]>
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; |
34dc7c2f BB |
191 | |
192 | if (newval < SPA_MINBLOCKSIZE || | |
193 | newval > SPA_MAXBLOCKSIZE || !ISP2(newval)) | |
194 | newval = SPA_MAXBLOCKSIZE; | |
195 | ||
3558fd73 | 196 | zsb->z_max_blksz = newval; |
34dc7c2f BB |
197 | } |
198 | ||
199 | static void | |
200 | readonly_changed_cb(void *arg, uint64_t newval) | |
201 | { | |
3558fd73 BB |
202 | zfs_sb_t *zsb = arg; |
203 | struct super_block *sb = zsb->z_sb; | |
34dc7c2f | 204 | |
2cf7f52b BB |
205 | if (sb == NULL) |
206 | return; | |
207 | ||
208 | if (newval) | |
3558fd73 | 209 | sb->s_flags |= MS_RDONLY; |
2cf7f52b | 210 | else |
3558fd73 | 211 | sb->s_flags &= ~MS_RDONLY; |
34dc7c2f BB |
212 | } |
213 | ||
214 | static void | |
215 | devices_changed_cb(void *arg, uint64_t newval) | |
216 | { | |
34dc7c2f BB |
217 | } |
218 | ||
219 | static void | |
220 | setuid_changed_cb(void *arg, uint64_t newval) | |
221 | { | |
34dc7c2f BB |
222 | } |
223 | ||
224 | static void | |
225 | exec_changed_cb(void *arg, uint64_t newval) | |
226 | { | |
34dc7c2f BB |
227 | } |
228 | ||
34dc7c2f BB |
229 | static void |
230 | nbmand_changed_cb(void *arg, uint64_t newval) | |
231 | { | |
3558fd73 BB |
232 | zfs_sb_t *zsb = arg; |
233 | struct super_block *sb = zsb->z_sb; | |
234 | ||
2cf7f52b BB |
235 | if (sb == NULL) |
236 | return; | |
237 | ||
238 | if (newval == TRUE) | |
3558fd73 | 239 | sb->s_flags |= MS_MANDLOCK; |
2cf7f52b | 240 | else |
3558fd73 | 241 | sb->s_flags &= ~MS_MANDLOCK; |
34dc7c2f BB |
242 | } |
243 | ||
244 | static void | |
245 | snapdir_changed_cb(void *arg, uint64_t newval) | |
246 | { | |
3558fd73 | 247 | ((zfs_sb_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
248 | } |
249 | ||
250 | static void | |
251 | vscan_changed_cb(void *arg, uint64_t newval) | |
252 | { | |
3558fd73 | 253 | ((zfs_sb_t *)arg)->z_vscan = newval; |
34dc7c2f BB |
254 | } |
255 | ||
34dc7c2f BB |
256 | static void |
257 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
258 | { | |
3558fd73 | 259 | ((zfs_sb_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
260 | } |
261 | ||
e5c39b95 | 262 | int |
3558fd73 | 263 | zfs_register_callbacks(zfs_sb_t *zsb) |
34dc7c2f BB |
264 | { |
265 | struct dsl_dataset *ds = NULL; | |
3558fd73 | 266 | objset_t *os = zsb->z_os; |
c9ada6d5 | 267 | boolean_t do_readonly = B_FALSE; |
34dc7c2f BB |
268 | int error = 0; |
269 | ||
2cf7f52b | 270 | if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) |
c9ada6d5 | 271 | do_readonly = B_TRUE; |
34dc7c2f BB |
272 | |
273 | /* | |
274 | * Register property callbacks. | |
275 | * | |
276 | * It would probably be fine to just check for i/o error from | |
277 | * the first prop_register(), but I guess I like to go | |
278 | * overboard... | |
279 | */ | |
280 | ds = dmu_objset_ds(os); | |
13fe0198 | 281 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
3558fd73 | 282 | error = dsl_prop_register(ds, |
13fe0198 | 283 | zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb); |
98fad862 | 284 | error = error ? error : dsl_prop_register(ds, |
6d111134 | 285 | zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb); |
34dc7c2f | 286 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 287 | zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb); |
34dc7c2f | 288 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 289 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb); |
34dc7c2f | 290 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 291 | zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb); |
34dc7c2f | 292 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 293 | zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb); |
34dc7c2f | 294 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 295 | zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb); |
34dc7c2f | 296 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 297 | zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb); |
34dc7c2f | 298 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 299 | zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb); |
34dc7c2f | 300 | error = error ? error : dsl_prop_register(ds, |
023699cd MM |
301 | zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb); |
302 | error = error ? error : dsl_prop_register(ds, | |
303 | zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb); | |
34dc7c2f | 304 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 305 | zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb); |
2cf7f52b | 306 | error = error ? error : dsl_prop_register(ds, |
13fe0198 MA |
307 | zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb); |
308 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
34dc7c2f BB |
309 | if (error) |
310 | goto unregister; | |
311 | ||
c9ada6d5 BB |
312 | if (do_readonly) |
313 | readonly_changed_cb(zsb, B_TRUE); | |
314 | ||
34dc7c2f BB |
315 | return (0); |
316 | ||
317 | unregister: | |
318 | /* | |
319 | * We may attempt to unregister some callbacks that are not | |
320 | * registered, but this is OK; it will simply return ENOMSG, | |
321 | * which we will ignore. | |
322 | */ | |
13fe0198 MA |
323 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME), |
324 | atime_changed_cb, zsb); | |
6d111134 TC |
325 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RELATIME), |
326 | relatime_changed_cb, zsb); | |
13fe0198 MA |
327 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR), |
328 | xattr_changed_cb, zsb); | |
329 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), | |
330 | blksz_changed_cb, zsb); | |
331 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY), | |
332 | readonly_changed_cb, zsb); | |
333 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES), | |
334 | devices_changed_cb, zsb); | |
335 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID), | |
336 | setuid_changed_cb, zsb); | |
337 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC), | |
338 | exec_changed_cb, zsb); | |
339 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR), | |
340 | snapdir_changed_cb, zsb); | |
023699cd MM |
341 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE), |
342 | acltype_changed_cb, zsb); | |
13fe0198 MA |
343 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT), |
344 | acl_inherit_changed_cb, zsb); | |
345 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN), | |
346 | vscan_changed_cb, zsb); | |
347 | (void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND), | |
348 | nbmand_changed_cb, zsb); | |
34dc7c2f | 349 | |
3558fd73 | 350 | return (error); |
34dc7c2f | 351 | } |
e5c39b95 | 352 | EXPORT_SYMBOL(zfs_register_callbacks); |
34dc7c2f | 353 | |
428870ff BB |
354 | static int |
355 | zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, | |
356 | uint64_t *userp, uint64_t *groupp) | |
9babb374 | 357 | { |
428870ff BB |
358 | /* |
359 | * Is it a valid type of object to track? | |
360 | */ | |
361 | if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) | |
2e528b49 | 362 | return (SET_ERROR(ENOENT)); |
9babb374 | 363 | |
428870ff BB |
364 | /* |
365 | * If we have a NULL data pointer | |
366 | * then assume the id's aren't changing and | |
367 | * return EEXIST to the dmu to let it know to | |
368 | * use the same ids | |
369 | */ | |
370 | if (data == NULL) | |
2e528b49 | 371 | return (SET_ERROR(EEXIST)); |
9babb374 | 372 | |
428870ff | 373 | if (bonustype == DMU_OT_ZNODE) { |
a94addd9 | 374 | znode_phys_t *znp = data; |
428870ff BB |
375 | *userp = znp->zp_uid; |
376 | *groupp = znp->zp_gid; | |
9babb374 | 377 | } else { |
428870ff | 378 | int hdrsize; |
a94addd9 MA |
379 | sa_hdr_phys_t *sap = data; |
380 | sa_hdr_phys_t sa = *sap; | |
381 | boolean_t swap = B_FALSE; | |
9babb374 | 382 | |
428870ff | 383 | ASSERT(bonustype == DMU_OT_SA); |
428870ff | 384 | |
a94addd9 | 385 | if (sa.sa_magic == 0) { |
428870ff BB |
386 | /* |
387 | * This should only happen for newly created | |
388 | * files that haven't had the znode data filled | |
389 | * in yet. | |
390 | */ | |
391 | *userp = 0; | |
392 | *groupp = 0; | |
a94addd9 MA |
393 | return (0); |
394 | } | |
395 | if (sa.sa_magic == BSWAP_32(SA_MAGIC)) { | |
396 | sa.sa_magic = SA_MAGIC; | |
397 | sa.sa_layout_info = BSWAP_16(sa.sa_layout_info); | |
398 | swap = B_TRUE; | |
399 | } else { | |
400 | VERIFY3U(sa.sa_magic, ==, SA_MAGIC); | |
401 | } | |
402 | ||
403 | hdrsize = sa_hdrsize(&sa); | |
404 | VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t)); | |
405 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
406 | SA_UID_OFFSET)); | |
407 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
408 | SA_GID_OFFSET)); | |
409 | if (swap) { | |
410 | *userp = BSWAP_64(*userp); | |
411 | *groupp = BSWAP_64(*groupp); | |
428870ff | 412 | } |
9babb374 | 413 | } |
13fe0198 | 414 | return (0); |
9babb374 BB |
415 | } |
416 | ||
417 | static void | |
3558fd73 | 418 | fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr, |
9babb374 BB |
419 | char *domainbuf, int buflen, uid_t *ridp) |
420 | { | |
9babb374 BB |
421 | uint64_t fuid; |
422 | const char *domain; | |
423 | ||
424 | fuid = strtonum(fuidstr, NULL); | |
425 | ||
3558fd73 | 426 | domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid)); |
9babb374 BB |
427 | if (domain) |
428 | (void) strlcpy(domainbuf, domain, buflen); | |
429 | else | |
430 | domainbuf[0] = '\0'; | |
431 | *ridp = FUID_RID(fuid); | |
432 | } | |
433 | ||
434 | static uint64_t | |
3558fd73 | 435 | zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type) |
9babb374 BB |
436 | { |
437 | switch (type) { | |
438 | case ZFS_PROP_USERUSED: | |
439 | return (DMU_USERUSED_OBJECT); | |
440 | case ZFS_PROP_GROUPUSED: | |
441 | return (DMU_GROUPUSED_OBJECT); | |
442 | case ZFS_PROP_USERQUOTA: | |
3558fd73 | 443 | return (zsb->z_userquota_obj); |
9babb374 | 444 | case ZFS_PROP_GROUPQUOTA: |
3558fd73 | 445 | return (zsb->z_groupquota_obj); |
149e873a | 446 | default: |
2e528b49 | 447 | return (SET_ERROR(ENOTSUP)); |
9babb374 BB |
448 | } |
449 | return (0); | |
450 | } | |
451 | ||
452 | int | |
3558fd73 | 453 | zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
454 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
455 | { | |
456 | int error; | |
457 | zap_cursor_t zc; | |
458 | zap_attribute_t za; | |
459 | zfs_useracct_t *buf = vbuf; | |
460 | uint64_t obj; | |
461 | ||
3558fd73 | 462 | if (!dmu_objset_userspace_present(zsb->z_os)) |
2e528b49 | 463 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 464 | |
3558fd73 | 465 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
466 | if (obj == 0) { |
467 | *bufsizep = 0; | |
468 | return (0); | |
469 | } | |
470 | ||
3558fd73 | 471 | for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep); |
9babb374 BB |
472 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
473 | zap_cursor_advance(&zc)) { | |
474 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
475 | *bufsizep) | |
476 | break; | |
477 | ||
3558fd73 | 478 | fuidstr_to_sid(zsb, za.za_name, |
9babb374 BB |
479 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
480 | ||
481 | buf->zu_space = za.za_first_integer; | |
482 | buf++; | |
483 | } | |
484 | if (error == ENOENT) | |
485 | error = 0; | |
486 | ||
487 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
488 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
489 | *cookiep = zap_cursor_serialize(&zc); | |
490 | zap_cursor_fini(&zc); | |
491 | return (error); | |
492 | } | |
e5c39b95 | 493 | EXPORT_SYMBOL(zfs_userspace_many); |
9babb374 BB |
494 | |
495 | /* | |
496 | * buf must be big enough (eg, 32 bytes) | |
497 | */ | |
498 | static int | |
3558fd73 | 499 | id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid, |
9babb374 BB |
500 | char *buf, boolean_t addok) |
501 | { | |
502 | uint64_t fuid; | |
503 | int domainid = 0; | |
504 | ||
505 | if (domain && domain[0]) { | |
3558fd73 | 506 | domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok); |
9babb374 | 507 | if (domainid == -1) |
2e528b49 | 508 | return (SET_ERROR(ENOENT)); |
9babb374 BB |
509 | } |
510 | fuid = FUID_ENCODE(domainid, rid); | |
511 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
512 | return (0); | |
513 | } | |
514 | ||
515 | int | |
3558fd73 | 516 | zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
517 | const char *domain, uint64_t rid, uint64_t *valp) |
518 | { | |
519 | char buf[32]; | |
520 | int err; | |
521 | uint64_t obj; | |
522 | ||
523 | *valp = 0; | |
524 | ||
3558fd73 | 525 | if (!dmu_objset_userspace_present(zsb->z_os)) |
2e528b49 | 526 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 527 | |
3558fd73 | 528 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
529 | if (obj == 0) |
530 | return (0); | |
531 | ||
3558fd73 | 532 | err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE); |
9babb374 BB |
533 | if (err) |
534 | return (err); | |
535 | ||
3558fd73 | 536 | err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
537 | if (err == ENOENT) |
538 | err = 0; | |
539 | return (err); | |
540 | } | |
e5c39b95 | 541 | EXPORT_SYMBOL(zfs_userspace_one); |
9babb374 BB |
542 | |
543 | int | |
3558fd73 | 544 | zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
545 | const char *domain, uint64_t rid, uint64_t quota) |
546 | { | |
547 | char buf[32]; | |
548 | int err; | |
549 | dmu_tx_t *tx; | |
550 | uint64_t *objp; | |
551 | boolean_t fuid_dirtied; | |
552 | ||
553 | if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) | |
2e528b49 | 554 | return (SET_ERROR(EINVAL)); |
9babb374 | 555 | |
3558fd73 | 556 | if (zsb->z_version < ZPL_VERSION_USERSPACE) |
2e528b49 | 557 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 558 | |
3558fd73 BB |
559 | objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj : |
560 | &zsb->z_groupquota_obj; | |
9babb374 | 561 | |
3558fd73 | 562 | err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE); |
9babb374 BB |
563 | if (err) |
564 | return (err); | |
3558fd73 | 565 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 566 | |
3558fd73 | 567 | tx = dmu_tx_create(zsb->z_os); |
9babb374 BB |
568 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
569 | if (*objp == 0) { | |
570 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
571 | zfs_userquota_prop_prefixes[type]); | |
572 | } | |
573 | if (fuid_dirtied) | |
3558fd73 | 574 | zfs_fuid_txhold(zsb, tx); |
9babb374 BB |
575 | err = dmu_tx_assign(tx, TXG_WAIT); |
576 | if (err) { | |
577 | dmu_tx_abort(tx); | |
578 | return (err); | |
579 | } | |
580 | ||
3558fd73 | 581 | mutex_enter(&zsb->z_lock); |
9babb374 | 582 | if (*objp == 0) { |
3558fd73 | 583 | *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 584 | DMU_OT_NONE, 0, tx); |
3558fd73 | 585 | VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
586 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
587 | } | |
3558fd73 | 588 | mutex_exit(&zsb->z_lock); |
9babb374 BB |
589 | |
590 | if (quota == 0) { | |
3558fd73 | 591 | err = zap_remove(zsb->z_os, *objp, buf, tx); |
9babb374 BB |
592 | if (err == ENOENT) |
593 | err = 0; | |
594 | } else { | |
3558fd73 | 595 | err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
596 | } |
597 | ASSERT(err == 0); | |
598 | if (fuid_dirtied) | |
3558fd73 | 599 | zfs_fuid_sync(zsb, tx); |
9babb374 BB |
600 | dmu_tx_commit(tx); |
601 | return (err); | |
602 | } | |
e5c39b95 | 603 | EXPORT_SYMBOL(zfs_set_userquota); |
9babb374 BB |
604 | |
605 | boolean_t | |
3558fd73 | 606 | zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid) |
9babb374 BB |
607 | { |
608 | char buf[32]; | |
609 | uint64_t used, quota, usedobj, quotaobj; | |
610 | int err; | |
611 | ||
612 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
3558fd73 | 613 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
9babb374 | 614 | |
3558fd73 | 615 | if (quotaobj == 0 || zsb->z_replay) |
9babb374 BB |
616 | return (B_FALSE); |
617 | ||
618 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
3558fd73 | 619 | err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
620 | if (err != 0) |
621 | return (B_FALSE); | |
622 | ||
3558fd73 | 623 | err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
624 | if (err != 0) |
625 | return (B_FALSE); | |
626 | return (used >= quota); | |
627 | } | |
e5c39b95 | 628 | EXPORT_SYMBOL(zfs_fuid_overquota); |
9babb374 | 629 | |
428870ff | 630 | boolean_t |
3558fd73 | 631 | zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup) |
428870ff BB |
632 | { |
633 | uint64_t fuid; | |
634 | uint64_t quotaobj; | |
428870ff | 635 | |
3558fd73 | 636 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
428870ff | 637 | |
572e2857 | 638 | fuid = isgroup ? zp->z_gid : zp->z_uid; |
428870ff | 639 | |
3558fd73 | 640 | if (quotaobj == 0 || zsb->z_replay) |
428870ff BB |
641 | return (B_FALSE); |
642 | ||
3558fd73 | 643 | return (zfs_fuid_overquota(zsb, isgroup, fuid)); |
428870ff | 644 | } |
e5c39b95 | 645 | EXPORT_SYMBOL(zfs_owner_overquota); |
428870ff | 646 | |
9babb374 | 647 | int |
3558fd73 | 648 | zfs_sb_create(const char *osname, zfs_sb_t **zsbp) |
9babb374 BB |
649 | { |
650 | objset_t *os; | |
3558fd73 | 651 | zfs_sb_t *zsb; |
9babb374 BB |
652 | uint64_t zval; |
653 | int i, error; | |
428870ff | 654 | uint64_t sa_obj; |
9babb374 | 655 | |
efcd79a8 | 656 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
9babb374 | 657 | |
428870ff BB |
658 | /* |
659 | * We claim to always be readonly so we can open snapshots; | |
660 | * other ZPL code will prevent us from writing to snapshots. | |
661 | */ | |
3558fd73 | 662 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); |
428870ff | 663 | if (error) { |
3558fd73 | 664 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 | 665 | return (error); |
428870ff | 666 | } |
9babb374 BB |
667 | |
668 | /* | |
669 | * Initialize the zfs-specific filesystem structure. | |
670 | * Should probably make this a kmem cache, shuffle fields, | |
671 | * and just bzero up to z_hold_mtx[]. | |
672 | */ | |
2cf7f52b | 673 | zsb->z_sb = NULL; |
3558fd73 BB |
674 | zsb->z_parent = zsb; |
675 | zsb->z_max_blksz = SPA_MAXBLOCKSIZE; | |
676 | zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; | |
677 | zsb->z_os = os; | |
9babb374 | 678 | |
3558fd73 | 679 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); |
9babb374 BB |
680 | if (error) { |
681 | goto out; | |
287be44f | 682 | } else if (zsb->z_version > ZPL_VERSION) { |
2e528b49 | 683 | error = SET_ERROR(ENOTSUP); |
9babb374 BB |
684 | goto out; |
685 | } | |
9babb374 BB |
686 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) |
687 | goto out; | |
3558fd73 | 688 | zsb->z_norm = (int)zval; |
9babb374 BB |
689 | |
690 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
691 | goto out; | |
3558fd73 | 692 | zsb->z_utf8 = (zval != 0); |
9babb374 BB |
693 | |
694 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
695 | goto out; | |
3558fd73 | 696 | zsb->z_case = (uint_t)zval; |
9babb374 | 697 | |
023699cd MM |
698 | if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0) |
699 | goto out; | |
700 | zsb->z_acl_type = (uint_t)zval; | |
701 | ||
9babb374 BB |
702 | /* |
703 | * Fold case on file systems that are always or sometimes case | |
704 | * insensitive. | |
705 | */ | |
3558fd73 BB |
706 | if (zsb->z_case == ZFS_CASE_INSENSITIVE || |
707 | zsb->z_case == ZFS_CASE_MIXED) | |
708 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 709 | |
3558fd73 BB |
710 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
711 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
428870ff | 712 | |
3558fd73 | 713 | if (zsb->z_use_sa) { |
428870ff BB |
714 | /* should either have both of these objects or none */ |
715 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
716 | &sa_obj); | |
717 | if (error) | |
591fb62f | 718 | goto out; |
82a37189 BB |
719 | |
720 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval); | |
721 | if ((error == 0) && (zval == ZFS_XATTR_SA)) | |
722 | zsb->z_xattr_sa = B_TRUE; | |
428870ff BB |
723 | } else { |
724 | /* | |
725 | * Pre SA versions file systems should never touch | |
726 | * either the attribute registration or layout objects. | |
727 | */ | |
728 | sa_obj = 0; | |
729 | } | |
730 | ||
572e2857 | 731 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
3558fd73 | 732 | &zsb->z_attr_table); |
572e2857 BB |
733 | if (error) |
734 | goto out; | |
428870ff | 735 | |
3558fd73 | 736 | if (zsb->z_version >= ZPL_VERSION_SA) |
428870ff | 737 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
738 | |
739 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
3558fd73 | 740 | &zsb->z_root); |
9babb374 BB |
741 | if (error) |
742 | goto out; | |
3558fd73 | 743 | ASSERT(zsb->z_root != 0); |
9babb374 BB |
744 | |
745 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
3558fd73 | 746 | &zsb->z_unlinkedobj); |
9babb374 BB |
747 | if (error) |
748 | goto out; | |
749 | ||
750 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
751 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
3558fd73 | 752 | 8, 1, &zsb->z_userquota_obj); |
9babb374 BB |
753 | if (error && error != ENOENT) |
754 | goto out; | |
755 | ||
756 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
757 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
3558fd73 | 758 | 8, 1, &zsb->z_groupquota_obj); |
9babb374 BB |
759 | if (error && error != ENOENT) |
760 | goto out; | |
761 | ||
762 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
3558fd73 | 763 | &zsb->z_fuid_obj); |
9babb374 BB |
764 | if (error && error != ENOENT) |
765 | goto out; | |
766 | ||
767 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
3558fd73 | 768 | &zsb->z_shares_dir); |
9babb374 BB |
769 | if (error && error != ENOENT) |
770 | goto out; | |
771 | ||
3558fd73 BB |
772 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
773 | mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
774 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
9babb374 | 775 | offsetof(znode_t, z_link_node)); |
13fe0198 | 776 | rrw_init(&zsb->z_teardown_lock, B_FALSE); |
3558fd73 BB |
777 | rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); |
778 | rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
9babb374 | 779 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 780 | mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); |
9babb374 | 781 | |
ebe7e575 BB |
782 | avl_create(&zsb->z_ctldir_snaps, snapentry_compare, |
783 | sizeof (zfs_snapentry_t), offsetof(zfs_snapentry_t, se_node)); | |
784 | mutex_init(&zsb->z_ctldir_lock, NULL, MUTEX_DEFAULT, NULL); | |
785 | ||
3558fd73 | 786 | *zsbp = zsb; |
9babb374 BB |
787 | return (0); |
788 | ||
789 | out: | |
3558fd73 BB |
790 | dmu_objset_disown(os, zsb); |
791 | *zsbp = NULL; | |
792 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
9babb374 BB |
793 | return (error); |
794 | } | |
86f35f34 | 795 | EXPORT_SYMBOL(zfs_sb_create); |
9babb374 | 796 | |
86f35f34 | 797 | int |
3558fd73 | 798 | zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) |
34dc7c2f | 799 | { |
34dc7c2f BB |
800 | int error; |
801 | ||
3558fd73 | 802 | error = zfs_register_callbacks(zsb); |
34dc7c2f BB |
803 | if (error) |
804 | return (error); | |
805 | ||
806 | /* | |
3558fd73 | 807 | * Set the objset user_ptr to track its zsb. |
34dc7c2f | 808 | */ |
3558fd73 BB |
809 | mutex_enter(&zsb->z_os->os_user_ptr_lock); |
810 | dmu_objset_set_user(zsb->z_os, zsb); | |
811 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
34dc7c2f | 812 | |
3558fd73 | 813 | zsb->z_log = zil_open(zsb->z_os, zfs_get_data); |
9babb374 | 814 | |
34dc7c2f BB |
815 | /* |
816 | * If we are not mounting (ie: online recv), then we don't | |
817 | * have to worry about replaying the log as we blocked all | |
818 | * operations out since we closed the ZIL. | |
819 | */ | |
820 | if (mounting) { | |
b128c09f BB |
821 | boolean_t readonly; |
822 | ||
34dc7c2f BB |
823 | /* |
824 | * During replay we remove the read only flag to | |
825 | * allow replays to succeed. | |
826 | */ | |
2cf7f52b | 827 | readonly = zfs_is_readonly(zsb); |
fb5f0bc8 | 828 | if (readonly != 0) |
2cf7f52b | 829 | readonly_changed_cb(zsb, B_FALSE); |
fb5f0bc8 | 830 | else |
3558fd73 | 831 | zfs_unlinked_drain(zsb); |
34dc7c2f | 832 | |
428870ff BB |
833 | /* |
834 | * Parse and replay the intent log. | |
835 | * | |
836 | * Because of ziltest, this must be done after | |
837 | * zfs_unlinked_drain(). (Further note: ziltest | |
838 | * doesn't use readonly mounts, where | |
839 | * zfs_unlinked_drain() isn't called.) This is because | |
840 | * ziltest causes spa_sync() to think it's committed, | |
841 | * but actually it is not, so the intent log contains | |
842 | * many txg's worth of changes. | |
843 | * | |
844 | * In particular, if object N is in the unlinked set in | |
845 | * the last txg to actually sync, then it could be | |
846 | * actually freed in a later txg and then reallocated | |
847 | * in a yet later txg. This would write a "create | |
848 | * object N" record to the intent log. Normally, this | |
849 | * would be fine because the spa_sync() would have | |
850 | * written out the fact that object N is free, before | |
851 | * we could write the "create object N" intent log | |
852 | * record. | |
853 | * | |
854 | * But when we are in ziltest mode, we advance the "open | |
855 | * txg" without actually spa_sync()-ing the changes to | |
856 | * disk. So we would see that object N is still | |
857 | * allocated and in the unlinked set, and there is an | |
858 | * intent log record saying to allocate it. | |
859 | */ | |
3558fd73 | 860 | if (spa_writeable(dmu_objset_spa(zsb->z_os))) { |
572e2857 | 861 | if (zil_replay_disable) { |
3558fd73 | 862 | zil_destroy(zsb->z_log, B_FALSE); |
572e2857 | 863 | } else { |
3558fd73 BB |
864 | zsb->z_replay = B_TRUE; |
865 | zil_replay(zsb->z_os, zsb, | |
572e2857 | 866 | zfs_replay_vector); |
3558fd73 | 867 | zsb->z_replay = B_FALSE; |
572e2857 | 868 | } |
fb5f0bc8 | 869 | } |
2cf7f52b BB |
870 | |
871 | /* restore readonly bit */ | |
872 | if (readonly != 0) | |
873 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
874 | } |
875 | ||
34dc7c2f BB |
876 | return (0); |
877 | } | |
86f35f34 | 878 | EXPORT_SYMBOL(zfs_sb_setup); |
34dc7c2f | 879 | |
9babb374 | 880 | void |
3558fd73 | 881 | zfs_sb_free(zfs_sb_t *zsb) |
34dc7c2f | 882 | { |
9babb374 | 883 | int i; |
9babb374 | 884 | |
3558fd73 | 885 | zfs_fuid_destroy(zsb); |
9babb374 | 886 | |
3558fd73 BB |
887 | mutex_destroy(&zsb->z_znodes_lock); |
888 | mutex_destroy(&zsb->z_lock); | |
889 | list_destroy(&zsb->z_all_znodes); | |
890 | rrw_destroy(&zsb->z_teardown_lock); | |
891 | rw_destroy(&zsb->z_teardown_inactive_lock); | |
892 | rw_destroy(&zsb->z_fuid_lock); | |
9babb374 | 893 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 894 | mutex_destroy(&zsb->z_hold_mtx[i]); |
ebe7e575 BB |
895 | mutex_destroy(&zsb->z_ctldir_lock); |
896 | avl_destroy(&zsb->z_ctldir_snaps); | |
3558fd73 | 897 | kmem_free(zsb, sizeof (zfs_sb_t)); |
34dc7c2f | 898 | } |
86f35f34 | 899 | EXPORT_SYMBOL(zfs_sb_free); |
34dc7c2f | 900 | |
9babb374 | 901 | static void |
3558fd73 | 902 | zfs_set_fuid_feature(zfs_sb_t *zsb) |
9babb374 | 903 | { |
3558fd73 BB |
904 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
905 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
9babb374 | 906 | } |
34dc7c2f BB |
907 | |
908 | void | |
3558fd73 | 909 | zfs_unregister_callbacks(zfs_sb_t *zsb) |
34dc7c2f | 910 | { |
3558fd73 | 911 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
912 | struct dsl_dataset *ds; |
913 | ||
914 | /* | |
915 | * Unregister properties. | |
916 | */ | |
917 | if (!dmu_objset_is_snapshot(os)) { | |
918 | ds = dmu_objset_ds(os); | |
919 | VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, | |
3558fd73 | 920 | zsb) == 0); |
34dc7c2f | 921 | |
6d111134 TC |
922 | VERIFY(dsl_prop_unregister(ds, "relatime", relatime_changed_cb, |
923 | zsb) == 0); | |
924 | ||
34dc7c2f | 925 | VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, |
3558fd73 | 926 | zsb) == 0); |
34dc7c2f BB |
927 | |
928 | VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, | |
3558fd73 | 929 | zsb) == 0); |
34dc7c2f BB |
930 | |
931 | VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, | |
3558fd73 | 932 | zsb) == 0); |
34dc7c2f BB |
933 | |
934 | VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, | |
3558fd73 | 935 | zsb) == 0); |
34dc7c2f BB |
936 | |
937 | VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, | |
3558fd73 | 938 | zsb) == 0); |
34dc7c2f BB |
939 | |
940 | VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, | |
3558fd73 | 941 | zsb) == 0); |
34dc7c2f BB |
942 | |
943 | VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, | |
3558fd73 | 944 | zsb) == 0); |
34dc7c2f | 945 | |
023699cd MM |
946 | VERIFY(dsl_prop_unregister(ds, "acltype", acltype_changed_cb, |
947 | zsb) == 0); | |
948 | ||
34dc7c2f | 949 | VERIFY(dsl_prop_unregister(ds, "aclinherit", |
3558fd73 | 950 | acl_inherit_changed_cb, zsb) == 0); |
34dc7c2f BB |
951 | |
952 | VERIFY(dsl_prop_unregister(ds, "vscan", | |
3558fd73 | 953 | vscan_changed_cb, zsb) == 0); |
2cf7f52b BB |
954 | |
955 | VERIFY(dsl_prop_unregister(ds, "nbmand", | |
956 | nbmand_changed_cb, zsb) == 0); | |
34dc7c2f BB |
957 | } |
958 | } | |
e5c39b95 | 959 | EXPORT_SYMBOL(zfs_unregister_callbacks); |
34dc7c2f | 960 | |
bc3e15e3 | 961 | #ifdef HAVE_MLSLABEL |
428870ff | 962 | /* |
d3cc8b15 WA |
963 | * Check that the hex label string is appropriate for the dataset being |
964 | * mounted into the global_zone proper. | |
428870ff | 965 | * |
d3cc8b15 WA |
966 | * Return an error if the hex label string is not default or |
967 | * admin_low/admin_high. For admin_low labels, the corresponding | |
968 | * dataset must be readonly. | |
428870ff BB |
969 | */ |
970 | int | |
971 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
972 | { | |
973 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
974 | return (0); | |
975 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
976 | return (0); | |
977 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
978 | /* must be readonly */ | |
979 | uint64_t rdonly; | |
980 | ||
981 | if (dsl_prop_get_integer(dsname, | |
982 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
2e528b49 | 983 | return (SET_ERROR(EACCES)); |
428870ff BB |
984 | return (rdonly ? 0 : EACCES); |
985 | } | |
2e528b49 | 986 | return (SET_ERROR(EACCES)); |
428870ff | 987 | } |
86f35f34 | 988 | EXPORT_SYMBOL(zfs_check_global_label); |
bc3e15e3 | 989 | #endif /* HAVE_MLSLABEL */ |
428870ff | 990 | |
e5c39b95 | 991 | int |
3558fd73 | 992 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 993 | { |
3558fd73 | 994 | zfs_sb_t *zsb = dentry->d_sb->s_fs_info; |
34dc7c2f | 995 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
04f9432d | 996 | uint64_t fsid; |
3558fd73 | 997 | uint32_t bshift; |
34dc7c2f | 998 | |
3558fd73 | 999 | ZFS_ENTER(zsb); |
34dc7c2f | 1000 | |
3558fd73 | 1001 | dmu_objset_space(zsb->z_os, |
34dc7c2f BB |
1002 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
1003 | ||
04f9432d | 1004 | fsid = dmu_objset_fsid_guid(zsb->z_os); |
34dc7c2f | 1005 | /* |
05ff35c6 BB |
1006 | * The underlying storage pool actually uses multiple block |
1007 | * size. Under Solaris frsize (fragment size) is reported as | |
1008 | * the smallest block size we support, and bsize (block size) | |
1009 | * as the filesystem's maximum block size. Unfortunately, | |
1010 | * under Linux the fragment size and block size are often used | |
1011 | * interchangeably. Thus we are forced to report both of them | |
1012 | * as the filesystem's maximum block size. | |
34dc7c2f | 1013 | */ |
05ff35c6 | 1014 | statp->f_frsize = zsb->z_max_blksz; |
3558fd73 BB |
1015 | statp->f_bsize = zsb->z_max_blksz; |
1016 | bshift = fls(statp->f_bsize) - 1; | |
34dc7c2f BB |
1017 | |
1018 | /* | |
3558fd73 BB |
1019 | * The following report "total" blocks of various kinds in |
1020 | * the file system, but reported in terms of f_bsize - the | |
1021 | * "preferred" size. | |
34dc7c2f BB |
1022 | */ |
1023 | ||
3558fd73 BB |
1024 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
1025 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
1026 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
1027 | ||
1028 | /* | |
1029 | * statvfs() should really be called statufs(), because it assumes | |
1030 | * static metadata. ZFS doesn't preallocate files, so the best | |
1031 | * we can do is report the max that could possibly fit in f_files, | |
1032 | * and that minus the number actually used in f_ffree. | |
1033 | * For f_ffree, report the smaller of the number of object available | |
1034 | * and the number of blocks (each object will take at least a block). | |
1035 | */ | |
baab0630 | 1036 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 1037 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
1038 | statp->f_fsid.val[0] = (uint32_t)fsid; |
1039 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 BB |
1040 | statp->f_type = ZFS_SUPER_MAGIC; |
1041 | statp->f_namelen = ZFS_MAXNAMELEN; | |
34dc7c2f BB |
1042 | |
1043 | /* | |
3558fd73 | 1044 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
1045 | * Is there anything useful we could/should provide? |
1046 | */ | |
3558fd73 | 1047 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 1048 | |
3558fd73 | 1049 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1050 | return (0); |
1051 | } | |
e5c39b95 | 1052 | EXPORT_SYMBOL(zfs_statvfs); |
34dc7c2f | 1053 | |
e5c39b95 | 1054 | int |
3558fd73 | 1055 | zfs_root(zfs_sb_t *zsb, struct inode **ipp) |
34dc7c2f | 1056 | { |
34dc7c2f BB |
1057 | znode_t *rootzp; |
1058 | int error; | |
1059 | ||
3558fd73 | 1060 | ZFS_ENTER(zsb); |
34dc7c2f | 1061 | |
3558fd73 | 1062 | error = zfs_zget(zsb, zsb->z_root, &rootzp); |
34dc7c2f | 1063 | if (error == 0) |
3558fd73 | 1064 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1065 | |
3558fd73 | 1066 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1067 | return (error); |
1068 | } | |
e5c39b95 | 1069 | EXPORT_SYMBOL(zfs_root); |
34dc7c2f | 1070 | |
ab26409d BB |
1071 | #ifdef HAVE_SHRINK |
1072 | int | |
1073 | zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) | |
1074 | { | |
1075 | zfs_sb_t *zsb = sb->s_fs_info; | |
1076 | struct shrinker *shrinker = &sb->s_shrink; | |
1077 | struct shrink_control sc = { | |
1078 | .nr_to_scan = nr_to_scan, | |
1079 | .gfp_mask = GFP_KERNEL, | |
1080 | }; | |
1081 | ||
1082 | ZFS_ENTER(zsb); | |
1083 | *objects = (*shrinker->shrink)(shrinker, &sc); | |
1084 | ZFS_EXIT(zsb); | |
1085 | ||
1086 | return (0); | |
1087 | } | |
1088 | EXPORT_SYMBOL(zfs_sb_prune); | |
1089 | #endif /* HAVE_SHRINK */ | |
1090 | ||
34dc7c2f | 1091 | /* |
7b3e34ba | 1092 | * Teardown the zfs_sb_t. |
34dc7c2f BB |
1093 | * |
1094 | * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' | |
1095 | * and 'z_teardown_inactive_lock' held. | |
1096 | */ | |
3558fd73 | 1097 | int |
86f35f34 | 1098 | zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting) |
34dc7c2f BB |
1099 | { |
1100 | znode_t *zp; | |
1101 | ||
fd23720a BB |
1102 | /* |
1103 | * If someone has not already unmounted this file system, | |
1104 | * drain the iput_taskq to ensure all active references to the | |
1105 | * zfs_sb_t have been handled only then can it be safely destroyed. | |
1106 | */ | |
1107 | if (zsb->z_os) | |
1108 | taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os))); | |
1109 | ||
3558fd73 | 1110 | rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1111 | |
1112 | if (!unmounting) { | |
1113 | /* | |
ceb43b93 BB |
1114 | * We purge the parent filesystem's super block as the |
1115 | * parent filesystem and all of its snapshots have their | |
1116 | * inode's super block set to the parent's filesystem's | |
1117 | * super block. Note, 'z_parent' is self referential | |
1118 | * for non-snapshots. | |
34dc7c2f | 1119 | */ |
ceb43b93 | 1120 | shrink_dcache_sb(zsb->z_parent->z_sb); |
34dc7c2f BB |
1121 | } |
1122 | ||
1123 | /* | |
1124 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1125 | * threads are blocked as zil_close can call zfs_inactive. | |
1126 | */ | |
3558fd73 BB |
1127 | if (zsb->z_log) { |
1128 | zil_close(zsb->z_log); | |
1129 | zsb->z_log = NULL; | |
34dc7c2f BB |
1130 | } |
1131 | ||
3558fd73 | 1132 | rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1133 | |
1134 | /* | |
1135 | * If we are not unmounting (ie: online recv) and someone already | |
1136 | * unmounted this file system while we were doing the switcheroo, | |
1137 | * or a reopen of z_os failed then just bail out now. | |
1138 | */ | |
3558fd73 BB |
1139 | if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { |
1140 | rw_exit(&zsb->z_teardown_inactive_lock); | |
1141 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
2e528b49 | 1142 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1143 | } |
1144 | ||
1145 | /* | |
7b3e34ba BB |
1146 | * At this point there are no VFS ops active, and any new VFS ops |
1147 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1148 | * relevant for forced unmount). | |
34dc7c2f BB |
1149 | * |
1150 | * Release all holds on dbufs. | |
1151 | */ | |
3558fd73 BB |
1152 | mutex_enter(&zsb->z_znodes_lock); |
1153 | for (zp = list_head(&zsb->z_all_znodes); zp != NULL; | |
7b3e34ba | 1154 | zp = list_next(&zsb->z_all_znodes, zp)) { |
2e40f094 | 1155 | if (zp->z_sa_hdl) |
34dc7c2f | 1156 | zfs_znode_dmu_fini(zp); |
7b3e34ba | 1157 | } |
3558fd73 | 1158 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f BB |
1159 | |
1160 | /* | |
7b3e34ba | 1161 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1162 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1163 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1164 | */ |
1165 | if (unmounting) { | |
3558fd73 BB |
1166 | zsb->z_unmounted = B_TRUE; |
1167 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
1168 | rw_exit(&zsb->z_teardown_inactive_lock); | |
34dc7c2f BB |
1169 | } |
1170 | ||
1171 | /* | |
1172 | * z_os will be NULL if there was an error in attempting to reopen | |
3558fd73 BB |
1173 | * zsb, so just return as the properties had already been |
1174 | * | |
34dc7c2f BB |
1175 | * unregistered and cached data had been evicted before. |
1176 | */ | |
3558fd73 | 1177 | if (zsb->z_os == NULL) |
34dc7c2f BB |
1178 | return (0); |
1179 | ||
1180 | /* | |
1181 | * Unregister properties. | |
1182 | */ | |
3558fd73 | 1183 | zfs_unregister_callbacks(zsb); |
34dc7c2f BB |
1184 | |
1185 | /* | |
1186 | * Evict cached data | |
1187 | */ | |
04434775 MA |
1188 | if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) && |
1189 | !zfs_is_readonly(zsb)) | |
1190 | txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); | |
13fe0198 | 1191 | dmu_objset_evict_dbufs(zsb->z_os); |
34dc7c2f BB |
1192 | |
1193 | return (0); | |
1194 | } | |
86f35f34 | 1195 | EXPORT_SYMBOL(zfs_sb_teardown); |
34dc7c2f | 1196 | |
5547c2f1 BB |
1197 | #if defined(HAVE_BDI) && !defined(HAVE_BDI_SETUP_AND_REGISTER) |
1198 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); | |
1199 | #endif /* HAVE_BDI && !HAVE_BDI_SETUP_AND_REGISTER */ | |
76659dc1 | 1200 | |
e5c39b95 | 1201 | int |
3558fd73 | 1202 | zfs_domount(struct super_block *sb, void *data, int silent) |
34dc7c2f | 1203 | { |
3558fd73 BB |
1204 | zpl_mount_data_t *zmd = data; |
1205 | const char *osname = zmd->z_osname; | |
1206 | zfs_sb_t *zsb; | |
1207 | struct inode *root_inode; | |
1208 | uint64_t recordsize; | |
1209 | int error; | |
34dc7c2f | 1210 | |
3558fd73 BB |
1211 | error = zfs_sb_create(osname, &zsb); |
1212 | if (error) | |
1213 | return (error); | |
1214 | ||
1215 | if ((error = dsl_prop_get_integer(osname, "recordsize", | |
1216 | &recordsize, NULL))) | |
1217 | goto out; | |
1218 | ||
1219 | zsb->z_sb = sb; | |
3558fd73 BB |
1220 | sb->s_fs_info = zsb; |
1221 | sb->s_magic = ZFS_SUPER_MAGIC; | |
1222 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1223 | sb->s_time_gran = 1; | |
1224 | sb->s_blocksize = recordsize; | |
1225 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 BB |
1226 | |
1227 | #ifdef HAVE_BDI | |
1228 | /* | |
1229 | * 2.6.32 API change, | |
1230 | * Added backing_device_info (BDI) per super block interfaces. A BDI | |
1231 | * must be configured when using a non-device backed filesystem for | |
1232 | * proper writeback. This is not required for older pdflush kernels. | |
1233 | * | |
1234 | * NOTE: Linux read-ahead is disabled in favor of zfs read-ahead. | |
1235 | */ | |
1236 | zsb->z_bdi.ra_pages = 0; | |
1237 | sb->s_bdi = &zsb->z_bdi; | |
1238 | ||
1239 | error = -bdi_setup_and_register(&zsb->z_bdi, "zfs", BDI_CAP_MAP_COPY); | |
1240 | if (error) | |
1241 | goto out; | |
1242 | #endif /* HAVE_BDI */ | |
3558fd73 BB |
1243 | |
1244 | /* Set callback operations for the file system. */ | |
1245 | sb->s_op = &zpl_super_operations; | |
1246 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1247 | sb->s_export_op = &zpl_export_operations; |
ee930353 BB |
1248 | #ifdef HAVE_S_D_OP |
1249 | sb->s_d_op = &zpl_dentry_operations; | |
1250 | #endif /* HAVE_S_D_OP */ | |
3558fd73 BB |
1251 | |
1252 | /* Set features for file system. */ | |
1253 | zfs_set_fuid_feature(zsb); | |
1254 | ||
1255 | if (dmu_objset_is_snapshot(zsb->z_os)) { | |
1256 | uint64_t pval; | |
1257 | ||
1258 | atime_changed_cb(zsb, B_FALSE); | |
1259 | readonly_changed_cb(zsb, B_TRUE); | |
d1d7e268 MK |
1260 | if ((error = dsl_prop_get_integer(osname, |
1261 | "xattr", &pval, NULL))) | |
3558fd73 BB |
1262 | goto out; |
1263 | xattr_changed_cb(zsb, pval); | |
d1d7e268 MK |
1264 | if ((error = dsl_prop_get_integer(osname, |
1265 | "acltype", &pval, NULL))) | |
023699cd MM |
1266 | goto out; |
1267 | acltype_changed_cb(zsb, pval); | |
3558fd73 BB |
1268 | zsb->z_issnap = B_TRUE; |
1269 | zsb->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1270 | ||
1271 | mutex_enter(&zsb->z_os->os_user_ptr_lock); | |
1272 | dmu_objset_set_user(zsb->z_os, zsb); | |
1273 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
1274 | } else { | |
1275 | error = zfs_sb_setup(zsb, B_TRUE); | |
34dc7c2f BB |
1276 | } |
1277 | ||
3558fd73 BB |
1278 | /* Allocate a root inode for the filesystem. */ |
1279 | error = zfs_root(zsb, &root_inode); | |
1280 | if (error) { | |
1281 | (void) zfs_umount(sb); | |
1282 | goto out; | |
34dc7c2f BB |
1283 | } |
1284 | ||
3558fd73 | 1285 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1286 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1287 | if (sb->s_root == NULL) { |
1288 | (void) zfs_umount(sb); | |
2e528b49 | 1289 | error = SET_ERROR(ENOMEM); |
3558fd73 BB |
1290 | goto out; |
1291 | } | |
ebe7e575 BB |
1292 | |
1293 | if (!zsb->z_issnap) | |
1294 | zfsctl_create(zsb); | |
3558fd73 BB |
1295 | out: |
1296 | if (error) { | |
1297 | dmu_objset_disown(zsb->z_os, zsb); | |
1298 | zfs_sb_free(zsb); | |
1299 | } | |
34dc7c2f | 1300 | |
3558fd73 BB |
1301 | return (error); |
1302 | } | |
1303 | EXPORT_SYMBOL(zfs_domount); | |
1304 | ||
ebe7e575 BB |
1305 | /* |
1306 | * Called when an unmount is requested and certain sanity checks have | |
1307 | * already passed. At this point no dentries or inodes have been reclaimed | |
1308 | * from their respective caches. We drop the extra reference on the .zfs | |
1309 | * control directory to allow everything to be reclaimed. All snapshots | |
1310 | * must already have been unmounted to reach this point. | |
1311 | */ | |
1312 | void | |
1313 | zfs_preumount(struct super_block *sb) | |
1314 | { | |
1315 | zfs_sb_t *zsb = sb->s_fs_info; | |
1316 | ||
1f0d8a56 | 1317 | if (zsb != NULL && zsb->z_ctldir != NULL) |
ebe7e575 BB |
1318 | zfsctl_destroy(zsb); |
1319 | } | |
1320 | EXPORT_SYMBOL(zfs_preumount); | |
1321 | ||
1322 | /* | |
1323 | * Called once all other unmount released tear down has occurred. | |
1324 | * It is our responsibility to release any remaining infrastructure. | |
1325 | */ | |
3558fd73 BB |
1326 | /*ARGSUSED*/ |
1327 | int | |
1328 | zfs_umount(struct super_block *sb) | |
1329 | { | |
1330 | zfs_sb_t *zsb = sb->s_fs_info; | |
1331 | objset_t *os; | |
1332 | ||
86f35f34 | 1333 | VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0); |
3558fd73 | 1334 | os = zsb->z_os; |
34dc7c2f | 1335 | |
5547c2f1 BB |
1336 | #ifdef HAVE_BDI |
1337 | bdi_destroy(sb->s_bdi); | |
1338 | #endif /* HAVE_BDI */ | |
76659dc1 | 1339 | |
34dc7c2f BB |
1340 | /* |
1341 | * z_os will be NULL if there was an error in | |
3558fd73 | 1342 | * attempting to reopen zsb. |
34dc7c2f BB |
1343 | */ |
1344 | if (os != NULL) { | |
1345 | /* | |
1346 | * Unset the objset user_ptr. | |
1347 | */ | |
428870ff | 1348 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1349 | dmu_objset_set_user(os, NULL); |
428870ff | 1350 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1351 | |
1352 | /* | |
b128c09f | 1353 | * Finally release the objset |
34dc7c2f | 1354 | */ |
3558fd73 | 1355 | dmu_objset_disown(os, zsb); |
34dc7c2f BB |
1356 | } |
1357 | ||
3558fd73 | 1358 | zfs_sb_free(zsb); |
34dc7c2f BB |
1359 | return (0); |
1360 | } | |
e5c39b95 | 1361 | EXPORT_SYMBOL(zfs_umount); |
34dc7c2f | 1362 | |
0de19dad BB |
1363 | int |
1364 | zfs_remount(struct super_block *sb, int *flags, char *data) | |
1365 | { | |
0de19dad | 1366 | /* |
2cf7f52b BB |
1367 | * All namespace flags (MNT_*) and super block flags (MS_*) will |
1368 | * be handled by the Linux VFS. Only handle custom options here. | |
0de19dad | 1369 | */ |
0de19dad BB |
1370 | return (0); |
1371 | } | |
1372 | EXPORT_SYMBOL(zfs_remount); | |
1373 | ||
e5c39b95 | 1374 | int |
2cf7f52b | 1375 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1376 | { |
2cf7f52b | 1377 | zfs_sb_t *zsb = sb->s_fs_info; |
34dc7c2f BB |
1378 | znode_t *zp; |
1379 | uint64_t object = 0; | |
1380 | uint64_t fid_gen = 0; | |
1381 | uint64_t gen_mask; | |
1382 | uint64_t zp_gen; | |
3558fd73 | 1383 | int i, err; |
34dc7c2f | 1384 | |
3558fd73 | 1385 | *ipp = NULL; |
34dc7c2f | 1386 | |
3558fd73 | 1387 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1388 | |
1389 | if (fidp->fid_len == LONG_FID_LEN) { | |
1390 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1391 | uint64_t objsetid = 0; | |
1392 | uint64_t setgen = 0; | |
1393 | ||
1394 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1395 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1396 | ||
1397 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1398 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1399 | ||
3558fd73 | 1400 | ZFS_EXIT(zsb); |
34dc7c2f | 1401 | |
ebe7e575 | 1402 | err = zfsctl_lookup_objset(sb, objsetid, &zsb); |
34dc7c2f | 1403 | if (err) |
2e528b49 | 1404 | return (SET_ERROR(EINVAL)); |
ebe7e575 | 1405 | |
3558fd73 | 1406 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1407 | } |
1408 | ||
1409 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { | |
1410 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1411 | ||
1412 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1413 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
1414 | ||
1415 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
1416 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1417 | } else { | |
3558fd73 | 1418 | ZFS_EXIT(zsb); |
2e528b49 | 1419 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1420 | } |
1421 | ||
1422 | /* A zero fid_gen means we are in the .zfs control directories */ | |
1423 | if (fid_gen == 0 && | |
1424 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
3558fd73 BB |
1425 | *ipp = zsb->z_ctldir; |
1426 | ASSERT(*ipp != NULL); | |
34dc7c2f | 1427 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1428 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1429 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1430 | } else { |
3558fd73 | 1431 | igrab(*ipp); |
34dc7c2f | 1432 | } |
3558fd73 | 1433 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1434 | return (0); |
1435 | } | |
1436 | ||
1437 | gen_mask = -1ULL >> (64 - 8 * i); | |
1438 | ||
29e57d15 | 1439 | dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); |
3558fd73 BB |
1440 | if ((err = zfs_zget(zsb, object, &zp))) { |
1441 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1442 | return (err); |
1443 | } | |
3558fd73 | 1444 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, |
428870ff BB |
1445 | sizeof (uint64_t)); |
1446 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1447 | if (zp_gen == 0) |
1448 | zp_gen = 1; | |
1449 | if (zp->z_unlinked || zp_gen != fid_gen) { | |
29e57d15 NB |
1450 | dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, |
1451 | fid_gen); | |
3558fd73 BB |
1452 | iput(ZTOI(zp)); |
1453 | ZFS_EXIT(zsb); | |
2e528b49 | 1454 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1455 | } |
1456 | ||
3558fd73 BB |
1457 | *ipp = ZTOI(zp); |
1458 | if (*ipp) | |
1459 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1460 | |
3558fd73 | 1461 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1462 | return (0); |
1463 | } | |
e5c39b95 | 1464 | EXPORT_SYMBOL(zfs_vget); |
34dc7c2f BB |
1465 | |
1466 | /* | |
7b3e34ba | 1467 | * Block out VFS ops and close zfs_sb_t |
34dc7c2f BB |
1468 | * |
1469 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
831baf06 KW |
1470 | * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying |
1471 | * dataset and objset intact so that they can be atomically handed off during | |
1472 | * a subsequent rollback or recv operation and the resume thereafter. | |
34dc7c2f BB |
1473 | */ |
1474 | int | |
3558fd73 | 1475 | zfs_suspend_fs(zfs_sb_t *zsb) |
34dc7c2f BB |
1476 | { |
1477 | int error; | |
1478 | ||
86f35f34 | 1479 | if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0) |
34dc7c2f | 1480 | return (error); |
7b3e34ba | 1481 | |
34dc7c2f BB |
1482 | return (0); |
1483 | } | |
e5c39b95 | 1484 | EXPORT_SYMBOL(zfs_suspend_fs); |
34dc7c2f BB |
1485 | |
1486 | /* | |
7b3e34ba | 1487 | * Reopen zfs_sb_t and release VFS ops. |
34dc7c2f BB |
1488 | */ |
1489 | int | |
3558fd73 | 1490 | zfs_resume_fs(zfs_sb_t *zsb, const char *osname) |
34dc7c2f | 1491 | { |
2008e920 | 1492 | int err, err2; |
831baf06 KW |
1493 | znode_t *zp; |
1494 | uint64_t sa_obj = 0; | |
34dc7c2f | 1495 | |
3558fd73 BB |
1496 | ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock)); |
1497 | ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); | |
34dc7c2f | 1498 | |
831baf06 KW |
1499 | /* |
1500 | * We already own this, so just hold and rele it to update the | |
1501 | * objset_t, as the one we had before may have been evicted. | |
1502 | */ | |
1503 | VERIFY0(dmu_objset_hold(osname, zsb, &zsb->z_os)); | |
1504 | VERIFY3P(zsb->z_os->os_dsl_dataset->ds_owner, ==, zsb); | |
1505 | VERIFY(dsl_dataset_long_held(zsb->z_os->os_dsl_dataset)); | |
1506 | dmu_objset_rele(zsb->z_os, zsb); | |
428870ff | 1507 | |
831baf06 KW |
1508 | /* |
1509 | * Make sure version hasn't changed | |
1510 | */ | |
428870ff | 1511 | |
831baf06 KW |
1512 | err = zfs_get_zplprop(zsb->z_os, ZFS_PROP_VERSION, |
1513 | &zsb->z_version); | |
c1fabe79 | 1514 | |
831baf06 KW |
1515 | if (err) |
1516 | goto bail; | |
428870ff | 1517 | |
831baf06 KW |
1518 | err = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, |
1519 | ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
c1fabe79 | 1520 | |
831baf06 KW |
1521 | if (err && zsb->z_version >= ZPL_VERSION_SA) |
1522 | goto bail; | |
428870ff | 1523 | |
831baf06 KW |
1524 | if ((err = sa_setup(zsb->z_os, sa_obj, |
1525 | zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) | |
1526 | goto bail; | |
34dc7c2f | 1527 | |
831baf06 KW |
1528 | if (zsb->z_version >= ZPL_VERSION_SA) |
1529 | sa_register_update_callback(zsb->z_os, | |
1530 | zfs_sa_upgrade); | |
c1fabe79 | 1531 | |
831baf06 | 1532 | VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); |
c1fabe79 | 1533 | |
831baf06 KW |
1534 | zfs_set_fuid_feature(zsb); |
1535 | zsb->z_rollback_time = jiffies; | |
34dc7c2f | 1536 | |
831baf06 KW |
1537 | /* |
1538 | * Attempt to re-establish all the active inodes with their | |
1539 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
1540 | * and mark it stale. This prevents a collision if a new | |
1541 | * inode/object is created which must use the same inode | |
1542 | * number. The stale inode will be be released when the | |
1543 | * VFS prunes the dentry holding the remaining references | |
1544 | * on the stale inode. | |
1545 | */ | |
1546 | mutex_enter(&zsb->z_znodes_lock); | |
1547 | for (zp = list_head(&zsb->z_all_znodes); zp; | |
1548 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
2008e920 TC |
1549 | err2 = zfs_rezget(zp); |
1550 | if (err2) { | |
831baf06 KW |
1551 | remove_inode_hash(ZTOI(zp)); |
1552 | zp->z_is_stale = B_TRUE; | |
34dc7c2f | 1553 | } |
34dc7c2f | 1554 | } |
831baf06 | 1555 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f | 1556 | |
428870ff | 1557 | bail: |
7b3e34ba | 1558 | /* release the VFS ops */ |
3558fd73 BB |
1559 | rw_exit(&zsb->z_teardown_inactive_lock); |
1560 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1561 | |
1562 | if (err) { | |
1563 | /* | |
831baf06 KW |
1564 | * Since we couldn't setup the sa framework, try to force |
1565 | * unmount this file system. | |
34dc7c2f | 1566 | */ |
ab68b6e5 TC |
1567 | if (zsb->z_os) |
1568 | (void) zfs_umount(zsb->z_sb); | |
34dc7c2f BB |
1569 | } |
1570 | return (err); | |
1571 | } | |
e5c39b95 | 1572 | EXPORT_SYMBOL(zfs_resume_fs); |
34dc7c2f | 1573 | |
34dc7c2f | 1574 | int |
3558fd73 | 1575 | zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) |
34dc7c2f BB |
1576 | { |
1577 | int error; | |
3558fd73 | 1578 | objset_t *os = zsb->z_os; |
34dc7c2f | 1579 | dmu_tx_t *tx; |
34dc7c2f BB |
1580 | |
1581 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
2e528b49 | 1582 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1583 | |
3558fd73 | 1584 | if (newvers < zsb->z_version) |
2e528b49 | 1585 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1586 | |
428870ff | 1587 | if (zfs_spa_version_map(newvers) > |
3558fd73 | 1588 | spa_version(dmu_objset_spa(zsb->z_os))) |
2e528b49 | 1589 | return (SET_ERROR(ENOTSUP)); |
428870ff | 1590 | |
34dc7c2f | 1591 | tx = dmu_tx_create(os); |
9babb374 | 1592 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
3558fd73 | 1593 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1594 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1595 | ZFS_SA_ATTRS); | |
1596 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
1597 | } | |
34dc7c2f BB |
1598 | error = dmu_tx_assign(tx, TXG_WAIT); |
1599 | if (error) { | |
1600 | dmu_tx_abort(tx); | |
9babb374 BB |
1601 | return (error); |
1602 | } | |
428870ff | 1603 | |
9babb374 BB |
1604 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
1605 | 8, 1, &newvers, tx); | |
1606 | ||
1607 | if (error) { | |
1608 | dmu_tx_commit(tx); | |
1609 | return (error); | |
34dc7c2f | 1610 | } |
34dc7c2f | 1611 | |
3558fd73 | 1612 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1613 | uint64_t sa_obj; |
1614 | ||
3558fd73 | 1615 | ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, |
428870ff BB |
1616 | SPA_VERSION_SA); |
1617 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1618 | DMU_OT_NONE, 0, tx); | |
1619 | ||
1620 | error = zap_add(os, MASTER_NODE_OBJ, | |
1621 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c99c9001 | 1622 | ASSERT0(error); |
428870ff BB |
1623 | |
1624 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
1625 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1626 | } | |
1627 | ||
6f1ffb06 MA |
1628 | spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, |
1629 | "from %llu to %llu", zsb->z_version, newvers); | |
9babb374 | 1630 | |
34dc7c2f BB |
1631 | dmu_tx_commit(tx); |
1632 | ||
3558fd73 | 1633 | zsb->z_version = newvers; |
9babb374 | 1634 | |
c1fabe79 | 1635 | zfs_set_fuid_feature(zsb); |
9babb374 BB |
1636 | |
1637 | return (0); | |
34dc7c2f | 1638 | } |
e5c39b95 | 1639 | EXPORT_SYMBOL(zfs_set_version); |
34dc7c2f BB |
1640 | |
1641 | /* | |
1642 | * Read a property stored within the master node. | |
1643 | */ | |
1644 | int | |
1645 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
1646 | { | |
1647 | const char *pname; | |
2e528b49 | 1648 | int error = SET_ERROR(ENOENT); |
34dc7c2f BB |
1649 | |
1650 | /* | |
1651 | * Look up the file system's value for the property. For the | |
1652 | * version property, we look up a slightly different string. | |
1653 | */ | |
1654 | if (prop == ZFS_PROP_VERSION) | |
1655 | pname = ZPL_VERSION_STR; | |
1656 | else | |
1657 | pname = zfs_prop_to_name(prop); | |
1658 | ||
b128c09f BB |
1659 | if (os != NULL) |
1660 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
1661 | |
1662 | if (error == ENOENT) { | |
1663 | /* No value set, use the default value */ | |
1664 | switch (prop) { | |
1665 | case ZFS_PROP_VERSION: | |
1666 | *value = ZPL_VERSION; | |
1667 | break; | |
1668 | case ZFS_PROP_NORMALIZE: | |
1669 | case ZFS_PROP_UTF8ONLY: | |
1670 | *value = 0; | |
1671 | break; | |
1672 | case ZFS_PROP_CASE: | |
1673 | *value = ZFS_CASE_SENSITIVE; | |
1674 | break; | |
023699cd MM |
1675 | case ZFS_PROP_ACLTYPE: |
1676 | *value = ZFS_ACLTYPE_OFF; | |
1677 | break; | |
34dc7c2f BB |
1678 | default: |
1679 | return (error); | |
1680 | } | |
1681 | error = 0; | |
1682 | } | |
1683 | return (error); | |
1684 | } | |
86f35f34 | 1685 | EXPORT_SYMBOL(zfs_get_zplprop); |
3558fd73 BB |
1686 | |
1687 | void | |
1688 | zfs_init(void) | |
1689 | { | |
ebe7e575 | 1690 | zfsctl_init(); |
3558fd73 BB |
1691 | zfs_znode_init(); |
1692 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
1693 | register_filesystem(&zpl_fs_type); | |
ab26409d | 1694 | (void) arc_add_prune_callback(zpl_prune_sbs, NULL); |
3558fd73 BB |
1695 | } |
1696 | ||
1697 | void | |
1698 | zfs_fini(void) | |
1699 | { | |
023699cd | 1700 | taskq_wait(system_taskq); |
3558fd73 BB |
1701 | unregister_filesystem(&zpl_fs_type); |
1702 | zfs_znode_fini(); | |
ebe7e575 | 1703 | zfsctl_fini(); |
3558fd73 | 1704 | } |