]>
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. |
a08ee875 | 23 | * Copyright (c) 2013 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 | ||
a08ee875 LG |
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 | ||
a08ee875 LG |
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: | |
173 | #ifdef CONFIG_FS_POSIX_ACL | |
174 | zsb->z_acl_type = ZFS_ACLTYPE_POSIXACL; | |
175 | zsb->z_sb->s_flags |= MS_POSIXACL; | |
176 | #else | |
177 | zsb->z_acl_type = ZFS_ACLTYPE_OFF; | |
178 | zsb->z_sb->s_flags &= ~MS_POSIXACL; | |
179 | #endif /* CONFIG_FS_POSIX_ACL */ | |
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; |
e10b0808 AX |
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; |
e10b0808 | 265 | zfs_mntopts_t *zmo = zsb->z_mntopts; |
34dc7c2f BB |
266 | int error = 0; |
267 | ||
e10b0808 AX |
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); | |
a08ee875 LG |
290 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
291 | error = dsl_prop_register(ds, | |
292 | zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb); | |
ea04106b | 293 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 294 | zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb); |
34dc7c2f | 295 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 296 | zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb); |
34dc7c2f | 297 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 298 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb); |
34dc7c2f | 299 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 300 | zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb); |
34dc7c2f | 301 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 302 | zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb); |
34dc7c2f | 303 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 304 | zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb); |
34dc7c2f | 305 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 306 | zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb); |
34dc7c2f | 307 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 308 | zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb); |
34dc7c2f | 309 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 310 | zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb); |
34dc7c2f | 311 | error = error ? error : dsl_prop_register(ds, |
a08ee875 | 312 | zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb); |
2cf7f52b | 313 | error = error ? error : dsl_prop_register(ds, |
a08ee875 LG |
314 | zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb); |
315 | error = error ? error : dsl_prop_register(ds, | |
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 | ||
e10b0808 AX |
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); | |
c06d4368 | 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 | */ | |
a08ee875 LG |
349 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME), |
350 | atime_changed_cb, zsb); | |
351 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RELATIME), | |
352 | relatime_changed_cb, zsb); | |
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); | |
367 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE), | |
368 | acltype_changed_cb, zsb); | |
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) | |
a08ee875 | 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) | |
a08ee875 | 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 | } |
a08ee875 | 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: |
a08ee875 | 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)) |
a08ee875 | 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) |
a08ee875 | 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)) |
a08ee875 | 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) | |
a08ee875 | 580 | return (SET_ERROR(EINVAL)); |
9babb374 | 581 | |
3558fd73 | 582 | if (zsb->z_version < ZPL_VERSION_USERSPACE) |
a08ee875 | 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 | |
e10b0808 AX |
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 |
e10b0808 | 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 | 696 | uint64_t zval; |
4d815aed | 697 | int i, size, error; |
428870ff | 698 | uint64_t sa_obj; |
9babb374 | 699 | |
ea04106b | 700 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
9babb374 | 701 | |
428870ff | 702 | /* |
68d83c55 | 703 | * Optional temporary mount options, free'd in zfs_sb_free(). |
428870ff | 704 | */ |
68d83c55 | 705 | zsb->z_mntopts = (zmo ? zmo : zfs_mntopts_alloc()); |
9babb374 | 706 | |
e10b0808 | 707 | /* |
68d83c55 AX |
708 | * We claim to always be readonly so we can open snapshots; |
709 | * other ZPL code will prevent us from writing to snapshots. | |
e10b0808 | 710 | */ |
68d83c55 AX |
711 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); |
712 | if (error) | |
713 | goto out_zmo; | |
e10b0808 | 714 | |
9babb374 BB |
715 | /* |
716 | * Initialize the zfs-specific filesystem structure. | |
4d815aed | 717 | * Should probably make this a kmem cache, shuffle fields. |
9babb374 | 718 | */ |
2cf7f52b | 719 | zsb->z_sb = NULL; |
3558fd73 | 720 | zsb->z_parent = zsb; |
e10b0808 | 721 | zsb->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; |
3558fd73 BB |
722 | zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; |
723 | zsb->z_os = os; | |
9babb374 | 724 | |
3558fd73 | 725 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); |
9babb374 BB |
726 | if (error) { |
727 | goto out; | |
ea04106b | 728 | } else if (zsb->z_version > ZPL_VERSION) { |
a08ee875 | 729 | error = SET_ERROR(ENOTSUP); |
9babb374 BB |
730 | goto out; |
731 | } | |
9babb374 BB |
732 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) |
733 | goto out; | |
3558fd73 | 734 | zsb->z_norm = (int)zval; |
9babb374 BB |
735 | |
736 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
737 | goto out; | |
3558fd73 | 738 | zsb->z_utf8 = (zval != 0); |
9babb374 BB |
739 | |
740 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
741 | goto out; | |
3558fd73 | 742 | zsb->z_case = (uint_t)zval; |
9babb374 | 743 | |
a08ee875 LG |
744 | if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0) |
745 | goto out; | |
746 | zsb->z_acl_type = (uint_t)zval; | |
747 | ||
9babb374 BB |
748 | /* |
749 | * Fold case on file systems that are always or sometimes case | |
750 | * insensitive. | |
751 | */ | |
3558fd73 BB |
752 | if (zsb->z_case == ZFS_CASE_INSENSITIVE || |
753 | zsb->z_case == ZFS_CASE_MIXED) | |
754 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 755 | |
3558fd73 BB |
756 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
757 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
428870ff | 758 | |
3558fd73 | 759 | if (zsb->z_use_sa) { |
428870ff BB |
760 | /* should either have both of these objects or none */ |
761 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
762 | &sa_obj); | |
763 | if (error) | |
591fb62f | 764 | goto out; |
82a37189 BB |
765 | |
766 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval); | |
767 | if ((error == 0) && (zval == ZFS_XATTR_SA)) | |
768 | zsb->z_xattr_sa = B_TRUE; | |
428870ff BB |
769 | } else { |
770 | /* | |
771 | * Pre SA versions file systems should never touch | |
772 | * either the attribute registration or layout objects. | |
773 | */ | |
774 | sa_obj = 0; | |
775 | } | |
776 | ||
572e2857 | 777 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
3558fd73 | 778 | &zsb->z_attr_table); |
572e2857 BB |
779 | if (error) |
780 | goto out; | |
428870ff | 781 | |
3558fd73 | 782 | if (zsb->z_version >= ZPL_VERSION_SA) |
428870ff | 783 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
784 | |
785 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
3558fd73 | 786 | &zsb->z_root); |
9babb374 BB |
787 | if (error) |
788 | goto out; | |
3558fd73 | 789 | ASSERT(zsb->z_root != 0); |
9babb374 BB |
790 | |
791 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
3558fd73 | 792 | &zsb->z_unlinkedobj); |
9babb374 BB |
793 | if (error) |
794 | goto out; | |
795 | ||
796 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
797 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
3558fd73 | 798 | 8, 1, &zsb->z_userquota_obj); |
9babb374 BB |
799 | if (error && error != ENOENT) |
800 | goto out; | |
801 | ||
802 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
803 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
3558fd73 | 804 | 8, 1, &zsb->z_groupquota_obj); |
9babb374 BB |
805 | if (error && error != ENOENT) |
806 | goto out; | |
807 | ||
808 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
3558fd73 | 809 | &zsb->z_fuid_obj); |
9babb374 BB |
810 | if (error && error != ENOENT) |
811 | goto out; | |
812 | ||
813 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
3558fd73 | 814 | &zsb->z_shares_dir); |
9babb374 BB |
815 | if (error && error != ENOENT) |
816 | goto out; | |
817 | ||
3558fd73 BB |
818 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
819 | mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
820 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
9babb374 | 821 | offsetof(znode_t, z_link_node)); |
e10b0808 | 822 | rrm_init(&zsb->z_teardown_lock, B_FALSE); |
3558fd73 BB |
823 | rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); |
824 | rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
ea04106b | 825 | |
4d815aed AX |
826 | size = MIN(1 << (highbit64(zfs_object_mutex_size)-1), ZFS_OBJ_MTX_MAX); |
827 | zsb->z_hold_size = size; | |
828 | zsb->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, KM_SLEEP); | |
829 | zsb->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
830 | for (i = 0; i != size; i++) { | |
831 | avl_create(&zsb->z_hold_trees[i], zfs_znode_hold_compare, | |
832 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); | |
833 | mutex_init(&zsb->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); | |
834 | } | |
9babb374 | 835 | |
3558fd73 | 836 | *zsbp = zsb; |
9babb374 BB |
837 | return (0); |
838 | ||
839 | out: | |
3558fd73 | 840 | dmu_objset_disown(os, zsb); |
68d83c55 | 841 | out_zmo: |
3558fd73 | 842 | *zsbp = NULL; |
68d83c55 | 843 | zfs_mntopts_free(zsb->z_mntopts); |
3558fd73 | 844 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 BB |
845 | return (error); |
846 | } | |
86f35f34 | 847 | EXPORT_SYMBOL(zfs_sb_create); |
9babb374 | 848 | |
86f35f34 | 849 | int |
3558fd73 | 850 | zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) |
34dc7c2f | 851 | { |
34dc7c2f BB |
852 | int error; |
853 | ||
3558fd73 | 854 | error = zfs_register_callbacks(zsb); |
34dc7c2f BB |
855 | if (error) |
856 | return (error); | |
857 | ||
858 | /* | |
3558fd73 | 859 | * Set the objset user_ptr to track its zsb. |
34dc7c2f | 860 | */ |
3558fd73 BB |
861 | mutex_enter(&zsb->z_os->os_user_ptr_lock); |
862 | dmu_objset_set_user(zsb->z_os, zsb); | |
863 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
34dc7c2f | 864 | |
3558fd73 | 865 | zsb->z_log = zil_open(zsb->z_os, zfs_get_data); |
9babb374 | 866 | |
34dc7c2f BB |
867 | /* |
868 | * If we are not mounting (ie: online recv), then we don't | |
869 | * have to worry about replaying the log as we blocked all | |
870 | * operations out since we closed the ZIL. | |
871 | */ | |
872 | if (mounting) { | |
b128c09f BB |
873 | boolean_t readonly; |
874 | ||
34dc7c2f BB |
875 | /* |
876 | * During replay we remove the read only flag to | |
877 | * allow replays to succeed. | |
878 | */ | |
2cf7f52b | 879 | readonly = zfs_is_readonly(zsb); |
fb5f0bc8 | 880 | if (readonly != 0) |
2cf7f52b | 881 | readonly_changed_cb(zsb, B_FALSE); |
fb5f0bc8 | 882 | else |
3558fd73 | 883 | zfs_unlinked_drain(zsb); |
34dc7c2f | 884 | |
428870ff BB |
885 | /* |
886 | * Parse and replay the intent log. | |
887 | * | |
888 | * Because of ziltest, this must be done after | |
889 | * zfs_unlinked_drain(). (Further note: ziltest | |
890 | * doesn't use readonly mounts, where | |
891 | * zfs_unlinked_drain() isn't called.) This is because | |
892 | * ziltest causes spa_sync() to think it's committed, | |
893 | * but actually it is not, so the intent log contains | |
894 | * many txg's worth of changes. | |
895 | * | |
896 | * In particular, if object N is in the unlinked set in | |
897 | * the last txg to actually sync, then it could be | |
898 | * actually freed in a later txg and then reallocated | |
899 | * in a yet later txg. This would write a "create | |
900 | * object N" record to the intent log. Normally, this | |
901 | * would be fine because the spa_sync() would have | |
902 | * written out the fact that object N is free, before | |
903 | * we could write the "create object N" intent log | |
904 | * record. | |
905 | * | |
906 | * But when we are in ziltest mode, we advance the "open | |
907 | * txg" without actually spa_sync()-ing the changes to | |
908 | * disk. So we would see that object N is still | |
909 | * allocated and in the unlinked set, and there is an | |
910 | * intent log record saying to allocate it. | |
911 | */ | |
3558fd73 | 912 | if (spa_writeable(dmu_objset_spa(zsb->z_os))) { |
572e2857 | 913 | if (zil_replay_disable) { |
3558fd73 | 914 | zil_destroy(zsb->z_log, B_FALSE); |
572e2857 | 915 | } else { |
3558fd73 BB |
916 | zsb->z_replay = B_TRUE; |
917 | zil_replay(zsb->z_os, zsb, | |
572e2857 | 918 | zfs_replay_vector); |
3558fd73 | 919 | zsb->z_replay = B_FALSE; |
572e2857 | 920 | } |
fb5f0bc8 | 921 | } |
2cf7f52b BB |
922 | |
923 | /* restore readonly bit */ | |
924 | if (readonly != 0) | |
925 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
926 | } |
927 | ||
34dc7c2f BB |
928 | return (0); |
929 | } | |
86f35f34 | 930 | EXPORT_SYMBOL(zfs_sb_setup); |
34dc7c2f | 931 | |
9babb374 | 932 | void |
3558fd73 | 933 | zfs_sb_free(zfs_sb_t *zsb) |
34dc7c2f | 934 | { |
4d815aed | 935 | int i, size = zsb->z_hold_size; |
9babb374 | 936 | |
3558fd73 | 937 | zfs_fuid_destroy(zsb); |
9babb374 | 938 | |
3558fd73 BB |
939 | mutex_destroy(&zsb->z_znodes_lock); |
940 | mutex_destroy(&zsb->z_lock); | |
941 | list_destroy(&zsb->z_all_znodes); | |
e10b0808 | 942 | rrm_destroy(&zsb->z_teardown_lock); |
3558fd73 BB |
943 | rw_destroy(&zsb->z_teardown_inactive_lock); |
944 | rw_destroy(&zsb->z_fuid_lock); | |
4d815aed AX |
945 | for (i = 0; i != size; i++) { |
946 | avl_destroy(&zsb->z_hold_trees[i]); | |
947 | mutex_destroy(&zsb->z_hold_locks[i]); | |
948 | } | |
949 | vmem_free(zsb->z_hold_trees, sizeof (avl_tree_t) * size); | |
950 | vmem_free(zsb->z_hold_locks, sizeof (kmutex_t) * size); | |
e10b0808 | 951 | zfs_mntopts_free(zsb->z_mntopts); |
3558fd73 | 952 | kmem_free(zsb, sizeof (zfs_sb_t)); |
34dc7c2f | 953 | } |
86f35f34 | 954 | EXPORT_SYMBOL(zfs_sb_free); |
34dc7c2f | 955 | |
9babb374 | 956 | static void |
3558fd73 | 957 | zfs_set_fuid_feature(zfs_sb_t *zsb) |
9babb374 | 958 | { |
3558fd73 BB |
959 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
960 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
9babb374 | 961 | } |
34dc7c2f BB |
962 | |
963 | void | |
3558fd73 | 964 | zfs_unregister_callbacks(zfs_sb_t *zsb) |
34dc7c2f | 965 | { |
3558fd73 | 966 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
967 | struct dsl_dataset *ds; |
968 | ||
969 | /* | |
970 | * Unregister properties. | |
971 | */ | |
972 | if (!dmu_objset_is_snapshot(os)) { | |
973 | ds = dmu_objset_ds(os); | |
974 | VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, | |
3558fd73 | 975 | zsb) == 0); |
34dc7c2f | 976 | |
a08ee875 LG |
977 | VERIFY(dsl_prop_unregister(ds, "relatime", relatime_changed_cb, |
978 | zsb) == 0); | |
979 | ||
34dc7c2f | 980 | VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, |
3558fd73 | 981 | zsb) == 0); |
34dc7c2f BB |
982 | |
983 | VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, | |
3558fd73 | 984 | zsb) == 0); |
34dc7c2f BB |
985 | |
986 | VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, | |
3558fd73 | 987 | zsb) == 0); |
34dc7c2f BB |
988 | |
989 | VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, | |
3558fd73 | 990 | zsb) == 0); |
34dc7c2f BB |
991 | |
992 | VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, | |
3558fd73 | 993 | zsb) == 0); |
34dc7c2f BB |
994 | |
995 | VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, | |
3558fd73 | 996 | zsb) == 0); |
34dc7c2f BB |
997 | |
998 | VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, | |
3558fd73 | 999 | zsb) == 0); |
34dc7c2f | 1000 | |
a08ee875 LG |
1001 | VERIFY(dsl_prop_unregister(ds, "acltype", acltype_changed_cb, |
1002 | zsb) == 0); | |
1003 | ||
34dc7c2f | 1004 | VERIFY(dsl_prop_unregister(ds, "aclinherit", |
3558fd73 | 1005 | acl_inherit_changed_cb, zsb) == 0); |
34dc7c2f BB |
1006 | |
1007 | VERIFY(dsl_prop_unregister(ds, "vscan", | |
3558fd73 | 1008 | vscan_changed_cb, zsb) == 0); |
2cf7f52b BB |
1009 | |
1010 | VERIFY(dsl_prop_unregister(ds, "nbmand", | |
1011 | nbmand_changed_cb, zsb) == 0); | |
34dc7c2f BB |
1012 | } |
1013 | } | |
e5c39b95 | 1014 | EXPORT_SYMBOL(zfs_unregister_callbacks); |
34dc7c2f | 1015 | |
bc3e15e3 | 1016 | #ifdef HAVE_MLSLABEL |
428870ff | 1017 | /* |
a08ee875 LG |
1018 | * Check that the hex label string is appropriate for the dataset being |
1019 | * mounted into the global_zone proper. | |
428870ff | 1020 | * |
a08ee875 LG |
1021 | * Return an error if the hex label string is not default or |
1022 | * admin_low/admin_high. For admin_low labels, the corresponding | |
1023 | * dataset must be readonly. | |
428870ff BB |
1024 | */ |
1025 | int | |
1026 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
1027 | { | |
1028 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
1029 | return (0); | |
1030 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
1031 | return (0); | |
1032 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
1033 | /* must be readonly */ | |
1034 | uint64_t rdonly; | |
1035 | ||
1036 | if (dsl_prop_get_integer(dsname, | |
1037 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
a08ee875 | 1038 | return (SET_ERROR(EACCES)); |
428870ff BB |
1039 | return (rdonly ? 0 : EACCES); |
1040 | } | |
a08ee875 | 1041 | return (SET_ERROR(EACCES)); |
428870ff | 1042 | } |
86f35f34 | 1043 | EXPORT_SYMBOL(zfs_check_global_label); |
bc3e15e3 | 1044 | #endif /* HAVE_MLSLABEL */ |
428870ff | 1045 | |
e5c39b95 | 1046 | int |
3558fd73 | 1047 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 1048 | { |
3558fd73 | 1049 | zfs_sb_t *zsb = dentry->d_sb->s_fs_info; |
34dc7c2f | 1050 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
04f9432d | 1051 | uint64_t fsid; |
3558fd73 | 1052 | uint32_t bshift; |
34dc7c2f | 1053 | |
3558fd73 | 1054 | ZFS_ENTER(zsb); |
34dc7c2f | 1055 | |
3558fd73 | 1056 | dmu_objset_space(zsb->z_os, |
34dc7c2f BB |
1057 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
1058 | ||
04f9432d | 1059 | fsid = dmu_objset_fsid_guid(zsb->z_os); |
34dc7c2f | 1060 | /* |
05ff35c6 BB |
1061 | * The underlying storage pool actually uses multiple block |
1062 | * size. Under Solaris frsize (fragment size) is reported as | |
1063 | * the smallest block size we support, and bsize (block size) | |
1064 | * as the filesystem's maximum block size. Unfortunately, | |
1065 | * under Linux the fragment size and block size are often used | |
1066 | * interchangeably. Thus we are forced to report both of them | |
1067 | * as the filesystem's maximum block size. | |
34dc7c2f | 1068 | */ |
05ff35c6 | 1069 | statp->f_frsize = zsb->z_max_blksz; |
3558fd73 BB |
1070 | statp->f_bsize = zsb->z_max_blksz; |
1071 | bshift = fls(statp->f_bsize) - 1; | |
34dc7c2f BB |
1072 | |
1073 | /* | |
3558fd73 BB |
1074 | * The following report "total" blocks of various kinds in |
1075 | * the file system, but reported in terms of f_bsize - the | |
1076 | * "preferred" size. | |
34dc7c2f BB |
1077 | */ |
1078 | ||
3558fd73 BB |
1079 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
1080 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
1081 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
1082 | ||
1083 | /* | |
1084 | * statvfs() should really be called statufs(), because it assumes | |
1085 | * static metadata. ZFS doesn't preallocate files, so the best | |
1086 | * we can do is report the max that could possibly fit in f_files, | |
1087 | * and that minus the number actually used in f_ffree. | |
1088 | * For f_ffree, report the smaller of the number of object available | |
1089 | * and the number of blocks (each object will take at least a block). | |
1090 | */ | |
baab0630 | 1091 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 1092 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
1093 | statp->f_fsid.val[0] = (uint32_t)fsid; |
1094 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 BB |
1095 | statp->f_type = ZFS_SUPER_MAGIC; |
1096 | statp->f_namelen = ZFS_MAXNAMELEN; | |
34dc7c2f BB |
1097 | |
1098 | /* | |
3558fd73 | 1099 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
1100 | * Is there anything useful we could/should provide? |
1101 | */ | |
3558fd73 | 1102 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 1103 | |
3558fd73 | 1104 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1105 | return (0); |
1106 | } | |
e5c39b95 | 1107 | EXPORT_SYMBOL(zfs_statvfs); |
34dc7c2f | 1108 | |
e5c39b95 | 1109 | int |
3558fd73 | 1110 | zfs_root(zfs_sb_t *zsb, struct inode **ipp) |
34dc7c2f | 1111 | { |
34dc7c2f BB |
1112 | znode_t *rootzp; |
1113 | int error; | |
1114 | ||
3558fd73 | 1115 | ZFS_ENTER(zsb); |
34dc7c2f | 1116 | |
3558fd73 | 1117 | error = zfs_zget(zsb, zsb->z_root, &rootzp); |
34dc7c2f | 1118 | if (error == 0) |
3558fd73 | 1119 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1120 | |
3558fd73 | 1121 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1122 | return (error); |
1123 | } | |
e5c39b95 | 1124 | EXPORT_SYMBOL(zfs_root); |
34dc7c2f | 1125 | |
87dac73d | 1126 | #ifdef HAVE_D_PRUNE_ALIASES |
ea04106b AX |
1127 | /* |
1128 | * Linux kernels older than 3.1 do not support a per-filesystem shrinker. | |
1129 | * To accommodate this we must improvise and manually walk the list of znodes | |
1130 | * attempting to prune dentries in order to be able to drop the inodes. | |
1131 | * | |
1132 | * To avoid scanning the same znodes multiple times they are always rotated | |
1133 | * to the end of the z_all_znodes list. New znodes are inserted at the | |
1134 | * end of the list so we're always scanning the oldest znodes first. | |
1135 | */ | |
1136 | static int | |
1137 | zfs_sb_prune_aliases(zfs_sb_t *zsb, unsigned long nr_to_scan) | |
1138 | { | |
1139 | znode_t **zp_array, *zp; | |
1140 | int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *)); | |
1141 | int objects = 0; | |
1142 | int i = 0, j = 0; | |
1143 | ||
1144 | zp_array = kmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP); | |
1145 | ||
1146 | mutex_enter(&zsb->z_znodes_lock); | |
1147 | while ((zp = list_head(&zsb->z_all_znodes)) != NULL) { | |
1148 | ||
1149 | if ((i++ > nr_to_scan) || (j >= max_array)) | |
1150 | break; | |
1151 | ||
1152 | ASSERT(list_link_active(&zp->z_link_node)); | |
1153 | list_remove(&zsb->z_all_znodes, zp); | |
1154 | list_insert_tail(&zsb->z_all_znodes, zp); | |
1155 | ||
1156 | /* Skip active znodes and .zfs entries */ | |
1157 | if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir) | |
1158 | continue; | |
1159 | ||
1160 | if (igrab(ZTOI(zp)) == NULL) | |
1161 | continue; | |
1162 | ||
1163 | zp_array[j] = zp; | |
1164 | j++; | |
1165 | } | |
1166 | mutex_exit(&zsb->z_znodes_lock); | |
1167 | ||
1168 | for (i = 0; i < j; i++) { | |
1169 | zp = zp_array[i]; | |
1170 | ||
1171 | ASSERT3P(zp, !=, NULL); | |
1172 | d_prune_aliases(ZTOI(zp)); | |
1173 | ||
1174 | if (atomic_read(&ZTOI(zp)->i_count) == 1) | |
1175 | objects++; | |
1176 | ||
1177 | iput(ZTOI(zp)); | |
1178 | } | |
1179 | ||
1180 | kmem_free(zp_array, max_array * sizeof (znode_t *)); | |
1181 | ||
1182 | return (objects); | |
1183 | } | |
1184 | #endif /* HAVE_D_PRUNE_ALIASES */ | |
1185 | ||
1186 | /* | |
1187 | * The ARC has requested that the filesystem drop entries from the dentry | |
1188 | * and inode caches. This can occur when the ARC needs to free meta data | |
1189 | * blocks but can't because they are all pinned by entries in these caches. | |
1190 | */ | |
ab26409d BB |
1191 | int |
1192 | zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) | |
1193 | { | |
1194 | zfs_sb_t *zsb = sb->s_fs_info; | |
ea04106b AX |
1195 | int error = 0; |
1196 | #if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK) | |
ab26409d BB |
1197 | struct shrinker *shrinker = &sb->s_shrink; |
1198 | struct shrink_control sc = { | |
1199 | .nr_to_scan = nr_to_scan, | |
1200 | .gfp_mask = GFP_KERNEL, | |
1201 | }; | |
ea04106b | 1202 | #endif |
ab26409d BB |
1203 | |
1204 | ZFS_ENTER(zsb); | |
ea04106b AX |
1205 | |
1206 | #if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \ | |
1207 | defined(SHRINK_CONTROL_HAS_NID) && \ | |
1208 | defined(SHRINKER_NUMA_AWARE) | |
1209 | if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) { | |
1210 | *objects = 0; | |
1211 | for_each_online_node(sc.nid) | |
1212 | *objects += (*shrinker->scan_objects)(shrinker, &sc); | |
1213 | } else { | |
1214 | *objects = (*shrinker->scan_objects)(shrinker, &sc); | |
1215 | } | |
87dac73d | 1216 | |
ea04106b AX |
1217 | #elif defined(HAVE_SPLIT_SHRINKER_CALLBACK) |
1218 | *objects = (*shrinker->scan_objects)(shrinker, &sc); | |
1219 | #elif defined(HAVE_SHRINK) | |
ab26409d | 1220 | *objects = (*shrinker->shrink)(shrinker, &sc); |
ea04106b | 1221 | #elif defined(HAVE_D_PRUNE_ALIASES) |
87dac73d | 1222 | #define D_PRUNE_ALIASES_IS_DEFAULT |
ea04106b AX |
1223 | *objects = zfs_sb_prune_aliases(zsb, nr_to_scan); |
1224 | #else | |
1225 | #error "No available dentry and inode cache pruning mechanism." | |
1226 | #endif | |
87dac73d AX |
1227 | |
1228 | #if defined(HAVE_D_PRUNE_ALIASES) && !defined(D_PRUNE_ALIASES_IS_DEFAULT) | |
1229 | #undef D_PRUNE_ALIASES_IS_DEFAULT | |
1230 | /* | |
1231 | * Fall back to zfs_sb_prune_aliases if the kernel's per-superblock | |
1232 | * shrinker couldn't free anything, possibly due to the inodes being | |
1233 | * allocated in a different memcg. | |
1234 | */ | |
1235 | if (*objects == 0) | |
1236 | *objects = zfs_sb_prune_aliases(zsb, nr_to_scan); | |
1237 | #endif | |
1238 | ||
ab26409d BB |
1239 | ZFS_EXIT(zsb); |
1240 | ||
ea04106b AX |
1241 | dprintf_ds(zsb->z_os->os_dsl_dataset, |
1242 | "pruning, nr_to_scan=%lu objects=%d error=%d\n", | |
1243 | nr_to_scan, *objects, error); | |
1244 | ||
1245 | return (error); | |
ab26409d BB |
1246 | } |
1247 | EXPORT_SYMBOL(zfs_sb_prune); | |
ab26409d | 1248 | |
34dc7c2f | 1249 | /* |
7b3e34ba | 1250 | * Teardown the zfs_sb_t. |
34dc7c2f BB |
1251 | * |
1252 | * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' | |
1253 | * and 'z_teardown_inactive_lock' held. | |
1254 | */ | |
3558fd73 | 1255 | int |
86f35f34 | 1256 | zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting) |
34dc7c2f BB |
1257 | { |
1258 | znode_t *zp; | |
1259 | ||
a08ee875 LG |
1260 | /* |
1261 | * If someone has not already unmounted this file system, | |
1262 | * drain the iput_taskq to ensure all active references to the | |
1263 | * zfs_sb_t have been handled only then can it be safely destroyed. | |
1264 | */ | |
ea04106b AX |
1265 | if (zsb->z_os) { |
1266 | /* | |
1267 | * If we're unmounting we have to wait for the list to | |
1268 | * drain completely. | |
1269 | * | |
1270 | * If we're not unmounting there's no guarantee the list | |
1271 | * will drain completely, but iputs run from the taskq | |
1272 | * may add the parents of dir-based xattrs to the taskq | |
1273 | * so we want to wait for these. | |
1274 | * | |
1275 | * We can safely read z_nr_znodes without locking because the | |
1276 | * VFS has already blocked operations which add to the | |
1277 | * z_all_znodes list and thus increment z_nr_znodes. | |
1278 | */ | |
1279 | int round = 0; | |
1280 | while (zsb->z_nr_znodes > 0) { | |
e10b0808 AX |
1281 | taskq_wait_outstanding(dsl_pool_iput_taskq( |
1282 | dmu_objset_pool(zsb->z_os)), 0); | |
ea04106b AX |
1283 | if (++round > 1 && !unmounting) |
1284 | break; | |
1285 | } | |
1286 | } | |
a08ee875 | 1287 | |
e10b0808 | 1288 | rrm_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1289 | |
1290 | if (!unmounting) { | |
1291 | /* | |
ceb43b93 BB |
1292 | * We purge the parent filesystem's super block as the |
1293 | * parent filesystem and all of its snapshots have their | |
1294 | * inode's super block set to the parent's filesystem's | |
1295 | * super block. Note, 'z_parent' is self referential | |
1296 | * for non-snapshots. | |
34dc7c2f | 1297 | */ |
ceb43b93 | 1298 | shrink_dcache_sb(zsb->z_parent->z_sb); |
34dc7c2f BB |
1299 | } |
1300 | ||
1301 | /* | |
1302 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1303 | * threads are blocked as zil_close can call zfs_inactive. | |
1304 | */ | |
3558fd73 BB |
1305 | if (zsb->z_log) { |
1306 | zil_close(zsb->z_log); | |
1307 | zsb->z_log = NULL; | |
34dc7c2f BB |
1308 | } |
1309 | ||
3558fd73 | 1310 | rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1311 | |
1312 | /* | |
1313 | * If we are not unmounting (ie: online recv) and someone already | |
1314 | * unmounted this file system while we were doing the switcheroo, | |
1315 | * or a reopen of z_os failed then just bail out now. | |
1316 | */ | |
3558fd73 BB |
1317 | if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { |
1318 | rw_exit(&zsb->z_teardown_inactive_lock); | |
e10b0808 | 1319 | rrm_exit(&zsb->z_teardown_lock, FTAG); |
a08ee875 | 1320 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1321 | } |
1322 | ||
1323 | /* | |
7b3e34ba BB |
1324 | * At this point there are no VFS ops active, and any new VFS ops |
1325 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1326 | * relevant for forced unmount). | |
34dc7c2f BB |
1327 | * |
1328 | * Release all holds on dbufs. | |
1329 | */ | |
ea04106b AX |
1330 | if (!unmounting) { |
1331 | mutex_enter(&zsb->z_znodes_lock); | |
1332 | for (zp = list_head(&zsb->z_all_znodes); zp != NULL; | |
1333 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
1334 | if (zp->z_sa_hdl) | |
1335 | zfs_znode_dmu_fini(zp); | |
1336 | } | |
1337 | mutex_exit(&zsb->z_znodes_lock); | |
7b3e34ba | 1338 | } |
34dc7c2f BB |
1339 | |
1340 | /* | |
7b3e34ba | 1341 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1342 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1343 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1344 | */ |
1345 | if (unmounting) { | |
3558fd73 | 1346 | zsb->z_unmounted = B_TRUE; |
e10b0808 | 1347 | rrm_exit(&zsb->z_teardown_lock, FTAG); |
3558fd73 | 1348 | rw_exit(&zsb->z_teardown_inactive_lock); |
34dc7c2f BB |
1349 | } |
1350 | ||
1351 | /* | |
1352 | * z_os will be NULL if there was an error in attempting to reopen | |
3558fd73 BB |
1353 | * zsb, so just return as the properties had already been |
1354 | * | |
34dc7c2f BB |
1355 | * unregistered and cached data had been evicted before. |
1356 | */ | |
3558fd73 | 1357 | if (zsb->z_os == NULL) |
34dc7c2f BB |
1358 | return (0); |
1359 | ||
1360 | /* | |
1361 | * Unregister properties. | |
1362 | */ | |
3558fd73 | 1363 | zfs_unregister_callbacks(zsb); |
34dc7c2f BB |
1364 | |
1365 | /* | |
1366 | * Evict cached data | |
1367 | */ | |
04434775 MA |
1368 | if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) && |
1369 | !zfs_is_readonly(zsb)) | |
1370 | txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); | |
a08ee875 | 1371 | dmu_objset_evict_dbufs(zsb->z_os); |
34dc7c2f BB |
1372 | |
1373 | return (0); | |
1374 | } | |
86f35f34 | 1375 | EXPORT_SYMBOL(zfs_sb_teardown); |
34dc7c2f | 1376 | |
ea04106b AX |
1377 | #if !defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER) && \ |
1378 | !defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER) | |
5547c2f1 | 1379 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); |
ea04106b | 1380 | #endif |
76659dc1 | 1381 | |
e5c39b95 | 1382 | int |
e10b0808 | 1383 | zfs_domount(struct super_block *sb, zfs_mntopts_t *zmo, int silent) |
34dc7c2f | 1384 | { |
e10b0808 | 1385 | const char *osname = zmo->z_osname; |
3558fd73 BB |
1386 | zfs_sb_t *zsb; |
1387 | struct inode *root_inode; | |
1388 | uint64_t recordsize; | |
1389 | int error; | |
34dc7c2f | 1390 | |
e10b0808 | 1391 | error = zfs_sb_create(osname, zmo, &zsb); |
3558fd73 BB |
1392 | if (error) |
1393 | return (error); | |
1394 | ||
1395 | if ((error = dsl_prop_get_integer(osname, "recordsize", | |
1396 | &recordsize, NULL))) | |
1397 | goto out; | |
1398 | ||
1399 | zsb->z_sb = sb; | |
3558fd73 BB |
1400 | sb->s_fs_info = zsb; |
1401 | sb->s_magic = ZFS_SUPER_MAGIC; | |
1402 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1403 | sb->s_time_gran = 1; | |
1404 | sb->s_blocksize = recordsize; | |
1405 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 BB |
1406 | zsb->z_bdi.ra_pages = 0; |
1407 | sb->s_bdi = &zsb->z_bdi; | |
1408 | ||
22929307 | 1409 | error = -zpl_bdi_setup(sb, "zfs"); |
5547c2f1 BB |
1410 | if (error) |
1411 | goto out; | |
3558fd73 | 1412 | |
22929307 AX |
1413 | sb->s_bdi->ra_pages = 0; |
1414 | ||
3558fd73 BB |
1415 | /* Set callback operations for the file system. */ |
1416 | sb->s_op = &zpl_super_operations; | |
1417 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1418 | sb->s_export_op = &zpl_export_operations; |
ee930353 BB |
1419 | #ifdef HAVE_S_D_OP |
1420 | sb->s_d_op = &zpl_dentry_operations; | |
1421 | #endif /* HAVE_S_D_OP */ | |
3558fd73 BB |
1422 | |
1423 | /* Set features for file system. */ | |
1424 | zfs_set_fuid_feature(zsb); | |
1425 | ||
1426 | if (dmu_objset_is_snapshot(zsb->z_os)) { | |
1427 | uint64_t pval; | |
1428 | ||
1429 | atime_changed_cb(zsb, B_FALSE); | |
1430 | readonly_changed_cb(zsb, B_TRUE); | |
a08ee875 LG |
1431 | if ((error = dsl_prop_get_integer(osname, |
1432 | "xattr", &pval, NULL))) | |
3558fd73 BB |
1433 | goto out; |
1434 | xattr_changed_cb(zsb, pval); | |
a08ee875 LG |
1435 | if ((error = dsl_prop_get_integer(osname, |
1436 | "acltype", &pval, NULL))) | |
1437 | goto out; | |
1438 | acltype_changed_cb(zsb, pval); | |
3558fd73 BB |
1439 | zsb->z_issnap = B_TRUE; |
1440 | zsb->z_os->os_sync = ZFS_SYNC_DISABLED; | |
e10b0808 | 1441 | zsb->z_snap_defer_time = jiffies; |
3558fd73 BB |
1442 | |
1443 | mutex_enter(&zsb->z_os->os_user_ptr_lock); | |
1444 | dmu_objset_set_user(zsb->z_os, zsb); | |
1445 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
1446 | } else { | |
1447 | error = zfs_sb_setup(zsb, B_TRUE); | |
34dc7c2f BB |
1448 | } |
1449 | ||
3558fd73 BB |
1450 | /* Allocate a root inode for the filesystem. */ |
1451 | error = zfs_root(zsb, &root_inode); | |
1452 | if (error) { | |
1453 | (void) zfs_umount(sb); | |
1454 | goto out; | |
34dc7c2f BB |
1455 | } |
1456 | ||
3558fd73 | 1457 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1458 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1459 | if (sb->s_root == NULL) { |
1460 | (void) zfs_umount(sb); | |
a08ee875 | 1461 | error = SET_ERROR(ENOMEM); |
3558fd73 BB |
1462 | goto out; |
1463 | } | |
ebe7e575 BB |
1464 | |
1465 | if (!zsb->z_issnap) | |
1466 | zfsctl_create(zsb); | |
ea04106b AX |
1467 | |
1468 | zsb->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); | |
3558fd73 BB |
1469 | out: |
1470 | if (error) { | |
1471 | dmu_objset_disown(zsb->z_os, zsb); | |
1472 | zfs_sb_free(zsb); | |
1473 | } | |
34dc7c2f | 1474 | |
3558fd73 BB |
1475 | return (error); |
1476 | } | |
1477 | EXPORT_SYMBOL(zfs_domount); | |
1478 | ||
ebe7e575 BB |
1479 | /* |
1480 | * Called when an unmount is requested and certain sanity checks have | |
1481 | * already passed. At this point no dentries or inodes have been reclaimed | |
1482 | * from their respective caches. We drop the extra reference on the .zfs | |
1483 | * control directory to allow everything to be reclaimed. All snapshots | |
1484 | * must already have been unmounted to reach this point. | |
1485 | */ | |
1486 | void | |
1487 | zfs_preumount(struct super_block *sb) | |
1488 | { | |
1489 | zfs_sb_t *zsb = sb->s_fs_info; | |
1490 | ||
e10b0808 AX |
1491 | if (zsb) |
1492 | zfsctl_destroy(sb->s_fs_info); | |
ebe7e575 BB |
1493 | } |
1494 | EXPORT_SYMBOL(zfs_preumount); | |
1495 | ||
1496 | /* | |
1497 | * Called once all other unmount released tear down has occurred. | |
1498 | * It is our responsibility to release any remaining infrastructure. | |
1499 | */ | |
3558fd73 BB |
1500 | /*ARGSUSED*/ |
1501 | int | |
1502 | zfs_umount(struct super_block *sb) | |
1503 | { | |
1504 | zfs_sb_t *zsb = sb->s_fs_info; | |
1505 | objset_t *os; | |
1506 | ||
ea04106b | 1507 | arc_remove_prune_callback(zsb->z_arc_prune); |
86f35f34 | 1508 | VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0); |
3558fd73 | 1509 | os = zsb->z_os; |
22929307 | 1510 | zpl_bdi_destroy(sb); |
76659dc1 | 1511 | |
34dc7c2f BB |
1512 | /* |
1513 | * z_os will be NULL if there was an error in | |
3558fd73 | 1514 | * attempting to reopen zsb. |
34dc7c2f BB |
1515 | */ |
1516 | if (os != NULL) { | |
1517 | /* | |
1518 | * Unset the objset user_ptr. | |
1519 | */ | |
428870ff | 1520 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1521 | dmu_objset_set_user(os, NULL); |
428870ff | 1522 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1523 | |
1524 | /* | |
b128c09f | 1525 | * Finally release the objset |
34dc7c2f | 1526 | */ |
3558fd73 | 1527 | dmu_objset_disown(os, zsb); |
34dc7c2f BB |
1528 | } |
1529 | ||
3558fd73 | 1530 | zfs_sb_free(zsb); |
34dc7c2f BB |
1531 | return (0); |
1532 | } | |
e5c39b95 | 1533 | EXPORT_SYMBOL(zfs_umount); |
34dc7c2f | 1534 | |
0de19dad | 1535 | int |
e10b0808 | 1536 | zfs_remount(struct super_block *sb, int *flags, zfs_mntopts_t *zmo) |
0de19dad | 1537 | { |
e10b0808 AX |
1538 | zfs_sb_t *zsb = sb->s_fs_info; |
1539 | int error; | |
1540 | ||
1541 | zfs_unregister_callbacks(zsb); | |
1542 | error = zfs_register_callbacks(zsb); | |
1543 | ||
1544 | return (error); | |
0de19dad BB |
1545 | } |
1546 | EXPORT_SYMBOL(zfs_remount); | |
1547 | ||
e5c39b95 | 1548 | int |
2cf7f52b | 1549 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1550 | { |
2cf7f52b | 1551 | zfs_sb_t *zsb = sb->s_fs_info; |
34dc7c2f BB |
1552 | znode_t *zp; |
1553 | uint64_t object = 0; | |
1554 | uint64_t fid_gen = 0; | |
1555 | uint64_t gen_mask; | |
1556 | uint64_t zp_gen; | |
3558fd73 | 1557 | int i, err; |
34dc7c2f | 1558 | |
3558fd73 | 1559 | *ipp = NULL; |
34dc7c2f | 1560 | |
3558fd73 | 1561 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1562 | |
1563 | if (fidp->fid_len == LONG_FID_LEN) { | |
1564 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1565 | uint64_t objsetid = 0; | |
1566 | uint64_t setgen = 0; | |
1567 | ||
1568 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1569 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1570 | ||
1571 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1572 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1573 | ||
3558fd73 | 1574 | ZFS_EXIT(zsb); |
34dc7c2f | 1575 | |
ebe7e575 | 1576 | err = zfsctl_lookup_objset(sb, objsetid, &zsb); |
34dc7c2f | 1577 | if (err) |
a08ee875 | 1578 | return (SET_ERROR(EINVAL)); |
ebe7e575 | 1579 | |
3558fd73 | 1580 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1581 | } |
1582 | ||
1583 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { | |
1584 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1585 | ||
1586 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1587 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
1588 | ||
1589 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
1590 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1591 | } else { | |
3558fd73 | 1592 | ZFS_EXIT(zsb); |
a08ee875 | 1593 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1594 | } |
1595 | ||
1596 | /* A zero fid_gen means we are in the .zfs control directories */ | |
1597 | if (fid_gen == 0 && | |
1598 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
3558fd73 BB |
1599 | *ipp = zsb->z_ctldir; |
1600 | ASSERT(*ipp != NULL); | |
34dc7c2f | 1601 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1602 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1603 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1604 | } else { |
3558fd73 | 1605 | igrab(*ipp); |
34dc7c2f | 1606 | } |
3558fd73 | 1607 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1608 | return (0); |
1609 | } | |
1610 | ||
1611 | gen_mask = -1ULL >> (64 - 8 * i); | |
1612 | ||
ea04106b | 1613 | dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); |
3558fd73 BB |
1614 | if ((err = zfs_zget(zsb, object, &zp))) { |
1615 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1616 | return (err); |
1617 | } | |
87dac73d AX |
1618 | |
1619 | /* Don't export xattr stuff */ | |
1620 | if (zp->z_pflags & ZFS_XATTR) { | |
1621 | iput(ZTOI(zp)); | |
1622 | ZFS_EXIT(zsb); | |
1623 | return (SET_ERROR(ENOENT)); | |
1624 | } | |
1625 | ||
3558fd73 | 1626 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, |
428870ff BB |
1627 | sizeof (uint64_t)); |
1628 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1629 | if (zp_gen == 0) |
1630 | zp_gen = 1; | |
e10b0808 AX |
1631 | if ((fid_gen == 0) && (zsb->z_root == object)) |
1632 | fid_gen = zp_gen; | |
34dc7c2f | 1633 | if (zp->z_unlinked || zp_gen != fid_gen) { |
ea04106b AX |
1634 | dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, |
1635 | fid_gen); | |
3558fd73 BB |
1636 | iput(ZTOI(zp)); |
1637 | ZFS_EXIT(zsb); | |
87dac73d | 1638 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1639 | } |
1640 | ||
3558fd73 BB |
1641 | *ipp = ZTOI(zp); |
1642 | if (*ipp) | |
1643 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1644 | |
3558fd73 | 1645 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1646 | return (0); |
1647 | } | |
e5c39b95 | 1648 | EXPORT_SYMBOL(zfs_vget); |
34dc7c2f BB |
1649 | |
1650 | /* | |
7b3e34ba | 1651 | * Block out VFS ops and close zfs_sb_t |
34dc7c2f BB |
1652 | * |
1653 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
a08ee875 LG |
1654 | * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying |
1655 | * dataset and objset intact so that they can be atomically handed off during | |
1656 | * a subsequent rollback or recv operation and the resume thereafter. | |
34dc7c2f BB |
1657 | */ |
1658 | int | |
3558fd73 | 1659 | zfs_suspend_fs(zfs_sb_t *zsb) |
34dc7c2f BB |
1660 | { |
1661 | int error; | |
1662 | ||
86f35f34 | 1663 | if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0) |
34dc7c2f | 1664 | return (error); |
7b3e34ba | 1665 | |
34dc7c2f BB |
1666 | return (0); |
1667 | } | |
e5c39b95 | 1668 | EXPORT_SYMBOL(zfs_suspend_fs); |
34dc7c2f BB |
1669 | |
1670 | /* | |
7b3e34ba | 1671 | * Reopen zfs_sb_t and release VFS ops. |
34dc7c2f BB |
1672 | */ |
1673 | int | |
3558fd73 | 1674 | zfs_resume_fs(zfs_sb_t *zsb, const char *osname) |
34dc7c2f | 1675 | { |
428870ff | 1676 | int err, err2; |
a08ee875 LG |
1677 | znode_t *zp; |
1678 | uint64_t sa_obj = 0; | |
34dc7c2f | 1679 | |
e10b0808 | 1680 | ASSERT(RRM_WRITE_HELD(&zsb->z_teardown_lock)); |
3558fd73 | 1681 | ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); |
34dc7c2f | 1682 | |
a08ee875 LG |
1683 | /* |
1684 | * We already own this, so just hold and rele it to update the | |
1685 | * objset_t, as the one we had before may have been evicted. | |
1686 | */ | |
1687 | VERIFY0(dmu_objset_hold(osname, zsb, &zsb->z_os)); | |
1688 | VERIFY3P(zsb->z_os->os_dsl_dataset->ds_owner, ==, zsb); | |
1689 | VERIFY(dsl_dataset_long_held(zsb->z_os->os_dsl_dataset)); | |
1690 | dmu_objset_rele(zsb->z_os, zsb); | |
428870ff | 1691 | |
a08ee875 LG |
1692 | /* |
1693 | * Make sure version hasn't changed | |
1694 | */ | |
428870ff | 1695 | |
a08ee875 LG |
1696 | err = zfs_get_zplprop(zsb->z_os, ZFS_PROP_VERSION, |
1697 | &zsb->z_version); | |
428870ff | 1698 | |
a08ee875 LG |
1699 | if (err) |
1700 | goto bail; | |
428870ff | 1701 | |
a08ee875 LG |
1702 | err = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, |
1703 | ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
34dc7c2f | 1704 | |
a08ee875 LG |
1705 | if (err && zsb->z_version >= ZPL_VERSION_SA) |
1706 | goto bail; | |
34dc7c2f | 1707 | |
a08ee875 LG |
1708 | if ((err = sa_setup(zsb->z_os, sa_obj, |
1709 | zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) | |
1710 | goto bail; | |
1711 | ||
1712 | if (zsb->z_version >= ZPL_VERSION_SA) | |
1713 | sa_register_update_callback(zsb->z_os, | |
1714 | zfs_sa_upgrade); | |
1715 | ||
1716 | VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); | |
1717 | ||
1718 | zfs_set_fuid_feature(zsb); | |
1719 | zsb->z_rollback_time = jiffies; | |
1720 | ||
1721 | /* | |
1722 | * Attempt to re-establish all the active inodes with their | |
1723 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
1724 | * and mark it stale. This prevents a collision if a new | |
1725 | * inode/object is created which must use the same inode | |
1726 | * number. The stale inode will be be released when the | |
1727 | * VFS prunes the dentry holding the remaining references | |
1728 | * on the stale inode. | |
1729 | */ | |
1730 | mutex_enter(&zsb->z_znodes_lock); | |
1731 | for (zp = list_head(&zsb->z_all_znodes); zp; | |
1732 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
1733 | err2 = zfs_rezget(zp); | |
1734 | if (err2) { | |
1735 | remove_inode_hash(ZTOI(zp)); | |
1736 | zp->z_is_stale = B_TRUE; | |
34dc7c2f | 1737 | } |
34dc7c2f | 1738 | } |
a08ee875 | 1739 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f | 1740 | |
428870ff | 1741 | bail: |
7b3e34ba | 1742 | /* release the VFS ops */ |
3558fd73 | 1743 | rw_exit(&zsb->z_teardown_inactive_lock); |
e10b0808 | 1744 | rrm_exit(&zsb->z_teardown_lock, FTAG); |
34dc7c2f BB |
1745 | |
1746 | if (err) { | |
1747 | /* | |
a08ee875 LG |
1748 | * Since we couldn't setup the sa framework, try to force |
1749 | * unmount this file system. | |
34dc7c2f | 1750 | */ |
c06d4368 AX |
1751 | if (zsb->z_os) |
1752 | (void) zfs_umount(zsb->z_sb); | |
34dc7c2f BB |
1753 | } |
1754 | return (err); | |
1755 | } | |
e5c39b95 | 1756 | EXPORT_SYMBOL(zfs_resume_fs); |
34dc7c2f | 1757 | |
34dc7c2f | 1758 | int |
3558fd73 | 1759 | zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) |
34dc7c2f BB |
1760 | { |
1761 | int error; | |
3558fd73 | 1762 | objset_t *os = zsb->z_os; |
34dc7c2f | 1763 | dmu_tx_t *tx; |
34dc7c2f BB |
1764 | |
1765 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
a08ee875 | 1766 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1767 | |
3558fd73 | 1768 | if (newvers < zsb->z_version) |
a08ee875 | 1769 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1770 | |
428870ff | 1771 | if (zfs_spa_version_map(newvers) > |
3558fd73 | 1772 | spa_version(dmu_objset_spa(zsb->z_os))) |
a08ee875 | 1773 | return (SET_ERROR(ENOTSUP)); |
428870ff | 1774 | |
34dc7c2f | 1775 | tx = dmu_tx_create(os); |
9babb374 | 1776 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
3558fd73 | 1777 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1778 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1779 | ZFS_SA_ATTRS); | |
1780 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
1781 | } | |
34dc7c2f BB |
1782 | error = dmu_tx_assign(tx, TXG_WAIT); |
1783 | if (error) { | |
1784 | dmu_tx_abort(tx); | |
9babb374 BB |
1785 | return (error); |
1786 | } | |
428870ff | 1787 | |
9babb374 BB |
1788 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
1789 | 8, 1, &newvers, tx); | |
1790 | ||
1791 | if (error) { | |
1792 | dmu_tx_commit(tx); | |
1793 | return (error); | |
34dc7c2f | 1794 | } |
34dc7c2f | 1795 | |
3558fd73 | 1796 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1797 | uint64_t sa_obj; |
1798 | ||
3558fd73 | 1799 | ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, |
428870ff BB |
1800 | SPA_VERSION_SA); |
1801 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1802 | DMU_OT_NONE, 0, tx); | |
1803 | ||
1804 | error = zap_add(os, MASTER_NODE_OBJ, | |
1805 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c06d4368 | 1806 | ASSERT0(error); |
428870ff BB |
1807 | |
1808 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
1809 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1810 | } | |
1811 | ||
a08ee875 LG |
1812 | spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, |
1813 | "from %llu to %llu", zsb->z_version, newvers); | |
9babb374 | 1814 | |
34dc7c2f BB |
1815 | dmu_tx_commit(tx); |
1816 | ||
3558fd73 | 1817 | zsb->z_version = newvers; |
9babb374 | 1818 | |
a08ee875 | 1819 | zfs_set_fuid_feature(zsb); |
9babb374 BB |
1820 | |
1821 | return (0); | |
34dc7c2f | 1822 | } |
e5c39b95 | 1823 | EXPORT_SYMBOL(zfs_set_version); |
34dc7c2f BB |
1824 | |
1825 | /* | |
1826 | * Read a property stored within the master node. | |
1827 | */ | |
1828 | int | |
1829 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
1830 | { | |
1831 | const char *pname; | |
a08ee875 | 1832 | int error = SET_ERROR(ENOENT); |
34dc7c2f BB |
1833 | |
1834 | /* | |
1835 | * Look up the file system's value for the property. For the | |
1836 | * version property, we look up a slightly different string. | |
1837 | */ | |
1838 | if (prop == ZFS_PROP_VERSION) | |
1839 | pname = ZPL_VERSION_STR; | |
1840 | else | |
1841 | pname = zfs_prop_to_name(prop); | |
1842 | ||
b128c09f BB |
1843 | if (os != NULL) |
1844 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
1845 | |
1846 | if (error == ENOENT) { | |
1847 | /* No value set, use the default value */ | |
1848 | switch (prop) { | |
1849 | case ZFS_PROP_VERSION: | |
1850 | *value = ZPL_VERSION; | |
1851 | break; | |
1852 | case ZFS_PROP_NORMALIZE: | |
1853 | case ZFS_PROP_UTF8ONLY: | |
1854 | *value = 0; | |
1855 | break; | |
1856 | case ZFS_PROP_CASE: | |
1857 | *value = ZFS_CASE_SENSITIVE; | |
1858 | break; | |
a08ee875 LG |
1859 | case ZFS_PROP_ACLTYPE: |
1860 | *value = ZFS_ACLTYPE_OFF; | |
1861 | break; | |
34dc7c2f BB |
1862 | default: |
1863 | return (error); | |
1864 | } | |
1865 | error = 0; | |
1866 | } | |
1867 | return (error); | |
1868 | } | |
86f35f34 | 1869 | EXPORT_SYMBOL(zfs_get_zplprop); |
3558fd73 BB |
1870 | |
1871 | void | |
1872 | zfs_init(void) | |
1873 | { | |
ebe7e575 | 1874 | zfsctl_init(); |
3558fd73 BB |
1875 | zfs_znode_init(); |
1876 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
1877 | register_filesystem(&zpl_fs_type); | |
1878 | } | |
1879 | ||
1880 | void | |
1881 | zfs_fini(void) | |
1882 | { | |
68d83c55 AX |
1883 | /* |
1884 | * we don't use outstanding because zpl_posix_acl_free might add more. | |
1885 | */ | |
1886 | taskq_wait(system_taskq); | |
3558fd73 BB |
1887 | unregister_filesystem(&zpl_fs_type); |
1888 | zfs_znode_fini(); | |
ebe7e575 | 1889 | zfsctl_fini(); |
3558fd73 | 1890 | } |