]>
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. |
fcff0f35 | 23 | * Copyright (c) 2012, 2014 by Delphix. All rights reserved. |
34dc7c2f BB |
24 | */ |
25 | ||
428870ff BB |
26 | /* Portions Copyright 2010 Robert Milkowski */ |
27 | ||
34dc7c2f BB |
28 | #include <sys/types.h> |
29 | #include <sys/param.h> | |
30 | #include <sys/systm.h> | |
31 | #include <sys/sysmacros.h> | |
32 | #include <sys/kmem.h> | |
33 | #include <sys/pathname.h> | |
34 | #include <sys/vnode.h> | |
35 | #include <sys/vfs.h> | |
36 | #include <sys/vfs_opreg.h> | |
37 | #include <sys/mntent.h> | |
38 | #include <sys/mount.h> | |
39 | #include <sys/cmn_err.h> | |
40 | #include "fs/fs_subr.h" | |
41 | #include <sys/zfs_znode.h> | |
3558fd73 | 42 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
43 | #include <sys/zfs_dir.h> |
44 | #include <sys/zil.h> | |
45 | #include <sys/fs/zfs.h> | |
46 | #include <sys/dmu.h> | |
47 | #include <sys/dsl_prop.h> | |
48 | #include <sys/dsl_dataset.h> | |
49 | #include <sys/dsl_deleg.h> | |
50 | #include <sys/spa.h> | |
51 | #include <sys/zap.h> | |
428870ff | 52 | #include <sys/sa.h> |
a94addd9 | 53 | #include <sys/sa_impl.h> |
34dc7c2f BB |
54 | #include <sys/varargs.h> |
55 | #include <sys/policy.h> | |
56 | #include <sys/atomic.h> | |
57 | #include <sys/mkdev.h> | |
58 | #include <sys/modctl.h> | |
59 | #include <sys/refstr.h> | |
60 | #include <sys/zfs_ioctl.h> | |
ebe7e575 | 61 | #include <sys/zfs_ctldir.h> |
34dc7c2f BB |
62 | #include <sys/zfs_fuid.h> |
63 | #include <sys/bootconf.h> | |
64 | #include <sys/sunddi.h> | |
65 | #include <sys/dnlc.h> | |
66 | #include <sys/dmu_objset.h> | |
67 | #include <sys/spa_boot.h> | |
3558fd73 | 68 | #include <sys/zpl.h> |
428870ff | 69 | #include "zfs_comutil.h" |
34dc7c2f | 70 | |
34dc7c2f BB |
71 | /*ARGSUSED*/ |
72 | int | |
03f9ba9d | 73 | zfs_sync(struct super_block *sb, int wait, cred_t *cr) |
34dc7c2f | 74 | { |
03f9ba9d BB |
75 | zfs_sb_t *zsb = sb->s_fs_info; |
76 | ||
34dc7c2f BB |
77 | /* |
78 | * Data integrity is job one. We don't want a compromised kernel | |
79 | * writing to the storage pool, so we never sync during panic. | |
80 | */ | |
d5e53f9d | 81 | if (unlikely(oops_in_progress)) |
34dc7c2f BB |
82 | return (0); |
83 | ||
03f9ba9d BB |
84 | /* |
85 | * Semantically, the only requirement is that the sync be initiated. | |
86 | * The DMU syncs out txgs frequently, so there's nothing to do. | |
87 | */ | |
88 | if (!wait) | |
89 | return (0); | |
90 | ||
3558fd73 | 91 | if (zsb != NULL) { |
34dc7c2f BB |
92 | /* |
93 | * Sync a specific filesystem. | |
94 | */ | |
9babb374 | 95 | dsl_pool_t *dp; |
34dc7c2f | 96 | |
3558fd73 BB |
97 | ZFS_ENTER(zsb); |
98 | dp = dmu_objset_pool(zsb->z_os); | |
9babb374 BB |
99 | |
100 | /* | |
101 | * If the system is shutting down, then skip any | |
102 | * filesystems which may exist on a suspended pool. | |
103 | */ | |
03f9ba9d | 104 | if (spa_suspended(dp->dp_spa)) { |
3558fd73 | 105 | ZFS_EXIT(zsb); |
9babb374 BB |
106 | return (0); |
107 | } | |
108 | ||
3558fd73 BB |
109 | if (zsb->z_log != NULL) |
110 | zil_commit(zsb->z_log, 0); | |
428870ff | 111 | |
3558fd73 | 112 | ZFS_EXIT(zsb); |
34dc7c2f BB |
113 | } else { |
114 | /* | |
115 | * Sync all ZFS filesystems. This is what happens when you | |
116 | * run sync(1M). Unlike other filesystems, ZFS honors the | |
117 | * request by waiting for all pools to commit all dirty data. | |
118 | */ | |
119 | spa_sync_allpools(); | |
120 | } | |
121 | ||
122 | return (0); | |
123 | } | |
e5c39b95 | 124 | EXPORT_SYMBOL(zfs_sync); |
34dc7c2f | 125 | |
2cf7f52b BB |
126 | boolean_t |
127 | zfs_is_readonly(zfs_sb_t *zsb) | |
128 | { | |
129 | return (!!(zsb->z_sb->s_flags & MS_RDONLY)); | |
130 | } | |
131 | EXPORT_SYMBOL(zfs_is_readonly); | |
132 | ||
34dc7c2f BB |
133 | static void |
134 | atime_changed_cb(void *arg, uint64_t newval) | |
135 | { | |
2cf7f52b | 136 | ((zfs_sb_t *)arg)->z_atime = newval; |
34dc7c2f BB |
137 | } |
138 | ||
6d111134 TC |
139 | static void |
140 | relatime_changed_cb(void *arg, uint64_t newval) | |
141 | { | |
142 | ((zfs_sb_t *)arg)->z_relatime = newval; | |
143 | } | |
144 | ||
34dc7c2f BB |
145 | static void |
146 | xattr_changed_cb(void *arg, uint64_t newval) | |
147 | { | |
3558fd73 | 148 | zfs_sb_t *zsb = arg; |
34dc7c2f | 149 | |
82a37189 | 150 | if (newval == ZFS_XATTR_OFF) { |
2cf7f52b | 151 | zsb->z_flags &= ~ZSB_XATTR; |
82a37189 BB |
152 | } else { |
153 | zsb->z_flags |= ZSB_XATTR; | |
154 | ||
155 | if (newval == ZFS_XATTR_SA) | |
156 | zsb->z_xattr_sa = B_TRUE; | |
157 | else | |
158 | zsb->z_xattr_sa = B_FALSE; | |
159 | } | |
34dc7c2f BB |
160 | } |
161 | ||
023699cd MM |
162 | static void |
163 | acltype_changed_cb(void *arg, uint64_t newval) | |
164 | { | |
165 | zfs_sb_t *zsb = arg; | |
166 | ||
167 | switch (newval) { | |
168 | case ZFS_ACLTYPE_OFF: | |
169 | zsb->z_acl_type = ZFS_ACLTYPE_OFF; | |
170 | zsb->z_sb->s_flags &= ~MS_POSIXACL; | |
171 | break; | |
172 | case ZFS_ACLTYPE_POSIXACL: | |
b695c34e | 173 | #ifdef CONFIG_FS_POSIX_ACL |
023699cd MM |
174 | zsb->z_acl_type = ZFS_ACLTYPE_POSIXACL; |
175 | zsb->z_sb->s_flags |= MS_POSIXACL; | |
b695c34e MM |
176 | #else |
177 | zsb->z_acl_type = ZFS_ACLTYPE_OFF; | |
178 | zsb->z_sb->s_flags &= ~MS_POSIXACL; | |
179 | #endif /* CONFIG_FS_POSIX_ACL */ | |
023699cd MM |
180 | break; |
181 | default: | |
182 | break; | |
183 | } | |
184 | } | |
185 | ||
34dc7c2f BB |
186 | static void |
187 | blksz_changed_cb(void *arg, uint64_t newval) | |
188 | { | |
3558fd73 | 189 | zfs_sb_t *zsb = arg; |
f1512ee6 MA |
190 | ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zsb->z_os))); |
191 | ASSERT3U(newval, >=, SPA_MINBLOCKSIZE); | |
192 | ASSERT(ISP2(newval)); | |
34dc7c2f | 193 | |
3558fd73 | 194 | zsb->z_max_blksz = newval; |
34dc7c2f BB |
195 | } |
196 | ||
197 | static void | |
198 | readonly_changed_cb(void *arg, uint64_t newval) | |
199 | { | |
3558fd73 BB |
200 | zfs_sb_t *zsb = arg; |
201 | struct super_block *sb = zsb->z_sb; | |
34dc7c2f | 202 | |
2cf7f52b BB |
203 | if (sb == NULL) |
204 | return; | |
205 | ||
206 | if (newval) | |
3558fd73 | 207 | sb->s_flags |= MS_RDONLY; |
2cf7f52b | 208 | else |
3558fd73 | 209 | sb->s_flags &= ~MS_RDONLY; |
34dc7c2f BB |
210 | } |
211 | ||
212 | static void | |
213 | devices_changed_cb(void *arg, uint64_t newval) | |
214 | { | |
34dc7c2f BB |
215 | } |
216 | ||
217 | static void | |
218 | setuid_changed_cb(void *arg, uint64_t newval) | |
219 | { | |
34dc7c2f BB |
220 | } |
221 | ||
222 | static void | |
223 | exec_changed_cb(void *arg, uint64_t newval) | |
224 | { | |
34dc7c2f BB |
225 | } |
226 | ||
34dc7c2f BB |
227 | static void |
228 | nbmand_changed_cb(void *arg, uint64_t newval) | |
229 | { | |
3558fd73 BB |
230 | zfs_sb_t *zsb = arg; |
231 | struct super_block *sb = zsb->z_sb; | |
232 | ||
2cf7f52b BB |
233 | if (sb == NULL) |
234 | return; | |
235 | ||
236 | if (newval == TRUE) | |
3558fd73 | 237 | sb->s_flags |= MS_MANDLOCK; |
2cf7f52b | 238 | else |
3558fd73 | 239 | sb->s_flags &= ~MS_MANDLOCK; |
34dc7c2f BB |
240 | } |
241 | ||
242 | static void | |
243 | snapdir_changed_cb(void *arg, uint64_t newval) | |
244 | { | |
3558fd73 | 245 | ((zfs_sb_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
246 | } |
247 | ||
248 | static void | |
249 | vscan_changed_cb(void *arg, uint64_t newval) | |
250 | { | |
3558fd73 | 251 | ((zfs_sb_t *)arg)->z_vscan = newval; |
34dc7c2f BB |
252 | } |
253 | ||
34dc7c2f BB |
254 | static void |
255 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
256 | { | |
3558fd73 | 257 | ((zfs_sb_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
258 | } |
259 | ||
e5c39b95 | 260 | int |
3558fd73 | 261 | zfs_register_callbacks(zfs_sb_t *zsb) |
34dc7c2f BB |
262 | { |
263 | struct dsl_dataset *ds = NULL; | |
3558fd73 | 264 | objset_t *os = zsb->z_os; |
0282c413 | 265 | zfs_mntopts_t *zmo = zsb->z_mntopts; |
34dc7c2f BB |
266 | int error = 0; |
267 | ||
0282c413 BB |
268 | ASSERT(zsb); |
269 | ASSERT(zmo); | |
270 | ||
271 | /* | |
272 | * The act of registering our callbacks will destroy any mount | |
273 | * options we may have. In order to enable temporary overrides | |
274 | * of mount options, we stash away the current values and | |
275 | * restore them after we register the callbacks. | |
276 | */ | |
277 | if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) { | |
278 | zmo->z_do_readonly = B_TRUE; | |
279 | zmo->z_readonly = B_TRUE; | |
280 | } | |
34dc7c2f BB |
281 | |
282 | /* | |
283 | * Register property callbacks. | |
284 | * | |
285 | * It would probably be fine to just check for i/o error from | |
286 | * the first prop_register(), but I guess I like to go | |
287 | * overboard... | |
288 | */ | |
289 | ds = dmu_objset_ds(os); | |
13fe0198 | 290 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
3558fd73 | 291 | error = dsl_prop_register(ds, |
13fe0198 | 292 | zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb); |
98fad862 | 293 | error = error ? error : dsl_prop_register(ds, |
6d111134 | 294 | zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb); |
34dc7c2f | 295 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 296 | zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb); |
34dc7c2f | 297 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 298 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb); |
34dc7c2f | 299 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 300 | zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb); |
34dc7c2f | 301 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 302 | zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb); |
34dc7c2f | 303 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 304 | zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb); |
34dc7c2f | 305 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 306 | zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb); |
34dc7c2f | 307 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 308 | zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb); |
34dc7c2f | 309 | error = error ? error : dsl_prop_register(ds, |
023699cd MM |
310 | zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb); |
311 | error = error ? error : dsl_prop_register(ds, | |
312 | zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb); | |
34dc7c2f | 313 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 314 | zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb); |
2cf7f52b | 315 | error = error ? error : dsl_prop_register(ds, |
13fe0198 MA |
316 | zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb); |
317 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
34dc7c2f BB |
318 | if (error) |
319 | goto unregister; | |
320 | ||
0282c413 BB |
321 | /* |
322 | * Invoke our callbacks to restore temporary mount options. | |
323 | */ | |
324 | if (zmo->z_do_readonly) | |
325 | readonly_changed_cb(zsb, zmo->z_readonly); | |
326 | if (zmo->z_do_setuid) | |
327 | setuid_changed_cb(zsb, zmo->z_setuid); | |
328 | if (zmo->z_do_exec) | |
329 | exec_changed_cb(zsb, zmo->z_exec); | |
330 | if (zmo->z_do_devices) | |
331 | devices_changed_cb(zsb, zmo->z_devices); | |
332 | if (zmo->z_do_xattr) | |
333 | xattr_changed_cb(zsb, zmo->z_xattr); | |
334 | if (zmo->z_do_atime) | |
335 | atime_changed_cb(zsb, zmo->z_atime); | |
336 | if (zmo->z_do_relatime) | |
337 | relatime_changed_cb(zsb, zmo->z_relatime); | |
338 | if (zmo->z_do_nbmand) | |
339 | nbmand_changed_cb(zsb, zmo->z_nbmand); | |
c9ada6d5 | 340 | |
34dc7c2f BB |
341 | return (0); |
342 | ||
343 | unregister: | |
344 | /* | |
345 | * We may attempt to unregister some callbacks that are not | |
346 | * registered, but this is OK; it will simply return ENOMSG, | |
347 | * which we will ignore. | |
348 | */ | |
13fe0198 MA |
349 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME), |
350 | atime_changed_cb, zsb); | |
6d111134 TC |
351 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RELATIME), |
352 | relatime_changed_cb, zsb); | |
13fe0198 MA |
353 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR), |
354 | xattr_changed_cb, zsb); | |
355 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
356 | blksz_changed_cb, zsb); | |
357 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY), | |
358 | readonly_changed_cb, zsb); | |
359 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES), | |
360 | devices_changed_cb, zsb); | |
361 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID), | |
362 | setuid_changed_cb, zsb); | |
363 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC), | |
364 | exec_changed_cb, zsb); | |
365 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR), | |
366 | snapdir_changed_cb, zsb); | |
023699cd MM |
367 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE), |
368 | acltype_changed_cb, zsb); | |
13fe0198 MA |
369 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT), |
370 | acl_inherit_changed_cb, zsb); | |
371 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN), | |
372 | vscan_changed_cb, zsb); | |
373 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND), | |
374 | nbmand_changed_cb, zsb); | |
34dc7c2f | 375 | |
3558fd73 | 376 | return (error); |
34dc7c2f | 377 | } |
e5c39b95 | 378 | EXPORT_SYMBOL(zfs_register_callbacks); |
34dc7c2f | 379 | |
428870ff BB |
380 | static int |
381 | zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, | |
382 | uint64_t *userp, uint64_t *groupp) | |
9babb374 | 383 | { |
428870ff BB |
384 | /* |
385 | * Is it a valid type of object to track? | |
386 | */ | |
387 | if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) | |
2e528b49 | 388 | return (SET_ERROR(ENOENT)); |
9babb374 | 389 | |
428870ff BB |
390 | /* |
391 | * If we have a NULL data pointer | |
392 | * then assume the id's aren't changing and | |
393 | * return EEXIST to the dmu to let it know to | |
394 | * use the same ids | |
395 | */ | |
396 | if (data == NULL) | |
2e528b49 | 397 | return (SET_ERROR(EEXIST)); |
9babb374 | 398 | |
428870ff | 399 | if (bonustype == DMU_OT_ZNODE) { |
a94addd9 | 400 | znode_phys_t *znp = data; |
428870ff BB |
401 | *userp = znp->zp_uid; |
402 | *groupp = znp->zp_gid; | |
9babb374 | 403 | } else { |
428870ff | 404 | int hdrsize; |
a94addd9 MA |
405 | sa_hdr_phys_t *sap = data; |
406 | sa_hdr_phys_t sa = *sap; | |
407 | boolean_t swap = B_FALSE; | |
9babb374 | 408 | |
428870ff | 409 | ASSERT(bonustype == DMU_OT_SA); |
428870ff | 410 | |
a94addd9 | 411 | if (sa.sa_magic == 0) { |
428870ff BB |
412 | /* |
413 | * This should only happen for newly created | |
414 | * files that haven't had the znode data filled | |
415 | * in yet. | |
416 | */ | |
417 | *userp = 0; | |
418 | *groupp = 0; | |
a94addd9 MA |
419 | return (0); |
420 | } | |
421 | if (sa.sa_magic == BSWAP_32(SA_MAGIC)) { | |
422 | sa.sa_magic = SA_MAGIC; | |
423 | sa.sa_layout_info = BSWAP_16(sa.sa_layout_info); | |
424 | swap = B_TRUE; | |
425 | } else { | |
426 | VERIFY3U(sa.sa_magic, ==, SA_MAGIC); | |
427 | } | |
428 | ||
429 | hdrsize = sa_hdrsize(&sa); | |
430 | VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t)); | |
431 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
432 | SA_UID_OFFSET)); | |
433 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
434 | SA_GID_OFFSET)); | |
435 | if (swap) { | |
436 | *userp = BSWAP_64(*userp); | |
437 | *groupp = BSWAP_64(*groupp); | |
428870ff | 438 | } |
9babb374 | 439 | } |
13fe0198 | 440 | return (0); |
9babb374 BB |
441 | } |
442 | ||
443 | static void | |
3558fd73 | 444 | fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr, |
9babb374 BB |
445 | char *domainbuf, int buflen, uid_t *ridp) |
446 | { | |
9babb374 BB |
447 | uint64_t fuid; |
448 | const char *domain; | |
449 | ||
450 | fuid = strtonum(fuidstr, NULL); | |
451 | ||
3558fd73 | 452 | domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid)); |
9babb374 BB |
453 | if (domain) |
454 | (void) strlcpy(domainbuf, domain, buflen); | |
455 | else | |
456 | domainbuf[0] = '\0'; | |
457 | *ridp = FUID_RID(fuid); | |
458 | } | |
459 | ||
460 | static uint64_t | |
3558fd73 | 461 | zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type) |
9babb374 BB |
462 | { |
463 | switch (type) { | |
464 | case ZFS_PROP_USERUSED: | |
465 | return (DMU_USERUSED_OBJECT); | |
466 | case ZFS_PROP_GROUPUSED: | |
467 | return (DMU_GROUPUSED_OBJECT); | |
468 | case ZFS_PROP_USERQUOTA: | |
3558fd73 | 469 | return (zsb->z_userquota_obj); |
9babb374 | 470 | case ZFS_PROP_GROUPQUOTA: |
3558fd73 | 471 | return (zsb->z_groupquota_obj); |
149e873a | 472 | default: |
2e528b49 | 473 | return (SET_ERROR(ENOTSUP)); |
9babb374 BB |
474 | } |
475 | return (0); | |
476 | } | |
477 | ||
478 | int | |
3558fd73 | 479 | zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
480 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
481 | { | |
482 | int error; | |
483 | zap_cursor_t zc; | |
484 | zap_attribute_t za; | |
485 | zfs_useracct_t *buf = vbuf; | |
486 | uint64_t obj; | |
487 | ||
3558fd73 | 488 | if (!dmu_objset_userspace_present(zsb->z_os)) |
2e528b49 | 489 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 490 | |
3558fd73 | 491 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
492 | if (obj == 0) { |
493 | *bufsizep = 0; | |
494 | return (0); | |
495 | } | |
496 | ||
3558fd73 | 497 | for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep); |
9babb374 BB |
498 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
499 | zap_cursor_advance(&zc)) { | |
500 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
501 | *bufsizep) | |
502 | break; | |
503 | ||
3558fd73 | 504 | fuidstr_to_sid(zsb, za.za_name, |
9babb374 BB |
505 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
506 | ||
507 | buf->zu_space = za.za_first_integer; | |
508 | buf++; | |
509 | } | |
510 | if (error == ENOENT) | |
511 | error = 0; | |
512 | ||
513 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
514 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
515 | *cookiep = zap_cursor_serialize(&zc); | |
516 | zap_cursor_fini(&zc); | |
517 | return (error); | |
518 | } | |
e5c39b95 | 519 | EXPORT_SYMBOL(zfs_userspace_many); |
9babb374 BB |
520 | |
521 | /* | |
522 | * buf must be big enough (eg, 32 bytes) | |
523 | */ | |
524 | static int | |
3558fd73 | 525 | id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid, |
9babb374 BB |
526 | char *buf, boolean_t addok) |
527 | { | |
528 | uint64_t fuid; | |
529 | int domainid = 0; | |
530 | ||
531 | if (domain && domain[0]) { | |
3558fd73 | 532 | domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok); |
9babb374 | 533 | if (domainid == -1) |
2e528b49 | 534 | return (SET_ERROR(ENOENT)); |
9babb374 BB |
535 | } |
536 | fuid = FUID_ENCODE(domainid, rid); | |
537 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
538 | return (0); | |
539 | } | |
540 | ||
541 | int | |
3558fd73 | 542 | zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
543 | const char *domain, uint64_t rid, uint64_t *valp) |
544 | { | |
545 | char buf[32]; | |
546 | int err; | |
547 | uint64_t obj; | |
548 | ||
549 | *valp = 0; | |
550 | ||
3558fd73 | 551 | if (!dmu_objset_userspace_present(zsb->z_os)) |
2e528b49 | 552 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 553 | |
3558fd73 | 554 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
555 | if (obj == 0) |
556 | return (0); | |
557 | ||
3558fd73 | 558 | err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE); |
9babb374 BB |
559 | if (err) |
560 | return (err); | |
561 | ||
3558fd73 | 562 | err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
563 | if (err == ENOENT) |
564 | err = 0; | |
565 | return (err); | |
566 | } | |
e5c39b95 | 567 | EXPORT_SYMBOL(zfs_userspace_one); |
9babb374 BB |
568 | |
569 | int | |
3558fd73 | 570 | zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
571 | const char *domain, uint64_t rid, uint64_t quota) |
572 | { | |
573 | char buf[32]; | |
574 | int err; | |
575 | dmu_tx_t *tx; | |
576 | uint64_t *objp; | |
577 | boolean_t fuid_dirtied; | |
578 | ||
579 | if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) | |
2e528b49 | 580 | return (SET_ERROR(EINVAL)); |
9babb374 | 581 | |
3558fd73 | 582 | if (zsb->z_version < ZPL_VERSION_USERSPACE) |
2e528b49 | 583 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 584 | |
3558fd73 BB |
585 | objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj : |
586 | &zsb->z_groupquota_obj; | |
9babb374 | 587 | |
3558fd73 | 588 | err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE); |
9babb374 BB |
589 | if (err) |
590 | return (err); | |
3558fd73 | 591 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 592 | |
3558fd73 | 593 | tx = dmu_tx_create(zsb->z_os); |
9babb374 BB |
594 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
595 | if (*objp == 0) { | |
596 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
597 | zfs_userquota_prop_prefixes[type]); | |
598 | } | |
599 | if (fuid_dirtied) | |
3558fd73 | 600 | zfs_fuid_txhold(zsb, tx); |
9babb374 BB |
601 | err = dmu_tx_assign(tx, TXG_WAIT); |
602 | if (err) { | |
603 | dmu_tx_abort(tx); | |
604 | return (err); | |
605 | } | |
606 | ||
3558fd73 | 607 | mutex_enter(&zsb->z_lock); |
9babb374 | 608 | if (*objp == 0) { |
3558fd73 | 609 | *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 610 | DMU_OT_NONE, 0, tx); |
3558fd73 | 611 | VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
612 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
613 | } | |
3558fd73 | 614 | mutex_exit(&zsb->z_lock); |
9babb374 BB |
615 | |
616 | if (quota == 0) { | |
3558fd73 | 617 | err = zap_remove(zsb->z_os, *objp, buf, tx); |
9babb374 BB |
618 | if (err == ENOENT) |
619 | err = 0; | |
620 | } else { | |
3558fd73 | 621 | err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
622 | } |
623 | ASSERT(err == 0); | |
624 | if (fuid_dirtied) | |
3558fd73 | 625 | zfs_fuid_sync(zsb, tx); |
9babb374 BB |
626 | dmu_tx_commit(tx); |
627 | return (err); | |
628 | } | |
e5c39b95 | 629 | EXPORT_SYMBOL(zfs_set_userquota); |
9babb374 BB |
630 | |
631 | boolean_t | |
3558fd73 | 632 | zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid) |
9babb374 BB |
633 | { |
634 | char buf[32]; | |
635 | uint64_t used, quota, usedobj, quotaobj; | |
636 | int err; | |
637 | ||
638 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
3558fd73 | 639 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
9babb374 | 640 | |
3558fd73 | 641 | if (quotaobj == 0 || zsb->z_replay) |
9babb374 BB |
642 | return (B_FALSE); |
643 | ||
644 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
3558fd73 | 645 | err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
646 | if (err != 0) |
647 | return (B_FALSE); | |
648 | ||
3558fd73 | 649 | err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
650 | if (err != 0) |
651 | return (B_FALSE); | |
652 | return (used >= quota); | |
653 | } | |
e5c39b95 | 654 | EXPORT_SYMBOL(zfs_fuid_overquota); |
9babb374 | 655 | |
428870ff | 656 | boolean_t |
3558fd73 | 657 | zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup) |
428870ff BB |
658 | { |
659 | uint64_t fuid; | |
660 | uint64_t quotaobj; | |
428870ff | 661 | |
3558fd73 | 662 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
428870ff | 663 | |
572e2857 | 664 | fuid = isgroup ? zp->z_gid : zp->z_uid; |
428870ff | 665 | |
3558fd73 | 666 | if (quotaobj == 0 || zsb->z_replay) |
428870ff BB |
667 | return (B_FALSE); |
668 | ||
3558fd73 | 669 | return (zfs_fuid_overquota(zsb, isgroup, fuid)); |
428870ff | 670 | } |
e5c39b95 | 671 | EXPORT_SYMBOL(zfs_owner_overquota); |
428870ff | 672 | |
0282c413 BB |
673 | zfs_mntopts_t * |
674 | zfs_mntopts_alloc(void) | |
675 | { | |
676 | return (kmem_zalloc(sizeof (zfs_mntopts_t), KM_SLEEP)); | |
677 | } | |
678 | ||
679 | void | |
680 | zfs_mntopts_free(zfs_mntopts_t *zmo) | |
681 | { | |
682 | if (zmo->z_osname) | |
683 | strfree(zmo->z_osname); | |
684 | ||
685 | if (zmo->z_mntpoint) | |
686 | strfree(zmo->z_mntpoint); | |
687 | ||
688 | kmem_free(zmo, sizeof (zfs_mntopts_t)); | |
689 | } | |
690 | ||
9babb374 | 691 | int |
0282c413 | 692 | zfs_sb_create(const char *osname, zfs_mntopts_t *zmo, zfs_sb_t **zsbp) |
9babb374 BB |
693 | { |
694 | objset_t *os; | |
3558fd73 | 695 | zfs_sb_t *zsb; |
9babb374 BB |
696 | uint64_t zval; |
697 | int i, error; | |
428870ff | 698 | uint64_t sa_obj; |
9babb374 | 699 | |
efcd79a8 | 700 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
9babb374 | 701 | |
428870ff BB |
702 | /* |
703 | * We claim to always be readonly so we can open snapshots; | |
704 | * other ZPL code will prevent us from writing to snapshots. | |
705 | */ | |
3558fd73 | 706 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); |
428870ff | 707 | if (error) { |
3558fd73 | 708 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 | 709 | return (error); |
428870ff | 710 | } |
9babb374 | 711 | |
0282c413 BB |
712 | /* |
713 | * Optional temporary mount options, free'd in zfs_sb_free(). | |
714 | */ | |
715 | zsb->z_mntopts = (zmo ? zmo : zfs_mntopts_alloc()); | |
716 | ||
9babb374 BB |
717 | /* |
718 | * Initialize the zfs-specific filesystem structure. | |
719 | * Should probably make this a kmem cache, shuffle fields, | |
720 | * and just bzero up to z_hold_mtx[]. | |
721 | */ | |
2cf7f52b | 722 | zsb->z_sb = NULL; |
3558fd73 | 723 | zsb->z_parent = zsb; |
f1512ee6 | 724 | zsb->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; |
3558fd73 BB |
725 | zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; |
726 | zsb->z_os = os; | |
9babb374 | 727 | |
3558fd73 | 728 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); |
9babb374 BB |
729 | if (error) { |
730 | goto out; | |
287be44f | 731 | } else if (zsb->z_version > ZPL_VERSION) { |
2e528b49 | 732 | error = SET_ERROR(ENOTSUP); |
9babb374 BB |
733 | goto out; |
734 | } | |
9babb374 BB |
735 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) |
736 | goto out; | |
3558fd73 | 737 | zsb->z_norm = (int)zval; |
9babb374 BB |
738 | |
739 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
740 | goto out; | |
3558fd73 | 741 | zsb->z_utf8 = (zval != 0); |
9babb374 BB |
742 | |
743 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
744 | goto out; | |
3558fd73 | 745 | zsb->z_case = (uint_t)zval; |
9babb374 | 746 | |
023699cd MM |
747 | if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0) |
748 | goto out; | |
749 | zsb->z_acl_type = (uint_t)zval; | |
750 | ||
9babb374 BB |
751 | /* |
752 | * Fold case on file systems that are always or sometimes case | |
753 | * insensitive. | |
754 | */ | |
3558fd73 BB |
755 | if (zsb->z_case == ZFS_CASE_INSENSITIVE || |
756 | zsb->z_case == ZFS_CASE_MIXED) | |
757 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 758 | |
3558fd73 BB |
759 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
760 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
428870ff | 761 | |
3558fd73 | 762 | if (zsb->z_use_sa) { |
428870ff BB |
763 | /* should either have both of these objects or none */ |
764 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
765 | &sa_obj); | |
766 | if (error) | |
591fb62f | 767 | goto out; |
82a37189 BB |
768 | |
769 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval); | |
770 | if ((error == 0) && (zval == ZFS_XATTR_SA)) | |
771 | zsb->z_xattr_sa = B_TRUE; | |
428870ff BB |
772 | } else { |
773 | /* | |
774 | * Pre SA versions file systems should never touch | |
775 | * either the attribute registration or layout objects. | |
776 | */ | |
777 | sa_obj = 0; | |
778 | } | |
779 | ||
572e2857 | 780 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
3558fd73 | 781 | &zsb->z_attr_table); |
572e2857 BB |
782 | if (error) |
783 | goto out; | |
428870ff | 784 | |
3558fd73 | 785 | if (zsb->z_version >= ZPL_VERSION_SA) |
428870ff | 786 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
787 | |
788 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
3558fd73 | 789 | &zsb->z_root); |
9babb374 BB |
790 | if (error) |
791 | goto out; | |
3558fd73 | 792 | ASSERT(zsb->z_root != 0); |
9babb374 BB |
793 | |
794 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
3558fd73 | 795 | &zsb->z_unlinkedobj); |
9babb374 BB |
796 | if (error) |
797 | goto out; | |
798 | ||
799 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
800 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
3558fd73 | 801 | 8, 1, &zsb->z_userquota_obj); |
9babb374 BB |
802 | if (error && error != ENOENT) |
803 | goto out; | |
804 | ||
805 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
806 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
3558fd73 | 807 | 8, 1, &zsb->z_groupquota_obj); |
9babb374 BB |
808 | if (error && error != ENOENT) |
809 | goto out; | |
810 | ||
811 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
3558fd73 | 812 | &zsb->z_fuid_obj); |
9babb374 BB |
813 | if (error && error != ENOENT) |
814 | goto out; | |
815 | ||
816 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
3558fd73 | 817 | &zsb->z_shares_dir); |
9babb374 BB |
818 | if (error && error != ENOENT) |
819 | goto out; | |
820 | ||
3558fd73 BB |
821 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
822 | mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
823 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
9babb374 | 824 | offsetof(znode_t, z_link_node)); |
e16b3fcc | 825 | rrm_init(&zsb->z_teardown_lock, B_FALSE); |
3558fd73 BB |
826 | rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); |
827 | rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
d07b7c7f CD |
828 | |
829 | zsb->z_hold_mtx = vmem_zalloc(sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ, | |
830 | KM_SLEEP); | |
9babb374 | 831 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 832 | mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); |
9babb374 | 833 | |
3558fd73 | 834 | *zsbp = zsb; |
9babb374 BB |
835 | return (0); |
836 | ||
837 | out: | |
3558fd73 BB |
838 | dmu_objset_disown(os, zsb); |
839 | *zsbp = NULL; | |
d07b7c7f CD |
840 | |
841 | vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ); | |
3558fd73 | 842 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 BB |
843 | return (error); |
844 | } | |
86f35f34 | 845 | EXPORT_SYMBOL(zfs_sb_create); |
9babb374 | 846 | |
86f35f34 | 847 | int |
3558fd73 | 848 | zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) |
34dc7c2f | 849 | { |
34dc7c2f BB |
850 | int error; |
851 | ||
3558fd73 | 852 | error = zfs_register_callbacks(zsb); |
34dc7c2f BB |
853 | if (error) |
854 | return (error); | |
855 | ||
856 | /* | |
3558fd73 | 857 | * Set the objset user_ptr to track its zsb. |
34dc7c2f | 858 | */ |
3558fd73 BB |
859 | mutex_enter(&zsb->z_os->os_user_ptr_lock); |
860 | dmu_objset_set_user(zsb->z_os, zsb); | |
861 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
34dc7c2f | 862 | |
3558fd73 | 863 | zsb->z_log = zil_open(zsb->z_os, zfs_get_data); |
9babb374 | 864 | |
34dc7c2f BB |
865 | /* |
866 | * If we are not mounting (ie: online recv), then we don't | |
867 | * have to worry about replaying the log as we blocked all | |
868 | * operations out since we closed the ZIL. | |
869 | */ | |
870 | if (mounting) { | |
b128c09f BB |
871 | boolean_t readonly; |
872 | ||
34dc7c2f BB |
873 | /* |
874 | * During replay we remove the read only flag to | |
875 | * allow replays to succeed. | |
876 | */ | |
2cf7f52b | 877 | readonly = zfs_is_readonly(zsb); |
fb5f0bc8 | 878 | if (readonly != 0) |
2cf7f52b | 879 | readonly_changed_cb(zsb, B_FALSE); |
fb5f0bc8 | 880 | else |
3558fd73 | 881 | zfs_unlinked_drain(zsb); |
34dc7c2f | 882 | |
428870ff BB |
883 | /* |
884 | * Parse and replay the intent log. | |
885 | * | |
886 | * Because of ziltest, this must be done after | |
887 | * zfs_unlinked_drain(). (Further note: ziltest | |
888 | * doesn't use readonly mounts, where | |
889 | * zfs_unlinked_drain() isn't called.) This is because | |
890 | * ziltest causes spa_sync() to think it's committed, | |
891 | * but actually it is not, so the intent log contains | |
892 | * many txg's worth of changes. | |
893 | * | |
894 | * In particular, if object N is in the unlinked set in | |
895 | * the last txg to actually sync, then it could be | |
896 | * actually freed in a later txg and then reallocated | |
897 | * in a yet later txg. This would write a "create | |
898 | * object N" record to the intent log. Normally, this | |
899 | * would be fine because the spa_sync() would have | |
900 | * written out the fact that object N is free, before | |
901 | * we could write the "create object N" intent log | |
902 | * record. | |
903 | * | |
904 | * But when we are in ziltest mode, we advance the "open | |
905 | * txg" without actually spa_sync()-ing the changes to | |
906 | * disk. So we would see that object N is still | |
907 | * allocated and in the unlinked set, and there is an | |
908 | * intent log record saying to allocate it. | |
909 | */ | |
3558fd73 | 910 | if (spa_writeable(dmu_objset_spa(zsb->z_os))) { |
572e2857 | 911 | if (zil_replay_disable) { |
3558fd73 | 912 | zil_destroy(zsb->z_log, B_FALSE); |
572e2857 | 913 | } else { |
3558fd73 BB |
914 | zsb->z_replay = B_TRUE; |
915 | zil_replay(zsb->z_os, zsb, | |
572e2857 | 916 | zfs_replay_vector); |
3558fd73 | 917 | zsb->z_replay = B_FALSE; |
572e2857 | 918 | } |
fb5f0bc8 | 919 | } |
2cf7f52b BB |
920 | |
921 | /* restore readonly bit */ | |
922 | if (readonly != 0) | |
923 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
924 | } |
925 | ||
34dc7c2f BB |
926 | return (0); |
927 | } | |
86f35f34 | 928 | EXPORT_SYMBOL(zfs_sb_setup); |
34dc7c2f | 929 | |
9babb374 | 930 | void |
3558fd73 | 931 | zfs_sb_free(zfs_sb_t *zsb) |
34dc7c2f | 932 | { |
9babb374 | 933 | int i; |
9babb374 | 934 | |
3558fd73 | 935 | zfs_fuid_destroy(zsb); |
9babb374 | 936 | |
3558fd73 BB |
937 | mutex_destroy(&zsb->z_znodes_lock); |
938 | mutex_destroy(&zsb->z_lock); | |
939 | list_destroy(&zsb->z_all_znodes); | |
e16b3fcc | 940 | rrm_destroy(&zsb->z_teardown_lock); |
3558fd73 BB |
941 | rw_destroy(&zsb->z_teardown_inactive_lock); |
942 | rw_destroy(&zsb->z_fuid_lock); | |
9babb374 | 943 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 944 | mutex_destroy(&zsb->z_hold_mtx[i]); |
d07b7c7f | 945 | vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ); |
0282c413 | 946 | zfs_mntopts_free(zsb->z_mntopts); |
3558fd73 | 947 | kmem_free(zsb, sizeof (zfs_sb_t)); |
34dc7c2f | 948 | } |
86f35f34 | 949 | EXPORT_SYMBOL(zfs_sb_free); |
34dc7c2f | 950 | |
9babb374 | 951 | static void |
3558fd73 | 952 | zfs_set_fuid_feature(zfs_sb_t *zsb) |
9babb374 | 953 | { |
3558fd73 BB |
954 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
955 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
9babb374 | 956 | } |
34dc7c2f BB |
957 | |
958 | void | |
3558fd73 | 959 | zfs_unregister_callbacks(zfs_sb_t *zsb) |
34dc7c2f | 960 | { |
3558fd73 | 961 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
962 | struct dsl_dataset *ds; |
963 | ||
964 | /* | |
965 | * Unregister properties. | |
966 | */ | |
967 | if (!dmu_objset_is_snapshot(os)) { | |
968 | ds = dmu_objset_ds(os); | |
969 | VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, | |
3558fd73 | 970 | zsb) == 0); |
34dc7c2f | 971 | |
6d111134 TC |
972 | VERIFY(dsl_prop_unregister(ds, "relatime", relatime_changed_cb, |
973 | zsb) == 0); | |
974 | ||
34dc7c2f | 975 | VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, |
3558fd73 | 976 | zsb) == 0); |
34dc7c2f BB |
977 | |
978 | VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, | |
3558fd73 | 979 | zsb) == 0); |
34dc7c2f BB |
980 | |
981 | VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, | |
3558fd73 | 982 | zsb) == 0); |
34dc7c2f BB |
983 | |
984 | VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, | |
3558fd73 | 985 | zsb) == 0); |
34dc7c2f BB |
986 | |
987 | VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, | |
3558fd73 | 988 | zsb) == 0); |
34dc7c2f BB |
989 | |
990 | VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, | |
3558fd73 | 991 | zsb) == 0); |
34dc7c2f BB |
992 | |
993 | VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, | |
3558fd73 | 994 | zsb) == 0); |
34dc7c2f | 995 | |
023699cd MM |
996 | VERIFY(dsl_prop_unregister(ds, "acltype", acltype_changed_cb, |
997 | zsb) == 0); | |
998 | ||
34dc7c2f | 999 | VERIFY(dsl_prop_unregister(ds, "aclinherit", |
3558fd73 | 1000 | acl_inherit_changed_cb, zsb) == 0); |
34dc7c2f BB |
1001 | |
1002 | VERIFY(dsl_prop_unregister(ds, "vscan", | |
3558fd73 | 1003 | vscan_changed_cb, zsb) == 0); |
2cf7f52b BB |
1004 | |
1005 | VERIFY(dsl_prop_unregister(ds, "nbmand", | |
1006 | nbmand_changed_cb, zsb) == 0); | |
34dc7c2f BB |
1007 | } |
1008 | } | |
e5c39b95 | 1009 | EXPORT_SYMBOL(zfs_unregister_callbacks); |
34dc7c2f | 1010 | |
bc3e15e3 | 1011 | #ifdef HAVE_MLSLABEL |
428870ff | 1012 | /* |
d3cc8b15 WA |
1013 | * Check that the hex label string is appropriate for the dataset being |
1014 | * mounted into the global_zone proper. | |
428870ff | 1015 | * |
d3cc8b15 WA |
1016 | * Return an error if the hex label string is not default or |
1017 | * admin_low/admin_high. For admin_low labels, the corresponding | |
1018 | * dataset must be readonly. | |
428870ff BB |
1019 | */ |
1020 | int | |
1021 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
1022 | { | |
1023 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
1024 | return (0); | |
1025 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
1026 | return (0); | |
1027 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
1028 | /* must be readonly */ | |
1029 | uint64_t rdonly; | |
1030 | ||
1031 | if (dsl_prop_get_integer(dsname, | |
1032 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
2e528b49 | 1033 | return (SET_ERROR(EACCES)); |
428870ff BB |
1034 | return (rdonly ? 0 : EACCES); |
1035 | } | |
2e528b49 | 1036 | return (SET_ERROR(EACCES)); |
428870ff | 1037 | } |
86f35f34 | 1038 | EXPORT_SYMBOL(zfs_check_global_label); |
bc3e15e3 | 1039 | #endif /* HAVE_MLSLABEL */ |
428870ff | 1040 | |
e5c39b95 | 1041 | int |
3558fd73 | 1042 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 1043 | { |
3558fd73 | 1044 | zfs_sb_t *zsb = dentry->d_sb->s_fs_info; |
34dc7c2f | 1045 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
04f9432d | 1046 | uint64_t fsid; |
3558fd73 | 1047 | uint32_t bshift; |
34dc7c2f | 1048 | |
3558fd73 | 1049 | ZFS_ENTER(zsb); |
34dc7c2f | 1050 | |
3558fd73 | 1051 | dmu_objset_space(zsb->z_os, |
34dc7c2f BB |
1052 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
1053 | ||
04f9432d | 1054 | fsid = dmu_objset_fsid_guid(zsb->z_os); |
34dc7c2f | 1055 | /* |
05ff35c6 BB |
1056 | * The underlying storage pool actually uses multiple block |
1057 | * size. Under Solaris frsize (fragment size) is reported as | |
1058 | * the smallest block size we support, and bsize (block size) | |
1059 | * as the filesystem's maximum block size. Unfortunately, | |
1060 | * under Linux the fragment size and block size are often used | |
1061 | * interchangeably. Thus we are forced to report both of them | |
1062 | * as the filesystem's maximum block size. | |
34dc7c2f | 1063 | */ |
05ff35c6 | 1064 | statp->f_frsize = zsb->z_max_blksz; |
3558fd73 BB |
1065 | statp->f_bsize = zsb->z_max_blksz; |
1066 | bshift = fls(statp->f_bsize) - 1; | |
34dc7c2f BB |
1067 | |
1068 | /* | |
3558fd73 BB |
1069 | * The following report "total" blocks of various kinds in |
1070 | * the file system, but reported in terms of f_bsize - the | |
1071 | * "preferred" size. | |
34dc7c2f BB |
1072 | */ |
1073 | ||
3558fd73 BB |
1074 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
1075 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
1076 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
1077 | ||
1078 | /* | |
1079 | * statvfs() should really be called statufs(), because it assumes | |
1080 | * static metadata. ZFS doesn't preallocate files, so the best | |
1081 | * we can do is report the max that could possibly fit in f_files, | |
1082 | * and that minus the number actually used in f_ffree. | |
1083 | * For f_ffree, report the smaller of the number of object available | |
1084 | * and the number of blocks (each object will take at least a block). | |
1085 | */ | |
baab0630 | 1086 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 1087 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
1088 | statp->f_fsid.val[0] = (uint32_t)fsid; |
1089 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 BB |
1090 | statp->f_type = ZFS_SUPER_MAGIC; |
1091 | statp->f_namelen = ZFS_MAXNAMELEN; | |
34dc7c2f BB |
1092 | |
1093 | /* | |
3558fd73 | 1094 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
1095 | * Is there anything useful we could/should provide? |
1096 | */ | |
3558fd73 | 1097 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 1098 | |
3558fd73 | 1099 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1100 | return (0); |
1101 | } | |
e5c39b95 | 1102 | EXPORT_SYMBOL(zfs_statvfs); |
34dc7c2f | 1103 | |
e5c39b95 | 1104 | int |
3558fd73 | 1105 | zfs_root(zfs_sb_t *zsb, struct inode **ipp) |
34dc7c2f | 1106 | { |
34dc7c2f BB |
1107 | znode_t *rootzp; |
1108 | int error; | |
1109 | ||
3558fd73 | 1110 | ZFS_ENTER(zsb); |
34dc7c2f | 1111 | |
3558fd73 | 1112 | error = zfs_zget(zsb, zsb->z_root, &rootzp); |
34dc7c2f | 1113 | if (error == 0) |
3558fd73 | 1114 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1115 | |
3558fd73 | 1116 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1117 | return (error); |
1118 | } | |
e5c39b95 | 1119 | EXPORT_SYMBOL(zfs_root); |
34dc7c2f | 1120 | |
218b4e0a BB |
1121 | #if !defined(HAVE_SPLIT_SHRINKER_CALLBACK) && !defined(HAVE_SHRINK) && \ |
1122 | defined(HAVE_D_PRUNE_ALIASES) | |
1123 | /* | |
1124 | * Linux kernels older than 3.1 do not support a per-filesystem shrinker. | |
1125 | * To accommodate this we must improvise and manually walk the list of znodes | |
1126 | * attempting to prune dentries in order to be able to drop the inodes. | |
1127 | * | |
1128 | * To avoid scanning the same znodes multiple times they are always rotated | |
1129 | * to the end of the z_all_znodes list. New znodes are inserted at the | |
1130 | * end of the list so we're always scanning the oldest znodes first. | |
1131 | */ | |
1132 | static int | |
1133 | zfs_sb_prune_aliases(zfs_sb_t *zsb, unsigned long nr_to_scan) | |
1134 | { | |
1135 | znode_t **zp_array, *zp; | |
1136 | int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *)); | |
1137 | int objects = 0; | |
1138 | int i = 0, j = 0; | |
1139 | ||
1140 | zp_array = kmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP); | |
1141 | ||
1142 | mutex_enter(&zsb->z_znodes_lock); | |
1143 | while ((zp = list_head(&zsb->z_all_znodes)) != NULL) { | |
1144 | ||
1145 | if ((i++ > nr_to_scan) || (j >= max_array)) | |
1146 | break; | |
1147 | ||
1148 | ASSERT(list_link_active(&zp->z_link_node)); | |
1149 | list_remove(&zsb->z_all_znodes, zp); | |
1150 | list_insert_tail(&zsb->z_all_znodes, zp); | |
1151 | ||
1152 | /* Skip active znodes and .zfs entries */ | |
1153 | if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir) | |
1154 | continue; | |
1155 | ||
1156 | if (igrab(ZTOI(zp)) == NULL) | |
1157 | continue; | |
1158 | ||
1159 | zp_array[j] = zp; | |
1160 | j++; | |
1161 | } | |
1162 | mutex_exit(&zsb->z_znodes_lock); | |
1163 | ||
1164 | for (i = 0; i < j; i++) { | |
1165 | zp = zp_array[i]; | |
1166 | ||
1167 | ASSERT3P(zp, !=, NULL); | |
1168 | d_prune_aliases(ZTOI(zp)); | |
1169 | ||
1170 | if (atomic_read(&ZTOI(zp)->i_count) == 1) | |
1171 | objects++; | |
1172 | ||
1173 | iput(ZTOI(zp)); | |
1174 | } | |
1175 | ||
1176 | kmem_free(zp_array, max_array * sizeof (znode_t *)); | |
1177 | ||
1178 | return (objects); | |
1179 | } | |
1180 | #endif /* HAVE_D_PRUNE_ALIASES */ | |
1181 | ||
2cbb06b5 BB |
1182 | /* |
1183 | * The ARC has requested that the filesystem drop entries from the dentry | |
1184 | * and inode caches. This can occur when the ARC needs to free meta data | |
1185 | * blocks but can't because they are all pinned by entries in these caches. | |
1186 | */ | |
ab26409d BB |
1187 | int |
1188 | zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) | |
1189 | { | |
1190 | zfs_sb_t *zsb = sb->s_fs_info; | |
2cbb06b5 BB |
1191 | int error = 0; |
1192 | #if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK) | |
ab26409d BB |
1193 | struct shrinker *shrinker = &sb->s_shrink; |
1194 | struct shrink_control sc = { | |
1195 | .nr_to_scan = nr_to_scan, | |
1196 | .gfp_mask = GFP_KERNEL, | |
1197 | }; | |
2cbb06b5 | 1198 | #endif |
ab26409d BB |
1199 | |
1200 | ZFS_ENTER(zsb); | |
2cbb06b5 | 1201 | |
90947b23 TC |
1202 | #if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \ |
1203 | defined(SHRINK_CONTROL_HAS_NID) && \ | |
1204 | defined(SHRINKER_NUMA_AWARE) | |
1205 | if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) { | |
1206 | *objects = 0; | |
1207 | for_each_online_node(sc.nid) | |
1208 | *objects += (*shrinker->scan_objects)(shrinker, &sc); | |
1209 | } else { | |
1210 | *objects = (*shrinker->scan_objects)(shrinker, &sc); | |
1211 | } | |
1212 | #elif defined(HAVE_SPLIT_SHRINKER_CALLBACK) | |
3c832b8c | 1213 | *objects = (*shrinker->scan_objects)(shrinker, &sc); |
2cbb06b5 | 1214 | #elif defined(HAVE_SHRINK) |
ab26409d | 1215 | *objects = (*shrinker->shrink)(shrinker, &sc); |
218b4e0a BB |
1216 | #elif defined(HAVE_D_PRUNE_ALIASES) |
1217 | *objects = zfs_sb_prune_aliases(zsb, nr_to_scan); | |
2cbb06b5 | 1218 | #else |
218b4e0a | 1219 | #error "No available dentry and inode cache pruning mechanism." |
3c832b8c | 1220 | #endif |
ab26409d BB |
1221 | ZFS_EXIT(zsb); |
1222 | ||
2cbb06b5 BB |
1223 | dprintf_ds(zsb->z_os->os_dsl_dataset, |
1224 | "pruning, nr_to_scan=%lu objects=%d error=%d\n", | |
1225 | nr_to_scan, *objects, error); | |
1226 | ||
1227 | return (error); | |
ab26409d BB |
1228 | } |
1229 | EXPORT_SYMBOL(zfs_sb_prune); | |
ab26409d | 1230 | |
34dc7c2f | 1231 | /* |
7b3e34ba | 1232 | * Teardown the zfs_sb_t. |
34dc7c2f BB |
1233 | * |
1234 | * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' | |
1235 | * and 'z_teardown_inactive_lock' held. | |
1236 | */ | |
3558fd73 | 1237 | int |
86f35f34 | 1238 | zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting) |
34dc7c2f BB |
1239 | { |
1240 | znode_t *zp; | |
1241 | ||
fd23720a BB |
1242 | /* |
1243 | * If someone has not already unmounted this file system, | |
1244 | * drain the iput_taskq to ensure all active references to the | |
1245 | * zfs_sb_t have been handled only then can it be safely destroyed. | |
1246 | */ | |
f0da4d15 CD |
1247 | if (zsb->z_os) { |
1248 | /* | |
1249 | * If we're unmounting we have to wait for the list to | |
1250 | * drain completely. | |
1251 | * | |
1252 | * If we're not unmounting there's no guarantee the list | |
1253 | * will drain completely, but iputs run from the taskq | |
1254 | * may add the parents of dir-based xattrs to the taskq | |
1255 | * so we want to wait for these. | |
1256 | * | |
1257 | * We can safely read z_nr_znodes without locking because the | |
1258 | * VFS has already blocked operations which add to the | |
1259 | * z_all_znodes list and thus increment z_nr_znodes. | |
1260 | */ | |
1261 | int round = 0; | |
1262 | while (zsb->z_nr_znodes > 0) { | |
c5528b9b BB |
1263 | taskq_wait_outstanding(dsl_pool_iput_taskq( |
1264 | dmu_objset_pool(zsb->z_os)), 0); | |
f0da4d15 CD |
1265 | if (++round > 1 && !unmounting) |
1266 | break; | |
1267 | } | |
1268 | } | |
fd23720a | 1269 | |
e16b3fcc | 1270 | rrm_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1271 | |
1272 | if (!unmounting) { | |
1273 | /* | |
ceb43b93 BB |
1274 | * We purge the parent filesystem's super block as the |
1275 | * parent filesystem and all of its snapshots have their | |
1276 | * inode's super block set to the parent's filesystem's | |
1277 | * super block. Note, 'z_parent' is self referential | |
1278 | * for non-snapshots. | |
34dc7c2f | 1279 | */ |
ceb43b93 | 1280 | shrink_dcache_sb(zsb->z_parent->z_sb); |
34dc7c2f BB |
1281 | } |
1282 | ||
1283 | /* | |
1284 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1285 | * threads are blocked as zil_close can call zfs_inactive. | |
1286 | */ | |
3558fd73 BB |
1287 | if (zsb->z_log) { |
1288 | zil_close(zsb->z_log); | |
1289 | zsb->z_log = NULL; | |
34dc7c2f BB |
1290 | } |
1291 | ||
3558fd73 | 1292 | rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1293 | |
1294 | /* | |
1295 | * If we are not unmounting (ie: online recv) and someone already | |
1296 | * unmounted this file system while we were doing the switcheroo, | |
1297 | * or a reopen of z_os failed then just bail out now. | |
1298 | */ | |
3558fd73 BB |
1299 | if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { |
1300 | rw_exit(&zsb->z_teardown_inactive_lock); | |
e16b3fcc | 1301 | rrm_exit(&zsb->z_teardown_lock, FTAG); |
2e528b49 | 1302 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1303 | } |
1304 | ||
1305 | /* | |
7b3e34ba BB |
1306 | * At this point there are no VFS ops active, and any new VFS ops |
1307 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1308 | * relevant for forced unmount). | |
34dc7c2f BB |
1309 | * |
1310 | * Release all holds on dbufs. | |
1311 | */ | |
f0da4d15 CD |
1312 | if (!unmounting) { |
1313 | mutex_enter(&zsb->z_znodes_lock); | |
1314 | for (zp = list_head(&zsb->z_all_znodes); zp != NULL; | |
1315 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
1316 | if (zp->z_sa_hdl) | |
1317 | zfs_znode_dmu_fini(zp); | |
1318 | } | |
1319 | mutex_exit(&zsb->z_znodes_lock); | |
7b3e34ba | 1320 | } |
34dc7c2f BB |
1321 | |
1322 | /* | |
7b3e34ba | 1323 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1324 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1325 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1326 | */ |
1327 | if (unmounting) { | |
3558fd73 | 1328 | zsb->z_unmounted = B_TRUE; |
e16b3fcc | 1329 | rrm_exit(&zsb->z_teardown_lock, FTAG); |
3558fd73 | 1330 | rw_exit(&zsb->z_teardown_inactive_lock); |
34dc7c2f BB |
1331 | } |
1332 | ||
1333 | /* | |
1334 | * z_os will be NULL if there was an error in attempting to reopen | |
3558fd73 BB |
1335 | * zsb, so just return as the properties had already been |
1336 | * | |
34dc7c2f BB |
1337 | * unregistered and cached data had been evicted before. |
1338 | */ | |
3558fd73 | 1339 | if (zsb->z_os == NULL) |
34dc7c2f BB |
1340 | return (0); |
1341 | ||
1342 | /* | |
1343 | * Unregister properties. | |
1344 | */ | |
3558fd73 | 1345 | zfs_unregister_callbacks(zsb); |
34dc7c2f BB |
1346 | |
1347 | /* | |
1348 | * Evict cached data | |
1349 | */ | |
04434775 MA |
1350 | if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) && |
1351 | !zfs_is_readonly(zsb)) | |
1352 | txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); | |
13fe0198 | 1353 | dmu_objset_evict_dbufs(zsb->z_os); |
34dc7c2f BB |
1354 | |
1355 | return (0); | |
1356 | } | |
86f35f34 | 1357 | EXPORT_SYMBOL(zfs_sb_teardown); |
34dc7c2f | 1358 | |
8c45def2 BB |
1359 | #if !defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER) && \ |
1360 | !defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER) | |
5547c2f1 | 1361 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); |
8c45def2 | 1362 | #endif |
76659dc1 | 1363 | |
e5c39b95 | 1364 | int |
0282c413 | 1365 | zfs_domount(struct super_block *sb, zfs_mntopts_t *zmo, int silent) |
34dc7c2f | 1366 | { |
0282c413 | 1367 | const char *osname = zmo->z_osname; |
3558fd73 BB |
1368 | zfs_sb_t *zsb; |
1369 | struct inode *root_inode; | |
1370 | uint64_t recordsize; | |
1371 | int error; | |
34dc7c2f | 1372 | |
0282c413 | 1373 | error = zfs_sb_create(osname, zmo, &zsb); |
3558fd73 BB |
1374 | if (error) |
1375 | return (error); | |
1376 | ||
1377 | if ((error = dsl_prop_get_integer(osname, "recordsize", | |
1378 | &recordsize, NULL))) | |
1379 | goto out; | |
1380 | ||
1381 | zsb->z_sb = sb; | |
3558fd73 BB |
1382 | sb->s_fs_info = zsb; |
1383 | sb->s_magic = ZFS_SUPER_MAGIC; | |
1384 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1385 | sb->s_time_gran = 1; | |
1386 | sb->s_blocksize = recordsize; | |
1387 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 BB |
1388 | zsb->z_bdi.ra_pages = 0; |
1389 | sb->s_bdi = &zsb->z_bdi; | |
1390 | ||
8c45def2 | 1391 | error = -zpl_bdi_setup_and_register(&zsb->z_bdi, "zfs"); |
5547c2f1 BB |
1392 | if (error) |
1393 | goto out; | |
3558fd73 BB |
1394 | |
1395 | /* Set callback operations for the file system. */ | |
1396 | sb->s_op = &zpl_super_operations; | |
1397 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1398 | sb->s_export_op = &zpl_export_operations; |
ee930353 BB |
1399 | #ifdef HAVE_S_D_OP |
1400 | sb->s_d_op = &zpl_dentry_operations; | |
1401 | #endif /* HAVE_S_D_OP */ | |
3558fd73 BB |
1402 | |
1403 | /* Set features for file system. */ | |
1404 | zfs_set_fuid_feature(zsb); | |
1405 | ||
1406 | if (dmu_objset_is_snapshot(zsb->z_os)) { | |
1407 | uint64_t pval; | |
1408 | ||
1409 | atime_changed_cb(zsb, B_FALSE); | |
1410 | readonly_changed_cb(zsb, B_TRUE); | |
d1d7e268 MK |
1411 | if ((error = dsl_prop_get_integer(osname, |
1412 | "xattr", &pval, NULL))) | |
3558fd73 BB |
1413 | goto out; |
1414 | xattr_changed_cb(zsb, pval); | |
d1d7e268 MK |
1415 | if ((error = dsl_prop_get_integer(osname, |
1416 | "acltype", &pval, NULL))) | |
023699cd MM |
1417 | goto out; |
1418 | acltype_changed_cb(zsb, pval); | |
3558fd73 BB |
1419 | zsb->z_issnap = B_TRUE; |
1420 | zsb->z_os->os_sync = ZFS_SYNC_DISABLED; | |
278bee93 | 1421 | zsb->z_snap_defer_time = jiffies; |
3558fd73 BB |
1422 | |
1423 | mutex_enter(&zsb->z_os->os_user_ptr_lock); | |
1424 | dmu_objset_set_user(zsb->z_os, zsb); | |
1425 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
1426 | } else { | |
1427 | error = zfs_sb_setup(zsb, B_TRUE); | |
34dc7c2f BB |
1428 | } |
1429 | ||
3558fd73 BB |
1430 | /* Allocate a root inode for the filesystem. */ |
1431 | error = zfs_root(zsb, &root_inode); | |
1432 | if (error) { | |
1433 | (void) zfs_umount(sb); | |
1434 | goto out; | |
34dc7c2f BB |
1435 | } |
1436 | ||
3558fd73 | 1437 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1438 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1439 | if (sb->s_root == NULL) { |
1440 | (void) zfs_umount(sb); | |
2e528b49 | 1441 | error = SET_ERROR(ENOMEM); |
3558fd73 BB |
1442 | goto out; |
1443 | } | |
ebe7e575 BB |
1444 | |
1445 | if (!zsb->z_issnap) | |
1446 | zfsctl_create(zsb); | |
2cbb06b5 BB |
1447 | |
1448 | zsb->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); | |
3558fd73 BB |
1449 | out: |
1450 | if (error) { | |
1451 | dmu_objset_disown(zsb->z_os, zsb); | |
1452 | zfs_sb_free(zsb); | |
1453 | } | |
34dc7c2f | 1454 | |
3558fd73 BB |
1455 | return (error); |
1456 | } | |
1457 | EXPORT_SYMBOL(zfs_domount); | |
1458 | ||
ebe7e575 BB |
1459 | /* |
1460 | * Called when an unmount is requested and certain sanity checks have | |
1461 | * already passed. At this point no dentries or inodes have been reclaimed | |
1462 | * from their respective caches. We drop the extra reference on the .zfs | |
1463 | * control directory to allow everything to be reclaimed. All snapshots | |
1464 | * must already have been unmounted to reach this point. | |
1465 | */ | |
1466 | void | |
1467 | zfs_preumount(struct super_block *sb) | |
1468 | { | |
1469 | zfs_sb_t *zsb = sb->s_fs_info; | |
1470 | ||
278bee93 BB |
1471 | if (zsb) |
1472 | zfsctl_destroy(sb->s_fs_info); | |
ebe7e575 BB |
1473 | } |
1474 | EXPORT_SYMBOL(zfs_preumount); | |
1475 | ||
1476 | /* | |
1477 | * Called once all other unmount released tear down has occurred. | |
1478 | * It is our responsibility to release any remaining infrastructure. | |
1479 | */ | |
3558fd73 BB |
1480 | /*ARGSUSED*/ |
1481 | int | |
1482 | zfs_umount(struct super_block *sb) | |
1483 | { | |
1484 | zfs_sb_t *zsb = sb->s_fs_info; | |
1485 | objset_t *os; | |
1486 | ||
2cbb06b5 | 1487 | arc_remove_prune_callback(zsb->z_arc_prune); |
86f35f34 | 1488 | VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0); |
3558fd73 | 1489 | os = zsb->z_os; |
5547c2f1 | 1490 | bdi_destroy(sb->s_bdi); |
76659dc1 | 1491 | |
34dc7c2f BB |
1492 | /* |
1493 | * z_os will be NULL if there was an error in | |
3558fd73 | 1494 | * attempting to reopen zsb. |
34dc7c2f BB |
1495 | */ |
1496 | if (os != NULL) { | |
1497 | /* | |
1498 | * Unset the objset user_ptr. | |
1499 | */ | |
428870ff | 1500 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1501 | dmu_objset_set_user(os, NULL); |
428870ff | 1502 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1503 | |
1504 | /* | |
b128c09f | 1505 | * Finally release the objset |
34dc7c2f | 1506 | */ |
3558fd73 | 1507 | dmu_objset_disown(os, zsb); |
34dc7c2f BB |
1508 | } |
1509 | ||
3558fd73 | 1510 | zfs_sb_free(zsb); |
34dc7c2f BB |
1511 | return (0); |
1512 | } | |
e5c39b95 | 1513 | EXPORT_SYMBOL(zfs_umount); |
34dc7c2f | 1514 | |
0de19dad | 1515 | int |
0282c413 | 1516 | zfs_remount(struct super_block *sb, int *flags, zfs_mntopts_t *zmo) |
0de19dad | 1517 | { |
0282c413 BB |
1518 | zfs_sb_t *zsb = sb->s_fs_info; |
1519 | int error; | |
1520 | ||
1521 | zfs_unregister_callbacks(zsb); | |
1522 | error = zfs_register_callbacks(zsb); | |
1523 | ||
1524 | return (error); | |
0de19dad BB |
1525 | } |
1526 | EXPORT_SYMBOL(zfs_remount); | |
1527 | ||
e5c39b95 | 1528 | int |
2cf7f52b | 1529 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1530 | { |
2cf7f52b | 1531 | zfs_sb_t *zsb = sb->s_fs_info; |
34dc7c2f BB |
1532 | znode_t *zp; |
1533 | uint64_t object = 0; | |
1534 | uint64_t fid_gen = 0; | |
1535 | uint64_t gen_mask; | |
1536 | uint64_t zp_gen; | |
3558fd73 | 1537 | int i, err; |
34dc7c2f | 1538 | |
3558fd73 | 1539 | *ipp = NULL; |
34dc7c2f | 1540 | |
3558fd73 | 1541 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1542 | |
1543 | if (fidp->fid_len == LONG_FID_LEN) { | |
1544 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1545 | uint64_t objsetid = 0; | |
1546 | uint64_t setgen = 0; | |
1547 | ||
1548 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1549 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1550 | ||
1551 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1552 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1553 | ||
3558fd73 | 1554 | ZFS_EXIT(zsb); |
34dc7c2f | 1555 | |
ebe7e575 | 1556 | err = zfsctl_lookup_objset(sb, objsetid, &zsb); |
34dc7c2f | 1557 | if (err) |
2e528b49 | 1558 | return (SET_ERROR(EINVAL)); |
ebe7e575 | 1559 | |
3558fd73 | 1560 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1561 | } |
1562 | ||
1563 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { | |
1564 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1565 | ||
1566 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1567 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
1568 | ||
1569 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
1570 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1571 | } else { | |
3558fd73 | 1572 | ZFS_EXIT(zsb); |
2e528b49 | 1573 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1574 | } |
1575 | ||
1576 | /* A zero fid_gen means we are in the .zfs control directories */ | |
1577 | if (fid_gen == 0 && | |
1578 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
3558fd73 BB |
1579 | *ipp = zsb->z_ctldir; |
1580 | ASSERT(*ipp != NULL); | |
34dc7c2f | 1581 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1582 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1583 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1584 | } else { |
3558fd73 | 1585 | igrab(*ipp); |
34dc7c2f | 1586 | } |
3558fd73 | 1587 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1588 | return (0); |
1589 | } | |
1590 | ||
1591 | gen_mask = -1ULL >> (64 - 8 * i); | |
1592 | ||
29e57d15 | 1593 | dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); |
3558fd73 BB |
1594 | if ((err = zfs_zget(zsb, object, &zp))) { |
1595 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1596 | return (err); |
1597 | } | |
3558fd73 | 1598 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, |
428870ff BB |
1599 | sizeof (uint64_t)); |
1600 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1601 | if (zp_gen == 0) |
1602 | zp_gen = 1; | |
0500e835 BB |
1603 | if ((fid_gen == 0) && (zsb->z_root == object)) |
1604 | fid_gen = zp_gen; | |
34dc7c2f | 1605 | if (zp->z_unlinked || zp_gen != fid_gen) { |
29e57d15 NB |
1606 | dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, |
1607 | fid_gen); | |
3558fd73 BB |
1608 | iput(ZTOI(zp)); |
1609 | ZFS_EXIT(zsb); | |
2e528b49 | 1610 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1611 | } |
1612 | ||
3558fd73 BB |
1613 | *ipp = ZTOI(zp); |
1614 | if (*ipp) | |
1615 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1616 | |
3558fd73 | 1617 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1618 | return (0); |
1619 | } | |
e5c39b95 | 1620 | EXPORT_SYMBOL(zfs_vget); |
34dc7c2f BB |
1621 | |
1622 | /* | |
7b3e34ba | 1623 | * Block out VFS ops and close zfs_sb_t |
34dc7c2f BB |
1624 | * |
1625 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
831baf06 KW |
1626 | * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying |
1627 | * dataset and objset intact so that they can be atomically handed off during | |
1628 | * a subsequent rollback or recv operation and the resume thereafter. | |
34dc7c2f BB |
1629 | */ |
1630 | int | |
3558fd73 | 1631 | zfs_suspend_fs(zfs_sb_t *zsb) |
34dc7c2f BB |
1632 | { |
1633 | int error; | |
1634 | ||
86f35f34 | 1635 | if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0) |
34dc7c2f | 1636 | return (error); |
7b3e34ba | 1637 | |
34dc7c2f BB |
1638 | return (0); |
1639 | } | |
e5c39b95 | 1640 | EXPORT_SYMBOL(zfs_suspend_fs); |
34dc7c2f BB |
1641 | |
1642 | /* | |
7b3e34ba | 1643 | * Reopen zfs_sb_t and release VFS ops. |
34dc7c2f BB |
1644 | */ |
1645 | int | |
3558fd73 | 1646 | zfs_resume_fs(zfs_sb_t *zsb, const char *osname) |
34dc7c2f | 1647 | { |
2008e920 | 1648 | int err, err2; |
831baf06 KW |
1649 | znode_t *zp; |
1650 | uint64_t sa_obj = 0; | |
34dc7c2f | 1651 | |
e16b3fcc | 1652 | ASSERT(RRM_WRITE_HELD(&zsb->z_teardown_lock)); |
3558fd73 | 1653 | ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); |
34dc7c2f | 1654 | |
831baf06 KW |
1655 | /* |
1656 | * We already own this, so just hold and rele it to update the | |
1657 | * objset_t, as the one we had before may have been evicted. | |
1658 | */ | |
1659 | VERIFY0(dmu_objset_hold(osname, zsb, &zsb->z_os)); | |
1660 | VERIFY3P(zsb->z_os->os_dsl_dataset->ds_owner, ==, zsb); | |
1661 | VERIFY(dsl_dataset_long_held(zsb->z_os->os_dsl_dataset)); | |
1662 | dmu_objset_rele(zsb->z_os, zsb); | |
428870ff | 1663 | |
831baf06 KW |
1664 | /* |
1665 | * Make sure version hasn't changed | |
1666 | */ | |
428870ff | 1667 | |
831baf06 KW |
1668 | err = zfs_get_zplprop(zsb->z_os, ZFS_PROP_VERSION, |
1669 | &zsb->z_version); | |
c1fabe79 | 1670 | |
831baf06 KW |
1671 | if (err) |
1672 | goto bail; | |
428870ff | 1673 | |
831baf06 KW |
1674 | err = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, |
1675 | ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
c1fabe79 | 1676 | |
831baf06 KW |
1677 | if (err && zsb->z_version >= ZPL_VERSION_SA) |
1678 | goto bail; | |
428870ff | 1679 | |
831baf06 KW |
1680 | if ((err = sa_setup(zsb->z_os, sa_obj, |
1681 | zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) | |
1682 | goto bail; | |
34dc7c2f | 1683 | |
831baf06 KW |
1684 | if (zsb->z_version >= ZPL_VERSION_SA) |
1685 | sa_register_update_callback(zsb->z_os, | |
1686 | zfs_sa_upgrade); | |
c1fabe79 | 1687 | |
831baf06 | 1688 | VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); |
c1fabe79 | 1689 | |
831baf06 KW |
1690 | zfs_set_fuid_feature(zsb); |
1691 | zsb->z_rollback_time = jiffies; | |
34dc7c2f | 1692 | |
831baf06 KW |
1693 | /* |
1694 | * Attempt to re-establish all the active inodes with their | |
1695 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
1696 | * and mark it stale. This prevents a collision if a new | |
1697 | * inode/object is created which must use the same inode | |
1698 | * number. The stale inode will be be released when the | |
1699 | * VFS prunes the dentry holding the remaining references | |
1700 | * on the stale inode. | |
1701 | */ | |
1702 | mutex_enter(&zsb->z_znodes_lock); | |
1703 | for (zp = list_head(&zsb->z_all_znodes); zp; | |
1704 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
2008e920 TC |
1705 | err2 = zfs_rezget(zp); |
1706 | if (err2) { | |
831baf06 KW |
1707 | remove_inode_hash(ZTOI(zp)); |
1708 | zp->z_is_stale = B_TRUE; | |
34dc7c2f | 1709 | } |
34dc7c2f | 1710 | } |
831baf06 | 1711 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f | 1712 | |
428870ff | 1713 | bail: |
7b3e34ba | 1714 | /* release the VFS ops */ |
3558fd73 | 1715 | rw_exit(&zsb->z_teardown_inactive_lock); |
e16b3fcc | 1716 | rrm_exit(&zsb->z_teardown_lock, FTAG); |
34dc7c2f BB |
1717 | |
1718 | if (err) { | |
1719 | /* | |
831baf06 KW |
1720 | * Since we couldn't setup the sa framework, try to force |
1721 | * unmount this file system. | |
34dc7c2f | 1722 | */ |
ab68b6e5 TC |
1723 | if (zsb->z_os) |
1724 | (void) zfs_umount(zsb->z_sb); | |
34dc7c2f BB |
1725 | } |
1726 | return (err); | |
1727 | } | |
e5c39b95 | 1728 | EXPORT_SYMBOL(zfs_resume_fs); |
34dc7c2f | 1729 | |
34dc7c2f | 1730 | int |
3558fd73 | 1731 | zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) |
34dc7c2f BB |
1732 | { |
1733 | int error; | |
3558fd73 | 1734 | objset_t *os = zsb->z_os; |
34dc7c2f | 1735 | dmu_tx_t *tx; |
34dc7c2f BB |
1736 | |
1737 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
2e528b49 | 1738 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1739 | |
3558fd73 | 1740 | if (newvers < zsb->z_version) |
2e528b49 | 1741 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1742 | |
428870ff | 1743 | if (zfs_spa_version_map(newvers) > |
3558fd73 | 1744 | spa_version(dmu_objset_spa(zsb->z_os))) |
2e528b49 | 1745 | return (SET_ERROR(ENOTSUP)); |
428870ff | 1746 | |
34dc7c2f | 1747 | tx = dmu_tx_create(os); |
9babb374 | 1748 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
3558fd73 | 1749 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1750 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1751 | ZFS_SA_ATTRS); | |
1752 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
1753 | } | |
34dc7c2f BB |
1754 | error = dmu_tx_assign(tx, TXG_WAIT); |
1755 | if (error) { | |
1756 | dmu_tx_abort(tx); | |
9babb374 BB |
1757 | return (error); |
1758 | } | |
428870ff | 1759 | |
9babb374 BB |
1760 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
1761 | 8, 1, &newvers, tx); | |
1762 | ||
1763 | if (error) { | |
1764 | dmu_tx_commit(tx); | |
1765 | return (error); | |
34dc7c2f | 1766 | } |
34dc7c2f | 1767 | |
3558fd73 | 1768 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1769 | uint64_t sa_obj; |
1770 | ||
3558fd73 | 1771 | ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, |
428870ff BB |
1772 | SPA_VERSION_SA); |
1773 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1774 | DMU_OT_NONE, 0, tx); | |
1775 | ||
1776 | error = zap_add(os, MASTER_NODE_OBJ, | |
1777 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c99c9001 | 1778 | ASSERT0(error); |
428870ff BB |
1779 | |
1780 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
1781 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1782 | } | |
1783 | ||
6f1ffb06 MA |
1784 | spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, |
1785 | "from %llu to %llu", zsb->z_version, newvers); | |
9babb374 | 1786 | |
34dc7c2f BB |
1787 | dmu_tx_commit(tx); |
1788 | ||
3558fd73 | 1789 | zsb->z_version = newvers; |
9babb374 | 1790 | |
c1fabe79 | 1791 | zfs_set_fuid_feature(zsb); |
9babb374 BB |
1792 | |
1793 | return (0); | |
34dc7c2f | 1794 | } |
e5c39b95 | 1795 | EXPORT_SYMBOL(zfs_set_version); |
34dc7c2f BB |
1796 | |
1797 | /* | |
1798 | * Read a property stored within the master node. | |
1799 | */ | |
1800 | int | |
1801 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
1802 | { | |
1803 | const char *pname; | |
2e528b49 | 1804 | int error = SET_ERROR(ENOENT); |
34dc7c2f BB |
1805 | |
1806 | /* | |
1807 | * Look up the file system's value for the property. For the | |
1808 | * version property, we look up a slightly different string. | |
1809 | */ | |
1810 | if (prop == ZFS_PROP_VERSION) | |
1811 | pname = ZPL_VERSION_STR; | |
1812 | else | |
1813 | pname = zfs_prop_to_name(prop); | |
1814 | ||
b128c09f BB |
1815 | if (os != NULL) |
1816 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
1817 | |
1818 | if (error == ENOENT) { | |
1819 | /* No value set, use the default value */ | |
1820 | switch (prop) { | |
1821 | case ZFS_PROP_VERSION: | |
1822 | *value = ZPL_VERSION; | |
1823 | break; | |
1824 | case ZFS_PROP_NORMALIZE: | |
1825 | case ZFS_PROP_UTF8ONLY: | |
1826 | *value = 0; | |
1827 | break; | |
1828 | case ZFS_PROP_CASE: | |
1829 | *value = ZFS_CASE_SENSITIVE; | |
1830 | break; | |
023699cd MM |
1831 | case ZFS_PROP_ACLTYPE: |
1832 | *value = ZFS_ACLTYPE_OFF; | |
1833 | break; | |
34dc7c2f BB |
1834 | default: |
1835 | return (error); | |
1836 | } | |
1837 | error = 0; | |
1838 | } | |
1839 | return (error); | |
1840 | } | |
86f35f34 | 1841 | EXPORT_SYMBOL(zfs_get_zplprop); |
3558fd73 BB |
1842 | |
1843 | void | |
1844 | zfs_init(void) | |
1845 | { | |
ebe7e575 | 1846 | zfsctl_init(); |
3558fd73 BB |
1847 | zfs_znode_init(); |
1848 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
1849 | register_filesystem(&zpl_fs_type); | |
1850 | } | |
1851 | ||
1852 | void | |
1853 | zfs_fini(void) | |
1854 | { | |
c5528b9b | 1855 | taskq_wait_outstanding(system_taskq, 0); |
3558fd73 BB |
1856 | unregister_filesystem(&zpl_fs_type); |
1857 | zfs_znode_fini(); | |
ebe7e575 | 1858 | zfsctl_fini(); |
3558fd73 | 1859 | } |