]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
21 | /* | |
428870ff | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
fcff0f35 | 23 | * Copyright (c) 2012, 2014 by Delphix. All rights reserved. |
34dc7c2f BB |
24 | */ |
25 | ||
428870ff BB |
26 | /* Portions Copyright 2010 Robert Milkowski */ |
27 | ||
34dc7c2f BB |
28 | #include <sys/types.h> |
29 | #include <sys/param.h> | |
30 | #include <sys/systm.h> | |
31 | #include <sys/sysmacros.h> | |
32 | #include <sys/kmem.h> | |
33 | #include <sys/pathname.h> | |
34 | #include <sys/vnode.h> | |
35 | #include <sys/vfs.h> | |
36 | #include <sys/vfs_opreg.h> | |
37 | #include <sys/mntent.h> | |
38 | #include <sys/mount.h> | |
39 | #include <sys/cmn_err.h> | |
40 | #include "fs/fs_subr.h" | |
41 | #include <sys/zfs_znode.h> | |
3558fd73 | 42 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
43 | #include <sys/zfs_dir.h> |
44 | #include <sys/zil.h> | |
45 | #include <sys/fs/zfs.h> | |
46 | #include <sys/dmu.h> | |
47 | #include <sys/dsl_prop.h> | |
48 | #include <sys/dsl_dataset.h> | |
49 | #include <sys/dsl_deleg.h> | |
50 | #include <sys/spa.h> | |
51 | #include <sys/zap.h> | |
428870ff | 52 | #include <sys/sa.h> |
a94addd9 | 53 | #include <sys/sa_impl.h> |
34dc7c2f BB |
54 | #include <sys/varargs.h> |
55 | #include <sys/policy.h> | |
56 | #include <sys/atomic.h> | |
57 | #include <sys/mkdev.h> | |
58 | #include <sys/modctl.h> | |
59 | #include <sys/refstr.h> | |
60 | #include <sys/zfs_ioctl.h> | |
ebe7e575 | 61 | #include <sys/zfs_ctldir.h> |
34dc7c2f BB |
62 | #include <sys/zfs_fuid.h> |
63 | #include <sys/bootconf.h> | |
64 | #include <sys/sunddi.h> | |
65 | #include <sys/dnlc.h> | |
66 | #include <sys/dmu_objset.h> | |
67 | #include <sys/spa_boot.h> | |
3558fd73 | 68 | #include <sys/zpl.h> |
428870ff | 69 | #include "zfs_comutil.h" |
34dc7c2f | 70 | |
34dc7c2f BB |
71 | /*ARGSUSED*/ |
72 | int | |
03f9ba9d | 73 | zfs_sync(struct super_block *sb, int wait, cred_t *cr) |
34dc7c2f | 74 | { |
03f9ba9d BB |
75 | zfs_sb_t *zsb = sb->s_fs_info; |
76 | ||
34dc7c2f BB |
77 | /* |
78 | * Data integrity is job one. We don't want a compromised kernel | |
79 | * writing to the storage pool, so we never sync during panic. | |
80 | */ | |
d5e53f9d | 81 | if (unlikely(oops_in_progress)) |
34dc7c2f BB |
82 | return (0); |
83 | ||
03f9ba9d BB |
84 | /* |
85 | * Semantically, the only requirement is that the sync be initiated. | |
86 | * The DMU syncs out txgs frequently, so there's nothing to do. | |
87 | */ | |
88 | if (!wait) | |
89 | return (0); | |
90 | ||
3558fd73 | 91 | if (zsb != NULL) { |
34dc7c2f BB |
92 | /* |
93 | * Sync a specific filesystem. | |
94 | */ | |
9babb374 | 95 | dsl_pool_t *dp; |
34dc7c2f | 96 | |
3558fd73 BB |
97 | ZFS_ENTER(zsb); |
98 | dp = dmu_objset_pool(zsb->z_os); | |
9babb374 BB |
99 | |
100 | /* | |
101 | * If the system is shutting down, then skip any | |
102 | * filesystems which may exist on a suspended pool. | |
103 | */ | |
03f9ba9d | 104 | if (spa_suspended(dp->dp_spa)) { |
3558fd73 | 105 | ZFS_EXIT(zsb); |
9babb374 BB |
106 | return (0); |
107 | } | |
108 | ||
3558fd73 BB |
109 | if (zsb->z_log != NULL) |
110 | zil_commit(zsb->z_log, 0); | |
428870ff | 111 | |
3558fd73 | 112 | ZFS_EXIT(zsb); |
34dc7c2f BB |
113 | } else { |
114 | /* | |
115 | * Sync all ZFS filesystems. This is what happens when you | |
116 | * run sync(1M). Unlike other filesystems, ZFS honors the | |
117 | * request by waiting for all pools to commit all dirty data. | |
118 | */ | |
119 | spa_sync_allpools(); | |
120 | } | |
121 | ||
122 | return (0); | |
123 | } | |
e5c39b95 | 124 | EXPORT_SYMBOL(zfs_sync); |
34dc7c2f | 125 | |
2cf7f52b BB |
126 | boolean_t |
127 | zfs_is_readonly(zfs_sb_t *zsb) | |
128 | { | |
129 | return (!!(zsb->z_sb->s_flags & MS_RDONLY)); | |
130 | } | |
131 | EXPORT_SYMBOL(zfs_is_readonly); | |
132 | ||
34dc7c2f BB |
133 | static void |
134 | atime_changed_cb(void *arg, uint64_t newval) | |
135 | { | |
2cf7f52b | 136 | ((zfs_sb_t *)arg)->z_atime = newval; |
34dc7c2f BB |
137 | } |
138 | ||
6d111134 TC |
139 | static void |
140 | relatime_changed_cb(void *arg, uint64_t newval) | |
141 | { | |
142 | ((zfs_sb_t *)arg)->z_relatime = newval; | |
143 | } | |
144 | ||
34dc7c2f BB |
145 | static void |
146 | xattr_changed_cb(void *arg, uint64_t newval) | |
147 | { | |
3558fd73 | 148 | zfs_sb_t *zsb = arg; |
34dc7c2f | 149 | |
82a37189 | 150 | if (newval == ZFS_XATTR_OFF) { |
2cf7f52b | 151 | zsb->z_flags &= ~ZSB_XATTR; |
82a37189 BB |
152 | } else { |
153 | zsb->z_flags |= ZSB_XATTR; | |
154 | ||
155 | if (newval == ZFS_XATTR_SA) | |
156 | zsb->z_xattr_sa = B_TRUE; | |
157 | else | |
158 | zsb->z_xattr_sa = B_FALSE; | |
159 | } | |
34dc7c2f BB |
160 | } |
161 | ||
023699cd MM |
162 | static void |
163 | acltype_changed_cb(void *arg, uint64_t newval) | |
164 | { | |
165 | zfs_sb_t *zsb = arg; | |
166 | ||
167 | switch (newval) { | |
168 | case ZFS_ACLTYPE_OFF: | |
169 | zsb->z_acl_type = ZFS_ACLTYPE_OFF; | |
170 | zsb->z_sb->s_flags &= ~MS_POSIXACL; | |
171 | break; | |
172 | case ZFS_ACLTYPE_POSIXACL: | |
b695c34e | 173 | #ifdef CONFIG_FS_POSIX_ACL |
023699cd MM |
174 | zsb->z_acl_type = ZFS_ACLTYPE_POSIXACL; |
175 | zsb->z_sb->s_flags |= MS_POSIXACL; | |
b695c34e MM |
176 | #else |
177 | zsb->z_acl_type = ZFS_ACLTYPE_OFF; | |
178 | zsb->z_sb->s_flags &= ~MS_POSIXACL; | |
179 | #endif /* CONFIG_FS_POSIX_ACL */ | |
023699cd MM |
180 | break; |
181 | default: | |
182 | break; | |
183 | } | |
184 | } | |
185 | ||
34dc7c2f BB |
186 | static void |
187 | blksz_changed_cb(void *arg, uint64_t newval) | |
188 | { | |
3558fd73 | 189 | zfs_sb_t *zsb = arg; |
f1512ee6 MA |
190 | ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zsb->z_os))); |
191 | ASSERT3U(newval, >=, SPA_MINBLOCKSIZE); | |
192 | ASSERT(ISP2(newval)); | |
34dc7c2f | 193 | |
3558fd73 | 194 | zsb->z_max_blksz = newval; |
34dc7c2f BB |
195 | } |
196 | ||
197 | static void | |
198 | readonly_changed_cb(void *arg, uint64_t newval) | |
199 | { | |
3558fd73 BB |
200 | zfs_sb_t *zsb = arg; |
201 | struct super_block *sb = zsb->z_sb; | |
34dc7c2f | 202 | |
2cf7f52b BB |
203 | if (sb == NULL) |
204 | return; | |
205 | ||
206 | if (newval) | |
3558fd73 | 207 | sb->s_flags |= MS_RDONLY; |
2cf7f52b | 208 | else |
3558fd73 | 209 | sb->s_flags &= ~MS_RDONLY; |
34dc7c2f BB |
210 | } |
211 | ||
212 | static void | |
213 | devices_changed_cb(void *arg, uint64_t newval) | |
214 | { | |
34dc7c2f BB |
215 | } |
216 | ||
217 | static void | |
218 | setuid_changed_cb(void *arg, uint64_t newval) | |
219 | { | |
34dc7c2f BB |
220 | } |
221 | ||
222 | static void | |
223 | exec_changed_cb(void *arg, uint64_t newval) | |
224 | { | |
34dc7c2f BB |
225 | } |
226 | ||
34dc7c2f BB |
227 | static void |
228 | nbmand_changed_cb(void *arg, uint64_t newval) | |
229 | { | |
3558fd73 BB |
230 | zfs_sb_t *zsb = arg; |
231 | struct super_block *sb = zsb->z_sb; | |
232 | ||
2cf7f52b BB |
233 | if (sb == NULL) |
234 | return; | |
235 | ||
236 | if (newval == TRUE) | |
3558fd73 | 237 | sb->s_flags |= MS_MANDLOCK; |
2cf7f52b | 238 | else |
3558fd73 | 239 | sb->s_flags &= ~MS_MANDLOCK; |
34dc7c2f BB |
240 | } |
241 | ||
242 | static void | |
243 | snapdir_changed_cb(void *arg, uint64_t newval) | |
244 | { | |
3558fd73 | 245 | ((zfs_sb_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
246 | } |
247 | ||
248 | static void | |
249 | vscan_changed_cb(void *arg, uint64_t newval) | |
250 | { | |
3558fd73 | 251 | ((zfs_sb_t *)arg)->z_vscan = newval; |
34dc7c2f BB |
252 | } |
253 | ||
34dc7c2f BB |
254 | static void |
255 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
256 | { | |
3558fd73 | 257 | ((zfs_sb_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
258 | } |
259 | ||
e5c39b95 | 260 | int |
3558fd73 | 261 | zfs_register_callbacks(zfs_sb_t *zsb) |
34dc7c2f BB |
262 | { |
263 | struct dsl_dataset *ds = NULL; | |
3558fd73 | 264 | objset_t *os = zsb->z_os; |
0282c413 | 265 | zfs_mntopts_t *zmo = zsb->z_mntopts; |
34dc7c2f BB |
266 | int error = 0; |
267 | ||
0282c413 BB |
268 | ASSERT(zsb); |
269 | ASSERT(zmo); | |
270 | ||
271 | /* | |
272 | * The act of registering our callbacks will destroy any mount | |
273 | * options we may have. In order to enable temporary overrides | |
274 | * of mount options, we stash away the current values and | |
275 | * restore them after we register the callbacks. | |
276 | */ | |
277 | if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) { | |
278 | zmo->z_do_readonly = B_TRUE; | |
279 | zmo->z_readonly = B_TRUE; | |
280 | } | |
34dc7c2f BB |
281 | |
282 | /* | |
283 | * Register property callbacks. | |
284 | * | |
285 | * It would probably be fine to just check for i/o error from | |
286 | * the first prop_register(), but I guess I like to go | |
287 | * overboard... | |
288 | */ | |
289 | ds = dmu_objset_ds(os); | |
13fe0198 | 290 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
3558fd73 | 291 | error = dsl_prop_register(ds, |
13fe0198 | 292 | zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb); |
98fad862 | 293 | error = error ? error : dsl_prop_register(ds, |
6d111134 | 294 | zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb); |
34dc7c2f | 295 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 296 | zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb); |
34dc7c2f | 297 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 298 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb); |
34dc7c2f | 299 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 300 | zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb); |
34dc7c2f | 301 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 302 | zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb); |
34dc7c2f | 303 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 304 | zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb); |
34dc7c2f | 305 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 306 | zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb); |
34dc7c2f | 307 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 308 | zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb); |
34dc7c2f | 309 | error = error ? error : dsl_prop_register(ds, |
023699cd MM |
310 | zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb); |
311 | error = error ? error : dsl_prop_register(ds, | |
312 | zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb); | |
34dc7c2f | 313 | error = error ? error : dsl_prop_register(ds, |
13fe0198 | 314 | zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb); |
2cf7f52b | 315 | error = error ? error : dsl_prop_register(ds, |
13fe0198 MA |
316 | zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb); |
317 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); | |
34dc7c2f BB |
318 | if (error) |
319 | goto unregister; | |
320 | ||
0282c413 BB |
321 | /* |
322 | * Invoke our callbacks to restore temporary mount options. | |
323 | */ | |
324 | if (zmo->z_do_readonly) | |
325 | readonly_changed_cb(zsb, zmo->z_readonly); | |
326 | if (zmo->z_do_setuid) | |
327 | setuid_changed_cb(zsb, zmo->z_setuid); | |
328 | if (zmo->z_do_exec) | |
329 | exec_changed_cb(zsb, zmo->z_exec); | |
330 | if (zmo->z_do_devices) | |
331 | devices_changed_cb(zsb, zmo->z_devices); | |
332 | if (zmo->z_do_xattr) | |
333 | xattr_changed_cb(zsb, zmo->z_xattr); | |
334 | if (zmo->z_do_atime) | |
335 | atime_changed_cb(zsb, zmo->z_atime); | |
336 | if (zmo->z_do_relatime) | |
337 | relatime_changed_cb(zsb, zmo->z_relatime); | |
338 | if (zmo->z_do_nbmand) | |
339 | nbmand_changed_cb(zsb, zmo->z_nbmand); | |
c9ada6d5 | 340 | |
34dc7c2f BB |
341 | return (0); |
342 | ||
343 | unregister: | |
0eb21616 | 344 | dsl_prop_unregister_all(ds, zsb); |
3558fd73 | 345 | return (error); |
34dc7c2f | 346 | } |
e5c39b95 | 347 | EXPORT_SYMBOL(zfs_register_callbacks); |
34dc7c2f | 348 | |
428870ff BB |
349 | static int |
350 | zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, | |
351 | uint64_t *userp, uint64_t *groupp) | |
9babb374 | 352 | { |
428870ff BB |
353 | /* |
354 | * Is it a valid type of object to track? | |
355 | */ | |
356 | if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) | |
2e528b49 | 357 | return (SET_ERROR(ENOENT)); |
9babb374 | 358 | |
428870ff BB |
359 | /* |
360 | * If we have a NULL data pointer | |
361 | * then assume the id's aren't changing and | |
362 | * return EEXIST to the dmu to let it know to | |
363 | * use the same ids | |
364 | */ | |
365 | if (data == NULL) | |
2e528b49 | 366 | return (SET_ERROR(EEXIST)); |
9babb374 | 367 | |
428870ff | 368 | if (bonustype == DMU_OT_ZNODE) { |
a94addd9 | 369 | znode_phys_t *znp = data; |
428870ff BB |
370 | *userp = znp->zp_uid; |
371 | *groupp = znp->zp_gid; | |
9babb374 | 372 | } else { |
428870ff | 373 | int hdrsize; |
a94addd9 MA |
374 | sa_hdr_phys_t *sap = data; |
375 | sa_hdr_phys_t sa = *sap; | |
376 | boolean_t swap = B_FALSE; | |
9babb374 | 377 | |
428870ff | 378 | ASSERT(bonustype == DMU_OT_SA); |
428870ff | 379 | |
a94addd9 | 380 | if (sa.sa_magic == 0) { |
428870ff BB |
381 | /* |
382 | * This should only happen for newly created | |
383 | * files that haven't had the znode data filled | |
384 | * in yet. | |
385 | */ | |
386 | *userp = 0; | |
387 | *groupp = 0; | |
a94addd9 MA |
388 | return (0); |
389 | } | |
390 | if (sa.sa_magic == BSWAP_32(SA_MAGIC)) { | |
391 | sa.sa_magic = SA_MAGIC; | |
392 | sa.sa_layout_info = BSWAP_16(sa.sa_layout_info); | |
393 | swap = B_TRUE; | |
394 | } else { | |
395 | VERIFY3U(sa.sa_magic, ==, SA_MAGIC); | |
396 | } | |
397 | ||
398 | hdrsize = sa_hdrsize(&sa); | |
399 | VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t)); | |
400 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
401 | SA_UID_OFFSET)); | |
402 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
403 | SA_GID_OFFSET)); | |
404 | if (swap) { | |
405 | *userp = BSWAP_64(*userp); | |
406 | *groupp = BSWAP_64(*groupp); | |
428870ff | 407 | } |
9babb374 | 408 | } |
13fe0198 | 409 | return (0); |
9babb374 BB |
410 | } |
411 | ||
412 | static void | |
3558fd73 | 413 | fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr, |
9babb374 BB |
414 | char *domainbuf, int buflen, uid_t *ridp) |
415 | { | |
9babb374 BB |
416 | uint64_t fuid; |
417 | const char *domain; | |
418 | ||
419 | fuid = strtonum(fuidstr, NULL); | |
420 | ||
3558fd73 | 421 | domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid)); |
9babb374 BB |
422 | if (domain) |
423 | (void) strlcpy(domainbuf, domain, buflen); | |
424 | else | |
425 | domainbuf[0] = '\0'; | |
426 | *ridp = FUID_RID(fuid); | |
427 | } | |
428 | ||
429 | static uint64_t | |
3558fd73 | 430 | zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type) |
9babb374 BB |
431 | { |
432 | switch (type) { | |
433 | case ZFS_PROP_USERUSED: | |
434 | return (DMU_USERUSED_OBJECT); | |
435 | case ZFS_PROP_GROUPUSED: | |
436 | return (DMU_GROUPUSED_OBJECT); | |
437 | case ZFS_PROP_USERQUOTA: | |
3558fd73 | 438 | return (zsb->z_userquota_obj); |
9babb374 | 439 | case ZFS_PROP_GROUPQUOTA: |
3558fd73 | 440 | return (zsb->z_groupquota_obj); |
149e873a | 441 | default: |
2e528b49 | 442 | return (SET_ERROR(ENOTSUP)); |
9babb374 BB |
443 | } |
444 | return (0); | |
445 | } | |
446 | ||
447 | int | |
3558fd73 | 448 | zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
449 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
450 | { | |
451 | int error; | |
452 | zap_cursor_t zc; | |
453 | zap_attribute_t za; | |
454 | zfs_useracct_t *buf = vbuf; | |
455 | uint64_t obj; | |
456 | ||
3558fd73 | 457 | if (!dmu_objset_userspace_present(zsb->z_os)) |
2e528b49 | 458 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 459 | |
3558fd73 | 460 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
461 | if (obj == 0) { |
462 | *bufsizep = 0; | |
463 | return (0); | |
464 | } | |
465 | ||
3558fd73 | 466 | for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep); |
9babb374 BB |
467 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
468 | zap_cursor_advance(&zc)) { | |
469 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
470 | *bufsizep) | |
471 | break; | |
472 | ||
3558fd73 | 473 | fuidstr_to_sid(zsb, za.za_name, |
9babb374 BB |
474 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
475 | ||
476 | buf->zu_space = za.za_first_integer; | |
477 | buf++; | |
478 | } | |
479 | if (error == ENOENT) | |
480 | error = 0; | |
481 | ||
482 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
483 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
484 | *cookiep = zap_cursor_serialize(&zc); | |
485 | zap_cursor_fini(&zc); | |
486 | return (error); | |
487 | } | |
e5c39b95 | 488 | EXPORT_SYMBOL(zfs_userspace_many); |
9babb374 BB |
489 | |
490 | /* | |
491 | * buf must be big enough (eg, 32 bytes) | |
492 | */ | |
493 | static int | |
3558fd73 | 494 | id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid, |
9babb374 BB |
495 | char *buf, boolean_t addok) |
496 | { | |
497 | uint64_t fuid; | |
498 | int domainid = 0; | |
499 | ||
500 | if (domain && domain[0]) { | |
3558fd73 | 501 | domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok); |
9babb374 | 502 | if (domainid == -1) |
2e528b49 | 503 | return (SET_ERROR(ENOENT)); |
9babb374 BB |
504 | } |
505 | fuid = FUID_ENCODE(domainid, rid); | |
506 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
507 | return (0); | |
508 | } | |
509 | ||
510 | int | |
3558fd73 | 511 | zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
512 | const char *domain, uint64_t rid, uint64_t *valp) |
513 | { | |
514 | char buf[32]; | |
515 | int err; | |
516 | uint64_t obj; | |
517 | ||
518 | *valp = 0; | |
519 | ||
3558fd73 | 520 | if (!dmu_objset_userspace_present(zsb->z_os)) |
2e528b49 | 521 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 522 | |
3558fd73 | 523 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
524 | if (obj == 0) |
525 | return (0); | |
526 | ||
3558fd73 | 527 | err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE); |
9babb374 BB |
528 | if (err) |
529 | return (err); | |
530 | ||
3558fd73 | 531 | err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
532 | if (err == ENOENT) |
533 | err = 0; | |
534 | return (err); | |
535 | } | |
e5c39b95 | 536 | EXPORT_SYMBOL(zfs_userspace_one); |
9babb374 BB |
537 | |
538 | int | |
3558fd73 | 539 | zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
540 | const char *domain, uint64_t rid, uint64_t quota) |
541 | { | |
542 | char buf[32]; | |
543 | int err; | |
544 | dmu_tx_t *tx; | |
545 | uint64_t *objp; | |
546 | boolean_t fuid_dirtied; | |
547 | ||
548 | if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) | |
2e528b49 | 549 | return (SET_ERROR(EINVAL)); |
9babb374 | 550 | |
3558fd73 | 551 | if (zsb->z_version < ZPL_VERSION_USERSPACE) |
2e528b49 | 552 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 553 | |
3558fd73 BB |
554 | objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj : |
555 | &zsb->z_groupquota_obj; | |
9babb374 | 556 | |
3558fd73 | 557 | err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE); |
9babb374 BB |
558 | if (err) |
559 | return (err); | |
3558fd73 | 560 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 561 | |
3558fd73 | 562 | tx = dmu_tx_create(zsb->z_os); |
9babb374 BB |
563 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
564 | if (*objp == 0) { | |
565 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
566 | zfs_userquota_prop_prefixes[type]); | |
567 | } | |
568 | if (fuid_dirtied) | |
3558fd73 | 569 | zfs_fuid_txhold(zsb, tx); |
9babb374 BB |
570 | err = dmu_tx_assign(tx, TXG_WAIT); |
571 | if (err) { | |
572 | dmu_tx_abort(tx); | |
573 | return (err); | |
574 | } | |
575 | ||
3558fd73 | 576 | mutex_enter(&zsb->z_lock); |
9babb374 | 577 | if (*objp == 0) { |
3558fd73 | 578 | *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 579 | DMU_OT_NONE, 0, tx); |
3558fd73 | 580 | VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
581 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
582 | } | |
3558fd73 | 583 | mutex_exit(&zsb->z_lock); |
9babb374 BB |
584 | |
585 | if (quota == 0) { | |
3558fd73 | 586 | err = zap_remove(zsb->z_os, *objp, buf, tx); |
9babb374 BB |
587 | if (err == ENOENT) |
588 | err = 0; | |
589 | } else { | |
3558fd73 | 590 | err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
591 | } |
592 | ASSERT(err == 0); | |
593 | if (fuid_dirtied) | |
3558fd73 | 594 | zfs_fuid_sync(zsb, tx); |
9babb374 BB |
595 | dmu_tx_commit(tx); |
596 | return (err); | |
597 | } | |
e5c39b95 | 598 | EXPORT_SYMBOL(zfs_set_userquota); |
9babb374 BB |
599 | |
600 | boolean_t | |
3558fd73 | 601 | zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid) |
9babb374 BB |
602 | { |
603 | char buf[32]; | |
604 | uint64_t used, quota, usedobj, quotaobj; | |
605 | int err; | |
606 | ||
607 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
3558fd73 | 608 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
9babb374 | 609 | |
3558fd73 | 610 | if (quotaobj == 0 || zsb->z_replay) |
9babb374 BB |
611 | return (B_FALSE); |
612 | ||
613 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
3558fd73 | 614 | err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
615 | if (err != 0) |
616 | return (B_FALSE); | |
617 | ||
3558fd73 | 618 | err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
619 | if (err != 0) |
620 | return (B_FALSE); | |
621 | return (used >= quota); | |
622 | } | |
e5c39b95 | 623 | EXPORT_SYMBOL(zfs_fuid_overquota); |
9babb374 | 624 | |
428870ff | 625 | boolean_t |
3558fd73 | 626 | zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup) |
428870ff BB |
627 | { |
628 | uint64_t fuid; | |
629 | uint64_t quotaobj; | |
428870ff | 630 | |
3558fd73 | 631 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
428870ff | 632 | |
572e2857 | 633 | fuid = isgroup ? zp->z_gid : zp->z_uid; |
428870ff | 634 | |
3558fd73 | 635 | if (quotaobj == 0 || zsb->z_replay) |
428870ff BB |
636 | return (B_FALSE); |
637 | ||
3558fd73 | 638 | return (zfs_fuid_overquota(zsb, isgroup, fuid)); |
428870ff | 639 | } |
e5c39b95 | 640 | EXPORT_SYMBOL(zfs_owner_overquota); |
428870ff | 641 | |
0282c413 BB |
642 | zfs_mntopts_t * |
643 | zfs_mntopts_alloc(void) | |
644 | { | |
645 | return (kmem_zalloc(sizeof (zfs_mntopts_t), KM_SLEEP)); | |
646 | } | |
647 | ||
648 | void | |
649 | zfs_mntopts_free(zfs_mntopts_t *zmo) | |
650 | { | |
651 | if (zmo->z_osname) | |
652 | strfree(zmo->z_osname); | |
653 | ||
654 | if (zmo->z_mntpoint) | |
655 | strfree(zmo->z_mntpoint); | |
656 | ||
657 | kmem_free(zmo, sizeof (zfs_mntopts_t)); | |
658 | } | |
659 | ||
9babb374 | 660 | int |
0282c413 | 661 | zfs_sb_create(const char *osname, zfs_mntopts_t *zmo, zfs_sb_t **zsbp) |
9babb374 BB |
662 | { |
663 | objset_t *os; | |
3558fd73 | 664 | zfs_sb_t *zsb; |
9babb374 | 665 | uint64_t zval; |
c96c36fa | 666 | int i, size, error; |
428870ff | 667 | uint64_t sa_obj; |
9babb374 | 668 | |
efcd79a8 | 669 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
9babb374 | 670 | |
428870ff BB |
671 | /* |
672 | * We claim to always be readonly so we can open snapshots; | |
673 | * other ZPL code will prevent us from writing to snapshots. | |
674 | */ | |
3558fd73 | 675 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); |
428870ff | 676 | if (error) { |
3558fd73 | 677 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 | 678 | return (error); |
428870ff | 679 | } |
9babb374 | 680 | |
0282c413 BB |
681 | /* |
682 | * Optional temporary mount options, free'd in zfs_sb_free(). | |
683 | */ | |
684 | zsb->z_mntopts = (zmo ? zmo : zfs_mntopts_alloc()); | |
685 | ||
9babb374 BB |
686 | /* |
687 | * Initialize the zfs-specific filesystem structure. | |
c96c36fa | 688 | * Should probably make this a kmem cache, shuffle fields. |
9babb374 | 689 | */ |
2cf7f52b | 690 | zsb->z_sb = NULL; |
3558fd73 | 691 | zsb->z_parent = zsb; |
f1512ee6 | 692 | zsb->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; |
3558fd73 BB |
693 | zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; |
694 | zsb->z_os = os; | |
9babb374 | 695 | |
3558fd73 | 696 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); |
9babb374 BB |
697 | if (error) { |
698 | goto out; | |
287be44f | 699 | } else if (zsb->z_version > ZPL_VERSION) { |
2e528b49 | 700 | error = SET_ERROR(ENOTSUP); |
9babb374 BB |
701 | goto out; |
702 | } | |
9babb374 BB |
703 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) |
704 | goto out; | |
3558fd73 | 705 | zsb->z_norm = (int)zval; |
9babb374 BB |
706 | |
707 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
708 | goto out; | |
3558fd73 | 709 | zsb->z_utf8 = (zval != 0); |
9babb374 BB |
710 | |
711 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
712 | goto out; | |
3558fd73 | 713 | zsb->z_case = (uint_t)zval; |
9babb374 | 714 | |
023699cd MM |
715 | if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0) |
716 | goto out; | |
717 | zsb->z_acl_type = (uint_t)zval; | |
718 | ||
9babb374 BB |
719 | /* |
720 | * Fold case on file systems that are always or sometimes case | |
721 | * insensitive. | |
722 | */ | |
3558fd73 BB |
723 | if (zsb->z_case == ZFS_CASE_INSENSITIVE || |
724 | zsb->z_case == ZFS_CASE_MIXED) | |
725 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 726 | |
3558fd73 BB |
727 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
728 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
428870ff | 729 | |
3558fd73 | 730 | if (zsb->z_use_sa) { |
428870ff BB |
731 | /* should either have both of these objects or none */ |
732 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
733 | &sa_obj); | |
734 | if (error) | |
591fb62f | 735 | goto out; |
82a37189 BB |
736 | |
737 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval); | |
738 | if ((error == 0) && (zval == ZFS_XATTR_SA)) | |
739 | zsb->z_xattr_sa = B_TRUE; | |
428870ff BB |
740 | } else { |
741 | /* | |
742 | * Pre SA versions file systems should never touch | |
743 | * either the attribute registration or layout objects. | |
744 | */ | |
745 | sa_obj = 0; | |
746 | } | |
747 | ||
572e2857 | 748 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
3558fd73 | 749 | &zsb->z_attr_table); |
572e2857 BB |
750 | if (error) |
751 | goto out; | |
428870ff | 752 | |
3558fd73 | 753 | if (zsb->z_version >= ZPL_VERSION_SA) |
428870ff | 754 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
755 | |
756 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
3558fd73 | 757 | &zsb->z_root); |
9babb374 BB |
758 | if (error) |
759 | goto out; | |
3558fd73 | 760 | ASSERT(zsb->z_root != 0); |
9babb374 BB |
761 | |
762 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
3558fd73 | 763 | &zsb->z_unlinkedobj); |
9babb374 BB |
764 | if (error) |
765 | goto out; | |
766 | ||
767 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
768 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
3558fd73 | 769 | 8, 1, &zsb->z_userquota_obj); |
9babb374 BB |
770 | if (error && error != ENOENT) |
771 | goto out; | |
772 | ||
773 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
774 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
3558fd73 | 775 | 8, 1, &zsb->z_groupquota_obj); |
9babb374 BB |
776 | if (error && error != ENOENT) |
777 | goto out; | |
778 | ||
779 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
3558fd73 | 780 | &zsb->z_fuid_obj); |
9babb374 BB |
781 | if (error && error != ENOENT) |
782 | goto out; | |
783 | ||
784 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
3558fd73 | 785 | &zsb->z_shares_dir); |
9babb374 BB |
786 | if (error && error != ENOENT) |
787 | goto out; | |
788 | ||
3558fd73 BB |
789 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
790 | mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
791 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
9babb374 | 792 | offsetof(znode_t, z_link_node)); |
e16b3fcc | 793 | rrm_init(&zsb->z_teardown_lock, B_FALSE); |
3558fd73 BB |
794 | rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); |
795 | rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
d07b7c7f | 796 | |
c96c36fa BB |
797 | size = MIN(1 << (highbit64(zfs_object_mutex_size)-1), ZFS_OBJ_MTX_MAX); |
798 | zsb->z_hold_size = size; | |
799 | zsb->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, KM_SLEEP); | |
800 | zsb->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
801 | for (i = 0; i != size; i++) { | |
802 | avl_create(&zsb->z_hold_trees[i], zfs_znode_hold_compare, | |
803 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); | |
804 | mutex_init(&zsb->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); | |
805 | } | |
9babb374 | 806 | |
3558fd73 | 807 | *zsbp = zsb; |
9babb374 BB |
808 | return (0); |
809 | ||
810 | out: | |
3558fd73 BB |
811 | dmu_objset_disown(os, zsb); |
812 | *zsbp = NULL; | |
d07b7c7f | 813 | |
3558fd73 | 814 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 BB |
815 | return (error); |
816 | } | |
86f35f34 | 817 | EXPORT_SYMBOL(zfs_sb_create); |
9babb374 | 818 | |
86f35f34 | 819 | int |
3558fd73 | 820 | zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) |
34dc7c2f | 821 | { |
34dc7c2f BB |
822 | int error; |
823 | ||
3558fd73 | 824 | error = zfs_register_callbacks(zsb); |
34dc7c2f BB |
825 | if (error) |
826 | return (error); | |
827 | ||
828 | /* | |
3558fd73 | 829 | * Set the objset user_ptr to track its zsb. |
34dc7c2f | 830 | */ |
3558fd73 BB |
831 | mutex_enter(&zsb->z_os->os_user_ptr_lock); |
832 | dmu_objset_set_user(zsb->z_os, zsb); | |
833 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
34dc7c2f | 834 | |
3558fd73 | 835 | zsb->z_log = zil_open(zsb->z_os, zfs_get_data); |
9babb374 | 836 | |
34dc7c2f BB |
837 | /* |
838 | * If we are not mounting (ie: online recv), then we don't | |
839 | * have to worry about replaying the log as we blocked all | |
840 | * operations out since we closed the ZIL. | |
841 | */ | |
842 | if (mounting) { | |
b128c09f BB |
843 | boolean_t readonly; |
844 | ||
34dc7c2f BB |
845 | /* |
846 | * During replay we remove the read only flag to | |
847 | * allow replays to succeed. | |
848 | */ | |
2cf7f52b | 849 | readonly = zfs_is_readonly(zsb); |
fb5f0bc8 | 850 | if (readonly != 0) |
2cf7f52b | 851 | readonly_changed_cb(zsb, B_FALSE); |
fb5f0bc8 | 852 | else |
3558fd73 | 853 | zfs_unlinked_drain(zsb); |
34dc7c2f | 854 | |
428870ff BB |
855 | /* |
856 | * Parse and replay the intent log. | |
857 | * | |
858 | * Because of ziltest, this must be done after | |
859 | * zfs_unlinked_drain(). (Further note: ziltest | |
860 | * doesn't use readonly mounts, where | |
861 | * zfs_unlinked_drain() isn't called.) This is because | |
862 | * ziltest causes spa_sync() to think it's committed, | |
863 | * but actually it is not, so the intent log contains | |
864 | * many txg's worth of changes. | |
865 | * | |
866 | * In particular, if object N is in the unlinked set in | |
867 | * the last txg to actually sync, then it could be | |
868 | * actually freed in a later txg and then reallocated | |
869 | * in a yet later txg. This would write a "create | |
870 | * object N" record to the intent log. Normally, this | |
871 | * would be fine because the spa_sync() would have | |
872 | * written out the fact that object N is free, before | |
873 | * we could write the "create object N" intent log | |
874 | * record. | |
875 | * | |
876 | * But when we are in ziltest mode, we advance the "open | |
877 | * txg" without actually spa_sync()-ing the changes to | |
878 | * disk. So we would see that object N is still | |
879 | * allocated and in the unlinked set, and there is an | |
880 | * intent log record saying to allocate it. | |
881 | */ | |
3558fd73 | 882 | if (spa_writeable(dmu_objset_spa(zsb->z_os))) { |
572e2857 | 883 | if (zil_replay_disable) { |
3558fd73 | 884 | zil_destroy(zsb->z_log, B_FALSE); |
572e2857 | 885 | } else { |
3558fd73 BB |
886 | zsb->z_replay = B_TRUE; |
887 | zil_replay(zsb->z_os, zsb, | |
572e2857 | 888 | zfs_replay_vector); |
3558fd73 | 889 | zsb->z_replay = B_FALSE; |
572e2857 | 890 | } |
fb5f0bc8 | 891 | } |
2cf7f52b BB |
892 | |
893 | /* restore readonly bit */ | |
894 | if (readonly != 0) | |
895 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
896 | } |
897 | ||
34dc7c2f BB |
898 | return (0); |
899 | } | |
86f35f34 | 900 | EXPORT_SYMBOL(zfs_sb_setup); |
34dc7c2f | 901 | |
9babb374 | 902 | void |
3558fd73 | 903 | zfs_sb_free(zfs_sb_t *zsb) |
34dc7c2f | 904 | { |
c96c36fa | 905 | int i, size = zsb->z_hold_size; |
9babb374 | 906 | |
3558fd73 | 907 | zfs_fuid_destroy(zsb); |
9babb374 | 908 | |
3558fd73 BB |
909 | mutex_destroy(&zsb->z_znodes_lock); |
910 | mutex_destroy(&zsb->z_lock); | |
911 | list_destroy(&zsb->z_all_znodes); | |
e16b3fcc | 912 | rrm_destroy(&zsb->z_teardown_lock); |
3558fd73 BB |
913 | rw_destroy(&zsb->z_teardown_inactive_lock); |
914 | rw_destroy(&zsb->z_fuid_lock); | |
c96c36fa BB |
915 | for (i = 0; i != size; i++) { |
916 | avl_destroy(&zsb->z_hold_trees[i]); | |
917 | mutex_destroy(&zsb->z_hold_locks[i]); | |
918 | } | |
919 | vmem_free(zsb->z_hold_trees, sizeof (avl_tree_t) * size); | |
920 | vmem_free(zsb->z_hold_locks, sizeof (kmutex_t) * size); | |
0282c413 | 921 | zfs_mntopts_free(zsb->z_mntopts); |
3558fd73 | 922 | kmem_free(zsb, sizeof (zfs_sb_t)); |
34dc7c2f | 923 | } |
86f35f34 | 924 | EXPORT_SYMBOL(zfs_sb_free); |
34dc7c2f | 925 | |
9babb374 | 926 | static void |
3558fd73 | 927 | zfs_set_fuid_feature(zfs_sb_t *zsb) |
9babb374 | 928 | { |
3558fd73 BB |
929 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
930 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
9babb374 | 931 | } |
34dc7c2f BB |
932 | |
933 | void | |
3558fd73 | 934 | zfs_unregister_callbacks(zfs_sb_t *zsb) |
34dc7c2f | 935 | { |
3558fd73 | 936 | objset_t *os = zsb->z_os; |
34dc7c2f | 937 | |
0eb21616 JG |
938 | if (!dmu_objset_is_snapshot(os)) |
939 | dsl_prop_unregister_all(dmu_objset_ds(os), zsb); | |
34dc7c2f | 940 | } |
e5c39b95 | 941 | EXPORT_SYMBOL(zfs_unregister_callbacks); |
34dc7c2f | 942 | |
bc3e15e3 | 943 | #ifdef HAVE_MLSLABEL |
428870ff | 944 | /* |
d3cc8b15 WA |
945 | * Check that the hex label string is appropriate for the dataset being |
946 | * mounted into the global_zone proper. | |
428870ff | 947 | * |
d3cc8b15 WA |
948 | * Return an error if the hex label string is not default or |
949 | * admin_low/admin_high. For admin_low labels, the corresponding | |
950 | * dataset must be readonly. | |
428870ff BB |
951 | */ |
952 | int | |
953 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
954 | { | |
955 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
956 | return (0); | |
957 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
958 | return (0); | |
959 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
960 | /* must be readonly */ | |
961 | uint64_t rdonly; | |
962 | ||
963 | if (dsl_prop_get_integer(dsname, | |
964 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
2e528b49 | 965 | return (SET_ERROR(EACCES)); |
428870ff BB |
966 | return (rdonly ? 0 : EACCES); |
967 | } | |
2e528b49 | 968 | return (SET_ERROR(EACCES)); |
428870ff | 969 | } |
86f35f34 | 970 | EXPORT_SYMBOL(zfs_check_global_label); |
bc3e15e3 | 971 | #endif /* HAVE_MLSLABEL */ |
428870ff | 972 | |
e5c39b95 | 973 | int |
3558fd73 | 974 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 975 | { |
3558fd73 | 976 | zfs_sb_t *zsb = dentry->d_sb->s_fs_info; |
34dc7c2f | 977 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
04f9432d | 978 | uint64_t fsid; |
3558fd73 | 979 | uint32_t bshift; |
34dc7c2f | 980 | |
3558fd73 | 981 | ZFS_ENTER(zsb); |
34dc7c2f | 982 | |
3558fd73 | 983 | dmu_objset_space(zsb->z_os, |
34dc7c2f BB |
984 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
985 | ||
04f9432d | 986 | fsid = dmu_objset_fsid_guid(zsb->z_os); |
34dc7c2f | 987 | /* |
05ff35c6 BB |
988 | * The underlying storage pool actually uses multiple block |
989 | * size. Under Solaris frsize (fragment size) is reported as | |
990 | * the smallest block size we support, and bsize (block size) | |
991 | * as the filesystem's maximum block size. Unfortunately, | |
992 | * under Linux the fragment size and block size are often used | |
993 | * interchangeably. Thus we are forced to report both of them | |
994 | * as the filesystem's maximum block size. | |
34dc7c2f | 995 | */ |
05ff35c6 | 996 | statp->f_frsize = zsb->z_max_blksz; |
3558fd73 BB |
997 | statp->f_bsize = zsb->z_max_blksz; |
998 | bshift = fls(statp->f_bsize) - 1; | |
34dc7c2f BB |
999 | |
1000 | /* | |
3558fd73 BB |
1001 | * The following report "total" blocks of various kinds in |
1002 | * the file system, but reported in terms of f_bsize - the | |
1003 | * "preferred" size. | |
34dc7c2f BB |
1004 | */ |
1005 | ||
3558fd73 BB |
1006 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
1007 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
1008 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
1009 | ||
1010 | /* | |
1011 | * statvfs() should really be called statufs(), because it assumes | |
1012 | * static metadata. ZFS doesn't preallocate files, so the best | |
1013 | * we can do is report the max that could possibly fit in f_files, | |
1014 | * and that minus the number actually used in f_ffree. | |
1015 | * For f_ffree, report the smaller of the number of object available | |
1016 | * and the number of blocks (each object will take at least a block). | |
1017 | */ | |
baab0630 | 1018 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 1019 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
1020 | statp->f_fsid.val[0] = (uint32_t)fsid; |
1021 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 | 1022 | statp->f_type = ZFS_SUPER_MAGIC; |
eca7b760 | 1023 | statp->f_namelen = MAXNAMELEN - 1; |
34dc7c2f BB |
1024 | |
1025 | /* | |
3558fd73 | 1026 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
1027 | * Is there anything useful we could/should provide? |
1028 | */ | |
3558fd73 | 1029 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 1030 | |
3558fd73 | 1031 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1032 | return (0); |
1033 | } | |
e5c39b95 | 1034 | EXPORT_SYMBOL(zfs_statvfs); |
34dc7c2f | 1035 | |
e5c39b95 | 1036 | int |
3558fd73 | 1037 | zfs_root(zfs_sb_t *zsb, struct inode **ipp) |
34dc7c2f | 1038 | { |
34dc7c2f BB |
1039 | znode_t *rootzp; |
1040 | int error; | |
1041 | ||
3558fd73 | 1042 | ZFS_ENTER(zsb); |
34dc7c2f | 1043 | |
3558fd73 | 1044 | error = zfs_zget(zsb, zsb->z_root, &rootzp); |
34dc7c2f | 1045 | if (error == 0) |
3558fd73 | 1046 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1047 | |
3558fd73 | 1048 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1049 | return (error); |
1050 | } | |
e5c39b95 | 1051 | EXPORT_SYMBOL(zfs_root); |
34dc7c2f | 1052 | |
09fb30e5 | 1053 | #ifdef HAVE_D_PRUNE_ALIASES |
218b4e0a BB |
1054 | /* |
1055 | * Linux kernels older than 3.1 do not support a per-filesystem shrinker. | |
1056 | * To accommodate this we must improvise and manually walk the list of znodes | |
1057 | * attempting to prune dentries in order to be able to drop the inodes. | |
1058 | * | |
1059 | * To avoid scanning the same znodes multiple times they are always rotated | |
1060 | * to the end of the z_all_znodes list. New znodes are inserted at the | |
1061 | * end of the list so we're always scanning the oldest znodes first. | |
1062 | */ | |
1063 | static int | |
1064 | zfs_sb_prune_aliases(zfs_sb_t *zsb, unsigned long nr_to_scan) | |
1065 | { | |
1066 | znode_t **zp_array, *zp; | |
1067 | int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *)); | |
1068 | int objects = 0; | |
1069 | int i = 0, j = 0; | |
1070 | ||
1071 | zp_array = kmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP); | |
1072 | ||
1073 | mutex_enter(&zsb->z_znodes_lock); | |
1074 | while ((zp = list_head(&zsb->z_all_znodes)) != NULL) { | |
1075 | ||
1076 | if ((i++ > nr_to_scan) || (j >= max_array)) | |
1077 | break; | |
1078 | ||
1079 | ASSERT(list_link_active(&zp->z_link_node)); | |
1080 | list_remove(&zsb->z_all_znodes, zp); | |
1081 | list_insert_tail(&zsb->z_all_znodes, zp); | |
1082 | ||
1083 | /* Skip active znodes and .zfs entries */ | |
1084 | if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir) | |
1085 | continue; | |
1086 | ||
1087 | if (igrab(ZTOI(zp)) == NULL) | |
1088 | continue; | |
1089 | ||
1090 | zp_array[j] = zp; | |
1091 | j++; | |
1092 | } | |
1093 | mutex_exit(&zsb->z_znodes_lock); | |
1094 | ||
1095 | for (i = 0; i < j; i++) { | |
1096 | zp = zp_array[i]; | |
1097 | ||
1098 | ASSERT3P(zp, !=, NULL); | |
1099 | d_prune_aliases(ZTOI(zp)); | |
1100 | ||
1101 | if (atomic_read(&ZTOI(zp)->i_count) == 1) | |
1102 | objects++; | |
1103 | ||
1104 | iput(ZTOI(zp)); | |
1105 | } | |
1106 | ||
1107 | kmem_free(zp_array, max_array * sizeof (znode_t *)); | |
1108 | ||
1109 | return (objects); | |
1110 | } | |
1111 | #endif /* HAVE_D_PRUNE_ALIASES */ | |
1112 | ||
2cbb06b5 BB |
1113 | /* |
1114 | * The ARC has requested that the filesystem drop entries from the dentry | |
1115 | * and inode caches. This can occur when the ARC needs to free meta data | |
1116 | * blocks but can't because they are all pinned by entries in these caches. | |
1117 | */ | |
ab26409d BB |
1118 | int |
1119 | zfs_sb_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) | |
1120 | { | |
1121 | zfs_sb_t *zsb = sb->s_fs_info; | |
2cbb06b5 BB |
1122 | int error = 0; |
1123 | #if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK) | |
ab26409d BB |
1124 | struct shrinker *shrinker = &sb->s_shrink; |
1125 | struct shrink_control sc = { | |
1126 | .nr_to_scan = nr_to_scan, | |
1127 | .gfp_mask = GFP_KERNEL, | |
1128 | }; | |
2cbb06b5 | 1129 | #endif |
ab26409d BB |
1130 | |
1131 | ZFS_ENTER(zsb); | |
2cbb06b5 | 1132 | |
90947b23 TC |
1133 | #if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \ |
1134 | defined(SHRINK_CONTROL_HAS_NID) && \ | |
1135 | defined(SHRINKER_NUMA_AWARE) | |
1136 | if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) { | |
1137 | *objects = 0; | |
1138 | for_each_online_node(sc.nid) | |
1139 | *objects += (*shrinker->scan_objects)(shrinker, &sc); | |
1140 | } else { | |
1141 | *objects = (*shrinker->scan_objects)(shrinker, &sc); | |
1142 | } | |
09fb30e5 | 1143 | |
90947b23 | 1144 | #elif defined(HAVE_SPLIT_SHRINKER_CALLBACK) |
3c832b8c | 1145 | *objects = (*shrinker->scan_objects)(shrinker, &sc); |
2cbb06b5 | 1146 | #elif defined(HAVE_SHRINK) |
ab26409d | 1147 | *objects = (*shrinker->shrink)(shrinker, &sc); |
218b4e0a | 1148 | #elif defined(HAVE_D_PRUNE_ALIASES) |
09fb30e5 | 1149 | #define D_PRUNE_ALIASES_IS_DEFAULT |
218b4e0a | 1150 | *objects = zfs_sb_prune_aliases(zsb, nr_to_scan); |
2cbb06b5 | 1151 | #else |
218b4e0a | 1152 | #error "No available dentry and inode cache pruning mechanism." |
3c832b8c | 1153 | #endif |
09fb30e5 TC |
1154 | |
1155 | #if defined(HAVE_D_PRUNE_ALIASES) && !defined(D_PRUNE_ALIASES_IS_DEFAULT) | |
1156 | #undef D_PRUNE_ALIASES_IS_DEFAULT | |
1157 | /* | |
1158 | * Fall back to zfs_sb_prune_aliases if the kernel's per-superblock | |
1159 | * shrinker couldn't free anything, possibly due to the inodes being | |
1160 | * allocated in a different memcg. | |
1161 | */ | |
1162 | if (*objects == 0) | |
1163 | *objects = zfs_sb_prune_aliases(zsb, nr_to_scan); | |
1164 | #endif | |
1165 | ||
ab26409d BB |
1166 | ZFS_EXIT(zsb); |
1167 | ||
2cbb06b5 BB |
1168 | dprintf_ds(zsb->z_os->os_dsl_dataset, |
1169 | "pruning, nr_to_scan=%lu objects=%d error=%d\n", | |
1170 | nr_to_scan, *objects, error); | |
1171 | ||
1172 | return (error); | |
ab26409d BB |
1173 | } |
1174 | EXPORT_SYMBOL(zfs_sb_prune); | |
ab26409d | 1175 | |
34dc7c2f | 1176 | /* |
7b3e34ba | 1177 | * Teardown the zfs_sb_t. |
34dc7c2f BB |
1178 | * |
1179 | * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' | |
1180 | * and 'z_teardown_inactive_lock' held. | |
1181 | */ | |
3558fd73 | 1182 | int |
86f35f34 | 1183 | zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting) |
34dc7c2f BB |
1184 | { |
1185 | znode_t *zp; | |
1186 | ||
fd23720a BB |
1187 | /* |
1188 | * If someone has not already unmounted this file system, | |
1189 | * drain the iput_taskq to ensure all active references to the | |
1190 | * zfs_sb_t have been handled only then can it be safely destroyed. | |
1191 | */ | |
f0da4d15 CD |
1192 | if (zsb->z_os) { |
1193 | /* | |
1194 | * If we're unmounting we have to wait for the list to | |
1195 | * drain completely. | |
1196 | * | |
1197 | * If we're not unmounting there's no guarantee the list | |
1198 | * will drain completely, but iputs run from the taskq | |
1199 | * may add the parents of dir-based xattrs to the taskq | |
1200 | * so we want to wait for these. | |
1201 | * | |
1202 | * We can safely read z_nr_znodes without locking because the | |
1203 | * VFS has already blocked operations which add to the | |
1204 | * z_all_znodes list and thus increment z_nr_znodes. | |
1205 | */ | |
1206 | int round = 0; | |
1207 | while (zsb->z_nr_znodes > 0) { | |
c5528b9b BB |
1208 | taskq_wait_outstanding(dsl_pool_iput_taskq( |
1209 | dmu_objset_pool(zsb->z_os)), 0); | |
f0da4d15 CD |
1210 | if (++round > 1 && !unmounting) |
1211 | break; | |
1212 | } | |
1213 | } | |
fd23720a | 1214 | |
e16b3fcc | 1215 | rrm_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1216 | |
1217 | if (!unmounting) { | |
1218 | /* | |
ceb43b93 BB |
1219 | * We purge the parent filesystem's super block as the |
1220 | * parent filesystem and all of its snapshots have their | |
1221 | * inode's super block set to the parent's filesystem's | |
1222 | * super block. Note, 'z_parent' is self referential | |
1223 | * for non-snapshots. | |
34dc7c2f | 1224 | */ |
ceb43b93 | 1225 | shrink_dcache_sb(zsb->z_parent->z_sb); |
34dc7c2f BB |
1226 | } |
1227 | ||
1228 | /* | |
1229 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1230 | * threads are blocked as zil_close can call zfs_inactive. | |
1231 | */ | |
3558fd73 BB |
1232 | if (zsb->z_log) { |
1233 | zil_close(zsb->z_log); | |
1234 | zsb->z_log = NULL; | |
34dc7c2f BB |
1235 | } |
1236 | ||
3558fd73 | 1237 | rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1238 | |
1239 | /* | |
1240 | * If we are not unmounting (ie: online recv) and someone already | |
1241 | * unmounted this file system while we were doing the switcheroo, | |
1242 | * or a reopen of z_os failed then just bail out now. | |
1243 | */ | |
3558fd73 BB |
1244 | if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { |
1245 | rw_exit(&zsb->z_teardown_inactive_lock); | |
e16b3fcc | 1246 | rrm_exit(&zsb->z_teardown_lock, FTAG); |
2e528b49 | 1247 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1248 | } |
1249 | ||
1250 | /* | |
7b3e34ba BB |
1251 | * At this point there are no VFS ops active, and any new VFS ops |
1252 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1253 | * relevant for forced unmount). | |
34dc7c2f BB |
1254 | * |
1255 | * Release all holds on dbufs. | |
1256 | */ | |
f0da4d15 CD |
1257 | if (!unmounting) { |
1258 | mutex_enter(&zsb->z_znodes_lock); | |
1259 | for (zp = list_head(&zsb->z_all_znodes); zp != NULL; | |
1260 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
1261 | if (zp->z_sa_hdl) | |
1262 | zfs_znode_dmu_fini(zp); | |
1263 | } | |
1264 | mutex_exit(&zsb->z_znodes_lock); | |
7b3e34ba | 1265 | } |
34dc7c2f BB |
1266 | |
1267 | /* | |
7b3e34ba | 1268 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1269 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1270 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1271 | */ |
1272 | if (unmounting) { | |
3558fd73 | 1273 | zsb->z_unmounted = B_TRUE; |
e16b3fcc | 1274 | rrm_exit(&zsb->z_teardown_lock, FTAG); |
3558fd73 | 1275 | rw_exit(&zsb->z_teardown_inactive_lock); |
34dc7c2f BB |
1276 | } |
1277 | ||
1278 | /* | |
1279 | * z_os will be NULL if there was an error in attempting to reopen | |
3558fd73 BB |
1280 | * zsb, so just return as the properties had already been |
1281 | * | |
34dc7c2f BB |
1282 | * unregistered and cached data had been evicted before. |
1283 | */ | |
3558fd73 | 1284 | if (zsb->z_os == NULL) |
34dc7c2f BB |
1285 | return (0); |
1286 | ||
1287 | /* | |
1288 | * Unregister properties. | |
1289 | */ | |
3558fd73 | 1290 | zfs_unregister_callbacks(zsb); |
34dc7c2f BB |
1291 | |
1292 | /* | |
1293 | * Evict cached data | |
1294 | */ | |
04434775 MA |
1295 | if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) && |
1296 | !zfs_is_readonly(zsb)) | |
1297 | txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); | |
13fe0198 | 1298 | dmu_objset_evict_dbufs(zsb->z_os); |
34dc7c2f BB |
1299 | |
1300 | return (0); | |
1301 | } | |
86f35f34 | 1302 | EXPORT_SYMBOL(zfs_sb_teardown); |
34dc7c2f | 1303 | |
8c45def2 BB |
1304 | #if !defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER) && \ |
1305 | !defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER) | |
5547c2f1 | 1306 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); |
8c45def2 | 1307 | #endif |
76659dc1 | 1308 | |
e5c39b95 | 1309 | int |
0282c413 | 1310 | zfs_domount(struct super_block *sb, zfs_mntopts_t *zmo, int silent) |
34dc7c2f | 1311 | { |
0282c413 | 1312 | const char *osname = zmo->z_osname; |
3558fd73 BB |
1313 | zfs_sb_t *zsb; |
1314 | struct inode *root_inode; | |
1315 | uint64_t recordsize; | |
1316 | int error; | |
34dc7c2f | 1317 | |
0282c413 | 1318 | error = zfs_sb_create(osname, zmo, &zsb); |
3558fd73 BB |
1319 | if (error) |
1320 | return (error); | |
1321 | ||
1322 | if ((error = dsl_prop_get_integer(osname, "recordsize", | |
1323 | &recordsize, NULL))) | |
1324 | goto out; | |
1325 | ||
1326 | zsb->z_sb = sb; | |
3558fd73 BB |
1327 | sb->s_fs_info = zsb; |
1328 | sb->s_magic = ZFS_SUPER_MAGIC; | |
1329 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1330 | sb->s_time_gran = 1; | |
1331 | sb->s_blocksize = recordsize; | |
1332 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 BB |
1333 | zsb->z_bdi.ra_pages = 0; |
1334 | sb->s_bdi = &zsb->z_bdi; | |
1335 | ||
8c45def2 | 1336 | error = -zpl_bdi_setup_and_register(&zsb->z_bdi, "zfs"); |
5547c2f1 BB |
1337 | if (error) |
1338 | goto out; | |
3558fd73 BB |
1339 | |
1340 | /* Set callback operations for the file system. */ | |
1341 | sb->s_op = &zpl_super_operations; | |
1342 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1343 | sb->s_export_op = &zpl_export_operations; |
ee930353 BB |
1344 | #ifdef HAVE_S_D_OP |
1345 | sb->s_d_op = &zpl_dentry_operations; | |
1346 | #endif /* HAVE_S_D_OP */ | |
3558fd73 BB |
1347 | |
1348 | /* Set features for file system. */ | |
1349 | zfs_set_fuid_feature(zsb); | |
1350 | ||
1351 | if (dmu_objset_is_snapshot(zsb->z_os)) { | |
1352 | uint64_t pval; | |
1353 | ||
1354 | atime_changed_cb(zsb, B_FALSE); | |
1355 | readonly_changed_cb(zsb, B_TRUE); | |
d1d7e268 MK |
1356 | if ((error = dsl_prop_get_integer(osname, |
1357 | "xattr", &pval, NULL))) | |
3558fd73 BB |
1358 | goto out; |
1359 | xattr_changed_cb(zsb, pval); | |
d1d7e268 MK |
1360 | if ((error = dsl_prop_get_integer(osname, |
1361 | "acltype", &pval, NULL))) | |
023699cd MM |
1362 | goto out; |
1363 | acltype_changed_cb(zsb, pval); | |
3558fd73 BB |
1364 | zsb->z_issnap = B_TRUE; |
1365 | zsb->z_os->os_sync = ZFS_SYNC_DISABLED; | |
278bee93 | 1366 | zsb->z_snap_defer_time = jiffies; |
3558fd73 BB |
1367 | |
1368 | mutex_enter(&zsb->z_os->os_user_ptr_lock); | |
1369 | dmu_objset_set_user(zsb->z_os, zsb); | |
1370 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
1371 | } else { | |
1372 | error = zfs_sb_setup(zsb, B_TRUE); | |
34dc7c2f BB |
1373 | } |
1374 | ||
3558fd73 BB |
1375 | /* Allocate a root inode for the filesystem. */ |
1376 | error = zfs_root(zsb, &root_inode); | |
1377 | if (error) { | |
1378 | (void) zfs_umount(sb); | |
1379 | goto out; | |
34dc7c2f BB |
1380 | } |
1381 | ||
3558fd73 | 1382 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1383 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1384 | if (sb->s_root == NULL) { |
1385 | (void) zfs_umount(sb); | |
2e528b49 | 1386 | error = SET_ERROR(ENOMEM); |
3558fd73 BB |
1387 | goto out; |
1388 | } | |
ebe7e575 BB |
1389 | |
1390 | if (!zsb->z_issnap) | |
1391 | zfsctl_create(zsb); | |
2cbb06b5 BB |
1392 | |
1393 | zsb->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); | |
3558fd73 BB |
1394 | out: |
1395 | if (error) { | |
1396 | dmu_objset_disown(zsb->z_os, zsb); | |
1397 | zfs_sb_free(zsb); | |
1398 | } | |
34dc7c2f | 1399 | |
3558fd73 BB |
1400 | return (error); |
1401 | } | |
1402 | EXPORT_SYMBOL(zfs_domount); | |
1403 | ||
ebe7e575 BB |
1404 | /* |
1405 | * Called when an unmount is requested and certain sanity checks have | |
1406 | * already passed. At this point no dentries or inodes have been reclaimed | |
1407 | * from their respective caches. We drop the extra reference on the .zfs | |
1408 | * control directory to allow everything to be reclaimed. All snapshots | |
1409 | * must already have been unmounted to reach this point. | |
1410 | */ | |
1411 | void | |
1412 | zfs_preumount(struct super_block *sb) | |
1413 | { | |
1414 | zfs_sb_t *zsb = sb->s_fs_info; | |
1415 | ||
278bee93 BB |
1416 | if (zsb) |
1417 | zfsctl_destroy(sb->s_fs_info); | |
ebe7e575 BB |
1418 | } |
1419 | EXPORT_SYMBOL(zfs_preumount); | |
1420 | ||
1421 | /* | |
1422 | * Called once all other unmount released tear down has occurred. | |
1423 | * It is our responsibility to release any remaining infrastructure. | |
1424 | */ | |
3558fd73 BB |
1425 | /*ARGSUSED*/ |
1426 | int | |
1427 | zfs_umount(struct super_block *sb) | |
1428 | { | |
1429 | zfs_sb_t *zsb = sb->s_fs_info; | |
1430 | objset_t *os; | |
1431 | ||
2cbb06b5 | 1432 | arc_remove_prune_callback(zsb->z_arc_prune); |
86f35f34 | 1433 | VERIFY(zfs_sb_teardown(zsb, B_TRUE) == 0); |
3558fd73 | 1434 | os = zsb->z_os; |
5547c2f1 | 1435 | bdi_destroy(sb->s_bdi); |
76659dc1 | 1436 | |
34dc7c2f BB |
1437 | /* |
1438 | * z_os will be NULL if there was an error in | |
3558fd73 | 1439 | * attempting to reopen zsb. |
34dc7c2f BB |
1440 | */ |
1441 | if (os != NULL) { | |
1442 | /* | |
1443 | * Unset the objset user_ptr. | |
1444 | */ | |
428870ff | 1445 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1446 | dmu_objset_set_user(os, NULL); |
428870ff | 1447 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1448 | |
1449 | /* | |
b128c09f | 1450 | * Finally release the objset |
34dc7c2f | 1451 | */ |
3558fd73 | 1452 | dmu_objset_disown(os, zsb); |
34dc7c2f BB |
1453 | } |
1454 | ||
3558fd73 | 1455 | zfs_sb_free(zsb); |
34dc7c2f BB |
1456 | return (0); |
1457 | } | |
e5c39b95 | 1458 | EXPORT_SYMBOL(zfs_umount); |
34dc7c2f | 1459 | |
0de19dad | 1460 | int |
0282c413 | 1461 | zfs_remount(struct super_block *sb, int *flags, zfs_mntopts_t *zmo) |
0de19dad | 1462 | { |
0282c413 BB |
1463 | zfs_sb_t *zsb = sb->s_fs_info; |
1464 | int error; | |
1465 | ||
1466 | zfs_unregister_callbacks(zsb); | |
1467 | error = zfs_register_callbacks(zsb); | |
1468 | ||
1469 | return (error); | |
0de19dad BB |
1470 | } |
1471 | EXPORT_SYMBOL(zfs_remount); | |
1472 | ||
e5c39b95 | 1473 | int |
2cf7f52b | 1474 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1475 | { |
2cf7f52b | 1476 | zfs_sb_t *zsb = sb->s_fs_info; |
34dc7c2f BB |
1477 | znode_t *zp; |
1478 | uint64_t object = 0; | |
1479 | uint64_t fid_gen = 0; | |
1480 | uint64_t gen_mask; | |
1481 | uint64_t zp_gen; | |
3558fd73 | 1482 | int i, err; |
34dc7c2f | 1483 | |
3558fd73 | 1484 | *ipp = NULL; |
34dc7c2f | 1485 | |
3558fd73 | 1486 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1487 | |
1488 | if (fidp->fid_len == LONG_FID_LEN) { | |
1489 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1490 | uint64_t objsetid = 0; | |
1491 | uint64_t setgen = 0; | |
1492 | ||
1493 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1494 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1495 | ||
1496 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1497 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1498 | ||
3558fd73 | 1499 | ZFS_EXIT(zsb); |
34dc7c2f | 1500 | |
ebe7e575 | 1501 | err = zfsctl_lookup_objset(sb, objsetid, &zsb); |
34dc7c2f | 1502 | if (err) |
2e528b49 | 1503 | return (SET_ERROR(EINVAL)); |
ebe7e575 | 1504 | |
3558fd73 | 1505 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1506 | } |
1507 | ||
1508 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { | |
1509 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1510 | ||
1511 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1512 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
1513 | ||
1514 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
1515 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1516 | } else { | |
3558fd73 | 1517 | ZFS_EXIT(zsb); |
2e528b49 | 1518 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1519 | } |
1520 | ||
1521 | /* A zero fid_gen means we are in the .zfs control directories */ | |
1522 | if (fid_gen == 0 && | |
1523 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
3558fd73 BB |
1524 | *ipp = zsb->z_ctldir; |
1525 | ASSERT(*ipp != NULL); | |
34dc7c2f | 1526 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1527 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1528 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1529 | } else { |
3558fd73 | 1530 | igrab(*ipp); |
34dc7c2f | 1531 | } |
3558fd73 | 1532 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1533 | return (0); |
1534 | } | |
1535 | ||
1536 | gen_mask = -1ULL >> (64 - 8 * i); | |
1537 | ||
29e57d15 | 1538 | dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); |
3558fd73 BB |
1539 | if ((err = zfs_zget(zsb, object, &zp))) { |
1540 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1541 | return (err); |
1542 | } | |
7938c2ac CC |
1543 | |
1544 | /* Don't export xattr stuff */ | |
1545 | if (zp->z_pflags & ZFS_XATTR) { | |
1546 | iput(ZTOI(zp)); | |
1547 | ZFS_EXIT(zsb); | |
1548 | return (SET_ERROR(ENOENT)); | |
1549 | } | |
1550 | ||
3558fd73 | 1551 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, |
428870ff BB |
1552 | sizeof (uint64_t)); |
1553 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1554 | if (zp_gen == 0) |
1555 | zp_gen = 1; | |
0500e835 BB |
1556 | if ((fid_gen == 0) && (zsb->z_root == object)) |
1557 | fid_gen = zp_gen; | |
34dc7c2f | 1558 | if (zp->z_unlinked || zp_gen != fid_gen) { |
29e57d15 NB |
1559 | dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, |
1560 | fid_gen); | |
3558fd73 BB |
1561 | iput(ZTOI(zp)); |
1562 | ZFS_EXIT(zsb); | |
2e528b49 | 1563 | return (SET_ERROR(EINVAL)); |
34dc7c2f BB |
1564 | } |
1565 | ||
3558fd73 BB |
1566 | *ipp = ZTOI(zp); |
1567 | if (*ipp) | |
1568 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1569 | |
3558fd73 | 1570 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1571 | return (0); |
1572 | } | |
e5c39b95 | 1573 | EXPORT_SYMBOL(zfs_vget); |
34dc7c2f BB |
1574 | |
1575 | /* | |
7b3e34ba | 1576 | * Block out VFS ops and close zfs_sb_t |
34dc7c2f BB |
1577 | * |
1578 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
831baf06 KW |
1579 | * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying |
1580 | * dataset and objset intact so that they can be atomically handed off during | |
1581 | * a subsequent rollback or recv operation and the resume thereafter. | |
34dc7c2f BB |
1582 | */ |
1583 | int | |
3558fd73 | 1584 | zfs_suspend_fs(zfs_sb_t *zsb) |
34dc7c2f BB |
1585 | { |
1586 | int error; | |
1587 | ||
86f35f34 | 1588 | if ((error = zfs_sb_teardown(zsb, B_FALSE)) != 0) |
34dc7c2f | 1589 | return (error); |
7b3e34ba | 1590 | |
34dc7c2f BB |
1591 | return (0); |
1592 | } | |
e5c39b95 | 1593 | EXPORT_SYMBOL(zfs_suspend_fs); |
34dc7c2f BB |
1594 | |
1595 | /* | |
7b3e34ba | 1596 | * Reopen zfs_sb_t and release VFS ops. |
34dc7c2f BB |
1597 | */ |
1598 | int | |
3558fd73 | 1599 | zfs_resume_fs(zfs_sb_t *zsb, const char *osname) |
34dc7c2f | 1600 | { |
2008e920 | 1601 | int err, err2; |
831baf06 KW |
1602 | znode_t *zp; |
1603 | uint64_t sa_obj = 0; | |
34dc7c2f | 1604 | |
e16b3fcc | 1605 | ASSERT(RRM_WRITE_HELD(&zsb->z_teardown_lock)); |
3558fd73 | 1606 | ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); |
34dc7c2f | 1607 | |
831baf06 KW |
1608 | /* |
1609 | * We already own this, so just hold and rele it to update the | |
1610 | * objset_t, as the one we had before may have been evicted. | |
1611 | */ | |
1612 | VERIFY0(dmu_objset_hold(osname, zsb, &zsb->z_os)); | |
1613 | VERIFY3P(zsb->z_os->os_dsl_dataset->ds_owner, ==, zsb); | |
1614 | VERIFY(dsl_dataset_long_held(zsb->z_os->os_dsl_dataset)); | |
1615 | dmu_objset_rele(zsb->z_os, zsb); | |
428870ff | 1616 | |
831baf06 KW |
1617 | /* |
1618 | * Make sure version hasn't changed | |
1619 | */ | |
428870ff | 1620 | |
831baf06 KW |
1621 | err = zfs_get_zplprop(zsb->z_os, ZFS_PROP_VERSION, |
1622 | &zsb->z_version); | |
c1fabe79 | 1623 | |
831baf06 KW |
1624 | if (err) |
1625 | goto bail; | |
428870ff | 1626 | |
831baf06 KW |
1627 | err = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, |
1628 | ZFS_SA_ATTRS, 8, 1, &sa_obj); | |
c1fabe79 | 1629 | |
831baf06 KW |
1630 | if (err && zsb->z_version >= ZPL_VERSION_SA) |
1631 | goto bail; | |
428870ff | 1632 | |
831baf06 KW |
1633 | if ((err = sa_setup(zsb->z_os, sa_obj, |
1634 | zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) | |
1635 | goto bail; | |
34dc7c2f | 1636 | |
831baf06 KW |
1637 | if (zsb->z_version >= ZPL_VERSION_SA) |
1638 | sa_register_update_callback(zsb->z_os, | |
1639 | zfs_sa_upgrade); | |
c1fabe79 | 1640 | |
831baf06 | 1641 | VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); |
c1fabe79 | 1642 | |
831baf06 KW |
1643 | zfs_set_fuid_feature(zsb); |
1644 | zsb->z_rollback_time = jiffies; | |
34dc7c2f | 1645 | |
831baf06 KW |
1646 | /* |
1647 | * Attempt to re-establish all the active inodes with their | |
1648 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
1649 | * and mark it stale. This prevents a collision if a new | |
1650 | * inode/object is created which must use the same inode | |
1651 | * number. The stale inode will be be released when the | |
1652 | * VFS prunes the dentry holding the remaining references | |
1653 | * on the stale inode. | |
1654 | */ | |
1655 | mutex_enter(&zsb->z_znodes_lock); | |
1656 | for (zp = list_head(&zsb->z_all_znodes); zp; | |
1657 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
2008e920 TC |
1658 | err2 = zfs_rezget(zp); |
1659 | if (err2) { | |
831baf06 KW |
1660 | remove_inode_hash(ZTOI(zp)); |
1661 | zp->z_is_stale = B_TRUE; | |
34dc7c2f | 1662 | } |
34dc7c2f | 1663 | } |
831baf06 | 1664 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f | 1665 | |
428870ff | 1666 | bail: |
7b3e34ba | 1667 | /* release the VFS ops */ |
3558fd73 | 1668 | rw_exit(&zsb->z_teardown_inactive_lock); |
e16b3fcc | 1669 | rrm_exit(&zsb->z_teardown_lock, FTAG); |
34dc7c2f BB |
1670 | |
1671 | if (err) { | |
1672 | /* | |
831baf06 KW |
1673 | * Since we couldn't setup the sa framework, try to force |
1674 | * unmount this file system. | |
34dc7c2f | 1675 | */ |
ab68b6e5 TC |
1676 | if (zsb->z_os) |
1677 | (void) zfs_umount(zsb->z_sb); | |
34dc7c2f BB |
1678 | } |
1679 | return (err); | |
1680 | } | |
e5c39b95 | 1681 | EXPORT_SYMBOL(zfs_resume_fs); |
34dc7c2f | 1682 | |
34dc7c2f | 1683 | int |
3558fd73 | 1684 | zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) |
34dc7c2f BB |
1685 | { |
1686 | int error; | |
3558fd73 | 1687 | objset_t *os = zsb->z_os; |
34dc7c2f | 1688 | dmu_tx_t *tx; |
34dc7c2f BB |
1689 | |
1690 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
2e528b49 | 1691 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1692 | |
3558fd73 | 1693 | if (newvers < zsb->z_version) |
2e528b49 | 1694 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1695 | |
428870ff | 1696 | if (zfs_spa_version_map(newvers) > |
3558fd73 | 1697 | spa_version(dmu_objset_spa(zsb->z_os))) |
2e528b49 | 1698 | return (SET_ERROR(ENOTSUP)); |
428870ff | 1699 | |
34dc7c2f | 1700 | tx = dmu_tx_create(os); |
9babb374 | 1701 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
3558fd73 | 1702 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1703 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1704 | ZFS_SA_ATTRS); | |
1705 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
1706 | } | |
34dc7c2f BB |
1707 | error = dmu_tx_assign(tx, TXG_WAIT); |
1708 | if (error) { | |
1709 | dmu_tx_abort(tx); | |
9babb374 BB |
1710 | return (error); |
1711 | } | |
428870ff | 1712 | |
9babb374 BB |
1713 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
1714 | 8, 1, &newvers, tx); | |
1715 | ||
1716 | if (error) { | |
1717 | dmu_tx_commit(tx); | |
1718 | return (error); | |
34dc7c2f | 1719 | } |
34dc7c2f | 1720 | |
3558fd73 | 1721 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1722 | uint64_t sa_obj; |
1723 | ||
3558fd73 | 1724 | ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, |
428870ff BB |
1725 | SPA_VERSION_SA); |
1726 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1727 | DMU_OT_NONE, 0, tx); | |
1728 | ||
1729 | error = zap_add(os, MASTER_NODE_OBJ, | |
1730 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c99c9001 | 1731 | ASSERT0(error); |
428870ff BB |
1732 | |
1733 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
1734 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1735 | } | |
1736 | ||
6f1ffb06 MA |
1737 | spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, |
1738 | "from %llu to %llu", zsb->z_version, newvers); | |
9babb374 | 1739 | |
34dc7c2f BB |
1740 | dmu_tx_commit(tx); |
1741 | ||
3558fd73 | 1742 | zsb->z_version = newvers; |
9babb374 | 1743 | |
c1fabe79 | 1744 | zfs_set_fuid_feature(zsb); |
9babb374 BB |
1745 | |
1746 | return (0); | |
34dc7c2f | 1747 | } |
e5c39b95 | 1748 | EXPORT_SYMBOL(zfs_set_version); |
34dc7c2f BB |
1749 | |
1750 | /* | |
1751 | * Read a property stored within the master node. | |
1752 | */ | |
1753 | int | |
1754 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
1755 | { | |
1756 | const char *pname; | |
2e528b49 | 1757 | int error = SET_ERROR(ENOENT); |
34dc7c2f BB |
1758 | |
1759 | /* | |
1760 | * Look up the file system's value for the property. For the | |
1761 | * version property, we look up a slightly different string. | |
1762 | */ | |
1763 | if (prop == ZFS_PROP_VERSION) | |
1764 | pname = ZPL_VERSION_STR; | |
1765 | else | |
1766 | pname = zfs_prop_to_name(prop); | |
1767 | ||
b128c09f BB |
1768 | if (os != NULL) |
1769 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
1770 | |
1771 | if (error == ENOENT) { | |
1772 | /* No value set, use the default value */ | |
1773 | switch (prop) { | |
1774 | case ZFS_PROP_VERSION: | |
1775 | *value = ZPL_VERSION; | |
1776 | break; | |
1777 | case ZFS_PROP_NORMALIZE: | |
1778 | case ZFS_PROP_UTF8ONLY: | |
1779 | *value = 0; | |
1780 | break; | |
1781 | case ZFS_PROP_CASE: | |
1782 | *value = ZFS_CASE_SENSITIVE; | |
1783 | break; | |
023699cd MM |
1784 | case ZFS_PROP_ACLTYPE: |
1785 | *value = ZFS_ACLTYPE_OFF; | |
1786 | break; | |
34dc7c2f BB |
1787 | default: |
1788 | return (error); | |
1789 | } | |
1790 | error = 0; | |
1791 | } | |
1792 | return (error); | |
1793 | } | |
86f35f34 | 1794 | EXPORT_SYMBOL(zfs_get_zplprop); |
3558fd73 BB |
1795 | |
1796 | void | |
1797 | zfs_init(void) | |
1798 | { | |
ebe7e575 | 1799 | zfsctl_init(); |
3558fd73 BB |
1800 | zfs_znode_init(); |
1801 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
1802 | register_filesystem(&zpl_fs_type); | |
1803 | } | |
1804 | ||
1805 | void | |
1806 | zfs_fini(void) | |
1807 | { | |
c5528b9b | 1808 | taskq_wait_outstanding(system_taskq, 0); |
3558fd73 BB |
1809 | unregister_filesystem(&zpl_fs_type); |
1810 | zfs_znode_fini(); | |
ebe7e575 | 1811 | zfsctl_fini(); |
3558fd73 | 1812 | } |