]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
21 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
2e528b49 | 23 | * Copyright (c) 2013 by Delphix. All rights reserved. |
34dc7c2f BB |
24 | */ |
25 | ||
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 | |
9babb374 | 71 | |
34dc7c2f BB |
72 | /*ARGSUSED*/ |
73 | int | |
03f9ba9d | 74 | zfs_sync(struct super_block *sb, int wait, cred_t *cr) |
34dc7c2f | 75 | { |
03f9ba9d BB |
76 | zfs_sb_t *zsb = sb->s_fs_info; |
77 | ||
34dc7c2f BB |
78 | /* |
79 | * Data integrity is job one. We don't want a compromised kernel | |
80 | * writing to the storage pool, so we never sync during panic. | |
81 | */ | |
d5e53f9d | 82 | if (unlikely(oops_in_progress)) |
34dc7c2f BB |
83 | return (0); |
84 | ||
03f9ba9d BB |
85 | /* |
86 | * Semantically, the only requirement is that the sync be initiated. | |
87 | * The DMU syncs out txgs frequently, so there's nothing to do. | |
88 | */ | |
89 | if (!wait) | |
90 | return (0); | |
91 | ||
3558fd73 | 92 | if (zsb != NULL) { |
34dc7c2f BB |
93 | /* |
94 | * Sync a specific filesystem. | |
95 | */ | |
9babb374 | 96 | dsl_pool_t *dp; |
34dc7c2f | 97 | |
3558fd73 BB |
98 | ZFS_ENTER(zsb); |
99 | dp = dmu_objset_pool(zsb->z_os); | |
9babb374 BB |
100 | |
101 | /* | |
102 | * If the system is shutting down, then skip any | |
103 | * filesystems which may exist on a suspended pool. | |
104 | */ | |
03f9ba9d | 105 | if (spa_suspended(dp->dp_spa)) { |
3558fd73 | 106 | ZFS_EXIT(zsb); |
9babb374 BB |
107 | return (0); |
108 | } | |
109 | ||
3558fd73 BB |
110 | if (zsb->z_log != NULL) |
111 | zil_commit(zsb->z_log, 0); | |
428870ff | 112 | |
3558fd73 | 113 | ZFS_EXIT(zsb); |
34dc7c2f BB |
114 | } else { |
115 | /* | |
116 | * Sync all ZFS filesystems. This is what happens when you | |
117 | * run sync(1M). Unlike other filesystems, ZFS honors the | |
118 | * request by waiting for all pools to commit all dirty data. | |
119 | */ | |
120 | spa_sync_allpools(); | |
121 | } | |
122 | ||
123 | return (0); | |
124 | } | |
e5c39b95 | 125 | EXPORT_SYMBOL(zfs_sync); |
34dc7c2f | 126 | |
2cf7f52b BB |
127 | boolean_t |
128 | zfs_is_readonly(zfs_sb_t *zsb) | |
129 | { | |
130 | return (!!(zsb->z_sb->s_flags & MS_RDONLY)); | |
131 | } | |
132 | EXPORT_SYMBOL(zfs_is_readonly); | |
133 | ||
34dc7c2f BB |
134 | static void |
135 | atime_changed_cb(void *arg, uint64_t newval) | |
136 | { | |
2cf7f52b | 137 | ((zfs_sb_t *)arg)->z_atime = newval; |
34dc7c2f BB |
138 | } |
139 | ||
6d111134 TC |
140 | static void |
141 | relatime_changed_cb(void *arg, uint64_t newval) | |
142 | { | |
143 | ((zfs_sb_t *)arg)->z_relatime = newval; | |
144 | } | |
145 | ||
34dc7c2f BB |
146 | static void |
147 | xattr_changed_cb(void *arg, uint64_t newval) | |
148 | { | |
3558fd73 | 149 | zfs_sb_t *zsb = arg; |
34dc7c2f | 150 | |
82a37189 | 151 | if (newval == ZFS_XATTR_OFF) { |
2cf7f52b | 152 | zsb->z_flags &= ~ZSB_XATTR; |
82a37189 BB |
153 | } else { |
154 | zsb->z_flags |= ZSB_XATTR; | |
155 | ||
156 | if (newval == ZFS_XATTR_SA) | |
157 | zsb->z_xattr_sa = B_TRUE; | |
158 | else | |
159 | zsb->z_xattr_sa = B_FALSE; | |
160 | } | |
34dc7c2f BB |
161 | } |
162 | ||
023699cd MM |
163 | static void |
164 | acltype_changed_cb(void *arg, uint64_t newval) | |
165 | { | |
166 | zfs_sb_t *zsb = arg; | |
167 | ||
168 | switch (newval) { | |
169 | case ZFS_ACLTYPE_OFF: | |
170 | zsb->z_acl_type = ZFS_ACLTYPE_OFF; | |
171 | zsb->z_sb->s_flags &= ~MS_POSIXACL; | |
172 | break; | |
173 | case ZFS_ACLTYPE_POSIXACL: | |
b695c34e | 174 | #ifdef CONFIG_FS_POSIX_ACL |
023699cd MM |
175 | zsb->z_acl_type = ZFS_ACLTYPE_POSIXACL; |
176 | zsb->z_sb->s_flags |= MS_POSIXACL; | |
b695c34e MM |
177 | #else |
178 | zsb->z_acl_type = ZFS_ACLTYPE_OFF; | |
179 | zsb->z_sb->s_flags &= ~MS_POSIXACL; | |
180 | #endif /* CONFIG_FS_POSIX_ACL */ | |
023699cd MM |
181 | break; |
182 | default: | |
183 | break; | |
184 | } | |
185 | } | |
186 | ||
34dc7c2f BB |
187 | static void |
188 | blksz_changed_cb(void *arg, uint64_t newval) | |
189 | { | |
3558fd73 | 190 | zfs_sb_t *zsb = arg; |
f1512ee6 MA |
191 | ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zsb->z_os))); |
192 | ASSERT3U(newval, >=, SPA_MINBLOCKSIZE); | |
193 | ASSERT(ISP2(newval)); | |
34dc7c2f | 194 | |
3558fd73 | 195 | zsb->z_max_blksz = newval; |
34dc7c2f BB |
196 | } |
197 | ||
198 | static void | |
199 | readonly_changed_cb(void *arg, uint64_t newval) | |
200 | { | |
3558fd73 BB |
201 | zfs_sb_t *zsb = arg; |
202 | struct super_block *sb = zsb->z_sb; | |
34dc7c2f | 203 | |
2cf7f52b BB |
204 | if (sb == NULL) |
205 | return; | |
206 | ||
207 | if (newval) | |
3558fd73 | 208 | sb->s_flags |= MS_RDONLY; |
2cf7f52b | 209 | else |
3558fd73 | 210 | sb->s_flags &= ~MS_RDONLY; |
34dc7c2f BB |
211 | } |
212 | ||
213 | static void | |
214 | devices_changed_cb(void *arg, uint64_t newval) | |
215 | { | |
34dc7c2f BB |
216 | } |
217 | ||
218 | static void | |
219 | setuid_changed_cb(void *arg, uint64_t newval) | |
220 | { | |
34dc7c2f BB |
221 | } |
222 | ||
223 | static void | |
224 | exec_changed_cb(void *arg, uint64_t newval) | |
225 | { | |
34dc7c2f BB |
226 | } |
227 | ||
34dc7c2f BB |
228 | static void |
229 | nbmand_changed_cb(void *arg, uint64_t newval) | |
230 | { | |
3558fd73 BB |
231 | zfs_sb_t *zsb = arg; |
232 | struct super_block *sb = zsb->z_sb; | |
233 | ||
2cf7f52b BB |
234 | if (sb == NULL) |
235 | return; | |
236 | ||
237 | if (newval == TRUE) | |
3558fd73 | 238 | sb->s_flags |= MS_MANDLOCK; |
2cf7f52b | 239 | else |
3558fd73 | 240 | sb->s_flags &= ~MS_MANDLOCK; |
34dc7c2f BB |
241 | } |
242 | ||
243 | static void | |
244 | snapdir_changed_cb(void *arg, uint64_t newval) | |
245 | { | |
3558fd73 | 246 | ((zfs_sb_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
247 | } |
248 | ||
249 | static void | |
250 | vscan_changed_cb(void *arg, uint64_t newval) | |
251 | { | |
3558fd73 | 252 | ((zfs_sb_t *)arg)->z_vscan = newval; |
34dc7c2f BB |
253 | } |
254 | ||
34dc7c2f BB |
255 | static void |
256 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
257 | { | |
3558fd73 | 258 | ((zfs_sb_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
259 | } |
260 | ||
e5c39b95 | 261 | int |
3558fd73 | 262 | zfs_register_callbacks(zfs_sb_t *zsb) |
34dc7c2f BB |
263 | { |
264 | struct dsl_dataset *ds = NULL; | |
3558fd73 | 265 | objset_t *os = zsb->z_os; |
c9ada6d5 | 266 | boolean_t do_readonly = B_FALSE; |
34dc7c2f BB |
267 | int error = 0; |
268 | ||
2cf7f52b | 269 | if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) |
c9ada6d5 | 270 | do_readonly = B_TRUE; |
34dc7c2f BB |
271 | |
272 | /* | |
273 | * Register property callbacks. | |
274 | * | |
275 | * It would probably be fine to just check for i/o error from | |
276 | * the first prop_register(), but I guess I like to go | |
277 | * overboard... | |
278 | */ | |
279 | ds = dmu_objset_ds(os); | |
13fe0198 | 280 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
3558fd73 | 281 | error = dsl_prop_register(ds, |
13fe0198 | 282 | zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb); |
98fad862 | 283 | error = error ? error : dsl_prop_register(ds, |
6d111134 | 284 | zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb); |
34dc7c2f | 285 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 286 | zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb); |
34dc7c2f | 287 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 288 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb); |
34dc7c2f | 289 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 290 | zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb); |
34dc7c2f | 291 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 292 | zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb); |
34dc7c2f | 293 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 294 | zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb); |
34dc7c2f | 295 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 296 | zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb); |
34dc7c2f | 297 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 298 | zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb); |
34dc7c2f | 299 | error = error ? error : dsl_prop_register(ds, |
023699cd MM |
300 | zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb); |
301 | error = error ? error : dsl_prop_register(ds, | |
302 | zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb); | |
34dc7c2f | 303 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 304 | zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb); |
2cf7f52b | 305 | error = error ? error : dsl_prop_register(ds, |
13fe0198 MA |
306 | zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb); |
307 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
34dc7c2f BB |
308 | if (error) |
309 | goto unregister; | |
310 | ||
c9ada6d5 BB |
311 | if (do_readonly) |
312 | readonly_changed_cb(zsb, B_TRUE); | |
313 | ||
34dc7c2f BB |
314 | return (0); |
315 | ||
316 | unregister: | |
317 | /* | |
318 | * We may attempt to unregister some callbacks that are not | |
319 | * registered, but this is OK; it will simply return ENOMSG, | |
320 | * which we will ignore. | |
321 | */ | |
13fe0198 MA |
322 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME), |
323 | atime_changed_cb, zsb); | |
6d111134 TC |
324 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RELATIME), |
325 | relatime_changed_cb, zsb); | |
13fe0198 MA |
326 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR), |
327 | xattr_changed_cb, zsb); | |
328 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
329 | blksz_changed_cb, zsb); | |
330 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY), | |
331 | readonly_changed_cb, zsb); | |
332 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES), | |
333 | devices_changed_cb, zsb); | |
334 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID), | |
335 | setuid_changed_cb, zsb); | |
336 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC), | |
337 | exec_changed_cb, zsb); | |
338 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR), | |
339 | snapdir_changed_cb, zsb); | |
023699cd MM |
340 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE), |
341 | acltype_changed_cb, zsb); | |
13fe0198 MA |
342 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT), |
343 | acl_inherit_changed_cb, zsb); | |
344 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN), | |
345 | vscan_changed_cb, zsb); | |
346 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND), | |
347 | nbmand_changed_cb, zsb); | |
34dc7c2f | 348 | |
3558fd73 | 349 | return (error); |
34dc7c2f | 350 | } |
e5c39b95 | 351 | EXPORT_SYMBOL(zfs_register_callbacks); |
34dc7c2f | 352 | |
428870ff BB |
353 | static int |
354 | zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, | |
355 | uint64_t *userp, uint64_t *groupp) | |
9babb374 | 356 | { |
428870ff BB |
357 | /* |
358 | * Is it a valid type of object to track? | |
359 | */ | |
360 | if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) | |
2e528b49 | 361 | return (SET_ERROR(ENOENT)); |
9babb374 | 362 | |
428870ff BB |
363 | /* |
364 | * If we have a NULL data pointer | |
365 | * then assume the id's aren't changing and | |
366 | * return EEXIST to the dmu to let it know to | |
367 | * use the same ids | |
368 | */ | |
369 | if (data == NULL) | |
2e528b49 | 370 | return (SET_ERROR(EEXIST)); |
9babb374 | 371 | |
428870ff | 372 | if (bonustype == DMU_OT_ZNODE) { |
a94addd9 | 373 | znode_phys_t *znp = data; |
428870ff BB |
374 | *userp = znp->zp_uid; |
375 | *groupp = znp->zp_gid; | |
9babb374 | 376 | } else { |
428870ff | 377 | int hdrsize; |
a94addd9 MA |
378 | sa_hdr_phys_t *sap = data; |
379 | sa_hdr_phys_t sa = *sap; | |
380 | boolean_t swap = B_FALSE; | |
9babb374 | 381 | |
428870ff | 382 | ASSERT(bonustype == DMU_OT_SA); |
428870ff | 383 | |
a94addd9 | 384 | if (sa.sa_magic == 0) { |
428870ff BB |
385 | /* |
386 | * This should only happen for newly created | |
387 | * files that haven't had the znode data filled | |
388 | * in yet. | |
389 | */ | |
390 | *userp = 0; | |
391 | *groupp = 0; | |
a94addd9 MA |
392 | return (0); |
393 | } | |
394 | if (sa.sa_magic == BSWAP_32(SA_MAGIC)) { | |
395 | sa.sa_magic = SA_MAGIC; | |
396 | sa.sa_layout_info = BSWAP_16(sa.sa_layout_info); | |
397 | swap = B_TRUE; | |
398 | } else { | |
399 | VERIFY3U(sa.sa_magic, ==, SA_MAGIC); | |
400 | } | |
401 | ||
402 | hdrsize = sa_hdrsize(&sa); | |
403 | VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t)); | |
404 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
405 | SA_UID_OFFSET)); | |
406 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
407 | SA_GID_OFFSET)); | |
408 | if (swap) { | |
409 | *userp = BSWAP_64(*userp); | |
410 | *groupp = BSWAP_64(*groupp); | |
428870ff | 411 | } |
9babb374 | 412 | } |
13fe0198 | 413 | return (0); |
9babb374 BB |
414 | } |
415 | ||
416 | static void | |
3558fd73 | 417 | fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr, |
9babb374 BB |
418 | char *domainbuf, int buflen, uid_t *ridp) |
419 | { | |
9babb374 BB |
420 | uint64_t fuid; |
421 | const char *domain; | |
422 | ||
423 | fuid = strtonum(fuidstr, NULL); | |
424 | ||
3558fd73 | 425 | domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid)); |
9babb374 BB |
426 | if (domain) |
427 | (void) strlcpy(domainbuf, domain, buflen); | |
428 | else | |
429 | domainbuf[0] = '\0'; | |
430 | *ridp = FUID_RID(fuid); | |
431 | } | |
432 | ||
433 | static uint64_t | |
3558fd73 | 434 | zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type) |
9babb374 BB |
435 | { |
436 | switch (type) { | |
437 | case ZFS_PROP_USERUSED: | |
438 | return (DMU_USERUSED_OBJECT); | |
439 | case ZFS_PROP_GROUPUSED: | |
440 | return (DMU_GROUPUSED_OBJECT); | |
441 | case ZFS_PROP_USERQUOTA: | |
3558fd73 | 442 | return (zsb->z_userquota_obj); |
9babb374 | 443 | case ZFS_PROP_GROUPQUOTA: |
3558fd73 | 444 | return (zsb->z_groupquota_obj); |
149e873a | 445 | default: |
2e528b49 | 446 | return (SET_ERROR(ENOTSUP)); |
9babb374 BB |
447 | } |
448 | return (0); | |
449 | } | |
450 | ||
451 | int | |
3558fd73 | 452 | zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
453 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
454 | { | |
455 | int error; | |
456 | zap_cursor_t zc; | |
457 | zap_attribute_t za; | |
458 | zfs_useracct_t *buf = vbuf; | |
459 | uint64_t obj; | |
460 | ||
3558fd73 | 461 | if (!dmu_objset_userspace_present(zsb->z_os)) |
2e528b49 | 462 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 463 | |
3558fd73 | 464 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
465 | if (obj == 0) { |
466 | *bufsizep = 0; | |
467 | return (0); | |
468 | } | |
469 | ||
3558fd73 | 470 | for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep); |
9babb374 BB |
471 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
472 | zap_cursor_advance(&zc)) { | |
473 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
474 | *bufsizep) | |
475 | break; | |
476 | ||
3558fd73 | 477 | fuidstr_to_sid(zsb, za.za_name, |
9babb374 BB |
478 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
479 | ||
480 | buf->zu_space = za.za_first_integer; | |
481 | buf++; | |
482 | } | |
483 | if (error == ENOENT) | |
484 | error = 0; | |
485 | ||
486 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
487 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
488 | *cookiep = zap_cursor_serialize(&zc); | |
489 | zap_cursor_fini(&zc); | |
490 | return (error); | |
491 | } | |
e5c39b95 | 492 | EXPORT_SYMBOL(zfs_userspace_many); |
9babb374 BB |
493 | |
494 | /* | |
495 | * buf must be big enough (eg, 32 bytes) | |
496 | */ | |
497 | static int | |
3558fd73 | 498 | id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid, |
9babb374 BB |
499 | char *buf, boolean_t addok) |
500 | { | |
501 | uint64_t fuid; | |
502 | int domainid = 0; | |
503 | ||
504 | if (domain && domain[0]) { | |
3558fd73 | 505 | domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok); |
9babb374 | 506 | if (domainid == -1) |
2e528b49 | 507 | return (SET_ERROR(ENOENT)); |
9babb374 BB |
508 | } |
509 | fuid = FUID_ENCODE(domainid, rid); | |
510 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
511 | return (0); | |
512 | } | |
513 | ||
514 | int | |
3558fd73 | 515 | zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
516 | const char *domain, uint64_t rid, uint64_t *valp) |
517 | { | |
518 | char buf[32]; | |
519 | int err; | |
520 | uint64_t obj; | |
521 | ||
522 | *valp = 0; | |
523 | ||
3558fd73 | 524 | if (!dmu_objset_userspace_present(zsb->z_os)) |
2e528b49 | 525 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 526 | |
3558fd73 | 527 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
528 | if (obj == 0) |
529 | return (0); | |
530 | ||
3558fd73 | 531 | err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE); |
9babb374 BB |
532 | if (err) |
533 | return (err); | |
534 | ||
3558fd73 | 535 | err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
536 | if (err == ENOENT) |
537 | err = 0; | |
538 | return (err); | |
539 | } | |
e5c39b95 | 540 | EXPORT_SYMBOL(zfs_userspace_one); |
9babb374 BB |
541 | |
542 | int | |
3558fd73 | 543 | zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
544 | const char *domain, uint64_t rid, uint64_t quota) |
545 | { | |
546 | char buf[32]; | |
547 | int err; | |
548 | dmu_tx_t *tx; | |
549 | uint64_t *objp; | |
550 | boolean_t fuid_dirtied; | |
551 | ||
552 | if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) | |
2e528b49 | 553 | return (SET_ERROR(EINVAL)); |
9babb374 | 554 | |
3558fd73 | 555 | if (zsb->z_version < ZPL_VERSION_USERSPACE) |
2e528b49 | 556 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 557 | |
3558fd73 BB |
558 | objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj : |
559 | &zsb->z_groupquota_obj; | |
9babb374 | 560 | |
3558fd73 | 561 | err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE); |
9babb374 BB |
562 | if (err) |
563 | return (err); | |
3558fd73 | 564 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 565 | |
3558fd73 | 566 | tx = dmu_tx_create(zsb->z_os); |
9babb374 BB |
567 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
568 | if (*objp == 0) { | |
569 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
570 | zfs_userquota_prop_prefixes[type]); | |
571 | } | |
572 | if (fuid_dirtied) | |
3558fd73 | 573 | zfs_fuid_txhold(zsb, tx); |
9babb374 BB |
574 | err = dmu_tx_assign(tx, TXG_WAIT); |
575 | if (err) { | |
576 | dmu_tx_abort(tx); | |
577 | return (err); | |
578 | } | |
579 | ||
3558fd73 | 580 | mutex_enter(&zsb->z_lock); |
9babb374 | 581 | if (*objp == 0) { |
3558fd73 | 582 | *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 583 | DMU_OT_NONE, 0, tx); |
3558fd73 | 584 | VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
585 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
586 | } | |
3558fd73 | 587 | mutex_exit(&zsb->z_lock); |
9babb374 BB |
588 | |
589 | if (quota == 0) { | |
3558fd73 | 590 | err = zap_remove(zsb->z_os, *objp, buf, tx); |
9babb374 BB |
591 | if (err == ENOENT) |
592 | err = 0; | |
593 | } else { | |
3558fd73 | 594 | err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
595 | } |
596 | ASSERT(err == 0); | |
597 | if (fuid_dirtied) | |
3558fd73 | 598 | zfs_fuid_sync(zsb, tx); |
9babb374 BB |
599 | dmu_tx_commit(tx); |
600 | return (err); | |
601 | } | |
e5c39b95 | 602 | EXPORT_SYMBOL(zfs_set_userquota); |
9babb374 BB |
603 | |
604 | boolean_t | |
3558fd73 | 605 | zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid) |
9babb374 BB |
606 | { |
607 | char buf[32]; | |
608 | uint64_t used, quota, usedobj, quotaobj; | |
609 | int err; | |
610 | ||
611 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
3558fd73 | 612 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
9babb374 | 613 | |
3558fd73 | 614 | if (quotaobj == 0 || zsb->z_replay) |
9babb374 BB |
615 | return (B_FALSE); |
616 | ||
617 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
3558fd73 | 618 | err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
619 | if (err != 0) |
620 | return (B_FALSE); | |
621 | ||
3558fd73 | 622 | err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
623 | if (err != 0) |
624 | return (B_FALSE); | |
625 | return (used >= quota); | |
626 | } | |
e5c39b95 | 627 | EXPORT_SYMBOL(zfs_fuid_overquota); |
9babb374 | 628 | |
428870ff | 629 | boolean_t |
3558fd73 | 630 | zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup) |
428870ff BB |
631 | { |
632 | uint64_t fuid; | |
633 | uint64_t quotaobj; | |
428870ff | 634 | |
3558fd73 | 635 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
428870ff | 636 | |
572e2857 | 637 | fuid = isgroup ? zp->z_gid : zp->z_uid; |
428870ff | 638 | |
3558fd73 | 639 | if (quotaobj == 0 || zsb->z_replay) |
428870ff BB |
640 | return (B_FALSE); |
641 | ||
3558fd73 | 642 | return (zfs_fuid_overquota(zsb, isgroup, fuid)); |
428870ff | 643 | } |
e5c39b95 | 644 | EXPORT_SYMBOL(zfs_owner_overquota); |
428870ff | 645 | |
9babb374 | 646 | int |
3558fd73 | 647 | zfs_sb_create(const char *osname, zfs_sb_t **zsbp) |
9babb374 BB |
648 | { |
649 | objset_t *os; | |
3558fd73 | 650 | zfs_sb_t *zsb; |
9babb374 BB |
651 | uint64_t zval; |
652 | int i, error; | |
428870ff | 653 | uint64_t sa_obj; |
9babb374 | 654 | |
efcd79a8 | 655 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
9babb374 | 656 | |
428870ff BB |
657 | /* |
658 | * We claim to always be readonly so we can open snapshots; | |
659 | * other ZPL code will prevent us from writing to snapshots. | |
660 | */ | |
3558fd73 | 661 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); |
428870ff | 662 | if (error) { |
3558fd73 | 663 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 | 664 | return (error); |
428870ff | 665 | } |
9babb374 BB |
666 | |
667 | /* | |
668 | * Initialize the zfs-specific filesystem structure. | |
669 | * Should probably make this a kmem cache, shuffle fields, | |
670 | * and just bzero up to z_hold_mtx[]. | |
671 | */ | |
2cf7f52b | 672 | zsb->z_sb = NULL; |
3558fd73 | 673 | zsb->z_parent = zsb; |
f1512ee6 | 674 | zsb->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; |
3558fd73 BB |
675 | zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; |
676 | zsb->z_os = os; | |
9babb374 | 677 | |
3558fd73 | 678 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); |
9babb374 BB |
679 | if (error) { |
680 | goto out; | |
287be44f | 681 | } else if (zsb->z_version > ZPL_VERSION) { |
2e528b49 | 682 | error = SET_ERROR(ENOTSUP); |
9babb374 BB |
683 | goto out; |
684 | } | |
9babb374 BB |
685 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) |
686 | goto out; | |
3558fd73 | 687 | zsb->z_norm = (int)zval; |
9babb374 BB |
688 | |
689 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
690 | goto out; | |
3558fd73 | 691 | zsb->z_utf8 = (zval != 0); |
9babb374 BB |
692 | |
693 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
694 | goto out; | |
3558fd73 | 695 | zsb->z_case = (uint_t)zval; |
9babb374 | 696 | |
023699cd MM |
697 | if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0) |
698 | goto out; | |
699 | zsb->z_acl_type = (uint_t)zval; | |
700 | ||
9babb374 BB |
701 | /* |
702 | * Fold case on file systems that are always or sometimes case | |
703 | * insensitive. | |
704 | */ | |
3558fd73 BB |
705 | if (zsb->z_case == ZFS_CASE_INSENSITIVE || |
706 | zsb->z_case == ZFS_CASE_MIXED) | |
707 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 708 | |
3558fd73 BB |
709 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
710 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
428870ff | 711 | |
3558fd73 | 712 | if (zsb->z_use_sa) { |
428870ff BB |
713 | /* should either have both of these objects or none */ |
714 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
715 | &sa_obj); | |
716 | if (error) | |
591fb62f | 717 | goto out; |
82a37189 BB |
718 | |
719 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval); | |
720 | if ((error == 0) && (zval == ZFS_XATTR_SA)) | |
721 | zsb->z_xattr_sa = B_TRUE; | |
428870ff BB |
722 | } else { |
723 | /* | |
724 | * Pre SA versions file systems should never touch | |
725 | * either the attribute registration or layout objects. | |
726 | */ | |
727 | sa_obj = 0; | |
728 | } | |
729 | ||
572e2857 | 730 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
3558fd73 | 731 | &zsb->z_attr_table); |
572e2857 BB |
732 | if (error) |
733 | goto out; | |
428870ff | 734 | |
3558fd73 | 735 | if (zsb->z_version >= ZPL_VERSION_SA) |
428870ff | 736 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
737 | |
738 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
3558fd73 | 739 | &zsb->z_root); |
9babb374 BB |
740 | if (error) |
741 | goto out; | |
3558fd73 | 742 | ASSERT(zsb->z_root != 0); |
9babb374 BB |
743 | |
744 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
3558fd73 | 745 | &zsb->z_unlinkedobj); |
9babb374 BB |
746 | if (error) |
747 | goto out; | |
748 | ||
749 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
750 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
3558fd73 | 751 | 8, 1, &zsb->z_userquota_obj); |
9babb374 BB |
752 | if (error && error != ENOENT) |
753 | goto out; | |
754 | ||
755 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
756 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
3558fd73 | 757 | 8, 1, &zsb->z_groupquota_obj); |
9babb374 BB |
758 | if (error && error != ENOENT) |
759 | goto out; | |
760 | ||
761 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
3558fd73 | 762 | &zsb->z_fuid_obj); |
9babb374 BB |
763 | if (error && error != ENOENT) |
764 | goto out; | |
765 | ||
766 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
3558fd73 | 767 | &zsb->z_shares_dir); |
9babb374 BB |
768 | if (error && error != ENOENT) |
769 | goto out; | |
770 | ||
3558fd73 BB |
771 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
772 | mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
773 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
9babb374 | 774 | offsetof(znode_t, z_link_node)); |
13fe0198 | 775 | rrw_init(&zsb->z_teardown_lock, B_FALSE); |
3558fd73 BB |
776 | rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); |
777 | rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
d07b7c7f CD |
778 | |
779 | zsb->z_hold_mtx = vmem_zalloc(sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ, | |
780 | KM_SLEEP); | |
9babb374 | 781 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 782 | mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); |
9babb374 | 783 | |
ebe7e575 BB |
784 | avl_create(&zsb->z_ctldir_snaps, snapentry_compare, |
785 | sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node)); | |
786 | mutex_init(&zsb->z_ctldir_lock, NULL, MUTEX_DEFAULT, NULL); | |
787 | ||
3558fd73 | 788 | *zsbp = zsb; |
9babb374 BB |
789 | return (0); |
790 | ||
791 | out: | |
3558fd73 BB |
792 | dmu_objset_disown(os, zsb); |
793 | *zsbp = NULL; | |
d07b7c7f CD |
794 | |
795 | vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ); | |
3558fd73 | 796 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 BB |
797 | return (error); |
798 | } | |
86f35f34 | 799 | EXPORT_SYMBOL(zfs_sb_create); |
9babb374 | 800 | |
86f35f34 | 801 | int |
3558fd73 | 802 | zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) |
34dc7c2f | 803 | { |
34dc7c2f BB |
804 | int error; |
805 | ||
3558fd73 | 806 | error = zfs_register_callbacks(zsb); |
34dc7c2f BB |
807 | if (error) |
808 | return (error); | |
809 | ||
810 | /* | |
3558fd73 | 811 | * Set the objset user_ptr to track its zsb. |
34dc7c2f | 812 | */ |
3558fd73 BB |
813 | mutex_enter(&zsb->z_os->os_user_ptr_lock); |
814 | dmu_objset_set_user(zsb->z_os, zsb); | |
815 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
34dc7c2f | 816 | |
3558fd73 | 817 | zsb->z_log = zil_open(zsb->z_os, zfs_get_data); |
9babb374 | 818 | |
34dc7c2f BB |
819 | /* |
820 | * If we are not mounting (ie: online recv), then we don't | |
821 | * have to worry about replaying the log as we blocked all | |
822 | * operations out since we closed the ZIL. | |
823 | */ | |
824 | if (mounting) { | |
b128c09f BB |
825 | boolean_t readonly; |
826 | ||
34dc7c2f BB |
827 | /* |
828 | * During replay we remove the read only flag to | |
829 | * allow replays to succeed. | |
830 | */ | |
2cf7f52b | 831 | readonly = zfs_is_readonly(zsb); |
fb5f0bc8 | 832 | if (readonly != 0) |
2cf7f52b | 833 | readonly_changed_cb(zsb, B_FALSE); |
fb5f0bc8 | 834 | else |
3558fd73 | 835 | zfs_unlinked_drain(zsb); |
34dc7c2f | 836 | |
428870ff BB |
837 | /* |
838 | * Parse and replay the intent log. | |
839 | * | |
840 | * Because of ziltest, this must be done after | |
841 | * zfs_unlinked_drain(). (Further note: ziltest | |
842 | * doesn't use readonly mounts, where | |
843 | * zfs_unlinked_drain() isn't called.) This is because | |
844 | * ziltest causes spa_sync() to think it's committed, | |
845 | * but actually it is not, so the intent log contains | |
846 | * many txg's worth of changes. | |
847 | * | |
848 | * In particular, if object N is in the unlinked set in | |
849 | * the last txg to actually sync, then it could be | |
850 | * actually freed in a later txg and then reallocated | |
851 | * in a yet later txg. This would write a "create | |
852 | * object N" record to the intent log. Normally, this | |
853 | * would be fine because the spa_sync() would have | |
854 | * written out the fact that object N is free, before | |
855 | * we could write the "create object N" intent log | |
856 | * record. | |
857 | * | |
858 | * But when we are in ziltest mode, we advance the "open | |
859 | * txg" without actually spa_sync()-ing the changes to | |
860 | * disk. So we would see that object N is still | |
861 | * allocated and in the unlinked set, and there is an | |
862 | * intent log record saying to allocate it. | |
863 | */ | |
3558fd73 | 864 | if (spa_writeable(dmu_objset_spa(zsb->z_os))) { |
572e2857 | 865 | if (zil_replay_disable) { |
3558fd73 | 866 | zil_destroy(zsb->z_log, B_FALSE); |
572e2857 | 867 | } else { |
3558fd73 BB |
868 | zsb->z_replay = B_TRUE; |
869 | zil_replay(zsb->z_os, zsb, | |
572e2857 | 870 | zfs_replay_vector); |
3558fd73 | 871 | zsb->z_replay = B_FALSE; |
572e2857 | 872 | } |
fb5f0bc8 | 873 | } |
2cf7f52b BB |
874 | |
875 | /* restore readonly bit */ | |
876 | if (readonly != 0) | |
877 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
878 | } |
879 | ||
34dc7c2f BB |
880 | return (0); |
881 | } | |
86f35f34 | 882 | EXPORT_SYMBOL(zfs_sb_setup); |
34dc7c2f | 883 | |
9babb374 | 884 | void |
3558fd73 | 885 | zfs_sb_free(zfs_sb_t *zsb) |
34dc7c2f | 886 | { |
9babb374 | 887 | int i; |
9babb374 | 888 | |
3558fd73 | 889 | zfs_fuid_destroy(zsb); |
9babb374 | 890 | |
3558fd73 BB |
891 | mutex_destroy(&zsb->z_znodes_lock); |
892 | mutex_destroy(&zsb->z_lock); | |
893 | list_destroy(&zsb->z_all_znodes); | |
894 | rrw_destroy(&zsb->z_teardown_lock); | |
895 | rw_destroy(&zsb->z_teardown_inactive_lock); | |
896 | rw_destroy(&zsb->z_fuid_lock); | |
9babb374 | 897 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 898 | mutex_destroy(&zsb->z_hold_mtx[i]); |
d07b7c7f | 899 | vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ); |
ebe7e575 BB |
900 | mutex_destroy(&zsb->z_ctldir_lock); |
901 | avl_destroy(&zsb->z_ctldir_snaps); | |
3558fd73 | 902 | kmem_free(zsb, sizeof (zfs_sb_t)); |
34dc7c2f | 903 | } |
86f35f34 | 904 | EXPORT_SYMBOL(zfs_sb_free); |
34dc7c2f | 905 | |
9babb374 | 906 | static void |
3558fd73 | 907 | zfs_set_fuid_feature(zfs_sb_t *zsb) |
9babb374 | 908 | { |
3558fd73 BB |
909 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
910 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
9babb374 | 911 | } |
34dc7c2f BB |
912 | |
913 | void | |
3558fd73 | 914 | zfs_unregister_callbacks(zfs_sb_t *zsb) |
34dc7c2f | 915 | { |
3558fd73 | 916 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
917 | struct dsl_dataset *ds; |
918 | ||
919 | /* | |
920 | * Unregister properties. | |
921 | */ | |
922 | if (!dmu_objset_is_snapshot(os)) { | |
923 | ds = dmu_objset_ds(os); | |
924 | VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, | |
3558fd73 | 925 | zsb) == 0); |
34dc7c2f | 926 | |
6d111134 TC |
927 | VERIFY(dsl_prop_unregister(ds, "relatime", relatime_changed_cb, |
928 | zsb) == 0); | |
929 | ||
34dc7c2f | 930 | VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, |
3558fd73 | 931 | zsb) == 0); |
34dc7c2f BB |
932 | |
933 | VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, | |
3558fd73 | 934 | zsb) == 0); |
34dc7c2f BB |
935 | |
936 | VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, | |
3558fd73 | 937 | zsb) == 0); |
34dc7c2f BB |
938 | |
939 | VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, | |
3558fd73 | 940 | zsb) == 0); |
34dc7c2f BB |
941 | |
942 | VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, | |
3558fd73 | 943 | zsb) == 0); |
34dc7c2f BB |
944 | |
945 | VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, | |
3558fd73 | 946 | zsb) == 0); |
34dc7c2f BB |
947 | |
948 | VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, | |
3558fd73 | 949 | zsb) == 0); |
34dc7c2f | 950 | |
023699cd MM |
951 | VERIFY(dsl_prop_unregister(ds, "acltype", acltype_changed_cb, |
952 | zsb) == 0); | |
953 | ||
34dc7c2f | 954 | VERIFY(dsl_prop_unregister(ds, "aclinherit", |
3558fd73 | 955 | acl_inherit_changed_cb, zsb) == 0); |
34dc7c2f BB |
956 | |
957 | VERIFY(dsl_prop_unregister(ds, "vscan", | |
3558fd73 | 958 | vscan_changed_cb, zsb) == 0); |
2cf7f52b BB |
959 | |
960 | VERIFY(dsl_prop_unregister(ds, "nbmand", | |
961 | nbmand_changed_cb, zsb) == 0); | |
34dc7c2f BB |
962 | } |
963 | } | |
e5c39b95 | 964 | EXPORT_SYMBOL(zfs_unregister_callbacks); |
34dc7c2f | 965 | |
bc3e15e3 | 966 | #ifdef HAVE_MLSLABEL |
428870ff | 967 | /* |
d3cc8b15 WA |
968 | * Check that the hex label string is appropriate for the dataset being |
969 | * mounted into the global_zone proper. | |
428870ff | 970 | * |
d3cc8b15 WA |
971 | * Return an error if the hex label string is not default or |
972 | * admin_low/admin_high. For admin_low labels, the corresponding | |
973 | * dataset must be readonly. | |
428870ff BB |
974 | */ |
975 | int | |
976 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
977 | { | |
978 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
979 | return (0); | |
980 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
981 | return (0); | |
982 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
983 | /* must be readonly */ | |
984 | uint64_t rdonly; | |
985 | ||
986 | if (dsl_prop_get_integer(dsname, | |
987 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
2e528b49 | 988 | return (SET_ERROR(EACCES)); |
428870ff BB |
989 | return (rdonly ? 0 : EACCES); |
990 | } | |
2e528b49 | 991 | return (SET_ERROR(EACCES)); |
428870ff | 992 | } |
86f35f34 | 993 | EXPORT_SYMBOL(zfs_check_global_label); |
bc3e15e3 | 994 | #endif /* HAVE_MLSLABEL */ |
428870ff | 995 | |
e5c39b95 | 996 | int |
3558fd73 | 997 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 998 | { |
3558fd73 | 999 | zfs_sb_t *zsb = dentry->d_sb->s_fs_info; |
34dc7c2f | 1000 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
04f9432d | 1001 | uint64_t fsid; |
3558fd73 | 1002 | uint32_t bshift; |
34dc7c2f | 1003 | |
3558fd73 | 1004 | ZFS_ENTER(zsb); |
34dc7c2f | 1005 | |
3558fd73 | 1006 | dmu_objset_space(zsb->z_os, |
34dc7c2f BB |
1007 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
1008 | ||
04f9432d | 1009 | fsid = dmu_objset_fsid_guid(zsb->z_os); |
34dc7c2f | 1010 | /* |
05ff35c6 BB |
1011 | * The underlying storage pool actually uses multiple block |
1012 | * size. Under Solaris frsize (fragment size) is reported as | |
1013 | * the smallest block size we support, and bsize (block size) | |
1014 | * as the filesystem's maximum block size. Unfortunately, | |
1015 | * under Linux the fragment size and block size are often used | |
1016 | * interchangeably. Thus we are forced to report both of them | |
1017 | * as the filesystem's maximum block size. | |
34dc7c2f | 1018 | */ |
05ff35c6 | 1019 | statp->f_frsize = zsb->z_max_blksz; |
3558fd73 BB |
1020 | statp->f_bsize = zsb->z_max_blksz; |
1021 | bshift = fls(statp->f_bsize) - 1; | |
34dc7c2f BB |
1022 | |
1023 | /* | |
3558fd73 BB |
1024 | * The following report "total" blocks of various kinds in |
1025 | * the file system, but reported in terms of f_bsize - the | |
1026 | * "preferred" size. | |
34dc7c2f BB |
1027 | */ |
1028 | ||
3558fd73 BB |
1029 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
1030 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
1031 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
1032 | ||
1033 | /* | |
1034 | * statvfs() should really be called statufs(), because it assumes | |
1035 | * static metadata. ZFS doesn't preallocate files, so the best | |
1036 | * we can do is report the max that could possibly fit in f_files, | |
1037 | * and that minus the number actually used in f_ffree. | |
1038 | * For f_ffree, report the smaller of the number of object available | |
1039 | * and the number of blocks (each object will take at least a block). | |
1040 | */ | |
baab0630 | 1041 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 1042 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
1043 | statp->f_fsid.val[0] = (uint32_t)fsid; |
1044 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 BB |
1045 | statp->f_type = ZFS_SUPER_MAGIC; |
1046 | statp->f_namelen = ZFS_MAXNAMELEN; | |
34dc7c2f BB |
1047 | |
1048 | /* | |
3558fd73 | 1049 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
1050 | * Is there anything useful we could/should provide? |
1051 | */ | |
3558fd73 | 1052 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 1053 | |
3558fd73 | 1054 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1055 | return (0); |
1056 | } | |
e5c39b95 | 1057 | EXPORT_SYMBOL(zfs_statvfs); |
34dc7c2f | 1058 | |
e5c39b95 | 1059 | int |
3558fd73 | 1060 | zfs_root(zfs_sb_t *zsb, struct inode **ipp) |
34dc7c2f | 1061 | { |
34dc7c2f BB |
1062 | znode_t *rootzp; |
1063 | int error; | |
1064 | ||
3558fd73 | 1065 | ZFS_ENTER(zsb); |
34dc7c2f | 1066 | |
3558fd73 | 1067 | error = zfs_zget(zsb, zsb->z_root, &rootzp); |
34dc7c2f | 1068 | if (error == 0) |
3558fd73 | 1069 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1070 | |
3558fd73 | 1071 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1072 | return (error); |
1073 | } | |
e5c39b95 | 1074 | EXPORT_SYMBOL(zfs_root); |
34dc7c2f | 1075 | |
2cbb06b5 BB |
1076 | /* |
1077 | * The ARC has requested that the filesystem drop entries from the dentry | |
1078 | * and inode caches. This can occur when the ARC needs to free meta data | |
1079 | * blocks but can't because they are all pinned by entries in these caches. | |
1080 | */ | |
ab26409d BB |
1081 | int |
1082 | zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) | |
1083 | { | |
1084 | zfs_sb_t *zsb = sb->s_fs_info; | |
2cbb06b5 BB |
1085 | int error = 0; |
1086 | #if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK) | |
ab26409d BB |
1087 | struct shrinker *shrinker = &sb->s_shrink; |
1088 | struct shrink_control sc = { | |
1089 | .nr_to_scan = nr_to_scan, | |
1090 | .gfp_mask = GFP_KERNEL, | |
1091 | }; | |
2cbb06b5 | 1092 | #endif |
ab26409d BB |
1093 | |
1094 | ZFS_ENTER(zsb); | |
2cbb06b5 BB |
1095 | |
1096 | #if defined(HAVE_SPLIT_SHRINKER_CALLBACK) | |
3c832b8c | 1097 | *objects = (*shrinker->scan_objects)(shrinker, &sc); |
2cbb06b5 | 1098 | #elif defined(HAVE_SHRINK) |
ab26409d | 1099 | *objects = (*shrinker->shrink)(shrinker, &sc); |
2cbb06b5 BB |
1100 | #else |
1101 | /* | |
1102 | * Linux kernels older than 3.1 do not support a per-filesystem | |
1103 | * shrinker. Therefore, we must fall back to the only available | |
1104 | * interface which is to discard all unused dentries and inodes. | |
1105 | * This behavior clearly isn't ideal but it's required so the ARC | |
1106 | * may free memory. The performance impact is mitigated by the | |
1107 | * fact that the frequently accessed dentry and inode buffers will | |
1108 | * still be in the ARC making them relatively cheap to recreate. | |
1109 | */ | |
1110 | *objects = 0; | |
1111 | shrink_dcache_parent(sb->s_root); | |
3c832b8c | 1112 | #endif |
ab26409d BB |
1113 | ZFS_EXIT(zsb); |
1114 | ||
2cbb06b5 BB |
1115 | dprintf_ds(zsb->z_os->os_dsl_dataset, |
1116 | "pruning, nr_to_scan=%lu objects=%d error=%d\n", | |
1117 | nr_to_scan, *objects, error); | |
1118 | ||
1119 | return (error); | |
ab26409d BB |
1120 | } |
1121 | EXPORT_SYMBOL(zfs_sb_prune); | |
ab26409d | 1122 | |
34dc7c2f | 1123 | /* |
7b3e34ba | 1124 | * Teardown the zfs_sb_t. |
34dc7c2f BB |
1125 | * |
1126 | * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' | |
1127 | * and 'z_teardown_inactive_lock' held. | |
1128 | */ | |
3558fd73 | 1129 | int |
86f35f34 | 1130 | zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting) |
34dc7c2f BB |
1131 | { |
1132 | znode_t *zp; | |
1133 | ||
fd23720a BB |
1134 | /* |
1135 | * If someone has not already unmounted this file system, | |
1136 | * drain the iput_taskq to ensure all active references to the | |
1137 | * zfs_sb_t have been handled only then can it be safely destroyed. | |
1138 | */ | |
f0da4d15 CD |
1139 | if (zsb->z_os) { |
1140 | /* | |
1141 | * If we're unmounting we have to wait for the list to | |
1142 | * drain completely. | |
1143 | * | |
1144 | * If we're not unmounting there's no guarantee the list | |
1145 | * will drain completely, but iputs run from the taskq | |
1146 | * may add the parents of dir-based xattrs to the taskq | |
1147 | * so we want to wait for these. | |
1148 | * | |
1149 | * We can safely read z_nr_znodes without locking because the | |
1150 | * VFS has already blocked operations which add to the | |
1151 | * z_all_znodes list and thus increment z_nr_znodes. | |
1152 | */ | |
1153 | int round = 0; | |
1154 | while (zsb->z_nr_znodes > 0) { | |
1155 | taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool( | |
1156 | zsb->z_os))); | |
1157 | if (++round > 1 && !unmounting) | |
1158 | break; | |
1159 | } | |
1160 | } | |
fd23720a | 1161 | |
3558fd73 | 1162 | rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1163 | |
1164 | if (!unmounting) { | |
1165 | /* | |
ceb43b93 BB |
1166 | * We purge the parent filesystem's super block as the |
1167 | * parent filesystem and all of its snapshots have their | |
1168 | * inode's super block set to the parent's filesystem's | |
1169 | * super block. Note, 'z_parent' is self referential | |
1170 | * for non-snapshots. | |
34dc7c2f | 1171 | */ |
ceb43b93 | 1172 | shrink_dcache_sb(zsb->z_parent->z_sb); |
34dc7c2f BB |
1173 | } |
1174 | ||
1175 | /* | |
1176 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1177 | * threads are blocked as zil_close can call zfs_inactive. | |
1178 | */ | |
3558fd73 BB |
1179 | if (zsb->z_log) { |
1180 | zil_close(zsb->z_log); | |
1181 | zsb->z_log = NULL; | |
34dc7c2f BB |
1182 | } |
1183 | ||
3558fd73 | 1184 | rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1185 | |
1186 | /* | |
1187 | * If we are not unmounting (ie: online recv) and someone already | |
1188 | * unmounted this file system while we were doing the switcheroo, | |
1189 | * or a reopen of z_os failed then just bail out now. | |
1190 | */ | |
3558fd73 BB |
1191 | if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { |
1192 | rw_exit(&zsb->z_teardown_inactive_lock); | |
1193 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
2e528b49 | 1194 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1195 | } |
1196 | ||
1197 | /* | |
7b3e34ba BB |
1198 | * At this point there are no VFS ops active, and any new VFS ops |
1199 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1200 | * relevant for forced unmount). | |
34dc7c2f BB |
1201 | * |
1202 | * Release all holds on dbufs. | |
1203 | */ | |
f0da4d15 CD |
1204 | if (!unmounting) { |
1205 | mutex_enter(&zsb->z_znodes_lock); | |
1206 | for (zp = list_head(&zsb->z_all_znodes); zp != NULL; | |
1207 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
1208 | if (zp->z_sa_hdl) | |
1209 | zfs_znode_dmu_fini(zp); | |
1210 | } | |
1211 | mutex_exit(&zsb->z_znodes_lock); | |
7b3e34ba | 1212 | } |
34dc7c2f BB |
1213 | |
1214 | /* | |
7b3e34ba | 1215 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1216 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1217 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1218 | */ |
1219 | if (unmounting) { | |
3558fd73 BB |
1220 | zsb->z_unmounted = B_TRUE; |
1221 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
1222 | rw_exit(&zsb->z_teardown_inactive_lock); | |
34dc7c2f BB |
1223 | } |
1224 | ||
1225 | /* | |
1226 | * z_os will be NULL if there was an error in attempting to reopen | |
3558fd73 BB |
1227 | * zsb, so just return as the properties had already been |
1228 | * | |
34dc7c2f BB |
1229 | * unregistered and cached data had been evicted before. |
1230 | */ | |
3558fd73 | 1231 | if (zsb->z_os == NULL) |
34dc7c2f BB |
1232 | return (0); |
1233 | ||
1234 | /* | |
1235 | * Unregister properties. | |
1236 | */ | |
3558fd73 | 1237 | zfs_unregister_callbacks(zsb); |
34dc7c2f BB |
1238 | |
1239 | /* | |
1240 | * Evict cached data | |
1241 | */ | |
04434775 MA |
1242 | if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) && |
1243 | !zfs_is_readonly(zsb)) | |
1244 | txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); | |
13fe0198 | 1245 | dmu_objset_evict_dbufs(zsb->z_os); |
34dc7c2f BB |
1246 | |
1247 | return (0); | |
1248 | } | |
86f35f34 | 1249 | EXPORT_SYMBOL(zfs_sb_teardown); |
34dc7c2f | 1250 | |
8c45def2 BB |
1251 | #if !defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER) && \ |
1252 | !defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER) | |
5547c2f1 | 1253 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); |
8c45def2 | 1254 | #endif |
76659dc1 | 1255 | |
e5c39b95 | 1256 | int |
3558fd73 | 1257 | zfs_domount(struct super_block *sb, void *data, int silent) |
34dc7c2f | 1258 | { |
3558fd73 BB |
1259 | zpl_mount_data_t *zmd = data; |
1260 | const char *osname = zmd->z_osname; | |
1261 | zfs_sb_t *zsb; | |
1262 | struct inode *root_inode; | |
1263 | uint64_t recordsize; | |
1264 | int error; | |
34dc7c2f | 1265 | |
3558fd73 BB |
1266 | error = zfs_sb_create(osname, &zsb); |
1267 | if (error) | |
1268 | return (error); | |
1269 | ||
1270 | if ((error = dsl_prop_get_integer(osname, "recordsize", | |
1271 | &recordsize, NULL))) | |
1272 | goto out; | |
1273 | ||
1274 | zsb->z_sb = sb; | |
3558fd73 BB |
1275 | sb->s_fs_info = zsb; |
1276 | sb->s_magic = ZFS_SUPER_MAGIC; | |
1277 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1278 | sb->s_time_gran = 1; | |
1279 | sb->s_blocksize = recordsize; | |
1280 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 BB |
1281 | zsb->z_bdi.ra_pages = 0; |
1282 | sb->s_bdi = &zsb->z_bdi; | |
1283 | ||
8c45def2 | 1284 | error = -zpl_bdi_setup_and_register(&zsb->z_bdi, "zfs"); |
5547c2f1 BB |
1285 | if (error) |
1286 | goto out; | |
3558fd73 BB |
1287 | |
1288 | /* Set callback operations for the file system. */ | |
1289 | sb->s_op = &zpl_super_operations; | |
1290 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1291 | sb->s_export_op = &zpl_export_operations; |
ee930353 BB |
1292 | #ifdef HAVE_S_D_OP |
1293 | sb->s_d_op = &zpl_dentry_operations; | |
1294 | #endif /* HAVE_S_D_OP */ | |
3558fd73 BB |
1295 | |
1296 | /* Set features for file system. */ | |
1297 | zfs_set_fuid_feature(zsb); | |
1298 | ||
1299 | if (dmu_objset_is_snapshot(zsb->z_os)) { | |
1300 | uint64_t pval; | |
1301 | ||
1302 | atime_changed_cb(zsb, B_FALSE); | |
1303 | readonly_changed_cb(zsb, B_TRUE); | |
d1d7e268 MK |
1304 | if ((error = dsl_prop_get_integer(osname, |
1305 | "xattr", &pval, NULL))) | |
3558fd73 BB |
1306 | goto out; |
1307 | xattr_changed_cb(zsb, pval); | |
d1d7e268 MK |
1308 | if ((error = dsl_prop_get_integer(osname, |
1309 | "acltype", &pval, NULL))) | |
023699cd MM |
1310 | goto out; |
1311 | acltype_changed_cb(zsb, pval); | |
3558fd73 BB |
1312 | zsb->z_issnap = B_TRUE; |
1313 | zsb->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1314 | ||
1315 | mutex_enter(&zsb->z_os->os_user_ptr_lock); | |
1316 | dmu_objset_set_user(zsb->z_os, zsb); | |
1317 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
1318 | } else { | |
1319 | error = zfs_sb_setup(zsb, B_TRUE); | |
34dc7c2f BB |
1320 | } |
1321 | ||
3558fd73 BB |
1322 | /* Allocate a root inode for the filesystem. */ |
1323 | error = zfs_root(zsb, &root_inode); | |
1324 | if (error) { | |
1325 | (void) zfs_umount(sb); | |
1326 | goto out; | |
34dc7c2f BB |
1327 | } |
1328 | ||
3558fd73 | 1329 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1330 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1331 | if (sb->s_root == NULL) { |
1332 | (void) zfs_umount(sb); | |
2e528b49 | 1333 | error = SET_ERROR(ENOMEM); |
3558fd73 BB |
1334 | goto out; |
1335 | } | |
ebe7e575 BB |
1336 | |
1337 | if (!zsb->z_issnap) | |
1338 | zfsctl_create(zsb); | |
2cbb06b5 BB |
1339 | |
1340 | zsb->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); | |
3558fd73 BB |
1341 | out: |
1342 | if (error) { | |
1343 | dmu_objset_disown(zsb->z_os, zsb); | |
1344 | zfs_sb_free(zsb); | |
1345 | } | |
34dc7c2f | 1346 | |
3558fd73 BB |
1347 | return (error); |
1348 | } | |
1349 | EXPORT_SYMBOL(zfs_domount); | |
1350 | ||
ebe7e575 BB |
1351 | /* |
1352 | * Called when an unmount is requested and certain sanity checks have | |
1353 | * already passed. At this point no dentries or inodes have been reclaimed | |
1354 | * from their respective caches. We drop the extra reference on the .zfs | |
1355 | * control directory to allow everything to be reclaimed. All snapshots | |
1356 | * must already have been unmounted to reach this point. | |
1357 | */ | |
1358 | void | |
1359 | zfs_preumount(struct super_block *sb) | |
1360 | { | |
1361 | zfs_sb_t *zsb = sb->s_fs_info; | |
1362 | ||
1f0d8a56 | 1363 | if (zsb != NULL && zsb->z_ctldir != NULL) |
ebe7e575 BB |
1364 | zfsctl_destroy(zsb); |
1365 | } | |
1366 | EXPORT_SYMBOL(zfs_preumount); | |
1367 | ||
1368 | /* | |
1369 | * Called once all other unmount released tear down has occurred. | |
1370 | * It is our responsibility to release any remaining infrastructure. | |
1371 | */ | |
3558fd73 BB |
1372 | /*ARGSUSED*/ |
1373 | int | |
1374 | zfs_umount(struct super_block *sb) | |
1375 | { | |
1376 | zfs_sb_t *zsb = sb->s_fs_info; | |
1377 | objset_t *os; | |
1378 | ||
2cbb06b5 | 1379 | arc_remove_prune_callback(zsb->z_arc_prune); |
86f35f34 | 1380 | VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0); |
3558fd73 | 1381 | os = zsb->z_os; |
5547c2f1 | 1382 | bdi_destroy(sb->s_bdi); |
76659dc1 | 1383 | |
34dc7c2f BB |
1384 | /* |
1385 | * z_os will be NULL if there was an error in | |
3558fd73 | 1386 | * attempting to reopen zsb. |
34dc7c2f BB |
1387 | */ |
1388 | if (os != NULL) { | |
1389 | /* | |
1390 | * Unset the objset user_ptr. | |
1391 | */ | |
428870ff | 1392 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1393 | dmu_objset_set_user(os, NULL); |
428870ff | 1394 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1395 | |
1396 | /* | |
b128c09f | 1397 | * Finally release the objset |
34dc7c2f | 1398 | */ |
3558fd73 | 1399 | dmu_objset_disown(os, zsb); |
34dc7c2f BB |
1400 | } |
1401 | ||
3558fd73 | 1402 | zfs_sb_free(zsb); |
34dc7c2f BB |
1403 | return (0); |
1404 | } | |
e5c39b95 | 1405 | EXPORT_SYMBOL(zfs_umount); |
34dc7c2f | 1406 | |
0de19dad BB |
1407 | int |
1408 | zfs_remount(struct super_block *sb, int *flags, char *data) | |
1409 | { | |
0de19dad | 1410 | /* |
2cf7f52b BB |
1411 | * All namespace flags (MNT_*) and super block flags (MS_*) will |
1412 | * be handled by the Linux VFS. Only handle custom options here. | |
0de19dad | 1413 | */ |
0de19dad BB |
1414 | return (0); |
1415 | } | |
1416 | EXPORT_SYMBOL(zfs_remount); | |
1417 | ||
e5c39b95 | 1418 | int |
2cf7f52b | 1419 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1420 | { |
2cf7f52b | 1421 | zfs_sb_t *zsb = sb->s_fs_info; |
34dc7c2f BB |
1422 | znode_t *zp; |
1423 | uint64_t object = 0; | |
1424 | uint64_t fid_gen = 0; | |
1425 | uint64_t gen_mask; | |
1426 | uint64_t zp_gen; | |
3558fd73 | 1427 | int i, err; |
34dc7c2f | 1428 | |
3558fd73 | 1429 | *ipp = NULL; |
34dc7c2f | 1430 | |
3558fd73 | 1431 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1432 | |
1433 | if (fidp->fid_len == LONG_FID_LEN) { | |
1434 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1435 | uint64_t objsetid = 0; | |
1436 | uint64_t setgen = 0; | |
1437 | ||
1438 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1439 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1440 | ||
1441 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1442 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1443 | ||
3558fd73 | 1444 | ZFS_EXIT(zsb); |
34dc7c2f | 1445 | |
ebe7e575 | 1446 | err = zfsctl_lookup_objset(sb, objsetid, &zsb); |
34dc7c2f | 1447 | if (err) |
2e528b49 | 1448 | return (SET_ERROR(EINVAL)); |
ebe7e575 | 1449 | |
3558fd73 | 1450 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1451 | } |
1452 | ||
1453 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { | |
1454 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1455 | ||
1456 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1457 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
1458 | ||
1459 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
1460 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1461 | } else { | |
3558fd73 | 1462 | ZFS_EXIT(zsb); |
2e528b49 | 1463 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1464 | } |
1465 | ||
1466 | /* A zero fid_gen means we are in the .zfs control directories */ | |
1467 | if (fid_gen == 0 && | |
1468 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
3558fd73 BB |
1469 | *ipp = zsb->z_ctldir; |
1470 | ASSERT(*ipp != NULL); | |
34dc7c2f | 1471 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1472 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1473 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1474 | } else { |
3558fd73 | 1475 | igrab(*ipp); |
34dc7c2f | 1476 | } |
3558fd73 | 1477 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1478 | return (0); |
1479 | } | |
1480 | ||
1481 | gen_mask = -1ULL >> (64 - 8 * i); | |
1482 | ||
29e57d15 | 1483 | dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); |
3558fd73 BB |
1484 | if ((err = zfs_zget(zsb, object, &zp))) { |
1485 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1486 | return (err); |
1487 | } | |
3558fd73 | 1488 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, |
428870ff BB |
1489 | sizeof (uint64_t)); |
1490 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1491 | if (zp_gen == 0) |
1492 | zp_gen = 1; | |
1493 | if (zp->z_unlinked || zp_gen != fid_gen) { | |
29e57d15 NB |
1494 | dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, |
1495 | fid_gen); | |
3558fd73 BB |
1496 | iput(ZTOI(zp)); |
1497 | ZFS_EXIT(zsb); | |
2e528b49 | 1498 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1499 | } |
1500 | ||
3558fd73 BB |
1501 | *ipp = ZTOI(zp); |
1502 | if (*ipp) | |
1503 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1504 | |
3558fd73 | 1505 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1506 | return (0); |
1507 | } | |
e5c39b95 | 1508 | EXPORT_SYMBOL(zfs_vget); |
34dc7c2f BB |
1509 | |
1510 | /* | |
7b3e34ba | 1511 | * Block out VFS ops and close zfs_sb_t |
34dc7c2f BB |
1512 | * |
1513 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
831baf06 KW |
1514 | * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying |
1515 | * dataset and objset intact so that they can be atomically handed off during | |
1516 | * a subsequent rollback or recv operation and the resume thereafter. | |
34dc7c2f BB |
1517 | */ |
1518 | int | |
3558fd73 | 1519 | zfs_suspend_fs(zfs_sb_t *zsb) |
34dc7c2f BB |
1520 | { |
1521 | int error; | |
1522 | ||
86f35f34 | 1523 | if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0) |
34dc7c2f | 1524 | return (error); |
7b3e34ba | 1525 | |
34dc7c2f BB |
1526 | return (0); |
1527 | } | |
e5c39b95 | 1528 | EXPORT_SYMBOL(zfs_suspend_fs); |
34dc7c2f BB |
1529 | |
1530 | /* | |
7b3e34ba | 1531 | * Reopen zfs_sb_t and release VFS ops. |
34dc7c2f BB |
1532 | */ |
1533 | int | |
3558fd73 | 1534 | zfs_resume_fs(zfs_sb_t *zsb, const char *osname) |
34dc7c2f | 1535 | { |
2008e920 | 1536 | int err, err2; |
831baf06 KW |
1537 | znode_t *zp; |
1538 | uint64_t sa_obj = 0; | |
34dc7c2f | 1539 | |
3558fd73 BB |
1540 | ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock)); |
1541 | ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); | |
34dc7c2f | 1542 | |
831baf06 KW |
1543 | /* |
1544 | * We already own this, so just hold and rele it to update the | |
1545 | * objset_t, as the one we had before may have been evicted. | |
1546 | */ | |
1547 | VERIFY0(dmu_objset_hold(osname, zsb, &zsb->z_os)); | |
1548 | VERIFY3P(zsb->z_os->os_dsl_dataset->ds_owner, ==, zsb); | |
1549 | VERIFY(dsl_dataset_long_held(zsb->z_os->os_dsl_dataset)); | |
1550 | dmu_objset_rele(zsb->z_os, zsb); | |
428870ff | 1551 | |
831baf06 KW |
1552 | /* |
1553 | * Make sure version hasn't changed | |
1554 | */ | |
428870ff | 1555 | |
831baf06 KW |
1556 | err = zfs_get_zplprop(zsb->z_os, ZFS_PROP_VERSION, |
1557 | &zsb->z_version); | |
c1fabe79 | 1558 | |
831baf06 KW |
1559 | if (err) |
1560 | goto bail; | |
428870ff | 1561 | |
831baf06 KW |
1562 | err = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, |
1563 | ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
c1fabe79 | 1564 | |
831baf06 KW |
1565 | if (err && zsb->z_version >= ZPL_VERSION_SA) |
1566 | goto bail; | |
428870ff | 1567 | |
831baf06 KW |
1568 | if ((err = sa_setup(zsb->z_os, sa_obj, |
1569 | zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) | |
1570 | goto bail; | |
34dc7c2f | 1571 | |
831baf06 KW |
1572 | if (zsb->z_version >= ZPL_VERSION_SA) |
1573 | sa_register_update_callback(zsb->z_os, | |
1574 | zfs_sa_upgrade); | |
c1fabe79 | 1575 | |
831baf06 | 1576 | VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); |
c1fabe79 | 1577 | |
831baf06 KW |
1578 | zfs_set_fuid_feature(zsb); |
1579 | zsb->z_rollback_time = jiffies; | |
34dc7c2f | 1580 | |
831baf06 KW |
1581 | /* |
1582 | * Attempt to re-establish all the active inodes with their | |
1583 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
1584 | * and mark it stale. This prevents a collision if a new | |
1585 | * inode/object is created which must use the same inode | |
1586 | * number. The stale inode will be be released when the | |
1587 | * VFS prunes the dentry holding the remaining references | |
1588 | * on the stale inode. | |
1589 | */ | |
1590 | mutex_enter(&zsb->z_znodes_lock); | |
1591 | for (zp = list_head(&zsb->z_all_znodes); zp; | |
1592 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
2008e920 TC |
1593 | err2 = zfs_rezget(zp); |
1594 | if (err2) { | |
831baf06 KW |
1595 | remove_inode_hash(ZTOI(zp)); |
1596 | zp->z_is_stale = B_TRUE; | |
34dc7c2f | 1597 | } |
34dc7c2f | 1598 | } |
831baf06 | 1599 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f | 1600 | |
428870ff | 1601 | bail: |
7b3e34ba | 1602 | /* release the VFS ops */ |
3558fd73 BB |
1603 | rw_exit(&zsb->z_teardown_inactive_lock); |
1604 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1605 | |
1606 | if (err) { | |
1607 | /* | |
831baf06 KW |
1608 | * Since we couldn't setup the sa framework, try to force |
1609 | * unmount this file system. | |
34dc7c2f | 1610 | */ |
ab68b6e5 TC |
1611 | if (zsb->z_os) |
1612 | (void) zfs_umount(zsb->z_sb); | |
34dc7c2f BB |
1613 | } |
1614 | return (err); | |
1615 | } | |
e5c39b95 | 1616 | EXPORT_SYMBOL(zfs_resume_fs); |
34dc7c2f | 1617 | |
34dc7c2f | 1618 | int |
3558fd73 | 1619 | zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) |
34dc7c2f BB |
1620 | { |
1621 | int error; | |
3558fd73 | 1622 | objset_t *os = zsb->z_os; |
34dc7c2f | 1623 | dmu_tx_t *tx; |
34dc7c2f BB |
1624 | |
1625 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
2e528b49 | 1626 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1627 | |
3558fd73 | 1628 | if (newvers < zsb->z_version) |
2e528b49 | 1629 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1630 | |
428870ff | 1631 | if (zfs_spa_version_map(newvers) > |
3558fd73 | 1632 | spa_version(dmu_objset_spa(zsb->z_os))) |
2e528b49 | 1633 | return (SET_ERROR(ENOTSUP)); |
428870ff | 1634 | |
34dc7c2f | 1635 | tx = dmu_tx_create(os); |
9babb374 | 1636 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
3558fd73 | 1637 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1638 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1639 | ZFS_SA_ATTRS); | |
1640 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
1641 | } | |
34dc7c2f BB |
1642 | error = dmu_tx_assign(tx, TXG_WAIT); |
1643 | if (error) { | |
1644 | dmu_tx_abort(tx); | |
9babb374 BB |
1645 | return (error); |
1646 | } | |
428870ff | 1647 | |
9babb374 BB |
1648 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
1649 | 8, 1, &newvers, tx); | |
1650 | ||
1651 | if (error) { | |
1652 | dmu_tx_commit(tx); | |
1653 | return (error); | |
34dc7c2f | 1654 | } |
34dc7c2f | 1655 | |
3558fd73 | 1656 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1657 | uint64_t sa_obj; |
1658 | ||
3558fd73 | 1659 | ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, |
428870ff BB |
1660 | SPA_VERSION_SA); |
1661 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1662 | DMU_OT_NONE, 0, tx); | |
1663 | ||
1664 | error = zap_add(os, MASTER_NODE_OBJ, | |
1665 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c99c9001 | 1666 | ASSERT0(error); |
428870ff BB |
1667 | |
1668 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
1669 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1670 | } | |
1671 | ||
6f1ffb06 MA |
1672 | spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, |
1673 | "from %llu to %llu", zsb->z_version, newvers); | |
9babb374 | 1674 | |
34dc7c2f BB |
1675 | dmu_tx_commit(tx); |
1676 | ||
3558fd73 | 1677 | zsb->z_version = newvers; |
9babb374 | 1678 | |
c1fabe79 | 1679 | zfs_set_fuid_feature(zsb); |
9babb374 BB |
1680 | |
1681 | return (0); | |
34dc7c2f | 1682 | } |
e5c39b95 | 1683 | EXPORT_SYMBOL(zfs_set_version); |
34dc7c2f BB |
1684 | |
1685 | /* | |
1686 | * Read a property stored within the master node. | |
1687 | */ | |
1688 | int | |
1689 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
1690 | { | |
1691 | const char *pname; | |
2e528b49 | 1692 | int error = SET_ERROR(ENOENT); |
34dc7c2f BB |
1693 | |
1694 | /* | |
1695 | * Look up the file system's value for the property. For the | |
1696 | * version property, we look up a slightly different string. | |
1697 | */ | |
1698 | if (prop == ZFS_PROP_VERSION) | |
1699 | pname = ZPL_VERSION_STR; | |
1700 | else | |
1701 | pname = zfs_prop_to_name(prop); | |
1702 | ||
b128c09f BB |
1703 | if (os != NULL) |
1704 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
1705 | |
1706 | if (error == ENOENT) { | |
1707 | /* No value set, use the default value */ | |
1708 | switch (prop) { | |
1709 | case ZFS_PROP_VERSION: | |
1710 | *value = ZPL_VERSION; | |
1711 | break; | |
1712 | case ZFS_PROP_NORMALIZE: | |
1713 | case ZFS_PROP_UTF8ONLY: | |
1714 | *value = 0; | |
1715 | break; | |
1716 | case ZFS_PROP_CASE: | |
1717 | *value = ZFS_CASE_SENSITIVE; | |
1718 | break; | |
023699cd MM |
1719 | case ZFS_PROP_ACLTYPE: |
1720 | *value = ZFS_ACLTYPE_OFF; | |
1721 | break; | |
34dc7c2f BB |
1722 | default: |
1723 | return (error); | |
1724 | } | |
1725 | error = 0; | |
1726 | } | |
1727 | return (error); | |
1728 | } | |
86f35f34 | 1729 | EXPORT_SYMBOL(zfs_get_zplprop); |
3558fd73 BB |
1730 | |
1731 | void | |
1732 | zfs_init(void) | |
1733 | { | |
ebe7e575 | 1734 | zfsctl_init(); |
3558fd73 BB |
1735 | zfs_znode_init(); |
1736 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
1737 | register_filesystem(&zpl_fs_type); | |
1738 | } | |
1739 | ||
1740 | void | |
1741 | zfs_fini(void) | |
1742 | { | |
023699cd | 1743 | taskq_wait(system_taskq); |
3558fd73 BB |
1744 | unregister_filesystem(&zpl_fs_type); |
1745 | zfs_znode_fini(); | |
ebe7e575 | 1746 | zfsctl_fini(); |
3558fd73 | 1747 | } |