]>
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 | |
5d43cc9a MA |
94 | /* |
95 | * This callback is invoked when acquiring a RL_WRITER or RL_APPEND lock on | |
96 | * z_rangelock. It will modify the offset and length of the lock to reflect | |
97 | * znode-specific information, and convert RL_APPEND to RL_WRITER. This is | |
98 | * called with the rangelock_t's rl_lock held, which avoids races. | |
99 | */ | |
100 | static void | |
101 | zfs_rangelock_cb(locked_range_t *new, void *arg) | |
102 | { | |
103 | znode_t *zp = arg; | |
104 | ||
105 | /* | |
106 | * If in append mode, convert to writer and lock starting at the | |
107 | * current end of file. | |
108 | */ | |
109 | if (new->lr_type == RL_APPEND) { | |
110 | new->lr_offset = zp->z_size; | |
111 | new->lr_type = RL_WRITER; | |
112 | } | |
113 | ||
114 | /* | |
115 | * If we need to grow the block size then lock the whole file range. | |
116 | */ | |
117 | uint64_t end_size = MAX(zp->z_size, new->lr_offset + new->lr_length); | |
118 | if (end_size > zp->z_blksz && (!ISP2(zp->z_blksz) || | |
119 | zp->z_blksz < ZTOZSB(zp)->z_max_blksz)) { | |
120 | new->lr_offset = 0; | |
121 | new->lr_length = UINT64_MAX; | |
122 | } | |
123 | } | |
124 | ||
34dc7c2f BB |
125 | /*ARGSUSED*/ |
126 | static int | |
b128c09f | 127 | zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) |
34dc7c2f BB |
128 | { |
129 | znode_t *zp = buf; | |
130 | ||
3558fd73 | 131 | inode_init_once(ZTOI(zp)); |
b128c09f BB |
132 | list_link_init(&zp->z_link_node); |
133 | ||
34dc7c2f | 134 | mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 135 | rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); |
448d7aaa | 136 | rw_init(&zp->z_name_lock, NULL, RW_NOLOCKDEP, NULL); |
34dc7c2f | 137 | mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); |
82a37189 | 138 | rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); |
34dc7c2f | 139 | |
5d43cc9a | 140 | rangelock_init(&zp->z_rangelock, zfs_rangelock_cb, zp); |
34dc7c2f | 141 | |
b128c09f | 142 | zp->z_dirlocks = NULL; |
45d1cae3 | 143 | zp->z_acl_cached = NULL; |
82a37189 | 144 | zp->z_xattr_cached = NULL; |
98701490 | 145 | zp->z_xattr_parent = 0; |
572e2857 | 146 | zp->z_moved = 0; |
34dc7c2f BB |
147 | return (0); |
148 | } | |
149 | ||
150 | /*ARGSUSED*/ | |
151 | static void | |
b128c09f | 152 | zfs_znode_cache_destructor(void *buf, void *arg) |
34dc7c2f BB |
153 | { |
154 | znode_t *zp = buf; | |
155 | ||
b128c09f | 156 | ASSERT(!list_link_active(&zp->z_link_node)); |
34dc7c2f | 157 | mutex_destroy(&zp->z_lock); |
34dc7c2f BB |
158 | rw_destroy(&zp->z_parent_lock); |
159 | rw_destroy(&zp->z_name_lock); | |
160 | mutex_destroy(&zp->z_acl_lock); | |
82a37189 | 161 | rw_destroy(&zp->z_xattr_lock); |
5d43cc9a | 162 | rangelock_fini(&zp->z_rangelock); |
34dc7c2f | 163 | |
b128c09f | 164 | ASSERT(zp->z_dirlocks == NULL); |
45d1cae3 | 165 | ASSERT(zp->z_acl_cached == NULL); |
82a37189 | 166 | ASSERT(zp->z_xattr_cached == NULL); |
b128c09f BB |
167 | } |
168 | ||
c96c36fa BB |
169 | static int |
170 | zfs_znode_hold_cache_constructor(void *buf, void *arg, int kmflags) | |
171 | { | |
172 | znode_hold_t *zh = buf; | |
173 | ||
174 | mutex_init(&zh->zh_lock, NULL, MUTEX_DEFAULT, NULL); | |
424fd7c3 | 175 | zfs_refcount_create(&zh->zh_refcount); |
c96c36fa BB |
176 | zh->zh_obj = ZFS_NO_OBJECT; |
177 | ||
178 | return (0); | |
179 | } | |
180 | ||
181 | static void | |
182 | zfs_znode_hold_cache_destructor(void *buf, void *arg) | |
183 | { | |
184 | znode_hold_t *zh = buf; | |
185 | ||
186 | mutex_destroy(&zh->zh_lock); | |
424fd7c3 | 187 | zfs_refcount_destroy(&zh->zh_refcount); |
c96c36fa BB |
188 | } |
189 | ||
34dc7c2f BB |
190 | void |
191 | zfs_znode_init(void) | |
192 | { | |
193 | /* | |
5074bfe8 TC |
194 | * Initialize zcache. The KMC_SLAB hint is used in order that it be |
195 | * backed by kmalloc() when on the Linux slab in order that any | |
196 | * wait_on_bit() operations on the related inode operate properly. | |
34dc7c2f BB |
197 | */ |
198 | ASSERT(znode_cache == NULL); | |
199 | znode_cache = kmem_cache_create("zfs_znode_cache", | |
200 | sizeof (znode_t), 0, zfs_znode_cache_constructor, | |
5074bfe8 | 201 | zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_SLAB); |
c96c36fa BB |
202 | |
203 | ASSERT(znode_hold_cache == NULL); | |
204 | znode_hold_cache = kmem_cache_create("zfs_znode_hold_cache", | |
205 | sizeof (znode_hold_t), 0, zfs_znode_hold_cache_constructor, | |
206 | zfs_znode_hold_cache_destructor, NULL, NULL, NULL, 0); | |
34dc7c2f BB |
207 | } |
208 | ||
209 | void | |
210 | zfs_znode_fini(void) | |
211 | { | |
34dc7c2f BB |
212 | /* |
213 | * Cleanup zcache | |
214 | */ | |
215 | if (znode_cache) | |
216 | kmem_cache_destroy(znode_cache); | |
217 | znode_cache = NULL; | |
c96c36fa BB |
218 | |
219 | if (znode_hold_cache) | |
220 | kmem_cache_destroy(znode_hold_cache); | |
221 | znode_hold_cache = NULL; | |
222 | } | |
223 | ||
224 | /* | |
225 | * The zfs_znode_hold_enter() / zfs_znode_hold_exit() functions are used to | |
226 | * serialize access to a znode and its SA buffer while the object is being | |
227 | * created or destroyed. This kind of locking would normally reside in the | |
228 | * znode itself but in this case that's impossible because the znode and SA | |
229 | * buffer may not yet exist. Therefore the locking is handled externally | |
230 | * with an array of mutexs and AVLs trees which contain per-object locks. | |
231 | * | |
232 | * In zfs_znode_hold_enter() a per-object lock is created as needed, inserted | |
233 | * in to the correct AVL tree and finally the per-object lock is held. In | |
234 | * zfs_znode_hold_exit() the process is reversed. The per-object lock is | |
235 | * released, removed from the AVL tree and destroyed if there are no waiters. | |
236 | * | |
237 | * This scheme has two important properties: | |
238 | * | |
239 | * 1) No memory allocations are performed while holding one of the z_hold_locks. | |
240 | * This ensures evict(), which can be called from direct memory reclaim, will | |
241 | * never block waiting on a z_hold_locks which just happens to have hashed | |
242 | * to the same index. | |
243 | * | |
244 | * 2) All locks used to serialize access to an object are per-object and never | |
245 | * shared. This minimizes lock contention without creating a large number | |
246 | * of dedicated locks. | |
247 | * | |
248 | * On the downside it does require znode_lock_t structures to be frequently | |
249 | * allocated and freed. However, because these are backed by a kmem cache | |
250 | * and very short lived this cost is minimal. | |
251 | */ | |
252 | int | |
253 | zfs_znode_hold_compare(const void *a, const void *b) | |
254 | { | |
ee36c709 GN |
255 | const znode_hold_t *zh_a = (const znode_hold_t *)a; |
256 | const znode_hold_t *zh_b = (const znode_hold_t *)b; | |
257 | ||
258 | return (AVL_CMP(zh_a->zh_obj, zh_b->zh_obj)); | |
c96c36fa BB |
259 | } |
260 | ||
261 | boolean_t | |
0037b49e | 262 | zfs_znode_held(zfsvfs_t *zfsvfs, uint64_t obj) |
c96c36fa BB |
263 | { |
264 | znode_hold_t *zh, search; | |
0037b49e | 265 | int i = ZFS_OBJ_HASH(zfsvfs, obj); |
37c56346 | 266 | boolean_t held; |
c96c36fa BB |
267 | |
268 | search.zh_obj = obj; | |
269 | ||
0037b49e BB |
270 | mutex_enter(&zfsvfs->z_hold_locks[i]); |
271 | zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL); | |
37c56346 | 272 | held = (zh && MUTEX_HELD(&zh->zh_lock)) ? B_TRUE : B_FALSE; |
0037b49e | 273 | mutex_exit(&zfsvfs->z_hold_locks[i]); |
c96c36fa | 274 | |
37c56346 | 275 | return (held); |
c96c36fa BB |
276 | } |
277 | ||
278 | static znode_hold_t * | |
0037b49e | 279 | zfs_znode_hold_enter(zfsvfs_t *zfsvfs, uint64_t obj) |
c96c36fa BB |
280 | { |
281 | znode_hold_t *zh, *zh_new, search; | |
0037b49e | 282 | int i = ZFS_OBJ_HASH(zfsvfs, obj); |
c96c36fa BB |
283 | boolean_t found = B_FALSE; |
284 | ||
285 | zh_new = kmem_cache_alloc(znode_hold_cache, KM_SLEEP); | |
286 | zh_new->zh_obj = obj; | |
287 | search.zh_obj = obj; | |
288 | ||
0037b49e BB |
289 | mutex_enter(&zfsvfs->z_hold_locks[i]); |
290 | zh = avl_find(&zfsvfs->z_hold_trees[i], &search, NULL); | |
c96c36fa BB |
291 | if (likely(zh == NULL)) { |
292 | zh = zh_new; | |
0037b49e | 293 | avl_add(&zfsvfs->z_hold_trees[i], zh); |
c96c36fa BB |
294 | } else { |
295 | ASSERT3U(zh->zh_obj, ==, obj); | |
296 | found = B_TRUE; | |
297 | } | |
c13060e4 | 298 | zfs_refcount_add(&zh->zh_refcount, NULL); |
0037b49e | 299 | mutex_exit(&zfsvfs->z_hold_locks[i]); |
c96c36fa BB |
300 | |
301 | if (found == B_TRUE) | |
302 | kmem_cache_free(znode_hold_cache, zh_new); | |
303 | ||
304 | ASSERT(MUTEX_NOT_HELD(&zh->zh_lock)); | |
424fd7c3 | 305 | ASSERT3S(zfs_refcount_count(&zh->zh_refcount), >, 0); |
c96c36fa BB |
306 | mutex_enter(&zh->zh_lock); |
307 | ||
308 | return (zh); | |
309 | } | |
310 | ||
311 | static void | |
0037b49e | 312 | zfs_znode_hold_exit(zfsvfs_t *zfsvfs, znode_hold_t *zh) |
c96c36fa | 313 | { |
0037b49e | 314 | int i = ZFS_OBJ_HASH(zfsvfs, zh->zh_obj); |
c96c36fa BB |
315 | boolean_t remove = B_FALSE; |
316 | ||
0037b49e | 317 | ASSERT(zfs_znode_held(zfsvfs, zh->zh_obj)); |
424fd7c3 | 318 | ASSERT3S(zfs_refcount_count(&zh->zh_refcount), >, 0); |
c96c36fa BB |
319 | mutex_exit(&zh->zh_lock); |
320 | ||
0037b49e | 321 | mutex_enter(&zfsvfs->z_hold_locks[i]); |
424fd7c3 | 322 | if (zfs_refcount_remove(&zh->zh_refcount, NULL) == 0) { |
0037b49e | 323 | avl_remove(&zfsvfs->z_hold_trees[i], zh); |
c96c36fa BB |
324 | remove = B_TRUE; |
325 | } | |
0037b49e | 326 | mutex_exit(&zfsvfs->z_hold_locks[i]); |
c96c36fa BB |
327 | |
328 | if (remove == B_TRUE) | |
329 | kmem_cache_free(znode_hold_cache, zh); | |
34dc7c2f BB |
330 | } |
331 | ||
34dc7c2f | 332 | int |
0037b49e | 333 | zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) |
34dc7c2f | 334 | { |
3c9609b3 | 335 | #ifdef HAVE_SMB_SHARE |
9babb374 BB |
336 | zfs_acl_ids_t acl_ids; |
337 | vattr_t vattr; | |
338 | znode_t *sharezp; | |
339 | vnode_t *vp; | |
340 | znode_t *zp; | |
341 | int error; | |
34dc7c2f | 342 | |
9babb374 | 343 | vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; |
3558fd73 | 344 | vattr.va_mode = S_IFDIR | 0555; |
9babb374 BB |
345 | vattr.va_uid = crgetuid(kcred); |
346 | vattr.va_gid = crgetgid(kcred); | |
34dc7c2f | 347 | |
79c76d5b | 348 | sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
572e2857 | 349 | sharezp->z_moved = 0; |
9babb374 BB |
350 | sharezp->z_unlinked = 0; |
351 | sharezp->z_atime_dirty = 0; | |
352 | sharezp->z_zfsvfs = zfsvfs; | |
428870ff | 353 | sharezp->z_is_sa = zfsvfs->z_use_sa; |
9c5167d1 | 354 | sharezp->z_pflags = 0; |
34dc7c2f | 355 | |
9babb374 BB |
356 | vp = ZTOV(sharezp); |
357 | vn_reinit(vp); | |
358 | vp->v_type = VDIR; | |
34dc7c2f | 359 | |
9babb374 BB |
360 | VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, |
361 | kcred, NULL, &acl_ids)); | |
428870ff | 362 | zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); |
9babb374 BB |
363 | ASSERT3P(zp, ==, sharezp); |
364 | ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */ | |
365 | POINTER_INVALIDATE(&sharezp->z_zfsvfs); | |
366 | error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, | |
367 | ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); | |
368 | zfsvfs->z_shares_dir = sharezp->z_id; | |
369 | ||
370 | zfs_acl_ids_free(&acl_ids); | |
3558fd73 | 371 | // ZTOV(sharezp)->v_count = 0; |
428870ff | 372 | sa_handle_destroy(sharezp->z_sa_hdl); |
9babb374 | 373 | kmem_cache_free(znode_cache, sharezp); |
34dc7c2f | 374 | |
9babb374 | 375 | return (error); |
9ee7fac5 BB |
376 | #else |
377 | return (0); | |
3c9609b3 | 378 | #endif /* HAVE_SMB_SHARE */ |
34dc7c2f BB |
379 | } |
380 | ||
34dc7c2f | 381 | static void |
0037b49e | 382 | zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, |
428870ff | 383 | dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) |
34dc7c2f | 384 | { |
0037b49e | 385 | ASSERT(zfs_znode_held(zfsvfs, zp->z_id)); |
34dc7c2f BB |
386 | |
387 | mutex_enter(&zp->z_lock); | |
388 | ||
428870ff BB |
389 | ASSERT(zp->z_sa_hdl == NULL); |
390 | ASSERT(zp->z_acl_cached == NULL); | |
391 | if (sa_hdl == NULL) { | |
0037b49e | 392 | VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp, |
428870ff BB |
393 | SA_HDL_SHARED, &zp->z_sa_hdl)); |
394 | } else { | |
395 | zp->z_sa_hdl = sa_hdl; | |
396 | sa_set_userp(sa_hdl, zp); | |
397 | } | |
34dc7c2f | 398 | |
428870ff | 399 | zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; |
34dc7c2f | 400 | |
34dc7c2f | 401 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
402 | } |
403 | ||
404 | void | |
405 | zfs_znode_dmu_fini(znode_t *zp) | |
406 | { | |
c96c36fa | 407 | ASSERT(zfs_znode_held(ZTOZSB(zp), zp->z_id) || zp->z_unlinked || |
3558fd73 | 408 | RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock)); |
428870ff BB |
409 | |
410 | sa_handle_destroy(zp->z_sa_hdl); | |
411 | zp->z_sa_hdl = NULL; | |
34dc7c2f BB |
412 | } |
413 | ||
414 | /* | |
3558fd73 BB |
415 | * Called by new_inode() to allocate a new inode. |
416 | */ | |
417 | int | |
418 | zfs_inode_alloc(struct super_block *sb, struct inode **ip) | |
419 | { | |
420 | znode_t *zp; | |
421 | ||
79c76d5b | 422 | zp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
3558fd73 BB |
423 | *ip = ZTOI(zp); |
424 | ||
425 | return (0); | |
426 | } | |
427 | ||
428 | /* | |
429 | * Called in multiple places when an inode should be destroyed. | |
430 | */ | |
431 | void | |
432 | zfs_inode_destroy(struct inode *ip) | |
433 | { | |
434 | znode_t *zp = ITOZ(ip); | |
0037b49e | 435 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
3558fd73 | 436 | |
0037b49e | 437 | mutex_enter(&zfsvfs->z_znodes_lock); |
7b3e34ba | 438 | if (list_link_active(&zp->z_link_node)) { |
0037b49e BB |
439 | list_remove(&zfsvfs->z_all_znodes, zp); |
440 | zfsvfs->z_nr_znodes--; | |
7b3e34ba | 441 | } |
0037b49e | 442 | mutex_exit(&zfsvfs->z_znodes_lock); |
3558fd73 BB |
443 | |
444 | if (zp->z_acl_cached) { | |
445 | zfs_acl_free(zp->z_acl_cached); | |
446 | zp->z_acl_cached = NULL; | |
447 | } | |
448 | ||
82a37189 BB |
449 | if (zp->z_xattr_cached) { |
450 | nvlist_free(zp->z_xattr_cached); | |
451 | zp->z_xattr_cached = NULL; | |
452 | } | |
453 | ||
3558fd73 BB |
454 | kmem_cache_free(znode_cache, zp); |
455 | } | |
456 | ||
457 | static void | |
0037b49e | 458 | zfs_inode_set_ops(zfsvfs_t *zfsvfs, struct inode *ip) |
3558fd73 | 459 | { |
aa6d8c10 | 460 | uint64_t rdev = 0; |
3558fd73 BB |
461 | |
462 | switch (ip->i_mode & S_IFMT) { | |
463 | case S_IFREG: | |
464 | ip->i_op = &zpl_inode_operations; | |
465 | ip->i_fop = &zpl_file_operations; | |
466 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
467 | break; | |
468 | ||
469 | case S_IFDIR: | |
470 | ip->i_op = &zpl_dir_inode_operations; | |
471 | ip->i_fop = &zpl_dir_file_operations; | |
472 | ITOZ(ip)->z_zn_prefetch = B_TRUE; | |
473 | break; | |
474 | ||
475 | case S_IFLNK: | |
476 | ip->i_op = &zpl_symlink_inode_operations; | |
477 | break; | |
478 | ||
aa6d8c10 NB |
479 | /* |
480 | * rdev is only stored in a SA only for device files. | |
481 | */ | |
3558fd73 BB |
482 | case S_IFCHR: |
483 | case S_IFBLK: | |
0037b49e | 484 | (void) sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zfsvfs), &rdev, |
53b1d979 | 485 | sizeof (rdev)); |
aa6d8c10 NB |
486 | /*FALLTHROUGH*/ |
487 | case S_IFIFO: | |
488 | case S_IFSOCK: | |
3558fd73 BB |
489 | init_special_inode(ip, ip->i_mode, rdev); |
490 | ip->i_op = &zpl_special_inode_operations; | |
491 | break; | |
492 | ||
493 | default: | |
53b1d979 BB |
494 | zfs_panic_recover("inode %llu has invalid mode: 0x%x\n", |
495 | (u_longlong_t)ip->i_ino, ip->i_mode); | |
496 | ||
497 | /* Assume the inode is a file and attempt to continue */ | |
498 | ip->i_mode = S_IFREG | 0644; | |
499 | ip->i_op = &zpl_inode_operations; | |
500 | ip->i_fop = &zpl_file_operations; | |
501 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
502 | break; | |
3558fd73 BB |
503 | } |
504 | } | |
505 | ||
7bb1325f CC |
506 | void |
507 | zfs_set_inode_flags(znode_t *zp, struct inode *ip) | |
508 | { | |
509 | /* | |
510 | * Linux and Solaris have different sets of file attributes, so we | |
511 | * restrict this conversion to the intersection of the two. | |
512 | */ | |
a5248129 CC |
513 | #ifdef HAVE_INODE_SET_FLAGS |
514 | unsigned int flags = 0; | |
515 | if (zp->z_pflags & ZFS_IMMUTABLE) | |
516 | flags |= S_IMMUTABLE; | |
517 | if (zp->z_pflags & ZFS_APPENDONLY) | |
518 | flags |= S_APPEND; | |
7bb1325f | 519 | |
a5248129 CC |
520 | inode_set_flags(ip, flags, S_IMMUTABLE|S_APPEND); |
521 | #else | |
7bb1325f CC |
522 | if (zp->z_pflags & ZFS_IMMUTABLE) |
523 | ip->i_flags |= S_IMMUTABLE; | |
524 | else | |
525 | ip->i_flags &= ~S_IMMUTABLE; | |
526 | ||
527 | if (zp->z_pflags & ZFS_APPENDONLY) | |
528 | ip->i_flags |= S_APPEND; | |
529 | else | |
530 | ip->i_flags &= ~S_APPEND; | |
a5248129 | 531 | #endif |
7bb1325f CC |
532 | } |
533 | ||
704cd075 CC |
534 | /* |
535 | * Update the embedded inode given the znode. We should work toward | |
536 | * eliminating this function as soon as possible by removing values | |
537 | * which are duplicated between the znode and inode. If the generic | |
538 | * inode has the correct field it should be used, and the ZFS code | |
539 | * updated to access the inode. This can be done incrementally. | |
540 | */ | |
9f5f0019 NB |
541 | void |
542 | zfs_inode_update(znode_t *zp) | |
704cd075 | 543 | { |
0037b49e | 544 | zfsvfs_t *zfsvfs; |
704cd075 CC |
545 | struct inode *ip; |
546 | uint32_t blksize; | |
547 | u_longlong_t i_blocks; | |
704cd075 CC |
548 | |
549 | ASSERT(zp != NULL); | |
0037b49e | 550 | zfsvfs = ZTOZSB(zp); |
704cd075 CC |
551 | ip = ZTOI(zp); |
552 | ||
553 | /* Skip .zfs control nodes which do not exist on disk. */ | |
554 | if (zfsctl_is_node(ip)) | |
555 | return; | |
556 | ||
704cd075 CC |
557 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, &i_blocks); |
558 | ||
559 | spin_lock(&ip->i_lock); | |
704cd075 | 560 | ip->i_blocks = i_blocks; |
704cd075 CC |
561 | i_size_write(ip, zp->z_size); |
562 | spin_unlock(&ip->i_lock); | |
563 | } | |
564 | ||
704cd075 | 565 | |
3558fd73 BB |
566 | /* |
567 | * Construct a znode+inode and initialize. | |
34dc7c2f BB |
568 | * |
569 | * This does not do a call to dmu_set_user() that is | |
570 | * up to the caller to do, in case you don't want to | |
571 | * return the znode | |
572 | */ | |
573 | static znode_t * | |
0037b49e | 574 | zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, |
31b6111f | 575 | dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl) |
34dc7c2f BB |
576 | { |
577 | znode_t *zp; | |
3558fd73 | 578 | struct inode *ip; |
7f89ae6b | 579 | uint64_t mode; |
428870ff | 580 | uint64_t parent; |
278f2236 | 581 | uint64_t tmp_gen; |
dfbc8630 | 582 | uint64_t links; |
2c6abf15 | 583 | uint64_t z_uid, z_gid; |
9f5f0019 | 584 | uint64_t atime[2], mtime[2], ctime[2]; |
9c5167d1 | 585 | uint64_t projid = ZFS_DEFAULT_PROJID; |
9f5f0019 | 586 | sa_bulk_attr_t bulk[11]; |
428870ff | 587 | int count = 0; |
34dc7c2f | 588 | |
0037b49e | 589 | ASSERT(zfsvfs != NULL); |
34dc7c2f | 590 | |
0037b49e | 591 | ip = new_inode(zfsvfs->z_sb); |
3558fd73 BB |
592 | if (ip == NULL) |
593 | return (NULL); | |
7304b6e5 | 594 | |
3558fd73 | 595 | zp = ITOZ(ip); |
34dc7c2f | 596 | ASSERT(zp->z_dirlocks == NULL); |
ebe7e575 BB |
597 | ASSERT3P(zp->z_acl_cached, ==, NULL); |
598 | ASSERT3P(zp->z_xattr_cached, ==, NULL); | |
572e2857 | 599 | zp->z_moved = 0; |
428870ff | 600 | zp->z_sa_hdl = NULL; |
34dc7c2f BB |
601 | zp->z_unlinked = 0; |
602 | zp->z_atime_dirty = 0; | |
603 | zp->z_mapcnt = 0; | |
34dc7c2f BB |
604 | zp->z_id = db->db_object; |
605 | zp->z_blksz = blksz; | |
606 | zp->z_seq = 0x7A4653; | |
607 | zp->z_sync_cnt = 0; | |
ebe7e575 BB |
608 | zp->z_is_mapped = B_FALSE; |
609 | zp->z_is_ctldir = B_FALSE; | |
7b3e34ba | 610 | zp->z_is_stale = B_FALSE; |
34dc7c2f | 611 | |
0037b49e | 612 | zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); |
3558fd73 | 613 | |
0037b49e BB |
614 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); |
615 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &tmp_gen, 8); | |
616 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, | |
617 | &zp->z_size, 8); | |
618 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); | |
619 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, | |
428870ff | 620 | &zp->z_pflags, 8); |
0037b49e | 621 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, |
7304b6e5 | 622 | &parent, 8); |
0037b49e BB |
623 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, &z_uid, 8); |
624 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, &z_gid, 8); | |
625 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, &atime, 16); | |
626 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); | |
627 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); | |
428870ff | 628 | |
9c5167d1 NF |
629 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || tmp_gen == 0 || |
630 | (dmu_objset_projectquota_enabled(zfsvfs->z_os) && | |
631 | (zp->z_pflags & ZFS_PROJID) && | |
632 | sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), &projid, 8) != 0)) { | |
428870ff BB |
633 | if (hdl == NULL) |
634 | sa_handle_destroy(zp->z_sa_hdl); | |
07d63f0c | 635 | zp->z_sa_hdl = NULL; |
3558fd73 | 636 | goto error; |
34dc7c2f | 637 | } |
7304b6e5 | 638 | |
9c5167d1 | 639 | zp->z_projid = projid; |
12fa7f34 | 640 | zp->z_mode = ip->i_mode = mode; |
278f2236 | 641 | ip->i_generation = (uint32_t)tmp_gen; |
ba2fe6af | 642 | ip->i_blkbits = SPA_MINBLOCKSHIFT; |
dfbc8630 | 643 | set_nlink(ip, (uint32_t)links); |
2c6abf15 NB |
644 | zfs_uid_write(ip, z_uid); |
645 | zfs_gid_write(ip, z_gid); | |
7bb1325f | 646 | zfs_set_inode_flags(zp, ip); |
7f89ae6b | 647 | |
98701490 CC |
648 | /* Cache the xattr parent id */ |
649 | if (zp->z_pflags & ZFS_XATTR) | |
650 | zp->z_xattr_parent = parent; | |
651 | ||
9f5f0019 NB |
652 | ZFS_TIME_DECODE(&ip->i_atime, atime); |
653 | ZFS_TIME_DECODE(&ip->i_mtime, mtime); | |
654 | ZFS_TIME_DECODE(&ip->i_ctime, ctime); | |
655 | ||
3558fd73 | 656 | ip->i_ino = obj; |
9f5f0019 | 657 | zfs_inode_update(zp); |
0037b49e | 658 | zfs_inode_set_ops(zfsvfs, ip); |
3558fd73 | 659 | |
7b3e34ba BB |
660 | /* |
661 | * The only way insert_inode_locked() can fail is if the ip->i_ino | |
662 | * number is already hashed for this super block. This can never | |
663 | * happen because the inode numbers map 1:1 with the object numbers. | |
664 | * | |
665 | * The one exception is rolling back a mounted file system, but in | |
666 | * this case all the active inode are unhashed during the rollback. | |
667 | */ | |
668 | VERIFY3S(insert_inode_locked(ip), ==, 0); | |
c85b224f | 669 | |
0037b49e BB |
670 | mutex_enter(&zfsvfs->z_znodes_lock); |
671 | list_insert_tail(&zfsvfs->z_all_znodes, zp); | |
672 | zfsvfs->z_nr_znodes++; | |
b128c09f | 673 | membar_producer(); |
0037b49e | 674 | mutex_exit(&zfsvfs->z_znodes_lock); |
b128c09f | 675 | |
3558fd73 | 676 | unlock_new_inode(ip); |
34dc7c2f | 677 | return (zp); |
3558fd73 BB |
678 | |
679 | error: | |
3558fd73 | 680 | iput(ip); |
d1d7e268 | 681 | return (NULL); |
34dc7c2f BB |
682 | } |
683 | ||
1e8db771 BB |
684 | /* |
685 | * Safely mark an inode dirty. Inodes which are part of a read-only | |
686 | * file system or snapshot may not be dirtied. | |
687 | */ | |
688 | void | |
689 | zfs_mark_inode_dirty(struct inode *ip) | |
690 | { | |
0037b49e | 691 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
1e8db771 | 692 | |
0037b49e | 693 | if (zfs_is_readonly(zfsvfs) || dmu_objset_is_snapshot(zfsvfs->z_os)) |
1e8db771 BB |
694 | return; |
695 | ||
696 | mark_inode_dirty(ip); | |
697 | } | |
698 | ||
428870ff BB |
699 | static uint64_t empty_xattr; |
700 | static uint64_t pad[4]; | |
701 | static zfs_acl_phys_t acl_phys; | |
34dc7c2f BB |
702 | /* |
703 | * Create a new DMU object to hold a zfs znode. | |
704 | * | |
705 | * IN: dzp - parent directory for new znode | |
706 | * vap - file attributes for new znode | |
707 | * tx - dmu transaction id for zap operations | |
708 | * cr - credentials of caller | |
709 | * flag - flags: | |
710 | * IS_ROOT_NODE - new object will be root | |
711 | * IS_XATTR - new object is an attribute | |
34dc7c2f BB |
712 | * bonuslen - length of bonus buffer |
713 | * setaclp - File/Dir initial ACL | |
714 | * fuidp - Tracks fuid allocation. | |
715 | * | |
716 | * OUT: zpp - allocated znode | |
717 | * | |
718 | */ | |
719 | void | |
720 | zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, | |
428870ff | 721 | uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) |
34dc7c2f | 722 | { |
428870ff BB |
723 | uint64_t crtime[2], atime[2], mtime[2], ctime[2]; |
724 | uint64_t mode, size, links, parent, pflags; | |
9c5167d1 | 725 | uint64_t projid = ZFS_DEFAULT_PROJID; |
428870ff | 726 | uint64_t rdev = 0; |
0037b49e | 727 | zfsvfs_t *zfsvfs = ZTOZSB(dzp); |
428870ff | 728 | dmu_buf_t *db; |
6413c95f | 729 | inode_timespec_t now; |
34dc7c2f | 730 | uint64_t gen, obj; |
428870ff | 731 | int bonuslen; |
50c957f7 | 732 | int dnodesize; |
428870ff BB |
733 | sa_handle_t *sa_hdl; |
734 | dmu_object_type_t obj_type; | |
f30484af | 735 | sa_bulk_attr_t *sa_attrs; |
428870ff BB |
736 | int cnt = 0; |
737 | zfs_acl_locator_cb_t locate = { 0 }; | |
c96c36fa | 738 | znode_hold_t *zh; |
34dc7c2f | 739 | |
0037b49e | 740 | if (zfsvfs->z_replay) { |
34dc7c2f | 741 | obj = vap->va_nodeid; |
34dc7c2f BB |
742 | now = vap->va_ctime; /* see zfs_replay_create() */ |
743 | gen = vap->va_nblocks; /* ditto */ | |
50c957f7 | 744 | dnodesize = vap->va_fsid; /* ditto */ |
34dc7c2f BB |
745 | } else { |
746 | obj = 0; | |
747 | gethrestime(&now); | |
748 | gen = dmu_tx_get_txg(tx); | |
0037b49e | 749 | dnodesize = dmu_objset_dnodesize(zfsvfs->z_os); |
34dc7c2f BB |
750 | } |
751 | ||
50c957f7 NB |
752 | if (dnodesize == 0) |
753 | dnodesize = DNODE_MIN_SIZE; | |
754 | ||
0037b49e | 755 | obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; |
50c957f7 | 756 | |
428870ff | 757 | bonuslen = (obj_type == DMU_OT_SA) ? |
50c957f7 | 758 | DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE; |
428870ff | 759 | |
34dc7c2f BB |
760 | /* |
761 | * Create a new DMU object. | |
762 | */ | |
763 | /* | |
764 | * There's currently no mechanism for pre-reading the blocks that will | |
572e2857 | 765 | * be needed to allocate a new object, so we accept the small chance |
34dc7c2f BB |
766 | * that there will be an i/o error and we will fail one of the |
767 | * assertions below. | |
768 | */ | |
3558fd73 | 769 | if (S_ISDIR(vap->va_mode)) { |
0037b49e BB |
770 | if (zfsvfs->z_replay) { |
771 | VERIFY0(zap_create_claim_norm_dnsize(zfsvfs->z_os, obj, | |
772 | zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
50c957f7 | 773 | obj_type, bonuslen, dnodesize, tx)); |
34dc7c2f | 774 | } else { |
0037b49e BB |
775 | obj = zap_create_norm_dnsize(zfsvfs->z_os, |
776 | zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
50c957f7 | 777 | obj_type, bonuslen, dnodesize, tx); |
34dc7c2f BB |
778 | } |
779 | } else { | |
0037b49e BB |
780 | if (zfsvfs->z_replay) { |
781 | VERIFY0(dmu_object_claim_dnsize(zfsvfs->z_os, obj, | |
34dc7c2f | 782 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
50c957f7 | 783 | obj_type, bonuslen, dnodesize, tx)); |
34dc7c2f | 784 | } else { |
0037b49e | 785 | obj = dmu_object_alloc_dnsize(zfsvfs->z_os, |
34dc7c2f | 786 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
50c957f7 | 787 | obj_type, bonuslen, dnodesize, tx); |
34dc7c2f BB |
788 | } |
789 | } | |
34dc7c2f | 790 | |
0037b49e | 791 | zh = zfs_znode_hold_enter(zfsvfs, obj); |
9631681b | 792 | VERIFY0(sa_buf_hold(zfsvfs->z_os, obj, NULL, &db)); |
34dc7c2f BB |
793 | |
794 | /* | |
795 | * If this is the root, fix up the half-initialized parent pointer | |
796 | * to reference the just-allocated physical data area. | |
797 | */ | |
798 | if (flag & IS_ROOT_NODE) { | |
34dc7c2f BB |
799 | dzp->z_id = obj; |
800 | } | |
801 | ||
802 | /* | |
803 | * If parent is an xattr, so am I. | |
804 | */ | |
9c5167d1 | 805 | if (dzp->z_pflags & ZFS_XATTR) { |
34dc7c2f | 806 | flag |= IS_XATTR; |
34dc7c2f BB |
807 | } |
808 | ||
0037b49e | 809 | if (zfsvfs->z_use_fuids) |
428870ff BB |
810 | pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; |
811 | else | |
812 | pflags = 0; | |
34dc7c2f | 813 | |
3558fd73 | 814 | if (S_ISDIR(vap->va_mode)) { |
428870ff | 815 | size = 2; /* contents ("." and "..") */ |
dfbc8630 | 816 | links = 2; |
428870ff | 817 | } else { |
dfbc8630 | 818 | size = 0; |
ace1eae8 | 819 | links = (flag & IS_TMPFILE) ? 0 : 1; |
34dc7c2f BB |
820 | } |
821 | ||
aa6d8c10 | 822 | if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode)) |
dc1d7665 | 823 | rdev = vap->va_rdev; |
428870ff BB |
824 | |
825 | parent = dzp->z_id; | |
826 | mode = acl_ids->z_mode; | |
34dc7c2f | 827 | if (flag & IS_XATTR) |
428870ff | 828 | pflags |= ZFS_XATTR; |
34dc7c2f | 829 | |
9c5167d1 NF |
830 | if (S_ISREG(vap->va_mode) || S_ISDIR(vap->va_mode)) { |
831 | /* | |
832 | * With ZFS_PROJID flag, we can easily know whether there is | |
833 | * project ID stored on disk or not. See zfs_space_delta_cb(). | |
834 | */ | |
835 | if (obj_type != DMU_OT_ZNODE && | |
836 | dmu_objset_projectquota_enabled(zfsvfs->z_os)) | |
837 | pflags |= ZFS_PROJID; | |
838 | ||
839 | /* | |
840 | * Inherit project ID from parent if required. | |
841 | */ | |
842 | projid = zfs_inherit_projid(dzp); | |
843 | if (dzp->z_pflags & ZFS_PROJINHERIT) | |
844 | pflags |= ZFS_PROJINHERIT; | |
845 | } | |
846 | ||
428870ff BB |
847 | /* |
848 | * No execs denied will be deterimed when zfs_mode_compute() is called. | |
849 | */ | |
850 | pflags |= acl_ids->z_aclp->z_hints & | |
851 | (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| | |
852 | ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); | |
34dc7c2f | 853 | |
428870ff BB |
854 | ZFS_TIME_ENCODE(&now, crtime); |
855 | ZFS_TIME_ENCODE(&now, ctime); | |
34dc7c2f | 856 | |
3558fd73 | 857 | if (vap->va_mask & ATTR_ATIME) { |
428870ff | 858 | ZFS_TIME_ENCODE(&vap->va_atime, atime); |
34dc7c2f | 859 | } else { |
428870ff | 860 | ZFS_TIME_ENCODE(&now, atime); |
34dc7c2f BB |
861 | } |
862 | ||
3558fd73 | 863 | if (vap->va_mask & ATTR_MTIME) { |
428870ff BB |
864 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
865 | } else { | |
866 | ZFS_TIME_ENCODE(&now, mtime); | |
867 | } | |
868 | ||
869 | /* Now add in all of the "SA" attributes */ | |
0037b49e | 870 | VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, |
428870ff BB |
871 | &sa_hdl)); |
872 | ||
873 | /* | |
874 | * Setup the array of attributes to be replaced/set on the new file | |
875 | * | |
876 | * order for DMU_OT_ZNODE is critical since it needs to be constructed | |
877 | * in the old znode_phys_t format. Don't change this ordering | |
878 | */ | |
79c76d5b | 879 | sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); |
428870ff BB |
880 | |
881 | if (obj_type == DMU_OT_ZNODE) { | |
0037b49e | 882 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), |
428870ff | 883 | NULL, &atime, 16); |
0037b49e | 884 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), |
428870ff | 885 | NULL, &mtime, 16); |
0037b49e | 886 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), |
428870ff | 887 | NULL, &ctime, 16); |
0037b49e | 888 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), |
428870ff | 889 | NULL, &crtime, 16); |
0037b49e | 890 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), |
428870ff | 891 | NULL, &gen, 8); |
0037b49e | 892 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), |
428870ff | 893 | NULL, &mode, 8); |
0037b49e | 894 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), |
428870ff | 895 | NULL, &size, 8); |
0037b49e | 896 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), |
428870ff | 897 | NULL, &parent, 8); |
34dc7c2f | 898 | } else { |
0037b49e | 899 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), |
428870ff | 900 | NULL, &mode, 8); |
0037b49e | 901 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), |
428870ff | 902 | NULL, &size, 8); |
0037b49e | 903 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), |
428870ff | 904 | NULL, &gen, 8); |
0037b49e | 905 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), |
3558fd73 | 906 | NULL, &acl_ids->z_fuid, 8); |
0037b49e | 907 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), |
3558fd73 | 908 | NULL, &acl_ids->z_fgid, 8); |
0037b49e | 909 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), |
428870ff | 910 | NULL, &parent, 8); |
0037b49e | 911 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 912 | NULL, &pflags, 8); |
0037b49e | 913 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), |
428870ff | 914 | NULL, &atime, 16); |
0037b49e | 915 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), |
428870ff | 916 | NULL, &mtime, 16); |
0037b49e | 917 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), |
428870ff | 918 | NULL, &ctime, 16); |
0037b49e | 919 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), |
428870ff BB |
920 | NULL, &crtime, 16); |
921 | } | |
922 | ||
0037b49e | 923 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); |
428870ff BB |
924 | |
925 | if (obj_type == DMU_OT_ZNODE) { | |
0037b49e | 926 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, |
428870ff | 927 | &empty_xattr, 8); |
9c5167d1 NF |
928 | } else if (dmu_objset_projectquota_enabled(zfsvfs->z_os) && |
929 | pflags & ZFS_PROJID) { | |
930 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PROJID(zfsvfs), | |
931 | NULL, &projid, 8); | |
34dc7c2f | 932 | } |
428870ff | 933 | if (obj_type == DMU_OT_ZNODE || |
aa6d8c10 | 934 | (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) { |
0037b49e | 935 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs), |
428870ff | 936 | NULL, &rdev, 8); |
428870ff BB |
937 | } |
938 | if (obj_type == DMU_OT_ZNODE) { | |
0037b49e | 939 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), |
428870ff | 940 | NULL, &pflags, 8); |
0037b49e | 941 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, |
428870ff | 942 | &acl_ids->z_fuid, 8); |
0037b49e | 943 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, |
428870ff | 944 | &acl_ids->z_fgid, 8); |
0037b49e | 945 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad, |
428870ff | 946 | sizeof (uint64_t) * 4); |
0037b49e | 947 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, |
428870ff BB |
948 | &acl_phys, sizeof (zfs_acl_phys_t)); |
949 | } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { | |
0037b49e | 950 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL, |
428870ff BB |
951 | &acl_ids->z_aclp->z_acl_count, 8); |
952 | locate.cb_aclp = acl_ids->z_aclp; | |
0037b49e | 953 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs), |
428870ff BB |
954 | zfs_acl_data_locator, &locate, |
955 | acl_ids->z_aclp->z_acl_bytes); | |
956 | mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, | |
957 | acl_ids->z_fuid, acl_ids->z_fgid); | |
958 | } | |
959 | ||
960 | VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); | |
34dc7c2f | 961 | |
34dc7c2f | 962 | if (!(flag & IS_ROOT_NODE)) { |
8d703987 BB |
963 | /* |
964 | * The call to zfs_znode_alloc() may fail if memory is low | |
965 | * via the call path: alloc_inode() -> inode_init_always() -> | |
966 | * security_inode_alloc() -> inode_alloc_security(). Since | |
967 | * the existing code is written such that zfs_mknode() can | |
968 | * not fail retry until sufficient memory has been reclaimed. | |
969 | */ | |
970 | do { | |
971 | *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, obj, | |
972 | sa_hdl); | |
973 | } while (*zpp == NULL); | |
974 | ||
7b3e34ba BB |
975 | VERIFY(*zpp != NULL); |
976 | VERIFY(dzp != NULL); | |
34dc7c2f BB |
977 | } else { |
978 | /* | |
979 | * If we are creating the root node, the "parent" we | |
980 | * passed in is the znode for the root. | |
981 | */ | |
982 | *zpp = dzp; | |
428870ff BB |
983 | |
984 | (*zpp)->z_sa_hdl = sa_hdl; | |
34dc7c2f | 985 | } |
428870ff BB |
986 | |
987 | (*zpp)->z_pflags = pflags; | |
12fa7f34 | 988 | (*zpp)->z_mode = ZTOI(*zpp)->i_mode = mode; |
50c957f7 | 989 | (*zpp)->z_dnodesize = dnodesize; |
9c5167d1 | 990 | (*zpp)->z_projid = projid; |
428870ff | 991 | |
428870ff BB |
992 | if (obj_type == DMU_OT_ZNODE || |
993 | acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { | |
b0bc7a84 | 994 | VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); |
428870ff | 995 | } |
d1d7e268 | 996 | kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); |
0037b49e | 997 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
998 | } |
999 | ||
5484965a | 1000 | /* |
d3cc8b15 WA |
1001 | * Update in-core attributes. It is assumed the caller will be doing an |
1002 | * sa_bulk_update to push the changes out. | |
5484965a BB |
1003 | */ |
1004 | void | |
1005 | zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) | |
1006 | { | |
1007 | xoptattr_t *xoap; | |
7bb1325f | 1008 | boolean_t update_inode = B_FALSE; |
5484965a BB |
1009 | |
1010 | xoap = xva_getxoptattr(xvap); | |
1011 | ASSERT(xoap); | |
1012 | ||
1013 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { | |
1014 | uint64_t times[2]; | |
1015 | ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); | |
1016 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)), | |
1017 | ×, sizeof (times), tx); | |
1018 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
1019 | } | |
1020 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
1021 | ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, | |
1022 | zp->z_pflags, tx); | |
1023 | XVA_SET_RTN(xvap, XAT_READONLY); | |
1024 | } | |
1025 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
1026 | ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, | |
1027 | zp->z_pflags, tx); | |
1028 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
1029 | } | |
1030 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
1031 | ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, | |
1032 | zp->z_pflags, tx); | |
1033 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
1034 | } | |
1035 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
1036 | ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, | |
1037 | zp->z_pflags, tx); | |
1038 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
1039 | } | |
1040 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
1041 | ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, | |
1042 | zp->z_pflags, tx); | |
1043 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
64c688d7 | 1044 | |
7bb1325f | 1045 | update_inode = B_TRUE; |
5484965a BB |
1046 | } |
1047 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
1048 | ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, | |
1049 | zp->z_pflags, tx); | |
1050 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
1051 | } | |
1052 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
1053 | ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, | |
1054 | zp->z_pflags, tx); | |
1055 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
64c688d7 | 1056 | |
7bb1325f | 1057 | update_inode = B_TRUE; |
5484965a BB |
1058 | } |
1059 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
1060 | ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, | |
1061 | zp->z_pflags, tx); | |
1062 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
1063 | } | |
1064 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
1065 | ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, | |
1066 | zp->z_pflags, tx); | |
1067 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
1068 | } | |
1069 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
1070 | ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, | |
1071 | xoap->xoa_av_quarantined, zp->z_pflags, tx); | |
1072 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
1073 | } | |
1074 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
1075 | ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, | |
1076 | zp->z_pflags, tx); | |
1077 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
1078 | } | |
1079 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { | |
1080 | zfs_sa_set_scanstamp(zp, xvap, tx); | |
1081 | XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); | |
1082 | } | |
1083 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
1084 | ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, | |
1085 | zp->z_pflags, tx); | |
1086 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
1087 | } | |
1088 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
1089 | ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, | |
1090 | zp->z_pflags, tx); | |
1091 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
1092 | } | |
1093 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
1094 | ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, | |
1095 | zp->z_pflags, tx); | |
1096 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
1097 | } | |
9c5167d1 NF |
1098 | if (XVA_ISSET_REQ(xvap, XAT_PROJINHERIT)) { |
1099 | ZFS_ATTR_SET(zp, ZFS_PROJINHERIT, xoap->xoa_projinherit, | |
1100 | zp->z_pflags, tx); | |
1101 | XVA_SET_RTN(xvap, XAT_PROJINHERIT); | |
1102 | } | |
7bb1325f CC |
1103 | |
1104 | if (update_inode) | |
1105 | zfs_set_inode_flags(zp, ZTOI(zp)); | |
5484965a BB |
1106 | } |
1107 | ||
34dc7c2f | 1108 | int |
0037b49e | 1109 | zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) |
34dc7c2f BB |
1110 | { |
1111 | dmu_object_info_t doi; | |
1112 | dmu_buf_t *db; | |
1113 | znode_t *zp; | |
c96c36fa | 1114 | znode_hold_t *zh; |
34dc7c2f | 1115 | int err; |
428870ff | 1116 | sa_handle_t *hdl; |
34dc7c2f BB |
1117 | |
1118 | *zpp = NULL; | |
1119 | ||
6f9548c4 | 1120 | again: |
0037b49e | 1121 | zh = zfs_znode_hold_enter(zfsvfs, obj_num); |
34dc7c2f | 1122 | |
0037b49e | 1123 | err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); |
34dc7c2f | 1124 | if (err) { |
0037b49e | 1125 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1126 | return (err); |
1127 | } | |
1128 | ||
1129 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1130 | if (doi.doi_bonus_type != DMU_OT_SA && |
1131 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1132 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1133 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1134 | sa_buf_rele(db, NULL); | |
0037b49e | 1135 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1136 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1137 | } |
1138 | ||
428870ff BB |
1139 | hdl = dmu_buf_get_user(db); |
1140 | if (hdl != NULL) { | |
36df2843 | 1141 | zp = sa_get_userdata(hdl); |
34dc7c2f | 1142 | |
8ac67298 | 1143 | |
34dc7c2f | 1144 | /* |
428870ff BB |
1145 | * Since "SA" does immediate eviction we |
1146 | * should never find a sa handle that doesn't | |
1147 | * know about the znode. | |
34dc7c2f | 1148 | */ |
428870ff BB |
1149 | |
1150 | ASSERT3P(zp, !=, NULL); | |
1151 | ||
1152 | mutex_enter(&zp->z_lock); | |
34dc7c2f | 1153 | ASSERT3U(zp->z_id, ==, obj_num); |
98701490 CC |
1154 | /* |
1155 | * If igrab() returns NULL the VFS has independently | |
1156 | * determined the inode should be evicted and has | |
1157 | * called iput_final() to start the eviction process. | |
1158 | * The SA handle is still valid but because the VFS | |
1159 | * requires that the eviction succeed we must drop | |
1160 | * our locks and references to allow the eviction to | |
1161 | * complete. The zfs_zget() may then be retried. | |
1162 | * | |
1163 | * This unlikely case could be optimized by registering | |
1164 | * a sops->drop_inode() callback. The callback would | |
1165 | * need to detect the active SA hold thereby informing | |
1166 | * the VFS that this inode should not be evicted. | |
1167 | */ | |
1168 | if (igrab(ZTOI(zp)) == NULL) { | |
1169 | mutex_exit(&zp->z_lock); | |
1170 | sa_buf_rele(db, NULL); | |
0037b49e | 1171 | zfs_znode_hold_exit(zfsvfs, zh); |
98701490 CC |
1172 | /* inode might need this to finish evict */ |
1173 | cond_resched(); | |
1174 | goto again; | |
34dc7c2f | 1175 | } |
98701490 CC |
1176 | *zpp = zp; |
1177 | err = 0; | |
34dc7c2f | 1178 | mutex_exit(&zp->z_lock); |
f3ad9cd6 | 1179 | sa_buf_rele(db, NULL); |
0037b49e | 1180 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1181 | return (err); |
1182 | } | |
1183 | ||
1184 | /* | |
3558fd73 | 1185 | * Not found create new znode/vnode but only if file exists. |
428870ff BB |
1186 | * |
1187 | * There is a small window where zfs_vget() could | |
1188 | * find this object while a file create is still in | |
1189 | * progress. This is checked for in zfs_znode_alloc() | |
1190 | * | |
1191 | * if zfs_znode_alloc() fails it will drop the hold on the | |
1192 | * bonus buffer. | |
34dc7c2f | 1193 | */ |
0037b49e | 1194 | zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size, |
31b6111f | 1195 | doi.doi_bonus_type, obj_num, NULL); |
428870ff | 1196 | if (zp == NULL) { |
2e528b49 | 1197 | err = SET_ERROR(ENOENT); |
428870ff BB |
1198 | } else { |
1199 | *zpp = zp; | |
1200 | } | |
0037b49e | 1201 | zfs_znode_hold_exit(zfsvfs, zh); |
428870ff | 1202 | return (err); |
34dc7c2f BB |
1203 | } |
1204 | ||
1205 | int | |
1206 | zfs_rezget(znode_t *zp) | |
1207 | { | |
0037b49e | 1208 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f BB |
1209 | dmu_object_info_t doi; |
1210 | dmu_buf_t *db; | |
1211 | uint64_t obj_num = zp->z_id; | |
428870ff | 1212 | uint64_t mode; |
dfbc8630 | 1213 | uint64_t links; |
9f5f0019 | 1214 | sa_bulk_attr_t bulk[10]; |
34dc7c2f | 1215 | int err; |
428870ff BB |
1216 | int count = 0; |
1217 | uint64_t gen; | |
2c6abf15 | 1218 | uint64_t z_uid, z_gid; |
9f5f0019 | 1219 | uint64_t atime[2], mtime[2], ctime[2]; |
9c5167d1 | 1220 | uint64_t projid = ZFS_DEFAULT_PROJID; |
c96c36fa | 1221 | znode_hold_t *zh; |
34dc7c2f | 1222 | |
cbecb4fb CC |
1223 | /* |
1224 | * skip ctldir, otherwise they will always get invalidated. This will | |
1225 | * cause funny behaviour for the mounted snapdirs. Especially for | |
1226 | * Linux >= 3.18, d_invalidate will detach the mountpoint and prevent | |
1227 | * anyone automount it again as long as someone is still using the | |
1228 | * detached mount. | |
1229 | */ | |
1230 | if (zp->z_is_ctldir) | |
1231 | return (0); | |
1232 | ||
0037b49e | 1233 | zh = zfs_znode_hold_enter(zfsvfs, obj_num); |
34dc7c2f | 1234 | |
428870ff BB |
1235 | mutex_enter(&zp->z_acl_lock); |
1236 | if (zp->z_acl_cached) { | |
1237 | zfs_acl_free(zp->z_acl_cached); | |
1238 | zp->z_acl_cached = NULL; | |
1239 | } | |
428870ff | 1240 | mutex_exit(&zp->z_acl_lock); |
7b3e34ba | 1241 | |
228b461b | 1242 | rw_enter(&zp->z_xattr_lock, RW_WRITER); |
7b3e34ba BB |
1243 | if (zp->z_xattr_cached) { |
1244 | nvlist_free(zp->z_xattr_cached); | |
1245 | zp->z_xattr_cached = NULL; | |
1246 | } | |
7b3e34ba BB |
1247 | rw_exit(&zp->z_xattr_lock); |
1248 | ||
428870ff | 1249 | ASSERT(zp->z_sa_hdl == NULL); |
0037b49e | 1250 | err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); |
34dc7c2f | 1251 | if (err) { |
0037b49e | 1252 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1253 | return (err); |
1254 | } | |
1255 | ||
1256 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1257 | if (doi.doi_bonus_type != DMU_OT_SA && |
1258 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1259 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1260 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1261 | sa_buf_rele(db, NULL); | |
0037b49e | 1262 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1263 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1264 | } |
1265 | ||
0037b49e | 1266 | zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL); |
428870ff BB |
1267 | |
1268 | /* reload cached values */ | |
0037b49e | 1269 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, |
428870ff | 1270 | &gen, sizeof (gen)); |
0037b49e | 1271 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, |
428870ff | 1272 | &zp->z_size, sizeof (zp->z_size)); |
0037b49e | 1273 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, |
dfbc8630 | 1274 | &links, sizeof (links)); |
0037b49e | 1275 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, |
428870ff | 1276 | &zp->z_pflags, sizeof (zp->z_pflags)); |
0037b49e | 1277 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, |
2c6abf15 | 1278 | &z_uid, sizeof (z_uid)); |
0037b49e | 1279 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, |
2c6abf15 | 1280 | &z_gid, sizeof (z_gid)); |
0037b49e | 1281 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, |
428870ff | 1282 | &mode, sizeof (mode)); |
0037b49e | 1283 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, |
9f5f0019 | 1284 | &atime, 16); |
0037b49e | 1285 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, |
9f5f0019 | 1286 | &mtime, 16); |
0037b49e | 1287 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, |
9f5f0019 | 1288 | &ctime, 16); |
428870ff | 1289 | |
428870ff BB |
1290 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { |
1291 | zfs_znode_dmu_fini(zp); | |
0037b49e | 1292 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1293 | return (SET_ERROR(EIO)); |
428870ff BB |
1294 | } |
1295 | ||
9c5167d1 NF |
1296 | if (dmu_objset_projectquota_enabled(zfsvfs->z_os)) { |
1297 | err = sa_lookup(zp->z_sa_hdl, SA_ZPL_PROJID(zfsvfs), | |
1298 | &projid, 8); | |
1299 | if (err != 0 && err != ENOENT) { | |
1300 | zfs_znode_dmu_fini(zp); | |
1301 | zfs_znode_hold_exit(zfsvfs, zh); | |
1302 | return (SET_ERROR(err)); | |
1303 | } | |
1304 | } | |
1305 | ||
1306 | zp->z_projid = projid; | |
12fa7f34 | 1307 | zp->z_mode = ZTOI(zp)->i_mode = mode; |
2c6abf15 NB |
1308 | zfs_uid_write(ZTOI(zp), z_uid); |
1309 | zfs_gid_write(ZTOI(zp), z_gid); | |
572e2857 | 1310 | |
9f5f0019 NB |
1311 | ZFS_TIME_DECODE(&ZTOI(zp)->i_atime, atime); |
1312 | ZFS_TIME_DECODE(&ZTOI(zp)->i_mtime, mtime); | |
1313 | ZFS_TIME_DECODE(&ZTOI(zp)->i_ctime, ctime); | |
1314 | ||
278f2236 | 1315 | if (gen != ZTOI(zp)->i_generation) { |
428870ff | 1316 | zfs_znode_dmu_fini(zp); |
0037b49e | 1317 | zfs_znode_hold_exit(zfsvfs, zh); |
2e528b49 | 1318 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1319 | } |
1320 | ||
dfbc8630 | 1321 | set_nlink(ZTOI(zp), (uint32_t)links); |
7bb1325f | 1322 | zfs_set_inode_flags(zp, ZTOI(zp)); |
dfbc8630 | 1323 | |
34dc7c2f | 1324 | zp->z_blksz = doi.doi_data_block_size; |
704cd075 | 1325 | zp->z_atime_dirty = 0; |
9f5f0019 | 1326 | zfs_inode_update(zp); |
34dc7c2f | 1327 | |
6a218566 AG |
1328 | /* |
1329 | * If the file has zero links, then it has been unlinked on the send | |
1330 | * side and it must be in the received unlinked set. | |
1331 | * We call zfs_znode_dmu_fini() now to prevent any accesses to the | |
1332 | * stale data and to prevent automatical removal of the file in | |
1333 | * zfs_zinactive(). The file will be removed either when it is removed | |
1334 | * on the send side and the next incremental stream is received or | |
1335 | * when the unlinked set gets processed. | |
1336 | */ | |
1337 | zp->z_unlinked = (ZTOI(zp)->i_nlink == 0); | |
1338 | if (zp->z_unlinked) | |
1339 | zfs_znode_dmu_fini(zp); | |
1340 | ||
0037b49e | 1341 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1342 | |
1343 | return (0); | |
1344 | } | |
1345 | ||
1346 | void | |
1347 | zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) | |
1348 | { | |
0037b49e BB |
1349 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
1350 | objset_t *os = zfsvfs->z_os; | |
34dc7c2f | 1351 | uint64_t obj = zp->z_id; |
572e2857 | 1352 | uint64_t acl_obj = zfs_external_acl(zp); |
c96c36fa | 1353 | znode_hold_t *zh; |
34dc7c2f | 1354 | |
0037b49e | 1355 | zh = zfs_znode_hold_enter(zfsvfs, obj); |
572e2857 BB |
1356 | if (acl_obj) { |
1357 | VERIFY(!zp->z_is_sa); | |
b128c09f | 1358 | VERIFY(0 == dmu_object_free(os, acl_obj, tx)); |
572e2857 | 1359 | } |
b128c09f | 1360 | VERIFY(0 == dmu_object_free(os, obj, tx)); |
34dc7c2f | 1361 | zfs_znode_dmu_fini(zp); |
0037b49e | 1362 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1363 | } |
1364 | ||
1365 | void | |
1366 | zfs_zinactive(znode_t *zp) | |
1367 | { | |
0037b49e | 1368 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
34dc7c2f | 1369 | uint64_t z_id = zp->z_id; |
c96c36fa | 1370 | znode_hold_t *zh; |
34dc7c2f | 1371 | |
428870ff | 1372 | ASSERT(zp->z_sa_hdl); |
34dc7c2f BB |
1373 | |
1374 | /* | |
d6bd8eaa | 1375 | * Don't allow a zfs_zget() while were trying to release this znode. |
34dc7c2f | 1376 | */ |
0037b49e | 1377 | zh = zfs_znode_hold_enter(zfsvfs, z_id); |
d6bd8eaa | 1378 | |
34dc7c2f | 1379 | mutex_enter(&zp->z_lock); |
34dc7c2f BB |
1380 | |
1381 | /* | |
6a218566 AG |
1382 | * If this was the last reference to a file with no links, remove |
1383 | * the file from the file system unless the file system is mounted | |
1384 | * read-only. That can happen, for example, if the file system was | |
1385 | * originally read-write, the file was opened, then unlinked and | |
1386 | * the file system was made read-only before the file was finally | |
1387 | * closed. The file will remain in the unlinked set. | |
34dc7c2f BB |
1388 | */ |
1389 | if (zp->z_unlinked) { | |
6a218566 AG |
1390 | ASSERT(!zfsvfs->z_issnap); |
1391 | if (!zfs_is_readonly(zfsvfs)) { | |
1392 | mutex_exit(&zp->z_lock); | |
1393 | zfs_znode_hold_exit(zfsvfs, zh); | |
1394 | zfs_rmnode(zp); | |
1395 | return; | |
1396 | } | |
34dc7c2f | 1397 | } |
428870ff | 1398 | |
34dc7c2f BB |
1399 | mutex_exit(&zp->z_lock); |
1400 | zfs_znode_dmu_fini(zp); | |
d6bd8eaa | 1401 | |
0037b49e | 1402 | zfs_znode_hold_exit(zfsvfs, zh); |
34dc7c2f BB |
1403 | } |
1404 | ||
6d111134 TC |
1405 | static inline int |
1406 | zfs_compare_timespec(struct timespec *t1, struct timespec *t2) | |
1407 | { | |
1408 | if (t1->tv_sec < t2->tv_sec) | |
1409 | return (-1); | |
1410 | ||
1411 | if (t1->tv_sec > t2->tv_sec) | |
1412 | return (1); | |
1413 | ||
1414 | return (t1->tv_nsec - t2->tv_nsec); | |
1415 | } | |
1416 | ||
6d111134 TC |
1417 | /* |
1418 | * Prepare to update znode time stamps. | |
1419 | * | |
1420 | * IN: zp - znode requiring timestamp update | |
0df9673f | 1421 | * flag - ATTR_MTIME, ATTR_CTIME flags |
6d111134 | 1422 | * |
0df9673f | 1423 | * OUT: zp - z_seq |
6d111134 TC |
1424 | * mtime - new mtime |
1425 | * ctime - new ctime | |
1426 | * | |
0df9673f CC |
1427 | * Note: We don't update atime here, because we rely on Linux VFS to do |
1428 | * atime updating. | |
6d111134 | 1429 | */ |
34dc7c2f | 1430 | void |
428870ff | 1431 | zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], |
0df9673f | 1432 | uint64_t ctime[2]) |
34dc7c2f | 1433 | { |
6413c95f | 1434 | inode_timespec_t now; |
34dc7c2f | 1435 | |
34dc7c2f BB |
1436 | gethrestime(&now); |
1437 | ||
0df9673f | 1438 | zp->z_seq++; |
34dc7c2f | 1439 | |
3558fd73 | 1440 | if (flag & ATTR_MTIME) { |
428870ff | 1441 | ZFS_TIME_ENCODE(&now, mtime); |
9f5f0019 | 1442 | ZFS_TIME_DECODE(&(ZTOI(zp)->i_mtime), mtime); |
3558fd73 | 1443 | if (ZTOZSB(zp)->z_use_fuids) { |
428870ff BB |
1444 | zp->z_pflags |= (ZFS_ARCHIVE | |
1445 | ZFS_AV_MODIFIED); | |
1446 | } | |
34dc7c2f BB |
1447 | } |
1448 | ||
3558fd73 | 1449 | if (flag & ATTR_CTIME) { |
428870ff | 1450 | ZFS_TIME_ENCODE(&now, ctime); |
9f5f0019 | 1451 | ZFS_TIME_DECODE(&(ZTOI(zp)->i_ctime), ctime); |
3558fd73 | 1452 | if (ZTOZSB(zp)->z_use_fuids) |
428870ff | 1453 | zp->z_pflags |= ZFS_ARCHIVE; |
34dc7c2f BB |
1454 | } |
1455 | } | |
1456 | ||
34dc7c2f BB |
1457 | /* |
1458 | * Grow the block size for a file. | |
1459 | * | |
1460 | * IN: zp - znode of file to free data in. | |
1461 | * size - requested block size | |
1462 | * tx - open transaction. | |
1463 | * | |
1464 | * NOTE: this function assumes that the znode is write locked. | |
1465 | */ | |
1466 | void | |
1467 | zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) | |
1468 | { | |
1469 | int error; | |
1470 | u_longlong_t dummy; | |
1471 | ||
1472 | if (size <= zp->z_blksz) | |
1473 | return; | |
1474 | /* | |
1475 | * If the file size is already greater than the current blocksize, | |
1476 | * we will not grow. If there is more than one block in a file, | |
1477 | * the blocksize cannot change. | |
1478 | */ | |
428870ff | 1479 | if (zp->z_blksz && zp->z_size > zp->z_blksz) |
34dc7c2f BB |
1480 | return; |
1481 | ||
3558fd73 | 1482 | error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id, |
34dc7c2f | 1483 | size, 0, tx); |
428870ff | 1484 | |
34dc7c2f BB |
1485 | if (error == ENOTSUP) |
1486 | return; | |
c99c9001 | 1487 | ASSERT0(error); |
34dc7c2f BB |
1488 | |
1489 | /* What blocksize did we actually get? */ | |
428870ff | 1490 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); |
34dc7c2f BB |
1491 | } |
1492 | ||
34dc7c2f | 1493 | /* |
b128c09f | 1494 | * Increase the file length |
34dc7c2f BB |
1495 | * |
1496 | * IN: zp - znode of file to free data in. | |
b128c09f | 1497 | * end - new end-of-file |
34dc7c2f | 1498 | * |
19d55079 | 1499 | * RETURN: 0 on success, error code on failure |
34dc7c2f | 1500 | */ |
b128c09f BB |
1501 | static int |
1502 | zfs_extend(znode_t *zp, uint64_t end) | |
34dc7c2f | 1503 | { |
0037b49e | 1504 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
b128c09f | 1505 | dmu_tx_t *tx; |
5d43cc9a | 1506 | locked_range_t *lr; |
b128c09f | 1507 | uint64_t newblksz; |
34dc7c2f BB |
1508 | int error; |
1509 | ||
34dc7c2f | 1510 | /* |
b128c09f | 1511 | * We will change zp_size, lock the whole file. |
34dc7c2f | 1512 | */ |
5d43cc9a | 1513 | lr = rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); |
34dc7c2f BB |
1514 | |
1515 | /* | |
1516 | * Nothing to do if file already at desired length. | |
1517 | */ | |
428870ff | 1518 | if (end <= zp->z_size) { |
5d43cc9a | 1519 | rangelock_exit(lr); |
34dc7c2f BB |
1520 | return (0); |
1521 | } | |
0037b49e | 1522 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1523 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1524 | zfs_sa_upgrade_txholds(tx, zp); | |
b128c09f | 1525 | if (end > zp->z_blksz && |
0037b49e | 1526 | (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { |
34dc7c2f BB |
1527 | /* |
1528 | * We are growing the file past the current block size. | |
1529 | */ | |
3558fd73 | 1530 | if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) { |
f1512ee6 MA |
1531 | /* |
1532 | * File's blocksize is already larger than the | |
1533 | * "recordsize" property. Only let it grow to | |
1534 | * the next power of 2. | |
1535 | */ | |
34dc7c2f | 1536 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 | 1537 | newblksz = MIN(end, 1 << highbit64(zp->z_blksz)); |
34dc7c2f | 1538 | } else { |
3558fd73 | 1539 | newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz); |
34dc7c2f | 1540 | } |
b128c09f BB |
1541 | dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); |
1542 | } else { | |
1543 | newblksz = 0; | |
34dc7c2f BB |
1544 | } |
1545 | ||
384f8a09 | 1546 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 1547 | if (error) { |
34dc7c2f | 1548 | dmu_tx_abort(tx); |
5d43cc9a | 1549 | rangelock_exit(lr); |
34dc7c2f BB |
1550 | return (error); |
1551 | } | |
1552 | ||
b128c09f BB |
1553 | if (newblksz) |
1554 | zfs_grow_blocksize(zp, newblksz, tx); | |
34dc7c2f | 1555 | |
428870ff BB |
1556 | zp->z_size = end; |
1557 | ||
3558fd73 | 1558 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), |
428870ff | 1559 | &zp->z_size, sizeof (zp->z_size), tx)); |
34dc7c2f | 1560 | |
5d43cc9a | 1561 | rangelock_exit(lr); |
34dc7c2f | 1562 | |
b128c09f | 1563 | dmu_tx_commit(tx); |
34dc7c2f | 1564 | |
b128c09f BB |
1565 | return (0); |
1566 | } | |
1567 | ||
223df016 TC |
1568 | /* |
1569 | * zfs_zero_partial_page - Modeled after update_pages() but | |
1570 | * with different arguments and semantics for use by zfs_freesp(). | |
1571 | * | |
1572 | * Zeroes a piece of a single page cache entry for zp at offset | |
1573 | * start and length len. | |
1574 | * | |
1575 | * Caller must acquire a range lock on the file for the region | |
1576 | * being zeroed in order that the ARC and page cache stay in sync. | |
1577 | */ | |
1578 | static void | |
1579 | zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len) | |
1580 | { | |
1581 | struct address_space *mp = ZTOI(zp)->i_mapping; | |
1582 | struct page *pp; | |
1583 | int64_t off; | |
1584 | void *pb; | |
1585 | ||
8b1899d3 | 1586 | ASSERT((start & PAGE_MASK) == ((start + len - 1) & PAGE_MASK)); |
223df016 | 1587 | |
8b1899d3 BB |
1588 | off = start & (PAGE_SIZE - 1); |
1589 | start &= PAGE_MASK; | |
223df016 | 1590 | |
8b1899d3 | 1591 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
223df016 TC |
1592 | if (pp) { |
1593 | if (mapping_writably_mapped(mp)) | |
1594 | flush_dcache_page(pp); | |
1595 | ||
1596 | pb = kmap(pp); | |
1597 | bzero(pb + off, len); | |
1598 | kunmap(pp); | |
1599 | ||
1600 | if (mapping_writably_mapped(mp)) | |
1601 | flush_dcache_page(pp); | |
1602 | ||
1603 | mark_page_accessed(pp); | |
1604 | SetPageUptodate(pp); | |
1605 | ClearPageError(pp); | |
1606 | unlock_page(pp); | |
8b1899d3 | 1607 | put_page(pp); |
223df016 TC |
1608 | } |
1609 | } | |
1610 | ||
b128c09f BB |
1611 | /* |
1612 | * Free space in a file. | |
1613 | * | |
1614 | * IN: zp - znode of file to free data in. | |
1615 | * off - start of section to free. | |
1616 | * len - length of section to free. | |
1617 | * | |
19d55079 | 1618 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1619 | */ |
1620 | static int | |
1621 | zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) | |
1622 | { | |
0037b49e | 1623 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
5d43cc9a | 1624 | locked_range_t *lr; |
b128c09f BB |
1625 | int error; |
1626 | ||
1627 | /* | |
1628 | * Lock the range being freed. | |
1629 | */ | |
5d43cc9a | 1630 | lr = rangelock_enter(&zp->z_rangelock, off, len, RL_WRITER); |
b128c09f BB |
1631 | |
1632 | /* | |
1633 | * Nothing to do if file already at desired length. | |
1634 | */ | |
428870ff | 1635 | if (off >= zp->z_size) { |
5d43cc9a | 1636 | rangelock_exit(lr); |
b128c09f | 1637 | return (0); |
34dc7c2f BB |
1638 | } |
1639 | ||
428870ff BB |
1640 | if (off + len > zp->z_size) |
1641 | len = zp->z_size - off; | |
b128c09f | 1642 | |
0037b49e | 1643 | error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); |
b128c09f | 1644 | |
223df016 TC |
1645 | /* |
1646 | * Zero partial page cache entries. This must be done under a | |
1647 | * range lock in order to keep the ARC and page cache in sync. | |
1648 | */ | |
1649 | if (zp->z_is_mapped) { | |
1650 | loff_t first_page, last_page, page_len; | |
1651 | loff_t first_page_offset, last_page_offset; | |
1652 | ||
1653 | /* first possible full page in hole */ | |
8b1899d3 | 1654 | first_page = (off + PAGE_SIZE - 1) >> PAGE_SHIFT; |
223df016 | 1655 | /* last page of hole */ |
8b1899d3 | 1656 | last_page = (off + len) >> PAGE_SHIFT; |
223df016 TC |
1657 | |
1658 | /* offset of first_page */ | |
8b1899d3 | 1659 | first_page_offset = first_page << PAGE_SHIFT; |
223df016 | 1660 | /* offset of last_page */ |
8b1899d3 | 1661 | last_page_offset = last_page << PAGE_SHIFT; |
223df016 | 1662 | |
cb08f063 TC |
1663 | /* truncate whole pages */ |
1664 | if (last_page_offset > first_page_offset) { | |
1665 | truncate_inode_pages_range(ZTOI(zp)->i_mapping, | |
1666 | first_page_offset, last_page_offset - 1); | |
1667 | } | |
1668 | ||
1669 | /* truncate sub-page ranges */ | |
223df016 TC |
1670 | if (first_page > last_page) { |
1671 | /* entire punched area within a single page */ | |
1672 | zfs_zero_partial_page(zp, off, len); | |
1673 | } else { | |
1674 | /* beginning of punched area at the end of a page */ | |
1675 | page_len = first_page_offset - off; | |
1676 | if (page_len > 0) | |
1677 | zfs_zero_partial_page(zp, off, page_len); | |
1678 | ||
1679 | /* end of punched area at the beginning of a page */ | |
1680 | page_len = off + len - last_page_offset; | |
1681 | if (page_len > 0) | |
1682 | zfs_zero_partial_page(zp, last_page_offset, | |
1683 | page_len); | |
1684 | } | |
1685 | } | |
5d43cc9a | 1686 | rangelock_exit(lr); |
34dc7c2f | 1687 | |
b128c09f BB |
1688 | return (error); |
1689 | } | |
1690 | ||
1691 | /* | |
1692 | * Truncate a file | |
1693 | * | |
1694 | * IN: zp - znode of file to free data in. | |
1695 | * end - new end-of-file. | |
1696 | * | |
19d55079 | 1697 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1698 | */ |
1699 | static int | |
1700 | zfs_trunc(znode_t *zp, uint64_t end) | |
1701 | { | |
0037b49e | 1702 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
b128c09f | 1703 | dmu_tx_t *tx; |
5d43cc9a | 1704 | locked_range_t *lr; |
b128c09f | 1705 | int error; |
572e2857 BB |
1706 | sa_bulk_attr_t bulk[2]; |
1707 | int count = 0; | |
b128c09f BB |
1708 | |
1709 | /* | |
1710 | * We will change zp_size, lock the whole file. | |
1711 | */ | |
5d43cc9a | 1712 | lr = rangelock_enter(&zp->z_rangelock, 0, UINT64_MAX, RL_WRITER); |
b128c09f BB |
1713 | |
1714 | /* | |
1715 | * Nothing to do if file already at desired length. | |
1716 | */ | |
428870ff | 1717 | if (end >= zp->z_size) { |
5d43cc9a | 1718 | rangelock_exit(lr); |
b128c09f BB |
1719 | return (0); |
1720 | } | |
1721 | ||
18a2485f FS |
1722 | error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, |
1723 | DMU_OBJECT_END); | |
b128c09f | 1724 | if (error) { |
5d43cc9a | 1725 | rangelock_exit(lr); |
b128c09f BB |
1726 | return (error); |
1727 | } | |
0037b49e | 1728 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1729 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1730 | zfs_sa_upgrade_txholds(tx, zp); | |
19d55079 | 1731 | dmu_tx_mark_netfree(tx); |
7a8f0e80 | 1732 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1733 | if (error) { |
b128c09f | 1734 | dmu_tx_abort(tx); |
5d43cc9a | 1735 | rangelock_exit(lr); |
b128c09f BB |
1736 | return (error); |
1737 | } | |
b128c09f | 1738 | |
428870ff | 1739 | zp->z_size = end; |
0037b49e | 1740 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), |
572e2857 | 1741 | NULL, &zp->z_size, sizeof (zp->z_size)); |
428870ff | 1742 | |
572e2857 BB |
1743 | if (end == 0) { |
1744 | zp->z_pflags &= ~ZFS_SPARSE; | |
0037b49e | 1745 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), |
572e2857 BB |
1746 | NULL, &zp->z_pflags, 8); |
1747 | } | |
1748 | VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); | |
b128c09f | 1749 | |
34dc7c2f | 1750 | dmu_tx_commit(tx); |
5d43cc9a | 1751 | rangelock_exit(lr); |
34dc7c2f BB |
1752 | |
1753 | return (0); | |
1754 | } | |
1755 | ||
b128c09f BB |
1756 | /* |
1757 | * Free space in a file | |
1758 | * | |
1759 | * IN: zp - znode of file to free data in. | |
1760 | * off - start of range | |
1761 | * len - end of range (0 => EOF) | |
1762 | * flag - current file open mode flags. | |
1763 | * log - TRUE if this action should be logged | |
1764 | * | |
19d55079 | 1765 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1766 | */ |
1767 | int | |
1768 | zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) | |
1769 | { | |
b128c09f | 1770 | dmu_tx_t *tx; |
0037b49e BB |
1771 | zfsvfs_t *zfsvfs = ZTOZSB(zp); |
1772 | zilog_t *zilog = zfsvfs->z_log; | |
428870ff BB |
1773 | uint64_t mode; |
1774 | uint64_t mtime[2], ctime[2]; | |
1775 | sa_bulk_attr_t bulk[3]; | |
1776 | int count = 0; | |
b128c09f BB |
1777 | int error; |
1778 | ||
0037b49e | 1779 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode, |
428870ff BB |
1780 | sizeof (mode))) != 0) |
1781 | return (error); | |
1782 | ||
1783 | if (off > zp->z_size) { | |
b128c09f BB |
1784 | error = zfs_extend(zp, off+len); |
1785 | if (error == 0 && log) | |
1786 | goto log; | |
223df016 | 1787 | goto out; |
b128c09f BB |
1788 | } |
1789 | ||
b128c09f BB |
1790 | if (len == 0) { |
1791 | error = zfs_trunc(zp, off); | |
1792 | } else { | |
1793 | if ((error = zfs_free_range(zp, off, len)) == 0 && | |
428870ff | 1794 | off + len > zp->z_size) |
b128c09f BB |
1795 | error = zfs_extend(zp, off+len); |
1796 | } | |
1797 | if (error || !log) | |
223df016 | 1798 | goto out; |
b128c09f | 1799 | log: |
0037b49e | 1800 | tx = dmu_tx_create(zfsvfs->z_os); |
428870ff BB |
1801 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1802 | zfs_sa_upgrade_txholds(tx, zp); | |
384f8a09 | 1803 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1804 | if (error) { |
b128c09f | 1805 | dmu_tx_abort(tx); |
223df016 | 1806 | goto out; |
b128c09f BB |
1807 | } |
1808 | ||
0037b49e BB |
1809 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16); |
1810 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16); | |
1811 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), | |
428870ff | 1812 | NULL, &zp->z_pflags, 8); |
0df9673f | 1813 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
428870ff BB |
1814 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
1815 | ASSERT(error == 0); | |
1816 | ||
b128c09f BB |
1817 | zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); |
1818 | ||
1819 | dmu_tx_commit(tx); | |
223df016 | 1820 | |
960e08fe | 1821 | zfs_inode_update(zp); |
223df016 TC |
1822 | error = 0; |
1823 | ||
1824 | out: | |
1825 | /* | |
1826 | * Truncate the page cache - for file truncate operations, use | |
1827 | * the purpose-built API for truncations. For punching operations, | |
cb08f063 | 1828 | * the truncation is handled under a range lock in zfs_free_range. |
223df016 TC |
1829 | */ |
1830 | if (len == 0) | |
1831 | truncate_setsize(ZTOI(zp), off); | |
223df016 | 1832 | return (error); |
b128c09f BB |
1833 | } |
1834 | ||
34dc7c2f BB |
1835 | void |
1836 | zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) | |
1837 | { | |
22872ff5 | 1838 | struct super_block *sb; |
0037b49e | 1839 | zfsvfs_t *zfsvfs; |
428870ff | 1840 | uint64_t moid, obj, sa_obj, version; |
22872ff5 | 1841 | uint64_t sense = ZFS_CASE_SENSITIVE; |
34dc7c2f BB |
1842 | uint64_t norm = 0; |
1843 | nvpair_t *elem; | |
c96c36fa | 1844 | int size; |
34dc7c2f | 1845 | int error; |
22872ff5 BB |
1846 | int i; |
1847 | znode_t *rootzp = NULL; | |
1848 | vattr_t vattr; | |
1849 | znode_t *zp; | |
1850 | zfs_acl_ids_t acl_ids; | |
34dc7c2f BB |
1851 | |
1852 | /* | |
1853 | * First attempt to create master node. | |
1854 | */ | |
1855 | /* | |
1856 | * In an empty objset, there are no blocks to read and thus | |
1857 | * there can be no i/o errors (which we assert below). | |
1858 | */ | |
1859 | moid = MASTER_NODE_OBJ; | |
1860 | error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, | |
1861 | DMU_OT_NONE, 0, tx); | |
1862 | ASSERT(error == 0); | |
1863 | ||
1864 | /* | |
1865 | * Set starting attributes. | |
1866 | */ | |
428870ff | 1867 | version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); |
34dc7c2f BB |
1868 | elem = NULL; |
1869 | while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { | |
1870 | /* For the moment we expect all zpl props to be uint64_ts */ | |
1871 | uint64_t val; | |
1872 | char *name; | |
1873 | ||
1874 | ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); | |
1875 | VERIFY(nvpair_value_uint64(elem, &val) == 0); | |
1876 | name = nvpair_name(elem); | |
1877 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { | |
9babb374 BB |
1878 | if (val < version) |
1879 | version = val; | |
34dc7c2f BB |
1880 | } else { |
1881 | error = zap_update(os, moid, name, 8, 1, &val, tx); | |
1882 | } | |
1883 | ASSERT(error == 0); | |
1884 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) | |
1885 | norm = val; | |
22872ff5 BB |
1886 | else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) |
1887 | sense = val; | |
34dc7c2f BB |
1888 | } |
1889 | ASSERT(version != 0); | |
9babb374 | 1890 | error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); |
34dc7c2f | 1891 | |
428870ff BB |
1892 | /* |
1893 | * Create zap object used for SA attribute registration | |
1894 | */ | |
1895 | ||
1896 | if (version >= ZPL_VERSION_SA) { | |
1897 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1898 | DMU_OT_NONE, 0, tx); | |
1899 | error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
1900 | ASSERT(error == 0); | |
1901 | } else { | |
1902 | sa_obj = 0; | |
1903 | } | |
34dc7c2f BB |
1904 | /* |
1905 | * Create a delete queue. | |
1906 | */ | |
9babb374 | 1907 | obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); |
34dc7c2f | 1908 | |
9babb374 | 1909 | error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); |
34dc7c2f BB |
1910 | ASSERT(error == 0); |
1911 | ||
9babb374 | 1912 | /* |
0037b49e | 1913 | * Create root znode. Create minimal znode/inode/zfsvfs/sb |
22872ff5 | 1914 | * to allow zfs_mknode to work. |
9babb374 | 1915 | */ |
22872ff5 BB |
1916 | vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID; |
1917 | vattr.va_mode = S_IFDIR|0755; | |
1918 | vattr.va_uid = crgetuid(cr); | |
1919 | vattr.va_gid = crgetgid(cr); | |
1920 | ||
79c76d5b | 1921 | rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
22872ff5 BB |
1922 | rootzp->z_moved = 0; |
1923 | rootzp->z_unlinked = 0; | |
1924 | rootzp->z_atime_dirty = 0; | |
1925 | rootzp->z_is_sa = USE_SA(version, os); | |
9c5167d1 | 1926 | rootzp->z_pflags = 0; |
22872ff5 | 1927 | |
0037b49e BB |
1928 | zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); |
1929 | zfsvfs->z_os = os; | |
1930 | zfsvfs->z_parent = zfsvfs; | |
1931 | zfsvfs->z_version = version; | |
1932 | zfsvfs->z_use_fuids = USE_FUIDS(version, os); | |
1933 | zfsvfs->z_use_sa = USE_SA(version, os); | |
1934 | zfsvfs->z_norm = norm; | |
22872ff5 | 1935 | |
79c76d5b | 1936 | sb = kmem_zalloc(sizeof (struct super_block), KM_SLEEP); |
0037b49e | 1937 | sb->s_fs_info = zfsvfs; |
22872ff5 BB |
1938 | |
1939 | ZTOI(rootzp)->i_sb = sb; | |
1940 | ||
1941 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, | |
0037b49e | 1942 | &zfsvfs->z_attr_table); |
9babb374 | 1943 | |
22872ff5 | 1944 | ASSERT(error == 0); |
9babb374 | 1945 | |
60101509 | 1946 | /* |
22872ff5 BB |
1947 | * Fold case on file systems that are always or sometimes case |
1948 | * insensitive. | |
60101509 | 1949 | */ |
22872ff5 | 1950 | if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) |
0037b49e | 1951 | zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; |
60101509 | 1952 | |
0037b49e BB |
1953 | mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
1954 | list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), | |
22872ff5 | 1955 | offsetof(znode_t, z_link_node)); |
60101509 | 1956 | |
c96c36fa | 1957 | size = MIN(1 << (highbit64(zfs_object_mutex_size)-1), ZFS_OBJ_MTX_MAX); |
0037b49e BB |
1958 | zfsvfs->z_hold_size = size; |
1959 | zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, | |
1960 | KM_SLEEP); | |
1961 | zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
c96c36fa | 1962 | for (i = 0; i != size; i++) { |
0037b49e | 1963 | avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare, |
c96c36fa | 1964 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); |
0037b49e | 1965 | mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); |
c96c36fa | 1966 | } |
60101509 | 1967 | |
22872ff5 BB |
1968 | VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, |
1969 | cr, NULL, &acl_ids)); | |
1970 | zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); | |
1971 | ASSERT3P(zp, ==, rootzp); | |
1972 | error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); | |
1973 | ASSERT(error == 0); | |
1974 | zfs_acl_ids_free(&acl_ids); | |
60101509 | 1975 | |
22872ff5 BB |
1976 | atomic_set(&ZTOI(rootzp)->i_count, 0); |
1977 | sa_handle_destroy(rootzp->z_sa_hdl); | |
22872ff5 BB |
1978 | kmem_cache_free(znode_cache, rootzp); |
1979 | ||
1980 | /* | |
1981 | * Create shares directory | |
1982 | */ | |
0037b49e | 1983 | error = zfs_create_share_dir(zfsvfs, tx); |
9babb374 | 1984 | ASSERT(error == 0); |
428870ff | 1985 | |
c96c36fa | 1986 | for (i = 0; i != size; i++) { |
0037b49e BB |
1987 | avl_destroy(&zfsvfs->z_hold_trees[i]); |
1988 | mutex_destroy(&zfsvfs->z_hold_locks[i]); | |
c96c36fa | 1989 | } |
2708f716 | 1990 | |
c17486b2 GN |
1991 | mutex_destroy(&zfsvfs->z_znodes_lock); |
1992 | ||
0037b49e BB |
1993 | vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size); |
1994 | vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size); | |
2708f716 | 1995 | kmem_free(sb, sizeof (struct super_block)); |
0037b49e | 1996 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); |
34dc7c2f | 1997 | } |
34dc7c2f | 1998 | #endif /* _KERNEL */ |
428870ff | 1999 | |
34dc7c2f | 2000 | static int |
572e2857 BB |
2001 | zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) |
2002 | { | |
2003 | uint64_t sa_obj = 0; | |
2004 | int error; | |
2005 | ||
2006 | error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
2007 | if (error != 0 && error != ENOENT) | |
2008 | return (error); | |
2009 | ||
2010 | error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); | |
2011 | return (error); | |
2012 | } | |
2013 | ||
2014 | static int | |
2015 | zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, | |
7b8518cb | 2016 | dmu_buf_t **db, void *tag) |
34dc7c2f | 2017 | { |
34dc7c2f | 2018 | dmu_object_info_t doi; |
34dc7c2f | 2019 | int error; |
428870ff | 2020 | |
7b8518cb | 2021 | if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) |
34dc7c2f BB |
2022 | return (error); |
2023 | ||
572e2857 | 2024 | dmu_object_info_from_db(*db, &doi); |
428870ff BB |
2025 | if ((doi.doi_bonus_type != DMU_OT_SA && |
2026 | doi.doi_bonus_type != DMU_OT_ZNODE) || | |
d6320ddb BB |
2027 | (doi.doi_bonus_type == DMU_OT_ZNODE && |
2028 | doi.doi_bonus_size < sizeof (znode_phys_t))) { | |
7b8518cb | 2029 | sa_buf_rele(*db, tag); |
2e528b49 | 2030 | return (SET_ERROR(ENOTSUP)); |
34dc7c2f BB |
2031 | } |
2032 | ||
572e2857 BB |
2033 | error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); |
2034 | if (error != 0) { | |
7b8518cb | 2035 | sa_buf_rele(*db, tag); |
428870ff BB |
2036 | return (error); |
2037 | } | |
2038 | ||
572e2857 BB |
2039 | return (0); |
2040 | } | |
2041 | ||
2042 | void | |
7b8518cb | 2043 | zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) |
572e2857 BB |
2044 | { |
2045 | sa_handle_destroy(hdl); | |
7b8518cb | 2046 | sa_buf_rele(db, tag); |
572e2857 BB |
2047 | } |
2048 | ||
2049 | /* | |
2050 | * Given an object number, return its parent object number and whether | |
2051 | * or not the object is an extended attribute directory. | |
2052 | */ | |
2053 | static int | |
b23ad7f3 JJ |
2054 | zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, |
2055 | uint64_t *pobjp, int *is_xattrdir) | |
572e2857 BB |
2056 | { |
2057 | uint64_t parent; | |
2058 | uint64_t pflags; | |
2059 | uint64_t mode; | |
b23ad7f3 | 2060 | uint64_t parent_mode; |
572e2857 | 2061 | sa_bulk_attr_t bulk[3]; |
b23ad7f3 JJ |
2062 | sa_handle_t *sa_hdl; |
2063 | dmu_buf_t *sa_db; | |
572e2857 BB |
2064 | int count = 0; |
2065 | int error; | |
2066 | ||
2067 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, | |
2068 | &parent, sizeof (parent)); | |
428870ff | 2069 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, |
572e2857 | 2070 | &pflags, sizeof (pflags)); |
428870ff | 2071 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, |
572e2857 | 2072 | &mode, sizeof (mode)); |
428870ff | 2073 | |
572e2857 | 2074 | if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) |
428870ff | 2075 | return (error); |
572e2857 | 2076 | |
b23ad7f3 JJ |
2077 | /* |
2078 | * When a link is removed its parent pointer is not changed and will | |
2079 | * be invalid. There are two cases where a link is removed but the | |
2080 | * file stays around, when it goes to the delete queue and when there | |
2081 | * are additional links. | |
2082 | */ | |
2083 | error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); | |
2084 | if (error != 0) | |
2085 | return (error); | |
2086 | ||
2087 | error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); | |
2088 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); | |
2089 | if (error != 0) | |
2090 | return (error); | |
2091 | ||
428870ff | 2092 | *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); |
34dc7c2f | 2093 | |
b23ad7f3 JJ |
2094 | /* |
2095 | * Extended attributes can be applied to files, directories, etc. | |
2096 | * Otherwise the parent must be a directory. | |
2097 | */ | |
2098 | if (!*is_xattrdir && !S_ISDIR(parent_mode)) | |
ecb2b7dc | 2099 | return (SET_ERROR(EINVAL)); |
b23ad7f3 JJ |
2100 | |
2101 | *pobjp = parent; | |
2102 | ||
34dc7c2f BB |
2103 | return (0); |
2104 | } | |
2105 | ||
572e2857 BB |
2106 | /* |
2107 | * Given an object number, return some zpl level statistics | |
2108 | */ | |
2109 | static int | |
2110 | zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, | |
2111 | zfs_stat_t *sb) | |
34dc7c2f | 2112 | { |
572e2857 BB |
2113 | sa_bulk_attr_t bulk[4]; |
2114 | int count = 0; | |
2115 | ||
2116 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, | |
2117 | &sb->zs_mode, sizeof (sb->zs_mode)); | |
2118 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, | |
2119 | &sb->zs_gen, sizeof (sb->zs_gen)); | |
2120 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, | |
2121 | &sb->zs_links, sizeof (sb->zs_links)); | |
2122 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, | |
2123 | &sb->zs_ctime, sizeof (sb->zs_ctime)); | |
2124 | ||
2125 | return (sa_bulk_lookup(hdl, bulk, count)); | |
2126 | } | |
2127 | ||
2128 | static int | |
2129 | zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, | |
2130 | sa_attr_type_t *sa_table, char *buf, int len) | |
2131 | { | |
2132 | sa_handle_t *sa_hdl; | |
2133 | sa_handle_t *prevhdl = NULL; | |
2134 | dmu_buf_t *prevdb = NULL; | |
2135 | dmu_buf_t *sa_db = NULL; | |
34dc7c2f BB |
2136 | char *path = buf + len - 1; |
2137 | int error; | |
2138 | ||
2139 | *path = '\0'; | |
572e2857 | 2140 | sa_hdl = hdl; |
428870ff | 2141 | |
64c1dcef PD |
2142 | uint64_t deleteq_obj; |
2143 | VERIFY0(zap_lookup(osp, MASTER_NODE_OBJ, | |
2144 | ZFS_UNLINKED_SET, sizeof (uint64_t), 1, &deleteq_obj)); | |
2145 | error = zap_lookup_int(osp, deleteq_obj, obj); | |
2146 | if (error == 0) { | |
2147 | return (ESTALE); | |
2148 | } else if (error != ENOENT) { | |
2149 | return (error); | |
2150 | } | |
2151 | error = 0; | |
2152 | ||
34dc7c2f | 2153 | for (;;) { |
17897ce2 | 2154 | uint64_t pobj = 0; |
34dc7c2f BB |
2155 | char component[MAXNAMELEN + 2]; |
2156 | size_t complen; | |
17897ce2 | 2157 | int is_xattrdir = 0; |
34dc7c2f | 2158 | |
572e2857 | 2159 | if (prevdb) |
7b8518cb | 2160 | zfs_release_sa_handle(prevhdl, prevdb, FTAG); |
572e2857 | 2161 | |
b23ad7f3 | 2162 | if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, |
572e2857 | 2163 | &is_xattrdir)) != 0) |
34dc7c2f BB |
2164 | break; |
2165 | ||
2166 | if (pobj == obj) { | |
2167 | if (path[0] != '/') | |
2168 | *--path = '/'; | |
2169 | break; | |
2170 | } | |
2171 | ||
2172 | component[0] = '/'; | |
2173 | if (is_xattrdir) { | |
2174 | (void) sprintf(component + 1, "<xattrdir>"); | |
2175 | } else { | |
2176 | error = zap_value_search(osp, pobj, obj, | |
2177 | ZFS_DIRENT_OBJ(-1ULL), component + 1); | |
2178 | if (error != 0) | |
2179 | break; | |
2180 | } | |
2181 | ||
2182 | complen = strlen(component); | |
2183 | path -= complen; | |
2184 | ASSERT(path >= buf); | |
2185 | bcopy(component, path, complen); | |
2186 | obj = pobj; | |
572e2857 BB |
2187 | |
2188 | if (sa_hdl != hdl) { | |
2189 | prevhdl = sa_hdl; | |
2190 | prevdb = sa_db; | |
2191 | } | |
7b8518cb | 2192 | error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); |
572e2857 BB |
2193 | if (error != 0) { |
2194 | sa_hdl = prevhdl; | |
2195 | sa_db = prevdb; | |
2196 | break; | |
2197 | } | |
2198 | } | |
2199 | ||
2200 | if (sa_hdl != NULL && sa_hdl != hdl) { | |
2201 | ASSERT(sa_db != NULL); | |
7b8518cb | 2202 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); |
34dc7c2f BB |
2203 | } |
2204 | ||
2205 | if (error == 0) | |
2206 | (void) memmove(buf, path, buf + len - path); | |
428870ff | 2207 | |
34dc7c2f BB |
2208 | return (error); |
2209 | } | |
572e2857 BB |
2210 | |
2211 | int | |
2212 | zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) | |
2213 | { | |
2214 | sa_attr_type_t *sa_table; | |
2215 | sa_handle_t *hdl; | |
2216 | dmu_buf_t *db; | |
2217 | int error; | |
2218 | ||
2219 | error = zfs_sa_setup(osp, &sa_table); | |
2220 | if (error != 0) | |
2221 | return (error); | |
2222 | ||
7b8518cb | 2223 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2224 | if (error != 0) |
2225 | return (error); | |
2226 | ||
2227 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2228 | ||
7b8518cb | 2229 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2230 | return (error); |
2231 | } | |
2232 | ||
2233 | int | |
2234 | zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, | |
2235 | char *buf, int len) | |
2236 | { | |
2237 | char *path = buf + len - 1; | |
2238 | sa_attr_type_t *sa_table; | |
2239 | sa_handle_t *hdl; | |
2240 | dmu_buf_t *db; | |
2241 | int error; | |
2242 | ||
2243 | *path = '\0'; | |
2244 | ||
2245 | error = zfs_sa_setup(osp, &sa_table); | |
2246 | if (error != 0) | |
2247 | return (error); | |
2248 | ||
7b8518cb | 2249 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2250 | if (error != 0) |
2251 | return (error); | |
2252 | ||
2253 | error = zfs_obj_to_stats_impl(hdl, sa_table, sb); | |
2254 | if (error != 0) { | |
7b8518cb | 2255 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2256 | return (error); |
2257 | } | |
2258 | ||
2259 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2260 | ||
7b8518cb | 2261 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2262 | return (error); |
2263 | } | |
c28b2279 | 2264 | |
93ce2b4c | 2265 | #if defined(_KERNEL) |
c28b2279 BB |
2266 | EXPORT_SYMBOL(zfs_create_fs); |
2267 | EXPORT_SYMBOL(zfs_obj_to_path); | |
0720116d | 2268 | |
02730c33 | 2269 | /* CSTYLED */ |
0720116d BB |
2270 | module_param(zfs_object_mutex_size, uint, 0644); |
2271 | MODULE_PARM_DESC(zfs_object_mutex_size, "Size of znode hold array"); | |
c28b2279 | 2272 | #endif |