]>
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. |
2e528b49 | 23 | * Copyright (c) 2013 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; |
34dc7c2f | 98 | |
34dc7c2f BB |
99 | /*ARGSUSED*/ |
100 | static int | |
b128c09f | 101 | zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) |
34dc7c2f BB |
102 | { |
103 | znode_t *zp = buf; | |
104 | ||
3558fd73 | 105 | inode_init_once(ZTOI(zp)); |
b128c09f BB |
106 | list_link_init(&zp->z_link_node); |
107 | ||
34dc7c2f | 108 | mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); |
34dc7c2f BB |
109 | rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); |
110 | rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL); | |
111 | mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); | |
82a37189 | 112 | rw_init(&zp->z_xattr_lock, NULL, RW_DEFAULT, NULL); |
34dc7c2f BB |
113 | |
114 | mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL); | |
115 | avl_create(&zp->z_range_avl, zfs_range_compare, | |
116 | sizeof (rl_t), offsetof(rl_t, r_node)); | |
117 | ||
b128c09f | 118 | zp->z_dirlocks = NULL; |
45d1cae3 | 119 | zp->z_acl_cached = NULL; |
82a37189 | 120 | zp->z_xattr_cached = NULL; |
e89260a1 | 121 | zp->z_xattr_parent = 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); |
34dc7c2f BB |
138 | avl_destroy(&zp->z_range_avl); |
139 | mutex_destroy(&zp->z_range_lock); | |
140 | ||
b128c09f | 141 | ASSERT(zp->z_dirlocks == NULL); |
45d1cae3 | 142 | ASSERT(zp->z_acl_cached == NULL); |
82a37189 | 143 | ASSERT(zp->z_xattr_cached == NULL); |
e89260a1 | 144 | ASSERT(zp->z_xattr_parent == NULL); |
b128c09f BB |
145 | } |
146 | ||
34dc7c2f BB |
147 | void |
148 | zfs_znode_init(void) | |
149 | { | |
150 | /* | |
5074bfe8 TC |
151 | * Initialize zcache. The KMC_SLAB hint is used in order that it be |
152 | * backed by kmalloc() when on the Linux slab in order that any | |
153 | * wait_on_bit() operations on the related inode operate properly. | |
34dc7c2f BB |
154 | */ |
155 | ASSERT(znode_cache == NULL); | |
156 | znode_cache = kmem_cache_create("zfs_znode_cache", | |
157 | sizeof (znode_t), 0, zfs_znode_cache_constructor, | |
5074bfe8 | 158 | zfs_znode_cache_destructor, NULL, NULL, NULL, KMC_SLAB); |
34dc7c2f BB |
159 | } |
160 | ||
161 | void | |
162 | zfs_znode_fini(void) | |
163 | { | |
34dc7c2f BB |
164 | /* |
165 | * Cleanup zcache | |
166 | */ | |
167 | if (znode_cache) | |
168 | kmem_cache_destroy(znode_cache); | |
169 | znode_cache = NULL; | |
170 | } | |
171 | ||
34dc7c2f | 172 | int |
3558fd73 | 173 | zfs_create_share_dir(zfs_sb_t *zsb, dmu_tx_t *tx) |
34dc7c2f | 174 | { |
3c9609b3 | 175 | #ifdef HAVE_SMB_SHARE |
9babb374 BB |
176 | zfs_acl_ids_t acl_ids; |
177 | vattr_t vattr; | |
178 | znode_t *sharezp; | |
179 | vnode_t *vp; | |
180 | znode_t *zp; | |
181 | int error; | |
34dc7c2f | 182 | |
9babb374 | 183 | vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; |
3558fd73 | 184 | vattr.va_mode = S_IFDIR | 0555; |
9babb374 BB |
185 | vattr.va_uid = crgetuid(kcred); |
186 | vattr.va_gid = crgetgid(kcred); | |
34dc7c2f | 187 | |
79c76d5b | 188 | sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
572e2857 | 189 | sharezp->z_moved = 0; |
9babb374 BB |
190 | sharezp->z_unlinked = 0; |
191 | sharezp->z_atime_dirty = 0; | |
192 | sharezp->z_zfsvfs = zfsvfs; | |
428870ff | 193 | sharezp->z_is_sa = zfsvfs->z_use_sa; |
34dc7c2f | 194 | |
9babb374 BB |
195 | vp = ZTOV(sharezp); |
196 | vn_reinit(vp); | |
197 | vp->v_type = VDIR; | |
34dc7c2f | 198 | |
9babb374 BB |
199 | VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, |
200 | kcred, NULL, &acl_ids)); | |
428870ff | 201 | zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); |
9babb374 BB |
202 | ASSERT3P(zp, ==, sharezp); |
203 | ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */ | |
204 | POINTER_INVALIDATE(&sharezp->z_zfsvfs); | |
205 | error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, | |
206 | ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); | |
207 | zfsvfs->z_shares_dir = sharezp->z_id; | |
208 | ||
209 | zfs_acl_ids_free(&acl_ids); | |
3558fd73 | 210 | // ZTOV(sharezp)->v_count = 0; |
428870ff | 211 | sa_handle_destroy(sharezp->z_sa_hdl); |
9babb374 | 212 | kmem_cache_free(znode_cache, sharezp); |
34dc7c2f | 213 | |
9babb374 | 214 | return (error); |
9ee7fac5 BB |
215 | #else |
216 | return (0); | |
3c9609b3 | 217 | #endif /* HAVE_SMB_SHARE */ |
34dc7c2f BB |
218 | } |
219 | ||
34dc7c2f | 220 | static void |
3558fd73 | 221 | zfs_znode_sa_init(zfs_sb_t *zsb, znode_t *zp, |
428870ff | 222 | dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) |
34dc7c2f | 223 | { |
3558fd73 | 224 | ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zsb, zp->z_id))); |
34dc7c2f BB |
225 | |
226 | mutex_enter(&zp->z_lock); | |
227 | ||
428870ff BB |
228 | ASSERT(zp->z_sa_hdl == NULL); |
229 | ASSERT(zp->z_acl_cached == NULL); | |
230 | if (sa_hdl == NULL) { | |
3558fd73 | 231 | VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, zp, |
428870ff BB |
232 | SA_HDL_SHARED, &zp->z_sa_hdl)); |
233 | } else { | |
234 | zp->z_sa_hdl = sa_hdl; | |
235 | sa_set_userp(sa_hdl, zp); | |
236 | } | |
34dc7c2f | 237 | |
428870ff | 238 | zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; |
34dc7c2f | 239 | |
34dc7c2f | 240 | mutex_exit(&zp->z_lock); |
34dc7c2f BB |
241 | } |
242 | ||
243 | void | |
244 | zfs_znode_dmu_fini(znode_t *zp) | |
245 | { | |
3558fd73 | 246 | ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(ZTOZSB(zp), zp->z_id)) || |
b128c09f | 247 | zp->z_unlinked || |
3558fd73 | 248 | RW_WRITE_HELD(&ZTOZSB(zp)->z_teardown_inactive_lock)); |
428870ff BB |
249 | |
250 | sa_handle_destroy(zp->z_sa_hdl); | |
251 | zp->z_sa_hdl = NULL; | |
34dc7c2f BB |
252 | } |
253 | ||
254 | /* | |
3558fd73 BB |
255 | * Called by new_inode() to allocate a new inode. |
256 | */ | |
257 | int | |
258 | zfs_inode_alloc(struct super_block *sb, struct inode **ip) | |
259 | { | |
260 | znode_t *zp; | |
261 | ||
79c76d5b | 262 | zp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
3558fd73 BB |
263 | *ip = ZTOI(zp); |
264 | ||
265 | return (0); | |
266 | } | |
267 | ||
268 | /* | |
269 | * Called in multiple places when an inode should be destroyed. | |
270 | */ | |
271 | void | |
272 | zfs_inode_destroy(struct inode *ip) | |
273 | { | |
274 | znode_t *zp = ITOZ(ip); | |
275 | zfs_sb_t *zsb = ZTOZSB(zp); | |
276 | ||
ebe7e575 BB |
277 | if (zfsctl_is_node(ip)) |
278 | zfsctl_inode_destroy(ip); | |
279 | ||
3558fd73 | 280 | mutex_enter(&zsb->z_znodes_lock); |
7b3e34ba BB |
281 | if (list_link_active(&zp->z_link_node)) { |
282 | list_remove(&zsb->z_all_znodes, zp); | |
283 | zsb->z_nr_znodes--; | |
284 | } | |
3558fd73 BB |
285 | mutex_exit(&zsb->z_znodes_lock); |
286 | ||
287 | if (zp->z_acl_cached) { | |
288 | zfs_acl_free(zp->z_acl_cached); | |
289 | zp->z_acl_cached = NULL; | |
290 | } | |
291 | ||
82a37189 BB |
292 | if (zp->z_xattr_cached) { |
293 | nvlist_free(zp->z_xattr_cached); | |
294 | zp->z_xattr_cached = NULL; | |
295 | } | |
296 | ||
e89260a1 | 297 | if (zp->z_xattr_parent) { |
0a50679c | 298 | zfs_iput_async(ZTOI(zp->z_xattr_parent)); |
e89260a1 BB |
299 | zp->z_xattr_parent = NULL; |
300 | } | |
301 | ||
3558fd73 BB |
302 | kmem_cache_free(znode_cache, zp); |
303 | } | |
304 | ||
305 | static void | |
306 | zfs_inode_set_ops(zfs_sb_t *zsb, struct inode *ip) | |
307 | { | |
aa6d8c10 | 308 | uint64_t rdev = 0; |
3558fd73 BB |
309 | |
310 | switch (ip->i_mode & S_IFMT) { | |
311 | case S_IFREG: | |
312 | ip->i_op = &zpl_inode_operations; | |
313 | ip->i_fop = &zpl_file_operations; | |
314 | ip->i_mapping->a_ops = &zpl_address_space_operations; | |
315 | break; | |
316 | ||
317 | case S_IFDIR: | |
318 | ip->i_op = &zpl_dir_inode_operations; | |
319 | ip->i_fop = &zpl_dir_file_operations; | |
320 | ITOZ(ip)->z_zn_prefetch = B_TRUE; | |
321 | break; | |
322 | ||
323 | case S_IFLNK: | |
324 | ip->i_op = &zpl_symlink_inode_operations; | |
325 | break; | |
326 | ||
aa6d8c10 NB |
327 | /* |
328 | * rdev is only stored in a SA only for device files. | |
329 | */ | |
3558fd73 BB |
330 | case S_IFCHR: |
331 | case S_IFBLK: | |
3558fd73 BB |
332 | VERIFY(sa_lookup(ITOZ(ip)->z_sa_hdl, SA_ZPL_RDEV(zsb), |
333 | &rdev, sizeof (rdev)) == 0); | |
aa6d8c10 NB |
334 | /*FALLTHROUGH*/ |
335 | case S_IFIFO: | |
336 | case S_IFSOCK: | |
3558fd73 BB |
337 | init_special_inode(ip, ip->i_mode, rdev); |
338 | ip->i_op = &zpl_special_inode_operations; | |
339 | break; | |
340 | ||
341 | default: | |
342 | printk("ZFS: Invalid mode: 0x%x\n", ip->i_mode); | |
343 | VERIFY(0); | |
344 | } | |
345 | } | |
346 | ||
347 | /* | |
348 | * Construct a znode+inode and initialize. | |
34dc7c2f BB |
349 | * |
350 | * This does not do a call to dmu_set_user() that is | |
351 | * up to the caller to do, in case you don't want to | |
352 | * return the znode | |
353 | */ | |
354 | static znode_t * | |
3558fd73 | 355 | zfs_znode_alloc(zfs_sb_t *zsb, dmu_buf_t *db, int blksz, |
c85b224f | 356 | dmu_object_type_t obj_type, uint64_t obj, sa_handle_t *hdl, |
7b3e34ba | 357 | struct inode *dip) |
34dc7c2f BB |
358 | { |
359 | znode_t *zp; | |
3558fd73 | 360 | struct inode *ip; |
7f89ae6b | 361 | uint64_t mode; |
428870ff | 362 | uint64_t parent; |
428870ff BB |
363 | sa_bulk_attr_t bulk[9]; |
364 | int count = 0; | |
34dc7c2f | 365 | |
3558fd73 | 366 | ASSERT(zsb != NULL); |
34dc7c2f | 367 | |
3558fd73 BB |
368 | ip = new_inode(zsb->z_sb); |
369 | if (ip == NULL) | |
370 | return (NULL); | |
7304b6e5 | 371 | |
3558fd73 | 372 | zp = ITOZ(ip); |
34dc7c2f | 373 | ASSERT(zp->z_dirlocks == NULL); |
ebe7e575 BB |
374 | ASSERT3P(zp->z_acl_cached, ==, NULL); |
375 | ASSERT3P(zp->z_xattr_cached, ==, NULL); | |
e89260a1 | 376 | ASSERT3P(zp->z_xattr_parent, ==, NULL); |
572e2857 | 377 | zp->z_moved = 0; |
428870ff | 378 | zp->z_sa_hdl = NULL; |
34dc7c2f BB |
379 | zp->z_unlinked = 0; |
380 | zp->z_atime_dirty = 0; | |
381 | zp->z_mapcnt = 0; | |
34dc7c2f BB |
382 | zp->z_id = db->db_object; |
383 | zp->z_blksz = blksz; | |
384 | zp->z_seq = 0x7A4653; | |
385 | zp->z_sync_cnt = 0; | |
ebe7e575 BB |
386 | zp->z_is_zvol = B_FALSE; |
387 | zp->z_is_mapped = B_FALSE; | |
388 | zp->z_is_ctldir = B_FALSE; | |
7b3e34ba | 389 | zp->z_is_stale = B_FALSE; |
34dc7c2f | 390 | |
3558fd73 BB |
391 | zfs_znode_sa_init(zsb, zp, db, obj_type, hdl); |
392 | ||
7f89ae6b | 393 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, &mode, 8); |
3558fd73 BB |
394 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, &zp->z_gen, 8); |
395 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, &zp->z_size, 8); | |
396 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, &zp->z_links, 8); | |
397 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, | |
428870ff | 398 | &zp->z_pflags, 8); |
3558fd73 | 399 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zsb), NULL, |
7304b6e5 | 400 | &parent, 8); |
3558fd73 | 401 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, |
428870ff | 402 | &zp->z_atime, 16); |
3558fd73 BB |
403 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, &zp->z_uid, 8); |
404 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, &zp->z_gid, 8); | |
428870ff BB |
405 | |
406 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) { | |
407 | if (hdl == NULL) | |
408 | sa_handle_destroy(zp->z_sa_hdl); | |
34dc7c2f | 409 | |
3558fd73 | 410 | goto error; |
34dc7c2f | 411 | } |
7304b6e5 | 412 | |
7f89ae6b BB |
413 | zp->z_mode = mode; |
414 | ||
e89260a1 BB |
415 | /* |
416 | * xattr znodes hold a reference on their unique parent | |
417 | */ | |
418 | if (dip && zp->z_pflags & ZFS_XATTR) { | |
419 | igrab(dip); | |
420 | zp->z_xattr_parent = ITOZ(dip); | |
421 | } | |
422 | ||
3558fd73 | 423 | ip->i_ino = obj; |
5f35b190 | 424 | zfs_inode_update(zp); |
3558fd73 BB |
425 | zfs_inode_set_ops(zsb, ip); |
426 | ||
7b3e34ba BB |
427 | /* |
428 | * The only way insert_inode_locked() can fail is if the ip->i_ino | |
429 | * number is already hashed for this super block. This can never | |
430 | * happen because the inode numbers map 1:1 with the object numbers. | |
431 | * | |
432 | * The one exception is rolling back a mounted file system, but in | |
433 | * this case all the active inode are unhashed during the rollback. | |
434 | */ | |
435 | VERIFY3S(insert_inode_locked(ip), ==, 0); | |
c85b224f | 436 | |
3558fd73 BB |
437 | mutex_enter(&zsb->z_znodes_lock); |
438 | list_insert_tail(&zsb->z_all_znodes, zp); | |
ab26409d | 439 | zsb->z_nr_znodes++; |
b128c09f | 440 | membar_producer(); |
3558fd73 | 441 | mutex_exit(&zsb->z_znodes_lock); |
b128c09f | 442 | |
3558fd73 | 443 | unlock_new_inode(ip); |
34dc7c2f | 444 | return (zp); |
3558fd73 BB |
445 | |
446 | error: | |
447 | unlock_new_inode(ip); | |
448 | iput(ip); | |
d1d7e268 | 449 | return (NULL); |
34dc7c2f BB |
450 | } |
451 | ||
9d317793 RY |
452 | void |
453 | zfs_set_inode_flags(znode_t *zp, struct inode *ip) | |
454 | { | |
455 | /* | |
456 | * Linux and Solaris have different sets of file attributes, so we | |
457 | * restrict this conversion to the intersection of the two. | |
458 | */ | |
459 | ||
460 | if (zp->z_pflags & ZFS_IMMUTABLE) | |
461 | ip->i_flags |= S_IMMUTABLE; | |
462 | else | |
463 | ip->i_flags &= ~S_IMMUTABLE; | |
464 | ||
465 | if (zp->z_pflags & ZFS_APPENDONLY) | |
466 | ip->i_flags |= S_APPEND; | |
467 | else | |
468 | ip->i_flags &= ~S_APPEND; | |
469 | } | |
470 | ||
960e08fe BB |
471 | /* |
472 | * Update the embedded inode given the znode. We should work toward | |
473 | * eliminating this function as soon as possible by removing values | |
474 | * which are duplicated between the znode and inode. If the generic | |
475 | * inode has the correct field it should be used, and the ZFS code | |
476 | * updated to access the inode. This can be done incrementally. | |
477 | */ | |
478 | void | |
479 | zfs_inode_update(znode_t *zp) | |
480 | { | |
3558fd73 BB |
481 | zfs_sb_t *zsb; |
482 | struct inode *ip; | |
960e08fe BB |
483 | uint32_t blksize; |
484 | uint64_t atime[2], mtime[2], ctime[2]; | |
485 | ||
486 | ASSERT(zp != NULL); | |
3558fd73 BB |
487 | zsb = ZTOZSB(zp); |
488 | ip = ZTOI(zp); | |
489 | ||
ebe7e575 BB |
490 | /* Skip .zfs control nodes which do not exist on disk. */ |
491 | if (zfsctl_is_node(ip)) | |
492 | return; | |
493 | ||
3558fd73 BB |
494 | sa_lookup(zp->z_sa_hdl, SA_ZPL_ATIME(zsb), &atime, 16); |
495 | sa_lookup(zp->z_sa_hdl, SA_ZPL_MTIME(zsb), &mtime, 16); | |
496 | sa_lookup(zp->z_sa_hdl, SA_ZPL_CTIME(zsb), &ctime, 16); | |
497 | ||
498 | spin_lock(&ip->i_lock); | |
499 | ip->i_generation = zp->z_gen; | |
570d6edf RY |
500 | ip->i_uid = SUID_TO_KUID(zp->z_uid); |
501 | ip->i_gid = SGID_TO_KGID(zp->z_gid); | |
28eb9213 | 502 | set_nlink(ip, zp->z_links); |
3558fd73 | 503 | ip->i_mode = zp->z_mode; |
9d317793 | 504 | zfs_set_inode_flags(zp, ip); |
3558fd73 | 505 | ip->i_blkbits = SPA_MINBLOCKSHIFT; |
960e08fe | 506 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &blksize, |
3558fd73 | 507 | (u_longlong_t *)&ip->i_blocks); |
960e08fe | 508 | |
3558fd73 BB |
509 | ZFS_TIME_DECODE(&ip->i_atime, atime); |
510 | ZFS_TIME_DECODE(&ip->i_mtime, mtime); | |
511 | ZFS_TIME_DECODE(&ip->i_ctime, ctime); | |
960e08fe | 512 | |
3558fd73 BB |
513 | i_size_write(ip, zp->z_size); |
514 | spin_unlock(&ip->i_lock); | |
960e08fe BB |
515 | } |
516 | ||
1e8db771 BB |
517 | /* |
518 | * Safely mark an inode dirty. Inodes which are part of a read-only | |
519 | * file system or snapshot may not be dirtied. | |
520 | */ | |
521 | void | |
522 | zfs_mark_inode_dirty(struct inode *ip) | |
523 | { | |
524 | zfs_sb_t *zsb = ITOZSB(ip); | |
525 | ||
526 | if (zfs_is_readonly(zsb) || dmu_objset_is_snapshot(zsb->z_os)) | |
527 | return; | |
528 | ||
529 | mark_inode_dirty(ip); | |
530 | } | |
531 | ||
428870ff BB |
532 | static uint64_t empty_xattr; |
533 | static uint64_t pad[4]; | |
534 | static zfs_acl_phys_t acl_phys; | |
34dc7c2f BB |
535 | /* |
536 | * Create a new DMU object to hold a zfs znode. | |
537 | * | |
538 | * IN: dzp - parent directory for new znode | |
539 | * vap - file attributes for new znode | |
540 | * tx - dmu transaction id for zap operations | |
541 | * cr - credentials of caller | |
542 | * flag - flags: | |
543 | * IS_ROOT_NODE - new object will be root | |
544 | * IS_XATTR - new object is an attribute | |
34dc7c2f BB |
545 | * bonuslen - length of bonus buffer |
546 | * setaclp - File/Dir initial ACL | |
547 | * fuidp - Tracks fuid allocation. | |
548 | * | |
549 | * OUT: zpp - allocated znode | |
550 | * | |
551 | */ | |
552 | void | |
553 | zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, | |
428870ff | 554 | uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) |
34dc7c2f | 555 | { |
428870ff BB |
556 | uint64_t crtime[2], atime[2], mtime[2], ctime[2]; |
557 | uint64_t mode, size, links, parent, pflags; | |
572e2857 | 558 | uint64_t dzp_pflags = 0; |
428870ff | 559 | uint64_t rdev = 0; |
3558fd73 | 560 | zfs_sb_t *zsb = ZTOZSB(dzp); |
428870ff | 561 | dmu_buf_t *db; |
34dc7c2f BB |
562 | timestruc_t now; |
563 | uint64_t gen, obj; | |
428870ff BB |
564 | int bonuslen; |
565 | sa_handle_t *sa_hdl; | |
566 | dmu_object_type_t obj_type; | |
f30484af | 567 | sa_bulk_attr_t *sa_attrs; |
428870ff BB |
568 | int cnt = 0; |
569 | zfs_acl_locator_cb_t locate = { 0 }; | |
34dc7c2f | 570 | |
3558fd73 | 571 | if (zsb->z_replay) { |
34dc7c2f | 572 | obj = vap->va_nodeid; |
34dc7c2f BB |
573 | now = vap->va_ctime; /* see zfs_replay_create() */ |
574 | gen = vap->va_nblocks; /* ditto */ | |
575 | } else { | |
576 | obj = 0; | |
577 | gethrestime(&now); | |
578 | gen = dmu_tx_get_txg(tx); | |
579 | } | |
580 | ||
3558fd73 | 581 | obj_type = zsb->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; |
428870ff BB |
582 | bonuslen = (obj_type == DMU_OT_SA) ? |
583 | DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE; | |
584 | ||
34dc7c2f BB |
585 | /* |
586 | * Create a new DMU object. | |
587 | */ | |
588 | /* | |
589 | * There's currently no mechanism for pre-reading the blocks that will | |
572e2857 | 590 | * be needed to allocate a new object, so we accept the small chance |
34dc7c2f BB |
591 | * that there will be an i/o error and we will fail one of the |
592 | * assertions below. | |
593 | */ | |
3558fd73 BB |
594 | if (S_ISDIR(vap->va_mode)) { |
595 | if (zsb->z_replay) { | |
b0bc7a84 | 596 | VERIFY0(zap_create_claim_norm(zsb->z_os, obj, |
3558fd73 | 597 | zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS, |
b0bc7a84 | 598 | obj_type, bonuslen, tx)); |
34dc7c2f | 599 | } else { |
3558fd73 BB |
600 | obj = zap_create_norm(zsb->z_os, |
601 | zsb->z_norm, DMU_OT_DIRECTORY_CONTENTS, | |
428870ff | 602 | obj_type, bonuslen, tx); |
34dc7c2f BB |
603 | } |
604 | } else { | |
3558fd73 | 605 | if (zsb->z_replay) { |
b0bc7a84 | 606 | VERIFY0(dmu_object_claim(zsb->z_os, obj, |
34dc7c2f | 607 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
b0bc7a84 | 608 | obj_type, bonuslen, tx)); |
34dc7c2f | 609 | } else { |
3558fd73 | 610 | obj = dmu_object_alloc(zsb->z_os, |
34dc7c2f | 611 | DMU_OT_PLAIN_FILE_CONTENTS, 0, |
428870ff | 612 | obj_type, bonuslen, tx); |
34dc7c2f BB |
613 | } |
614 | } | |
34dc7c2f | 615 | |
3558fd73 BB |
616 | ZFS_OBJ_HOLD_ENTER(zsb, obj); |
617 | VERIFY(0 == sa_buf_hold(zsb->z_os, obj, NULL, &db)); | |
34dc7c2f BB |
618 | |
619 | /* | |
620 | * If this is the root, fix up the half-initialized parent pointer | |
621 | * to reference the just-allocated physical data area. | |
622 | */ | |
623 | if (flag & IS_ROOT_NODE) { | |
34dc7c2f | 624 | dzp->z_id = obj; |
428870ff BB |
625 | } else { |
626 | dzp_pflags = dzp->z_pflags; | |
34dc7c2f BB |
627 | } |
628 | ||
629 | /* | |
630 | * If parent is an xattr, so am I. | |
631 | */ | |
428870ff | 632 | if (dzp_pflags & ZFS_XATTR) { |
34dc7c2f | 633 | flag |= IS_XATTR; |
34dc7c2f BB |
634 | } |
635 | ||
3558fd73 | 636 | if (zsb->z_use_fuids) |
428870ff BB |
637 | pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; |
638 | else | |
639 | pflags = 0; | |
34dc7c2f | 640 | |
3558fd73 | 641 | if (S_ISDIR(vap->va_mode)) { |
428870ff BB |
642 | size = 2; /* contents ("." and "..") */ |
643 | links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; | |
644 | } else { | |
645 | size = links = 0; | |
34dc7c2f BB |
646 | } |
647 | ||
aa6d8c10 | 648 | if (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode)) |
dc1d7665 | 649 | rdev = vap->va_rdev; |
428870ff BB |
650 | |
651 | parent = dzp->z_id; | |
652 | mode = acl_ids->z_mode; | |
34dc7c2f | 653 | if (flag & IS_XATTR) |
428870ff | 654 | pflags |= ZFS_XATTR; |
34dc7c2f | 655 | |
428870ff BB |
656 | /* |
657 | * No execs denied will be deterimed when zfs_mode_compute() is called. | |
658 | */ | |
659 | pflags |= acl_ids->z_aclp->z_hints & | |
660 | (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| | |
661 | ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); | |
34dc7c2f | 662 | |
428870ff BB |
663 | ZFS_TIME_ENCODE(&now, crtime); |
664 | ZFS_TIME_ENCODE(&now, ctime); | |
34dc7c2f | 665 | |
3558fd73 | 666 | if (vap->va_mask & ATTR_ATIME) { |
428870ff | 667 | ZFS_TIME_ENCODE(&vap->va_atime, atime); |
34dc7c2f | 668 | } else { |
428870ff | 669 | ZFS_TIME_ENCODE(&now, atime); |
34dc7c2f BB |
670 | } |
671 | ||
3558fd73 | 672 | if (vap->va_mask & ATTR_MTIME) { |
428870ff BB |
673 | ZFS_TIME_ENCODE(&vap->va_mtime, mtime); |
674 | } else { | |
675 | ZFS_TIME_ENCODE(&now, mtime); | |
676 | } | |
677 | ||
678 | /* Now add in all of the "SA" attributes */ | |
3558fd73 | 679 | VERIFY(0 == sa_handle_get_from_db(zsb->z_os, db, NULL, SA_HDL_SHARED, |
428870ff BB |
680 | &sa_hdl)); |
681 | ||
682 | /* | |
683 | * Setup the array of attributes to be replaced/set on the new file | |
684 | * | |
685 | * order for DMU_OT_ZNODE is critical since it needs to be constructed | |
686 | * in the old znode_phys_t format. Don't change this ordering | |
687 | */ | |
79c76d5b | 688 | sa_attrs = kmem_alloc(sizeof (sa_bulk_attr_t) * ZPL_END, KM_SLEEP); |
428870ff BB |
689 | |
690 | if (obj_type == DMU_OT_ZNODE) { | |
3558fd73 | 691 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb), |
428870ff | 692 | NULL, &atime, 16); |
3558fd73 | 693 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb), |
428870ff | 694 | NULL, &mtime, 16); |
3558fd73 | 695 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb), |
428870ff | 696 | NULL, &ctime, 16); |
3558fd73 | 697 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb), |
428870ff | 698 | NULL, &crtime, 16); |
3558fd73 | 699 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb), |
428870ff | 700 | NULL, &gen, 8); |
3558fd73 | 701 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb), |
428870ff | 702 | NULL, &mode, 8); |
3558fd73 | 703 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb), |
428870ff | 704 | NULL, &size, 8); |
3558fd73 | 705 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb), |
428870ff | 706 | NULL, &parent, 8); |
34dc7c2f | 707 | } else { |
3558fd73 | 708 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zsb), |
428870ff | 709 | NULL, &mode, 8); |
3558fd73 | 710 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zsb), |
428870ff | 711 | NULL, &size, 8); |
3558fd73 | 712 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zsb), |
428870ff | 713 | NULL, &gen, 8); |
3558fd73 BB |
714 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), |
715 | NULL, &acl_ids->z_fuid, 8); | |
716 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), | |
717 | NULL, &acl_ids->z_fgid, 8); | |
718 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zsb), | |
428870ff | 719 | NULL, &parent, 8); |
3558fd73 | 720 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb), |
428870ff | 721 | NULL, &pflags, 8); |
3558fd73 | 722 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zsb), |
428870ff | 723 | NULL, &atime, 16); |
3558fd73 | 724 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zsb), |
428870ff | 725 | NULL, &mtime, 16); |
3558fd73 | 726 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zsb), |
428870ff | 727 | NULL, &ctime, 16); |
3558fd73 | 728 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zsb), |
428870ff BB |
729 | NULL, &crtime, 16); |
730 | } | |
731 | ||
3558fd73 | 732 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zsb), NULL, &links, 8); |
428870ff BB |
733 | |
734 | if (obj_type == DMU_OT_ZNODE) { | |
3558fd73 | 735 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zsb), NULL, |
428870ff | 736 | &empty_xattr, 8); |
34dc7c2f | 737 | } |
428870ff | 738 | if (obj_type == DMU_OT_ZNODE || |
aa6d8c10 | 739 | (S_ISBLK(vap->va_mode) || S_ISCHR(vap->va_mode))) { |
3558fd73 | 740 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zsb), |
428870ff | 741 | NULL, &rdev, 8); |
428870ff BB |
742 | } |
743 | if (obj_type == DMU_OT_ZNODE) { | |
3558fd73 | 744 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zsb), |
428870ff | 745 | NULL, &pflags, 8); |
3558fd73 | 746 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zsb), NULL, |
428870ff | 747 | &acl_ids->z_fuid, 8); |
3558fd73 | 748 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zsb), NULL, |
428870ff | 749 | &acl_ids->z_fgid, 8); |
3558fd73 | 750 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zsb), NULL, pad, |
428870ff | 751 | sizeof (uint64_t) * 4); |
3558fd73 | 752 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zsb), NULL, |
428870ff BB |
753 | &acl_phys, sizeof (zfs_acl_phys_t)); |
754 | } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { | |
3558fd73 | 755 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zsb), NULL, |
428870ff BB |
756 | &acl_ids->z_aclp->z_acl_count, 8); |
757 | locate.cb_aclp = acl_ids->z_aclp; | |
3558fd73 | 758 | SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zsb), |
428870ff BB |
759 | zfs_acl_data_locator, &locate, |
760 | acl_ids->z_aclp->z_acl_bytes); | |
761 | mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, | |
762 | acl_ids->z_fuid, acl_ids->z_fgid); | |
763 | } | |
764 | ||
765 | VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); | |
34dc7c2f | 766 | |
34dc7c2f | 767 | if (!(flag & IS_ROOT_NODE)) { |
c85b224f | 768 | *zpp = zfs_znode_alloc(zsb, db, 0, obj_type, obj, sa_hdl, |
7b3e34ba BB |
769 | ZTOI(dzp)); |
770 | VERIFY(*zpp != NULL); | |
771 | VERIFY(dzp != NULL); | |
34dc7c2f BB |
772 | } else { |
773 | /* | |
774 | * If we are creating the root node, the "parent" we | |
775 | * passed in is the znode for the root. | |
776 | */ | |
777 | *zpp = dzp; | |
428870ff BB |
778 | |
779 | (*zpp)->z_sa_hdl = sa_hdl; | |
34dc7c2f | 780 | } |
428870ff BB |
781 | |
782 | (*zpp)->z_pflags = pflags; | |
783 | (*zpp)->z_mode = mode; | |
784 | ||
428870ff BB |
785 | if (obj_type == DMU_OT_ZNODE || |
786 | acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { | |
b0bc7a84 | 787 | VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); |
428870ff | 788 | } |
d1d7e268 | 789 | kmem_free(sa_attrs, sizeof (sa_bulk_attr_t) * ZPL_END); |
3558fd73 | 790 | ZFS_OBJ_HOLD_EXIT(zsb, obj); |
34dc7c2f BB |
791 | } |
792 | ||
5484965a | 793 | /* |
d3cc8b15 WA |
794 | * Update in-core attributes. It is assumed the caller will be doing an |
795 | * sa_bulk_update to push the changes out. | |
5484965a BB |
796 | */ |
797 | void | |
798 | zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) | |
799 | { | |
800 | xoptattr_t *xoap; | |
801 | ||
802 | xoap = xva_getxoptattr(xvap); | |
803 | ASSERT(xoap); | |
804 | ||
805 | if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { | |
806 | uint64_t times[2]; | |
807 | ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); | |
808 | (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(ZTOZSB(zp)), | |
809 | ×, sizeof (times), tx); | |
810 | XVA_SET_RTN(xvap, XAT_CREATETIME); | |
811 | } | |
812 | if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { | |
813 | ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, | |
814 | zp->z_pflags, tx); | |
815 | XVA_SET_RTN(xvap, XAT_READONLY); | |
816 | } | |
817 | if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { | |
818 | ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, | |
819 | zp->z_pflags, tx); | |
820 | XVA_SET_RTN(xvap, XAT_HIDDEN); | |
821 | } | |
822 | if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { | |
823 | ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, | |
824 | zp->z_pflags, tx); | |
825 | XVA_SET_RTN(xvap, XAT_SYSTEM); | |
826 | } | |
827 | if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { | |
828 | ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, | |
829 | zp->z_pflags, tx); | |
830 | XVA_SET_RTN(xvap, XAT_ARCHIVE); | |
831 | } | |
832 | if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { | |
833 | ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, | |
834 | zp->z_pflags, tx); | |
835 | XVA_SET_RTN(xvap, XAT_IMMUTABLE); | |
836 | } | |
837 | if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { | |
838 | ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, | |
839 | zp->z_pflags, tx); | |
840 | XVA_SET_RTN(xvap, XAT_NOUNLINK); | |
841 | } | |
842 | if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { | |
843 | ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, | |
844 | zp->z_pflags, tx); | |
845 | XVA_SET_RTN(xvap, XAT_APPENDONLY); | |
846 | } | |
847 | if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { | |
848 | ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, | |
849 | zp->z_pflags, tx); | |
850 | XVA_SET_RTN(xvap, XAT_NODUMP); | |
851 | } | |
852 | if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { | |
853 | ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, | |
854 | zp->z_pflags, tx); | |
855 | XVA_SET_RTN(xvap, XAT_OPAQUE); | |
856 | } | |
857 | if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { | |
858 | ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, | |
859 | xoap->xoa_av_quarantined, zp->z_pflags, tx); | |
860 | XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); | |
861 | } | |
862 | if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { | |
863 | ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, | |
864 | zp->z_pflags, tx); | |
865 | XVA_SET_RTN(xvap, XAT_AV_MODIFIED); | |
866 | } | |
867 | if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { | |
868 | zfs_sa_set_scanstamp(zp, xvap, tx); | |
869 | XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); | |
870 | } | |
871 | if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { | |
872 | ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, | |
873 | zp->z_pflags, tx); | |
874 | XVA_SET_RTN(xvap, XAT_REPARSE); | |
875 | } | |
876 | if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { | |
877 | ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, | |
878 | zp->z_pflags, tx); | |
879 | XVA_SET_RTN(xvap, XAT_OFFLINE); | |
880 | } | |
881 | if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { | |
882 | ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, | |
883 | zp->z_pflags, tx); | |
884 | XVA_SET_RTN(xvap, XAT_SPARSE); | |
885 | } | |
886 | } | |
887 | ||
34dc7c2f | 888 | int |
3558fd73 | 889 | zfs_zget(zfs_sb_t *zsb, uint64_t obj_num, znode_t **zpp) |
34dc7c2f BB |
890 | { |
891 | dmu_object_info_t doi; | |
892 | dmu_buf_t *db; | |
893 | znode_t *zp; | |
894 | int err; | |
428870ff | 895 | sa_handle_t *hdl; |
34dc7c2f BB |
896 | |
897 | *zpp = NULL; | |
898 | ||
6f9548c4 | 899 | again: |
3558fd73 | 900 | ZFS_OBJ_HOLD_ENTER(zsb, obj_num); |
34dc7c2f | 901 | |
3558fd73 | 902 | err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db); |
34dc7c2f | 903 | if (err) { |
3558fd73 | 904 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); |
34dc7c2f BB |
905 | return (err); |
906 | } | |
907 | ||
908 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
909 | if (doi.doi_bonus_type != DMU_OT_SA && |
910 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
911 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
912 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
913 | sa_buf_rele(db, NULL); | |
3558fd73 | 914 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); |
2e528b49 | 915 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
916 | } |
917 | ||
428870ff BB |
918 | hdl = dmu_buf_get_user(db); |
919 | if (hdl != NULL) { | |
36df2843 | 920 | zp = sa_get_userdata(hdl); |
34dc7c2f | 921 | |
8ac67298 | 922 | |
34dc7c2f | 923 | /* |
428870ff BB |
924 | * Since "SA" does immediate eviction we |
925 | * should never find a sa handle that doesn't | |
926 | * know about the znode. | |
34dc7c2f | 927 | */ |
428870ff BB |
928 | |
929 | ASSERT3P(zp, !=, NULL); | |
930 | ||
931 | mutex_enter(&zp->z_lock); | |
34dc7c2f BB |
932 | ASSERT3U(zp->z_id, ==, obj_num); |
933 | if (zp->z_unlinked) { | |
2e528b49 | 934 | err = SET_ERROR(ENOENT); |
34dc7c2f | 935 | } else { |
6f9548c4 RY |
936 | /* |
937 | * If igrab() returns NULL the VFS has independently | |
938 | * determined the inode should be evicted and has | |
939 | * called iput_final() to start the eviction process. | |
940 | * The SA handle is still valid but because the VFS | |
941 | * requires that the eviction succeed we must drop | |
942 | * our locks and references to allow the eviction to | |
943 | * complete. The zfs_zget() may then be retried. | |
944 | * | |
945 | * This unlikely case could be optimized by registering | |
946 | * a sops->drop_inode() callback. The callback would | |
947 | * need to detect the active SA hold thereby informing | |
948 | * the VFS that this inode should not be evicted. | |
949 | */ | |
950 | if (igrab(ZTOI(zp)) == NULL) { | |
951 | mutex_exit(&zp->z_lock); | |
952 | sa_buf_rele(db, NULL); | |
953 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); | |
6102d037 | 954 | /* inode might need this to finish evict */ |
955 | cond_resched(); | |
6f9548c4 RY |
956 | goto again; |
957 | } | |
34dc7c2f BB |
958 | *zpp = zp; |
959 | err = 0; | |
960 | } | |
34dc7c2f | 961 | mutex_exit(&zp->z_lock); |
f3ad9cd6 | 962 | sa_buf_rele(db, NULL); |
3558fd73 | 963 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); |
34dc7c2f BB |
964 | return (err); |
965 | } | |
966 | ||
967 | /* | |
3558fd73 | 968 | * Not found create new znode/vnode but only if file exists. |
428870ff BB |
969 | * |
970 | * There is a small window where zfs_vget() could | |
971 | * find this object while a file create is still in | |
972 | * progress. This is checked for in zfs_znode_alloc() | |
973 | * | |
974 | * if zfs_znode_alloc() fails it will drop the hold on the | |
975 | * bonus buffer. | |
34dc7c2f | 976 | */ |
3558fd73 | 977 | zp = zfs_znode_alloc(zsb, db, doi.doi_data_block_size, |
7b3e34ba | 978 | doi.doi_bonus_type, obj_num, NULL, NULL); |
428870ff | 979 | if (zp == NULL) { |
2e528b49 | 980 | err = SET_ERROR(ENOENT); |
428870ff BB |
981 | } else { |
982 | *zpp = zp; | |
983 | } | |
3558fd73 | 984 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); |
428870ff | 985 | return (err); |
34dc7c2f BB |
986 | } |
987 | ||
988 | int | |
989 | zfs_rezget(znode_t *zp) | |
990 | { | |
3558fd73 | 991 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
992 | dmu_object_info_t doi; |
993 | dmu_buf_t *db; | |
994 | uint64_t obj_num = zp->z_id; | |
428870ff | 995 | uint64_t mode; |
428870ff | 996 | sa_bulk_attr_t bulk[8]; |
34dc7c2f | 997 | int err; |
428870ff BB |
998 | int count = 0; |
999 | uint64_t gen; | |
34dc7c2f | 1000 | |
3558fd73 | 1001 | ZFS_OBJ_HOLD_ENTER(zsb, obj_num); |
34dc7c2f | 1002 | |
428870ff BB |
1003 | mutex_enter(&zp->z_acl_lock); |
1004 | if (zp->z_acl_cached) { | |
1005 | zfs_acl_free(zp->z_acl_cached); | |
1006 | zp->z_acl_cached = NULL; | |
1007 | } | |
428870ff | 1008 | mutex_exit(&zp->z_acl_lock); |
7b3e34ba BB |
1009 | |
1010 | rw_enter(&zp->z_xattr_lock, RW_WRITER); | |
1011 | if (zp->z_xattr_cached) { | |
1012 | nvlist_free(zp->z_xattr_cached); | |
1013 | zp->z_xattr_cached = NULL; | |
1014 | } | |
1015 | ||
1016 | if (zp->z_xattr_parent) { | |
2d501583 | 1017 | zfs_iput_async(ZTOI(zp->z_xattr_parent)); |
7b3e34ba BB |
1018 | zp->z_xattr_parent = NULL; |
1019 | } | |
1020 | rw_exit(&zp->z_xattr_lock); | |
1021 | ||
428870ff | 1022 | ASSERT(zp->z_sa_hdl == NULL); |
3558fd73 | 1023 | err = sa_buf_hold(zsb->z_os, obj_num, NULL, &db); |
34dc7c2f | 1024 | if (err) { |
3558fd73 | 1025 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); |
34dc7c2f BB |
1026 | return (err); |
1027 | } | |
1028 | ||
1029 | dmu_object_info_from_db(db, &doi); | |
428870ff BB |
1030 | if (doi.doi_bonus_type != DMU_OT_SA && |
1031 | (doi.doi_bonus_type != DMU_OT_ZNODE || | |
1032 | (doi.doi_bonus_type == DMU_OT_ZNODE && | |
1033 | doi.doi_bonus_size < sizeof (znode_phys_t)))) { | |
1034 | sa_buf_rele(db, NULL); | |
3558fd73 | 1035 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); |
2e528b49 | 1036 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1037 | } |
1038 | ||
3558fd73 | 1039 | zfs_znode_sa_init(zsb, zp, db, doi.doi_bonus_type, NULL); |
428870ff BB |
1040 | |
1041 | /* reload cached values */ | |
3558fd73 | 1042 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zsb), NULL, |
428870ff | 1043 | &gen, sizeof (gen)); |
3558fd73 | 1044 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), NULL, |
428870ff | 1045 | &zp->z_size, sizeof (zp->z_size)); |
3558fd73 | 1046 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zsb), NULL, |
428870ff | 1047 | &zp->z_links, sizeof (zp->z_links)); |
3558fd73 | 1048 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), NULL, |
428870ff | 1049 | &zp->z_pflags, sizeof (zp->z_pflags)); |
3558fd73 | 1050 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zsb), NULL, |
428870ff | 1051 | &zp->z_atime, sizeof (zp->z_atime)); |
3558fd73 | 1052 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zsb), NULL, |
572e2857 | 1053 | &zp->z_uid, sizeof (zp->z_uid)); |
3558fd73 | 1054 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zsb), NULL, |
572e2857 | 1055 | &zp->z_gid, sizeof (zp->z_gid)); |
3558fd73 | 1056 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zsb), NULL, |
428870ff BB |
1057 | &mode, sizeof (mode)); |
1058 | ||
428870ff BB |
1059 | if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { |
1060 | zfs_znode_dmu_fini(zp); | |
3558fd73 | 1061 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); |
2e528b49 | 1062 | return (SET_ERROR(EIO)); |
428870ff BB |
1063 | } |
1064 | ||
572e2857 BB |
1065 | zp->z_mode = mode; |
1066 | ||
428870ff BB |
1067 | if (gen != zp->z_gen) { |
1068 | zfs_znode_dmu_fini(zp); | |
3558fd73 | 1069 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); |
2e528b49 | 1070 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1071 | } |
1072 | ||
428870ff | 1073 | zp->z_unlinked = (zp->z_links == 0); |
34dc7c2f | 1074 | zp->z_blksz = doi.doi_data_block_size; |
7b3e34ba | 1075 | zfs_inode_update(zp); |
34dc7c2f | 1076 | |
3558fd73 | 1077 | ZFS_OBJ_HOLD_EXIT(zsb, obj_num); |
34dc7c2f BB |
1078 | |
1079 | return (0); | |
1080 | } | |
1081 | ||
1082 | void | |
1083 | zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) | |
1084 | { | |
3558fd73 BB |
1085 | zfs_sb_t *zsb = ZTOZSB(zp); |
1086 | objset_t *os = zsb->z_os; | |
34dc7c2f | 1087 | uint64_t obj = zp->z_id; |
572e2857 | 1088 | uint64_t acl_obj = zfs_external_acl(zp); |
34dc7c2f | 1089 | |
3558fd73 | 1090 | ZFS_OBJ_HOLD_ENTER(zsb, obj); |
572e2857 BB |
1091 | if (acl_obj) { |
1092 | VERIFY(!zp->z_is_sa); | |
b128c09f | 1093 | VERIFY(0 == dmu_object_free(os, acl_obj, tx)); |
572e2857 | 1094 | } |
b128c09f | 1095 | VERIFY(0 == dmu_object_free(os, obj, tx)); |
34dc7c2f | 1096 | zfs_znode_dmu_fini(zp); |
3558fd73 | 1097 | ZFS_OBJ_HOLD_EXIT(zsb, obj); |
34dc7c2f BB |
1098 | } |
1099 | ||
1100 | void | |
1101 | zfs_zinactive(znode_t *zp) | |
1102 | { | |
3558fd73 | 1103 | zfs_sb_t *zsb = ZTOZSB(zp); |
34dc7c2f BB |
1104 | uint64_t z_id = zp->z_id; |
1105 | ||
428870ff | 1106 | ASSERT(zp->z_sa_hdl); |
34dc7c2f BB |
1107 | |
1108 | /* | |
d6bd8eaa | 1109 | * Don't allow a zfs_zget() while were trying to release this znode. |
34dc7c2f | 1110 | */ |
40d06e3c | 1111 | ZFS_OBJ_HOLD_ENTER(zsb, z_id); |
d6bd8eaa | 1112 | |
34dc7c2f | 1113 | mutex_enter(&zp->z_lock); |
34dc7c2f BB |
1114 | |
1115 | /* | |
1116 | * If this was the last reference to a file with no links, | |
1117 | * remove the file from the file system. | |
1118 | */ | |
1119 | if (zp->z_unlinked) { | |
1120 | mutex_exit(&zp->z_lock); | |
d6bd8eaa | 1121 | |
40d06e3c | 1122 | ZFS_OBJ_HOLD_EXIT(zsb, z_id); |
d6bd8eaa | 1123 | |
34dc7c2f BB |
1124 | zfs_rmnode(zp); |
1125 | return; | |
1126 | } | |
428870ff | 1127 | |
34dc7c2f BB |
1128 | mutex_exit(&zp->z_lock); |
1129 | zfs_znode_dmu_fini(zp); | |
d6bd8eaa | 1130 | |
40d06e3c | 1131 | ZFS_OBJ_HOLD_EXIT(zsb, z_id); |
34dc7c2f BB |
1132 | } |
1133 | ||
6d111134 TC |
1134 | static inline int |
1135 | zfs_compare_timespec(struct timespec *t1, struct timespec *t2) | |
1136 | { | |
1137 | if (t1->tv_sec < t2->tv_sec) | |
1138 | return (-1); | |
1139 | ||
1140 | if (t1->tv_sec > t2->tv_sec) | |
1141 | return (1); | |
1142 | ||
1143 | return (t1->tv_nsec - t2->tv_nsec); | |
1144 | } | |
1145 | ||
1146 | /* | |
1147 | * Determine whether the znode's atime must be updated. The logic mostly | |
1148 | * duplicates the Linux kernel's relatime_need_update() functionality. | |
1149 | * This function is only called if the underlying filesystem actually has | |
1150 | * atime updates enabled. | |
1151 | */ | |
1152 | static inline boolean_t | |
1153 | zfs_atime_need_update(znode_t *zp, timestruc_t *now) | |
1154 | { | |
1155 | if (!ZTOZSB(zp)->z_relatime) | |
1156 | return (B_TRUE); | |
1157 | ||
1158 | /* | |
1159 | * In relatime mode, only update the atime if the previous atime | |
1160 | * is earlier than either the ctime or mtime or if at least a day | |
1161 | * has passed since the last update of atime. | |
1162 | */ | |
1163 | if (zfs_compare_timespec(&ZTOI(zp)->i_mtime, &ZTOI(zp)->i_atime) >= 0) | |
1164 | return (B_TRUE); | |
1165 | ||
1166 | if (zfs_compare_timespec(&ZTOI(zp)->i_ctime, &ZTOI(zp)->i_atime) >= 0) | |
1167 | return (B_TRUE); | |
1168 | ||
1169 | if ((long)now->tv_sec - ZTOI(zp)->i_atime.tv_sec >= 24*60*60) | |
1170 | return (B_TRUE); | |
1171 | ||
1172 | return (B_FALSE); | |
1173 | } | |
1174 | ||
1175 | /* | |
1176 | * Prepare to update znode time stamps. | |
1177 | * | |
1178 | * IN: zp - znode requiring timestamp update | |
1179 | * flag - ATTR_MTIME, ATTR_CTIME, ATTR_ATIME flags | |
1180 | * have_tx - true of caller is creating a new txg | |
1181 | * | |
1182 | * OUT: zp - new atime (via underlying inode's i_atime) | |
1183 | * mtime - new mtime | |
1184 | * ctime - new ctime | |
1185 | * | |
1186 | * NOTE: The arguments are somewhat redundant. The following condition | |
1187 | * is always true: | |
1188 | * | |
1189 | * have_tx == !(flag & ATTR_ATIME) | |
1190 | */ | |
34dc7c2f | 1191 | void |
428870ff BB |
1192 | zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], |
1193 | uint64_t ctime[2], boolean_t have_tx) | |
34dc7c2f BB |
1194 | { |
1195 | timestruc_t now; | |
1196 | ||
6d111134 | 1197 | ASSERT(have_tx == !(flag & ATTR_ATIME)); |
34dc7c2f BB |
1198 | gethrestime(&now); |
1199 | ||
6d111134 TC |
1200 | /* |
1201 | * NOTE: The following test intentionally does not update z_atime_dirty | |
1202 | * in the case where an ATIME update has been requested but for which | |
1203 | * the update is omitted due to relatime logic. The rationale being | |
1204 | * that if the flag was set somewhere else, we should leave it alone | |
1205 | * here. | |
1206 | */ | |
1207 | if (flag & ATTR_ATIME) { | |
1208 | if (zfs_atime_need_update(zp, &now)) { | |
1209 | ZFS_TIME_ENCODE(&now, zp->z_atime); | |
1210 | ZTOI(zp)->i_atime.tv_sec = zp->z_atime[0]; | |
1211 | ZTOI(zp)->i_atime.tv_nsec = zp->z_atime[1]; | |
1212 | zp->z_atime_dirty = 1; | |
1213 | } | |
1214 | } else { | |
34dc7c2f BB |
1215 | zp->z_atime_dirty = 0; |
1216 | zp->z_seq++; | |
428870ff | 1217 | } |
34dc7c2f | 1218 | |
3558fd73 | 1219 | if (flag & ATTR_MTIME) { |
428870ff | 1220 | ZFS_TIME_ENCODE(&now, mtime); |
3558fd73 | 1221 | if (ZTOZSB(zp)->z_use_fuids) { |
428870ff BB |
1222 | zp->z_pflags |= (ZFS_ARCHIVE | |
1223 | ZFS_AV_MODIFIED); | |
1224 | } | |
34dc7c2f BB |
1225 | } |
1226 | ||
3558fd73 | 1227 | if (flag & ATTR_CTIME) { |
428870ff | 1228 | ZFS_TIME_ENCODE(&now, ctime); |
3558fd73 | 1229 | if (ZTOZSB(zp)->z_use_fuids) |
428870ff | 1230 | zp->z_pflags |= ZFS_ARCHIVE; |
34dc7c2f BB |
1231 | } |
1232 | } | |
1233 | ||
34dc7c2f BB |
1234 | /* |
1235 | * Grow the block size for a file. | |
1236 | * | |
1237 | * IN: zp - znode of file to free data in. | |
1238 | * size - requested block size | |
1239 | * tx - open transaction. | |
1240 | * | |
1241 | * NOTE: this function assumes that the znode is write locked. | |
1242 | */ | |
1243 | void | |
1244 | zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) | |
1245 | { | |
1246 | int error; | |
1247 | u_longlong_t dummy; | |
1248 | ||
1249 | if (size <= zp->z_blksz) | |
1250 | return; | |
1251 | /* | |
1252 | * If the file size is already greater than the current blocksize, | |
1253 | * we will not grow. If there is more than one block in a file, | |
1254 | * the blocksize cannot change. | |
1255 | */ | |
428870ff | 1256 | if (zp->z_blksz && zp->z_size > zp->z_blksz) |
34dc7c2f BB |
1257 | return; |
1258 | ||
3558fd73 | 1259 | error = dmu_object_set_blocksize(ZTOZSB(zp)->z_os, zp->z_id, |
34dc7c2f | 1260 | size, 0, tx); |
428870ff | 1261 | |
34dc7c2f BB |
1262 | if (error == ENOTSUP) |
1263 | return; | |
c99c9001 | 1264 | ASSERT0(error); |
34dc7c2f BB |
1265 | |
1266 | /* What blocksize did we actually get? */ | |
428870ff | 1267 | dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); |
34dc7c2f BB |
1268 | } |
1269 | ||
34dc7c2f | 1270 | /* |
b128c09f | 1271 | * Increase the file length |
34dc7c2f BB |
1272 | * |
1273 | * IN: zp - znode of file to free data in. | |
b128c09f | 1274 | * end - new end-of-file |
34dc7c2f | 1275 | * |
d3cc8b15 | 1276 | * RETURN: 0 on success, error code on failure |
34dc7c2f | 1277 | */ |
b128c09f BB |
1278 | static int |
1279 | zfs_extend(znode_t *zp, uint64_t end) | |
34dc7c2f | 1280 | { |
3558fd73 | 1281 | zfs_sb_t *zsb = ZTOZSB(zp); |
b128c09f | 1282 | dmu_tx_t *tx; |
34dc7c2f | 1283 | rl_t *rl; |
b128c09f | 1284 | uint64_t newblksz; |
34dc7c2f BB |
1285 | int error; |
1286 | ||
34dc7c2f | 1287 | /* |
b128c09f | 1288 | * We will change zp_size, lock the whole file. |
34dc7c2f | 1289 | */ |
b128c09f | 1290 | rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); |
34dc7c2f BB |
1291 | |
1292 | /* | |
1293 | * Nothing to do if file already at desired length. | |
1294 | */ | |
428870ff | 1295 | if (end <= zp->z_size) { |
34dc7c2f BB |
1296 | zfs_range_unlock(rl); |
1297 | return (0); | |
1298 | } | |
3558fd73 | 1299 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
1300 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1301 | zfs_sa_upgrade_txholds(tx, zp); | |
b128c09f | 1302 | if (end > zp->z_blksz && |
3558fd73 | 1303 | (!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) { |
34dc7c2f BB |
1304 | /* |
1305 | * We are growing the file past the current block size. | |
1306 | */ | |
3558fd73 | 1307 | if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) { |
f1512ee6 MA |
1308 | /* |
1309 | * File's blocksize is already larger than the | |
1310 | * "recordsize" property. Only let it grow to | |
1311 | * the next power of 2. | |
1312 | */ | |
34dc7c2f | 1313 | ASSERT(!ISP2(zp->z_blksz)); |
f1512ee6 | 1314 | newblksz = MIN(end, 1 << highbit64(zp->z_blksz)); |
34dc7c2f | 1315 | } else { |
3558fd73 | 1316 | newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz); |
34dc7c2f | 1317 | } |
b128c09f BB |
1318 | dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); |
1319 | } else { | |
1320 | newblksz = 0; | |
34dc7c2f BB |
1321 | } |
1322 | ||
384f8a09 | 1323 | error = dmu_tx_assign(tx, TXG_WAIT); |
34dc7c2f | 1324 | if (error) { |
34dc7c2f BB |
1325 | dmu_tx_abort(tx); |
1326 | zfs_range_unlock(rl); | |
1327 | return (error); | |
1328 | } | |
1329 | ||
b128c09f BB |
1330 | if (newblksz) |
1331 | zfs_grow_blocksize(zp, newblksz, tx); | |
34dc7c2f | 1332 | |
428870ff BB |
1333 | zp->z_size = end; |
1334 | ||
3558fd73 | 1335 | VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)), |
428870ff | 1336 | &zp->z_size, sizeof (zp->z_size), tx)); |
34dc7c2f | 1337 | |
b128c09f | 1338 | zfs_range_unlock(rl); |
34dc7c2f | 1339 | |
b128c09f | 1340 | dmu_tx_commit(tx); |
34dc7c2f | 1341 | |
b128c09f BB |
1342 | return (0); |
1343 | } | |
1344 | ||
223df016 TC |
1345 | /* |
1346 | * zfs_zero_partial_page - Modeled after update_pages() but | |
1347 | * with different arguments and semantics for use by zfs_freesp(). | |
1348 | * | |
1349 | * Zeroes a piece of a single page cache entry for zp at offset | |
1350 | * start and length len. | |
1351 | * | |
1352 | * Caller must acquire a range lock on the file for the region | |
1353 | * being zeroed in order that the ARC and page cache stay in sync. | |
1354 | */ | |
1355 | static void | |
1356 | zfs_zero_partial_page(znode_t *zp, uint64_t start, uint64_t len) | |
1357 | { | |
1358 | struct address_space *mp = ZTOI(zp)->i_mapping; | |
1359 | struct page *pp; | |
1360 | int64_t off; | |
1361 | void *pb; | |
1362 | ||
1363 | ASSERT((start & PAGE_CACHE_MASK) == | |
1364 | ((start + len - 1) & PAGE_CACHE_MASK)); | |
1365 | ||
1366 | off = start & (PAGE_CACHE_SIZE - 1); | |
1367 | start &= PAGE_CACHE_MASK; | |
1368 | ||
1369 | pp = find_lock_page(mp, start >> PAGE_CACHE_SHIFT); | |
1370 | if (pp) { | |
1371 | if (mapping_writably_mapped(mp)) | |
1372 | flush_dcache_page(pp); | |
1373 | ||
1374 | pb = kmap(pp); | |
1375 | bzero(pb + off, len); | |
1376 | kunmap(pp); | |
1377 | ||
1378 | if (mapping_writably_mapped(mp)) | |
1379 | flush_dcache_page(pp); | |
1380 | ||
1381 | mark_page_accessed(pp); | |
1382 | SetPageUptodate(pp); | |
1383 | ClearPageError(pp); | |
1384 | unlock_page(pp); | |
1385 | page_cache_release(pp); | |
1386 | } | |
1387 | } | |
1388 | ||
b128c09f BB |
1389 | /* |
1390 | * Free space in a file. | |
1391 | * | |
1392 | * IN: zp - znode of file to free data in. | |
1393 | * off - start of section to free. | |
1394 | * len - length of section to free. | |
1395 | * | |
d3cc8b15 | 1396 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1397 | */ |
1398 | static int | |
1399 | zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) | |
1400 | { | |
3558fd73 | 1401 | zfs_sb_t *zsb = ZTOZSB(zp); |
b128c09f BB |
1402 | rl_t *rl; |
1403 | int error; | |
1404 | ||
1405 | /* | |
1406 | * Lock the range being freed. | |
1407 | */ | |
1408 | rl = zfs_range_lock(zp, off, len, RL_WRITER); | |
1409 | ||
1410 | /* | |
1411 | * Nothing to do if file already at desired length. | |
1412 | */ | |
428870ff | 1413 | if (off >= zp->z_size) { |
b128c09f BB |
1414 | zfs_range_unlock(rl); |
1415 | return (0); | |
34dc7c2f BB |
1416 | } |
1417 | ||
428870ff BB |
1418 | if (off + len > zp->z_size) |
1419 | len = zp->z_size - off; | |
b128c09f | 1420 | |
3558fd73 | 1421 | error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len); |
b128c09f | 1422 | |
223df016 TC |
1423 | /* |
1424 | * Zero partial page cache entries. This must be done under a | |
1425 | * range lock in order to keep the ARC and page cache in sync. | |
1426 | */ | |
1427 | if (zp->z_is_mapped) { | |
1428 | loff_t first_page, last_page, page_len; | |
1429 | loff_t first_page_offset, last_page_offset; | |
1430 | ||
1431 | /* first possible full page in hole */ | |
1432 | first_page = (off + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; | |
1433 | /* last page of hole */ | |
1434 | last_page = (off + len) >> PAGE_CACHE_SHIFT; | |
1435 | ||
1436 | /* offset of first_page */ | |
1437 | first_page_offset = first_page << PAGE_CACHE_SHIFT; | |
1438 | /* offset of last_page */ | |
1439 | last_page_offset = last_page << PAGE_CACHE_SHIFT; | |
1440 | ||
cb08f063 TC |
1441 | /* truncate whole pages */ |
1442 | if (last_page_offset > first_page_offset) { | |
1443 | truncate_inode_pages_range(ZTOI(zp)->i_mapping, | |
1444 | first_page_offset, last_page_offset - 1); | |
1445 | } | |
1446 | ||
1447 | /* truncate sub-page ranges */ | |
223df016 TC |
1448 | if (first_page > last_page) { |
1449 | /* entire punched area within a single page */ | |
1450 | zfs_zero_partial_page(zp, off, len); | |
1451 | } else { | |
1452 | /* beginning of punched area at the end of a page */ | |
1453 | page_len = first_page_offset - off; | |
1454 | if (page_len > 0) | |
1455 | zfs_zero_partial_page(zp, off, page_len); | |
1456 | ||
1457 | /* end of punched area at the beginning of a page */ | |
1458 | page_len = off + len - last_page_offset; | |
1459 | if (page_len > 0) | |
1460 | zfs_zero_partial_page(zp, last_page_offset, | |
1461 | page_len); | |
1462 | } | |
1463 | } | |
34dc7c2f BB |
1464 | zfs_range_unlock(rl); |
1465 | ||
b128c09f BB |
1466 | return (error); |
1467 | } | |
1468 | ||
1469 | /* | |
1470 | * Truncate a file | |
1471 | * | |
1472 | * IN: zp - znode of file to free data in. | |
1473 | * end - new end-of-file. | |
1474 | * | |
d3cc8b15 | 1475 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1476 | */ |
1477 | static int | |
1478 | zfs_trunc(znode_t *zp, uint64_t end) | |
1479 | { | |
3558fd73 | 1480 | zfs_sb_t *zsb = ZTOZSB(zp); |
b128c09f BB |
1481 | dmu_tx_t *tx; |
1482 | rl_t *rl; | |
1483 | int error; | |
572e2857 BB |
1484 | sa_bulk_attr_t bulk[2]; |
1485 | int count = 0; | |
b128c09f BB |
1486 | |
1487 | /* | |
1488 | * We will change zp_size, lock the whole file. | |
1489 | */ | |
1490 | rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); | |
1491 | ||
1492 | /* | |
1493 | * Nothing to do if file already at desired length. | |
1494 | */ | |
428870ff | 1495 | if (end >= zp->z_size) { |
b128c09f BB |
1496 | zfs_range_unlock(rl); |
1497 | return (0); | |
1498 | } | |
1499 | ||
3558fd73 | 1500 | error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1); |
b128c09f BB |
1501 | if (error) { |
1502 | zfs_range_unlock(rl); | |
1503 | return (error); | |
1504 | } | |
3558fd73 | 1505 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
1506 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1507 | zfs_sa_upgrade_txholds(tx, zp); | |
7a8f0e80 | 1508 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1509 | if (error) { |
b128c09f BB |
1510 | dmu_tx_abort(tx); |
1511 | zfs_range_unlock(rl); | |
1512 | return (error); | |
1513 | } | |
b128c09f | 1514 | |
428870ff | 1515 | zp->z_size = end; |
3558fd73 | 1516 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb), |
572e2857 | 1517 | NULL, &zp->z_size, sizeof (zp->z_size)); |
428870ff | 1518 | |
572e2857 BB |
1519 | if (end == 0) { |
1520 | zp->z_pflags &= ~ZFS_SPARSE; | |
3558fd73 | 1521 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), |
572e2857 BB |
1522 | NULL, &zp->z_pflags, 8); |
1523 | } | |
1524 | VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); | |
b128c09f | 1525 | |
34dc7c2f BB |
1526 | dmu_tx_commit(tx); |
1527 | ||
d164b209 | 1528 | zfs_range_unlock(rl); |
34dc7c2f BB |
1529 | |
1530 | return (0); | |
1531 | } | |
1532 | ||
b128c09f BB |
1533 | /* |
1534 | * Free space in a file | |
1535 | * | |
1536 | * IN: zp - znode of file to free data in. | |
1537 | * off - start of range | |
1538 | * len - end of range (0 => EOF) | |
1539 | * flag - current file open mode flags. | |
1540 | * log - TRUE if this action should be logged | |
1541 | * | |
d3cc8b15 | 1542 | * RETURN: 0 on success, error code on failure |
b128c09f BB |
1543 | */ |
1544 | int | |
1545 | zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) | |
1546 | { | |
b128c09f | 1547 | dmu_tx_t *tx; |
3558fd73 BB |
1548 | zfs_sb_t *zsb = ZTOZSB(zp); |
1549 | zilog_t *zilog = zsb->z_log; | |
428870ff BB |
1550 | uint64_t mode; |
1551 | uint64_t mtime[2], ctime[2]; | |
1552 | sa_bulk_attr_t bulk[3]; | |
1553 | int count = 0; | |
b128c09f BB |
1554 | int error; |
1555 | ||
3558fd73 | 1556 | if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zsb), &mode, |
428870ff BB |
1557 | sizeof (mode))) != 0) |
1558 | return (error); | |
1559 | ||
1560 | if (off > zp->z_size) { | |
b128c09f BB |
1561 | error = zfs_extend(zp, off+len); |
1562 | if (error == 0 && log) | |
1563 | goto log; | |
223df016 | 1564 | goto out; |
b128c09f BB |
1565 | } |
1566 | ||
b128c09f BB |
1567 | if (len == 0) { |
1568 | error = zfs_trunc(zp, off); | |
1569 | } else { | |
1570 | if ((error = zfs_free_range(zp, off, len)) == 0 && | |
428870ff | 1571 | off + len > zp->z_size) |
b128c09f BB |
1572 | error = zfs_extend(zp, off+len); |
1573 | } | |
1574 | if (error || !log) | |
223df016 | 1575 | goto out; |
b128c09f | 1576 | log: |
3558fd73 | 1577 | tx = dmu_tx_create(zsb->z_os); |
428870ff BB |
1578 | dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); |
1579 | zfs_sa_upgrade_txholds(tx, zp); | |
384f8a09 | 1580 | error = dmu_tx_assign(tx, TXG_WAIT); |
b128c09f | 1581 | if (error) { |
b128c09f | 1582 | dmu_tx_abort(tx); |
223df016 | 1583 | goto out; |
b128c09f BB |
1584 | } |
1585 | ||
3558fd73 BB |
1586 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zsb), NULL, mtime, 16); |
1587 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zsb), NULL, ctime, 16); | |
1588 | SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb), | |
428870ff BB |
1589 | NULL, &zp->z_pflags, 8); |
1590 | zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); | |
1591 | error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); | |
1592 | ASSERT(error == 0); | |
1593 | ||
b128c09f BB |
1594 | zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); |
1595 | ||
1596 | dmu_tx_commit(tx); | |
223df016 | 1597 | |
960e08fe | 1598 | zfs_inode_update(zp); |
223df016 TC |
1599 | error = 0; |
1600 | ||
1601 | out: | |
1602 | /* | |
1603 | * Truncate the page cache - for file truncate operations, use | |
1604 | * the purpose-built API for truncations. For punching operations, | |
cb08f063 | 1605 | * the truncation is handled under a range lock in zfs_free_range. |
223df016 TC |
1606 | */ |
1607 | if (len == 0) | |
1608 | truncate_setsize(ZTOI(zp), off); | |
223df016 | 1609 | return (error); |
b128c09f BB |
1610 | } |
1611 | ||
34dc7c2f BB |
1612 | void |
1613 | zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) | |
1614 | { | |
22872ff5 BB |
1615 | struct super_block *sb; |
1616 | zfs_sb_t *zsb; | |
428870ff | 1617 | uint64_t moid, obj, sa_obj, version; |
22872ff5 | 1618 | uint64_t sense = ZFS_CASE_SENSITIVE; |
34dc7c2f BB |
1619 | uint64_t norm = 0; |
1620 | nvpair_t *elem; | |
1621 | int error; | |
22872ff5 BB |
1622 | int i; |
1623 | znode_t *rootzp = NULL; | |
1624 | vattr_t vattr; | |
1625 | znode_t *zp; | |
1626 | zfs_acl_ids_t acl_ids; | |
34dc7c2f BB |
1627 | |
1628 | /* | |
1629 | * First attempt to create master node. | |
1630 | */ | |
1631 | /* | |
1632 | * In an empty objset, there are no blocks to read and thus | |
1633 | * there can be no i/o errors (which we assert below). | |
1634 | */ | |
1635 | moid = MASTER_NODE_OBJ; | |
1636 | error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, | |
1637 | DMU_OT_NONE, 0, tx); | |
1638 | ASSERT(error == 0); | |
1639 | ||
1640 | /* | |
1641 | * Set starting attributes. | |
1642 | */ | |
428870ff | 1643 | version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); |
34dc7c2f BB |
1644 | elem = NULL; |
1645 | while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { | |
1646 | /* For the moment we expect all zpl props to be uint64_ts */ | |
1647 | uint64_t val; | |
1648 | char *name; | |
1649 | ||
1650 | ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); | |
1651 | VERIFY(nvpair_value_uint64(elem, &val) == 0); | |
1652 | name = nvpair_name(elem); | |
1653 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { | |
9babb374 BB |
1654 | if (val < version) |
1655 | version = val; | |
34dc7c2f BB |
1656 | } else { |
1657 | error = zap_update(os, moid, name, 8, 1, &val, tx); | |
1658 | } | |
1659 | ASSERT(error == 0); | |
1660 | if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) | |
1661 | norm = val; | |
22872ff5 BB |
1662 | else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) |
1663 | sense = val; | |
34dc7c2f BB |
1664 | } |
1665 | ASSERT(version != 0); | |
9babb374 | 1666 | error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); |
34dc7c2f | 1667 | |
428870ff BB |
1668 | /* |
1669 | * Create zap object used for SA attribute registration | |
1670 | */ | |
1671 | ||
1672 | if (version >= ZPL_VERSION_SA) { | |
1673 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1674 | DMU_OT_NONE, 0, tx); | |
1675 | error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
1676 | ASSERT(error == 0); | |
1677 | } else { | |
1678 | sa_obj = 0; | |
1679 | } | |
34dc7c2f BB |
1680 | /* |
1681 | * Create a delete queue. | |
1682 | */ | |
9babb374 | 1683 | obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); |
34dc7c2f | 1684 | |
9babb374 | 1685 | error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); |
34dc7c2f BB |
1686 | ASSERT(error == 0); |
1687 | ||
9babb374 | 1688 | /* |
22872ff5 BB |
1689 | * Create root znode. Create minimal znode/inode/zsb/sb |
1690 | * to allow zfs_mknode to work. | |
9babb374 | 1691 | */ |
22872ff5 BB |
1692 | vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID; |
1693 | vattr.va_mode = S_IFDIR|0755; | |
1694 | vattr.va_uid = crgetuid(cr); | |
1695 | vattr.va_gid = crgetgid(cr); | |
1696 | ||
79c76d5b | 1697 | rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); |
22872ff5 BB |
1698 | rootzp->z_moved = 0; |
1699 | rootzp->z_unlinked = 0; | |
1700 | rootzp->z_atime_dirty = 0; | |
1701 | rootzp->z_is_sa = USE_SA(version, os); | |
1702 | ||
79c76d5b | 1703 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
22872ff5 BB |
1704 | zsb->z_os = os; |
1705 | zsb->z_parent = zsb; | |
1706 | zsb->z_version = version; | |
1707 | zsb->z_use_fuids = USE_FUIDS(version, os); | |
1708 | zsb->z_use_sa = USE_SA(version, os); | |
1709 | zsb->z_norm = norm; | |
1710 | ||
79c76d5b | 1711 | sb = kmem_zalloc(sizeof (struct super_block), KM_SLEEP); |
22872ff5 BB |
1712 | sb->s_fs_info = zsb; |
1713 | ||
1714 | ZTOI(rootzp)->i_sb = sb; | |
1715 | ||
1716 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, | |
1717 | &zsb->z_attr_table); | |
9babb374 | 1718 | |
22872ff5 | 1719 | ASSERT(error == 0); |
9babb374 | 1720 | |
60101509 | 1721 | /* |
22872ff5 BB |
1722 | * Fold case on file systems that are always or sometimes case |
1723 | * insensitive. | |
60101509 | 1724 | */ |
22872ff5 BB |
1725 | if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) |
1726 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
60101509 | 1727 | |
22872ff5 BB |
1728 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
1729 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
1730 | offsetof(znode_t, z_link_node)); | |
60101509 | 1731 | |
d07b7c7f CD |
1732 | zsb->z_hold_mtx = vmem_zalloc(sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ, |
1733 | KM_SLEEP); | |
22872ff5 BB |
1734 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
1735 | mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); | |
60101509 | 1736 | |
22872ff5 BB |
1737 | VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, |
1738 | cr, NULL, &acl_ids)); | |
1739 | zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); | |
1740 | ASSERT3P(zp, ==, rootzp); | |
1741 | error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); | |
1742 | ASSERT(error == 0); | |
1743 | zfs_acl_ids_free(&acl_ids); | |
60101509 | 1744 | |
22872ff5 BB |
1745 | atomic_set(&ZTOI(rootzp)->i_count, 0); |
1746 | sa_handle_destroy(rootzp->z_sa_hdl); | |
22872ff5 BB |
1747 | kmem_cache_free(znode_cache, rootzp); |
1748 | ||
1749 | /* | |
1750 | * Create shares directory | |
1751 | */ | |
22872ff5 | 1752 | error = zfs_create_share_dir(zsb, tx); |
9babb374 | 1753 | ASSERT(error == 0); |
428870ff | 1754 | |
22872ff5 BB |
1755 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
1756 | mutex_destroy(&zsb->z_hold_mtx[i]); | |
2708f716 | 1757 | |
d07b7c7f | 1758 | vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ); |
2708f716 BB |
1759 | kmem_free(sb, sizeof (struct super_block)); |
1760 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
34dc7c2f | 1761 | } |
34dc7c2f | 1762 | #endif /* _KERNEL */ |
428870ff | 1763 | |
34dc7c2f | 1764 | static int |
572e2857 BB |
1765 | zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) |
1766 | { | |
1767 | uint64_t sa_obj = 0; | |
1768 | int error; | |
1769 | ||
1770 | error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
1771 | if (error != 0 && error != ENOENT) | |
1772 | return (error); | |
1773 | ||
1774 | error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); | |
1775 | return (error); | |
1776 | } | |
1777 | ||
1778 | static int | |
1779 | zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, | |
7b8518cb | 1780 | dmu_buf_t **db, void *tag) |
34dc7c2f | 1781 | { |
34dc7c2f | 1782 | dmu_object_info_t doi; |
34dc7c2f | 1783 | int error; |
428870ff | 1784 | |
7b8518cb | 1785 | if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) |
34dc7c2f BB |
1786 | return (error); |
1787 | ||
572e2857 | 1788 | dmu_object_info_from_db(*db, &doi); |
428870ff BB |
1789 | if ((doi.doi_bonus_type != DMU_OT_SA && |
1790 | doi.doi_bonus_type != DMU_OT_ZNODE) || | |
d6320ddb BB |
1791 | (doi.doi_bonus_type == DMU_OT_ZNODE && |
1792 | doi.doi_bonus_size < sizeof (znode_phys_t))) { | |
7b8518cb | 1793 | sa_buf_rele(*db, tag); |
2e528b49 | 1794 | return (SET_ERROR(ENOTSUP)); |
34dc7c2f BB |
1795 | } |
1796 | ||
572e2857 BB |
1797 | error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); |
1798 | if (error != 0) { | |
7b8518cb | 1799 | sa_buf_rele(*db, tag); |
428870ff BB |
1800 | return (error); |
1801 | } | |
1802 | ||
572e2857 BB |
1803 | return (0); |
1804 | } | |
1805 | ||
1806 | void | |
7b8518cb | 1807 | zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) |
572e2857 BB |
1808 | { |
1809 | sa_handle_destroy(hdl); | |
7b8518cb | 1810 | sa_buf_rele(db, tag); |
572e2857 BB |
1811 | } |
1812 | ||
1813 | /* | |
1814 | * Given an object number, return its parent object number and whether | |
1815 | * or not the object is an extended attribute directory. | |
1816 | */ | |
1817 | static int | |
1818 | zfs_obj_to_pobj(sa_handle_t *hdl, sa_attr_type_t *sa_table, uint64_t *pobjp, | |
1819 | int *is_xattrdir) | |
1820 | { | |
1821 | uint64_t parent; | |
1822 | uint64_t pflags; | |
1823 | uint64_t mode; | |
1824 | sa_bulk_attr_t bulk[3]; | |
1825 | int count = 0; | |
1826 | int error; | |
1827 | ||
1828 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, | |
1829 | &parent, sizeof (parent)); | |
428870ff | 1830 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, |
572e2857 | 1831 | &pflags, sizeof (pflags)); |
428870ff | 1832 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, |
572e2857 | 1833 | &mode, sizeof (mode)); |
428870ff | 1834 | |
572e2857 | 1835 | if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) |
428870ff | 1836 | return (error); |
572e2857 | 1837 | |
428870ff BB |
1838 | *pobjp = parent; |
1839 | *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); | |
34dc7c2f BB |
1840 | |
1841 | return (0); | |
1842 | } | |
1843 | ||
572e2857 BB |
1844 | /* |
1845 | * Given an object number, return some zpl level statistics | |
1846 | */ | |
1847 | static int | |
1848 | zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, | |
1849 | zfs_stat_t *sb) | |
34dc7c2f | 1850 | { |
572e2857 BB |
1851 | sa_bulk_attr_t bulk[4]; |
1852 | int count = 0; | |
1853 | ||
1854 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, | |
1855 | &sb->zs_mode, sizeof (sb->zs_mode)); | |
1856 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, | |
1857 | &sb->zs_gen, sizeof (sb->zs_gen)); | |
1858 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, | |
1859 | &sb->zs_links, sizeof (sb->zs_links)); | |
1860 | SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, | |
1861 | &sb->zs_ctime, sizeof (sb->zs_ctime)); | |
1862 | ||
1863 | return (sa_bulk_lookup(hdl, bulk, count)); | |
1864 | } | |
1865 | ||
1866 | static int | |
1867 | zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, | |
1868 | sa_attr_type_t *sa_table, char *buf, int len) | |
1869 | { | |
1870 | sa_handle_t *sa_hdl; | |
1871 | sa_handle_t *prevhdl = NULL; | |
1872 | dmu_buf_t *prevdb = NULL; | |
1873 | dmu_buf_t *sa_db = NULL; | |
34dc7c2f BB |
1874 | char *path = buf + len - 1; |
1875 | int error; | |
1876 | ||
1877 | *path = '\0'; | |
572e2857 | 1878 | sa_hdl = hdl; |
428870ff | 1879 | |
34dc7c2f | 1880 | for (;;) { |
17897ce2 | 1881 | uint64_t pobj = 0; |
34dc7c2f BB |
1882 | char component[MAXNAMELEN + 2]; |
1883 | size_t complen; | |
17897ce2 | 1884 | int is_xattrdir = 0; |
34dc7c2f | 1885 | |
572e2857 | 1886 | if (prevdb) |
7b8518cb | 1887 | zfs_release_sa_handle(prevhdl, prevdb, FTAG); |
572e2857 BB |
1888 | |
1889 | if ((error = zfs_obj_to_pobj(sa_hdl, sa_table, &pobj, | |
1890 | &is_xattrdir)) != 0) | |
34dc7c2f BB |
1891 | break; |
1892 | ||
1893 | if (pobj == obj) { | |
1894 | if (path[0] != '/') | |
1895 | *--path = '/'; | |
1896 | break; | |
1897 | } | |
1898 | ||
1899 | component[0] = '/'; | |
1900 | if (is_xattrdir) { | |
1901 | (void) sprintf(component + 1, "<xattrdir>"); | |
1902 | } else { | |
1903 | error = zap_value_search(osp, pobj, obj, | |
1904 | ZFS_DIRENT_OBJ(-1ULL), component + 1); | |
1905 | if (error != 0) | |
1906 | break; | |
1907 | } | |
1908 | ||
1909 | complen = strlen(component); | |
1910 | path -= complen; | |
1911 | ASSERT(path >= buf); | |
1912 | bcopy(component, path, complen); | |
1913 | obj = pobj; | |
572e2857 BB |
1914 | |
1915 | if (sa_hdl != hdl) { | |
1916 | prevhdl = sa_hdl; | |
1917 | prevdb = sa_db; | |
1918 | } | |
7b8518cb | 1919 | error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); |
572e2857 BB |
1920 | if (error != 0) { |
1921 | sa_hdl = prevhdl; | |
1922 | sa_db = prevdb; | |
1923 | break; | |
1924 | } | |
1925 | } | |
1926 | ||
1927 | if (sa_hdl != NULL && sa_hdl != hdl) { | |
1928 | ASSERT(sa_db != NULL); | |
7b8518cb | 1929 | zfs_release_sa_handle(sa_hdl, sa_db, FTAG); |
34dc7c2f BB |
1930 | } |
1931 | ||
1932 | if (error == 0) | |
1933 | (void) memmove(buf, path, buf + len - path); | |
428870ff | 1934 | |
34dc7c2f BB |
1935 | return (error); |
1936 | } | |
572e2857 BB |
1937 | |
1938 | int | |
1939 | zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) | |
1940 | { | |
1941 | sa_attr_type_t *sa_table; | |
1942 | sa_handle_t *hdl; | |
1943 | dmu_buf_t *db; | |
1944 | int error; | |
1945 | ||
1946 | error = zfs_sa_setup(osp, &sa_table); | |
1947 | if (error != 0) | |
1948 | return (error); | |
1949 | ||
7b8518cb | 1950 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
1951 | if (error != 0) |
1952 | return (error); | |
1953 | ||
1954 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
1955 | ||
7b8518cb | 1956 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
1957 | return (error); |
1958 | } | |
1959 | ||
1960 | int | |
1961 | zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, | |
1962 | char *buf, int len) | |
1963 | { | |
1964 | char *path = buf + len - 1; | |
1965 | sa_attr_type_t *sa_table; | |
1966 | sa_handle_t *hdl; | |
1967 | dmu_buf_t *db; | |
1968 | int error; | |
1969 | ||
1970 | *path = '\0'; | |
1971 | ||
1972 | error = zfs_sa_setup(osp, &sa_table); | |
1973 | if (error != 0) | |
1974 | return (error); | |
1975 | ||
7b8518cb | 1976 | error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); |
572e2857 BB |
1977 | if (error != 0) |
1978 | return (error); | |
1979 | ||
1980 | error = zfs_obj_to_stats_impl(hdl, sa_table, sb); | |
1981 | if (error != 0) { | |
7b8518cb | 1982 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
1983 | return (error); |
1984 | } | |
1985 | ||
1986 | error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); | |
1987 | ||
7b8518cb | 1988 | zfs_release_sa_handle(hdl, db, FTAG); |
572e2857 BB |
1989 | return (error); |
1990 | } | |
c28b2279 BB |
1991 | |
1992 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
1993 | EXPORT_SYMBOL(zfs_create_fs); | |
1994 | EXPORT_SYMBOL(zfs_obj_to_path); | |
1995 | #endif |