]>
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> |
50c957f7 | 65 | #include <sys/dmu_tx.h> |
34dc7c2f BB |
66 | #include <sys/refcount.h> |
67 | #include <sys/stat.h> | |
68 | #include <sys/zap.h> | |
69 | #include <sys/zfs_znode.h> | |
428870ff BB |
70 | #include <sys/sa.h> |
71 | #include <sys/zfs_sa.h> | |
572e2857 | 72 | #include <sys/zfs_stat.h> |
34dc7c2f BB |
73 | |
74 | #include "zfs_prop.h" | |
428870ff | 75 | #include "zfs_comutil.h" |
34dc7c2f | 76 | |
b128c09f BB |
77 | /* |
78 | * Define ZNODE_STATS to turn on statistic gathering. By default, it is only | |
79 | * turned on when DEBUG is also defined. | |
80 | */ | |
81 | #ifdef DEBUG | |
82 | #define ZNODE_STATS | |
83 | #endif /* DEBUG */ | |
84 | ||
85 | #ifdef ZNODE_STATS | |
86 | #define ZNODE_STAT_ADD(stat) ((stat)++) | |
87 | #else | |
88 | #define ZNODE_STAT_ADD(stat) /* nothing */ | |
89 | #endif /* ZNODE_STATS */ | |
90 | ||
34dc7c2f BB |
91 | /* |
92 | * Functions needed for userland (ie: libzpool) are not put under | |
93 | * #ifdef_KERNEL; the rest of the functions have dependencies | |
94 | * (such as VFS logic) that will not compile easily in userland. | |
95 | */ | |
96 | #ifdef _KERNEL | |
9babb374 | 97 | |
b128c09f | 98 | static kmem_cache_t *znode_cache = NULL; |
c96c36fa | 99 | static kmem_cache_t *znode_hold_cache = NULL; |
0720116d | 100 | unsigned int zfs_object_mutex_size = ZFS_OBJ_MTX_SZ; |
34dc7c2f | 101 | |
34dc7c2f BB |
102 | /*ARGSUSED*/ |
103 | static int | |
b128c09f | 104 | zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) |
34dc7c2f BB |
105 | { |
106 | znode_t *zp = buf; | |
107 | ||
3558fd73 | 108 | inode_init_once(ZTOI(zp)); |
b128c09f BB |
109 | list_link_init(&zp->z_link_node); |
110 | ||
34dc7c2f | 111 | mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f | 112 | rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); |
448d7aaa | 113 | rw_init(&zp->z_name_lock, NULL, RW_NOLOCKDEP, NULL); |
34dc7c2f | 114 | mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); |
82a37189 | 115 | rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); |
34dc7c2f | 116 | |
d88895a0 | 117 | zfs_rlock_init(&zp->z_range_lock); |
34dc7c2f | 118 | |
b128c09f | 119 | zp->z_dirlocks = NULL; |
45d1cae3 | 120 | zp->z_acl_cached = NULL; |
82a37189 | 121 | zp->z_xattr_cached = NULL; |
572e2857 | 122 | zp->z_moved = 0; |
34dc7c2f BB |
123 | return (0); |
124 | } | |
125 | ||
126 | /*ARGSUSED*/ | |
127 | static void | |
b128c09f | 128 | zfs_znode_cache_destructor(void *buf, void *arg) |
34dc7c2f BB |
129 | { |
130 | znode_t *zp = buf; | |
131 | ||
b128c09f | 132 | ASSERT(!list_link_active(&zp->z_link_node)); |
34dc7c2f | 133 | mutex_destroy(&zp->z_lock); |
34dc7c2f BB |
134 | rw_destroy(&zp->z_parent_lock); |
135 | rw_destroy(&zp->z_name_lock); | |
136 | mutex_destroy(&zp->z_acl_lock); | |
82a37189 | 137 | rw_destroy(&zp->z_xattr_lock); |
d88895a0 | 138 | zfs_rlock_destroy(&zp->z_range_lock); |
34dc7c2f | 139 | |
b128c09f | 140 | ASSERT(zp->z_dirlocks == NULL); |
45d1cae3 | 141 | ASSERT(zp->z_acl_cached == NULL); |
82a37189 | 142 | ASSERT(zp->z_xattr_cached == NULL); |
b128c09f BB |
143 | } |
144 | ||
c96c36fa BB |
145 | static int |
146 | zfs_znode_hold_cache_constructor(void *buf, void *arg, int kmflags) | |
147 | { | |
148 | znode_hold_t *zh = buf; | |
149 | ||
150 | mutex_init(&zh->zh_lock, NULL, MUTEX_DEFAULT, NULL); | |
151 | refcount_create(&zh->zh_refcount); | |
152 | zh->zh_obj = ZFS_NO_OBJECT; | |
153 | ||
154 | return (0); | |
155 | } | |
156 | ||
157 | static void | |
158 | zfs_znode_hold_cache_destructor(void *buf, void *arg) | |
159 | { | |
160 | znode_hold_t *zh = buf; | |
161 | ||
162 | mutex_destroy(&zh->zh_lock); | |
163 | refcount_destroy(&zh->zh_refcount); | |
164 | } | |
165 | ||
34dc7c2f BB |
166 | void |
167 | zfs_znode_init(void) | |
168 | { | |
169 | /* | |
5074bfe8 TC |
170 | * Initialize zcache. The KMC_SLAB hint is used in order that it be |
171 | * backed by kmalloc() when on the Linux slab in order that any | |
172 | * wait_on_bit() operations on the related inode operate properly. | |
34dc7c2f BB |
173 | */ |
174 | ASSERT(znode_cache == NULL); | |
175 | znode_cache = kmem_cache_create("zfs_znode_cache", | |
176 | sizeof (znode_t), 0, zfs_znode_cache_constructor, | |
5074bfe8 | 177 | zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_SLAB); |
c96c36fa BB |
178 | |
179 | ASSERT(znode_hold_cache == NULL); | |
180 | znode_hold_cache = kmem_cache_create("zfs_znode_hold_cache", | |
181 | sizeof (znode_hold_t), 0, zfs_znode_hold_cache_constructor, | |
182 | zfs_znode_hold_cache_destructor, NULL, NULL, NULL, 0); | |
34dc7c2f BB |
183 | } |
184 | ||
185 | void | |
186 | zfs_znode_fini(void) | |
187 | { | |
34dc7c2f BB |
188 | /* |
189 | * Cleanup zcache | |
190 | */ | |
191 | if (znode_cache) | |
192 | kmem_cache_destroy(znode_cache); | |
193 | znode_cache = NULL; | |
c96c36fa BB |
194 | |
195 | if (znode_hold_cache) | |
196 | kmem_cache_destroy(znode_hold_cache); | |
197 | znode_hold_cache = NULL; | |
198 | } | |
199 | ||
200 | /* | |
201 | * The zfs_znode_hold_enter() / zfs_znode_hold_exit() functions are used to | |
202 | * serialize access to a znode and its SA buffer while the object is being | |
203 | * created or destroyed. This kind of locking would normally reside in the | |
204 | * znode itself but in this case that's impossible because the znode and SA | |
205 | * buffer may not yet exist. Therefore the locking is handled externally | |
206 | * with an array of mutexs and AVLs trees which contain per-object locks. | |
207 | * | |
208 | * In zfs_znode_hold_enter() a per-object lock is created as needed, inserted | |
209 | * in to the correct AVL tree and finally the per-object lock is held. In | |
210 | * zfs_znode_hold_exit() the process is reversed. The per-object lock is | |
211 | * released, removed from the AVL tree and destroyed if there are no waiters. | |
212 | * | |
213 | * This scheme has two important properties: | |
214 | * | |
215 | * 1) No memory allocations are performed while holding one of the z_hold_locks. | |
216 | * This ensures evict(), which can be called from direct memory reclaim, will | |
217 | * never block waiting on a z_hold_locks which just happens to have hashed | |
218 | * to the same index. | |
219 | * | |
220 | * 2) All locks used to serialize access to an object are per-object and never | |
221 | * shared. This minimizes lock contention without creating a large number | |
222 | * of dedicated locks. | |
223 | * | |
224 | * On the downside it does require znode_lock_t structures to be frequently | |
225 | * allocated and freed. However, because these are backed by a kmem cache | |
226 | * and very short lived this cost is minimal. | |
227 | */ | |
228 | int | |
229 | zfs_znode_hold_compare(const void *a, const void *b) | |
230 | { | |
231 | const znode_hold_t *zh_a = a; | |
232 | const znode_hold_t *zh_b = b; | |
233 | ||
234 | if (zh_a->zh_obj < zh_b->zh_obj) | |
235 | return (-1); | |
236 | else if (zh_a->zh_obj > zh_b->zh_obj) | |
237 | return (1); | |
238 | else | |
239 | return (0); | |
240 | } | |
241 | ||
242 | boolean_t | |
243 | zfs_znode_held(zfs_sb_t *zsb, uint64_t obj) | |
244 | { | |
245 | znode_hold_t *zh, search; | |
246 | int i = ZFS_OBJ_HASH(zsb, obj); | |
37c56346 | 247 | boolean_t held; |
c96c36fa BB |
248 | |
249 | search.zh_obj = obj; | |
250 | ||
251 | mutex_enter(&zsb->z_hold_locks[i]); | |
252 | zh = avl_find(&zsb->z_hold_trees[i], &search, NULL); | |
37c56346 | 253 | held = (zh && MUTEX_HELD(&zh->zh_lock)) ? B_TRUE : B_FALSE; |
c96c36fa BB |
254 | mutex_exit(&zsb->z_hold_locks[i]); |
255 | ||
37c56346 | 256 | return (held); |
c96c36fa BB |
257 | } |
258 | ||
259 | static znode_hold_t * | |
260 | zfs_znode_hold_enter(zfs_sb_t *zsb, uint64_t obj) | |
261 | { | |
262 | znode_hold_t *zh, *zh_new, search; | |
263 | int i = ZFS_OBJ_HASH(zsb, obj); | |
264 | boolean_t found = B_FALSE; | |
265 | ||
266 | zh_new = kmem_cache_alloc(znode_hold_cache, KM_SLEEP); | |
267 | zh_new->zh_obj = obj; | |
268 | search.zh_obj = obj; | |
269 | ||
270 | mutex_enter(&zsb->z_hold_locks[i]); | |
271 | zh = avl_find(&zsb->z_hold_trees[i], &search, NULL); | |
272 | if (likely(zh == NULL)) { | |
273 | zh = zh_new; | |
274 | avl_add(&zsb->z_hold_trees[i], zh); | |
275 | } else { | |
276 | ASSERT3U(zh->zh_obj, ==, obj); | |
277 | found = B_TRUE; | |
278 | } | |
279 | refcount_add(&zh->zh_refcount, NULL); | |
280 | mutex_exit(&zsb->z_hold_locks[i]); | |
281 | ||
282 | if (found == B_TRUE) | |
283 | kmem_cache_free(znode_hold_cache, zh_new); | |
284 | ||
285 | ASSERT(MUTEX_NOT_HELD(&zh->zh_lock)); | |
286 | ASSERT3S(refcount_count(&zh->zh_refcount), >, 0); | |
287 | mutex_enter(&zh->zh_lock); | |
288 | ||
289 | return (zh); | |
290 | } | |
291 | ||
292 | static void | |
293 | zfs_znode_hold_exit(zfs_sb_t *zsb, znode_hold_t *zh) | |
294 | { | |
295 | int i = ZFS_OBJ_HASH(zsb, zh->zh_obj); | |
296 | boolean_t remove = B_FALSE; | |
297 | ||
298 | ASSERT(zfs_znode_held(zsb, zh->zh_obj)); | |
299 | ASSERT3S(refcount_count(&zh->zh_refcount), >, 0); | |
300 | mutex_exit(&zh->zh_lock); | |
301 | ||
302 | mutex_enter(&zsb->z_hold_locks[i]); | |
303 | if (refcount_remove(&zh->zh_refcount, NULL) == 0) { | |
304 | avl_remove(&zsb->z_hold_trees[i], zh); | |
305 | remove = B_TRUE; | |
306 | } | |
307 | mutex_exit(&zsb->z_hold_locks[i]); | |
308 | ||
309 | if (remove == B_TRUE) | |
310 | kmem_cache_free(znode_hold_cache, zh); | |
34dc7c2f BB |
311 | } |
312 | ||
34dc7c2f | 313 | int |
3558fd73 | 314 | zfs_create_share_dir(zfs_sb_t *zsb, dmu_tx_t *tx) |
34dc7c2f | 315 | { |
3c9609b3 | 316 | #ifdef HAVE_SMB_SHARE |
9babb374 BB |
317 | zfs_acl_ids_t acl_ids; |
318 | vattr_t vattr; | |
319 | znode_t *sharezp; | |
320 | vnode_t *vp; | |
321 | znode_t *zp; | |
322 | int error; | |
34dc7c2f | 323 | |
9babb374 | 324 | vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; |
3558fd73 | 325 | vattr.va_mode = S_IFDIR | 0555; |
9babb374 BB |
326 | vattr.va_uid = crgetuid(kcred); |
327 | vattr.va_gid = crgetgid(kcred); | |
34dc7c2f | 328 | |
79c76d5b | 329 | sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
572e2857 | 330 | sharezp->z_moved = 0; |
9babb374 BB |
331 | sharezp->z_unlinked = 0; |
332 | sharezp->z_atime_dirty = 0; | |
333 | sharezp->z_zfsvfs = zfsvfs; | |
428870ff | 334 | sharezp->z_is_sa = zfsvfs->z_use_sa; |
34dc7c2f | 335 | |
9babb374 BB |
336 | vp = ZTOV(sharezp); |
337 | vn_reinit(vp); | |
338 | vp->v_type = VDIR; | |
34dc7c2f | 339 | |
9babb374 BB |
340 | VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, |
341 | kcred, NULL, &acl_ids)); | |
428870ff | 342 | zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); |
9babb374 BB |
343 | ASSERT3P(zp, ==, sharezp); |
344 | ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */ | |
345 | POINTER_INVALIDATE(&sharezp->z_zfsvfs); | |
346 | error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, | |
347 | ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); | |
348 | zfsvfs->z_shares_dir = sharezp->z_id; | |
349 | ||
350 | zfs_acl_ids_free(&acl_ids); | |
3558fd73 | 351 | // ZTOV(sharezp)->v_count = 0; |
428870ff | 352 | sa_handle_destroy(sharezp->z_sa_hdl); |
9babb374 | 353 | kmem_cache_free(znode_cache, sharezp); |
34dc7c2f | 354 | |
9babb374 | 355 | return (error); |
9ee7fac5 BB |
356 | #else |
357 | return (0); | |
3c9609b3 | 358 | #endif /* HAVE_SMB_SHARE */ |
34dc7c2f BB |
359 | } |
360 | ||
34dc7c2f | 361 | static void |
3558fd73 | 362 | zfs_znode_sa_init(zfs_sb_t *zsb, znode_t *zp, |
428870ff | 363 | dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) |
34dc7c2f | 364 | { |
c96c36fa | 365 | ASSERT(zfs_znode_held(zsb, zp->z_id)); |
34dc7c2f BB |
366 | |
367 | mutex_enter(&zp->z_lock); | |
368 | ||
428870ff BB |
369 | ASSERT(zp->z_sa_hdl == NULL); |
370 | ASSERT(zp->z_acl_cached == NULL); | |
371 | if (sa_hdl == NULL) { | |
3558fd73 | 372 | VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, zp, |
428870ff BB |
373 | SA_HDL_SHARED, &zp->z_sa_hdl)); |
374 | } else { | |
375 | zp->z_sa_hdl = sa_hdl; | |
376 | sa_set_userp(sa_hdl, zp); | |
377 | } | |
34dc7c2f | 378 | |
428870ff | 379 | zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; |
34dc7c2f | 380 | |
34dc7c2f | 381 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
382 | } |
383 | ||
384 | void | |
385 | zfs_znode_dmu_fini(znode_t *zp) | |
386 | { | |
c96c36fa | 387 | ASSERT(zfs_znode_held(ZTOZSB(zp), zp->z_id) || zp->z_unlinked || |
3558fd73 | 388 | RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock)); |
428870ff BB |
389 | |
390 | sa_handle_destroy(zp->z_sa_hdl); | |
391 | zp->z_sa_hdl = NULL; | |
34dc7c2f BB |
392 | } |
393 | ||
394 | /* | |
3558fd73 BB |
395 | * Called by new_inode() to allocate a new inode. |
396 | */ | |
397 | int | |
398 | zfs_inode_alloc(struct super_block *sb, struct inode **ip) | |
399 | { | |
400 | znode_t *zp; | |
401 | ||
79c76d5b | 402 | zp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
3558fd73 BB |
403 | *ip = ZTOI(zp); |
404 | ||
405 | return (0); | |
406 | } | |
407 | ||
408 | /* | |
409 | * Called in multiple places when an inode should be destroyed. | |
410 | */ | |
411 | void | |
412 | zfs_inode_destroy(struct inode *ip) | |
413 | { | |
414 | znode_t *zp = ITOZ(ip); | |
415 | zfs_sb_t *zsb = ZTOZSB(zp); | |
416 | ||
417 | mutex_enter(&zsb->z_znodes_lock); | |
7b3e34ba BB |
418 | if (list_link_active(&zp->z_link_node)) { |
419 | list_remove(&zsb->z_all_znodes, zp); | |
420 | zsb->z_nr_znodes--; | |
421 | } | |
3558fd73 BB |
422 | mutex_exit(&zsb->z_znodes_lock); |
423 | ||
424 | if (zp->z_acl_cached) { | |
425 | zfs_acl_free(zp->z_acl_cached); | |
426 | zp->z_acl_cached = NULL; | |
427 | } | |
428 | ||
82a37189 BB |
429 | if (zp->z_xattr_cached) { |
430 | nvlist_free(zp->z_xattr_cached); | |
431 | zp->z_xattr_cached = NULL; | |
432 | } | |
433 | ||
3558fd73 BB |
434 | kmem_cache_free(znode_cache, zp); |
435 | } | |
436 | ||
437 | static void | |
438 | zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip) | |
439 | { | |
aa6d8c10 | 440 | uint64_t rdev = 0; |
3558fd73 BB |
441 | |
442 | switch (ip->i_mode & S_IFMT) { | |
443 | case S_IFREG: | |
444 | ip->i_op = &zpl_inode_operations; | |
445 | ip->i_fop = &zpl_file_operations; | |
446 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
447 | break; | |
448 | ||
449 | case S_IFDIR: | |
450 | ip->i_op = &zpl_dir_inode_operations; | |
451 | ip->i_fop = &zpl_dir_file_operations; | |
452 | ITOZ(ip)->z_zn_prefetch = B_TRUE; | |
453 | break; | |
454 | ||
455 | case S_IFLNK: | |
456 | ip->i_op = &zpl_symlink_inode_operations; | |
457 | break; | |
458 | ||
aa6d8c10 NB |
459 | /* |
460 | * rdev is only stored in a SA only for device files. | |
461 | */ | |
3558fd73 BB |
462 | case S_IFCHR: |
463 | case S_IFBLK: | |
53b1d979 BB |
464 | sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb), &rdev, |
465 | sizeof (rdev)); | |
aa6d8c10 NB |
466 | /*FALLTHROUGH*/ |
467 | case S_IFIFO: | |
468 | case S_IFSOCK: | |
3558fd73 BB |
469 | init_special_inode(ip, ip->i_mode, rdev); |
470 | ip->i_op = &zpl_special_inode_operations; | |
471 | break; | |
472 | ||
473 | default: | |
53b1d979 BB |
474 | zfs_panic_recover("inode %llu has invalid mode: 0x%x\n", |
475 | (u_longlong_t)ip->i_ino, ip->i_mode); | |
476 | ||
477 | /* Assume the inode is a file and attempt to continue */ | |
478 | ip->i_mode = S_IFREG | 0644; | |
479 | ip->i_op = &zpl_inode_operations; | |
480 | ip->i_fop = &zpl_file_operations; | |
481 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
482 | break; | |
3558fd73 BB |
483 | } |
484 | } | |
485 | ||
704cd075 CC |
486 | void |
487 | zfs_set_inode_flags(znode_t *zp, struct inode *ip) | |
488 | { | |
489 | /* | |
490 | * Linux and Solaris have different sets of file attributes, so we | |
491 | * restrict this conversion to the intersection of the two. | |
492 | */ | |
493 | ||
494 | if (zp->z_pflags & ZFS_IMMUTABLE) | |
495 | ip->i_flags |= S_IMMUTABLE; | |
496 | else | |
497 | ip->i_flags &= ~S_IMMUTABLE; | |
498 | ||
499 | if (zp->z_pflags & ZFS_APPENDONLY) | |
500 | ip->i_flags |= S_APPEND; | |
501 | else | |
502 | ip->i_flags &= ~S_APPEND; | |
503 | } | |
504 | ||
505 | /* | |
506 | * Update the embedded inode given the znode. We should work toward | |
507 | * eliminating this function as soon as possible by removing values | |
508 | * which are duplicated between the znode and inode. If the generic | |
509 | * inode has the correct field it should be used, and the ZFS code | |
510 | * updated to access the inode. This can be done incrementally. | |
511 | */ | |
512 | static void | |
513 | zfs_inode_update_impl(znode_t *zp, boolean_t new) | |
514 | { | |
515 | zfs_sb_t *zsb; | |
516 | struct inode *ip; | |
517 | uint32_t blksize; | |
518 | u_longlong_t i_blocks; | |
519 | uint64_t atime[2], mtime[2], ctime[2]; | |
520 | ||
521 | ASSERT(zp != NULL); | |
522 | zsb = ZTOZSB(zp); | |
523 | ip = ZTOI(zp); | |
524 | ||
525 | /* Skip .zfs control nodes which do not exist on disk. */ | |
526 | if (zfsctl_is_node(ip)) | |
527 | return; | |
528 | ||
529 | sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16); | |
530 | sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16); | |
531 | sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16); | |
532 | ||
533 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, &i_blocks); | |
534 | ||
535 | spin_lock(&ip->i_lock); | |
704cd075 CC |
536 | ip->i_mode = zp->z_mode; |
537 | zfs_set_inode_flags(zp, ip); | |
538 | ip->i_blkbits = SPA_MINBLOCKSHIFT; | |
539 | ip->i_blocks = i_blocks; | |
540 | ||
541 | /* | |
542 | * Only read atime from SA if we are newly created inode (or rezget), | |
543 | * otherwise i_atime might be dirty. | |
544 | */ | |
545 | if (new) | |
546 | ZFS_TIME_DECODE(&ip->i_atime, atime); | |
547 | ZFS_TIME_DECODE(&ip->i_mtime, mtime); | |
548 | ZFS_TIME_DECODE(&ip->i_ctime, ctime); | |
549 | ||
550 | i_size_write(ip, zp->z_size); | |
551 | spin_unlock(&ip->i_lock); | |
552 | } | |
553 | ||
554 | static void | |
555 | zfs_inode_update_new(znode_t *zp) | |
556 | { | |
557 | zfs_inode_update_impl(zp, B_TRUE); | |
558 | } | |
559 | ||
560 | void | |
561 | zfs_inode_update(znode_t *zp) | |
562 | { | |
563 | zfs_inode_update_impl(zp, B_FALSE); | |
564 | } | |
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 * | |
3558fd73 | 574 | zfs_znode_alloc(zfs_sb_t *zsb, 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; |
0df9673f | 584 | sa_bulk_attr_t bulk[8]; |
428870ff | 585 | int count = 0; |
34dc7c2f | 586 | |
3558fd73 | 587 | ASSERT(zsb != NULL); |
34dc7c2f | 588 | |
3558fd73 BB |
589 | ip = new_inode(zsb->z_sb); |
590 | if (ip == NULL) | |
591 | return (NULL); | |
7304b6e5 | 592 | |
3558fd73 | 593 | zp = ITOZ(ip); |
34dc7c2f | 594 | ASSERT(zp->z_dirlocks == NULL); |
ebe7e575 BB |
595 | ASSERT3P(zp->z_acl_cached, ==, NULL); |
596 | ASSERT3P(zp->z_xattr_cached, ==, NULL); | |
572e2857 | 597 | zp->z_moved = 0; |
428870ff | 598 | zp->z_sa_hdl = NULL; |
34dc7c2f BB |
599 | zp->z_unlinked = 0; |
600 | zp->z_atime_dirty = 0; | |
601 | zp->z_mapcnt = 0; | |
34dc7c2f BB |
602 | zp->z_id = db->db_object; |
603 | zp->z_blksz = blksz; | |
604 | zp->z_seq = 0x7A4653; | |
605 | zp->z_sync_cnt = 0; | |
ebe7e575 BB |
606 | zp->z_is_mapped = B_FALSE; |
607 | zp->z_is_ctldir = B_FALSE; | |
7b3e34ba | 608 | zp->z_is_stale = B_FALSE; |
d88895a0 CC |
609 | zp->z_range_lock.zr_size = &zp->z_size; |
610 | zp->z_range_lock.zr_blksz = &zp->z_blksz; | |
611 | zp->z_range_lock.zr_max_blksz = &ZTOZSB(zp)->z_max_blksz; | |
34dc7c2f | 612 | |
3558fd73 BB |
613 | zfs_znode_sa_init(zsb, zp, db, obj_type, hdl); |
614 | ||
7f89ae6b | 615 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8); |
278f2236 | 616 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &tmp_gen, 8); |
3558fd73 | 617 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); |
dfbc8630 | 618 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &links, 8); |
3558fd73 | 619 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff | 620 | &zp->z_pflags, 8); |
3558fd73 | 621 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, |
7304b6e5 | 622 | &parent, 8); |
2c6abf15 NB |
623 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &z_uid, 8); |
624 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &z_gid, 8); | |
428870ff | 625 | |
278f2236 NB |
626 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || |
627 | tmp_gen == 0) { | |
628 | ||
428870ff BB |
629 | if (hdl == NULL) |
630 | sa_handle_destroy(zp->z_sa_hdl); | |
07d63f0c | 631 | zp->z_sa_hdl = NULL; |
3558fd73 | 632 | goto error; |
34dc7c2f | 633 | } |
7304b6e5 | 634 | |
7f89ae6b | 635 | zp->z_mode = mode; |
278f2236 | 636 | ip->i_generation = (uint32_t)tmp_gen; |
dfbc8630 | 637 | set_nlink(ip, (uint32_t)links); |
2c6abf15 NB |
638 | zfs_uid_write(ip, z_uid); |
639 | zfs_gid_write(ip, z_gid); | |
7f89ae6b | 640 | |
3558fd73 | 641 | ip->i_ino = obj; |
704cd075 | 642 | zfs_inode_update_new(zp); |
3558fd73 BB |
643 | zfs_inode_set_ops(zsb, ip); |
644 | ||
7b3e34ba BB |
645 | /* |
646 | * The only way insert_inode_locked() can fail is if the ip->i_ino | |
647 | * number is already hashed for this super block. This can never | |
648 | * happen because the inode numbers map 1:1 with the object numbers. | |
649 | * | |
650 | * The one exception is rolling back a mounted file system, but in | |
651 | * this case all the active inode are unhashed during the rollback. | |
652 | */ | |
653 | VERIFY3S(insert_inode_locked(ip), ==, 0); | |
c85b224f | 654 | |
3558fd73 BB |
655 | mutex_enter(&zsb->z_znodes_lock); |
656 | list_insert_tail(&zsb->z_all_znodes, zp); | |
ab26409d | 657 | zsb->z_nr_znodes++; |
b128c09f | 658 | membar_producer(); |
3558fd73 | 659 | mutex_exit(&zsb->z_znodes_lock); |
b128c09f | 660 | |
3558fd73 | 661 | unlock_new_inode(ip); |
34dc7c2f | 662 | return (zp); |
3558fd73 BB |
663 | |
664 | error: | |
3558fd73 | 665 | iput(ip); |
d1d7e268 | 666 | return (NULL); |
34dc7c2f BB |
667 | } |
668 | ||
1e8db771 BB |
669 | /* |
670 | * Safely mark an inode dirty. Inodes which are part of a read-only | |
671 | * file system or snapshot may not be dirtied. | |
672 | */ | |
673 | void | |
674 | zfs_mark_inode_dirty(struct inode *ip) | |
675 | { | |
676 | zfs_sb_t *zsb = ITOZSB(ip); | |
677 | ||
678 | if (zfs_is_readonly(zsb) || dmu_objset_is_snapshot(zsb->z_os)) | |
679 | return; | |
680 | ||
681 | mark_inode_dirty(ip); | |
682 | } | |
683 | ||
428870ff BB |
684 | static uint64_t empty_xattr; |
685 | static uint64_t pad[4]; | |
686 | static zfs_acl_phys_t acl_phys; | |
34dc7c2f BB |
687 | /* |
688 | * Create a new DMU object to hold a zfs znode. | |
689 | * | |
690 | * IN: dzp - parent directory for new znode | |
691 | * vap - file attributes for new znode | |
692 | * tx - dmu transaction id for zap operations | |
693 | * cr - credentials of caller | |
694 | * flag - flags: | |
695 | * IS_ROOT_NODE - new object will be root | |
696 | * IS_XATTR - new object is an attribute | |
34dc7c2f BB |
697 | * bonuslen - length of bonus buffer |
698 | * setaclp - File/Dir initial ACL | |
699 | * fuidp - Tracks fuid allocation. | |
700 | * | |
701 | * OUT: zpp - allocated znode | |
702 | * | |
703 | */ | |
704 | void | |
705 | zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, | |
428870ff | 706 | uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) |
34dc7c2f | 707 | { |
428870ff BB |
708 | uint64_t crtime[2], atime[2], mtime[2], ctime[2]; |
709 | uint64_t mode, size, links, parent, pflags; | |
572e2857 | 710 | uint64_t dzp_pflags = 0; |
428870ff | 711 | uint64_t rdev = 0; |
3558fd73 | 712 | zfs_sb_t *zsb = ZTOZSB(dzp); |
428870ff | 713 | dmu_buf_t *db; |
34dc7c2f BB |
714 | timestruc_t now; |
715 | uint64_t gen, obj; | |
428870ff | 716 | int bonuslen; |
50c957f7 | 717 | int dnodesize; |
428870ff BB |
718 | sa_handle_t *sa_hdl; |
719 | dmu_object_type_t obj_type; | |
f30484af | 720 | sa_bulk_attr_t *sa_attrs; |
428870ff BB |
721 | int cnt = 0; |
722 | zfs_acl_locator_cb_t locate = { 0 }; | |
c96c36fa | 723 | znode_hold_t *zh; |
34dc7c2f | 724 | |
3558fd73 | 725 | if (zsb->z_replay) { |
34dc7c2f | 726 | obj = vap->va_nodeid; |
34dc7c2f BB |
727 | now = vap->va_ctime; /* see zfs_replay_create() */ |
728 | gen = vap->va_nblocks; /* ditto */ | |
50c957f7 | 729 | dnodesize = vap->va_fsid; /* ditto */ |
34dc7c2f BB |
730 | } else { |
731 | obj = 0; | |
732 | gethrestime(&now); | |
733 | gen = dmu_tx_get_txg(tx); | |
50c957f7 | 734 | dnodesize = dmu_objset_dnodesize(zsb->z_os); |
34dc7c2f BB |
735 | } |
736 | ||
50c957f7 NB |
737 | if (dnodesize == 0) |
738 | dnodesize = DNODE_MIN_SIZE; | |
739 | ||
3558fd73 | 740 | obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; |
50c957f7 | 741 | |
428870ff | 742 | bonuslen = (obj_type == DMU_OT_SA) ? |
50c957f7 | 743 | DN_BONUS_SIZE(dnodesize) : ZFS_OLD_ZNODE_PHYS_SIZE; |
428870ff | 744 | |
34dc7c2f BB |
745 | /* |
746 | * Create a new DMU object. | |
747 | */ | |
748 | /* | |
749 | * There's currently no mechanism for pre-reading the blocks that will | |
572e2857 | 750 | * be needed to allocate a new object, so we accept the small chance |
34dc7c2f BB |
751 | * that there will be an i/o error and we will fail one of the |
752 | * assertions below. | |
753 | */ | |
3558fd73 BB |
754 | if (S_ISDIR(vap->va_mode)) { |
755 | if (zsb->z_replay) { | |
50c957f7 | 756 | VERIFY0(zap_create_claim_norm_dnsize(zsb->z_os, obj, |
3558fd73 | 757 | zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS, |
50c957f7 | 758 | obj_type, bonuslen, dnodesize, tx)); |
34dc7c2f | 759 | } else { |
50c957f7 | 760 | obj = zap_create_norm_dnsize(zsb->z_os, |
3558fd73 | 761 | zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS, |
50c957f7 | 762 | obj_type, bonuslen, dnodesize, tx); |
34dc7c2f BB |
763 | } |
764 | } else { | |
3558fd73 | 765 | if (zsb->z_replay) { |
50c957f7 | 766 | VERIFY0(dmu_object_claim_dnsize(zsb->z_os, obj, |
34dc7c2f | 767 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
50c957f7 | 768 | obj_type, bonuslen, dnodesize, tx)); |
34dc7c2f | 769 | } else { |
50c957f7 | 770 | obj = dmu_object_alloc_dnsize(zsb->z_os, |
34dc7c2f | 771 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
50c957f7 | 772 | obj_type, bonuslen, dnodesize, tx); |
34dc7c2f BB |
773 | } |
774 | } | |
34dc7c2f | 775 | |
c96c36fa | 776 | zh = zfs_znode_hold_enter(zsb, obj); |
3558fd73 | 777 | VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db)); |
34dc7c2f BB |
778 | |
779 | /* | |
780 | * If this is the root, fix up the half-initialized parent pointer | |
781 | * to reference the just-allocated physical data area. | |
782 | */ | |
783 | if (flag & IS_ROOT_NODE) { | |
34dc7c2f | 784 | dzp->z_id = obj; |
428870ff BB |
785 | } else { |
786 | dzp_pflags = dzp->z_pflags; | |
34dc7c2f BB |
787 | } |
788 | ||
789 | /* | |
790 | * If parent is an xattr, so am I. | |
791 | */ | |
428870ff | 792 | if (dzp_pflags & ZFS_XATTR) { |
34dc7c2f | 793 | flag |= IS_XATTR; |
34dc7c2f BB |
794 | } |
795 | ||
3558fd73 | 796 | if (zsb->z_use_fuids) |
428870ff BB |
797 | pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; |
798 | else | |
799 | pflags = 0; | |
34dc7c2f | 800 | |
3558fd73 | 801 | if (S_ISDIR(vap->va_mode)) { |
428870ff | 802 | size = 2; /* contents ("." and "..") */ |
dfbc8630 | 803 | links = 2; |
428870ff | 804 | } else { |
dfbc8630 CD |
805 | size = 0; |
806 | links = 1; | |
34dc7c2f BB |
807 | } |
808 | ||
aa6d8c10 | 809 | if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode)) |
dc1d7665 | 810 | rdev = vap->va_rdev; |
428870ff BB |
811 | |
812 | parent = dzp->z_id; | |
813 | mode = acl_ids->z_mode; | |
34dc7c2f | 814 | if (flag & IS_XATTR) |
428870ff | 815 | pflags |= ZFS_XATTR; |
34dc7c2f | 816 | |
428870ff BB |
817 | /* |
818 | * No execs denied will be deterimed when zfs_mode_compute() is called. | |
819 | */ | |
820 | pflags |= acl_ids->z_aclp->z_hints & | |
821 | (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| | |
822 | ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); | |
34dc7c2f | 823 | |
428870ff BB |
824 | ZFS_TIME_ENCODE(&now, crtime); |
825 | ZFS_TIME_ENCODE(&now, ctime); | |
34dc7c2f | 826 | |
3558fd73 | 827 | if (vap->va_mask & ATTR_ATIME) { |
428870ff | 828 | ZFS_TIME_ENCODE(&vap->va_atime, atime); |
34dc7c2f | 829 | } else { |
428870ff | 830 | ZFS_TIME_ENCODE(&now, atime); |
34dc7c2f BB |
831 | } |
832 | ||
3558fd73 | 833 | if (vap->va_mask & ATTR_MTIME) { |
428870ff BB |
834 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
835 | } else { | |
836 | ZFS_TIME_ENCODE(&now, mtime); | |
837 | } | |
838 | ||
839 | /* Now add in all of the "SA" attributes */ | |
3558fd73 | 840 | VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED, |
428870ff BB |
841 | &sa_hdl)); |
842 | ||
843 | /* | |
844 | * Setup the array of attributes to be replaced/set on the new file | |
845 | * | |
846 | * order for DMU_OT_ZNODE is critical since it needs to be constructed | |
847 | * in the old znode_phys_t format. Don't change this ordering | |
848 | */ | |
79c76d5b | 849 | sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); |
428870ff BB |
850 | |
851 | if (obj_type == DMU_OT_ZNODE) { | |
3558fd73 | 852 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb), |
428870ff | 853 | NULL, &atime, 16); |
3558fd73 | 854 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb), |
428870ff | 855 | NULL, &mtime, 16); |
3558fd73 | 856 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb), |
428870ff | 857 | NULL, &ctime, 16); |
3558fd73 | 858 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb), |
428870ff | 859 | NULL, &crtime, 16); |
3558fd73 | 860 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb), |
428870ff | 861 | NULL, &gen, 8); |
3558fd73 | 862 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb), |
428870ff | 863 | NULL, &mode, 8); |
3558fd73 | 864 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb), |
428870ff | 865 | NULL, &size, 8); |
3558fd73 | 866 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb), |
428870ff | 867 | NULL, &parent, 8); |
34dc7c2f | 868 | } else { |
3558fd73 | 869 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb), |
428870ff | 870 | NULL, &mode, 8); |
3558fd73 | 871 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb), |
428870ff | 872 | NULL, &size, 8); |
3558fd73 | 873 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb), |
428870ff | 874 | NULL, &gen, 8); |
3558fd73 BB |
875 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), |
876 | NULL, &acl_ids->z_fuid, 8); | |
877 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), | |
878 | NULL, &acl_ids->z_fgid, 8); | |
879 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb), | |
428870ff | 880 | NULL, &parent, 8); |
3558fd73 | 881 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb), |
428870ff | 882 | NULL, &pflags, 8); |
3558fd73 | 883 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb), |
428870ff | 884 | NULL, &atime, 16); |
3558fd73 | 885 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb), |
428870ff | 886 | NULL, &mtime, 16); |
3558fd73 | 887 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb), |
428870ff | 888 | NULL, &ctime, 16); |
3558fd73 | 889 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb), |
428870ff BB |
890 | NULL, &crtime, 16); |
891 | } | |
892 | ||
3558fd73 | 893 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8); |
428870ff BB |
894 | |
895 | if (obj_type == DMU_OT_ZNODE) { | |
3558fd73 | 896 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL, |
428870ff | 897 | &empty_xattr, 8); |
34dc7c2f | 898 | } |
428870ff | 899 | if (obj_type == DMU_OT_ZNODE || |
aa6d8c10 | 900 | (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) { |
3558fd73 | 901 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb), |
428870ff | 902 | NULL, &rdev, 8); |
428870ff BB |
903 | } |
904 | if (obj_type == DMU_OT_ZNODE) { | |
3558fd73 | 905 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb), |
428870ff | 906 | NULL, &pflags, 8); |
3558fd73 | 907 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL, |
428870ff | 908 | &acl_ids->z_fuid, 8); |
3558fd73 | 909 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL, |
428870ff | 910 | &acl_ids->z_fgid, 8); |
3558fd73 | 911 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad, |
428870ff | 912 | sizeof (uint64_t) * 4); |
3558fd73 | 913 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL, |
428870ff BB |
914 | &acl_phys, sizeof (zfs_acl_phys_t)); |
915 | } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { | |
3558fd73 | 916 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL, |
428870ff BB |
917 | &acl_ids->z_aclp->z_acl_count, 8); |
918 | locate.cb_aclp = acl_ids->z_aclp; | |
3558fd73 | 919 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb), |
428870ff BB |
920 | zfs_acl_data_locator, &locate, |
921 | acl_ids->z_aclp->z_acl_bytes); | |
922 | mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, | |
923 | acl_ids->z_fuid, acl_ids->z_fgid); | |
924 | } | |
925 | ||
926 | VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); | |
34dc7c2f | 927 | |
34dc7c2f | 928 | if (!(flag & IS_ROOT_NODE)) { |
31b6111f | 929 | *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl); |
7b3e34ba BB |
930 | VERIFY(*zpp != NULL); |
931 | VERIFY(dzp != NULL); | |
34dc7c2f BB |
932 | } else { |
933 | /* | |
934 | * If we are creating the root node, the "parent" we | |
935 | * passed in is the znode for the root. | |
936 | */ | |
937 | *zpp = dzp; | |
428870ff BB |
938 | |
939 | (*zpp)->z_sa_hdl = sa_hdl; | |
34dc7c2f | 940 | } |
428870ff BB |
941 | |
942 | (*zpp)->z_pflags = pflags; | |
943 | (*zpp)->z_mode = mode; | |
50c957f7 | 944 | (*zpp)->z_dnodesize = dnodesize; |
428870ff | 945 | |
428870ff BB |
946 | if (obj_type == DMU_OT_ZNODE || |
947 | acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { | |
b0bc7a84 | 948 | VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); |
428870ff | 949 | } |
d1d7e268 | 950 | kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); |
c96c36fa | 951 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
952 | } |
953 | ||
5484965a | 954 | /* |
d3cc8b15 WA |
955 | * Update in-core attributes. It is assumed the caller will be doing an |
956 | * sa_bulk_update to push the changes out. | |
5484965a BB |
957 | */ |
958 | void | |
959 | zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) | |
960 | { | |
961 | xoptattr_t *xoap; | |
962 | ||
963 | xoap = xva_getxoptattr(xvap); | |
964 | ASSERT(xoap); | |
965 | ||
966 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { | |
967 | uint64_t times[2]; | |
968 | ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); | |
969 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)), | |
970 | ×, sizeof (times), tx); | |
971 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
972 | } | |
973 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
974 | ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, | |
975 | zp->z_pflags, tx); | |
976 | XVA_SET_RTN(xvap, XAT_READONLY); | |
977 | } | |
978 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
979 | ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, | |
980 | zp->z_pflags, tx); | |
981 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
982 | } | |
983 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
984 | ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, | |
985 | zp->z_pflags, tx); | |
986 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
987 | } | |
988 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
989 | ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, | |
990 | zp->z_pflags, tx); | |
991 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
992 | } | |
993 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
994 | ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, | |
995 | zp->z_pflags, tx); | |
996 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
997 | } | |
998 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
999 | ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, | |
1000 | zp->z_pflags, tx); | |
1001 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
1002 | } | |
1003 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
1004 | ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, | |
1005 | zp->z_pflags, tx); | |
1006 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
1007 | } | |
1008 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
1009 | ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, | |
1010 | zp->z_pflags, tx); | |
1011 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
1012 | } | |
1013 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
1014 | ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, | |
1015 | zp->z_pflags, tx); | |
1016 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
1017 | } | |
1018 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
1019 | ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, | |
1020 | xoap->xoa_av_quarantined, zp->z_pflags, tx); | |
1021 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
1022 | } | |
1023 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
1024 | ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, | |
1025 | zp->z_pflags, tx); | |
1026 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
1027 | } | |
1028 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { | |
1029 | zfs_sa_set_scanstamp(zp, xvap, tx); | |
1030 | XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); | |
1031 | } | |
1032 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
1033 | ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, | |
1034 | zp->z_pflags, tx); | |
1035 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
1036 | } | |
1037 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
1038 | ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, | |
1039 | zp->z_pflags, tx); | |
1040 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
1041 | } | |
1042 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
1043 | ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, | |
1044 | zp->z_pflags, tx); | |
1045 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
1046 | } | |
1047 | } | |
1048 | ||
34dc7c2f | 1049 | int |
3558fd73 | 1050 | zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp) |
34dc7c2f BB |
1051 | { |
1052 | dmu_object_info_t doi; | |
1053 | dmu_buf_t *db; | |
1054 | znode_t *zp; | |
c96c36fa | 1055 | znode_hold_t *zh; |
34dc7c2f | 1056 | int err; |
428870ff | 1057 | sa_handle_t *hdl; |
34dc7c2f BB |
1058 | |
1059 | *zpp = NULL; | |
1060 | ||
6f9548c4 | 1061 | again: |
c96c36fa | 1062 | zh = zfs_znode_hold_enter(zsb, obj_num); |
34dc7c2f | 1063 | |
3558fd73 | 1064 | err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db); |
34dc7c2f | 1065 | if (err) { |
c96c36fa | 1066 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1067 | return (err); |
1068 | } | |
1069 | ||
1070 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1071 | if (doi.doi_bonus_type != DMU_OT_SA && |
1072 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1073 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1074 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1075 | sa_buf_rele(db, NULL); | |
c96c36fa | 1076 | zfs_znode_hold_exit(zsb, zh); |
2e528b49 | 1077 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1078 | } |
1079 | ||
428870ff BB |
1080 | hdl = dmu_buf_get_user(db); |
1081 | if (hdl != NULL) { | |
36df2843 | 1082 | zp = sa_get_userdata(hdl); |
34dc7c2f | 1083 | |
8ac67298 | 1084 | |
34dc7c2f | 1085 | /* |
428870ff BB |
1086 | * Since "SA" does immediate eviction we |
1087 | * should never find a sa handle that doesn't | |
1088 | * know about the znode. | |
34dc7c2f | 1089 | */ |
428870ff BB |
1090 | |
1091 | ASSERT3P(zp, !=, NULL); | |
1092 | ||
1093 | mutex_enter(&zp->z_lock); | |
34dc7c2f BB |
1094 | ASSERT3U(zp->z_id, ==, obj_num); |
1095 | if (zp->z_unlinked) { | |
2e528b49 | 1096 | err = SET_ERROR(ENOENT); |
34dc7c2f | 1097 | } else { |
6f9548c4 RY |
1098 | /* |
1099 | * If igrab() returns NULL the VFS has independently | |
1100 | * determined the inode should be evicted and has | |
1101 | * called iput_final() to start the eviction process. | |
1102 | * The SA handle is still valid but because the VFS | |
1103 | * requires that the eviction succeed we must drop | |
1104 | * our locks and references to allow the eviction to | |
1105 | * complete. The zfs_zget() may then be retried. | |
1106 | * | |
1107 | * This unlikely case could be optimized by registering | |
1108 | * a sops->drop_inode() callback. The callback would | |
1109 | * need to detect the active SA hold thereby informing | |
1110 | * the VFS that this inode should not be evicted. | |
1111 | */ | |
1112 | if (igrab(ZTOI(zp)) == NULL) { | |
1113 | mutex_exit(&zp->z_lock); | |
1114 | sa_buf_rele(db, NULL); | |
c96c36fa | 1115 | zfs_znode_hold_exit(zsb, zh); |
6102d037 | 1116 | /* inode might need this to finish evict */ |
1117 | cond_resched(); | |
6f9548c4 RY |
1118 | goto again; |
1119 | } | |
34dc7c2f BB |
1120 | *zpp = zp; |
1121 | err = 0; | |
1122 | } | |
34dc7c2f | 1123 | mutex_exit(&zp->z_lock); |
f3ad9cd6 | 1124 | sa_buf_rele(db, NULL); |
c96c36fa | 1125 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1126 | return (err); |
1127 | } | |
1128 | ||
1129 | /* | |
3558fd73 | 1130 | * Not found create new znode/vnode but only if file exists. |
428870ff BB |
1131 | * |
1132 | * There is a small window where zfs_vget() could | |
1133 | * find this object while a file create is still in | |
1134 | * progress. This is checked for in zfs_znode_alloc() | |
1135 | * | |
1136 | * if zfs_znode_alloc() fails it will drop the hold on the | |
1137 | * bonus buffer. | |
34dc7c2f | 1138 | */ |
3558fd73 | 1139 | zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size, |
31b6111f | 1140 | doi.doi_bonus_type, obj_num, NULL); |
428870ff | 1141 | if (zp == NULL) { |
2e528b49 | 1142 | err = SET_ERROR(ENOENT); |
428870ff BB |
1143 | } else { |
1144 | *zpp = zp; | |
1145 | } | |
c96c36fa | 1146 | zfs_znode_hold_exit(zsb, zh); |
428870ff | 1147 | return (err); |
34dc7c2f BB |
1148 | } |
1149 | ||
1150 | int | |
1151 | zfs_rezget(znode_t *zp) | |
1152 | { | |
3558fd73 | 1153 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
1154 | dmu_object_info_t doi; |
1155 | dmu_buf_t *db; | |
1156 | uint64_t obj_num = zp->z_id; | |
428870ff | 1157 | uint64_t mode; |
dfbc8630 | 1158 | uint64_t links; |
0df9673f | 1159 | sa_bulk_attr_t bulk[7]; |
34dc7c2f | 1160 | int err; |
428870ff BB |
1161 | int count = 0; |
1162 | uint64_t gen; | |
2c6abf15 | 1163 | uint64_t z_uid, z_gid; |
c96c36fa | 1164 | znode_hold_t *zh; |
34dc7c2f | 1165 | |
cbecb4fb CC |
1166 | /* |
1167 | * skip ctldir, otherwise they will always get invalidated. This will | |
1168 | * cause funny behaviour for the mounted snapdirs. Especially for | |
1169 | * Linux >= 3.18, d_invalidate will detach the mountpoint and prevent | |
1170 | * anyone automount it again as long as someone is still using the | |
1171 | * detached mount. | |
1172 | */ | |
1173 | if (zp->z_is_ctldir) | |
1174 | return (0); | |
1175 | ||
c96c36fa | 1176 | zh = zfs_znode_hold_enter(zsb, obj_num); |
34dc7c2f | 1177 | |
428870ff BB |
1178 | mutex_enter(&zp->z_acl_lock); |
1179 | if (zp->z_acl_cached) { | |
1180 | zfs_acl_free(zp->z_acl_cached); | |
1181 | zp->z_acl_cached = NULL; | |
1182 | } | |
428870ff | 1183 | mutex_exit(&zp->z_acl_lock); |
7b3e34ba | 1184 | |
228b461b | 1185 | rw_enter(&zp->z_xattr_lock, RW_WRITER); |
7b3e34ba BB |
1186 | if (zp->z_xattr_cached) { |
1187 | nvlist_free(zp->z_xattr_cached); | |
1188 | zp->z_xattr_cached = NULL; | |
1189 | } | |
7b3e34ba BB |
1190 | rw_exit(&zp->z_xattr_lock); |
1191 | ||
428870ff | 1192 | ASSERT(zp->z_sa_hdl == NULL); |
3558fd73 | 1193 | err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db); |
34dc7c2f | 1194 | if (err) { |
c96c36fa | 1195 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1196 | return (err); |
1197 | } | |
1198 | ||
1199 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1200 | if (doi.doi_bonus_type != DMU_OT_SA && |
1201 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1202 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1203 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1204 | sa_buf_rele(db, NULL); | |
c96c36fa | 1205 | zfs_znode_hold_exit(zsb, zh); |
2e528b49 | 1206 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1207 | } |
1208 | ||
3558fd73 | 1209 | zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL); |
428870ff BB |
1210 | |
1211 | /* reload cached values */ | |
3558fd73 | 1212 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, |
428870ff | 1213 | &gen, sizeof (gen)); |
3558fd73 | 1214 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, |
428870ff | 1215 | &zp->z_size, sizeof (zp->z_size)); |
3558fd73 | 1216 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, |
dfbc8630 | 1217 | &links, sizeof (links)); |
3558fd73 | 1218 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff | 1219 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3558fd73 | 1220 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, |
2c6abf15 | 1221 | &z_uid, sizeof (z_uid)); |
3558fd73 | 1222 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, |
2c6abf15 | 1223 | &z_gid, sizeof (z_gid)); |
3558fd73 | 1224 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, |
428870ff BB |
1225 | &mode, sizeof (mode)); |
1226 | ||
428870ff BB |
1227 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { |
1228 | zfs_znode_dmu_fini(zp); | |
c96c36fa | 1229 | zfs_znode_hold_exit(zsb, zh); |
2e528b49 | 1230 | return (SET_ERROR(EIO)); |
428870ff BB |
1231 | } |
1232 | ||
572e2857 | 1233 | zp->z_mode = mode; |
2c6abf15 NB |
1234 | zfs_uid_write(ZTOI(zp), z_uid); |
1235 | zfs_gid_write(ZTOI(zp), z_gid); | |
572e2857 | 1236 | |
278f2236 | 1237 | if (gen != ZTOI(zp)->i_generation) { |
428870ff | 1238 | zfs_znode_dmu_fini(zp); |
c96c36fa | 1239 | zfs_znode_hold_exit(zsb, zh); |
2e528b49 | 1240 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1241 | } |
1242 | ||
dfbc8630 CD |
1243 | zp->z_unlinked = (ZTOI(zp)->i_nlink == 0); |
1244 | set_nlink(ZTOI(zp), (uint32_t)links); | |
1245 | ||
34dc7c2f | 1246 | zp->z_blksz = doi.doi_data_block_size; |
704cd075 CC |
1247 | zp->z_atime_dirty = 0; |
1248 | zfs_inode_update_new(zp); | |
34dc7c2f | 1249 | |
c96c36fa | 1250 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1251 | |
1252 | return (0); | |
1253 | } | |
1254 | ||
1255 | void | |
1256 | zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) | |
1257 | { | |
3558fd73 BB |
1258 | zfs_sb_t *zsb = ZTOZSB(zp); |
1259 | objset_t *os = zsb->z_os; | |
34dc7c2f | 1260 | uint64_t obj = zp->z_id; |
572e2857 | 1261 | uint64_t acl_obj = zfs_external_acl(zp); |
c96c36fa | 1262 | znode_hold_t *zh; |
34dc7c2f | 1263 | |
c96c36fa | 1264 | zh = zfs_znode_hold_enter(zsb, obj); |
572e2857 BB |
1265 | if (acl_obj) { |
1266 | VERIFY(!zp->z_is_sa); | |
b128c09f | 1267 | VERIFY(0 == dmu_object_free(os, acl_obj, tx)); |
572e2857 | 1268 | } |
b128c09f | 1269 | VERIFY(0 == dmu_object_free(os, obj, tx)); |
34dc7c2f | 1270 | zfs_znode_dmu_fini(zp); |
c96c36fa | 1271 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1272 | } |
1273 | ||
1274 | void | |
1275 | zfs_zinactive(znode_t *zp) | |
1276 | { | |
3558fd73 | 1277 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f | 1278 | uint64_t z_id = zp->z_id; |
c96c36fa | 1279 | znode_hold_t *zh; |
34dc7c2f | 1280 | |
428870ff | 1281 | ASSERT(zp->z_sa_hdl); |
34dc7c2f BB |
1282 | |
1283 | /* | |
d6bd8eaa | 1284 | * Don't allow a zfs_zget() while were trying to release this znode. |
34dc7c2f | 1285 | */ |
c96c36fa | 1286 | zh = zfs_znode_hold_enter(zsb, z_id); |
d6bd8eaa | 1287 | |
34dc7c2f | 1288 | mutex_enter(&zp->z_lock); |
34dc7c2f BB |
1289 | |
1290 | /* | |
1291 | * If this was the last reference to a file with no links, | |
1292 | * remove the file from the file system. | |
1293 | */ | |
1294 | if (zp->z_unlinked) { | |
1295 | mutex_exit(&zp->z_lock); | |
c96c36fa | 1296 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1297 | zfs_rmnode(zp); |
1298 | return; | |
1299 | } | |
428870ff | 1300 | |
34dc7c2f BB |
1301 | mutex_exit(&zp->z_lock); |
1302 | zfs_znode_dmu_fini(zp); | |
d6bd8eaa | 1303 | |
c96c36fa | 1304 | zfs_znode_hold_exit(zsb, zh); |
34dc7c2f BB |
1305 | } |
1306 | ||
6d111134 TC |
1307 | static inline int |
1308 | zfs_compare_timespec(struct timespec *t1, struct timespec *t2) | |
1309 | { | |
1310 | if (t1->tv_sec < t2->tv_sec) | |
1311 | return (-1); | |
1312 | ||
1313 | if (t1->tv_sec > t2->tv_sec) | |
1314 | return (1); | |
1315 | ||
1316 | return (t1->tv_nsec - t2->tv_nsec); | |
1317 | } | |
1318 | ||
6d111134 TC |
1319 | /* |
1320 | * Prepare to update znode time stamps. | |
1321 | * | |
1322 | * IN: zp - znode requiring timestamp update | |
0df9673f | 1323 | * flag - ATTR_MTIME, ATTR_CTIME flags |
6d111134 | 1324 | * |
0df9673f | 1325 | * OUT: zp - z_seq |
6d111134 TC |
1326 | * mtime - new mtime |
1327 | * ctime - new ctime | |
1328 | * | |
0df9673f CC |
1329 | * Note: We don't update atime here, because we rely on Linux VFS to do |
1330 | * atime updating. | |
6d111134 | 1331 | */ |
34dc7c2f | 1332 | void |
428870ff | 1333 | zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], |
0df9673f | 1334 | uint64_t ctime[2]) |
34dc7c2f BB |
1335 | { |
1336 | timestruc_t now; | |
1337 | ||
34dc7c2f BB |
1338 | gethrestime(&now); |
1339 | ||
0df9673f | 1340 | zp->z_seq++; |
34dc7c2f | 1341 | |
3558fd73 | 1342 | if (flag & ATTR_MTIME) { |
428870ff | 1343 | ZFS_TIME_ENCODE(&now, mtime); |
3558fd73 | 1344 | if (ZTOZSB(zp)->z_use_fuids) { |
428870ff BB |
1345 | zp->z_pflags |= (ZFS_ARCHIVE | |
1346 | ZFS_AV_MODIFIED); | |
1347 | } | |
34dc7c2f BB |
1348 | } |
1349 | ||
3558fd73 | 1350 | if (flag & ATTR_CTIME) { |
428870ff | 1351 | ZFS_TIME_ENCODE(&now, ctime); |
3558fd73 | 1352 | if (ZTOZSB(zp)->z_use_fuids) |
428870ff | 1353 | zp->z_pflags |= ZFS_ARCHIVE; |
34dc7c2f BB |
1354 | } |
1355 | } | |
1356 | ||
34dc7c2f BB |
1357 | /* |
1358 | * Grow the block size for a file. | |
1359 | * | |
1360 | * IN: zp - znode of file to free data in. | |
1361 | * size - requested block size | |
1362 | * tx - open transaction. | |
1363 | * | |
1364 | * NOTE: this function assumes that the znode is write locked. | |
1365 | */ | |
1366 | void | |
1367 | zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) | |
1368 | { | |
1369 | int error; | |
1370 | u_longlong_t dummy; | |
1371 | ||
1372 | if (size <= zp->z_blksz) | |
1373 | return; | |
1374 | /* | |
1375 | * If the file size is already greater than the current blocksize, | |
1376 | * we will not grow. If there is more than one block in a file, | |
1377 | * the blocksize cannot change. | |
1378 | */ | |
428870ff | 1379 | if (zp->z_blksz && zp->z_size > zp->z_blksz) |
34dc7c2f BB |
1380 | return; |
1381 | ||
3558fd73 | 1382 | error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id, |
34dc7c2f | 1383 | size, 0, tx); |
428870ff | 1384 | |
34dc7c2f BB |
1385 | if (error == ENOTSUP) |
1386 | return; | |
c99c9001 | 1387 | ASSERT0(error); |
34dc7c2f BB |
1388 | |
1389 | /* What blocksize did we actually get? */ | |
428870ff | 1390 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); |
34dc7c2f BB |
1391 | } |
1392 | ||
34dc7c2f | 1393 | /* |
b128c09f | 1394 | * Increase the file length |
34dc7c2f BB |
1395 | * |
1396 | * IN: zp - znode of file to free data in. | |
b128c09f | 1397 | * end - new end-of-file |
34dc7c2f | 1398 | * |
19d55079 | 1399 | * RETURN: 0 on success, error code on failure |
34dc7c2f | 1400 | */ |
b128c09f BB |
1401 | static int |
1402 | zfs_extend(znode_t *zp, uint64_t end) | |
34dc7c2f | 1403 | { |
3558fd73 | 1404 | zfs_sb_t *zsb = ZTOZSB(zp); |
b128c09f | 1405 | dmu_tx_t *tx; |
34dc7c2f | 1406 | rl_t *rl; |
b128c09f | 1407 | uint64_t newblksz; |
34dc7c2f BB |
1408 | int error; |
1409 | ||
34dc7c2f | 1410 | /* |
b128c09f | 1411 | * We will change zp_size, lock the whole file. |
34dc7c2f | 1412 | */ |
d88895a0 | 1413 | rl = zfs_range_lock(&zp->z_range_lock, 0, UINT64_MAX, RL_WRITER); |
34dc7c2f BB |
1414 | |
1415 | /* | |
1416 | * Nothing to do if file already at desired length. | |
1417 | */ | |
428870ff | 1418 | if (end <= zp->z_size) { |
34dc7c2f BB |
1419 | zfs_range_unlock(rl); |
1420 | return (0); | |
1421 | } | |
3558fd73 | 1422 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
1423 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1424 | zfs_sa_upgrade_txholds(tx, zp); | |
b128c09f | 1425 | if (end > zp->z_blksz && |
3558fd73 | 1426 | (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) { |
34dc7c2f BB |
1427 | /* |
1428 | * We are growing the file past the current block size. | |
1429 | */ | |
3558fd73 | 1430 | if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) { |
f1512ee6 MA |
1431 | /* |
1432 | * File's blocksize is already larger than the | |
1433 | * "recordsize" property. Only let it grow to | |
1434 | * the next power of 2. | |
1435 | */ | |
34dc7c2f | 1436 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 | 1437 | newblksz = MIN(end, 1 << highbit64(zp->z_blksz)); |
34dc7c2f | 1438 | } else { |
3558fd73 | 1439 | newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz); |
34dc7c2f | 1440 | } |
b128c09f BB |
1441 | dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); |
1442 | } else { | |
1443 | newblksz = 0; | |
34dc7c2f BB |
1444 | } |
1445 | ||
384f8a09 | 1446 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 1447 | if (error) { |
34dc7c2f BB |
1448 | dmu_tx_abort(tx); |
1449 | zfs_range_unlock(rl); | |
1450 | return (error); | |
1451 | } | |
1452 | ||
b128c09f BB |
1453 | if (newblksz) |
1454 | zfs_grow_blocksize(zp, newblksz, tx); | |
34dc7c2f | 1455 | |
428870ff BB |
1456 | zp->z_size = end; |
1457 | ||
3558fd73 | 1458 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), |
428870ff | 1459 | &zp->z_size, sizeof (zp->z_size), tx)); |
34dc7c2f | 1460 | |
b128c09f | 1461 | zfs_range_unlock(rl); |
34dc7c2f | 1462 | |
b128c09f | 1463 | dmu_tx_commit(tx); |
34dc7c2f | 1464 | |
b128c09f BB |
1465 | return (0); |
1466 | } | |
1467 | ||
223df016 TC |
1468 | /* |
1469 | * zfs_zero_partial_page - Modeled after update_pages() but | |
1470 | * with different arguments and semantics for use by zfs_freesp(). | |
1471 | * | |
1472 | * Zeroes a piece of a single page cache entry for zp at offset | |
1473 | * start and length len. | |
1474 | * | |
1475 | * Caller must acquire a range lock on the file for the region | |
1476 | * being zeroed in order that the ARC and page cache stay in sync. | |
1477 | */ | |
1478 | static void | |
1479 | zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len) | |
1480 | { | |
1481 | struct address_space *mp = ZTOI(zp)->i_mapping; | |
1482 | struct page *pp; | |
1483 | int64_t off; | |
1484 | void *pb; | |
1485 | ||
8b1899d3 | 1486 | ASSERT((start & PAGE_MASK) == ((start + len - 1) & PAGE_MASK)); |
223df016 | 1487 | |
8b1899d3 BB |
1488 | off = start & (PAGE_SIZE - 1); |
1489 | start &= PAGE_MASK; | |
223df016 | 1490 | |
8b1899d3 | 1491 | pp = find_lock_page(mp, start >> PAGE_SHIFT); |
223df016 TC |
1492 | if (pp) { |
1493 | if (mapping_writably_mapped(mp)) | |
1494 | flush_dcache_page(pp); | |
1495 | ||
1496 | pb = kmap(pp); | |
1497 | bzero(pb + off, len); | |
1498 | kunmap(pp); | |
1499 | ||
1500 | if (mapping_writably_mapped(mp)) | |
1501 | flush_dcache_page(pp); | |
1502 | ||
1503 | mark_page_accessed(pp); | |
1504 | SetPageUptodate(pp); | |
1505 | ClearPageError(pp); | |
1506 | unlock_page(pp); | |
8b1899d3 | 1507 | put_page(pp); |
223df016 TC |
1508 | } |
1509 | } | |
1510 | ||
b128c09f BB |
1511 | /* |
1512 | * Free space in a file. | |
1513 | * | |
1514 | * IN: zp - znode of file to free data in. | |
1515 | * off - start of section to free. | |
1516 | * len - length of section to free. | |
1517 | * | |
19d55079 | 1518 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1519 | */ |
1520 | static int | |
1521 | zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) | |
1522 | { | |
3558fd73 | 1523 | zfs_sb_t *zsb = ZTOZSB(zp); |
b128c09f BB |
1524 | rl_t *rl; |
1525 | int error; | |
1526 | ||
1527 | /* | |
1528 | * Lock the range being freed. | |
1529 | */ | |
d88895a0 | 1530 | rl = zfs_range_lock(&zp->z_range_lock, off, len, RL_WRITER); |
b128c09f BB |
1531 | |
1532 | /* | |
1533 | * Nothing to do if file already at desired length. | |
1534 | */ | |
428870ff | 1535 | if (off >= zp->z_size) { |
b128c09f BB |
1536 | zfs_range_unlock(rl); |
1537 | return (0); | |
34dc7c2f BB |
1538 | } |
1539 | ||
428870ff BB |
1540 | if (off + len > zp->z_size) |
1541 | len = zp->z_size - off; | |
b128c09f | 1542 | |
3558fd73 | 1543 | error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len); |
b128c09f | 1544 | |
223df016 TC |
1545 | /* |
1546 | * Zero partial page cache entries. This must be done under a | |
1547 | * range lock in order to keep the ARC and page cache in sync. | |
1548 | */ | |
1549 | if (zp->z_is_mapped) { | |
1550 | loff_t first_page, last_page, page_len; | |
1551 | loff_t first_page_offset, last_page_offset; | |
1552 | ||
1553 | /* first possible full page in hole */ | |
8b1899d3 | 1554 | first_page = (off + PAGE_SIZE - 1) >> PAGE_SHIFT; |
223df016 | 1555 | /* last page of hole */ |
8b1899d3 | 1556 | last_page = (off + len) >> PAGE_SHIFT; |
223df016 TC |
1557 | |
1558 | /* offset of first_page */ | |
8b1899d3 | 1559 | first_page_offset = first_page << PAGE_SHIFT; |
223df016 | 1560 | /* offset of last_page */ |
8b1899d3 | 1561 | last_page_offset = last_page << PAGE_SHIFT; |
223df016 | 1562 | |
cb08f063 TC |
1563 | /* truncate whole pages */ |
1564 | if (last_page_offset > first_page_offset) { | |
1565 | truncate_inode_pages_range(ZTOI(zp)->i_mapping, | |
1566 | first_page_offset, last_page_offset - 1); | |
1567 | } | |
1568 | ||
1569 | /* truncate sub-page ranges */ | |
223df016 TC |
1570 | if (first_page > last_page) { |
1571 | /* entire punched area within a single page */ | |
1572 | zfs_zero_partial_page(zp, off, len); | |
1573 | } else { | |
1574 | /* beginning of punched area at the end of a page */ | |
1575 | page_len = first_page_offset - off; | |
1576 | if (page_len > 0) | |
1577 | zfs_zero_partial_page(zp, off, page_len); | |
1578 | ||
1579 | /* end of punched area at the beginning of a page */ | |
1580 | page_len = off + len - last_page_offset; | |
1581 | if (page_len > 0) | |
1582 | zfs_zero_partial_page(zp, last_page_offset, | |
1583 | page_len); | |
1584 | } | |
1585 | } | |
34dc7c2f BB |
1586 | zfs_range_unlock(rl); |
1587 | ||
b128c09f BB |
1588 | return (error); |
1589 | } | |
1590 | ||
1591 | /* | |
1592 | * Truncate a file | |
1593 | * | |
1594 | * IN: zp - znode of file to free data in. | |
1595 | * end - new end-of-file. | |
1596 | * | |
19d55079 | 1597 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1598 | */ |
1599 | static int | |
1600 | zfs_trunc(znode_t *zp, uint64_t end) | |
1601 | { | |
3558fd73 | 1602 | zfs_sb_t *zsb = ZTOZSB(zp); |
b128c09f BB |
1603 | dmu_tx_t *tx; |
1604 | rl_t *rl; | |
1605 | int error; | |
572e2857 BB |
1606 | sa_bulk_attr_t bulk[2]; |
1607 | int count = 0; | |
b128c09f BB |
1608 | |
1609 | /* | |
1610 | * We will change zp_size, lock the whole file. | |
1611 | */ | |
d88895a0 | 1612 | rl = zfs_range_lock(&zp->z_range_lock, 0, UINT64_MAX, RL_WRITER); |
b128c09f BB |
1613 | |
1614 | /* | |
1615 | * Nothing to do if file already at desired length. | |
1616 | */ | |
428870ff | 1617 | if (end >= zp->z_size) { |
b128c09f BB |
1618 | zfs_range_unlock(rl); |
1619 | return (0); | |
1620 | } | |
1621 | ||
3558fd73 | 1622 | error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1); |
b128c09f BB |
1623 | if (error) { |
1624 | zfs_range_unlock(rl); | |
1625 | return (error); | |
1626 | } | |
3558fd73 | 1627 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
1628 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1629 | zfs_sa_upgrade_txholds(tx, zp); | |
19d55079 | 1630 | dmu_tx_mark_netfree(tx); |
7a8f0e80 | 1631 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1632 | if (error) { |
b128c09f BB |
1633 | dmu_tx_abort(tx); |
1634 | zfs_range_unlock(rl); | |
1635 | return (error); | |
1636 | } | |
b128c09f | 1637 | |
428870ff | 1638 | zp->z_size = end; |
3558fd73 | 1639 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), |
572e2857 | 1640 | NULL, &zp->z_size, sizeof (zp->z_size)); |
428870ff | 1641 | |
572e2857 BB |
1642 | if (end == 0) { |
1643 | zp->z_pflags &= ~ZFS_SPARSE; | |
3558fd73 | 1644 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), |
572e2857 BB |
1645 | NULL, &zp->z_pflags, 8); |
1646 | } | |
1647 | VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); | |
b128c09f | 1648 | |
34dc7c2f BB |
1649 | dmu_tx_commit(tx); |
1650 | ||
d164b209 | 1651 | zfs_range_unlock(rl); |
34dc7c2f BB |
1652 | |
1653 | return (0); | |
1654 | } | |
1655 | ||
b128c09f BB |
1656 | /* |
1657 | * Free space in a file | |
1658 | * | |
1659 | * IN: zp - znode of file to free data in. | |
1660 | * off - start of range | |
1661 | * len - end of range (0 => EOF) | |
1662 | * flag - current file open mode flags. | |
1663 | * log - TRUE if this action should be logged | |
1664 | * | |
19d55079 | 1665 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1666 | */ |
1667 | int | |
1668 | zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) | |
1669 | { | |
b128c09f | 1670 | dmu_tx_t *tx; |
3558fd73 BB |
1671 | zfs_sb_t *zsb = ZTOZSB(zp); |
1672 | zilog_t *zilog = zsb->z_log; | |
428870ff BB |
1673 | uint64_t mode; |
1674 | uint64_t mtime[2], ctime[2]; | |
1675 | sa_bulk_attr_t bulk[3]; | |
1676 | int count = 0; | |
b128c09f BB |
1677 | int error; |
1678 | ||
3558fd73 | 1679 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode, |
428870ff BB |
1680 | sizeof (mode))) != 0) |
1681 | return (error); | |
1682 | ||
1683 | if (off > zp->z_size) { | |
b128c09f BB |
1684 | error = zfs_extend(zp, off+len); |
1685 | if (error == 0 && log) | |
1686 | goto log; | |
223df016 | 1687 | goto out; |
b128c09f BB |
1688 | } |
1689 | ||
b128c09f BB |
1690 | if (len == 0) { |
1691 | error = zfs_trunc(zp, off); | |
1692 | } else { | |
1693 | if ((error = zfs_free_range(zp, off, len)) == 0 && | |
428870ff | 1694 | off + len > zp->z_size) |
b128c09f BB |
1695 | error = zfs_extend(zp, off+len); |
1696 | } | |
1697 | if (error || !log) | |
223df016 | 1698 | goto out; |
b128c09f | 1699 | log: |
3558fd73 | 1700 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
1701 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1702 | zfs_sa_upgrade_txholds(tx, zp); | |
384f8a09 | 1703 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1704 | if (error) { |
b128c09f | 1705 | dmu_tx_abort(tx); |
223df016 | 1706 | goto out; |
b128c09f BB |
1707 | } |
1708 | ||
3558fd73 BB |
1709 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16); |
1710 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16); | |
1711 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), | |
428870ff | 1712 | NULL, &zp->z_pflags, 8); |
0df9673f | 1713 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime); |
428870ff BB |
1714 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); |
1715 | ASSERT(error == 0); | |
1716 | ||
b128c09f BB |
1717 | zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); |
1718 | ||
1719 | dmu_tx_commit(tx); | |
223df016 | 1720 | |
960e08fe | 1721 | zfs_inode_update(zp); |
223df016 TC |
1722 | error = 0; |
1723 | ||
1724 | out: | |
1725 | /* | |
1726 | * Truncate the page cache - for file truncate operations, use | |
1727 | * the purpose-built API for truncations. For punching operations, | |
cb08f063 | 1728 | * the truncation is handled under a range lock in zfs_free_range. |
223df016 TC |
1729 | */ |
1730 | if (len == 0) | |
1731 | truncate_setsize(ZTOI(zp), off); | |
223df016 | 1732 | return (error); |
b128c09f BB |
1733 | } |
1734 | ||
34dc7c2f BB |
1735 | void |
1736 | zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) | |
1737 | { | |
22872ff5 BB |
1738 | struct super_block *sb; |
1739 | zfs_sb_t *zsb; | |
428870ff | 1740 | uint64_t moid, obj, sa_obj, version; |
22872ff5 | 1741 | uint64_t sense = ZFS_CASE_SENSITIVE; |
34dc7c2f BB |
1742 | uint64_t norm = 0; |
1743 | nvpair_t *elem; | |
c96c36fa | 1744 | int size; |
34dc7c2f | 1745 | int error; |
22872ff5 BB |
1746 | int i; |
1747 | znode_t *rootzp = NULL; | |
1748 | vattr_t vattr; | |
1749 | znode_t *zp; | |
1750 | zfs_acl_ids_t acl_ids; | |
34dc7c2f BB |
1751 | |
1752 | /* | |
1753 | * First attempt to create master node. | |
1754 | */ | |
1755 | /* | |
1756 | * In an empty objset, there are no blocks to read and thus | |
1757 | * there can be no i/o errors (which we assert below). | |
1758 | */ | |
1759 | moid = MASTER_NODE_OBJ; | |
1760 | error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, | |
1761 | DMU_OT_NONE, 0, tx); | |
1762 | ASSERT(error == 0); | |
1763 | ||
50c957f7 NB |
1764 | /* |
1765 | * Give dmu_object_alloc() a hint about where to start | |
1766 | * allocating new objects. Otherwise, since the metadnode's | |
1767 | * dnode_phys_t structure isn't initialized yet, dmu_object_next() | |
1768 | * would fail and we'd have to skip to the next dnode block. | |
1769 | */ | |
1770 | os->os_obj_next = moid + 1; | |
1771 | ||
34dc7c2f BB |
1772 | /* |
1773 | * Set starting attributes. | |
1774 | */ | |
428870ff | 1775 | version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); |
34dc7c2f BB |
1776 | elem = NULL; |
1777 | while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { | |
1778 | /* For the moment we expect all zpl props to be uint64_ts */ | |
1779 | uint64_t val; | |
1780 | char *name; | |
1781 | ||
1782 | ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); | |
1783 | VERIFY(nvpair_value_uint64(elem, &val) == 0); | |
1784 | name = nvpair_name(elem); | |
1785 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { | |
9babb374 BB |
1786 | if (val < version) |
1787 | version = val; | |
34dc7c2f BB |
1788 | } else { |
1789 | error = zap_update(os, moid, name, 8, 1, &val, tx); | |
1790 | } | |
1791 | ASSERT(error == 0); | |
1792 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) | |
1793 | norm = val; | |
22872ff5 BB |
1794 | else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) |
1795 | sense = val; | |
34dc7c2f BB |
1796 | } |
1797 | ASSERT(version != 0); | |
9babb374 | 1798 | error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); |
34dc7c2f | 1799 | |
428870ff BB |
1800 | /* |
1801 | * Create zap object used for SA attribute registration | |
1802 | */ | |
1803 | ||
1804 | if (version >= ZPL_VERSION_SA) { | |
1805 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1806 | DMU_OT_NONE, 0, tx); | |
1807 | error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
1808 | ASSERT(error == 0); | |
1809 | } else { | |
1810 | sa_obj = 0; | |
1811 | } | |
34dc7c2f BB |
1812 | /* |
1813 | * Create a delete queue. | |
1814 | */ | |
9babb374 | 1815 | obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); |
34dc7c2f | 1816 | |
9babb374 | 1817 | error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); |
34dc7c2f BB |
1818 | ASSERT(error == 0); |
1819 | ||
9babb374 | 1820 | /* |
22872ff5 BB |
1821 | * Create root znode. Create minimal znode/inode/zsb/sb |
1822 | * to allow zfs_mknode to work. | |
9babb374 | 1823 | */ |
22872ff5 BB |
1824 | vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID; |
1825 | vattr.va_mode = S_IFDIR|0755; | |
1826 | vattr.va_uid = crgetuid(cr); | |
1827 | vattr.va_gid = crgetgid(cr); | |
1828 | ||
79c76d5b | 1829 | rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
22872ff5 BB |
1830 | rootzp->z_moved = 0; |
1831 | rootzp->z_unlinked = 0; | |
1832 | rootzp->z_atime_dirty = 0; | |
1833 | rootzp->z_is_sa = USE_SA(version, os); | |
1834 | ||
79c76d5b | 1835 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
22872ff5 BB |
1836 | zsb->z_os = os; |
1837 | zsb->z_parent = zsb; | |
1838 | zsb->z_version = version; | |
1839 | zsb->z_use_fuids = USE_FUIDS(version, os); | |
1840 | zsb->z_use_sa = USE_SA(version, os); | |
1841 | zsb->z_norm = norm; | |
1842 | ||
79c76d5b | 1843 | sb = kmem_zalloc(sizeof (struct super_block), KM_SLEEP); |
22872ff5 BB |
1844 | sb->s_fs_info = zsb; |
1845 | ||
1846 | ZTOI(rootzp)->i_sb = sb; | |
1847 | ||
1848 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, | |
1849 | &zsb->z_attr_table); | |
9babb374 | 1850 | |
22872ff5 | 1851 | ASSERT(error == 0); |
9babb374 | 1852 | |
60101509 | 1853 | /* |
22872ff5 BB |
1854 | * Fold case on file systems that are always or sometimes case |
1855 | * insensitive. | |
60101509 | 1856 | */ |
22872ff5 BB |
1857 | if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) |
1858 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
60101509 | 1859 | |
22872ff5 BB |
1860 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
1861 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
1862 | offsetof(znode_t, z_link_node)); | |
60101509 | 1863 | |
c96c36fa BB |
1864 | size = MIN(1 << (highbit64(zfs_object_mutex_size)-1), ZFS_OBJ_MTX_MAX); |
1865 | zsb->z_hold_size = size; | |
1866 | zsb->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, KM_SLEEP); | |
1867 | zsb->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
1868 | for (i = 0; i != size; i++) { | |
1869 | avl_create(&zsb->z_hold_trees[i], zfs_znode_hold_compare, | |
1870 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); | |
1871 | mutex_init(&zsb->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); | |
1872 | } | |
60101509 | 1873 | |
22872ff5 BB |
1874 | VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, |
1875 | cr, NULL, &acl_ids)); | |
1876 | zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); | |
1877 | ASSERT3P(zp, ==, rootzp); | |
1878 | error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); | |
1879 | ASSERT(error == 0); | |
1880 | zfs_acl_ids_free(&acl_ids); | |
60101509 | 1881 | |
22872ff5 BB |
1882 | atomic_set(&ZTOI(rootzp)->i_count, 0); |
1883 | sa_handle_destroy(rootzp->z_sa_hdl); | |
22872ff5 BB |
1884 | kmem_cache_free(znode_cache, rootzp); |
1885 | ||
1886 | /* | |
1887 | * Create shares directory | |
1888 | */ | |
22872ff5 | 1889 | error = zfs_create_share_dir(zsb, tx); |
9babb374 | 1890 | ASSERT(error == 0); |
428870ff | 1891 | |
c96c36fa BB |
1892 | for (i = 0; i != size; i++) { |
1893 | avl_destroy(&zsb->z_hold_trees[i]); | |
1894 | mutex_destroy(&zsb->z_hold_locks[i]); | |
1895 | } | |
2708f716 | 1896 | |
c96c36fa BB |
1897 | vmem_free(zsb->z_hold_trees, sizeof (avl_tree_t) * size); |
1898 | vmem_free(zsb->z_hold_locks, sizeof (kmutex_t) * size); | |
2708f716 BB |
1899 | kmem_free(sb, sizeof (struct super_block)); |
1900 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
34dc7c2f | 1901 | } |
34dc7c2f | 1902 | #endif /* _KERNEL */ |
428870ff | 1903 | |
34dc7c2f | 1904 | static int |
572e2857 BB |
1905 | zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) |
1906 | { | |
1907 | uint64_t sa_obj = 0; | |
1908 | int error; | |
1909 | ||
1910 | error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
1911 | if (error != 0 && error != ENOENT) | |
1912 | return (error); | |
1913 | ||
1914 | error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); | |
1915 | return (error); | |
1916 | } | |
1917 | ||
1918 | static int | |
1919 | zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, | |
7b8518cb | 1920 | dmu_buf_t **db, void *tag) |
34dc7c2f | 1921 | { |
34dc7c2f | 1922 | dmu_object_info_t doi; |
34dc7c2f | 1923 | int error; |
428870ff | 1924 | |
7b8518cb | 1925 | if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) |
34dc7c2f BB |
1926 | return (error); |
1927 | ||
572e2857 | 1928 | dmu_object_info_from_db(*db, &doi); |
428870ff BB |
1929 | if ((doi.doi_bonus_type != DMU_OT_SA && |
1930 | doi.doi_bonus_type != DMU_OT_ZNODE) || | |
d6320ddb BB |
1931 | (doi.doi_bonus_type == DMU_OT_ZNODE && |
1932 | doi.doi_bonus_size < sizeof (znode_phys_t))) { | |
7b8518cb | 1933 | sa_buf_rele(*db, tag); |
2e528b49 | 1934 | return (SET_ERROR(ENOTSUP)); |
34dc7c2f BB |
1935 | } |
1936 | ||
572e2857 BB |
1937 | error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); |
1938 | if (error != 0) { | |
7b8518cb | 1939 | sa_buf_rele(*db, tag); |
428870ff BB |
1940 | return (error); |
1941 | } | |
1942 | ||
572e2857 BB |
1943 | return (0); |
1944 | } | |
1945 | ||
1946 | void | |
7b8518cb | 1947 | zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) |
572e2857 BB |
1948 | { |
1949 | sa_handle_destroy(hdl); | |
7b8518cb | 1950 | sa_buf_rele(db, tag); |
572e2857 BB |
1951 | } |
1952 | ||
1953 | /* | |
1954 | * Given an object number, return its parent object number and whether | |
1955 | * or not the object is an extended attribute directory. | |
1956 | */ | |
1957 | static int | |
b23ad7f3 JJ |
1958 | zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, |
1959 | uint64_t *pobjp, int *is_xattrdir) | |
572e2857 BB |
1960 | { |
1961 | uint64_t parent; | |
1962 | uint64_t pflags; | |
1963 | uint64_t mode; | |
b23ad7f3 | 1964 | uint64_t parent_mode; |
572e2857 | 1965 | sa_bulk_attr_t bulk[3]; |
b23ad7f3 JJ |
1966 | sa_handle_t *sa_hdl; |
1967 | dmu_buf_t *sa_db; | |
572e2857 BB |
1968 | int count = 0; |
1969 | int error; | |
1970 | ||
1971 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, | |
1972 | &parent, sizeof (parent)); | |
428870ff | 1973 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, |
572e2857 | 1974 | &pflags, sizeof (pflags)); |
428870ff | 1975 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, |
572e2857 | 1976 | &mode, sizeof (mode)); |
428870ff | 1977 | |
572e2857 | 1978 | if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) |
428870ff | 1979 | return (error); |
572e2857 | 1980 | |
b23ad7f3 JJ |
1981 | /* |
1982 | * When a link is removed its parent pointer is not changed and will | |
1983 | * be invalid. There are two cases where a link is removed but the | |
1984 | * file stays around, when it goes to the delete queue and when there | |
1985 | * are additional links. | |
1986 | */ | |
1987 | error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); | |
1988 | if (error != 0) | |
1989 | return (error); | |
1990 | ||
1991 | error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); | |
1992 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); | |
1993 | if (error != 0) | |
1994 | return (error); | |
1995 | ||
428870ff | 1996 | *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); |
34dc7c2f | 1997 | |
b23ad7f3 JJ |
1998 | /* |
1999 | * Extended attributes can be applied to files, directories, etc. | |
2000 | * Otherwise the parent must be a directory. | |
2001 | */ | |
2002 | if (!*is_xattrdir && !S_ISDIR(parent_mode)) | |
2003 | return (EINVAL); | |
2004 | ||
2005 | *pobjp = parent; | |
2006 | ||
34dc7c2f BB |
2007 | return (0); |
2008 | } | |
2009 | ||
572e2857 BB |
2010 | /* |
2011 | * Given an object number, return some zpl level statistics | |
2012 | */ | |
2013 | static int | |
2014 | zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, | |
2015 | zfs_stat_t *sb) | |
34dc7c2f | 2016 | { |
572e2857 BB |
2017 | sa_bulk_attr_t bulk[4]; |
2018 | int count = 0; | |
2019 | ||
2020 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, | |
2021 | &sb->zs_mode, sizeof (sb->zs_mode)); | |
2022 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, | |
2023 | &sb->zs_gen, sizeof (sb->zs_gen)); | |
2024 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, | |
2025 | &sb->zs_links, sizeof (sb->zs_links)); | |
2026 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, | |
2027 | &sb->zs_ctime, sizeof (sb->zs_ctime)); | |
2028 | ||
2029 | return (sa_bulk_lookup(hdl, bulk, count)); | |
2030 | } | |
2031 | ||
2032 | static int | |
2033 | zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, | |
2034 | sa_attr_type_t *sa_table, char *buf, int len) | |
2035 | { | |
2036 | sa_handle_t *sa_hdl; | |
2037 | sa_handle_t *prevhdl = NULL; | |
2038 | dmu_buf_t *prevdb = NULL; | |
2039 | dmu_buf_t *sa_db = NULL; | |
34dc7c2f BB |
2040 | char *path = buf + len - 1; |
2041 | int error; | |
2042 | ||
2043 | *path = '\0'; | |
572e2857 | 2044 | sa_hdl = hdl; |
428870ff | 2045 | |
34dc7c2f | 2046 | for (;;) { |
17897ce2 | 2047 | uint64_t pobj = 0; |
34dc7c2f BB |
2048 | char component[MAXNAMELEN + 2]; |
2049 | size_t complen; | |
17897ce2 | 2050 | int is_xattrdir = 0; |
34dc7c2f | 2051 | |
572e2857 | 2052 | if (prevdb) |
7b8518cb | 2053 | zfs_release_sa_handle(prevhdl, prevdb, FTAG); |
572e2857 | 2054 | |
b23ad7f3 | 2055 | if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, |
572e2857 | 2056 | &is_xattrdir)) != 0) |
34dc7c2f BB |
2057 | break; |
2058 | ||
2059 | if (pobj == obj) { | |
2060 | if (path[0] != '/') | |
2061 | *--path = '/'; | |
2062 | break; | |
2063 | } | |
2064 | ||
2065 | component[0] = '/'; | |
2066 | if (is_xattrdir) { | |
2067 | (void) sprintf(component + 1, "<xattrdir>"); | |
2068 | } else { | |
2069 | error = zap_value_search(osp, pobj, obj, | |
2070 | ZFS_DIRENT_OBJ(-1ULL), component + 1); | |
2071 | if (error != 0) | |
2072 | break; | |
2073 | } | |
2074 | ||
2075 | complen = strlen(component); | |
2076 | path -= complen; | |
2077 | ASSERT(path >= buf); | |
2078 | bcopy(component, path, complen); | |
2079 | obj = pobj; | |
572e2857 BB |
2080 | |
2081 | if (sa_hdl != hdl) { | |
2082 | prevhdl = sa_hdl; | |
2083 | prevdb = sa_db; | |
2084 | } | |
7b8518cb | 2085 | error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); |
572e2857 BB |
2086 | if (error != 0) { |
2087 | sa_hdl = prevhdl; | |
2088 | sa_db = prevdb; | |
2089 | break; | |
2090 | } | |
2091 | } | |
2092 | ||
2093 | if (sa_hdl != NULL && sa_hdl != hdl) { | |
2094 | ASSERT(sa_db != NULL); | |
7b8518cb | 2095 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); |
34dc7c2f BB |
2096 | } |
2097 | ||
2098 | if (error == 0) | |
2099 | (void) memmove(buf, path, buf + len - path); | |
428870ff | 2100 | |
34dc7c2f BB |
2101 | return (error); |
2102 | } | |
572e2857 BB |
2103 | |
2104 | int | |
2105 | zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) | |
2106 | { | |
2107 | sa_attr_type_t *sa_table; | |
2108 | sa_handle_t *hdl; | |
2109 | dmu_buf_t *db; | |
2110 | int error; | |
2111 | ||
2112 | error = zfs_sa_setup(osp, &sa_table); | |
2113 | if (error != 0) | |
2114 | return (error); | |
2115 | ||
7b8518cb | 2116 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2117 | if (error != 0) |
2118 | return (error); | |
2119 | ||
2120 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2121 | ||
7b8518cb | 2122 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2123 | return (error); |
2124 | } | |
2125 | ||
2126 | int | |
2127 | zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, | |
2128 | char *buf, int len) | |
2129 | { | |
2130 | char *path = buf + len - 1; | |
2131 | sa_attr_type_t *sa_table; | |
2132 | sa_handle_t *hdl; | |
2133 | dmu_buf_t *db; | |
2134 | int error; | |
2135 | ||
2136 | *path = '\0'; | |
2137 | ||
2138 | error = zfs_sa_setup(osp, &sa_table); | |
2139 | if (error != 0) | |
2140 | return (error); | |
2141 | ||
7b8518cb | 2142 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
2143 | if (error != 0) |
2144 | return (error); | |
2145 | ||
2146 | error = zfs_obj_to_stats_impl(hdl, sa_table, sb); | |
2147 | if (error != 0) { | |
7b8518cb | 2148 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2149 | return (error); |
2150 | } | |
2151 | ||
2152 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
2153 | ||
7b8518cb | 2154 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
2155 | return (error); |
2156 | } | |
c28b2279 BB |
2157 | |
2158 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
2159 | EXPORT_SYMBOL(zfs_create_fs); | |
2160 | EXPORT_SYMBOL(zfs_obj_to_path); | |
0720116d BB |
2161 | |
2162 | module_param(zfs_object_mutex_size, uint, 0644); | |
2163 | MODULE_PARM_DESC(zfs_object_mutex_size, "Size of znode hold array"); | |
c28b2279 | 2164 | #endif |