr); |
34dc7c2f | 1644 | |
b128c09f BB |
1645 | return (error); |
1646 | } | |
1647 | ||
1648 | /* | |
1649 | * Truncate a file | |
1650 | * | |
1651 | * IN: zp - znode of file to free data in. | |
1652 | * end - new end-of-file. | |
1653 | * | |
19d55079 | 1654 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1655 | */ |
1656 | static int | |
1657 | zfs_trunc(znode_t *zp, uint64_t end) | |
1658 | { | |
0037b49e | 1659 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
b128c09f | 1660 | dmu_tx_t *tx; |
5d43cc9a | 1661 | locked_range_t *lr; |
b128c09f | 1662 | int error; |
572e2857 BB |
1663 | sa_bulk_attr_t bulk[2]; |
1664 | int count = 0; | |
b128c09f BB |
1665 | |
1666 | /* | |
1667 | * We will change zp_size, lock the whole file. | |
1668 | */ | |
5d43cc9a | 1669 | lr = rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); |
b128c09f BB |
1670 | |
1671 | /* | |
1672 | * Nothing to do if file already at desired length. | |
1673 | */ | |
428870ff | 1674 | if (end >= zp->z_size) { |
5d43cc9a | 1675 | rangelock_exit(lr); |
b128c09f BB |
1676 | return (0); |
1677 | } | |
1678 | ||
18a2485f FS |
1679 | error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, |
1680 | DMU_OBJECT_END); | |
b128c09f | 1681 | if (error) { |
5d43cc9a | 1682 | rangelock_exit(lr); |
b128c09f BB |
1683 | return (error); |
1684 | } | |
0037b49e | 1685 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1686 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1687 | zfs_sa_upgrade_txholds(tx, zp); | |
19d55079 | 1688 | dmu_tx_mark_netfree(tx); |
7a8f0e80 | 1689 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1690 | if (error) { |
b128c09f | 1691 | dmu_tx_abort(tx); |
5d43cc9a | 1692 | rangelock_exit(lr); |
b128c09f BB |
1693 | return (error); |
1694 | } | |
b128c09f | 1695 | |
428870ff | 1696 | zp->z_size = end; |
0037b49e | 1697 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), |
572e2857 | 1698 | NULL, &zp->z_size, sizeof (zp->z_size)); |
428870ff | 1699 | |
572e2857 BB |
1700 | if (end == 0) { |
1701 | zp->z_pflags &= ~ZFS_SPARSE; | |
0037b49e | 1702 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), |
572e2857 BB |
1703 | NULL, &zp->z_pflags, 8); |
1704 | } | |
1705 | VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); | |
b128c09f | 1706 | |
34dc7c2f | 1707 | dmu_tx_commit(tx); |
5d43cc9a | 1708 | rangelock_exit(lr); |
34dc7c2f BB |
1709 | |
1710 | return (0); | |
1711 | } | |
1712 | ||
b128c09f BB |
1713 | /* |
1714 | * Free space in a file | |
1715 | * | |
1716 | * IN: zp - znode of file to free data in. | |
1717 | * off - start of range | |
1718 | * len - end of range (0 => EOF) | |
1719 | * flag - current file open mode flags. | |
1720 | * log - TRUE if this action should be logged | |
1721 | * | |
19d55079 | 1722 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1723 | */ |
1724 | int | |
1725 | zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) | |
1726 | { | |
b128c09f | 1727 | dmu_tx_t *tx; |
0037b49e BB |
1728 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
1729 | zilog_t *zilog = zfsvfs->z_log; | |
428870ff BB |
1730 | uint64_t mode; |
1731 | uint64_t mtime[2], ctime[2]; | |
1732 | sa_bulk_attr_t bulk[3]; | |
1733 | int count = 0; | |
b128c09f BB |
1734 | int error; |
1735 | ||
0037b49e | 1736 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode, |
428870ff BB |
1737 | sizeof (mode))) != 0) |
1738 | return (error); | |
1739 | ||
1740 | if (off > zp->z_size) { | |
b128c09f BB |
1741 | error = zfs_extend(zp, off+len); |
1742 | if (error == 0 && log) | |
1743 | goto log; | |
223df016 | 1744 | goto out; |
b128c09f BB |
1745 | } |
1746 | ||
b128c09f BB |
1747 | if (len == 0) { |
1748 | error = zfs_trunc(zp, off); | |
1749 | } else { | |
1750 | if ((error = zfs_free_range(zp, off, len)) == 0 && | |
428870ff | 1751 | off + len > zp->z_size) |
b128c09f BB |
1752 | error = zfs_extend(zp, off+len); |
1753 | } | |
1754 | if (error || !log) | |
223df016 | 1755 | goto out; |
b128c09f | 1756 | log: |
0037b49e | 1757 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1758 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1759 | zfs_sa_upgrade_txholds(tx, zp); | |
384f8a09 | 1760 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1761 | if (error) { |
b128c09f | 1762 | dmu_tx_abort(tx); |
223df016 | 1763 | goto out; |
b128c09f BB |
1764 | } |
1765 | ||
0037b49e BB |
1766 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16); |
1767 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16); | |
1768 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), | |
428870ff | 1769 | NULL, &zp->z_pflags, 8); |
0df9673f | 1770 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
428870ff BB |
1771 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
1772 | ASSERT(error == 0); | |
1773 | ||
b128c09f BB |
1774 | zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); |
1775 | ||
1776 | dmu_tx_commit(tx); | |
223df016 | 1777 | |
960e08fe | 1778 | zfs_inode_update(zp); |
223df016 TC |
1779 | error = 0; |
1780 | ||
1781 | out: | |
1782 | /* | |
1783 | * Truncate the page cache - for file truncate operations, use | |
1784 | * the purpose-built API for truncations. For punching operations, | |
cb08f063 | 1785 | * the truncation is handled under a range lock in zfs_free_range. |
223df016 TC |
1786 | */ |
1787 | if (len == 0) | |
1788 | truncate_setsize(ZTOI(zp), off); | |
223df016 | 1789 | return (error); |
b128c09f BB |
1790 | } |
1791 | ||
34dc7c2f BB |
1792 | void |
1793 | zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) | |
1794 | { | |
22872ff5 | 1795 | struct super_block *sb; |
0037b49e | 1796 | zfsvfs_t *zfsvfs; |
428870ff | 1797 | uint64_t moid, obj, sa_obj, version; |
22872ff5 | 1798 | uint64_t sense = ZFS_CASE_SENSITIVE; |
34dc7c2f BB |
1799 | uint64_t norm = 0; |
1800 | nvpair_t *elem; | |
c96c36fa | 1801 | int size; |
34dc7c2f | 1802 | int error; |
22872ff5 BB |
1803 | int i; |
1804 | znode_t *rootzp = NULL; | |
1805 | vattr_t vattr; | |
1806 | znode_t *zp; | |
1807 | zfs_acl_ids_t acl_ids; | |
34dc7c2f BB |
1808 | |
1809 | /* | |
1810 | * First attempt to create master node. | |
1811 | */ | |
1812 | /* | |
1813 | * In an empty objset, there are no blocks to read and thus | |
1814 | * there can be no i/o errors (which we assert below). | |
1815 | */ | |
1816 | moid = MASTER_NODE_OBJ; | |
1817 | error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, | |
1818 | DMU_OT_NONE, 0, tx); | |
1819 | ASSERT(error == 0); | |
1820 | ||
1821 | /* | |
1822 | * Set starting attributes. | |
1823 | */ | |
428870ff | 1824 | version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); |
34dc7c2f BB |
1825 | elem = NULL; |
1826 | while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { | |
1827 | /* For the moment we expect all zpl props to be uint64_ts */ | |
1828 | uint64_t val; | |
1829 | char *name; | |
1830 | ||
1831 | ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); | |
1832 | VERIFY(nvpair_value_uint64(elem, &val) == 0); | |
1833 | name = nvpair_name(elem); | |
1834 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { | |
9babb374 BB |
1835 | if (val < version) |
1836 | version = val; | |
34dc7c2f BB |
1837 | } else { |
1838 | error = zap_update(os, moid, name, 8, 1, &val, tx); | |
1839 | } | |
1840 | ASSERT(error == 0); | |
1841 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) | |
1842 | norm = val; | |
22872ff5 BB |
1843 | else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) |
1844 | sense = val; | |
34dc7c2f BB |
1845 | } |
1846 | ASSERT(version != 0); | |
9babb374 | 1847 | error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); |
34dc7c2f | 1848 | |
428870ff BB |
1849 | /* |
1850 | * Create zap object used for SA attribute registration | |
1851 | */ | |
1852 | ||
1853 | if (version >= ZPL_VERSION_SA) { | |
1854 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1855 | DMU_OT_NONE, 0, tx); | |
1856 | error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
1857 | ASSERT(error == 0); | |
1858 | } else { | |
1859 | sa_obj = 0; | |
1860 | } | |
34dc7c2f BB |
1861 | /* |
1862 | * Create a delete queue. | |
1863 | */ | |
9babb374 | 1864 | obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); |
34dc7c2f | 1865 | |
9babb374 | 1866 | error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); |
34dc7c2f BB |
1867 | ASSERT(error == 0); |
1868 | ||
9babb374 | 1869 | /* |
0037b49e | 1870 | * Create root znode. Create minimal znode/inode/zfsvfs/sb |
22872ff5 | 1871 | * to allow zfs_mknode to work. |
9babb374 | 1872 | */ |
22872ff5 BB |
1873 | vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID; |
1874 | vattr.va_mode = S_IFDIR|0755; | |
1875 | vattr.va_uid = crgetuid(cr); | |
1876 | vattr.va_gid = crgetgid(cr); | |
1877 | ||
79c76d5b | 1878 | rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
22872ff5 BB |
1879 | rootzp->z_moved = 0; |
1880 | rootzp->z_unlinked = 0; | |
1881 | rootzp->z_atime_dirty = 0; | |
1882 | rootzp->z_is_sa = USE_SA(version, os); | |
9c5167d1 | 1883 | rootzp->z_pflags = 0; |
22872ff5 | 1884 | |
0037b49e BB |
1885 | zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); |
1886 | zfsvfs->z_os = os; | |
1887 | zfsvfs->z_parent = zfsvfs; | |
1888 | zfsvfs->z_version = version; | |
1889 | zfsvfs->z_use_fuids = USE_FUIDS(version, os); | |
1890 | zfsvfs->z_use_sa = USE_SA(version, os); | |
1891 | zfsvfs->z_norm = norm; | |
22872ff5 | 1892 | |
79c76d5b | 1893 | sb = kmem_zalloc(sizeof (struct super_block), KM_SLEEP); |
0037b49e | 1894 | sb->s_fs_info = zfsvfs; |
22872ff5 BB |
1895 | |
1896 | ZTOI(rootzp)->i_sb = sb; | |
1897 | ||
1898 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, | |
0037b49e | 1899 | &zfsvfs->z_attr_table); |
9babb374 | 1900 | |
22872ff5 | 1901 | ASSERT(error == 0); |
9babb374 | 1902 | |
60101509 | 1903 | /* |
22872ff5 BB |
1904 | * Fold case on file systems that are always or sometimes case |
1905 | * insensitive. | |
60101509 | 1906 | */ |
22872ff5 | 1907 | if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) |
0037b49e | 1908 | zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; |
60101509 | 1909 | |
0037b49e BB |
1910 | mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
1911 | list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), | |
22872ff5 | 1912 | offsetof(znode_t, z_link_node)); |
60101509 | 1913 | |
c96c36fa | 1914 | size = MIN(1 << (highbit64(zfs_object_mutex_size)-1), ZFS_OBJ_MTX_MAX); |
0037b49e BB |
1915 | zfsvfs->z_hold_size = size; |
1916 | zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, | |
1917 | KM_SLEEP); | |
1918 | zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
c96c36fa | 1919 | for (i = 0; i != size; i++) { |
0037b49e | 1920 | avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare, |
c96c36fa | 1921 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); |
0037b49e | 1922 | mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); |
c96c36fa | 1923 | } |
60101509 | 1924 | |
22872ff5 BB |
1925 | VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, |
1926 | cr, NULL, &acl_ids)); | |
1927 | zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); | |
1928 | ASSERT3P(zp, ==, rootzp); | |
1929 | error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); | |
1930 | ASSERT(error == 0); | |
1931 | zfs_acl_ids_free(&acl_ids); | |
60101509 | 1932 | |
22872ff5 BB |
1933 | atomic_set(&ZTOI(rootzp)->i_count, 0); |
1934 | sa_handle_destroy(rootzp->z_sa_hdl); | |
22872ff5 BB |
1935 | kmem_cache_free(znode_cache, rootzp); |
1936 | ||
c96c36fa | 1937 | for (i = 0; i != size; i++) { |
0037b49e BB |
1938 | avl_destroy(&zfsvfs->z_hold_trees[i]); |
1939 | mutex_destroy(&zfsvfs->z_hold_locks[i]); | |
c96c36fa | 1940 | } |
2708f716 | 1941 | |
c17486b2 GN |
1942 | mutex_destroy(&zfsvfs->z_znodes_lock); |
1943 | ||
0037b49e BB |
1944 | vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size); |
1945 | vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size); | |
2708f716 | 1946 | kmem_free(sb, sizeof (struct super_block)); |
0037b49e | 1947 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); |
34dc7c2f | 1948 | } |
34dc7c2f | 1949 | #endif /* _KERNEL */ |
428870ff | 1950 | |
34dc7c2f | 1951 | static int |
572e2857 BB |
1952 | zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) |
1953 | { | |
1954 | uint64_t sa_obj = 0; | |
1955 | int error; | |
1956 | ||
1957 | error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
1958 | if (error != 0 && error != ENOENT) | |
1959 | return (error); | |
1960 | ||
1961 | error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); | |
1962 | return (error); | |
1963 | } | |
1964 | ||
1965 | static int | |
1966 | zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, | |
7b8518cb | 1967 | dmu_buf_t **db, void *tag) |
34dc7c2f | 1968 | { |
34dc7c2f | 1969 | dmu_object_info_t doi; |
34dc7c2f | 1970 | int error; |
428870ff | 1971 | |
7b8518cb | 1972 | if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) |
34dc7c2f BB |
1973 | return (error); |
1974 | ||
572e2857 | 1975 | dmu_object_info_from_db(*db, &doi); |
428870ff BB |
1976 | if ((doi.doi_bonus_type != DMU_OT_SA && |
1977 | doi.doi_bonus_type != DMU_OT_ZNODE) || | |
d6320ddb BB |
1978 | (doi.doi_bonus_type == DMU_OT_ZNODE && |
1979 | doi.doi_bonus_size < sizeof (znode_phys_t))) { | |
7b8518cb | 1980 | sa_buf_rele(*db, tag); |
2e528b49 | 1981 | return (SET_ERROR(ENOTSUP)); |
34dc7c2f BB |
1982 | } |
1983 | ||
572e2857 BB |
1984 | error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); |
1985 | if (error != 0) { | |
7b8518cb | 1986 | sa_buf_rele(*db, tag); |
428870ff BB |
1987 | return (error); |
1988 | } | |
1989 | ||
572e2857 BB |
1990 | return (0); |
1991 | } | |
1992 | ||
1993 | void | |
7b8518cb | 1994 | zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) |
572e2857 BB |
1995 | { |
1996 | sa_handle_destroy(hdl); | |
7b8518cb | 1997 | sa_buf_rele(db, tag); |
572e2857 BB |
1998 | } |
1999 | ||
2000 | /* | |
2001 | * Given an object number, return its parent object number and whether | |
2002 | * or not the object is an extended attribute directory. | |
2003 | */ | |
2004 | static int | |
b23ad7f3 JJ |
2005 | zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, |
2006 | uint64_t *pobjp, int *is_xattrdir) | |
572e2857 BB |
2007 | { |
2008 | uint64_t parent; | |
2009 | uint64_t pflags; | |
2010 | uint64_t mode; | |
b23ad7f3 | 2011 | uint64_t parent_mode; |
572e2857 | 2012 | sa_bulk_attr_t bulk[3]; |
b23ad7f3 JJ |
2013 | sa_handle_t *sa_hdl; |
2014 | dmu_buf_t *sa_db; | |
572e2857 BB |
2015 | int count = 0; |
2016 | int error; | |
2017 | ||
2018 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, | |
2019 | &parent, sizeof (parent)); | |
428870ff | 2020 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, |
572e2857 | 2021 | &pflags, sizeof (pflags)); |
428870ff | 2022 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, |
572e2857 | 2023 | &mode, sizeof (mode)); |
428870ff | 2024 | |
572e2857 | 2025 | if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) |
428870ff | 2026 | return (error); |
572e2857 | 2027 | |
b23ad7f3 JJ |
2028 | /* |
2029 | * When a link is removed its parent pointer is not changed and will | |
2030 | * be invalid. There are two cases where a link is removed but the | |
2031 | * file stays around, when it goes to the delete queue and when there | |
2032 | * are additional links. | |
2033 | */ | |
2034 | error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); | |
2035 | if (error != 0) | |
2036 | return (error); | |
2037 | ||
2038 | error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); | |
2039 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); | |
2040 | if (error != 0) | |
2041 | return (error); | |
2042 | ||
428870ff | 2043 | *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); |
34dc7c2f | 2044 | |
b23ad7f3 JJ |
2045 | /* |
2046 | * Extended attributes can be applied to files, directories, etc. | |
2047 | * Otherwise the parent must be a directory. | |
2048 | */ | |
2049 | if (!*is_xattrdir && !S_ISDIR(parent_mode)) | |
ecb2b7dc | 2050 | return (SET_ERROR(EINVAL)); |
b23ad7f3 JJ |
2051 | |
2052 | *pobjp = parent; | |
2053 | ||
34dc7c2f BB |
2054 | return (0); |
2055 | } | |
2056 | ||
572e2857 BB |
2057 | /* |
2058 | * Given an object number, return some zpl level statistics | |
2059 | */ | |
2060 | static int | |
2061 | zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, | |
2062 | zfs_stat_t *sb) | |
34dc7c2f | 2063 | { |
572e2857 BB |
2064 | sa_bulk_attr_t bulk[4]; |
2065 | int count = 0; | |
2066 | ||
2067 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, | |
2068 | &sb->zs_mode, sizeof (sb->zs_mode)); | |
2069 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, | |
2070 | &sb->zs_gen, sizeof (sb->zs_gen)); | |
2071 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, | |
2072 | &sb->zs_links, sizeof (sb->zs_links)); | |
2073 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, | |
2074 | &sb->zs_ctime, sizeof (sb->zs_ctime)); | |
2075 | ||
2076 | return (sa_bulk_lookup(hdl, bulk, count)); | |
2077 | } | |
2078 | ||
2079 | static int | |
2080 | zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, | |
2081 | sa_attr_type_t *sa_table, char *buf, int len) | |
2082 | { | |
2083 | sa_handle_t *sa_hdl; | |
2084 | sa_handle_t *prevhdl = NULL; | |
2085 | dmu_buf_t *prevdb = NULL; | |
2086 | dmu_buf_t *sa_db = NULL; | |
34dc7c2f BB |
2087 | char *path = buf + len - 1; |
2088 | int error; | |
2089 | ||
2090 | *path = '\0'; | |
572e2857 | 2091 | sa_hdl = hdl; |
428870ff | 2092 | |
64c1dcef PD |
2093 | uint64_t deleteq_obj; |
2094 | VERIFY0(zap_lookup(osp, MASTER_NODE_OBJ, | |
2095 | ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj)); | |
2096 | error = zap_lookup_int(osp, deleteq_obj, obj); | |
2097 | if (error == 0) { | |
2098 | return (ESTALE); | |
2099 | } else if (error != ENOENT) { | |
2100 | return (error); | |
2101 | } | |
2102 | error = 0; | |
2103 | ||
34dc7c2f | 2104 | for (;;) { |
17897ce2 | 2105 | uint64_t pobj = 0; |
34dc7c2f BB |
2106 | char component[MAXNAMELEN + 2]; |
2107 | size_t complen; | |
17897ce2 | 2108 | int is_xattrdir = 0; |
34dc7c2f | 2109 | |
572e2857 | 2110 | if (prevdb) |
7b8518cb | 2111 | zfs_release_sa_handle(prevhdl, prevdb, FTAG); |
572e2857 | 2112 | |
b23ad7f3 | 2113 | if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, |
572e2857 | 2114 | &is_xattrdir)) != 0) |
34dc7c2f BB |
2115 | break; |
2116 | ||
2117 | if (pobj == obj) { | |
2118 | if (path[0] != '/') | |
2119 | *--path = '/'; | |
2120 | break; | |
2121 | } | |
2122 | ||
2123 | component[0] = '/'; | |
2124 | if (is_xattrdir) { | |
2125 | (void) sprintf(component + 1, "<xattrdir>"); | |
2126 | } else { | |
2127 | error = zap_value_search(osp, pobj, obj, | |
2128 | ZFS_DIRENT_OBJ(-1ULL), component + 1); | |
2129 | if (error != 0) | |
2130 | break; | |
2131 | } | |
2132 | ||
2133 | complen = strlen(component); | |
2134 | path -= complen; | |
2135 | ASSERT(path >= buf); | |
2136 | bcopy(component, path, complen); | |
2137 | obj = pobj; | |
572e2857 BB |
2138 | |
2139 | if (sa_hdl != hdl) { | |
2140 | prevhdl = sa_hdl; | |
2141 | prevdb = sa_db; | |
2142 | } | |
7b8518cb | 2143 | error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); |
572e2857 BB |
2144 | if (error != 0) { |
2145 | sa_hdl = prevhdl; | |
2146 | sa_db = prevdb; | |
2147 | break; | |
2148 | } | |
2149 | } | |
2150 | ||
2151 | if (sa_hdl != NULL && sa_hdl != hdl) { | |
2152 | ASSERT(sa_db != NULL); | |
7b8518cb | 2153 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); |
34dc7c2f BB |
2154 | } |
2155 | ||
2156 | if (error == 0) | |
2157 | (void) memmove(buf, path, buf + len - path); | |
428870ff | 2158 | |
34dc7c2f BB |
2159 | return (error); |
2160 | } | |
572e2857 BB |
2161 | |
2162 | int | |
2163 | zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) | |
2164 | { | |
2165 | sa_attr_type_t *sa_table; | |
2166 | sa_handle_t *hdl; | |
2167 | dmu_buf_t *db; | |
2168 | int error; | |
2169 | ||
2170 | error = zfs_sa_setup(osp, &sa_table); | |
2171 | if (error != 0) | |
2172 | return (error); | |
2173 | ||
7b8518cb | 2174 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2175 | if (error != 0) |
2176 | return (error); | |
2177 | ||
2178 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2179 | ||
7b8518cb | 2180 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2181 | return (error); |
2182 | } | |
2183 | ||
2184 | int | |
2185 | zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, | |
2186 | char *buf, int len) | |
2187 | { | |
2188 | char *path = buf + len - 1; | |
2189 | sa_attr_type_t *sa_table; | |
2190 | sa_handle_t *hdl; | |
2191 | dmu_buf_t *db; | |
2192 | int error; | |
2193 | ||
2194 | *path = '\0'; | |
2195 | ||
2196 | error = zfs_sa_setup(osp, &sa_table); | |
2197 | if (error != 0) | |
2198 | return (error); | |
2199 | ||
7b8518cb | 2200 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2201 | if (error != 0) |
2202 | return (error); | |
2203 | ||
2204 | error = zfs_obj_to_stats_impl(hdl, sa_table, sb); | |
2205 | if (error != 0) { | |
7b8518cb | 2206 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2207 | return (error); |
2208 | } | |
2209 | ||
2210 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2211 | ||
7b8518cb | 2212 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2213 | return (error); |
2214 | } | |
c28b2279 | 2215 | |
93ce2b4c | 2216 | #if defined(_KERNEL) |
c28b2279 BB |
2217 | EXPORT_SYMBOL(zfs_create_fs); |
2218 | EXPORT_SYMBOL(zfs_obj_to_path); | |
0720116d | 2219 | |
02730c33 | 2220 | /* CSTYLED */ |
0720116d BB |
2221 | module_param(zfs_object_mutex_size, uint, 0644); |
2222 | MODULE_PARM_DESC(zfs_object_mutex_size, "Size of znode hold array"); | |
dcec0a12 AP |
2223 | module_param(zfs_unlink_suspend_progress, int, 0644); |
2224 | MODULE_PARM_DESC(zfs_unlink_suspend_progress, "Set to prevent async unlinks " | |
2225 | "(debug - leaks space into the unlinked set)"); | |
c28b2279 | 2226 | #endif |