]>
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 BB |
68 | |
69 | /* | |
70 | * Functions needed for userland (ie: libzpool) are not put under | |
71 | * #ifdef_KERNEL; the rest of the functions have dependencies | |
72 | * (such as VFS logic) that will not compile easily in userland. | |
73 | */ | |
74 | #ifdef _KERNEL | |
9babb374 | 75 | |
b128c09f | 76 | static kmem_cache_t *znode_cache = NULL; |
c96c36fa | 77 | static kmem_cache_t *znode_hold_cache = NULL; |
0720116d | 78 | unsigned int zfs_object_mutex_size = ZFS_OBJ_MTX_SZ; |
34dc7c2f | 79 | |
dcec0a12 AP |
80 | /* |
81 | * This is used by the test suite so that it can delay znodes from being | |
82 | * freed in order to inspect the unlinked set. | |
83 | */ | |
84 | int zfs_unlink_suspend_progress = 0; | |
85 | ||
5d43cc9a MA |
86 | /* |
87 | * This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on | |
88 | * z_rangelock. It will modify the offset and length of the lock to reflect | |
89 | * znode-specific information, and convert RL_APPEND to RL_WRITER. This is | |
90 | * called with the rangelock_t's rl_lock held, which avoids races. | |
91 | */ | |
92 | static void | |
93 | zfs_rangelock_cb(locked_range_t *new, void *arg) | |
94 | { | |
95 | znode_t *zp = arg; | |
96 | ||
97 | /* | |
98 | * If in append mode, convert to writer and lock starting at the | |
99 | * current end of file. | |
100 | */ | |
101 | if (new->lr_type == RL_APPEND) { | |
102 | new->lr_offset = zp->z_size; | |
103 | new->lr_type = RL_WRITER; | |
104 | } | |
105 | ||
106 | /* | |
107 | * If we need to grow the block size then lock the whole file range. | |
108 | */ | |
109 | uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length); | |
110 | if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) || | |
111 | zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) { | |
112 | new->lr_offset = 0; | |
113 | new->lr_length = UINT64_MAX; | |
114 | } | |
115 | } | |
116 | ||
34dc7c2f BB |
117 | /*ARGSUSED*/ |
118 | static int | |
b128c09f | 119 | zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) |
34dc7c2f BB |
120 | { |
121 | znode_t *zp = buf; | |
122 | ||
3558fd73 | 123 | inode_init_once(ZTOI(zp)); |
b128c09f BB |
124 | list_link_init(&zp->z_link_node); |
125 | ||
34dc7c2f | 126 | mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 127 | rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); |
448d7aaa | 128 | rw_init(&zp->z_name_lock, NULL, RW_NOLOCKDEP, NULL); |
34dc7c2f | 129 | mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); |
82a37189 | 130 | rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); |
34dc7c2f | 131 | |
5d43cc9a | 132 | rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp); |
34dc7c2f | 133 | |
b128c09f | 134 | zp->z_dirlocks = NULL; |
45d1cae3 | 135 | zp->z_acl_cached = NULL; |
82a37189 | 136 | zp->z_xattr_cached = NULL; |
98701490 | 137 | zp->z_xattr_parent = 0; |
572e2857 | 138 | zp->z_moved = 0; |
34dc7c2f BB |
139 | return (0); |
140 | } | |
141 | ||
142 | /*ARGSUSED*/ | |
143 | static void | |
b128c09f | 144 | zfs_znode_cache_destructor(void *buf, void *arg) |
34dc7c2f BB |
145 | { |
146 | znode_t *zp = buf; | |
147 | ||
b128c09f | 148 | ASSERT(!list_link_active(&zp->z_link_node)); |
34dc7c2f | 149 | mutex_destroy(&zp->z_lock); |
34dc7c2f BB |
150 | rw_destroy(&zp->z_parent_lock); |
151 | rw_destroy(&zp->z_name_lock); | |
152 | mutex_destroy(&zp->z_acl_lock); | |
82a37189 | 153 | rw_destroy(&zp->z_xattr_lock); |
5d43cc9a | 154 | rangelock_fini(&zp->z_rangelock); |
34dc7c2f | 155 | |
b128c09f | 156 | ASSERT(zp->z_dirlocks == NULL); |
45d1cae3 | 157 | ASSERT(zp->z_acl_cached == NULL); |
82a37189 | 158 | ASSERT(zp->z_xattr_cached == NULL); |
b128c09f BB |
159 | } |
160 | ||
c96c36fa BB |
161 | static int |
162 | zfs_znode_hold_cache_constructor(void *buf, void *arg, int kmflags) | |
163 | { | |
164 | znode_hold_t *zh = buf; | |
165 | ||
166 | mutex_init(&zh->zh_lock, NULL, MUTEX_DEFAULT, NULL); | |
424fd7c3 | 167 | zfs_refcount_create(&zh->zh_refcount); |
c96c36fa BB |
168 | zh->zh_obj = ZFS_NO_OBJECT; |
169 | ||
170 | return (0); | |
171 | } | |
172 | ||
173 | static void | |
174 | zfs_znode_hold_cache_destructor(void *buf, void *arg) | |
175 | { | |
176 | znode_hold_t *zh = buf; | |
177 | ||
178 | mutex_destroy(&zh->zh_lock); | |
424fd7c3 | 179 | zfs_refcount_destroy(&zh->zh_refcount); |
c96c36fa BB |
180 | } |
181 | ||
34dc7c2f BB |
182 | void |
183 | zfs_znode_init(void) | |
184 | { | |
185 | /* | |
5074bfe8 TC |
186 | * Initialize zcache. The KMC_SLAB hint is used in order that it be |
187 | * backed by kmalloc() when on the Linux slab in order that any | |
188 | * wait_on_bit() operations on the related inode operate properly. | |
34dc7c2f BB |
189 | */ |
190 | ASSERT(znode_cache == NULL); | |
191 | znode_cache = kmem_cache_create("zfs_znode_cache", | |
192 | sizeof (znode_t), 0, zfs_znode_cache_constructor, | |
5074bfe8 | 193 | zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_SLAB); |
c96c36fa BB |
194 | |
195 | ASSERT(znode_hold_cache == NULL); | |
196 | znode_hold_cache = kmem_cache_create("zfs_znode_hold_cache", | |
197 | sizeof (znode_hold_t), 0, zfs_znode_hold_cache_constructor, | |
198 | zfs_znode_hold_cache_destructor, NULL, NULL, NULL, 0); | |
34dc7c2f BB |
199 | } |
200 | ||
201 | void | |
202 | zfs_znode_fini(void) | |
203 | { | |
34dc7c2f BB |
204 | /* |
205 | * Cleanup zcache | |
206 | */ | |
207 | if (znode_cache) | |
208 | kmem_cache_destroy(znode_cache); | |
209 | znode_cache = NULL; | |
c96c36fa BB |
210 | |
211 | if (znode_hold_cache) | |
212 | kmem_cache_destroy(znode_hold_cache); | |
213 | znode_hold_cache = NULL; | |
214 | } | |
215 | ||
216 | /* | |
217 | * The zfs_znode_hold_enter() / zfs_znode_hold_exit() functions are used to | |
218 | * serialize access to a znode and its SA buffer while the object is being | |
219 | * created or destroyed. This kind of locking would normally reside in the | |
220 | * znode itself but in this case that's impossible because the znode and SA | |
221 | * buffer may not yet exist. Therefore the locking is handled externally | |
222 | * with an array of mutexs and AVLs trees which contain per-object locks. | |
223 | * | |
224 | * In zfs_znode_hold_enter() a per-object lock is created as needed, inserted | |
225 | * in to the correct AVL tree and finally the per-object lock is held. In | |
226 | * zfs_znode_hold_exit() the process is reversed. The per-object lock is | |
227 | * released, removed from the AVL tree and destroyed if there are no waiters. | |
228 | * | |
229 | * This scheme has two important properties: | |
230 | * | |
231 | * 1) No memory allocations are performed while holding one of the z_hold_locks. | |
232 | * This ensures evict(), which can be called from direct memory reclaim, will | |
233 | * never block waiting on a z_hold_locks which just happens to have hashed | |
234 | * to the same index. | |
235 | * | |
236 | * 2) All locks used to serialize access to an object are per-object and never | |
237 | * shared. This minimizes lock contention without creating a large number | |
238 | * of dedicated locks. | |
239 | * | |
240 | * On the downside it does require znode_lock_t structures to be frequently | |
241 | * allocated and freed. However, because these are backed by a kmem cache | |
242 | * and very short lived this cost is minimal. | |
243 | */ | |
244 | int | |
245 | zfs_znode_hold_compare(const void *a, const void *b) | |
246 | { | |
ee36c709 GN |
247 | const znode_hold_t *zh_a = (const znode_hold_t *)a; |
248 | const znode_hold_t *zh_b = (const znode_hold_t *)b; | |
249 | ||
250 | return (AVL_CMP(zh_a->zh_obj, zh_b->zh_obj)); | |
c96c36fa BB |
251 | } |
252 | ||
253 | boolean_t | |
0037b49e | 254 | zfs_znode_held(zfsvfs_t *zfsvfs, uint64_t obj) |
c96c36fa BB |
255 | { |
256 | znode_hold_t *zh, search; | |
0037b49e | 257 | int i = ZFS_OBJ_HASH(zfsvfs, obj); |
37c56346 | 258 | boolean_t held; |
c96c36fa BB |
259 | |
260 | search.zh_obj = obj; | |
261 | ||
0037b49e BB |
262 | mutex_enter(&zfsvfs->z_hold_locks[i]); |
263 | zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL); | |
37c56346 | 264 | held = (zh && MUTEX_HELD(&zh->zh_lock)) ? B_TRUE : B_FALSE; |
0037b49e | 265 | mutex_exit(&zfsvfs->z_hold_locks[i]); |
c96c36fa | 266 | |
37c56346 | 267 | return (held); |
c96c36fa BB |
268 | } |
269 | ||
270 | static znode_hold_t * | |
0037b49e | 271 | zfs_znode_hold_enter(zfsvfs_t *zfsvfs, uint64_t obj) |
c96c36fa BB |
272 | { |
273 | znode_hold_t *zh, *zh_new, search; | |
0037b49e | 274 | int i = ZFS_OBJ_HASH(zfsvfs, obj); |
c96c36fa BB |
275 | boolean_t found = B_FALSE; |
276 | ||
277 | zh_new = kmem_cache_alloc(znode_hold_cache, KM_SLEEP); | |
278 | zh_new->zh_obj = obj; | |
279 | search.zh_obj = obj; | |
280 | ||
0037b49e BB |
281 | mutex_enter(&zfsvfs->z_hold_locks[i]); |
282 | zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL); | |
c96c36fa BB |
283 | if (likely(zh == NULL)) { |
284 | zh = zh_new; | |
0037b49e | 285 | avl_add(&zfsvfs->z_hold_trees[i], zh); |
c96c36fa BB |
286 | } else { |
287 | ASSERT3U(zh->zh_obj, ==, obj); | |
288 | found = B_TRUE; | |
289 | } | |
c13060e4 | 290 | zfs_refcount_add(&zh->zh_refcount, NULL); |
0037b49e | 291 | mutex_exit(&zfsvfs->z_hold_locks[i]); |
c96c36fa BB |
292 | |
293 | if (found == B_TRUE) | |
294 | kmem_cache_free(znode_hold_cache, zh_new); | |
295 | ||
296 | ASSERT(MUTEX_NOT_HELD(&zh->zh_lock)); | |
424fd7c3 | 297 | ASSERT3S(zfs_refcount_count(&zh->zh_refcount), >, 0); |
c96c36fa BB |
298 | mutex_enter(&zh->zh_lock); |
299 | ||
300 | return (zh); | |
301 | } | |
302 | ||
303 | static void | |
0037b49e | 304 | zfs_znode_hold_exit(zfsvfs_t *zfsvfs, znode_hold_t *zh) |
c96c36fa | 305 | { |
0037b49e | 306 | int i = ZFS_OBJ_HASH(zfsvfs, zh->zh_obj); |
c96c36fa BB |
307 | boolean_t remove = B_FALSE; |
308 | ||
0037b49e | 309 | ASSERT(zfs_znode_held(zfsvfs, zh->zh_obj)); |
424fd7c3 | 310 | ASSERT3S(zfs_refcount_count(&zh->zh_refcount), >, 0); |
c96c36fa BB |
311 | mutex_exit(&zh->zh_lock); |
312 | ||
0037b49e | 313 | mutex_enter(&zfsvfs->z_hold_locks[i]); |
424fd7c3 | 314 | if (zfs_refcount_remove(&zh->zh_refcount, NULL) == 0) { |
0037b49e | 315 | avl_remove(&zfsvfs->z_hold_trees[i], zh); |
c96c36fa BB |
316 | remove = B_TRUE; |
317 | } | |
0037b49e | 318 | mutex_exit(&zfsvfs->z_hold_locks[i]); |
c96c36fa BB |
319 | |
320 | if (remove == B_TRUE) | |
321 | kmem_cache_free(znode_hold_cache, zh); | |
34dc7c2f BB |
322 | } |
323 | ||
34dc7c2f | 324 | static void |
0037b49e | 325 | zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, |
428870ff | 326 | dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) |
34dc7c2f | 327 | { |
0037b49e | 328 | ASSERT(zfs_znode_held(zfsvfs, zp->z_id)); |
34dc7c2f BB |
329 | |
330 | mutex_enter(&zp->z_lock); | |
331 | ||
428870ff BB |
332 | ASSERT(zp->z_sa_hdl == NULL); |
333 | ASSERT(zp->z_acl_cached == NULL); | |
334 | if (sa_hdl == NULL) { | |
0037b49e | 335 | VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp, |
428870ff BB |
336 | SA_HDL_SHARED, &zp->z_sa_hdl)); |
337 | } else { | |
338 | zp->z_sa_hdl = sa_hdl; | |
339 | sa_set_userp(sa_hdl, zp); | |
340 | } | |
34dc7c2f | 341 | |
428870ff | 342 | zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; |
34dc7c2f | 343 | |
34dc7c2f | 344 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
345 | } |
346 | ||
347 | void | |
348 | zfs_znode_dmu_fini(znode_t *zp) | |
349 | { | |
c96c36fa | 350 | ASSERT(zfs_znode_held(ZTOZSB(zp), zp->z_id) || zp->z_unlinked || |
3558fd73 | 351 | RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock)); |
428870ff BB |
352 | |
353 | sa_handle_destroy(zp->z_sa_hdl); | |
354 | zp->z_sa_hdl = NULL; | |
34dc7c2f BB |
355 | } |
356 | ||
357 | /* | |
3558fd73 BB |
358 | * Called by new_inode() to allocate a new inode. |
359 | */ | |
360 | int | |
361 | zfs_inode_alloc(struct super_block *sb, struct inode **ip) | |
362 | { | |
363 | znode_t *zp; | |
364 | ||
79c76d5b | 365 | zp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
3558fd73 BB |
366 | *ip = ZTOI(zp); |
367 | ||
368 | return (0); | |
369 | } | |
370 | ||
371 | /* | |
372 | * Called in multiple places when an inode should be destroyed. | |
373 | */ | |
374 | void | |
375 | zfs_inode_destroy(struct inode *ip) | |
376 | { | |
377 | znode_t *zp = ITOZ(ip); | |
0037b49e | 378 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
3558fd73 | 379 | |
0037b49e | 380 | mutex_enter(&zfsvfs->z_znodes_lock); |
7b3e34ba | 381 | if (list_link_active(&zp->z_link_node)) { |
0037b49e BB |
382 | list_remove(&zfsvfs->z_all_znodes, zp); |
383 | zfsvfs->z_nr_znodes--; | |
7b3e34ba | 384 | } |
0037b49e | 385 | mutex_exit(&zfsvfs->z_znodes_lock); |
3558fd73 BB |
386 | |
387 | if (zp->z_acl_cached) { | |
388 | zfs_acl_free(zp->z_acl_cached); | |
389 | zp->z_acl_cached = NULL; | |
390 | } | |
391 | ||
82a37189 BB |
392 | if (zp->z_xattr_cached) { |
393 | nvlist_free(zp->z_xattr_cached); | |
394 | zp->z_xattr_cached = NULL; | |
395 | } | |
396 | ||
3558fd73 BB |
397 | kmem_cache_free(znode_cache, zp); |
398 | } | |
399 | ||
400 | static void | |
0037b49e | 401 | zfs_inode_set_ops(zfsvfs_t *zfsvfs, struct inode *ip) |
3558fd73 | 402 | { |
aa6d8c10 | 403 | uint64_t rdev = 0; |
3558fd73 BB |
404 | |
405 | switch (ip->i_mode & S_IFMT) { | |
406 | case S_IFREG: | |
407 | ip->i_op = &zpl_inode_operations; | |
408 | ip->i_fop = &zpl_file_operations; | |
409 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
410 | break; | |
411 | ||
412 | case S_IFDIR: | |
413 | ip->i_op = &zpl_dir_inode_operations; | |
414 | ip->i_fop = &zpl_dir_file_operations; | |
415 | ITOZ(ip)->z_zn_prefetch = B_TRUE; | |
416 | break; | |
417 | ||
418 | case S_IFLNK: | |
419 | ip->i_op = &zpl_symlink_inode_operations; | |
420 | break; | |
421 | ||
aa6d8c10 NB |
422 | /* |
423 | * rdev is only stored in a SA only for device files. | |
424 | */ | |
3558fd73 BB |
425 | case S_IFCHR: |
426 | case S_IFBLK: | |
0037b49e | 427 | (void) sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zfsvfs), &rdev, |
53b1d979 | 428 | sizeof (rdev)); |
aa6d8c10 NB |
429 | /*FALLTHROUGH*/ |
430 | case S_IFIFO: | |
431 | case S_IFSOCK: | |
3558fd73 BB |
432 | init_special_inode(ip, ip->i_mode, rdev); |
433 | ip->i_op = &zpl_special_inode_operations; | |
434 | break; | |
435 | ||
436 | default: | |
53b1d979 BB |
437 | zfs_panic_recover("inode %llu has invalid mode: 0x%x\n", |
438 | (u_longlong_t)ip->i_ino, ip->i_mode); | |
439 | ||
440 | /* Assume the inode is a file and attempt to continue */ | |
441 | ip->i_mode = S_IFREG | 0644; | |
442 | ip->i_op = &zpl_inode_operations; | |
443 | ip->i_fop = &zpl_file_operations; | |
444 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
445 | break; | |
3558fd73 BB |
446 | } |
447 | } | |
448 | ||
7bb1325f CC |
449 | void |
450 | zfs_set_inode_flags(znode_t *zp, struct inode *ip) | |
451 | { | |
452 | /* | |
453 | * Linux and Solaris have different sets of file attributes, so we | |
454 | * restrict this conversion to the intersection of the two. | |
455 | */ | |
a5248129 CC |
456 | #ifdef HAVE_INODE_SET_FLAGS |
457 | unsigned int flags = 0; | |
458 | if (zp->z_pflags & ZFS_IMMUTABLE) | |
459 | flags |= S_IMMUTABLE; | |
460 | if (zp->z_pflags & ZFS_APPENDONLY) | |
461 | flags |= S_APPEND; | |
7bb1325f | 462 | |
a5248129 CC |
463 | inode_set_flags(ip, flags, S_IMMUTABLE|S_APPEND); |
464 | #else | |
7bb1325f CC |
465 | if (zp->z_pflags & ZFS_IMMUTABLE) |
466 | ip->i_flags |= S_IMMUTABLE; | |
467 | else | |
468 | ip->i_flags &= ~S_IMMUTABLE; | |
469 | ||
470 | if (zp->z_pflags & ZFS_APPENDONLY) | |
471 | ip->i_flags |= S_APPEND; | |
472 | else | |
473 | ip->i_flags &= ~S_APPEND; | |
a5248129 | 474 | #endif |
7bb1325f CC |
475 | } |
476 | ||
704cd075 CC |
477 | /* |
478 | * Update the embedded inode given the znode. We should work toward | |
479 | * eliminating this function as soon as possible by removing values | |
480 | * which are duplicated between the znode and inode. If the generic | |
481 | * inode has the correct field it should be used, and the ZFS code | |
482 | * updated to access the inode. This can be done incrementally. | |
483 | */ | |
9f5f0019 NB |
484 | void |
485 | zfs_inode_update(znode_t *zp) | |
704cd075 | 486 | { |
0037b49e | 487 | zfsvfs_t *zfsvfs; |
704cd075 CC |
488 | struct inode *ip; |
489 | uint32_t blksize; | |
490 | u_longlong_t i_blocks; | |
704cd075 CC |
491 | |
492 | ASSERT(zp != NULL); | |
0037b49e | 493 | zfsvfs = ZTOZSB(zp); |
704cd075 CC |
494 | ip = ZTOI(zp); |
495 | ||
496 | /* Skip .zfs control nodes which do not exist on disk. */ | |
497 | if (zfsctl_is_node(ip)) | |
498 | return; | |
499 | ||
704cd075 CC |
500 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, &i_blocks); |
501 | ||
502 | spin_lock(&ip->i_lock); | |
704cd075 | 503 | ip->i_blocks = i_blocks; |
704cd075 CC |
504 | i_size_write(ip, zp->z_size); |
505 | spin_unlock(&ip->i_lock); | |
506 | } | |
507 | ||
704cd075 | 508 | |
3558fd73 BB |
509 | /* |
510 | * Construct a znode+inode and initialize. | |
34dc7c2f BB |
511 | * |
512 | * This does not do a call to dmu_set_user() that is | |
513 | * up to the caller to do, in case you don't want to | |
514 | * return the znode | |
515 | */ | |
516 | static znode_t * | |
0037b49e | 517 | zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, |
31b6111f | 518 | dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl) |
34dc7c2f BB |
519 | { |
520 | znode_t *zp; | |
3558fd73 | 521 | struct inode *ip; |
7f89ae6b | 522 | uint64_t mode; |
428870ff | 523 | uint64_t parent; |
278f2236 | 524 | uint64_t tmp_gen; |
dfbc8630 | 525 | uint64_t links; |
2c6abf15 | 526 | uint64_t z_uid, z_gid; |
9f5f0019 | 527 | uint64_t atime[2], mtime[2], ctime[2]; |
9c5167d1 | 528 | uint64_t projid = ZFS_DEFAULT_PROJID; |
9f5f0019 | 529 | sa_bulk_attr_t bulk[11]; |
428870ff | 530 | int count = 0; |
34dc7c2f | 531 | |
0037b49e | 532 | ASSERT(zfsvfs != NULL); |
34dc7c2f | 533 | |
0037b49e | 534 | ip = new_inode(zfsvfs->z_sb); |
3558fd73 BB |
535 | if (ip == NULL) |
536 | return (NULL); | |
7304b6e5 | 537 | |
3558fd73 | 538 | zp = ITOZ(ip); |
34dc7c2f | 539 | ASSERT(zp->z_dirlocks == NULL); |
ebe7e575 BB |
540 | ASSERT3P(zp->z_acl_cached, ==, NULL); |
541 | ASSERT3P(zp->z_xattr_cached, ==, NULL); | |
572e2857 | 542 | zp->z_moved = 0; |
428870ff | 543 | zp->z_sa_hdl = NULL; |
34dc7c2f BB |
544 | zp->z_unlinked = 0; |
545 | zp->z_atime_dirty = 0; | |
546 | zp->z_mapcnt = 0; | |
34dc7c2f BB |
547 | zp->z_id = db->db_object; |
548 | zp->z_blksz = blksz; | |
549 | zp->z_seq = 0x7A4653; | |
550 | zp->z_sync_cnt = 0; | |
ebe7e575 BB |
551 | zp->z_is_mapped = B_FALSE; |
552 | zp->z_is_ctldir = B_FALSE; | |
7b3e34ba | 553 | zp->z_is_stale = B_FALSE; |
34dc7c2f | 554 | |
0037b49e | 555 | zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); |
3558fd73 | 556 | |
0037b49e BB |
557 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
558 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8); | |
559 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
560 | &zp->z_size, 8); | |
561 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); | |
562 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
428870ff | 563 | &zp->z_pflags, 8); |
0037b49e | 564 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, |
7304b6e5 | 565 | &parent, 8); |
0037b49e BB |
566 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &z_uid, 8); |
567 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &z_gid, 8); | |
568 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
569 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
570 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
428870ff | 571 | |
9c5167d1 NF |
572 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || tmp_gen == 0 || |
573 | (dmu_objset_projectquota_enabled(zfsvfs->z_os) && | |
574 | (zp->z_pflags & ZFS_PROJID) && | |
575 | sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) { | |
428870ff BB |
576 | if (hdl == NULL) |
577 | sa_handle_destroy(zp->z_sa_hdl); | |
07d63f0c | 578 | zp->z_sa_hdl = NULL; |
3558fd73 | 579 | goto error; |
34dc7c2f | 580 | } |
7304b6e5 | 581 | |
9c5167d1 | 582 | zp->z_projid = projid; |
12fa7f34 | 583 | zp->z_mode = ip->i_mode = mode; |
278f2236 | 584 | ip->i_generation = (uint32_t)tmp_gen; |
ba2fe6af | 585 | ip->i_blkbits = SPA_MINBLOCKSHIFT; |
dfbc8630 | 586 | set_nlink(ip, (uint32_t)links); |
2c6abf15 NB |
587 | zfs_uid_write(ip, z_uid); |
588 | zfs_gid_write(ip, z_gid); | |
7bb1325f | 589 | zfs_set_inode_flags(zp, ip); |
7f89ae6b | 590 | |
98701490 CC |
591 | /* Cache the xattr parent id */ |
592 | if (zp->z_pflags & ZFS_XATTR) | |
593 | zp->z_xattr_parent = parent; | |
594 | ||
9f5f0019 NB |
595 | ZFS_TIME_DECODE(&ip->i_atime, atime); |
596 | ZFS_TIME_DECODE(&ip->i_mtime, mtime); | |
597 | ZFS_TIME_DECODE(&ip->i_ctime, ctime); | |
598 | ||
3558fd73 | 599 | ip->i_ino = obj; |
9f5f0019 | 600 | zfs_inode_update(zp); |
0037b49e | 601 | zfs_inode_set_ops(zfsvfs, ip); |
3558fd73 | 602 | |
7b3e34ba BB |
603 | /* |
604 | * The only way insert_inode_locked() can fail is if the ip->i_ino | |
605 | * number is already hashed for this super block. This can never | |
606 | * happen because the inode numbers map 1:1 with the object numbers. | |
607 | * | |
608 | * The one exception is rolling back a mounted file system, but in | |
609 | * this case all the active inode are unhashed during the rollback. | |
610 | */ | |
611 | VERIFY3S(insert_inode_locked(ip), ==, 0); | |
c85b224f | 612 | |
0037b49e BB |
613 | mutex_enter(&zfsvfs->z_znodes_lock); |
614 | list_insert_tail(&zfsvfs->z_all_znodes, zp); | |
615 | zfsvfs->z_nr_znodes++; | |
b128c09f | 616 | membar_producer(); |
0037b49e | 617 | mutex_exit(&zfsvfs->z_znodes_lock); |
b128c09f | 618 | |
3558fd73 | 619 | unlock_new_inode(ip); |
34dc7c2f | 620 | return (zp); |
3558fd73 BB |
621 | |
622 | error: | |
3558fd73 | 623 | iput(ip); |
d1d7e268 | 624 | return (NULL); |
34dc7c2f BB |
625 | } |
626 | ||
1e8db771 BB |
627 | /* |
628 | * Safely mark an inode dirty. Inodes which are part of a read-only | |
629 | * file system or snapshot may not be dirtied. | |
630 | */ | |
631 | void | |
632 | zfs_mark_inode_dirty(struct inode *ip) | |
633 | { | |
0037b49e | 634 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
1e8db771 | 635 | |
0037b49e | 636 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
1e8db771 BB |
637 | return; |
638 | ||
639 | mark_inode_dirty(ip); | |
640 | } | |
641 | ||
428870ff BB |
642 | static uint64_t empty_xattr; |
643 | static uint64_t pad[4]; | |
644 | static zfs_acl_phys_t acl_phys; | |
34dc7c2f BB |
645 | /* |
646 | * Create a new DMU object to hold a zfs znode. | |
647 | * | |
648 | * IN: dzp - parent directory for new znode | |
649 | * vap - file attributes for new znode | |
650 | * tx - dmu transaction id for zap operations | |
651 | * cr - credentials of caller | |
652 | * flag - flags: | |
653 | * IS_ROOT_NODE - new object will be root | |
654 | * IS_XATTR - new object is an attribute | |
34dc7c2f BB |
655 | * bonuslen - length of bonus buffer |
656 | * setaclp - File/Dir initial ACL | |
657 | * fuidp - Tracks fuid allocation. | |
658 | * | |
659 | * OUT: zpp - allocated znode | |
660 | * | |
661 | */ | |
662 | void | |
663 | zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, | |
428870ff | 664 | uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) |
34dc7c2f | 665 | { |
428870ff BB |
666 | uint64_t crtime[2], atime[2], mtime[2], ctime[2]; |
667 | uint64_t mode, size, links, parent, pflags; | |
9c5167d1 | 668 | uint64_t projid = ZFS_DEFAULT_PROJID; |
428870ff | 669 | uint64_t rdev = 0; |
0037b49e | 670 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
428870ff | 671 | dmu_buf_t *db; |
6413c95f | 672 | inode_timespec_t now; |
34dc7c2f | 673 | uint64_t gen, obj; |
428870ff | 674 | int bonuslen; |
50c957f7 | 675 | int dnodesize; |
428870ff BB |
676 | sa_handle_t *sa_hdl; |
677 | dmu_object_type_t obj_type; | |
f30484af | 678 | sa_bulk_attr_t *sa_attrs; |
428870ff BB |
679 | int cnt = 0; |
680 | zfs_acl_locator_cb_t locate = { 0 }; | |
c96c36fa | 681 | znode_hold_t *zh; |
34dc7c2f | 682 | |
0037b49e | 683 | if (zfsvfs->z_replay) { |
34dc7c2f | 684 | obj = vap->va_nodeid; |
34dc7c2f BB |
685 | now = vap->va_ctime; /* see zfs_replay_create() */ |
686 | gen = vap->va_nblocks; /* ditto */ | |
50c957f7 | 687 | dnodesize = vap->va_fsid; /* ditto */ |
34dc7c2f BB |
688 | } else { |
689 | obj = 0; | |
690 | gethrestime(&now); | |
691 | gen = dmu_tx_get_txg(tx); | |
0037b49e | 692 | dnodesize = dmu_objset_dnodesize(zfsvfs->z_os); |
34dc7c2f BB |
693 | } |
694 | ||
50c957f7 NB |
695 | if (dnodesize == 0) |
696 | dnodesize = DNODE_MIN_SIZE; | |
697 | ||
0037b49e | 698 | obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; |
50c957f7 | 699 | |
428870ff | 700 | bonuslen = (obj_type == DMU_OT_SA) ? |
50c957f7 | 701 | DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE; |
428870ff | 702 | |
34dc7c2f BB |
703 | /* |
704 | * Create a new DMU object. | |
705 | */ | |
706 | /* | |
707 | * There's currently no mechanism for pre-reading the blocks that will | |
572e2857 | 708 | * be needed to allocate a new object, so we accept the small chance |
34dc7c2f BB |
709 | * that there will be an i/o error and we will fail one of the |
710 | * assertions below. | |
711 | */ | |
3558fd73 | 712 | if (S_ISDIR(vap->va_mode)) { |
0037b49e BB |
713 | if (zfsvfs->z_replay) { |
714 | VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj, | |
715 | zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
50c957f7 | 716 | obj_type, bonuslen, dnodesize, tx)); |
34dc7c2f | 717 | } else { |
0037b49e BB |
718 | obj = zap_create_norm_dnsize(zfsvfs->z_os, |
719 | zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
50c957f7 | 720 | obj_type, bonuslen, dnodesize, tx); |
34dc7c2f BB |
721 | } |
722 | } else { | |
0037b49e BB |
723 | if (zfsvfs->z_replay) { |
724 | VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj, | |
34dc7c2f | 725 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
50c957f7 | 726 | obj_type, bonuslen, dnodesize, tx)); |
34dc7c2f | 727 | } else { |
0037b49e | 728 | obj = dmu_object_alloc_dnsize(zfsvfs->z_os, |
34dc7c2f | 729 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
50c957f7 | 730 | obj_type, bonuslen, dnodesize, tx); |
34dc7c2f BB |
731 | } |
732 | } | |
34dc7c2f | 733 | |
0037b49e | 734 | zh = zfs_znode_hold_enter(zfsvfs, obj); |
9631681b | 735 | VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db)); |
34dc7c2f BB |
736 | |
737 | /* | |
738 | * If this is the root, fix up the half-initialized parent pointer | |
739 | * to reference the just-allocated physical data area. | |
740 | */ | |
741 | if (flag & IS_ROOT_NODE) { | |
34dc7c2f BB |
742 | dzp->z_id = obj; |
743 | } | |
744 | ||
745 | /* | |
746 | * If parent is an xattr, so am I. | |
747 | */ | |
9c5167d1 | 748 | if (dzp->z_pflags & ZFS_XATTR) { |
34dc7c2f | 749 | flag |= IS_XATTR; |
34dc7c2f BB |
750 | } |
751 | ||
0037b49e | 752 | if (zfsvfs->z_use_fuids) |
428870ff BB |
753 | pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; |
754 | else | |
755 | pflags = 0; | |
34dc7c2f | 756 | |
3558fd73 | 757 | if (S_ISDIR(vap->va_mode)) { |
428870ff | 758 | size = 2; /* contents ("." and "..") */ |
dfbc8630 | 759 | links = 2; |
428870ff | 760 | } else { |
dfbc8630 | 761 | size = 0; |
ace1eae8 | 762 | links = (flag & IS_TMPFILE) ? 0 : 1; |
34dc7c2f BB |
763 | } |
764 | ||
aa6d8c10 | 765 | if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode)) |
dc1d7665 | 766 | rdev = vap->va_rdev; |
428870ff BB |
767 | |
768 | parent = dzp->z_id; | |
769 | mode = acl_ids->z_mode; | |
34dc7c2f | 770 | if (flag & IS_XATTR) |
428870ff | 771 | pflags |= ZFS_XATTR; |
34dc7c2f | 772 | |
9c5167d1 NF |
773 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) { |
774 | /* | |
775 | * With ZFS_PROJID flag, we can easily know whether there is | |
776 | * project ID stored on disk or not. See zfs_space_delta_cb(). | |
777 | */ | |
778 | if (obj_type != DMU_OT_ZNODE && | |
779 | dmu_objset_projectquota_enabled(zfsvfs->z_os)) | |
780 | pflags |= ZFS_PROJID; | |
781 | ||
782 | /* | |
783 | * Inherit project ID from parent if required. | |
784 | */ | |
785 | projid = zfs_inherit_projid(dzp); | |
786 | if (dzp->z_pflags & ZFS_PROJINHERIT) | |
787 | pflags |= ZFS_PROJINHERIT; | |
788 | } | |
789 | ||
428870ff BB |
790 | /* |
791 | * No execs denied will be deterimed when zfs_mode_compute() is called. | |
792 | */ | |
793 | pflags |= acl_ids->z_aclp->z_hints & | |
794 | (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| | |
795 | ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); | |
34dc7c2f | 796 | |
428870ff BB |
797 | ZFS_TIME_ENCODE(&now, crtime); |
798 | ZFS_TIME_ENCODE(&now, ctime); | |
34dc7c2f | 799 | |
3558fd73 | 800 | if (vap->va_mask & ATTR_ATIME) { |
428870ff | 801 | ZFS_TIME_ENCODE(&vap->va_atime, atime); |
34dc7c2f | 802 | } else { |
428870ff | 803 | ZFS_TIME_ENCODE(&now, atime); |
34dc7c2f BB |
804 | } |
805 | ||
3558fd73 | 806 | if (vap->va_mask & ATTR_MTIME) { |
428870ff BB |
807 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
808 | } else { | |
809 | ZFS_TIME_ENCODE(&now, mtime); | |
810 | } | |
811 | ||
812 | /* Now add in all of the "SA" attributes */ | |
0037b49e | 813 | VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, |
428870ff BB |
814 | &sa_hdl)); |
815 | ||
816 | /* | |
817 | * Setup the array of attributes to be replaced/set on the new file | |
818 | * | |
819 | * order for DMU_OT_ZNODE is critical since it needs to be constructed | |
820 | * in the old znode_phys_t format. Don't change this ordering | |
821 | */ | |
79c76d5b | 822 | sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); |
428870ff BB |
823 | |
824 | if (obj_type == DMU_OT_ZNODE) { | |
0037b49e | 825 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), |
428870ff | 826 | NULL, &atime, 16); |
0037b49e | 827 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), |
428870ff | 828 | NULL, &mtime, 16); |
0037b49e | 829 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), |
428870ff | 830 | NULL, &ctime, 16); |
0037b49e | 831 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), |
428870ff | 832 | NULL, &crtime, 16); |
0037b49e | 833 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), |
428870ff | 834 | NULL, &gen, 8); |
0037b49e | 835 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), |
428870ff | 836 | NULL, &mode, 8); |
0037b49e | 837 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), |
428870ff | 838 | NULL, &size, 8); |
0037b49e | 839 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), |
428870ff | 840 | NULL, &parent, 8); |
34dc7c2f | 841 | } else { |
0037b49e | 842 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), |
428870ff | 843 | NULL, &mode, 8); |
0037b49e | 844 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), |
428870ff | 845 | NULL, &size, 8); |
0037b49e | 846 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), |
428870ff | 847 | NULL, &gen, 8); |
0037b49e | 848 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), |
3558fd73 | 849 | NULL, &acl_ids->z_fuid, 8); |
0037b49e | 850 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), |
3558fd73 | 851 | NULL, &acl_ids->z_fgid, 8); |
0037b49e | 852 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), |
428870ff | 853 | NULL, &parent, 8); |
0037b49e | 854 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 855 | NULL, &pflags, 8); |
0037b49e | 856 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), |
428870ff | 857 | NULL, &atime, 16); |
0037b49e | 858 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), |
428870ff | 859 | NULL, &mtime, 16); |
0037b49e | 860 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), |
428870ff | 861 | NULL, &ctime, 16); |
0037b49e | 862 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), |
428870ff BB |
863 | NULL, &crtime, 16); |
864 | } | |
865 | ||
0037b49e | 866 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); |
428870ff BB |
867 | |
868 | if (obj_type == DMU_OT_ZNODE) { | |
0037b49e | 869 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, |
428870ff | 870 | &empty_xattr, 8); |
9c5167d1 NF |
871 | } else if (dmu_objset_projectquota_enabled(zfsvfs->z_os) && |
872 | pflags & ZFS_PROJID) { | |
873 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PROJID(zfsvfs), | |
874 | NULL, &projid, 8); | |
34dc7c2f | 875 | } |
428870ff | 876 | if (obj_type == DMU_OT_ZNODE || |
aa6d8c10 | 877 | (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) { |
0037b49e | 878 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs), |
428870ff | 879 | NULL, &rdev, 8); |
428870ff BB |
880 | } |
881 | if (obj_type == DMU_OT_ZNODE) { | |
0037b49e | 882 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 883 | NULL, &pflags, 8); |
0037b49e | 884 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 885 | &acl_ids->z_fuid, 8); |
0037b49e | 886 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, |
428870ff | 887 | &acl_ids->z_fgid, 8); |
0037b49e | 888 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad, |
428870ff | 889 | sizeof (uint64_t) * 4); |
0037b49e | 890 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, |
428870ff BB |
891 | &acl_phys, sizeof (zfs_acl_phys_t)); |
892 | } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { | |
0037b49e | 893 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL, |
428870ff BB |
894 | &acl_ids->z_aclp->z_acl_count, 8); |
895 | locate.cb_aclp = acl_ids->z_aclp; | |
0037b49e | 896 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs), |
428870ff BB |
897 | zfs_acl_data_locator, &locate, |
898 | acl_ids->z_aclp->z_acl_bytes); | |
899 | mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, | |
900 | acl_ids->z_fuid, acl_ids->z_fgid); | |
901 | } | |
902 | ||
903 | VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); | |
34dc7c2f | 904 | |
34dc7c2f | 905 | if (!(flag & IS_ROOT_NODE)) { |
8d703987 BB |
906 | /* |
907 | * The call to zfs_znode_alloc() may fail if memory is low | |
908 | * via the call path: alloc_inode() -> inode_init_always() -> | |
909 | * security_inode_alloc() -> inode_alloc_security(). Since | |
910 | * the existing code is written such that zfs_mknode() can | |
911 | * not fail retry until sufficient memory has been reclaimed. | |
912 | */ | |
913 | do { | |
914 | *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, obj, | |
915 | sa_hdl); | |
916 | } while (*zpp == NULL); | |
917 | ||
7b3e34ba BB |
918 | VERIFY(*zpp != NULL); |
919 | VERIFY(dzp != NULL); | |
34dc7c2f BB |
920 | } else { |
921 | /* | |
922 | * If we are creating the root node, the "parent" we | |
923 | * passed in is the znode for the root. | |
924 | */ | |
925 | *zpp = dzp; | |
428870ff BB |
926 | |
927 | (*zpp)->z_sa_hdl = sa_hdl; | |
34dc7c2f | 928 | } |
428870ff BB |
929 | |
930 | (*zpp)->z_pflags = pflags; | |
12fa7f34 | 931 | (*zpp)->z_mode = ZTOI(*zpp)->i_mode = mode; |
50c957f7 | 932 | (*zpp)->z_dnodesize = dnodesize; |
9c5167d1 | 933 | (*zpp)->z_projid = projid; |
428870ff | 934 | |
428870ff BB |
935 | if (obj_type == DMU_OT_ZNODE || |
936 | acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { | |
b0bc7a84 | 937 | VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); |
428870ff | 938 | } |
d1d7e268 | 939 | kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); |
0037b49e | 940 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
941 | } |
942 | ||
5484965a | 943 | /* |
d3cc8b15 WA |
944 | * Update in-core attributes. It is assumed the caller will be doing an |
945 | * sa_bulk_update to push the changes out. | |
5484965a BB |
946 | */ |
947 | void | |
948 | zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) | |
949 | { | |
950 | xoptattr_t *xoap; | |
7bb1325f | 951 | boolean_t update_inode = B_FALSE; |
5484965a BB |
952 | |
953 | xoap = xva_getxoptattr(xvap); | |
954 | ASSERT(xoap); | |
955 | ||
956 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { | |
957 | uint64_t times[2]; | |
958 | ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); | |
959 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)), | |
960 | ×, sizeof (times), tx); | |
961 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
962 | } | |
963 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
964 | ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, | |
965 | zp->z_pflags, tx); | |
966 | XVA_SET_RTN(xvap, XAT_READONLY); | |
967 | } | |
968 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
969 | ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, | |
970 | zp->z_pflags, tx); | |
971 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
972 | } | |
973 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
974 | ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, | |
975 | zp->z_pflags, tx); | |
976 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
977 | } | |
978 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
979 | ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, | |
980 | zp->z_pflags, tx); | |
981 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
982 | } | |
983 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
984 | ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, | |
985 | zp->z_pflags, tx); | |
986 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
64c688d7 | 987 | |
7bb1325f | 988 | update_inode = B_TRUE; |
5484965a BB |
989 | } |
990 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
991 | ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, | |
992 | zp->z_pflags, tx); | |
993 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
994 | } | |
995 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
996 | ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, | |
997 | zp->z_pflags, tx); | |
998 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
64c688d7 | 999 | |
7bb1325f | 1000 | update_inode = B_TRUE; |
5484965a BB |
1001 | } |
1002 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
1003 | ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, | |
1004 | zp->z_pflags, tx); | |
1005 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
1006 | } | |
1007 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
1008 | ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, | |
1009 | zp->z_pflags, tx); | |
1010 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
1011 | } | |
1012 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
1013 | ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, | |
1014 | xoap->xoa_av_quarantined, zp->z_pflags, tx); | |
1015 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
1016 | } | |
1017 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
1018 | ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, | |
1019 | zp->z_pflags, tx); | |
1020 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
1021 | } | |
1022 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { | |
1023 | zfs_sa_set_scanstamp(zp, xvap, tx); | |
1024 | XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); | |
1025 | } | |
1026 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
1027 | ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, | |
1028 | zp->z_pflags, tx); | |
1029 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
1030 | } | |
1031 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
1032 | ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, | |
1033 | zp->z_pflags, tx); | |
1034 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
1035 | } | |
1036 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
1037 | ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, | |
1038 | zp->z_pflags, tx); | |
1039 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
1040 | } | |
9c5167d1 NF |
1041 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { |
1042 | ZFS_ATTR_SET(zp, ZFS_PROJINHERIT, xoap->xoa_projinherit, | |
1043 | zp->z_pflags, tx); | |
1044 | XVA_SET_RTN(xvap, XAT_PROJINHERIT); | |
1045 | } | |
7bb1325f CC |
1046 | |
1047 | if (update_inode) | |
1048 | zfs_set_inode_flags(zp, ZTOI(zp)); | |
5484965a BB |
1049 | } |
1050 | ||
34dc7c2f | 1051 | int |
0037b49e | 1052 | zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) |
34dc7c2f BB |
1053 | { |
1054 | dmu_object_info_t doi; | |
1055 | dmu_buf_t *db; | |
1056 | znode_t *zp; | |
c96c36fa | 1057 | znode_hold_t *zh; |
34dc7c2f | 1058 | int err; |
428870ff | 1059 | sa_handle_t *hdl; |
34dc7c2f BB |
1060 | |
1061 | *zpp = NULL; | |
1062 | ||
6f9548c4 | 1063 | again: |
0037b49e | 1064 | zh = zfs_znode_hold_enter(zfsvfs, obj_num); |
34dc7c2f | 1065 | |
0037b49e | 1066 | err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); |
34dc7c2f | 1067 | if (err) { |
0037b49e | 1068 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1069 | return (err); |
1070 | } | |
1071 | ||
1072 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1073 | if (doi.doi_bonus_type != DMU_OT_SA && |
1074 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1075 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1076 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1077 | sa_buf_rele(db, NULL); | |
0037b49e | 1078 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1079 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1080 | } |
1081 | ||
428870ff BB |
1082 | hdl = dmu_buf_get_user(db); |
1083 | if (hdl != NULL) { | |
36df2843 | 1084 | zp = sa_get_userdata(hdl); |
34dc7c2f | 1085 | |
8ac67298 | 1086 | |
34dc7c2f | 1087 | /* |
428870ff BB |
1088 | * Since "SA" does immediate eviction we |
1089 | * should never find a sa handle that doesn't | |
1090 | * know about the znode. | |
34dc7c2f | 1091 | */ |
428870ff BB |
1092 | |
1093 | ASSERT3P(zp, !=, NULL); | |
1094 | ||
1095 | mutex_enter(&zp->z_lock); | |
34dc7c2f | 1096 | ASSERT3U(zp->z_id, ==, obj_num); |
98701490 CC |
1097 | /* |
1098 | * If igrab() returns NULL the VFS has independently | |
1099 | * determined the inode should be evicted and has | |
1100 | * called iput_final() to start the eviction process. | |
1101 | * The SA handle is still valid but because the VFS | |
1102 | * requires that the eviction succeed we must drop | |
1103 | * our locks and references to allow the eviction to | |
1104 | * complete. The zfs_zget() may then be retried. | |
1105 | * | |
1106 | * This unlikely case could be optimized by registering | |
1107 | * a sops->drop_inode() callback. The callback would | |
1108 | * need to detect the active SA hold thereby informing | |
1109 | * the VFS that this inode should not be evicted. | |
1110 | */ | |
1111 | if (igrab(ZTOI(zp)) == NULL) { | |
1112 | mutex_exit(&zp->z_lock); | |
1113 | sa_buf_rele(db, NULL); | |
0037b49e | 1114 | zfs_znode_hold_exit(zfsvfs, zh); |
98701490 CC |
1115 | /* inode might need this to finish evict */ |
1116 | cond_resched(); | |
1117 | goto again; | |
34dc7c2f | 1118 | } |
98701490 CC |
1119 | *zpp = zp; |
1120 | err = 0; | |
34dc7c2f | 1121 | mutex_exit(&zp->z_lock); |
f3ad9cd6 | 1122 | sa_buf_rele(db, NULL); |
0037b49e | 1123 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1124 | return (err); |
1125 | } | |
1126 | ||
1127 | /* | |
3558fd73 | 1128 | * Not found create new znode/vnode but only if file exists. |
428870ff BB |
1129 | * |
1130 | * There is a small window where zfs_vget() could | |
1131 | * find this object while a file create is still in | |
1132 | * progress. This is checked for in zfs_znode_alloc() | |
1133 | * | |
1134 | * if zfs_znode_alloc() fails it will drop the hold on the | |
1135 | * bonus buffer. | |
34dc7c2f | 1136 | */ |
0037b49e | 1137 | zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size, |
31b6111f | 1138 | doi.doi_bonus_type, obj_num, NULL); |
428870ff | 1139 | if (zp == NULL) { |
2e528b49 | 1140 | err = SET_ERROR(ENOENT); |
428870ff BB |
1141 | } else { |
1142 | *zpp = zp; | |
1143 | } | |
0037b49e | 1144 | zfs_znode_hold_exit(zfsvfs, zh); |
428870ff | 1145 | return (err); |
34dc7c2f BB |
1146 | } |
1147 | ||
1148 | int | |
1149 | zfs_rezget(znode_t *zp) | |
1150 | { | |
0037b49e | 1151 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
1152 | dmu_object_info_t doi; |
1153 | dmu_buf_t *db; | |
1154 | uint64_t obj_num = zp->z_id; | |
428870ff | 1155 | uint64_t mode; |
dfbc8630 | 1156 | uint64_t links; |
9f5f0019 | 1157 | sa_bulk_attr_t bulk[10]; |
34dc7c2f | 1158 | int err; |
428870ff BB |
1159 | int count = 0; |
1160 | uint64_t gen; | |
2c6abf15 | 1161 | uint64_t z_uid, z_gid; |
9f5f0019 | 1162 | uint64_t atime[2], mtime[2], ctime[2]; |
9c5167d1 | 1163 | uint64_t projid = ZFS_DEFAULT_PROJID; |
c96c36fa | 1164 | znode_hold_t *zh; |
34dc7c2f | 1165 | |
cbecb4fb CC |
1166 | /* |
1167 | * skip ctldir, otherwise they will always get invalidated. This will | |
1168 | * cause funny behaviour for the mounted snapdirs. Especially for | |
1169 | * Linux >= 3.18, d_invalidate will detach the mountpoint and prevent | |
1170 | * anyone automount it again as long as someone is still using the | |
1171 | * detached mount. | |
1172 | */ | |
1173 | if (zp->z_is_ctldir) | |
1174 | return (0); | |
1175 | ||
0037b49e | 1176 | zh = zfs_znode_hold_enter(zfsvfs, obj_num); |
34dc7c2f | 1177 | |
428870ff BB |
1178 | mutex_enter(&zp->z_acl_lock); |
1179 | if (zp->z_acl_cached) { | |
1180 | zfs_acl_free(zp->z_acl_cached); | |
1181 | zp->z_acl_cached = NULL; | |
1182 | } | |
428870ff | 1183 | mutex_exit(&zp->z_acl_lock); |
7b3e34ba | 1184 | |
228b461b | 1185 | rw_enter(&zp->z_xattr_lock, RW_WRITER); |
7b3e34ba BB |
1186 | if (zp->z_xattr_cached) { |
1187 | nvlist_free(zp->z_xattr_cached); | |
1188 | zp->z_xattr_cached = NULL; | |
1189 | } | |
7b3e34ba BB |
1190 | rw_exit(&zp->z_xattr_lock); |
1191 | ||
428870ff | 1192 | ASSERT(zp->z_sa_hdl == NULL); |
0037b49e | 1193 | err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); |
34dc7c2f | 1194 | if (err) { |
0037b49e | 1195 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1196 | return (err); |
1197 | } | |
1198 | ||
1199 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1200 | if (doi.doi_bonus_type != DMU_OT_SA && |
1201 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1202 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1203 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1204 | sa_buf_rele(db, NULL); | |
0037b49e | 1205 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1206 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1207 | } |
1208 | ||
0037b49e | 1209 | zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL); |
428870ff BB |
1210 | |
1211 | /* reload cached values */ | |
0037b49e | 1212 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, |
428870ff | 1213 | &gen, sizeof (gen)); |
0037b49e | 1214 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, |
428870ff | 1215 | &zp->z_size, sizeof (zp->z_size)); |
0037b49e | 1216 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, |
dfbc8630 | 1217 | &links, sizeof (links)); |
0037b49e | 1218 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff | 1219 | &zp->z_pflags, sizeof (zp->z_pflags)); |
0037b49e | 1220 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
2c6abf15 | 1221 | &z_uid, sizeof (z_uid)); |
0037b49e | 1222 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, |
2c6abf15 | 1223 | &z_gid, sizeof (z_gid)); |
0037b49e | 1224 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 1225 | &mode, sizeof (mode)); |
0037b49e | 1226 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
9f5f0019 | 1227 | &atime, 16); |
0037b49e | 1228 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
9f5f0019 | 1229 | &mtime, 16); |
0037b49e | 1230 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
9f5f0019 | 1231 | &ctime, 16); |
428870ff | 1232 | |
428870ff BB |
1233 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { |
1234 | zfs_znode_dmu_fini(zp); | |
0037b49e | 1235 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1236 | return (SET_ERROR(EIO)); |
428870ff BB |
1237 | } |
1238 | ||
9c5167d1 NF |
1239 | if (dmu_objset_projectquota_enabled(zfsvfs->z_os)) { |
1240 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), | |
1241 | &projid, 8); | |
1242 | if (err != 0 && err != ENOENT) { | |
1243 | zfs_znode_dmu_fini(zp); | |
1244 | zfs_znode_hold_exit(zfsvfs, zh); | |
1245 | return (SET_ERROR(err)); | |
1246 | } | |
1247 | } | |
1248 | ||
1249 | zp->z_projid = projid; | |
12fa7f34 | 1250 | zp->z_mode = ZTOI(zp)->i_mode = mode; |
2c6abf15 NB |
1251 | zfs_uid_write(ZTOI(zp), z_uid); |
1252 | zfs_gid_write(ZTOI(zp), z_gid); | |
572e2857 | 1253 | |
9f5f0019 NB |
1254 | ZFS_TIME_DECODE(&ZTOI(zp)->i_atime, atime); |
1255 | ZFS_TIME_DECODE(&ZTOI(zp)->i_mtime, mtime); | |
1256 | ZFS_TIME_DECODE(&ZTOI(zp)->i_ctime, ctime); | |
1257 | ||
278f2236 | 1258 | if (gen != ZTOI(zp)->i_generation) { |
428870ff | 1259 | zfs_znode_dmu_fini(zp); |
0037b49e | 1260 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1261 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1262 | } |
1263 | ||
dfbc8630 | 1264 | set_nlink(ZTOI(zp), (uint32_t)links); |
7bb1325f | 1265 | zfs_set_inode_flags(zp, ZTOI(zp)); |
dfbc8630 | 1266 | |
34dc7c2f | 1267 | zp->z_blksz = doi.doi_data_block_size; |
704cd075 | 1268 | zp->z_atime_dirty = 0; |
9f5f0019 | 1269 | zfs_inode_update(zp); |
34dc7c2f | 1270 | |
6a218566 AG |
1271 | /* |
1272 | * If the file has zero links, then it has been unlinked on the send | |
1273 | * side and it must be in the received unlinked set. | |
1274 | * We call zfs_znode_dmu_fini() now to prevent any accesses to the | |
1275 | * stale data and to prevent automatical removal of the file in | |
1276 | * zfs_zinactive(). The file will be removed either when it is removed | |
1277 | * on the send side and the next incremental stream is received or | |
1278 | * when the unlinked set gets processed. | |
1279 | */ | |
1280 | zp->z_unlinked = (ZTOI(zp)->i_nlink == 0); | |
1281 | if (zp->z_unlinked) | |
1282 | zfs_znode_dmu_fini(zp); | |
1283 | ||
0037b49e | 1284 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1285 | |
1286 | return (0); | |
1287 | } | |
1288 | ||
1289 | void | |
1290 | zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) | |
1291 | { | |
0037b49e BB |
1292 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
1293 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 1294 | uint64_t obj = zp->z_id; |
572e2857 | 1295 | uint64_t acl_obj = zfs_external_acl(zp); |
c96c36fa | 1296 | znode_hold_t *zh; |
34dc7c2f | 1297 | |
0037b49e | 1298 | zh = zfs_znode_hold_enter(zfsvfs, obj); |
572e2857 BB |
1299 | if (acl_obj) { |
1300 | VERIFY(!zp->z_is_sa); | |
b128c09f | 1301 | VERIFY(0 == dmu_object_free(os, acl_obj, tx)); |
572e2857 | 1302 | } |
b128c09f | 1303 | VERIFY(0 == dmu_object_free(os, obj, tx)); |
34dc7c2f | 1304 | zfs_znode_dmu_fini(zp); |
0037b49e | 1305 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1306 | } |
1307 | ||
1308 | void | |
1309 | zfs_zinactive(znode_t *zp) | |
1310 | { | |
0037b49e | 1311 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f | 1312 | uint64_t z_id = zp->z_id; |
c96c36fa | 1313 | znode_hold_t *zh; |
34dc7c2f | 1314 | |
428870ff | 1315 | ASSERT(zp->z_sa_hdl); |
34dc7c2f BB |
1316 | |
1317 | /* | |
d6bd8eaa | 1318 | * Don't allow a zfs_zget() while were trying to release this znode. |
34dc7c2f | 1319 | */ |
0037b49e | 1320 | zh = zfs_znode_hold_enter(zfsvfs, z_id); |
d6bd8eaa | 1321 | |
34dc7c2f | 1322 | mutex_enter(&zp->z_lock); |
34dc7c2f BB |
1323 | |
1324 | /* | |
6a218566 AG |
1325 | * If this was the last reference to a file with no links, remove |
1326 | * the file from the file system unless the file system is mounted | |
1327 | * read-only. That can happen, for example, if the file system was | |
1328 | * originally read-write, the file was opened, then unlinked and | |
1329 | * the file system was made read-only before the file was finally | |
1330 | * closed. The file will remain in the unlinked set. | |
34dc7c2f BB |
1331 | */ |
1332 | if (zp->z_unlinked) { | |
6a218566 | 1333 | ASSERT(!zfsvfs->z_issnap); |
dcec0a12 | 1334 | if (!zfs_is_readonly(zfsvfs) && !zfs_unlink_suspend_progress) { |
6a218566 AG |
1335 | mutex_exit(&zp->z_lock); |
1336 | zfs_znode_hold_exit(zfsvfs, zh); | |
1337 | zfs_rmnode(zp); | |
1338 | return; | |
1339 | } | |
34dc7c2f | 1340 | } |
428870ff | 1341 | |
34dc7c2f BB |
1342 | mutex_exit(&zp->z_lock); |
1343 | zfs_znode_dmu_fini(zp); | |
d6bd8eaa | 1344 | |
0037b49e | 1345 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1346 | } |
1347 | ||
9c53e516 TK |
1348 | #if defined(HAVE_INODE_TIMESPEC64_TIMES) |
1349 | #define zfs_compare_timespec timespec64_compare | |
1350 | #else | |
1351 | #define zfs_compare_timespec timespec_compare | |
1352 | #endif | |
1353 | ||
1354 | /* | |
1355 | * Determine whether the znode's atime must be updated. The logic mostly | |
1356 | * duplicates the Linux kernel's relatime_need_update() functionality. | |
1357 | * This function is only called if the underlying filesystem actually has | |
1358 | * atime updates enabled. | |
1359 | */ | |
1360 | boolean_t | |
1361 | zfs_relatime_need_update(const struct inode *ip) | |
6d111134 | 1362 | { |
9c53e516 TK |
1363 | inode_timespec_t now; |
1364 | ||
1365 | gethrestime(&now); | |
1366 | /* | |
1367 | * In relatime mode, only update the atime if the previous atime | |
1368 | * is earlier than either the ctime or mtime or if at least a day | |
1369 | * has passed since the last update of atime. | |
1370 | */ | |
1371 | if (zfs_compare_timespec(&ip->i_mtime, &ip->i_atime) >= 0) | |
1372 | return (B_TRUE); | |
1373 | ||
1374 | if (zfs_compare_timespec(&ip->i_ctime, &ip->i_atime) >= 0) | |
1375 | return (B_TRUE); | |
6d111134 | 1376 | |
9c53e516 TK |
1377 | if ((hrtime_t)now.tv_sec - (hrtime_t)ip->i_atime.tv_sec >= 24*60*60) |
1378 | return (B_TRUE); | |
6d111134 | 1379 | |
9c53e516 | 1380 | return (B_FALSE); |
6d111134 TC |
1381 | } |
1382 | ||
6d111134 TC |
1383 | /* |
1384 | * Prepare to update znode time stamps. | |
1385 | * | |
1386 | * IN: zp - znode requiring timestamp update | |
0df9673f | 1387 | * flag - ATTR_MTIME, ATTR_CTIME flags |
6d111134 | 1388 | * |
0df9673f | 1389 | * OUT: zp - z_seq |
6d111134 TC |
1390 | * mtime - new mtime |
1391 | * ctime - new ctime | |
1392 | * | |
0df9673f CC |
1393 | * Note: We don't update atime here, because we rely on Linux VFS to do |
1394 | * atime updating. | |
6d111134 | 1395 | */ |
34dc7c2f | 1396 | void |
428870ff | 1397 | zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], |
0df9673f | 1398 | uint64_t ctime[2]) |
34dc7c2f | 1399 | { |
6413c95f | 1400 | inode_timespec_t now; |
34dc7c2f | 1401 | |
34dc7c2f BB |
1402 | gethrestime(&now); |
1403 | ||
0df9673f | 1404 | zp->z_seq++; |
34dc7c2f | 1405 | |
3558fd73 | 1406 | if (flag & ATTR_MTIME) { |
428870ff | 1407 | ZFS_TIME_ENCODE(&now, mtime); |
9f5f0019 | 1408 | ZFS_TIME_DECODE(&(ZTOI(zp)->i_mtime), mtime); |
3558fd73 | 1409 | if (ZTOZSB(zp)->z_use_fuids) { |
428870ff BB |
1410 | zp->z_pflags |= (ZFS_ARCHIVE | |
1411 | ZFS_AV_MODIFIED); | |
1412 | } | |
34dc7c2f BB |
1413 | } |
1414 | ||
3558fd73 | 1415 | if (flag & ATTR_CTIME) { |
428870ff | 1416 | ZFS_TIME_ENCODE(&now, ctime); |
9f5f0019 | 1417 | ZFS_TIME_DECODE(&(ZTOI(zp)->i_ctime), ctime); |
3558fd73 | 1418 | if (ZTOZSB(zp)->z_use_fuids) |
428870ff | 1419 | zp->z_pflags |= ZFS_ARCHIVE; |
34dc7c2f BB |
1420 | } |
1421 | } | |
1422 | ||
34dc7c2f BB |
1423 | /* |
1424 | * Grow the block size for a file. | |
1425 | * | |
1426 | * IN: zp - znode of file to free data in. | |
1427 | * size - requested block size | |
1428 | * tx - open transaction. | |
1429 | * | |
1430 | * NOTE: this function assumes that the znode is write locked. | |
1431 | */ | |
1432 | void | |
1433 | zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) | |
1434 | { | |
1435 | int error; | |
1436 | u_longlong_t dummy; | |
1437 | ||
1438 | if (size <= zp->z_blksz) | |
1439 | return; | |
1440 | /* | |
1441 | * If the file size is already greater than the current blocksize, | |
1442 | * we will not grow. If there is more than one block in a file, | |
1443 | * the blocksize cannot change. | |
1444 | */ | |
428870ff | 1445 | if (zp->z_blksz && zp->z_size > zp->z_blksz) |
34dc7c2f BB |
1446 | return; |
1447 | ||
3558fd73 | 1448 | error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id, |
34dc7c2f | 1449 | size, 0, tx); |
428870ff | 1450 | |
34dc7c2f BB |
1451 | if (error == ENOTSUP) |
1452 | return; | |
c99c9001 | 1453 | ASSERT0(error); |
34dc7c2f BB |
1454 | |
1455 | /* What blocksize did we actually get? */ | |
428870ff | 1456 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); |
34dc7c2f BB |
1457 | } |
1458 | ||
34dc7c2f | 1459 | /* |
b128c09f | 1460 | * Increase the file length |
34dc7c2f BB |
1461 | * |
1462 | * IN: zp - znode of file to free data in. | |
b128c09f | 1463 | * end - new end-of-file |
34dc7c2f | 1464 | * |
19d55079 | 1465 | * RETURN: 0 on success, error code on failure |
34dc7c2f | 1466 | */ |
b128c09f BB |
1467 | static int |
1468 | zfs_extend(znode_t *zp, uint64_t end) | |
34dc7c2f | 1469 | { |
0037b49e | 1470 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
b128c09f | 1471 | dmu_tx_t *tx; |
5d43cc9a | 1472 | locked_range_t *lr; |
b128c09f | 1473 | uint64_t newblksz; |
34dc7c2f BB |
1474 | int error; |
1475 | ||
34dc7c2f | 1476 | /* |
b128c09f | 1477 | * We will change zp_size, lock the whole file. |
34dc7c2f | 1478 | */ |
5d43cc9a | 1479 | lr = rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); |
34dc7c2f BB |
1480 | |
1481 | /* | |
1482 | * Nothing to do if file already at desired length. | |
1483 | */ | |
428870ff | 1484 | if (end <= zp->z_size) { |
5d43cc9a | 1485 | rangelock_exit(lr); |
34dc7c2f BB |
1486 | return (0); |
1487 | } | |
0037b49e | 1488 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1489 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1490 | zfs_sa_upgrade_txholds(tx, zp); | |
b128c09f | 1491 | if (end > zp->z_blksz && |
0037b49e | 1492 | (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { |
34dc7c2f BB |
1493 | /* |
1494 | * We are growing the file past the current block size. | |
1495 | */ | |
3558fd73 | 1496 | if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) { |
f1512ee6 MA |
1497 | /* |
1498 | * File's blocksize is already larger than the | |
1499 | * "recordsize" property. Only let it grow to | |
1500 | * the next power of 2. | |
1501 | */ | |
34dc7c2f | 1502 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 | 1503 | newblksz = MIN(end, 1 << highbit64(zp->z_blksz)); |
34dc7c2f | 1504 | } else { |
3558fd73 | 1505 | newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz); |
34dc7c2f | 1506 | } |
b128c09f BB |
1507 | dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); |
1508 | } else { | |
1509 | newblksz = 0; | |
34dc7c2f BB |
1510 | } |
1511 | ||
384f8a09 | 1512 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 1513 | if (error) { |
34dc7c2f | 1514 | dmu_tx_abort(tx); |
5d43cc9a | 1515 | rangelock_exit(lr); |
34dc7c2f BB |
1516 | return (error); |
1517 | } | |
1518 | ||
b128c09f BB |
1519 | if (newblksz) |
1520 | zfs_grow_blocksize(zp, newblksz, tx); | |
34dc7c2f | 1521 | |
428870ff BB |
1522 | zp->z_size = end; |
1523 | ||
3558fd73 | 1524 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), |
428870ff | 1525 | &zp->z_size, sizeof (zp->z_size), tx)); |
34dc7c2f | 1526 | |
5d43cc9a | 1527 | rangelock_exit(lr); |
34dc7c2f | 1528 | |
b128c09f | 1529 | dmu_tx_commit(tx); |
34dc7c2f | 1530 | |
b128c09f BB |
1531 | return (0); |
1532 | } | |
1533 | ||
223df016 TC |
1534 | /* |
1535 | * zfs_zero_partial_page - Modeled after update_pages() but | |
1536 | * with different arguments and semantics for use by zfs_freesp(). | |
1537 | * | |
1538 | * Zeroes a piece of a single page cache entry for zp at offset | |
1539 | * start and length len. | |
1540 | * | |
1541 | * Caller must acquire a range lock on the file for the region | |
1542 | * being zeroed in order that the ARC and page cache stay in sync. | |
1543 | */ | |
1544 | static void | |
1545 | zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len) | |
1546 | { | |
1547 | struct address_space *mp = ZTOI(zp)->i_mapping; | |
1548 | struct page *pp; | |
1549 | int64_t off; | |
1550 | void *pb; | |
1551 | ||
8b1899d3 | 1552 | ASSERT((start & PAGE_MASK) == ((start + len - 1) & PAGE_MASK)); |
223df016 | 1553 | |
8b1899d3 BB |
1554 | off = start & (PAGE_SIZE - 1); |
1555 | start &= PAGE_MASK; | |
223df016 | 1556 | |
8b1899d3 | 1557 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
223df016 TC |
1558 | if (pp) { |
1559 | if (mapping_writably_mapped(mp)) | |
1560 | flush_dcache_page(pp); | |
1561 | ||
1562 | pb = kmap(pp); | |
1563 | bzero(pb + off, len); | |
1564 | kunmap(pp); | |
1565 | ||
1566 | if (mapping_writably_mapped(mp)) | |
1567 | flush_dcache_page(pp); | |
1568 | ||
1569 | mark_page_accessed(pp); | |
1570 | SetPageUptodate(pp); | |
1571 | ClearPageError(pp); | |
1572 | unlock_page(pp); | |
8b1899d3 | 1573 | put_page(pp); |
223df016 TC |
1574 | } |
1575 | } | |
1576 | ||
b128c09f BB |
1577 | /* |
1578 | * Free space in a file. | |
1579 | * | |
1580 | * IN: zp - znode of file to free data in. | |
1581 | * off - start of section to free. | |
1582 | * len - length of section to free. | |
1583 | * | |
19d55079 | 1584 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1585 | */ |
1586 | static int | |
1587 | zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) | |
1588 | { | |
0037b49e | 1589 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
5d43cc9a | 1590 | locked_range_t *lr; |
b128c09f BB |
1591 | int error; |
1592 | ||
1593 | /* | |
1594 | * Lock the range being freed. | |
1595 | */ | |
5d43cc9a | 1596 | lr = rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER); |
b128c09f BB |
1597 | |
1598 | /* | |
1599 | * Nothing to do if file already at desired length. | |
1600 | */ | |
428870ff | 1601 | if (off >= zp->z_size) { |
5d43cc9a | 1602 | rangelock_exit(lr); |
b128c09f | 1603 | return (0); |
34dc7c2f BB |
1604 | } |
1605 | ||
428870ff BB |
1606 | if (off + len > zp->z_size) |
1607 | len = zp->z_size - off; | |
b128c09f | 1608 | |
0037b49e | 1609 | error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); |
b128c09f | 1610 | |
223df016 TC |
1611 | /* |
1612 | * Zero partial page cache entries. This must be done under a | |
1613 | * range lock in order to keep the ARC and page cache in sync. | |
1614 | */ | |
1615 | if (zp->z_is_mapped) { | |
1616 | loff_t first_page, last_page, page_len; | |
1617 | loff_t first_page_offset, last_page_offset; | |
1618 | ||
1619 | /* first possible full page in hole */ | |
8b1899d3 | 1620 | first_page = (off + PAGE_SIZE - 1) >> PAGE_SHIFT; |
223df016 | 1621 | /* last page of hole */ |
8b1899d3 | 1622 | last_page = (off + len) >> PAGE_SHIFT; |
223df016 TC |
1623 | |
1624 | /* offset of first_page */ | |
8b1899d3 | 1625 | first_page_offset = first_page << PAGE_SHIFT; |
223df016 | 1626 | /* offset of last_page */ |
8b1899d3 | 1627 | last_page_offset = last_page << PAGE_SHIFT; |
223df016 | 1628 | |
cb08f063 TC |
1629 | /* truncate whole pages */ |
1630 | if (last_page_offset > first_page_offset) { | |
1631 | truncate_inode_pages_range(ZTOI(zp)->i_mapping, | |
1632 | first_page_offset, last_page_offset - 1); | |
1633 | } | |
1634 | ||
1635 | /* truncate sub-page ranges */ | |
223df016 TC |
1636 | if (first_page > last_page) { |
1637 | /* entire punched area within a single page */ | |
1638 | zfs_zero_partial_page(zp, off, len); | |
1639 | } else { | |
1640 | /* beginning of punched area at the end of a page */ | |
1641 | page_len = first_page_offset - off; | |
1642 | if (page_len > 0) | |
1643 | zfs_zero_partial_page(zp, off, page_len); | |
1644 | ||
1645 | /* end of punched area at the beginning of a page */ | |
1646 | page_len = off + len - last_page_offset; | |
1647 | if (page_len > 0) | |
1648 | zfs_zero_partial_page(zp, last_page_offset, | |
1649 | page_len); | |
1650 | } | |
1651 | } | |
5d43cc9a | 1652 | rangelock_exit(lr); |
34dc7c2f | 1653 | |
b128c09f BB |
1654 | return (error); |
1655 | } | |
1656 | ||
1657 | /* | |
1658 | * Truncate a file | |
1659 | * | |
1660 | * IN: zp - znode of file to free data in. | |
1661 | * end - new end-of-file. | |
1662 | * | |
19d55079 | 1663 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1664 | */ |
1665 | static int | |
1666 | zfs_trunc(znode_t *zp, uint64_t end) | |
1667 | { | |
0037b49e | 1668 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
b128c09f | 1669 | dmu_tx_t *tx; |
5d43cc9a | 1670 | locked_range_t *lr; |
b128c09f | 1671 | int error; |
572e2857 BB |
1672 | sa_bulk_attr_t bulk[2]; |
1673 | int count = 0; | |
b128c09f BB |
1674 | |
1675 | /* | |
1676 | * We will change zp_size, lock the whole file. | |
1677 | */ | |
5d43cc9a | 1678 | lr = rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); |
b128c09f BB |
1679 | |
1680 | /* | |
1681 | * Nothing to do if file already at desired length. | |
1682 | */ | |
428870ff | 1683 | if (end >= zp->z_size) { |
5d43cc9a | 1684 | rangelock_exit(lr); |
b128c09f BB |
1685 | return (0); |
1686 | } | |
1687 | ||
18a2485f FS |
1688 | error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, |
1689 | DMU_OBJECT_END); | |
b128c09f | 1690 | if (error) { |
5d43cc9a | 1691 | rangelock_exit(lr); |
b128c09f BB |
1692 | return (error); |
1693 | } | |
0037b49e | 1694 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1695 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1696 | zfs_sa_upgrade_txholds(tx, zp); | |
19d55079 | 1697 | dmu_tx_mark_netfree(tx); |
7a8f0e80 | 1698 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1699 | if (error) { |
b128c09f | 1700 | dmu_tx_abort(tx); |
5d43cc9a | 1701 | rangelock_exit(lr); |
b128c09f BB |
1702 | return (error); |
1703 | } | |
b128c09f | 1704 | |
428870ff | 1705 | zp->z_size = end; |
0037b49e | 1706 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), |
572e2857 | 1707 | NULL, &zp->z_size, sizeof (zp->z_size)); |
428870ff | 1708 | |
572e2857 BB |
1709 | if (end == 0) { |
1710 | zp->z_pflags &= ~ZFS_SPARSE; | |
0037b49e | 1711 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), |
572e2857 BB |
1712 | NULL, &zp->z_pflags, 8); |
1713 | } | |
1714 | VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); | |
b128c09f | 1715 | |
34dc7c2f | 1716 | dmu_tx_commit(tx); |
5d43cc9a | 1717 | rangelock_exit(lr); |
34dc7c2f BB |
1718 | |
1719 | return (0); | |
1720 | } | |
1721 | ||
b128c09f BB |
1722 | /* |
1723 | * Free space in a file | |
1724 | * | |
1725 | * IN: zp - znode of file to free data in. | |
1726 | * off - start of range | |
1727 | * len - end of range (0 => EOF) | |
1728 | * flag - current file open mode flags. | |
1729 | * log - TRUE if this action should be logged | |
1730 | * | |
19d55079 | 1731 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1732 | */ |
1733 | int | |
1734 | zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) | |
1735 | { | |
b128c09f | 1736 | dmu_tx_t *tx; |
0037b49e BB |
1737 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
1738 | zilog_t *zilog = zfsvfs->z_log; | |
428870ff BB |
1739 | uint64_t mode; |
1740 | uint64_t mtime[2], ctime[2]; | |
1741 | sa_bulk_attr_t bulk[3]; | |
1742 | int count = 0; | |
b128c09f BB |
1743 | int error; |
1744 | ||
0037b49e | 1745 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode, |
428870ff BB |
1746 | sizeof (mode))) != 0) |
1747 | return (error); | |
1748 | ||
1749 | if (off > zp->z_size) { | |
b128c09f BB |
1750 | error = zfs_extend(zp, off+len); |
1751 | if (error == 0 && log) | |
1752 | goto log; | |
223df016 | 1753 | goto out; |
b128c09f BB |
1754 | } |
1755 | ||
b128c09f BB |
1756 | if (len == 0) { |
1757 | error = zfs_trunc(zp, off); | |
1758 | } else { | |
1759 | if ((error = zfs_free_range(zp, off, len)) == 0 && | |
428870ff | 1760 | off + len > zp->z_size) |
b128c09f BB |
1761 | error = zfs_extend(zp, off+len); |
1762 | } | |
1763 | if (error || !log) | |
223df016 | 1764 | goto out; |
b128c09f | 1765 | log: |
0037b49e | 1766 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1767 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1768 | zfs_sa_upgrade_txholds(tx, zp); | |
384f8a09 | 1769 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1770 | if (error) { |
b128c09f | 1771 | dmu_tx_abort(tx); |
223df016 | 1772 | goto out; |
b128c09f BB |
1773 | } |
1774 | ||
0037b49e BB |
1775 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16); |
1776 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16); | |
1777 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), | |
428870ff | 1778 | NULL, &zp->z_pflags, 8); |
0df9673f | 1779 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
428870ff BB |
1780 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
1781 | ASSERT(error == 0); | |
1782 | ||
b128c09f BB |
1783 | zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); |
1784 | ||
1785 | dmu_tx_commit(tx); | |
223df016 | 1786 | |
960e08fe | 1787 | zfs_inode_update(zp); |
223df016 TC |
1788 | error = 0; |
1789 | ||
1790 | out: | |
1791 | /* | |
1792 | * Truncate the page cache - for file truncate operations, use | |
1793 | * the purpose-built API for truncations. For punching operations, | |
cb08f063 | 1794 | * the truncation is handled under a range lock in zfs_free_range. |
223df016 TC |
1795 | */ |
1796 | if (len == 0) | |
1797 | truncate_setsize(ZTOI(zp), off); | |
223df016 | 1798 | return (error); |
b128c09f BB |
1799 | } |
1800 | ||
34dc7c2f BB |
1801 | void |
1802 | zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) | |
1803 | { | |
22872ff5 | 1804 | struct super_block *sb; |
0037b49e | 1805 | zfsvfs_t *zfsvfs; |
428870ff | 1806 | uint64_t moid, obj, sa_obj, version; |
22872ff5 | 1807 | uint64_t sense = ZFS_CASE_SENSITIVE; |
34dc7c2f BB |
1808 | uint64_t norm = 0; |
1809 | nvpair_t *elem; | |
c96c36fa | 1810 | int size; |
34dc7c2f | 1811 | int error; |
22872ff5 BB |
1812 | int i; |
1813 | znode_t *rootzp = NULL; | |
1814 | vattr_t vattr; | |
1815 | znode_t *zp; | |
1816 | zfs_acl_ids_t acl_ids; | |
34dc7c2f BB |
1817 | |
1818 | /* | |
1819 | * First attempt to create master node. | |
1820 | */ | |
1821 | /* | |
1822 | * In an empty objset, there are no blocks to read and thus | |
1823 | * there can be no i/o errors (which we assert below). | |
1824 | */ | |
1825 | moid = MASTER_NODE_OBJ; | |
1826 | error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, | |
1827 | DMU_OT_NONE, 0, tx); | |
1828 | ASSERT(error == 0); | |
1829 | ||
1830 | /* | |
1831 | * Set starting attributes. | |
1832 | */ | |
428870ff | 1833 | version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); |
34dc7c2f BB |
1834 | elem = NULL; |
1835 | while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { | |
1836 | /* For the moment we expect all zpl props to be uint64_ts */ | |
1837 | uint64_t val; | |
1838 | char *name; | |
1839 | ||
1840 | ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); | |
1841 | VERIFY(nvpair_value_uint64(elem, &val) == 0); | |
1842 | name = nvpair_name(elem); | |
1843 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { | |
9babb374 BB |
1844 | if (val < version) |
1845 | version = val; | |
34dc7c2f BB |
1846 | } else { |
1847 | error = zap_update(os, moid, name, 8, 1, &val, tx); | |
1848 | } | |
1849 | ASSERT(error == 0); | |
1850 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) | |
1851 | norm = val; | |
22872ff5 BB |
1852 | else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) |
1853 | sense = val; | |
34dc7c2f BB |
1854 | } |
1855 | ASSERT(version != 0); | |
9babb374 | 1856 | error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); |
34dc7c2f | 1857 | |
428870ff BB |
1858 | /* |
1859 | * Create zap object used for SA attribute registration | |
1860 | */ | |
1861 | ||
1862 | if (version >= ZPL_VERSION_SA) { | |
1863 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1864 | DMU_OT_NONE, 0, tx); | |
1865 | error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
1866 | ASSERT(error == 0); | |
1867 | } else { | |
1868 | sa_obj = 0; | |
1869 | } | |
34dc7c2f BB |
1870 | /* |
1871 | * Create a delete queue. | |
1872 | */ | |
9babb374 | 1873 | obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); |
34dc7c2f | 1874 | |
9babb374 | 1875 | error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); |
34dc7c2f BB |
1876 | ASSERT(error == 0); |
1877 | ||
9babb374 | 1878 | /* |
0037b49e | 1879 | * Create root znode. Create minimal znode/inode/zfsvfs/sb |
22872ff5 | 1880 | * to allow zfs_mknode to work. |
9babb374 | 1881 | */ |
22872ff5 BB |
1882 | vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID; |
1883 | vattr.va_mode = S_IFDIR|0755; | |
1884 | vattr.va_uid = crgetuid(cr); | |
1885 | vattr.va_gid = crgetgid(cr); | |
1886 | ||
79c76d5b | 1887 | rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
22872ff5 BB |
1888 | rootzp->z_moved = 0; |
1889 | rootzp->z_unlinked = 0; | |
1890 | rootzp->z_atime_dirty = 0; | |
1891 | rootzp->z_is_sa = USE_SA(version, os); | |
9c5167d1 | 1892 | rootzp->z_pflags = 0; |
22872ff5 | 1893 | |
0037b49e BB |
1894 | zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); |
1895 | zfsvfs->z_os = os; | |
1896 | zfsvfs->z_parent = zfsvfs; | |
1897 | zfsvfs->z_version = version; | |
1898 | zfsvfs->z_use_fuids = USE_FUIDS(version, os); | |
1899 | zfsvfs->z_use_sa = USE_SA(version, os); | |
1900 | zfsvfs->z_norm = norm; | |
22872ff5 | 1901 | |
79c76d5b | 1902 | sb = kmem_zalloc(sizeof (struct super_block), KM_SLEEP); |
0037b49e | 1903 | sb->s_fs_info = zfsvfs; |
22872ff5 BB |
1904 | |
1905 | ZTOI(rootzp)->i_sb = sb; | |
1906 | ||
1907 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, | |
0037b49e | 1908 | &zfsvfs->z_attr_table); |
9babb374 | 1909 | |
22872ff5 | 1910 | ASSERT(error == 0); |
9babb374 | 1911 | |
60101509 | 1912 | /* |
22872ff5 BB |
1913 | * Fold case on file systems that are always or sometimes case |
1914 | * insensitive. | |
60101509 | 1915 | */ |
22872ff5 | 1916 | if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) |
0037b49e | 1917 | zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; |
60101509 | 1918 | |
0037b49e BB |
1919 | mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
1920 | list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), | |
22872ff5 | 1921 | offsetof(znode_t, z_link_node)); |
60101509 | 1922 | |
c96c36fa | 1923 | size = MIN(1 << (highbit64(zfs_object_mutex_size)-1), ZFS_OBJ_MTX_MAX); |
0037b49e BB |
1924 | zfsvfs->z_hold_size = size; |
1925 | zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, | |
1926 | KM_SLEEP); | |
1927 | zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
c96c36fa | 1928 | for (i = 0; i != size; i++) { |
0037b49e | 1929 | avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare, |
c96c36fa | 1930 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); |
0037b49e | 1931 | mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); |
c96c36fa | 1932 | } |
60101509 | 1933 | |
22872ff5 BB |
1934 | VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, |
1935 | cr, NULL, &acl_ids)); | |
1936 | zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); | |
1937 | ASSERT3P(zp, ==, rootzp); | |
1938 | error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); | |
1939 | ASSERT(error == 0); | |
1940 | zfs_acl_ids_free(&acl_ids); | |
60101509 | 1941 | |
22872ff5 BB |
1942 | atomic_set(&ZTOI(rootzp)->i_count, 0); |
1943 | sa_handle_destroy(rootzp->z_sa_hdl); | |
22872ff5 BB |
1944 | kmem_cache_free(znode_cache, rootzp); |
1945 | ||
c96c36fa | 1946 | for (i = 0; i != size; i++) { |
0037b49e BB |
1947 | avl_destroy(&zfsvfs->z_hold_trees[i]); |
1948 | mutex_destroy(&zfsvfs->z_hold_locks[i]); | |
c96c36fa | 1949 | } |
2708f716 | 1950 | |
c17486b2 GN |
1951 | mutex_destroy(&zfsvfs->z_znodes_lock); |
1952 | ||
0037b49e BB |
1953 | vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size); |
1954 | vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size); | |
2708f716 | 1955 | kmem_free(sb, sizeof (struct super_block)); |
0037b49e | 1956 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); |
34dc7c2f | 1957 | } |
34dc7c2f | 1958 | #endif /* _KERNEL */ |
428870ff | 1959 | |
34dc7c2f | 1960 | static int |
572e2857 BB |
1961 | zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) |
1962 | { | |
1963 | uint64_t sa_obj = 0; | |
1964 | int error; | |
1965 | ||
1966 | error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
1967 | if (error != 0 && error != ENOENT) | |
1968 | return (error); | |
1969 | ||
1970 | error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); | |
1971 | return (error); | |
1972 | } | |
1973 | ||
1974 | static int | |
1975 | zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, | |
7b8518cb | 1976 | dmu_buf_t **db, void *tag) |
34dc7c2f | 1977 | { |
34dc7c2f | 1978 | dmu_object_info_t doi; |
34dc7c2f | 1979 | int error; |
428870ff | 1980 | |
7b8518cb | 1981 | if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) |
34dc7c2f BB |
1982 | return (error); |
1983 | ||
572e2857 | 1984 | dmu_object_info_from_db(*db, &doi); |
428870ff BB |
1985 | if ((doi.doi_bonus_type != DMU_OT_SA && |
1986 | doi.doi_bonus_type != DMU_OT_ZNODE) || | |
d6320ddb BB |
1987 | (doi.doi_bonus_type == DMU_OT_ZNODE && |
1988 | doi.doi_bonus_size < sizeof (znode_phys_t))) { | |
7b8518cb | 1989 | sa_buf_rele(*db, tag); |
2e528b49 | 1990 | return (SET_ERROR(ENOTSUP)); |
34dc7c2f BB |
1991 | } |
1992 | ||
572e2857 BB |
1993 | error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); |
1994 | if (error != 0) { | |
7b8518cb | 1995 | sa_buf_rele(*db, tag); |
428870ff BB |
1996 | return (error); |
1997 | } | |
1998 | ||
572e2857 BB |
1999 | return (0); |
2000 | } | |
2001 | ||
2002 | void | |
7b8518cb | 2003 | zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) |
572e2857 BB |
2004 | { |
2005 | sa_handle_destroy(hdl); | |
7b8518cb | 2006 | sa_buf_rele(db, tag); |
572e2857 BB |
2007 | } |
2008 | ||
2009 | /* | |
2010 | * Given an object number, return its parent object number and whether | |
2011 | * or not the object is an extended attribute directory. | |
2012 | */ | |
2013 | static int | |
b23ad7f3 JJ |
2014 | zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, |
2015 | uint64_t *pobjp, int *is_xattrdir) | |
572e2857 BB |
2016 | { |
2017 | uint64_t parent; | |
2018 | uint64_t pflags; | |
2019 | uint64_t mode; | |
b23ad7f3 | 2020 | uint64_t parent_mode; |
572e2857 | 2021 | sa_bulk_attr_t bulk[3]; |
b23ad7f3 JJ |
2022 | sa_handle_t *sa_hdl; |
2023 | dmu_buf_t *sa_db; | |
572e2857 BB |
2024 | int count = 0; |
2025 | int error; | |
2026 | ||
2027 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, | |
2028 | &parent, sizeof (parent)); | |
428870ff | 2029 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, |
572e2857 | 2030 | &pflags, sizeof (pflags)); |
428870ff | 2031 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, |
572e2857 | 2032 | &mode, sizeof (mode)); |
428870ff | 2033 | |
572e2857 | 2034 | if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) |
428870ff | 2035 | return (error); |
572e2857 | 2036 | |
b23ad7f3 JJ |
2037 | /* |
2038 | * When a link is removed its parent pointer is not changed and will | |
2039 | * be invalid. There are two cases where a link is removed but the | |
2040 | * file stays around, when it goes to the delete queue and when there | |
2041 | * are additional links. | |
2042 | */ | |
2043 | error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); | |
2044 | if (error != 0) | |
2045 | return (error); | |
2046 | ||
2047 | error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); | |
2048 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); | |
2049 | if (error != 0) | |
2050 | return (error); | |
2051 | ||
428870ff | 2052 | *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); |
34dc7c2f | 2053 | |
b23ad7f3 JJ |
2054 | /* |
2055 | * Extended attributes can be applied to files, directories, etc. | |
2056 | * Otherwise the parent must be a directory. | |
2057 | */ | |
2058 | if (!*is_xattrdir && !S_ISDIR(parent_mode)) | |
ecb2b7dc | 2059 | return (SET_ERROR(EINVAL)); |
b23ad7f3 JJ |
2060 | |
2061 | *pobjp = parent; | |
2062 | ||
34dc7c2f BB |
2063 | return (0); |
2064 | } | |
2065 | ||
572e2857 BB |
2066 | /* |
2067 | * Given an object number, return some zpl level statistics | |
2068 | */ | |
2069 | static int | |
2070 | zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, | |
2071 | zfs_stat_t *sb) | |
34dc7c2f | 2072 | { |
572e2857 BB |
2073 | sa_bulk_attr_t bulk[4]; |
2074 | int count = 0; | |
2075 | ||
2076 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, | |
2077 | &sb->zs_mode, sizeof (sb->zs_mode)); | |
2078 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, | |
2079 | &sb->zs_gen, sizeof (sb->zs_gen)); | |
2080 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, | |
2081 | &sb->zs_links, sizeof (sb->zs_links)); | |
2082 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, | |
2083 | &sb->zs_ctime, sizeof (sb->zs_ctime)); | |
2084 | ||
2085 | return (sa_bulk_lookup(hdl, bulk, count)); | |
2086 | } | |
2087 | ||
2088 | static int | |
2089 | zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, | |
2090 | sa_attr_type_t *sa_table, char *buf, int len) | |
2091 | { | |
2092 | sa_handle_t *sa_hdl; | |
2093 | sa_handle_t *prevhdl = NULL; | |
2094 | dmu_buf_t *prevdb = NULL; | |
2095 | dmu_buf_t *sa_db = NULL; | |
34dc7c2f BB |
2096 | char *path = buf + len - 1; |
2097 | int error; | |
2098 | ||
2099 | *path = '\0'; | |
572e2857 | 2100 | sa_hdl = hdl; |
428870ff | 2101 | |
64c1dcef PD |
2102 | uint64_t deleteq_obj; |
2103 | VERIFY0(zap_lookup(osp, MASTER_NODE_OBJ, | |
2104 | ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj)); | |
2105 | error = zap_lookup_int(osp, deleteq_obj, obj); | |
2106 | if (error == 0) { | |
2107 | return (ESTALE); | |
2108 | } else if (error != ENOENT) { | |
2109 | return (error); | |
2110 | } | |
2111 | error = 0; | |
2112 | ||
34dc7c2f | 2113 | for (;;) { |
17897ce2 | 2114 | uint64_t pobj = 0; |
34dc7c2f BB |
2115 | char component[MAXNAMELEN + 2]; |
2116 | size_t complen; | |
17897ce2 | 2117 | int is_xattrdir = 0; |
34dc7c2f | 2118 | |
572e2857 | 2119 | if (prevdb) |
7b8518cb | 2120 | zfs_release_sa_handle(prevhdl, prevdb, FTAG); |
572e2857 | 2121 | |
b23ad7f3 | 2122 | if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, |
572e2857 | 2123 | &is_xattrdir)) != 0) |
34dc7c2f BB |
2124 | break; |
2125 | ||
2126 | if (pobj == obj) { | |
2127 | if (path[0] != '/') | |
2128 | *--path = '/'; | |
2129 | break; | |
2130 | } | |
2131 | ||
2132 | component[0] = '/'; | |
2133 | if (is_xattrdir) { | |
2134 | (void) sprintf(component + 1, "<xattrdir>"); | |
2135 | } else { | |
2136 | error = zap_value_search(osp, pobj, obj, | |
2137 | ZFS_DIRENT_OBJ(-1ULL), component + 1); | |
2138 | if (error != 0) | |
2139 | break; | |
2140 | } | |
2141 | ||
2142 | complen = strlen(component); | |
2143 | path -= complen; | |
2144 | ASSERT(path >= buf); | |
2145 | bcopy(component, path, complen); | |
2146 | obj = pobj; | |
572e2857 BB |
2147 | |
2148 | if (sa_hdl != hdl) { | |
2149 | prevhdl = sa_hdl; | |
2150 | prevdb = sa_db; | |
2151 | } | |
7b8518cb | 2152 | error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); |
572e2857 BB |
2153 | if (error != 0) { |
2154 | sa_hdl = prevhdl; | |
2155 | sa_db = prevdb; | |
2156 | break; | |
2157 | } | |
2158 | } | |
2159 | ||
2160 | if (sa_hdl != NULL && sa_hdl != hdl) { | |
2161 | ASSERT(sa_db != NULL); | |
7b8518cb | 2162 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); |
34dc7c2f BB |
2163 | } |
2164 | ||
2165 | if (error == 0) | |
2166 | (void) memmove(buf, path, buf + len - path); | |
428870ff | 2167 | |
34dc7c2f BB |
2168 | return (error); |
2169 | } | |
572e2857 BB |
2170 | |
2171 | int | |
2172 | zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) | |
2173 | { | |
2174 | sa_attr_type_t *sa_table; | |
2175 | sa_handle_t *hdl; | |
2176 | dmu_buf_t *db; | |
2177 | int error; | |
2178 | ||
2179 | error = zfs_sa_setup(osp, &sa_table); | |
2180 | if (error != 0) | |
2181 | return (error); | |
2182 | ||
7b8518cb | 2183 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2184 | if (error != 0) |
2185 | return (error); | |
2186 | ||
2187 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2188 | ||
7b8518cb | 2189 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2190 | return (error); |
2191 | } | |
2192 | ||
2193 | int | |
2194 | zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, | |
2195 | char *buf, int len) | |
2196 | { | |
2197 | char *path = buf + len - 1; | |
2198 | sa_attr_type_t *sa_table; | |
2199 | sa_handle_t *hdl; | |
2200 | dmu_buf_t *db; | |
2201 | int error; | |
2202 | ||
2203 | *path = '\0'; | |
2204 | ||
2205 | error = zfs_sa_setup(osp, &sa_table); | |
2206 | if (error != 0) | |
2207 | return (error); | |
2208 | ||
7b8518cb | 2209 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2210 | if (error != 0) |
2211 | return (error); | |
2212 | ||
2213 | error = zfs_obj_to_stats_impl(hdl, sa_table, sb); | |
2214 | if (error != 0) { | |
7b8518cb | 2215 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2216 | return (error); |
2217 | } | |
2218 | ||
2219 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2220 | ||
7b8518cb | 2221 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2222 | return (error); |
2223 | } | |
c28b2279 | 2224 | |
93ce2b4c | 2225 | #if defined(_KERNEL) |
c28b2279 BB |
2226 | EXPORT_SYMBOL(zfs_create_fs); |
2227 | EXPORT_SYMBOL(zfs_obj_to_path); | |
0720116d | 2228 | |
02730c33 | 2229 | /* CSTYLED */ |
0720116d BB |
2230 | module_param(zfs_object_mutex_size, uint, 0644); |
2231 | MODULE_PARM_DESC(zfs_object_mutex_size, "Size of znode hold array"); | |
dcec0a12 AP |
2232 | module_param(zfs_unlink_suspend_progress, int, 0644); |
2233 | MODULE_PARM_DESC(zfs_unlink_suspend_progress, "Set to prevent async unlinks " | |
2234 | "(debug - leaks space into the unlinked set)"); | |
c28b2279 | 2235 | #endif |