]>
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. |
19d55079 | 23 | * Copyright (c) 2012, 2014 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> | |
32 | #include <sys/systm.h> | |
33 | #include <sys/sysmacros.h> | |
34 | #include <sys/resource.h> | |
35 | #include <sys/mntent.h> | |
36 | #include <sys/mkdev.h> | |
37 | #include <sys/u8_textprep.h> | |
38 | #include <sys/dsl_dataset.h> | |
39 | #include <sys/vfs.h> | |
40 | #include <sys/vfs_opreg.h> | |
41 | #include <sys/vnode.h> | |
42 | #include <sys/file.h> | |
43 | #include <sys/kmem.h> | |
44 | #include <sys/errno.h> | |
45 | #include <sys/unistd.h> | |
46 | #include <sys/mode.h> | |
47 | #include <sys/atomic.h> | |
48 | #include <vm/pvn.h> | |
49 | #include "fs/fs_subr.h" | |
50 | #include <sys/zfs_dir.h> | |
51 | #include <sys/zfs_acl.h> | |
52 | #include <sys/zfs_ioctl.h> | |
53 | #include <sys/zfs_rlock.h> | |
54 | #include <sys/zfs_fuid.h> | |
3558fd73 | 55 | #include <sys/zfs_vnops.h> |
ebe7e575 | 56 | #include <sys/zfs_ctldir.h> |
428870ff | 57 | #include <sys/dnode.h> |
34dc7c2f BB |
58 | #include <sys/fs/zfs.h> |
59 | #include <sys/kidmap.h> | |
3558fd73 | 60 | #include <sys/zpl.h> |
34dc7c2f BB |
61 | #endif /* _KERNEL */ |
62 | ||
63 | #include <sys/dmu.h> | |
f1512ee6 | 64 | #include <sys/dmu_objset.h> |
34dc7c2f BB |
65 | #include <sys/refcount.h> |
66 | #include <sys/stat.h> | |
67 | #include <sys/zap.h> | |
68 | #include <sys/zfs_znode.h> | |
428870ff BB |
69 | #include <sys/sa.h> |
70 | #include <sys/zfs_sa.h> | |
572e2857 | 71 | #include <sys/zfs_stat.h> |
34dc7c2f BB |
72 | |
73 | #include "zfs_prop.h" | |
428870ff | 74 | #include "zfs_comutil.h" |
34dc7c2f | 75 | |
b128c09f BB |
76 | /* |
77 | * Define ZNODE_STATS to turn on statistic gathering. By default, it is only | |
78 | * turned on when DEBUG is also defined. | |
79 | */ | |
80 | #ifdef DEBUG | |
81 | #define ZNODE_STATS | |
82 | #endif /* DEBUG */ | |
83 | ||
84 | #ifdef ZNODE_STATS | |
85 | #define ZNODE_STAT_ADD(stat) ((stat)++) | |
86 | #else | |
87 | #define ZNODE_STAT_ADD(stat) /* nothing */ | |
88 | #endif /* ZNODE_STATS */ | |
89 | ||
34dc7c2f BB |
90 | /* |
91 | * Functions needed for userland (ie: libzpool) are not put under | |
92 | * #ifdef_KERNEL; the rest of the functions have dependencies | |
93 | * (such as VFS logic) that will not compile easily in userland. | |
94 | */ | |
95 | #ifdef _KERNEL | |
9babb374 | 96 | |
b128c09f | 97 | static kmem_cache_t *znode_cache = NULL; |
c96c36fa | 98 | static kmem_cache_t *znode_hold_cache = NULL; |
0720116d | 99 | unsigned int zfs_object_mutex_size = ZFS_OBJ_MTX_SZ; |
34dc7c2f | 100 | |
34dc7c2f BB |
101 | /*ARGSUSED*/ |
102 | static int | |
b128c09f | 103 | zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) |
34dc7c2f BB |
104 | { |
105 | znode_t *zp = buf; | |
106 | ||
3558fd73 | 107 | inode_init_once(ZTOI(zp)); |
b128c09f BB |
108 | list_link_init(&zp->z_link_node); |
109 | ||
34dc7c2f | 110 | mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 111 | rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); |
448d7aaa | 112 | rw_init(&zp->z_name_lock, NULL, RW_NOLOCKDEP, NULL); |
34dc7c2f | 113 | mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); |
82a37189 | 114 | rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); |
34dc7c2f BB |
115 | |
116 | mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL); | |
117 | avl_create(&zp->z_range_avl, zfs_range_compare, | |
118 | sizeof (rl_t), offsetof(rl_t, r_node)); | |
119 | ||
b128c09f | 120 | zp->z_dirlocks = NULL; |
45d1cae3 | 121 | zp->z_acl_cached = NULL; |
82a37189 | 122 | zp->z_xattr_cached = NULL; |
e89260a1 | 123 | zp->z_xattr_parent = NULL; |
572e2857 | 124 | zp->z_moved = 0; |
34dc7c2f BB |
125 | return (0); |
126 | } | |
127 | ||
128 | /*ARGSUSED*/ | |
129 | static void | |
b128c09f | 130 | zfs_znode_cache_destructor(void *buf, void *arg) |
34dc7c2f BB |
131 | { |
132 | znode_t *zp = buf; | |
133 | ||
b128c09f | 134 | ASSERT(!list_link_active(&zp->z_link_node)); |
34dc7c2f | 135 | mutex_destroy(&zp->z_lock); |
34dc7c2f BB |
136 | rw_destroy(&zp->z_parent_lock); |
137 | rw_destroy(&zp->z_name_lock); | |
138 | mutex_destroy(&zp->z_acl_lock); | |
82a37189 | 139 | rw_destroy(&zp->z_xattr_lock); |
34dc7c2f BB |
140 | avl_destroy(&zp->z_range_avl); |
141 | mutex_destroy(&zp->z_range_lock); | |
142 | ||
b128c09f | 143 | ASSERT(zp->z_dirlocks == NULL); |
45d1cae3 | 144 | ASSERT(zp->z_acl_cached == NULL); |
82a37189 | 145 | ASSERT(zp->z_xattr_cached == NULL); |
e89260a1 | 146 | ASSERT(zp->z_xattr_parent == NULL); |
b128c09f BB |
147 | } |
148 | ||
c96c36fa BB |
149 | static int |
150 | zfs_znode_hold_cache_constructor(void *buf, void *arg, int kmflags) | |
151 | { | |
152 | znode_hold_t *zh = buf; | |
153 | ||
154 | mutex_init(&zh->zh_lock, NULL, MUTEX_DEFAULT, NULL); | |
155 | refcount_create(&zh->zh_refcount); | |
156 | zh->zh_obj = ZFS_NO_OBJECT; | |
157 | ||
158 | return (0); | |
159 | } | |
160 | ||
161 | static void | |
162 | zfs_znode_hold_cache_destructor(void *buf, void *arg) | |
163 | { | |
164 | znode_hold_t *zh = buf; | |
165 | ||
166 | mutex_destroy(&zh->zh_lock); | |
167 | refcount_destroy(&zh->zh_refcount); | |
168 | } | |
169 | ||
34dc7c2f BB |
170 | void |
171 | zfs_znode_init(void) | |
172 | { | |
173 | /* | |
5074bfe8 TC |
174 | * Initialize zcache. The KMC_SLAB hint is used in order that it be |
175 | * backed by kmalloc() when on the Linux slab in order that any | |
176 | * wait_on_bit() operations on the related inode operate properly. | |
34dc7c2f BB |
177 | */ |
178 | ASSERT(znode_cache == NULL); | |
179 | znode_cache = kmem_cache_create("zfs_znode_cache", | |
180 | sizeof (znode_t), 0, zfs_znode_cache_constructor, | |
5074bfe8 | 181 | zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_SLAB); |
c96c36fa BB |
182 | |
183 | ASSERT(znode_hold_cache == NULL); | |
184 | znode_hold_cache = kmem_cache_create("zfs_znode_hold_cache", | |
185 | sizeof (znode_hold_t), 0, zfs_znode_hold_cache_constructor, | |
186 | zfs_znode_hold_cache_destructor, NULL, NULL, NULL, 0); | |
34dc7c2f BB |
187 | } |
188 | ||
189 | void | |
190 | zfs_znode_fini(void) | |
191 | { | |
34dc7c2f BB |
192 | /* |
193 | * Cleanup zcache | |
194 | */ | |
195 | if (znode_cache) | |
196 | kmem_cache_destroy(znode_cache); | |
197 | znode_cache = NULL; | |
c96c36fa BB |
198 | |
199 | if (znode_hold_cache) | |
200 | kmem_cache_destroy(znode_hold_cache); | |
201 | znode_hold_cache = NULL; | |
202 | } | |
203 | ||
204 | /* | |
205 | * The zfs_znode_hold_enter() / zfs_znode_hold_exit() functions are used to | |
206 | * serialize access to a znode and its SA buffer while the object is being | |
207 | * created or destroyed. This kind of locking would normally reside in the | |
208 | * znode itself but in this case that's impossible because the znode and SA | |
209 | * buffer may not yet exist. Therefore the locking is handled externally | |
210 | * with an array of mutexs and AVLs trees which contain per-object locks. | |
211 | * | |
212 | * In zfs_znode_hold_enter() a per-object lock is created as needed, inserted | |
213 | * in to the correct AVL tree and finally the per-object lock is held. In | |
214 | * zfs_znode_hold_exit() the process is reversed. The per-object lock is | |
215 | * released, removed from the AVL tree and destroyed if there are no waiters. | |
216 | * | |
217 | * This scheme has two important properties: | |
218 | * | |
219 | * 1) No memory allocations are performed while holding one of the z_hold_locks. | |
220 | * This ensures evict(), which can be called from direct memory reclaim, will | |
221 | * never block waiting on a z_hold_locks which just happens to have hashed | |
222 | * to the same index. | |
223 | * | |
224 | * 2) All locks used to serialize access to an object are per-object and never | |
225 | * shared. This minimizes lock contention without creating a large number | |
226 | * of dedicated locks. | |
227 | * | |
228 | * On the downside it does require znode_lock_t structures to be frequently | |
229 | * allocated and freed. However, because these are backed by a kmem cache | |
230 | * and very short lived this cost is minimal. | |
231 | */ | |
232 | int | |
233 | zfs_znode_hold_compare(const void *a, const void *b) | |
234 | { | |
235 | const znode_hold_t *zh_a = a; | |
236 | const znode_hold_t *zh_b = b; | |
237 | ||
238 | if (zh_a->zh_obj < zh_b->zh_obj) | |
239 | return (-1); | |
240 | else if (zh_a->zh_obj > zh_b->zh_obj) | |
241 | return (1); | |
242 | else | |
243 | return (0); | |
244 | } | |
245 | ||
246 | boolean_t | |
247 | zfs_znode_held(zfs_sb_t *zsb, uint64_t obj) | |
248 | { | |
249 | znode_hold_t *zh, search; | |
250 | int i = ZFS_OBJ_HASH(zsb, obj); | |
37c56346 | 251 | boolean_t held; |
c96c36fa BB |
252 | |
253 | search.zh_obj = obj; | |
254 | ||
255 | mutex_enter(&zsb->z_hold_locks[i]); | |
256 | zh = avl_find(&zsb->z_hold_trees[i], &search, NULL); | |
37c56346 | 257 | held = (zh && MUTEX_HELD(&zh->zh_lock)) ? B_TRUE : B_FALSE; |
c96c36fa BB |
258 | mutex_exit(&zsb->z_hold_locks[i]); |
259 | ||
37c56346 | 260 | return (held); |
c96c36fa BB |
261 | } |
262 | ||
263 | static znode_hold_t * | |
264 | zfs_znode_hold_enter(zfs_sb_t *zsb, uint64_t obj) | |
265 | { | |
266 | znode_hold_t *zh, *zh_new, search; | |
267 | int i = ZFS_OBJ_HASH(zsb, obj); | |
268 | boolean_t found = B_FALSE; | |
269 | ||
270 | zh_new = kmem_cache_alloc(znode_hold_cache, KM_SLEEP); | |
271 | zh_new->zh_obj = obj; | |
272 | search.zh_obj = obj; | |
273 | ||
274 | mutex_enter(&zsb->z_hold_locks[i]); | |
275 | zh = avl_find(&zsb->z_hold_trees[i], &search, NULL); | |
276 | if (likely(zh == NULL)) { | |
277 | zh = zh_new; | |
278 | avl_add(&zsb->z_hold_trees[i], zh); | |
279 | } else { | |
280 | ASSERT3U(zh->zh_obj, ==, obj); | |
281 | found = B_TRUE; | |
282 | } | |
283 | refcount_add(&zh->zh_refcount, NULL); | |
284 | mutex_exit(&zsb->z_hold_locks[i]); | |
285 | ||
286 | if (found == B_TRUE) | |
287 | kmem_cache_free(znode_hold_cache, zh_new); | |
288 | ||
289 | ASSERT(MUTEX_NOT_HELD(&zh->zh_lock)); | |
290 | ASSERT3S(refcount_count(&zh->zh_refcount), >, 0); | |
291 | mutex_enter(&zh->zh_lock); | |
292 | ||
293 | return (zh); | |
294 | } | |
295 | ||
296 | static void | |
297 | zfs_znode_hold_exit(zfs_sb_t *zsb, znode_hold_t *zh) | |
298 | { | |
299 | int i = ZFS_OBJ_HASH(zsb, zh->zh_obj); | |
300 | boolean_t remove = B_FALSE; | |
301 | ||
302 | ASSERT(zfs_znode_held(zsb, zh->zh_obj)); | |
303 | ASSERT3S(refcount_count(&zh->zh_refcount), >, 0); | |
304 | mutex_exit(&zh->zh_lock); | |
305 | ||
306 | mutex_enter(&zsb->z_hold_locks[i]); | |
307 | if (refcount_remove(&zh->zh_refcount, NULL) == 0) { | |
308 | avl_remove(&zsb->z_hold_trees[i], zh); | |
309 | remove = B_TRUE; | |
310 | } | |
311 | mutex_exit(&zsb->z_hold_locks[i]); | |
312 | ||
313 | if (remove == B_TRUE) | |
314 | kmem_cache_free(znode_hold_cache, zh); | |
34dc7c2f BB |
315 | } |
316 | ||
34dc7c2f | 317 | int |
3558fd73 | 318 | zfs_create_share_dir(zfs_sb_t *zsb, dmu_tx_t *tx) |
34dc7c2f | 319 | { |
3c9609b3 | 320 | #ifdef HAVE_SMB_SHARE |
9babb374 BB |
321 | zfs_acl_ids_t acl_ids; |
322 | vattr_t vattr; | |
323 | znode_t *sharezp; | |
324 | vnode_t *vp; | |
325 | znode_t *zp; | |
326 | int error; | |
34dc7c2f | 327 | |
9babb374 | 328 | vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; |
3558fd73 | 329 | vattr.va_mode = S_IFDIR | 0555; |
9babb374 BB |
330 | vattr.va_uid = crgetuid(kcred); |
331 | vattr.va_gid = crgetgid(kcred); | |
34dc7c2f | 332 | |
79c76d5b | 333 | sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
572e2857 | 334 | sharezp->z_moved = 0; |
9babb374 BB |
335 | sharezp->z_unlinked = 0; |
336 | sharezp->z_atime_dirty = 0; | |
337 | sharezp->z_zfsvfs = zfsvfs; | |
428870ff | 338 | sharezp->z_is_sa = zfsvfs->z_use_sa; |
34dc7c2f | 339 | |
9babb374 BB |
340 | vp = ZTOV(sharezp); |
341 | vn_reinit(vp); | |
342 | vp->v_type = VDIR; | |
34dc7c2f | 343 | |
9babb374 BB |
344 | VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, |
345 | kcred, NULL, &acl_ids)); | |
428870ff | 346 | zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); |
9babb374 BB |
347 | ASSERT3P(zp, ==, sharezp); |
348 | ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */ | |
349 | POINTER_INVALIDATE(&sharezp->z_zfsvfs); | |
350 | error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, | |
351 | ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); | |
352 | zfsvfs->z_shares_dir = sharezp->z_id; | |
353 | ||
354 | zfs_acl_ids_free(&acl_ids); | |
3558fd73 | 355 | // ZTOV(sharezp)->v_count = 0; |
428870ff | 356 | sa_handle_destroy(sharezp->z_sa_hdl); |
9babb374 | 357 | kmem_cache_free(znode_cache, sharezp); |
34dc7c2f | 358 | |
9babb374 | 359 | return (error); |
9ee7fac5 BB |
360 | #else |
361 | return (0); | |
3c9609b3 | 362 | #endif /* HAVE_SMB_SHARE */ |
34dc7c2f BB |
363 | } |
364 | ||
34dc7c2f | 365 | static void |
3558fd73 | 366 | zfs_znode_sa_init(zfs_sb_t *zsb, znode_t *zp, |
428870ff | 367 | dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) |
34dc7c2f | 368 | { |
c96c36fa | 369 | ASSERT(zfs_znode_held(zsb, zp->z_id)); |
34dc7c2f BB |
370 | |
371 | mutex_enter(&zp->z_lock); | |
372 | ||
428870ff BB |
373 | ASSERT(zp->z_sa_hdl == NULL); |
374 | ASSERT(zp->z_acl_cached == NULL); | |
375 | if (sa_hdl == NULL) { | |
3558fd73 | 376 | VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, zp, |
428870ff BB |
377 | SA_HDL_SHARED, &zp->z_sa_hdl)); |
378 | } else { | |
379 | zp->z_sa_hdl = sa_hdl; | |
380 | sa_set_userp(sa_hdl, zp); | |
381 | } | |
34dc7c2f | 382 | |
428870ff | 383 | zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; |
34dc7c2f | 384 | |
34dc7c2f | 385 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
386 | } |
387 | ||
388 | void | |
389 | zfs_znode_dmu_fini(znode_t *zp) | |
390 | { | |
c96c36fa | 391 | ASSERT(zfs_znode_held(ZTOZSB(zp), zp->z_id) || zp->z_unlinked || |
3558fd73 | 392 | RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock)); |
428870ff BB |
393 | |
394 | sa_handle_destroy(zp->z_sa_hdl); | |
395 | zp->z_sa_hdl = NULL; | |
34dc7c2f BB |
396 | } |
397 | ||
398 | /* | |
3558fd73 BB |
399 | * Called by new_inode() to allocate a new inode. |
400 | */ | |
401 | int | |
402 | zfs_inode_alloc(struct super_block *sb, struct inode **ip) | |
403 | { | |
404 | znode_t *zp; | |
405 | ||
79c76d5b | 406 | zp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
3558fd73 BB |
407 | *ip = ZTOI(zp); |
408 | ||
409 | return (0); | |
410 | } | |
411 | ||
412 | /* | |
413 | * Called in multiple places when an inode should be destroyed. | |
414 | */ | |
415 | void | |
416 | zfs_inode_destroy(struct inode *ip) | |
417 | { | |
418 | znode_t *zp = ITOZ(ip); | |
419 | zfs_sb_t *zsb = ZTOZSB(zp); | |
420 | ||
421 | mutex_enter(&zsb->z_znodes_lock); | |
7b3e34ba BB |
422 | if (list_link_active(&zp->z_link_node)) { |
423 | list_remove(&zsb->z_all_znodes, zp); | |
424 | zsb->z_nr_znodes--; | |
425 | } | |
3558fd73 BB |
426 | mutex_exit(&zsb->z_znodes_lock); |
427 | ||
428 | if (zp->z_acl_cached) { | |
429 | zfs_acl_free(zp->z_acl_cached); | |
430 | zp->z_acl_cached = NULL; | |
431 | } | |
432 | ||
82a37189 BB |
433 | if (zp->z_xattr_cached) { |
434 | nvlist_free(zp->z_xattr_cached); | |
435 | zp->z_xattr_cached = NULL; | |
436 | } | |
437 | ||
e89260a1 | 438 | if (zp->z_xattr_parent) { |
0a50679c | 439 | zfs_iput_async(ZTOI(zp->z_xattr_parent)); |
e89260a1 BB |
440 | zp->z_xattr_parent = NULL; |
441 | } | |
442 | ||
3558fd73 BB |
443 | kmem_cache_free(znode_cache, zp); |
444 | } | |
445 | ||
446 | static void | |
447 | zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip) | |
448 | { | |
aa6d8c10 | 449 | uint64_t rdev = 0; |
3558fd73 BB |
450 | |
451 | switch (ip->i_mode & S_IFMT) { | |
452 | case S_IFREG: | |
453 | ip->i_op = &zpl_inode_operations; | |
454 | ip->i_fop = &zpl_file_operations; | |
455 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
456 | break; | |
457 | ||
458 | case S_IFDIR: | |
459 | ip->i_op = &zpl_dir_inode_operations; | |
460 | ip->i_fop = &zpl_dir_file_operations; | |
461 | ITOZ(ip)->z_zn_prefetch = B_TRUE; | |
462 | break; | |
463 | ||
464 | case S_IFLNK: | |
465 | ip->i_op = &zpl_symlink_inode_operations; | |
466 | break; | |
467 | ||
aa6d8c10 NB |
468 | /* |
469 | * rdev is only stored in a SA only for device files. | |
470 | */ | |
3558fd73 BB |
471 | case S_IFCHR: |
472 | case S_IFBLK: | |
53b1d979 BB |
473 | sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb), &rdev, |
474 | sizeof (rdev)); | |
aa6d8c10 NB |
475 | /*FALLTHROUGH*/ |
476 | case S_IFIFO: | |
477 | case S_IFSOCK: | |
3558fd73 BB |
478 | init_special_inode(ip, ip->i_mode, rdev); |
479 | ip->i_op = &zpl_special_inode_operations; | |
480 | break; | |
481 | ||
482 | default: | |
53b1d979 BB |
483 | zfs_panic_recover("inode %llu has invalid mode: 0x%x\n", |
484 | (u_longlong_t)ip->i_ino, ip->i_mode); | |
485 | ||
486 | /* Assume the inode is a file and attempt to continue */ | |
487 | ip->i_mode = S_IFREG | 0644; | |
488 | ip->i_op = &zpl_inode_operations; | |
489 | ip->i_fop = &zpl_file_operations; | |
490 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
491 | break; | |
3558fd73 BB |
492 | } |
493 | } | |
494 | ||
704cd075 CC |
495 | void |
496 | zfs_set_inode_flags(znode_t *zp, struct inode *ip) | |
497 | { | |
498 | /* | |
499 | * Linux and Solaris have different sets of file attributes, so we | |
500 | * restrict this conversion to the intersection of the two. | |
501 | */ | |
502 | ||
503 | if (zp->z_pflags & ZFS_IMMUTABLE) | |
504 | ip->i_flags |= S_IMMUTABLE; | |
505 | else | |
506 | ip->i_flags &= ~S_IMMUTABLE; | |
507 | ||
508 | if (zp->z_pflags & ZFS_APPENDONLY) | |
509 | ip->i_flags |= S_APPEND; | |
510 | else | |
511 | ip->i_flags &= ~S_APPEND; | |
512 | } | |
513 | ||
514 | /* | |
515 | * Update the embedded inode given the znode. We should work toward | |
516 | * eliminating this function as soon as possible by removing values | |
517 | * which are duplicated between the znode and inode. If the generic | |
518 | * inode has the correct field it should be used, and the ZFS code | |
519 | * updated to access the inode. This can be done incrementally. | |
520 | */ | |
521 | static void | |
522 | zfs_inode_update_impl(znode_t *zp, boolean_t new) | |
523 | { | |
524 | zfs_sb_t *zsb; | |
525 | struct inode *ip; | |
526 | uint32_t blksize; | |
527 | u_longlong_t i_blocks; | |
528 | uint64_t atime[2], mtime[2], ctime[2]; | |
529 | ||
530 | ASSERT(zp != NULL); | |
531 | zsb = ZTOZSB(zp); | |
532 | ip = ZTOI(zp); | |
533 | ||
534 | /* Skip .zfs control nodes which do not exist on disk. */ | |
535 | if (zfsctl_is_node(ip)) | |
536 | return; | |
537 | ||
538 | sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16); | |
539 | sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16); | |
540 | sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16); | |
541 | ||
542 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, &i_blocks); | |
543 | ||
544 | spin_lock(&ip->i_lock); | |
c1570649 | 545 | ip->i_generation = zp->z_gen; |
704cd075 CC |
546 | ip->i_uid = SUID_TO_KUID(zp->z_uid); |
547 | ip->i_gid = SGID_TO_KGID(zp->z_gid); | |
548 | set_nlink(ip, zp->z_links); | |
549 | ip->i_mode = zp->z_mode; | |
550 | zfs_set_inode_flags(zp, ip); | |
551 | ip->i_blkbits = SPA_MINBLOCKSHIFT; | |
552 | ip->i_blocks = i_blocks; | |
553 | ||
554 | /* | |
555 | * Only read atime from SA if we are newly created inode (or rezget), | |
556 | * otherwise i_atime might be dirty. | |
557 | */ | |
558 | if (new) | |
559 | ZFS_TIME_DECODE(&ip->i_atime, atime); | |
560 | ZFS_TIME_DECODE(&ip->i_mtime, mtime); | |
561 | ZFS_TIME_DECODE(&ip->i_ctime, ctime); | |
562 | ||
563 | i_size_write(ip, zp->z_size); | |
564 | spin_unlock(&ip->i_lock); | |
565 | } | |
566 | ||
567 | static void | |
568 | zfs_inode_update_new(znode_t *zp) | |
569 | { | |
570 | zfs_inode_update_impl(zp, B_TRUE); | |
571 | } | |
572 | ||
573 | void | |
574 | zfs_inode_update(znode_t *zp) | |
575 | { | |
576 | zfs_inode_update_impl(zp, B_FALSE); | |
577 | } | |
578 | ||
3558fd73 BB |
579 | /* |
580 | * Construct a znode+inode and initialize. | |
34dc7c2f BB |
581 | * |
582 | * This does not do a call to dmu_set_user() that is | |
583 | * up to the caller to do, in case you don't want to | |
584 | * return the znode | |
585 | */ | |
586 | static znode_t * | |
3558fd73 | 587 | zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz, |
c85b224f | 588 | dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl, |
7b3e34ba | 589 | struct inode *dip) |
34dc7c2f BB |
590 | { |
591 | znode_t *zp; | |
3558fd73 | 592 | struct inode *ip; |
7f89ae6b | 593 | uint64_t mode; |
428870ff | 594 | uint64_t parent; |
0df9673f | 595 | sa_bulk_attr_t bulk[8]; |
428870ff | 596 | int count = 0; |
34dc7c2f | 597 | |
3558fd73 | 598 | ASSERT(zsb != NULL); |
34dc7c2f | 599 | |
3558fd73 BB |
600 | ip = new_inode(zsb->z_sb); |
601 | if (ip == NULL) | |
602 | return (NULL); | |
7304b6e5 | 603 | |
3558fd73 | 604 | zp = ITOZ(ip); |
34dc7c2f | 605 | ASSERT(zp->z_dirlocks == NULL); |
ebe7e575 BB |
606 | ASSERT3P(zp->z_acl_cached, ==, NULL); |
607 | ASSERT3P(zp->z_xattr_cached, ==, NULL); | |
e89260a1 | 608 | ASSERT3P(zp->z_xattr_parent, ==, NULL); |
572e2857 | 609 | zp->z_moved = 0; |
428870ff | 610 | zp->z_sa_hdl = NULL; |
34dc7c2f BB |
611 | zp->z_unlinked = 0; |
612 | zp->z_atime_dirty = 0; | |
613 | zp->z_mapcnt = 0; | |
34dc7c2f BB |
614 | zp->z_id = db->db_object; |
615 | zp->z_blksz = blksz; | |
616 | zp->z_seq = 0x7A4653; | |
617 | zp->z_sync_cnt = 0; | |
ebe7e575 BB |
618 | zp->z_is_zvol = B_FALSE; |
619 | zp->z_is_mapped = B_FALSE; | |
620 | zp->z_is_ctldir = B_FALSE; | |
7b3e34ba | 621 | zp->z_is_stale = B_FALSE; |
34dc7c2f | 622 | |
3558fd73 BB |
623 | zfs_znode_sa_init(zsb, zp, db, obj_type, hdl); |
624 | ||
7f89ae6b | 625 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8); |
c1570649 | 626 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8); |
3558fd73 BB |
627 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); |
628 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8); | |
629 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, | |
428870ff | 630 | &zp->z_pflags, 8); |
3558fd73 | 631 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, |
7304b6e5 | 632 | &parent, 8); |
3558fd73 BB |
633 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8); |
634 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8); | |
428870ff | 635 | |
c1570649 | 636 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) { |
428870ff BB |
637 | if (hdl == NULL) |
638 | sa_handle_destroy(zp->z_sa_hdl); | |
07d63f0c | 639 | zp->z_sa_hdl = NULL; |
3558fd73 | 640 | goto error; |
34dc7c2f | 641 | } |
7304b6e5 | 642 | |
7f89ae6b BB |
643 | zp->z_mode = mode; |
644 | ||
e89260a1 BB |
645 | /* |
646 | * xattr znodes hold a reference on their unique parent | |
647 | */ | |
648 | if (dip && zp->z_pflags & ZFS_XATTR) { | |
649 | igrab(dip); | |
650 | zp->z_xattr_parent = ITOZ(dip); | |
651 | } | |
652 | ||
3558fd73 | 653 | ip->i_ino = obj; |
704cd075 | 654 | zfs_inode_update_new(zp); |
3558fd73 BB |
655 | zfs_inode_set_ops(zsb, ip); |
656 | ||
7b3e34ba BB |
657 | /* |
658 | * The only way insert_inode_locked() can fail is if the ip->i_ino | |
659 | * number is already hashed for this super block. This can never | |
660 | * happen because the inode numbers map 1:1 with the object numbers. | |
661 | * | |
662 | * The one exception is rolling back a mounted file system, but in | |
663 | * this case all the active inode are unhashed during the rollback. | |
664 | */ | |
665 | VERIFY3S(insert_inode_locked(ip), ==, 0); | |
c85b224f | 666 | |
3558fd73 BB |
667 | mutex_enter(&zsb->z_znodes_lock); |
668 | list_insert_tail(&zsb->z_all_znodes, zp); | |
ab26409d | 669 | zsb->z_nr_znodes++; |
b128c09f | 670 | membar_producer(); |
3558fd73 | 671 | mutex_exit(&zsb->z_znodes_lock); |
b128c09f | 672 | |
3558fd73 | 673 | unlock_new_inode(ip); |
34dc7c2f | 674 | return (zp); |
3558fd73 BB |
675 | |
676 | error: | |
3558fd73 | 677 | iput(ip); |
d1d7e268 | 678 | return (NULL); |
34dc7c2f BB |
679 | } |
680 | ||
1e8db771 BB |
681 | /* |
682 | * Safely mark an inode dirty. Inodes which are part of a read-only | |
683 | * file system or snapshot may not be dirtied. | |
684 | */ | |
685 | void | |
686 | zfs_mark_inode_dirty(struct inode *ip) | |
687 | { | |
688 | zfs_sb_t *zsb = ITOZSB(ip); | |
689 | ||
690 | if (zfs_is_readonly(zsb) || dmu_objset_is_snapshot(zsb->z_os)) | |
691 | return; | |
692 | ||
693 | mark_inode_dirty(ip); | |
694 | } | |
695 | ||
428870ff BB |
696 | static uint64_t empty_xattr; |
697 | static uint64_t pad[4]; | |
698 | static zfs_acl_phys_t acl_phys; | |
34dc7c2f BB |
699 | /* |
700 | * Create a new DMU object to hold a zfs znode. | |
701 | * | |
702 | * IN: dzp - parent directory for new znode | |
703 | * vap - file attributes for new znode | |
704 | * tx - dmu transaction id for zap operations | |
705 | * cr - credentials of caller | |
706 | * flag - flags: | |
707 | * IS_ROOT_NODE - new object will be root | |
708 | * IS_XATTR - new object is an attribute | |
34dc7c2f BB |
709 | * bonuslen - length of bonus buffer |
710 | * setaclp - File/Dir initial ACL | |
711 | * fuidp - Tracks fuid allocation. | |
712 | * | |
713 | * OUT: zpp - allocated znode | |
714 | * | |
715 | */ | |
716 | void | |
717 | zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, | |
428870ff | 718 | uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) |
34dc7c2f | 719 | { |
428870ff BB |
720 | uint64_t crtime[2], atime[2], mtime[2], ctime[2]; |
721 | uint64_t mode, size, links, parent, pflags; | |
572e2857 | 722 | uint64_t dzp_pflags = 0; |
428870ff | 723 | uint64_t rdev = 0; |
3558fd73 | 724 | zfs_sb_t *zsb = ZTOZSB(dzp); |
428870ff | 725 | dmu_buf_t *db; |
34dc7c2f BB |
726 | timestruc_t now; |
727 | uint64_t gen, obj; | |
428870ff BB |
728 | int bonuslen; |
729 | sa_handle_t *sa_hdl; | |
730 | dmu_object_type_t obj_type; | |
f30484af | 731 | sa_bulk_attr_t *sa_attrs; |
428870ff BB |
732 | int cnt = 0; |
733 | zfs_acl_locator_cb_t locate = { 0 }; | |
c96c36fa | 734 | znode_hold_t *zh; |
34dc7c2f | 735 | |
3558fd73 | 736 | if (zsb->z_replay) { |
34dc7c2f | 737 | obj = vap->va_nodeid; |
34dc7c2f BB |
738 | now = vap->va_ctime; /* see zfs_replay_create() */ |
739 | gen = vap->va_nblocks; /* ditto */ | |
740 | } else { | |
741 | obj = 0; | |
742 | gethrestime(&now); | |
743 | gen = dmu_tx_get_txg(tx); | |
744 | } | |
745 | ||
3558fd73 | 746 | obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; |
428870ff BB |
747 | bonuslen = (obj_type == DMU_OT_SA) ? |
748 | DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE; | |
749 | ||
34dc7c2f BB |
750 | /* |
751 | * Create a new DMU object. | |
752 | */ | |
753 | /* | |
754 | * There's currently no mechanism for pre-reading the blocks that will | |
572e2857 | 755 | * be needed to allocate a new object, so we accept the small chance |
34dc7c2f BB |
756 | * that there will be an i/o error and we will fail one of the |
757 | * assertions below. | |
758 | */ | |
3558fd73 BB |
759 | if (S_ISDIR(vap->va_mode)) { |
760 | if (zsb->z_replay) { | |
b0bc7a84 | 761 | VERIFY0(zap_create_claim_norm(zsb->z_os, obj, |
3558fd73 | 762 | zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS, |
b0bc7a84 | 763 | obj_type, bonuslen, tx)); |
34dc7c2f | 764 | } else { |
3558fd73 BB |
765 | obj = zap_create_norm(zsb->z_os, |
766 | zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
428870ff | 767 | obj_type, bonuslen, tx); |
34dc7c2f BB |
768 | } |
769 | } else { | |
3558fd73 | 770 | if (zsb->z_replay) { |
b0bc7a84 | 771 | VERIFY0(dmu_object_claim(zsb->z_os, obj, |
34dc7c2f | 772 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
b0bc7a84 | 773 | obj_type, bonuslen, tx)); |
34dc7c2f | 774 | } else { |
3558fd73 | 775 | obj = dmu_object_alloc(zsb->z_os, |
34dc7c2f | 776 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
428870ff | 777 | obj_type, bonuslen, tx); |
34dc7c2f BB |
778 | } |
779 | } | |
34dc7c2f | 780 | |
c96c36fa | 781 | zh = zfs_znode_hold_enter(zsb, obj); |
3558fd73 | 782 | VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db)); |
34dc7c2f BB |
783 | |
784 | /* | |
785 | * If this is the root, fix up the half-initialized parent pointer | |
786 | * to reference the just-allocated physical data area. | |
787 | */ | |
788 | if (flag & IS_ROOT_NODE) { | |
34dc7c2f | 789 | dzp->z_id = obj; |
428870ff BB |
790 | } else { |
791 | dzp_pflags = dzp->z_pflags; | |
34dc7c2f BB |
792 | } |
793 | ||
794 | /* | |
795 | * If parent is an xattr, so am I. | |
796 | */ | |
428870ff | 797 | if (dzp_pflags & ZFS_XATTR) { |
34dc7c2f | 798 | flag |= IS_XATTR; |
34dc7c2f BB |
799 | } |
800 | ||
3558fd73 | 801 | if (zsb->z_use_fuids) |
428870ff BB |
802 | pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; |
803 | else | |
804 | pflags = 0; | |
34dc7c2f | 805 | |
3558fd73 | 806 | if (S_ISDIR(vap->va_mode)) { |
428870ff BB |
807 | size = 2; /* contents ("." and "..") */ |
808 | links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; | |
809 | } else { | |
810 | size = links = 0; | |
34dc7c2f BB |
811 | } |
812 | ||
aa6d8c10 | 813 | if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode)) |
dc1d7665 | 814 | rdev = vap->va_rdev; |
428870ff BB |
815 | |
816 | parent = dzp->z_id; | |
817 | mode = acl_ids->z_mode; | |
34dc7c2f | 818 | if (flag & IS_XATTR) |
428870ff | 819 | pflags |= ZFS_XATTR; |
34dc7c2f | 820 | |
428870ff BB |
821 | /* |
822 | * No execs denied will be deterimed when zfs_mode_compute() is called. | |
823 | */ | |
824 | pflags |= acl_ids->z_aclp->z_hints & | |
825 | (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| | |
826 | ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); | |
34dc7c2f | 827 | |
428870ff BB |
828 | ZFS_TIME_ENCODE(&now, crtime); |
829 | ZFS_TIME_ENCODE(&now, ctime); | |
34dc7c2f | 830 | |
3558fd73 | 831 | if (vap->va_mask & ATTR_ATIME) { |
428870ff | 832 | ZFS_TIME_ENCODE(&vap->va_atime, atime); |
34dc7c2f | 833 | } else { |
428870ff | 834 | ZFS_TIME_ENCODE(&now, atime); |
34dc7c2f BB |
835 | } |
836 | ||
3558fd73 | 837 | if (vap->va_mask & ATTR_MTIME) { |
428870ff BB |
838 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
839 | } else { | |
840 | ZFS_TIME_ENCODE(&now, mtime); | |
841 | } | |
842 | ||
843 | /* Now add in all of the "SA" attributes */ | |
3558fd73 | 844 | VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED, |
428870ff BB |
845 | &sa_hdl)); |
846 | ||
847 | /* | |
848 | * Setup the array of attributes to be replaced/set on the new file | |
849 | * | |
850 | * order for DMU_OT_ZNODE is critical since it needs to be constructed | |
851 | * in the old znode_phys_t format. Don't change this ordering | |
852 | */ | |
79c76d5b | 853 | sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); |
428870ff BB |
854 | |
855 | if (obj_type == DMU_OT_ZNODE) { | |
3558fd73 | 856 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb), |
428870ff | 857 | NULL, &atime, 16); |
3558fd73 | 858 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb), |
428870ff | 859 | NULL, &mtime, 16); |
3558fd73 | 860 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb), |
428870ff | 861 | NULL, &ctime, 16); |
3558fd73 | 862 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb), |
428870ff | 863 | NULL, &crtime, 16); |
3558fd73 | 864 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb), |
428870ff | 865 | NULL, &gen, 8); |
3558fd73 | 866 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb), |
428870ff | 867 | NULL, &mode, 8); |
3558fd73 | 868 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb), |
428870ff | 869 | NULL, &size, 8); |
3558fd73 | 870 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb), |
428870ff | 871 | NULL, &parent, 8); |
34dc7c2f | 872 | } else { |
3558fd73 | 873 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb), |
428870ff | 874 | NULL, &mode, 8); |
3558fd73 | 875 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb), |
428870ff | 876 | NULL, &size, 8); |
3558fd73 | 877 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb), |
428870ff | 878 | NULL, &gen, 8); |
3558fd73 BB |
879 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), |
880 | NULL, &acl_ids->z_fuid, 8); | |
881 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), | |
882 | NULL, &acl_ids->z_fgid, 8); | |
883 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb), | |
428870ff | 884 | NULL, &parent, 8); |
3558fd73 | 885 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb), |
428870ff | 886 | NULL, &pflags, 8); |
3558fd73 | 887 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb), |
428870ff | 888 | NULL, &atime, 16); |
3558fd73 | 889 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb), |
428870ff | 890 | NULL, &mtime, 16); |
3558fd73 | 891 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb), |
428870ff | 892 | NULL, &ctime, 16); |
3558fd73 | 893 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb), |
428870ff BB |
894 | NULL, &crtime, 16); |
895 | } | |
896 | ||
3558fd73 | 897 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8); |
428870ff BB |
898 | |
899 | if (obj_type == DMU_OT_ZNODE) { | |
3558fd73 | 900 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL, |
428870ff | 901 | &empty_xattr, 8); |
34dc7c2f | 902 | } |
428870ff | 903 | if (obj_type == DMU_OT_ZNODE || |
aa6d8c10 | 904 | (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) { |
3558fd73 | 905 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb), |
428870ff | 906 | NULL, &rdev, 8); |
428870ff BB |
907 | } |
908 | if (obj_type == DMU_OT_ZNODE) { | |
3558fd73 | 909 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb), |
428870ff | 910 | NULL, &pflags, 8); |
3558fd73 | 911 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL, |
428870ff | 912 | &acl_ids->z_fuid, 8); |
3558fd73 | 913 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL, |
428870ff | 914 | &acl_ids->z_fgid, 8); |
3558fd73 | 915 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad, |
428870ff | 916 | sizeof (uint64_t) * 4); |
3558fd73 | 917 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL, |
428870ff BB |
918 | &acl_phys, sizeof (zfs_acl_phys_t)); |
919 | } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { | |
3558fd73 | 920 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL, |
428870ff BB |
921 | &acl_ids->z_aclp->z_acl_count, 8); |
922 | locate.cb_aclp = acl_ids->z_aclp; | |
3558fd73 | 923 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb), |
428870ff BB |
924 | zfs_acl_data_locator, &locate, |
925 | acl_ids->z_aclp->z_acl_bytes); | |
926 | mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, | |
927 | acl_ids->z_fuid, acl_ids->z_fgid); | |
928 | } | |
929 | ||
930 | VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); | |
34dc7c2f | 931 | |
34dc7c2f | 932 | if (!(flag & IS_ROOT_NODE)) { |
c85b224f | 933 | *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl, |
7b3e34ba BB |
934 | ZTOI(dzp)); |
935 | VERIFY(*zpp != NULL); | |
936 | VERIFY(dzp != NULL); | |
34dc7c2f BB |
937 | } else { |
938 | /* | |
939 | * If we are creating the root node, the "parent" we | |
940 | * passed in is the znode for the root. | |
941 | */ | |
942 | *zpp = dzp; | |
428870ff BB |
943 | |
944 | (*zpp)->z_sa_hdl = sa_hdl; | |
34dc7c2f | 945 | } |
428870ff BB |
946 | |
947 | (*zpp)->z_pflags = pflags; | |
948 | (*zpp)->z_mode = mode; | |
949 | ||
428870ff BB |
950 | if (obj_type == DMU_OT_ZNODE || |
951 | acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { | |
b0bc7a84 | 952 | VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); |
428870ff | 953 | } |
d1d7e268 | 954 | kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); |
c96c36fa | 955 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
956 | } |
957 | ||
5484965a | 958 | /* |
d3cc8b15 WA |
959 | * Update in-core attributes. It is assumed the caller will be doing an |
960 | * sa_bulk_update to push the changes out. | |
5484965a BB |
961 | */ |
962 | void | |
963 | zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) | |
964 | { | |
965 | xoptattr_t *xoap; | |
966 | ||
967 | xoap = xva_getxoptattr(xvap); | |
968 | ASSERT(xoap); | |
969 | ||
970 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { | |
971 | uint64_t times[2]; | |
972 | ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); | |
973 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)), | |
974 | ×, sizeof (times), tx); | |
975 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
976 | } | |
977 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
978 | ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, | |
979 | zp->z_pflags, tx); | |
980 | XVA_SET_RTN(xvap, XAT_READONLY); | |
981 | } | |
982 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
983 | ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, | |
984 | zp->z_pflags, tx); | |
985 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
986 | } | |
987 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
988 | ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, | |
989 | zp->z_pflags, tx); | |
990 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
991 | } | |
992 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
993 | ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, | |
994 | zp->z_pflags, tx); | |
995 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
996 | } | |
997 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
998 | ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, | |
999 | zp->z_pflags, tx); | |
1000 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
1001 | } | |
1002 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
1003 | ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, | |
1004 | zp->z_pflags, tx); | |
1005 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
1006 | } | |
1007 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
1008 | ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, | |
1009 | zp->z_pflags, tx); | |
1010 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
1011 | } | |
1012 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
1013 | ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, | |
1014 | zp->z_pflags, tx); | |
1015 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
1016 | } | |
1017 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
1018 | ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, | |
1019 | zp->z_pflags, tx); | |
1020 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
1021 | } | |
1022 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
1023 | ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, | |
1024 | xoap->xoa_av_quarantined, zp->z_pflags, tx); | |
1025 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
1026 | } | |
1027 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
1028 | ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, | |
1029 | zp->z_pflags, tx); | |
1030 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
1031 | } | |
1032 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { | |
1033 | zfs_sa_set_scanstamp(zp, xvap, tx); | |
1034 | XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); | |
1035 | } | |
1036 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
1037 | ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, | |
1038 | zp->z_pflags, tx); | |
1039 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
1040 | } | |
1041 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
1042 | ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, | |
1043 | zp->z_pflags, tx); | |
1044 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
1045 | } | |
1046 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
1047 | ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, | |
1048 | zp->z_pflags, tx); | |
1049 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
1050 | } | |
1051 | } | |
1052 | ||
34dc7c2f | 1053 | int |
3558fd73 | 1054 | zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp) |
34dc7c2f BB |
1055 | { |
1056 | dmu_object_info_t doi; | |
1057 | dmu_buf_t *db; | |
1058 | znode_t *zp; | |
c96c36fa | 1059 | znode_hold_t *zh; |
34dc7c2f | 1060 | int err; |
428870ff | 1061 | sa_handle_t *hdl; |
34dc7c2f BB |
1062 | |
1063 | *zpp = NULL; | |
1064 | ||
6f9548c4 | 1065 | again: |
c96c36fa | 1066 | zh = zfs_znode_hold_enter(zsb, obj_num); |
34dc7c2f | 1067 | |
3558fd73 | 1068 | err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db); |
34dc7c2f | 1069 | if (err) { |
c96c36fa | 1070 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1071 | return (err); |
1072 | } | |
1073 | ||
1074 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1075 | if (doi.doi_bonus_type != DMU_OT_SA && |
1076 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1077 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1078 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1079 | sa_buf_rele(db, NULL); | |
c96c36fa | 1080 | zfs_znode_hold_exit(zsb, zh); |
2e528b49 | 1081 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1082 | } |
1083 | ||
428870ff BB |
1084 | hdl = dmu_buf_get_user(db); |
1085 | if (hdl != NULL) { | |
36df2843 | 1086 | zp = sa_get_userdata(hdl); |
34dc7c2f | 1087 | |
8ac67298 | 1088 | |
34dc7c2f | 1089 | /* |
428870ff BB |
1090 | * Since "SA" does immediate eviction we |
1091 | * should never find a sa handle that doesn't | |
1092 | * know about the znode. | |
34dc7c2f | 1093 | */ |
428870ff BB |
1094 | |
1095 | ASSERT3P(zp, !=, NULL); | |
1096 | ||
1097 | mutex_enter(&zp->z_lock); | |
34dc7c2f BB |
1098 | ASSERT3U(zp->z_id, ==, obj_num); |
1099 | if (zp->z_unlinked) { | |
2e528b49 | 1100 | err = SET_ERROR(ENOENT); |
34dc7c2f | 1101 | } else { |
6f9548c4 RY |
1102 | /* |
1103 | * If igrab() returns NULL the VFS has independently | |
1104 | * determined the inode should be evicted and has | |
1105 | * called iput_final() to start the eviction process. | |
1106 | * The SA handle is still valid but because the VFS | |
1107 | * requires that the eviction succeed we must drop | |
1108 | * our locks and references to allow the eviction to | |
1109 | * complete. The zfs_zget() may then be retried. | |
1110 | * | |
1111 | * This unlikely case could be optimized by registering | |
1112 | * a sops->drop_inode() callback. The callback would | |
1113 | * need to detect the active SA hold thereby informing | |
1114 | * the VFS that this inode should not be evicted. | |
1115 | */ | |
1116 | if (igrab(ZTOI(zp)) == NULL) { | |
1117 | mutex_exit(&zp->z_lock); | |
1118 | sa_buf_rele(db, NULL); | |
c96c36fa | 1119 | zfs_znode_hold_exit(zsb, zh); |
6102d037 | 1120 | /* inode might need this to finish evict */ |
1121 | cond_resched(); | |
6f9548c4 RY |
1122 | goto again; |
1123 | } | |
34dc7c2f BB |
1124 | *zpp = zp; |
1125 | err = 0; | |
1126 | } | |
34dc7c2f | 1127 | mutex_exit(&zp->z_lock); |
f3ad9cd6 | 1128 | sa_buf_rele(db, NULL); |
c96c36fa | 1129 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1130 | return (err); |
1131 | } | |
1132 | ||
1133 | /* | |
3558fd73 | 1134 | * Not found create new znode/vnode but only if file exists. |
428870ff BB |
1135 | * |
1136 | * There is a small window where zfs_vget() could | |
1137 | * find this object while a file create is still in | |
1138 | * progress. This is checked for in zfs_znode_alloc() | |
1139 | * | |
1140 | * if zfs_znode_alloc() fails it will drop the hold on the | |
1141 | * bonus buffer. | |
34dc7c2f | 1142 | */ |
3558fd73 | 1143 | zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size, |
7b3e34ba | 1144 | doi.doi_bonus_type, obj_num, NULL, NULL); |
428870ff | 1145 | if (zp == NULL) { |
2e528b49 | 1146 | err = SET_ERROR(ENOENT); |
428870ff BB |
1147 | } else { |
1148 | *zpp = zp; | |
1149 | } | |
c96c36fa | 1150 | zfs_znode_hold_exit(zsb, zh); |
428870ff | 1151 | return (err); |
34dc7c2f BB |
1152 | } |
1153 | ||
1154 | int | |
1155 | zfs_rezget(znode_t *zp) | |
1156 | { | |
3558fd73 | 1157 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
1158 | dmu_object_info_t doi; |
1159 | dmu_buf_t *db; | |
1160 | uint64_t obj_num = zp->z_id; | |
428870ff | 1161 | uint64_t mode; |
0df9673f | 1162 | sa_bulk_attr_t bulk[7]; |
34dc7c2f | 1163 | int err; |
428870ff BB |
1164 | int count = 0; |
1165 | uint64_t gen; | |
c96c36fa | 1166 | znode_hold_t *zh; |
34dc7c2f | 1167 | |
c96c36fa | 1168 | zh = zfs_znode_hold_enter(zsb, obj_num); |
34dc7c2f | 1169 | |
428870ff BB |
1170 | mutex_enter(&zp->z_acl_lock); |
1171 | if (zp->z_acl_cached) { | |
1172 | zfs_acl_free(zp->z_acl_cached); | |
1173 | zp->z_acl_cached = NULL; | |
1174 | } | |
428870ff | 1175 | mutex_exit(&zp->z_acl_lock); |
7b3e34ba | 1176 | |
228b461b | 1177 | rw_enter(&zp->z_xattr_lock, RW_WRITER); |
7b3e34ba BB |
1178 | if (zp->z_xattr_cached) { |
1179 | nvlist_free(zp->z_xattr_cached); | |
1180 | zp->z_xattr_cached = NULL; | |
1181 | } | |
1182 | ||
1183 | if (zp->z_xattr_parent) { | |
2d501583 | 1184 | zfs_iput_async(ZTOI(zp->z_xattr_parent)); |
7b3e34ba BB |
1185 | zp->z_xattr_parent = NULL; |
1186 | } | |
1187 | rw_exit(&zp->z_xattr_lock); | |
1188 | ||
428870ff | 1189 | ASSERT(zp->z_sa_hdl == NULL); |
3558fd73 | 1190 | err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db); |
34dc7c2f | 1191 | if (err) { |
c96c36fa | 1192 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1193 | return (err); |
1194 | } | |
1195 | ||
1196 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1197 | if (doi.doi_bonus_type != DMU_OT_SA && |
1198 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1199 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1200 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1201 | sa_buf_rele(db, NULL); | |
c96c36fa | 1202 | zfs_znode_hold_exit(zsb, zh); |
2e528b49 | 1203 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1204 | } |
1205 | ||
3558fd73 | 1206 | zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL); |
428870ff BB |
1207 | |
1208 | /* reload cached values */ | |
3558fd73 | 1209 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, |
428870ff | 1210 | &gen, sizeof (gen)); |
3558fd73 | 1211 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, |
428870ff | 1212 | &zp->z_size, sizeof (zp->z_size)); |
3558fd73 | 1213 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, |
428870ff | 1214 | &zp->z_links, sizeof (zp->z_links)); |
3558fd73 | 1215 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff | 1216 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3558fd73 | 1217 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, |
572e2857 | 1218 | &zp->z_uid, sizeof (zp->z_uid)); |
3558fd73 | 1219 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, |
572e2857 | 1220 | &zp->z_gid, sizeof (zp->z_gid)); |
3558fd73 | 1221 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, |
428870ff BB |
1222 | &mode, sizeof (mode)); |
1223 | ||
428870ff BB |
1224 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { |
1225 | zfs_znode_dmu_fini(zp); | |
c96c36fa | 1226 | zfs_znode_hold_exit(zsb, zh); |
2e528b49 | 1227 | return (SET_ERROR(EIO)); |
428870ff BB |
1228 | } |
1229 | ||
572e2857 BB |
1230 | zp->z_mode = mode; |
1231 | ||
c1570649 | 1232 | if (gen != zp->z_gen) { |
428870ff | 1233 | zfs_znode_dmu_fini(zp); |
c96c36fa | 1234 | zfs_znode_hold_exit(zsb, zh); |
2e528b49 | 1235 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1236 | } |
1237 | ||
428870ff | 1238 | zp->z_unlinked = (zp->z_links == 0); |
34dc7c2f | 1239 | zp->z_blksz = doi.doi_data_block_size; |
704cd075 CC |
1240 | zp->z_atime_dirty = 0; |
1241 | zfs_inode_update_new(zp); | |
34dc7c2f | 1242 | |
c96c36fa | 1243 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1244 | |
1245 | return (0); | |
1246 | } | |
1247 | ||
1248 | void | |
1249 | zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) | |
1250 | { | |
3558fd73 BB |
1251 | zfs_sb_t *zsb = ZTOZSB(zp); |
1252 | objset_t *os = zsb->z_os; | |
34dc7c2f | 1253 | uint64_t obj = zp->z_id; |
572e2857 | 1254 | uint64_t acl_obj = zfs_external_acl(zp); |
c96c36fa | 1255 | znode_hold_t *zh; |
34dc7c2f | 1256 | |
c96c36fa | 1257 | zh = zfs_znode_hold_enter(zsb, obj); |
572e2857 BB |
1258 | if (acl_obj) { |
1259 | VERIFY(!zp->z_is_sa); | |
b128c09f | 1260 | VERIFY(0 == dmu_object_free(os, acl_obj, tx)); |
572e2857 | 1261 | } |
b128c09f | 1262 | VERIFY(0 == dmu_object_free(os, obj, tx)); |
34dc7c2f | 1263 | zfs_znode_dmu_fini(zp); |
c96c36fa | 1264 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1265 | } |
1266 | ||
1267 | void | |
1268 | zfs_zinactive(znode_t *zp) | |
1269 | { | |
3558fd73 | 1270 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f | 1271 | uint64_t z_id = zp->z_id; |
c96c36fa | 1272 | znode_hold_t *zh; |
34dc7c2f | 1273 | |
428870ff | 1274 | ASSERT(zp->z_sa_hdl); |
34dc7c2f BB |
1275 | |
1276 | /* | |
d6bd8eaa | 1277 | * Don't allow a zfs_zget() while were trying to release this znode. |
34dc7c2f | 1278 | */ |
c96c36fa | 1279 | zh = zfs_znode_hold_enter(zsb, z_id); |
d6bd8eaa | 1280 | |
34dc7c2f | 1281 | mutex_enter(&zp->z_lock); |
34dc7c2f BB |
1282 | |
1283 | /* | |
1284 | * If this was the last reference to a file with no links, | |
1285 | * remove the file from the file system. | |
1286 | */ | |
1287 | if (zp->z_unlinked) { | |
1288 | mutex_exit(&zp->z_lock); | |
c96c36fa | 1289 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1290 | zfs_rmnode(zp); |
1291 | return; | |
1292 | } | |
428870ff | 1293 | |
34dc7c2f BB |
1294 | mutex_exit(&zp->z_lock); |
1295 | zfs_znode_dmu_fini(zp); | |
d6bd8eaa | 1296 | |
c96c36fa | 1297 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1298 | } |
1299 | ||
6d111134 TC |
1300 | static inline int |
1301 | zfs_compare_timespec(struct timespec *t1, struct timespec *t2) | |
1302 | { | |
1303 | if (t1->tv_sec < t2->tv_sec) | |
1304 | return (-1); | |
1305 | ||
1306 | if (t1->tv_sec > t2->tv_sec) | |
1307 | return (1); | |
1308 | ||
1309 | return (t1->tv_nsec - t2->tv_nsec); | |
1310 | } | |
1311 | ||
6d111134 TC |
1312 | /* |
1313 | * Prepare to update znode time stamps. | |
1314 | * | |
1315 | * IN: zp - znode requiring timestamp update | |
0df9673f | 1316 | * flag - ATTR_MTIME, ATTR_CTIME flags |
6d111134 | 1317 | * |
0df9673f | 1318 | * OUT: zp - z_seq |
6d111134 TC |
1319 | * mtime - new mtime |
1320 | * ctime - new ctime | |
1321 | * | |
0df9673f CC |
1322 | * Note: We don't update atime here, because we rely on Linux VFS to do |
1323 | * atime updating. | |
6d111134 | 1324 | */ |
34dc7c2f | 1325 | void |
428870ff | 1326 | zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], |
0df9673f | 1327 | uint64_t ctime[2]) |
34dc7c2f BB |
1328 | { |
1329 | timestruc_t now; | |
1330 | ||
34dc7c2f BB |
1331 | gethrestime(&now); |
1332 | ||
0df9673f | 1333 | zp->z_seq++; |
34dc7c2f | 1334 | |
3558fd73 | 1335 | if (flag & ATTR_MTIME) { |
428870ff | 1336 | ZFS_TIME_ENCODE(&now, mtime); |
3558fd73 | 1337 | if (ZTOZSB(zp)->z_use_fuids) { |
428870ff BB |
1338 | zp->z_pflags |= (ZFS_ARCHIVE | |
1339 | ZFS_AV_MODIFIED); | |
1340 | } | |
34dc7c2f BB |
1341 | } |
1342 | ||
3558fd73 | 1343 | if (flag & ATTR_CTIME) { |
428870ff | 1344 | ZFS_TIME_ENCODE(&now, ctime); |
3558fd73 | 1345 | if (ZTOZSB(zp)->z_use_fuids) |
428870ff | 1346 | zp->z_pflags |= ZFS_ARCHIVE; |
34dc7c2f BB |
1347 | } |
1348 | } | |
1349 | ||
34dc7c2f BB |
1350 | /* |
1351 | * Grow the block size for a file. | |
1352 | * | |
1353 | * IN: zp - znode of file to free data in. | |
1354 | * size - requested block size | |
1355 | * tx - open transaction. | |
1356 | * | |
1357 | * NOTE: this function assumes that the znode is write locked. | |
1358 | */ | |
1359 | void | |
1360 | zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) | |
1361 | { | |
1362 | int error; | |
1363 | u_longlong_t dummy; | |
1364 | ||
1365 | if (size <= zp->z_blksz) | |
1366 | return; | |
1367 | /* | |
1368 | * If the file size is already greater than the current blocksize, | |
1369 | * we will not grow. If there is more than one block in a file, | |
1370 | * the blocksize cannot change. | |
1371 | */ | |
428870ff | 1372 | if (zp->z_blksz && zp->z_size > zp->z_blksz) |
34dc7c2f BB |
1373 | return; |
1374 | ||
3558fd73 | 1375 | error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id, |
34dc7c2f | 1376 | size, 0, tx); |
428870ff | 1377 | |
34dc7c2f BB |
1378 | if (error == ENOTSUP) |
1379 | return; | |
c99c9001 | 1380 | ASSERT0(error); |
34dc7c2f BB |
1381 | |
1382 | /* What blocksize did we actually get? */ | |
428870ff | 1383 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); |
34dc7c2f BB |
1384 | } |
1385 | ||
34dc7c2f | 1386 | /* |
b128c09f | 1387 | * Increase the file length |
34dc7c2f BB |
1388 | * |
1389 | * IN: zp - znode of file to free data in. | |
b128c09f | 1390 | * end - new end-of-file |
34dc7c2f | 1391 | * |
19d55079 | 1392 | * RETURN: 0 on success, error code on failure |
34dc7c2f | 1393 | */ |
b128c09f BB |
1394 | static int |
1395 | zfs_extend(znode_t *zp, uint64_t end) | |
34dc7c2f | 1396 | { |
3558fd73 | 1397 | zfs_sb_t *zsb = ZTOZSB(zp); |
b128c09f | 1398 | dmu_tx_t *tx; |
34dc7c2f | 1399 | rl_t *rl; |
b128c09f | 1400 | uint64_t newblksz; |
34dc7c2f BB |
1401 | int error; |
1402 | ||
34dc7c2f | 1403 | /* |
b128c09f | 1404 | * We will change zp_size, lock the whole file. |
34dc7c2f | 1405 | */ |
b128c09f | 1406 | rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); |
34dc7c2f BB |
1407 | |
1408 | /* | |
1409 | * Nothing to do if file already at desired length. | |
1410 | */ | |
428870ff | 1411 | if (end <= zp->z_size) { |
34dc7c2f BB |
1412 | zfs_range_unlock(rl); |
1413 | return (0); | |
1414 | } | |
3558fd73 | 1415 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
1416 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1417 | zfs_sa_upgrade_txholds(tx, zp); | |
b128c09f | 1418 | if (end > zp->z_blksz && |
3558fd73 | 1419 | (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) { |
34dc7c2f BB |
1420 | /* |
1421 | * We are growing the file past the current block size. | |
1422 | */ | |
3558fd73 | 1423 | if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) { |
f1512ee6 MA |
1424 | /* |
1425 | * File's blocksize is already larger than the | |
1426 | * "recordsize" property. Only let it grow to | |
1427 | * the next power of 2. | |
1428 | */ | |
34dc7c2f | 1429 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 | 1430 | newblksz = MIN(end, 1 << highbit64(zp->z_blksz)); |
34dc7c2f | 1431 | } else { |
3558fd73 | 1432 | newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz); |
34dc7c2f | 1433 | } |
b128c09f BB |
1434 | dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); |
1435 | } else { | |
1436 | newblksz = 0; | |
34dc7c2f BB |
1437 | } |
1438 | ||
384f8a09 | 1439 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 1440 | if (error) { |
34dc7c2f BB |
1441 | dmu_tx_abort(tx); |
1442 | zfs_range_unlock(rl); | |
1443 | return (error); | |
1444 | } | |
1445 | ||
b128c09f BB |
1446 | if (newblksz) |
1447 | zfs_grow_blocksize(zp, newblksz, tx); | |
34dc7c2f | 1448 | |
428870ff BB |
1449 | zp->z_size = end; |
1450 | ||
3558fd73 | 1451 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), |
428870ff | 1452 | &zp->z_size, sizeof (zp->z_size), tx)); |
34dc7c2f | 1453 | |
b128c09f | 1454 | zfs_range_unlock(rl); |
34dc7c2f | 1455 | |
b128c09f | 1456 | dmu_tx_commit(tx); |
34dc7c2f | 1457 | |
b128c09f BB |
1458 | return (0); |
1459 | } | |
1460 | ||
223df016 TC |
1461 | /* |
1462 | * zfs_zero_partial_page - Modeled after update_pages() but | |
1463 | * with different arguments and semantics for use by zfs_freesp(). | |
1464 | * | |
1465 | * Zeroes a piece of a single page cache entry for zp at offset | |
1466 | * start and length len. | |
1467 | * | |
1468 | * Caller must acquire a range lock on the file for the region | |
1469 | * being zeroed in order that the ARC and page cache stay in sync. | |
1470 | */ | |
1471 | static void | |
1472 | zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len) | |
1473 | { | |
1474 | struct address_space *mp = ZTOI(zp)->i_mapping; | |
1475 | struct page *pp; | |
1476 | int64_t off; | |
1477 | void *pb; | |
1478 | ||
8b1899d3 | 1479 | ASSERT((start & PAGE_MASK) == ((start + len - 1) & PAGE_MASK)); |
223df016 | 1480 | |
8b1899d3 BB |
1481 | off = start & (PAGE_SIZE - 1); |
1482 | start &= PAGE_MASK; | |
223df016 | 1483 | |
8b1899d3 | 1484 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
223df016 TC |
1485 | if (pp) { |
1486 | if (mapping_writably_mapped(mp)) | |
1487 | flush_dcache_page(pp); | |
1488 | ||
1489 | pb = kmap(pp); | |
1490 | bzero(pb + off, len); | |
1491 | kunmap(pp); | |
1492 | ||
1493 | if (mapping_writably_mapped(mp)) | |
1494 | flush_dcache_page(pp); | |
1495 | ||
1496 | mark_page_accessed(pp); | |
1497 | SetPageUptodate(pp); | |
1498 | ClearPageError(pp); | |
1499 | unlock_page(pp); | |
8b1899d3 | 1500 | put_page(pp); |
223df016 TC |
1501 | } |
1502 | } | |
1503 | ||
b128c09f BB |
1504 | /* |
1505 | * Free space in a file. | |
1506 | * | |
1507 | * IN: zp - znode of file to free data in. | |
1508 | * off - start of section to free. | |
1509 | * len - length of section to free. | |
1510 | * | |
19d55079 | 1511 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1512 | */ |
1513 | static int | |
1514 | zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) | |
1515 | { | |
3558fd73 | 1516 | zfs_sb_t *zsb = ZTOZSB(zp); |
b128c09f BB |
1517 | rl_t *rl; |
1518 | int error; | |
1519 | ||
1520 | /* | |
1521 | * Lock the range being freed. | |
1522 | */ | |
1523 | rl = zfs_range_lock(zp, off, len, RL_WRITER); | |
1524 | ||
1525 | /* | |
1526 | * Nothing to do if file already at desired length. | |
1527 | */ | |
428870ff | 1528 | if (off >= zp->z_size) { |
b128c09f BB |
1529 | zfs_range_unlock(rl); |
1530 | return (0); | |
34dc7c2f BB |
1531 | } |
1532 | ||
428870ff BB |
1533 | if (off + len > zp->z_size) |
1534 | len = zp->z_size - off; | |
b128c09f | 1535 | |
3558fd73 | 1536 | error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len); |
b128c09f | 1537 | |
223df016 TC |
1538 | /* |
1539 | * Zero partial page cache entries. This must be done under a | |
1540 | * range lock in order to keep the ARC and page cache in sync. | |
1541 | */ | |
1542 | if (zp->z_is_mapped) { | |
1543 | loff_t first_page, last_page, page_len; | |
1544 | loff_t first_page_offset, last_page_offset; | |
1545 | ||
1546 | /* first possible full page in hole */ | |
8b1899d3 | 1547 | first_page = (off + PAGE_SIZE - 1) >> PAGE_SHIFT; |
223df016 | 1548 | /* last page of hole */ |
8b1899d3 | 1549 | last_page = (off + len) >> PAGE_SHIFT; |
223df016 TC |
1550 | |
1551 | /* offset of first_page */ | |
8b1899d3 | 1552 | first_page_offset = first_page << PAGE_SHIFT; |
223df016 | 1553 | /* offset of last_page */ |
8b1899d3 | 1554 | last_page_offset = last_page << PAGE_SHIFT; |
223df016 | 1555 | |
cb08f063 TC |
1556 | /* truncate whole pages */ |
1557 | if (last_page_offset > first_page_offset) { | |
1558 | truncate_inode_pages_range(ZTOI(zp)->i_mapping, | |
1559 | first_page_offset, last_page_offset - 1); | |
1560 | } | |
1561 | ||
1562 | /* truncate sub-page ranges */ | |
223df016 TC |
1563 | if (first_page > last_page) { |
1564 | /* entire punched area within a single page */ | |
1565 | zfs_zero_partial_page(zp, off, len); | |
1566 | } else { | |
1567 | /* beginning of punched area at the end of a page */ | |
1568 | page_len = first_page_offset - off; | |
1569 | if (page_len > 0) | |
1570 | zfs_zero_partial_page(zp, off, page_len); | |
1571 | ||
1572 | /* end of punched area at the beginning of a page */ | |
1573 | page_len = off + len - last_page_offset; | |
1574 | if (page_len > 0) | |
1575 | zfs_zero_partial_page(zp, last_page_offset, | |
1576 | page_len); | |
1577 | } | |
1578 | } | |
34dc7c2f BB |
1579 | zfs_range_unlock(rl); |
1580 | ||
b128c09f BB |
1581 | return (error); |
1582 | } | |
1583 | ||
1584 | /* | |
1585 | * Truncate a file | |
1586 | * | |
1587 | * IN: zp - znode of file to free data in. | |
1588 | * end - new end-of-file. | |
1589 | * | |
19d55079 | 1590 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1591 | */ |
1592 | static int | |
1593 | zfs_trunc(znode_t *zp, uint64_t end) | |
1594 | { | |
3558fd73 | 1595 | zfs_sb_t *zsb = ZTOZSB(zp); |
b128c09f BB |
1596 | dmu_tx_t *tx; |
1597 | rl_t *rl; | |
1598 | int error; | |
572e2857 BB |
1599 | sa_bulk_attr_t bulk[2]; |
1600 | int count = 0; | |
b128c09f BB |
1601 | |
1602 | /* | |
1603 | * We will change zp_size, lock the whole file. | |
1604 | */ | |
1605 | rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); | |
1606 | ||
1607 | /* | |
1608 | * Nothing to do if file already at desired length. | |
1609 | */ | |
428870ff | 1610 | if (end >= zp->z_size) { |
b128c09f BB |
1611 | zfs_range_unlock(rl); |
1612 | return (0); | |
1613 | } | |
1614 | ||
3558fd73 | 1615 | error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1); |
b128c09f BB |
1616 | if (error) { |
1617 | zfs_range_unlock(rl); | |
1618 | return (error); | |
1619 | } | |
3558fd73 | 1620 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
1621 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1622 | zfs_sa_upgrade_txholds(tx, zp); | |
19d55079 | 1623 | dmu_tx_mark_netfree(tx); |
7a8f0e80 | 1624 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1625 | if (error) { |
b128c09f BB |
1626 | dmu_tx_abort(tx); |
1627 | zfs_range_unlock(rl); | |
1628 | return (error); | |
1629 | } | |
b128c09f | 1630 | |
428870ff | 1631 | zp->z_size = end; |
3558fd73 | 1632 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), |
572e2857 | 1633 | NULL, &zp->z_size, sizeof (zp->z_size)); |
428870ff | 1634 | |
572e2857 BB |
1635 | if (end == 0) { |
1636 | zp->z_pflags &= ~ZFS_SPARSE; | |
3558fd73 | 1637 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), |
572e2857 BB |
1638 | NULL, &zp->z_pflags, 8); |
1639 | } | |
1640 | VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); | |
b128c09f | 1641 | |
34dc7c2f BB |
1642 | dmu_tx_commit(tx); |
1643 | ||
d164b209 | 1644 | zfs_range_unlock(rl); |
34dc7c2f BB |
1645 | |
1646 | return (0); | |
1647 | } | |
1648 | ||
b128c09f BB |
1649 | /* |
1650 | * Free space in a file | |
1651 | * | |
1652 | * IN: zp - znode of file to free data in. | |
1653 | * off - start of range | |
1654 | * len - end of range (0 => EOF) | |
1655 | * flag - current file open mode flags. | |
1656 | * log - TRUE if this action should be logged | |
1657 | * | |
19d55079 | 1658 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1659 | */ |
1660 | int | |
1661 | zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) | |
1662 | { | |
b128c09f | 1663 | dmu_tx_t *tx; |
3558fd73 BB |
1664 | zfs_sb_t *zsb = ZTOZSB(zp); |
1665 | zilog_t *zilog = zsb->z_log; | |
428870ff BB |
1666 | uint64_t mode; |
1667 | uint64_t mtime[2], ctime[2]; | |
1668 | sa_bulk_attr_t bulk[3]; | |
1669 | int count = 0; | |
b128c09f BB |
1670 | int error; |
1671 | ||
3558fd73 | 1672 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode, |
428870ff BB |
1673 | sizeof (mode))) != 0) |
1674 | return (error); | |
1675 | ||
1676 | if (off > zp->z_size) { | |
b128c09f BB |
1677 | error = zfs_extend(zp, off+len); |
1678 | if (error == 0 && log) | |
1679 | goto log; | |
223df016 | 1680 | goto out; |
b128c09f BB |
1681 | } |
1682 | ||
b128c09f BB |
1683 | if (len == 0) { |
1684 | error = zfs_trunc(zp, off); | |
1685 | } else { | |
1686 | if ((error = zfs_free_range(zp, off, len)) == 0 && | |
428870ff | 1687 | off + len > zp->z_size) |
b128c09f BB |
1688 | error = zfs_extend(zp, off+len); |
1689 | } | |
1690 | if (error || !log) | |
223df016 | 1691 | goto out; |
b128c09f | 1692 | log: |
3558fd73 | 1693 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
1694 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1695 | zfs_sa_upgrade_txholds(tx, zp); | |
384f8a09 | 1696 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1697 | if (error) { |
b128c09f | 1698 | dmu_tx_abort(tx); |
223df016 | 1699 | goto out; |
b128c09f BB |
1700 | } |
1701 | ||
3558fd73 BB |
1702 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16); |
1703 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16); | |
1704 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), | |
428870ff | 1705 | NULL, &zp->z_pflags, 8); |
0df9673f | 1706 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
428870ff BB |
1707 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
1708 | ASSERT(error == 0); | |
1709 | ||
b128c09f BB |
1710 | zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); |
1711 | ||
1712 | dmu_tx_commit(tx); | |
223df016 | 1713 | |
960e08fe | 1714 | zfs_inode_update(zp); |
223df016 TC |
1715 | error = 0; |
1716 | ||
1717 | out: | |
1718 | /* | |
1719 | * Truncate the page cache - for file truncate operations, use | |
1720 | * the purpose-built API for truncations. For punching operations, | |
cb08f063 | 1721 | * the truncation is handled under a range lock in zfs_free_range. |
223df016 TC |
1722 | */ |
1723 | if (len == 0) | |
1724 | truncate_setsize(ZTOI(zp), off); | |
223df016 | 1725 | return (error); |
b128c09f BB |
1726 | } |
1727 | ||
34dc7c2f BB |
1728 | void |
1729 | zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) | |
1730 | { | |
22872ff5 BB |
1731 | struct super_block *sb; |
1732 | zfs_sb_t *zsb; | |
428870ff | 1733 | uint64_t moid, obj, sa_obj, version; |
22872ff5 | 1734 | uint64_t sense = ZFS_CASE_SENSITIVE; |
34dc7c2f BB |
1735 | uint64_t norm = 0; |
1736 | nvpair_t *elem; | |
c96c36fa | 1737 | int size; |
34dc7c2f | 1738 | int error; |
22872ff5 BB |
1739 | int i; |
1740 | znode_t *rootzp = NULL; | |
1741 | vattr_t vattr; | |
1742 | znode_t *zp; | |
1743 | zfs_acl_ids_t acl_ids; | |
34dc7c2f BB |
1744 | |
1745 | /* | |
1746 | * First attempt to create master node. | |
1747 | */ | |
1748 | /* | |
1749 | * In an empty objset, there are no blocks to read and thus | |
1750 | * there can be no i/o errors (which we assert below). | |
1751 | */ | |
1752 | moid = MASTER_NODE_OBJ; | |
1753 | error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, | |
1754 | DMU_OT_NONE, 0, tx); | |
1755 | ASSERT(error == 0); | |
1756 | ||
1757 | /* | |
1758 | * Set starting attributes. | |
1759 | */ | |
428870ff | 1760 | version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); |
34dc7c2f BB |
1761 | elem = NULL; |
1762 | while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { | |
1763 | /* For the moment we expect all zpl props to be uint64_ts */ | |
1764 | uint64_t val; | |
1765 | char *name; | |
1766 | ||
1767 | ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); | |
1768 | VERIFY(nvpair_value_uint64(elem, &val) == 0); | |
1769 | name = nvpair_name(elem); | |
1770 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { | |
9babb374 BB |
1771 | if (val < version) |
1772 | version = val; | |
34dc7c2f BB |
1773 | } else { |
1774 | error = zap_update(os, moid, name, 8, 1, &val, tx); | |
1775 | } | |
1776 | ASSERT(error == 0); | |
1777 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) | |
1778 | norm = val; | |
22872ff5 BB |
1779 | else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) |
1780 | sense = val; | |
34dc7c2f BB |
1781 | } |
1782 | ASSERT(version != 0); | |
9babb374 | 1783 | error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); |
34dc7c2f | 1784 | |
428870ff BB |
1785 | /* |
1786 | * Create zap object used for SA attribute registration | |
1787 | */ | |
1788 | ||
1789 | if (version >= ZPL_VERSION_SA) { | |
1790 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1791 | DMU_OT_NONE, 0, tx); | |
1792 | error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
1793 | ASSERT(error == 0); | |
1794 | } else { | |
1795 | sa_obj = 0; | |
1796 | } | |
34dc7c2f BB |
1797 | /* |
1798 | * Create a delete queue. | |
1799 | */ | |
9babb374 | 1800 | obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); |
34dc7c2f | 1801 | |
9babb374 | 1802 | error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); |
34dc7c2f BB |
1803 | ASSERT(error == 0); |
1804 | ||
9babb374 | 1805 | /* |
22872ff5 BB |
1806 | * Create root znode. Create minimal znode/inode/zsb/sb |
1807 | * to allow zfs_mknode to work. | |
9babb374 | 1808 | */ |
22872ff5 BB |
1809 | vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID; |
1810 | vattr.va_mode = S_IFDIR|0755; | |
1811 | vattr.va_uid = crgetuid(cr); | |
1812 | vattr.va_gid = crgetgid(cr); | |
1813 | ||
79c76d5b | 1814 | rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
22872ff5 BB |
1815 | rootzp->z_moved = 0; |
1816 | rootzp->z_unlinked = 0; | |
1817 | rootzp->z_atime_dirty = 0; | |
1818 | rootzp->z_is_sa = USE_SA(version, os); | |
1819 | ||
79c76d5b | 1820 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
22872ff5 BB |
1821 | zsb->z_os = os; |
1822 | zsb->z_parent = zsb; | |
1823 | zsb->z_version = version; | |
1824 | zsb->z_use_fuids = USE_FUIDS(version, os); | |
1825 | zsb->z_use_sa = USE_SA(version, os); | |
1826 | zsb->z_norm = norm; | |
1827 | ||
79c76d5b | 1828 | sb = kmem_zalloc(sizeof (struct super_block), KM_SLEEP); |
22872ff5 BB |
1829 | sb->s_fs_info = zsb; |
1830 | ||
1831 | ZTOI(rootzp)->i_sb = sb; | |
1832 | ||
1833 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, | |
1834 | &zsb->z_attr_table); | |
9babb374 | 1835 | |
22872ff5 | 1836 | ASSERT(error == 0); |
9babb374 | 1837 | |
60101509 | 1838 | /* |
22872ff5 BB |
1839 | * Fold case on file systems that are always or sometimes case |
1840 | * insensitive. | |
60101509 | 1841 | */ |
22872ff5 BB |
1842 | if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) |
1843 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
60101509 | 1844 | |
22872ff5 BB |
1845 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
1846 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
1847 | offsetof(znode_t, z_link_node)); | |
60101509 | 1848 | |
c96c36fa BB |
1849 | size = MIN(1 << (highbit64(zfs_object_mutex_size)-1), ZFS_OBJ_MTX_MAX); |
1850 | zsb->z_hold_size = size; | |
1851 | zsb->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, KM_SLEEP); | |
1852 | zsb->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
1853 | for (i = 0; i != size; i++) { | |
1854 | avl_create(&zsb->z_hold_trees[i], zfs_znode_hold_compare, | |
1855 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); | |
1856 | mutex_init(&zsb->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); | |
1857 | } | |
60101509 | 1858 | |
22872ff5 BB |
1859 | VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, |
1860 | cr, NULL, &acl_ids)); | |
1861 | zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); | |
1862 | ASSERT3P(zp, ==, rootzp); | |
1863 | error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); | |
1864 | ASSERT(error == 0); | |
1865 | zfs_acl_ids_free(&acl_ids); | |
60101509 | 1866 | |
22872ff5 BB |
1867 | atomic_set(&ZTOI(rootzp)->i_count, 0); |
1868 | sa_handle_destroy(rootzp->z_sa_hdl); | |
22872ff5 BB |
1869 | kmem_cache_free(znode_cache, rootzp); |
1870 | ||
1871 | /* | |
1872 | * Create shares directory | |
1873 | */ | |
22872ff5 | 1874 | error = zfs_create_share_dir(zsb, tx); |
9babb374 | 1875 | ASSERT(error == 0); |
428870ff | 1876 | |
c96c36fa BB |
1877 | for (i = 0; i != size; i++) { |
1878 | avl_destroy(&zsb->z_hold_trees[i]); | |
1879 | mutex_destroy(&zsb->z_hold_locks[i]); | |
1880 | } | |
2708f716 | 1881 | |
c96c36fa BB |
1882 | vmem_free(zsb->z_hold_trees, sizeof (avl_tree_t) * size); |
1883 | vmem_free(zsb->z_hold_locks, sizeof (kmutex_t) * size); | |
2708f716 BB |
1884 | kmem_free(sb, sizeof (struct super_block)); |
1885 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
34dc7c2f | 1886 | } |
34dc7c2f | 1887 | #endif /* _KERNEL */ |
428870ff | 1888 | |
34dc7c2f | 1889 | static int |
572e2857 BB |
1890 | zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) |
1891 | { | |
1892 | uint64_t sa_obj = 0; | |
1893 | int error; | |
1894 | ||
1895 | error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
1896 | if (error != 0 && error != ENOENT) | |
1897 | return (error); | |
1898 | ||
1899 | error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); | |
1900 | return (error); | |
1901 | } | |
1902 | ||
1903 | static int | |
1904 | zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, | |
7b8518cb | 1905 | dmu_buf_t **db, void *tag) |
34dc7c2f | 1906 | { |
34dc7c2f | 1907 | dmu_object_info_t doi; |
34dc7c2f | 1908 | int error; |
428870ff | 1909 | |
7b8518cb | 1910 | if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) |
34dc7c2f BB |
1911 | return (error); |
1912 | ||
572e2857 | 1913 | dmu_object_info_from_db(*db, &doi); |
428870ff BB |
1914 | if ((doi.doi_bonus_type != DMU_OT_SA && |
1915 | doi.doi_bonus_type != DMU_OT_ZNODE) || | |
d6320ddb BB |
1916 | (doi.doi_bonus_type == DMU_OT_ZNODE && |
1917 | doi.doi_bonus_size < sizeof (znode_phys_t))) { | |
7b8518cb | 1918 | sa_buf_rele(*db, tag); |
2e528b49 | 1919 | return (SET_ERROR(ENOTSUP)); |
34dc7c2f BB |
1920 | } |
1921 | ||
572e2857 BB |
1922 | error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); |
1923 | if (error != 0) { | |
7b8518cb | 1924 | sa_buf_rele(*db, tag); |
428870ff BB |
1925 | return (error); |
1926 | } | |
1927 | ||
572e2857 BB |
1928 | return (0); |
1929 | } | |
1930 | ||
1931 | void | |
7b8518cb | 1932 | zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) |
572e2857 BB |
1933 | { |
1934 | sa_handle_destroy(hdl); | |
7b8518cb | 1935 | sa_buf_rele(db, tag); |
572e2857 BB |
1936 | } |
1937 | ||
1938 | /* | |
1939 | * Given an object number, return its parent object number and whether | |
1940 | * or not the object is an extended attribute directory. | |
1941 | */ | |
1942 | static int | |
b23ad7f3 JJ |
1943 | zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, |
1944 | uint64_t *pobjp, int *is_xattrdir) | |
572e2857 BB |
1945 | { |
1946 | uint64_t parent; | |
1947 | uint64_t pflags; | |
1948 | uint64_t mode; | |
b23ad7f3 | 1949 | uint64_t parent_mode; |
572e2857 | 1950 | sa_bulk_attr_t bulk[3]; |
b23ad7f3 JJ |
1951 | sa_handle_t *sa_hdl; |
1952 | dmu_buf_t *sa_db; | |
572e2857 BB |
1953 | int count = 0; |
1954 | int error; | |
1955 | ||
1956 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, | |
1957 | &parent, sizeof (parent)); | |
428870ff | 1958 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, |
572e2857 | 1959 | &pflags, sizeof (pflags)); |
428870ff | 1960 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, |
572e2857 | 1961 | &mode, sizeof (mode)); |
428870ff | 1962 | |
572e2857 | 1963 | if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) |
428870ff | 1964 | return (error); |
572e2857 | 1965 | |
b23ad7f3 JJ |
1966 | /* |
1967 | * When a link is removed its parent pointer is not changed and will | |
1968 | * be invalid. There are two cases where a link is removed but the | |
1969 | * file stays around, when it goes to the delete queue and when there | |
1970 | * are additional links. | |
1971 | */ | |
1972 | error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); | |
1973 | if (error != 0) | |
1974 | return (error); | |
1975 | ||
1976 | error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); | |
1977 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); | |
1978 | if (error != 0) | |
1979 | return (error); | |
1980 | ||
428870ff | 1981 | *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); |
34dc7c2f | 1982 | |
b23ad7f3 JJ |
1983 | /* |
1984 | * Extended attributes can be applied to files, directories, etc. | |
1985 | * Otherwise the parent must be a directory. | |
1986 | */ | |
1987 | if (!*is_xattrdir && !S_ISDIR(parent_mode)) | |
1988 | return (EINVAL); | |
1989 | ||
1990 | *pobjp = parent; | |
1991 | ||
34dc7c2f BB |
1992 | return (0); |
1993 | } | |
1994 | ||
572e2857 BB |
1995 | /* |
1996 | * Given an object number, return some zpl level statistics | |
1997 | */ | |
1998 | static int | |
1999 | zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, | |
2000 | zfs_stat_t *sb) | |
34dc7c2f | 2001 | { |
572e2857 BB |
2002 | sa_bulk_attr_t bulk[4]; |
2003 | int count = 0; | |
2004 | ||
2005 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, | |
2006 | &sb->zs_mode, sizeof (sb->zs_mode)); | |
2007 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, | |
2008 | &sb->zs_gen, sizeof (sb->zs_gen)); | |
2009 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, | |
2010 | &sb->zs_links, sizeof (sb->zs_links)); | |
2011 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, | |
2012 | &sb->zs_ctime, sizeof (sb->zs_ctime)); | |
2013 | ||
2014 | return (sa_bulk_lookup(hdl, bulk, count)); | |
2015 | } | |
2016 | ||
2017 | static int | |
2018 | zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, | |
2019 | sa_attr_type_t *sa_table, char *buf, int len) | |
2020 | { | |
2021 | sa_handle_t *sa_hdl; | |
2022 | sa_handle_t *prevhdl = NULL; | |
2023 | dmu_buf_t *prevdb = NULL; | |
2024 | dmu_buf_t *sa_db = NULL; | |
34dc7c2f BB |
2025 | char *path = buf + len - 1; |
2026 | int error; | |
2027 | ||
2028 | *path = '\0'; | |
572e2857 | 2029 | sa_hdl = hdl; |
428870ff | 2030 | |
34dc7c2f | 2031 | for (;;) { |
17897ce2 | 2032 | uint64_t pobj = 0; |
34dc7c2f BB |
2033 | char component[MAXNAMELEN + 2]; |
2034 | size_t complen; | |
17897ce2 | 2035 | int is_xattrdir = 0; |
34dc7c2f | 2036 | |
572e2857 | 2037 | if (prevdb) |
7b8518cb | 2038 | zfs_release_sa_handle(prevhdl, prevdb, FTAG); |
572e2857 | 2039 | |
b23ad7f3 | 2040 | if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, |
572e2857 | 2041 | &is_xattrdir)) != 0) |
34dc7c2f BB |
2042 | break; |
2043 | ||
2044 | if (pobj == obj) { | |
2045 | if (path[0] != '/') | |
2046 | *--path = '/'; | |
2047 | break; | |
2048 | } | |
2049 | ||
2050 | component[0] = '/'; | |
2051 | if (is_xattrdir) { | |
2052 | (void) sprintf(component + 1, "<xattrdir>"); | |
2053 | } else { | |
2054 | error = zap_value_search(osp, pobj, obj, | |
2055 | ZFS_DIRENT_OBJ(-1ULL), component + 1); | |
2056 | if (error != 0) | |
2057 | break; | |
2058 | } | |
2059 | ||
2060 | complen = strlen(component); | |
2061 | path -= complen; | |
2062 | ASSERT(path >= buf); | |
2063 | bcopy(component, path, complen); | |
2064 | obj = pobj; | |
572e2857 BB |
2065 | |
2066 | if (sa_hdl != hdl) { | |
2067 | prevhdl = sa_hdl; | |
2068 | prevdb = sa_db; | |
2069 | } | |
7b8518cb | 2070 | error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); |
572e2857 BB |
2071 | if (error != 0) { |
2072 | sa_hdl = prevhdl; | |
2073 | sa_db = prevdb; | |
2074 | break; | |
2075 | } | |
2076 | } | |
2077 | ||
2078 | if (sa_hdl != NULL && sa_hdl != hdl) { | |
2079 | ASSERT(sa_db != NULL); | |
7b8518cb | 2080 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); |
34dc7c2f BB |
2081 | } |
2082 | ||
2083 | if (error == 0) | |
2084 | (void) memmove(buf, path, buf + len - path); | |
428870ff | 2085 | |
34dc7c2f BB |
2086 | return (error); |
2087 | } | |
572e2857 BB |
2088 | |
2089 | int | |
2090 | zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) | |
2091 | { | |
2092 | sa_attr_type_t *sa_table; | |
2093 | sa_handle_t *hdl; | |
2094 | dmu_buf_t *db; | |
2095 | int error; | |
2096 | ||
2097 | error = zfs_sa_setup(osp, &sa_table); | |
2098 | if (error != 0) | |
2099 | return (error); | |
2100 | ||
7b8518cb | 2101 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2102 | if (error != 0) |
2103 | return (error); | |
2104 | ||
2105 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2106 | ||
7b8518cb | 2107 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2108 | return (error); |
2109 | } | |
2110 | ||
2111 | int | |
2112 | zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, | |
2113 | char *buf, int len) | |
2114 | { | |
2115 | char *path = buf + len - 1; | |
2116 | sa_attr_type_t *sa_table; | |
2117 | sa_handle_t *hdl; | |
2118 | dmu_buf_t *db; | |
2119 | int error; | |
2120 | ||
2121 | *path = '\0'; | |
2122 | ||
2123 | error = zfs_sa_setup(osp, &sa_table); | |
2124 | if (error != 0) | |
2125 | return (error); | |
2126 | ||
7b8518cb | 2127 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2128 | if (error != 0) |
2129 | return (error); | |
2130 | ||
2131 | error = zfs_obj_to_stats_impl(hdl, sa_table, sb); | |
2132 | if (error != 0) { | |
7b8518cb | 2133 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2134 | return (error); |
2135 | } | |
2136 | ||
2137 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2138 | ||
7b8518cb | 2139 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2140 | return (error); |
2141 | } | |
c28b2279 BB |
2142 | |
2143 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
2144 | EXPORT_SYMBOL(zfs_create_fs); | |
2145 | EXPORT_SYMBOL(zfs_obj_to_path); | |
0720116d BB |
2146 | |
2147 | module_param(zfs_object_mutex_size, uint, 0644); | |
2148 | MODULE_PARM_DESC(zfs_object_mutex_size, "Size of znode hold array"); | |
c28b2279 | 2149 | #endif |