]>
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. |
8614ddf9 | 23 | * Copyright (c) 2012, 2015 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 | |
1c2555ef BB |
71 | enum { |
72 | TOKEN_RO, | |
73 | TOKEN_RW, | |
74 | TOKEN_SETUID, | |
75 | TOKEN_NOSETUID, | |
76 | TOKEN_EXEC, | |
77 | TOKEN_NOEXEC, | |
78 | TOKEN_DEVICES, | |
79 | TOKEN_NODEVICES, | |
80 | TOKEN_DIRXATTR, | |
81 | TOKEN_SAXATTR, | |
82 | TOKEN_XATTR, | |
83 | TOKEN_NOXATTR, | |
84 | TOKEN_ATIME, | |
85 | TOKEN_NOATIME, | |
86 | TOKEN_RELATIME, | |
87 | TOKEN_NORELATIME, | |
88 | TOKEN_NBMAND, | |
89 | TOKEN_NONBMAND, | |
90 | TOKEN_MNTPOINT, | |
91 | TOKEN_LAST, | |
92 | }; | |
93 | ||
94 | static const match_table_t zpl_tokens = { | |
95 | { TOKEN_RO, MNTOPT_RO }, | |
96 | { TOKEN_RW, MNTOPT_RW }, | |
97 | { TOKEN_SETUID, MNTOPT_SETUID }, | |
98 | { TOKEN_NOSETUID, MNTOPT_NOSETUID }, | |
99 | { TOKEN_EXEC, MNTOPT_EXEC }, | |
100 | { TOKEN_NOEXEC, MNTOPT_NOEXEC }, | |
101 | { TOKEN_DEVICES, MNTOPT_DEVICES }, | |
102 | { TOKEN_NODEVICES, MNTOPT_NODEVICES }, | |
103 | { TOKEN_DIRXATTR, MNTOPT_DIRXATTR }, | |
104 | { TOKEN_SAXATTR, MNTOPT_SAXATTR }, | |
105 | { TOKEN_XATTR, MNTOPT_XATTR }, | |
106 | { TOKEN_NOXATTR, MNTOPT_NOXATTR }, | |
107 | { TOKEN_ATIME, MNTOPT_ATIME }, | |
108 | { TOKEN_NOATIME, MNTOPT_NOATIME }, | |
109 | { TOKEN_RELATIME, MNTOPT_RELATIME }, | |
110 | { TOKEN_NORELATIME, MNTOPT_NORELATIME }, | |
111 | { TOKEN_NBMAND, MNTOPT_NBMAND }, | |
112 | { TOKEN_NONBMAND, MNTOPT_NONBMAND }, | |
113 | { TOKEN_MNTPOINT, MNTOPT_MNTPOINT "=%s" }, | |
114 | { TOKEN_LAST, NULL }, | |
115 | }; | |
116 | ||
117 | static void | |
118 | zfsvfs_vfs_free(vfs_t *vfsp) | |
119 | { | |
120 | if (vfsp != NULL) { | |
121 | if (vfsp->vfs_mntpoint != NULL) | |
122 | strfree(vfsp->vfs_mntpoint); | |
123 | ||
124 | kmem_free(vfsp, sizeof (vfs_t)); | |
125 | } | |
126 | } | |
127 | ||
128 | static int | |
129 | zfsvfs_parse_option(char *option, int token, substring_t *args, vfs_t *vfsp) | |
130 | { | |
131 | switch (token) { | |
132 | case TOKEN_RO: | |
133 | vfsp->vfs_readonly = B_TRUE; | |
134 | vfsp->vfs_do_readonly = B_TRUE; | |
135 | break; | |
136 | case TOKEN_RW: | |
137 | vfsp->vfs_readonly = B_FALSE; | |
138 | vfsp->vfs_do_readonly = B_TRUE; | |
139 | break; | |
140 | case TOKEN_SETUID: | |
141 | vfsp->vfs_setuid = B_TRUE; | |
142 | vfsp->vfs_do_setuid = B_TRUE; | |
143 | break; | |
144 | case TOKEN_NOSETUID: | |
145 | vfsp->vfs_setuid = B_FALSE; | |
146 | vfsp->vfs_do_setuid = B_TRUE; | |
147 | break; | |
148 | case TOKEN_EXEC: | |
149 | vfsp->vfs_exec = B_TRUE; | |
150 | vfsp->vfs_do_exec = B_TRUE; | |
151 | break; | |
152 | case TOKEN_NOEXEC: | |
153 | vfsp->vfs_exec = B_FALSE; | |
154 | vfsp->vfs_do_exec = B_TRUE; | |
155 | break; | |
156 | case TOKEN_DEVICES: | |
157 | vfsp->vfs_devices = B_TRUE; | |
158 | vfsp->vfs_do_devices = B_TRUE; | |
159 | break; | |
160 | case TOKEN_NODEVICES: | |
161 | vfsp->vfs_devices = B_FALSE; | |
162 | vfsp->vfs_do_devices = B_TRUE; | |
163 | break; | |
164 | case TOKEN_DIRXATTR: | |
165 | vfsp->vfs_xattr = ZFS_XATTR_DIR; | |
166 | vfsp->vfs_do_xattr = B_TRUE; | |
167 | break; | |
168 | case TOKEN_SAXATTR: | |
169 | vfsp->vfs_xattr = ZFS_XATTR_SA; | |
170 | vfsp->vfs_do_xattr = B_TRUE; | |
171 | break; | |
172 | case TOKEN_XATTR: | |
173 | vfsp->vfs_xattr = ZFS_XATTR_DIR; | |
174 | vfsp->vfs_do_xattr = B_TRUE; | |
175 | break; | |
176 | case TOKEN_NOXATTR: | |
177 | vfsp->vfs_xattr = ZFS_XATTR_OFF; | |
178 | vfsp->vfs_do_xattr = B_TRUE; | |
179 | break; | |
180 | case TOKEN_ATIME: | |
181 | vfsp->vfs_atime = B_TRUE; | |
182 | vfsp->vfs_do_atime = B_TRUE; | |
183 | break; | |
184 | case TOKEN_NOATIME: | |
185 | vfsp->vfs_atime = B_FALSE; | |
186 | vfsp->vfs_do_atime = B_TRUE; | |
187 | break; | |
188 | case TOKEN_RELATIME: | |
189 | vfsp->vfs_relatime = B_TRUE; | |
190 | vfsp->vfs_do_relatime = B_TRUE; | |
191 | break; | |
192 | case TOKEN_NORELATIME: | |
193 | vfsp->vfs_relatime = B_FALSE; | |
194 | vfsp->vfs_do_relatime = B_TRUE; | |
195 | break; | |
196 | case TOKEN_NBMAND: | |
197 | vfsp->vfs_nbmand = B_TRUE; | |
198 | vfsp->vfs_do_nbmand = B_TRUE; | |
199 | break; | |
200 | case TOKEN_NONBMAND: | |
201 | vfsp->vfs_nbmand = B_FALSE; | |
202 | vfsp->vfs_do_nbmand = B_TRUE; | |
203 | break; | |
204 | case TOKEN_MNTPOINT: | |
205 | vfsp->vfs_mntpoint = match_strdup(&args[0]); | |
206 | if (vfsp->vfs_mntpoint == NULL) | |
207 | return (SET_ERROR(ENOMEM)); | |
208 | ||
209 | break; | |
210 | default: | |
211 | break; | |
212 | } | |
213 | ||
214 | return (0); | |
215 | } | |
216 | ||
217 | /* | |
218 | * Parse the raw mntopts and return a vfs_t describing the options. | |
219 | */ | |
220 | static int | |
221 | zfsvfs_parse_options(char *mntopts, vfs_t **vfsp) | |
222 | { | |
223 | vfs_t *tmp_vfsp; | |
224 | int error; | |
225 | ||
226 | tmp_vfsp = kmem_zalloc(sizeof (vfs_t), KM_SLEEP); | |
227 | ||
228 | if (mntopts != NULL) { | |
229 | substring_t args[MAX_OPT_ARGS]; | |
230 | char *tmp_mntopts, *p, *t; | |
231 | int token; | |
232 | ||
233 | tmp_mntopts = t = strdup(mntopts); | |
234 | if (tmp_mntopts == NULL) | |
235 | return (SET_ERROR(ENOMEM)); | |
236 | ||
237 | while ((p = strsep(&t, ",")) != NULL) { | |
238 | if (!*p) | |
239 | continue; | |
240 | ||
241 | args[0].to = args[0].from = NULL; | |
242 | token = match_token(p, zpl_tokens, args); | |
243 | error = zfsvfs_parse_option(p, token, args, tmp_vfsp); | |
244 | if (error) { | |
245 | strfree(tmp_mntopts); | |
246 | zfsvfs_vfs_free(tmp_vfsp); | |
247 | return (error); | |
248 | } | |
249 | } | |
250 | ||
251 | strfree(tmp_mntopts); | |
252 | } | |
253 | ||
254 | *vfsp = tmp_vfsp; | |
255 | ||
256 | return (0); | |
257 | } | |
258 | ||
259 | boolean_t | |
260 | zfs_is_readonly(zfsvfs_t *zfsvfs) | |
261 | { | |
262 | return (!!(zfsvfs->z_sb->s_flags & MS_RDONLY)); | |
263 | } | |
264 | ||
34dc7c2f BB |
265 | /*ARGSUSED*/ |
266 | int | |
03f9ba9d | 267 | zfs_sync(struct super_block *sb, int wait, cred_t *cr) |
34dc7c2f | 268 | { |
0037b49e | 269 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
03f9ba9d | 270 | |
34dc7c2f BB |
271 | /* |
272 | * Data integrity is job one. We don't want a compromised kernel | |
273 | * writing to the storage pool, so we never sync during panic. | |
274 | */ | |
d5e53f9d | 275 | if (unlikely(oops_in_progress)) |
34dc7c2f BB |
276 | return (0); |
277 | ||
03f9ba9d BB |
278 | /* |
279 | * Semantically, the only requirement is that the sync be initiated. | |
280 | * The DMU syncs out txgs frequently, so there's nothing to do. | |
281 | */ | |
282 | if (!wait) | |
283 | return (0); | |
284 | ||
0037b49e | 285 | if (zfsvfs != NULL) { |
34dc7c2f BB |
286 | /* |
287 | * Sync a specific filesystem. | |
288 | */ | |
9babb374 | 289 | dsl_pool_t *dp; |
34dc7c2f | 290 | |
0037b49e BB |
291 | ZFS_ENTER(zfsvfs); |
292 | dp = dmu_objset_pool(zfsvfs->z_os); | |
9babb374 BB |
293 | |
294 | /* | |
295 | * If the system is shutting down, then skip any | |
296 | * filesystems which may exist on a suspended pool. | |
297 | */ | |
03f9ba9d | 298 | if (spa_suspended(dp->dp_spa)) { |
0037b49e | 299 | ZFS_EXIT(zfsvfs); |
9babb374 BB |
300 | return (0); |
301 | } | |
302 | ||
0037b49e BB |
303 | if (zfsvfs->z_log != NULL) |
304 | zil_commit(zfsvfs->z_log, 0); | |
428870ff | 305 | |
0037b49e | 306 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
307 | } else { |
308 | /* | |
309 | * Sync all ZFS filesystems. This is what happens when you | |
310 | * run sync(1M). Unlike other filesystems, ZFS honors the | |
311 | * request by waiting for all pools to commit all dirty data. | |
312 | */ | |
313 | spa_sync_allpools(); | |
314 | } | |
315 | ||
316 | return (0); | |
317 | } | |
318 | ||
34dc7c2f BB |
319 | static void |
320 | atime_changed_cb(void *arg, uint64_t newval) | |
321 | { | |
0037b49e | 322 | ((zfsvfs_t *)arg)->z_atime = newval; |
34dc7c2f BB |
323 | } |
324 | ||
6d111134 TC |
325 | static void |
326 | relatime_changed_cb(void *arg, uint64_t newval) | |
327 | { | |
0037b49e | 328 | ((zfsvfs_t *)arg)->z_relatime = newval; |
6d111134 TC |
329 | } |
330 | ||
34dc7c2f BB |
331 | static void |
332 | xattr_changed_cb(void *arg, uint64_t newval) | |
333 | { | |
0037b49e | 334 | zfsvfs_t *zfsvfs = arg; |
34dc7c2f | 335 | |
82a37189 | 336 | if (newval == ZFS_XATTR_OFF) { |
0037b49e | 337 | zfsvfs->z_flags &= ~ZSB_XATTR; |
82a37189 | 338 | } else { |
0037b49e | 339 | zfsvfs->z_flags |= ZSB_XATTR; |
82a37189 BB |
340 | |
341 | if (newval == ZFS_XATTR_SA) | |
0037b49e | 342 | zfsvfs->z_xattr_sa = B_TRUE; |
82a37189 | 343 | else |
0037b49e | 344 | zfsvfs->z_xattr_sa = B_FALSE; |
82a37189 | 345 | } |
34dc7c2f BB |
346 | } |
347 | ||
023699cd MM |
348 | static void |
349 | acltype_changed_cb(void *arg, uint64_t newval) | |
350 | { | |
0037b49e | 351 | zfsvfs_t *zfsvfs = arg; |
023699cd MM |
352 | |
353 | switch (newval) { | |
354 | case ZFS_ACLTYPE_OFF: | |
0037b49e BB |
355 | zfsvfs->z_acl_type = ZFS_ACLTYPE_OFF; |
356 | zfsvfs->z_sb->s_flags &= ~MS_POSIXACL; | |
023699cd MM |
357 | break; |
358 | case ZFS_ACLTYPE_POSIXACL: | |
b695c34e | 359 | #ifdef CONFIG_FS_POSIX_ACL |
0037b49e BB |
360 | zfsvfs->z_acl_type = ZFS_ACLTYPE_POSIXACL; |
361 | zfsvfs->z_sb->s_flags |= MS_POSIXACL; | |
b695c34e | 362 | #else |
0037b49e BB |
363 | zfsvfs->z_acl_type = ZFS_ACLTYPE_OFF; |
364 | zfsvfs->z_sb->s_flags &= ~MS_POSIXACL; | |
b695c34e | 365 | #endif /* CONFIG_FS_POSIX_ACL */ |
023699cd MM |
366 | break; |
367 | default: | |
368 | break; | |
369 | } | |
370 | } | |
371 | ||
34dc7c2f BB |
372 | static void |
373 | blksz_changed_cb(void *arg, uint64_t newval) | |
374 | { | |
0037b49e BB |
375 | zfsvfs_t *zfsvfs = arg; |
376 | ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zfsvfs->z_os))); | |
f1512ee6 MA |
377 | ASSERT3U(newval, >=, SPA_MINBLOCKSIZE); |
378 | ASSERT(ISP2(newval)); | |
34dc7c2f | 379 | |
0037b49e | 380 | zfsvfs->z_max_blksz = newval; |
34dc7c2f BB |
381 | } |
382 | ||
383 | static void | |
384 | readonly_changed_cb(void *arg, uint64_t newval) | |
385 | { | |
0037b49e BB |
386 | zfsvfs_t *zfsvfs = arg; |
387 | struct super_block *sb = zfsvfs->z_sb; | |
34dc7c2f | 388 | |
2cf7f52b BB |
389 | if (sb == NULL) |
390 | return; | |
391 | ||
392 | if (newval) | |
3558fd73 | 393 | sb->s_flags |= MS_RDONLY; |
2cf7f52b | 394 | else |
3558fd73 | 395 | sb->s_flags &= ~MS_RDONLY; |
34dc7c2f BB |
396 | } |
397 | ||
398 | static void | |
399 | devices_changed_cb(void *arg, uint64_t newval) | |
400 | { | |
34dc7c2f BB |
401 | } |
402 | ||
403 | static void | |
404 | setuid_changed_cb(void *arg, uint64_t newval) | |
405 | { | |
34dc7c2f BB |
406 | } |
407 | ||
408 | static void | |
409 | exec_changed_cb(void *arg, uint64_t newval) | |
410 | { | |
34dc7c2f BB |
411 | } |
412 | ||
34dc7c2f BB |
413 | static void |
414 | nbmand_changed_cb(void *arg, uint64_t newval) | |
415 | { | |
0037b49e BB |
416 | zfsvfs_t *zfsvfs = arg; |
417 | struct super_block *sb = zfsvfs->z_sb; | |
3558fd73 | 418 | |
2cf7f52b BB |
419 | if (sb == NULL) |
420 | return; | |
421 | ||
422 | if (newval == TRUE) | |
3558fd73 | 423 | sb->s_flags |= MS_MANDLOCK; |
2cf7f52b | 424 | else |
3558fd73 | 425 | sb->s_flags &= ~MS_MANDLOCK; |
34dc7c2f BB |
426 | } |
427 | ||
428 | static void | |
429 | snapdir_changed_cb(void *arg, uint64_t newval) | |
430 | { | |
0037b49e | 431 | ((zfsvfs_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
432 | } |
433 | ||
434 | static void | |
435 | vscan_changed_cb(void *arg, uint64_t newval) | |
436 | { | |
0037b49e | 437 | ((zfsvfs_t *)arg)->z_vscan = newval; |
34dc7c2f BB |
438 | } |
439 | ||
34dc7c2f BB |
440 | static void |
441 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
442 | { | |
0037b49e | 443 | ((zfsvfs_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
444 | } |
445 | ||
1c2555ef BB |
446 | static int |
447 | zfs_register_callbacks(vfs_t *vfsp) | |
34dc7c2f BB |
448 | { |
449 | struct dsl_dataset *ds = NULL; | |
1c2555ef BB |
450 | objset_t *os = NULL; |
451 | zfsvfs_t *zfsvfs = NULL; | |
34dc7c2f BB |
452 | int error = 0; |
453 | ||
1c2555ef BB |
454 | ASSERT(vfsp); |
455 | zfsvfs = vfsp->vfs_data; | |
0037b49e | 456 | ASSERT(zfsvfs); |
1c2555ef | 457 | os = zfsvfs->z_os; |
0282c413 BB |
458 | |
459 | /* | |
460 | * The act of registering our callbacks will destroy any mount | |
461 | * options we may have. In order to enable temporary overrides | |
462 | * of mount options, we stash away the current values and | |
463 | * restore them after we register the callbacks. | |
464 | */ | |
0037b49e | 465 | if (zfs_is_readonly(zfsvfs) || !spa_writeable(dmu_objset_spa(os))) { |
1c2555ef BB |
466 | vfsp->vfs_do_readonly = B_TRUE; |
467 | vfsp->vfs_readonly = B_TRUE; | |
0282c413 | 468 | } |
34dc7c2f BB |
469 | |
470 | /* | |
471 | * Register property callbacks. | |
472 | * | |
473 | * It would probably be fine to just check for i/o error from | |
474 | * the first prop_register(), but I guess I like to go | |
475 | * overboard... | |
476 | */ | |
477 | ds = dmu_objset_ds(os); | |
13fe0198 | 478 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
3558fd73 | 479 | error = dsl_prop_register(ds, |
0037b49e | 480 | zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zfsvfs); |
98fad862 | 481 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 482 | zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zfsvfs); |
34dc7c2f | 483 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 484 | zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zfsvfs); |
34dc7c2f | 485 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 486 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zfsvfs); |
34dc7c2f | 487 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 488 | zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zfsvfs); |
34dc7c2f | 489 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 490 | zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zfsvfs); |
34dc7c2f | 491 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 492 | zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zfsvfs); |
34dc7c2f | 493 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 494 | zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zfsvfs); |
34dc7c2f | 495 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 496 | zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zfsvfs); |
34dc7c2f | 497 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 498 | zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zfsvfs); |
023699cd | 499 | error = error ? error : dsl_prop_register(ds, |
0037b49e BB |
500 | zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, |
501 | zfsvfs); | |
34dc7c2f | 502 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 503 | zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zfsvfs); |
2cf7f52b | 504 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 505 | zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zfsvfs); |
13fe0198 | 506 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); |
34dc7c2f BB |
507 | if (error) |
508 | goto unregister; | |
509 | ||
0282c413 BB |
510 | /* |
511 | * Invoke our callbacks to restore temporary mount options. | |
512 | */ | |
1c2555ef BB |
513 | if (vfsp->vfs_do_readonly) |
514 | readonly_changed_cb(zfsvfs, vfsp->vfs_readonly); | |
515 | if (vfsp->vfs_do_setuid) | |
516 | setuid_changed_cb(zfsvfs, vfsp->vfs_setuid); | |
517 | if (vfsp->vfs_do_exec) | |
518 | exec_changed_cb(zfsvfs, vfsp->vfs_exec); | |
519 | if (vfsp->vfs_do_devices) | |
520 | devices_changed_cb(zfsvfs, vfsp->vfs_devices); | |
521 | if (vfsp->vfs_do_xattr) | |
522 | xattr_changed_cb(zfsvfs, vfsp->vfs_xattr); | |
523 | if (vfsp->vfs_do_atime) | |
524 | atime_changed_cb(zfsvfs, vfsp->vfs_atime); | |
525 | if (vfsp->vfs_do_relatime) | |
526 | relatime_changed_cb(zfsvfs, vfsp->vfs_relatime); | |
527 | if (vfsp->vfs_do_nbmand) | |
528 | nbmand_changed_cb(zfsvfs, vfsp->vfs_nbmand); | |
c9ada6d5 | 529 | |
34dc7c2f BB |
530 | return (0); |
531 | ||
532 | unregister: | |
0037b49e | 533 | dsl_prop_unregister_all(ds, zfsvfs); |
3558fd73 | 534 | return (error); |
34dc7c2f BB |
535 | } |
536 | ||
428870ff BB |
537 | static int |
538 | zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, | |
539 | uint64_t *userp, uint64_t *groupp) | |
9babb374 | 540 | { |
428870ff BB |
541 | /* |
542 | * Is it a valid type of object to track? | |
543 | */ | |
544 | if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) | |
2e528b49 | 545 | return (SET_ERROR(ENOENT)); |
9babb374 | 546 | |
428870ff BB |
547 | /* |
548 | * If we have a NULL data pointer | |
549 | * then assume the id's aren't changing and | |
550 | * return EEXIST to the dmu to let it know to | |
551 | * use the same ids | |
552 | */ | |
553 | if (data == NULL) | |
2e528b49 | 554 | return (SET_ERROR(EEXIST)); |
9babb374 | 555 | |
428870ff | 556 | if (bonustype == DMU_OT_ZNODE) { |
a94addd9 | 557 | znode_phys_t *znp = data; |
428870ff BB |
558 | *userp = znp->zp_uid; |
559 | *groupp = znp->zp_gid; | |
9babb374 | 560 | } else { |
428870ff | 561 | int hdrsize; |
a94addd9 MA |
562 | sa_hdr_phys_t *sap = data; |
563 | sa_hdr_phys_t sa = *sap; | |
564 | boolean_t swap = B_FALSE; | |
9babb374 | 565 | |
428870ff | 566 | ASSERT(bonustype == DMU_OT_SA); |
428870ff | 567 | |
a94addd9 | 568 | if (sa.sa_magic == 0) { |
428870ff BB |
569 | /* |
570 | * This should only happen for newly created | |
571 | * files that haven't had the znode data filled | |
572 | * in yet. | |
573 | */ | |
574 | *userp = 0; | |
575 | *groupp = 0; | |
a94addd9 MA |
576 | return (0); |
577 | } | |
578 | if (sa.sa_magic == BSWAP_32(SA_MAGIC)) { | |
579 | sa.sa_magic = SA_MAGIC; | |
580 | sa.sa_layout_info = BSWAP_16(sa.sa_layout_info); | |
581 | swap = B_TRUE; | |
582 | } else { | |
583 | VERIFY3U(sa.sa_magic, ==, SA_MAGIC); | |
584 | } | |
585 | ||
586 | hdrsize = sa_hdrsize(&sa); | |
587 | VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t)); | |
588 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
589 | SA_UID_OFFSET)); | |
590 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
591 | SA_GID_OFFSET)); | |
592 | if (swap) { | |
593 | *userp = BSWAP_64(*userp); | |
594 | *groupp = BSWAP_64(*groupp); | |
428870ff | 595 | } |
9babb374 | 596 | } |
13fe0198 | 597 | return (0); |
9babb374 BB |
598 | } |
599 | ||
600 | static void | |
0037b49e | 601 | fuidstr_to_sid(zfsvfs_t *zfsvfs, const char *fuidstr, |
9babb374 BB |
602 | char *domainbuf, int buflen, uid_t *ridp) |
603 | { | |
9babb374 BB |
604 | uint64_t fuid; |
605 | const char *domain; | |
606 | ||
e19572e4 | 607 | fuid = zfs_strtonum(fuidstr, NULL); |
9babb374 | 608 | |
0037b49e | 609 | domain = zfs_fuid_find_by_idx(zfsvfs, FUID_INDEX(fuid)); |
9babb374 BB |
610 | if (domain) |
611 | (void) strlcpy(domainbuf, domain, buflen); | |
612 | else | |
613 | domainbuf[0] = '\0'; | |
614 | *ridp = FUID_RID(fuid); | |
615 | } | |
616 | ||
617 | static uint64_t | |
0037b49e | 618 | zfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type) |
9babb374 BB |
619 | { |
620 | switch (type) { | |
621 | case ZFS_PROP_USERUSED: | |
1de321e6 | 622 | case ZFS_PROP_USEROBJUSED: |
9babb374 BB |
623 | return (DMU_USERUSED_OBJECT); |
624 | case ZFS_PROP_GROUPUSED: | |
1de321e6 | 625 | case ZFS_PROP_GROUPOBJUSED: |
9babb374 BB |
626 | return (DMU_GROUPUSED_OBJECT); |
627 | case ZFS_PROP_USERQUOTA: | |
0037b49e | 628 | return (zfsvfs->z_userquota_obj); |
9babb374 | 629 | case ZFS_PROP_GROUPQUOTA: |
0037b49e | 630 | return (zfsvfs->z_groupquota_obj); |
1de321e6 | 631 | case ZFS_PROP_USEROBJQUOTA: |
0037b49e | 632 | return (zfsvfs->z_userobjquota_obj); |
1de321e6 | 633 | case ZFS_PROP_GROUPOBJQUOTA: |
0037b49e | 634 | return (zfsvfs->z_groupobjquota_obj); |
149e873a | 635 | default: |
1de321e6 | 636 | return (ZFS_NO_OBJECT); |
9babb374 | 637 | } |
9babb374 BB |
638 | } |
639 | ||
640 | int | |
0037b49e | 641 | zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, |
9babb374 BB |
642 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
643 | { | |
644 | int error; | |
645 | zap_cursor_t zc; | |
646 | zap_attribute_t za; | |
647 | zfs_useracct_t *buf = vbuf; | |
648 | uint64_t obj; | |
1de321e6 | 649 | int offset = 0; |
9babb374 | 650 | |
0037b49e | 651 | if (!dmu_objset_userspace_present(zfsvfs->z_os)) |
2e528b49 | 652 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 653 | |
1de321e6 JX |
654 | if ((type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || |
655 | type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA) && | |
0037b49e | 656 | !dmu_objset_userobjspace_present(zfsvfs->z_os)) |
1de321e6 JX |
657 | return (SET_ERROR(ENOTSUP)); |
658 | ||
0037b49e | 659 | obj = zfs_userquota_prop_to_obj(zfsvfs, type); |
1de321e6 | 660 | if (obj == ZFS_NO_OBJECT) { |
9babb374 BB |
661 | *bufsizep = 0; |
662 | return (0); | |
663 | } | |
664 | ||
1de321e6 JX |
665 | if (type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED) |
666 | offset = DMU_OBJACCT_PREFIX_LEN; | |
667 | ||
0037b49e | 668 | for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep); |
9babb374 BB |
669 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
670 | zap_cursor_advance(&zc)) { | |
671 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
672 | *bufsizep) | |
673 | break; | |
674 | ||
1de321e6 JX |
675 | /* |
676 | * skip object quota (with zap name prefix DMU_OBJACCT_PREFIX) | |
677 | * when dealing with block quota and vice versa. | |
678 | */ | |
679 | if ((offset > 0) != (strncmp(za.za_name, DMU_OBJACCT_PREFIX, | |
680 | DMU_OBJACCT_PREFIX_LEN) == 0)) | |
681 | continue; | |
682 | ||
0037b49e | 683 | fuidstr_to_sid(zfsvfs, za.za_name + offset, |
9babb374 BB |
684 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
685 | ||
686 | buf->zu_space = za.za_first_integer; | |
687 | buf++; | |
688 | } | |
689 | if (error == ENOENT) | |
690 | error = 0; | |
691 | ||
692 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
693 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
694 | *cookiep = zap_cursor_serialize(&zc); | |
695 | zap_cursor_fini(&zc); | |
696 | return (error); | |
697 | } | |
698 | ||
699 | /* | |
700 | * buf must be big enough (eg, 32 bytes) | |
701 | */ | |
702 | static int | |
0037b49e | 703 | id_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid, |
9babb374 BB |
704 | char *buf, boolean_t addok) |
705 | { | |
706 | uint64_t fuid; | |
707 | int domainid = 0; | |
708 | ||
709 | if (domain && domain[0]) { | |
0037b49e | 710 | domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok); |
9babb374 | 711 | if (domainid == -1) |
2e528b49 | 712 | return (SET_ERROR(ENOENT)); |
9babb374 BB |
713 | } |
714 | fuid = FUID_ENCODE(domainid, rid); | |
715 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
716 | return (0); | |
717 | } | |
718 | ||
719 | int | |
0037b49e | 720 | zfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, |
9babb374 BB |
721 | const char *domain, uint64_t rid, uint64_t *valp) |
722 | { | |
1de321e6 JX |
723 | char buf[20 + DMU_OBJACCT_PREFIX_LEN]; |
724 | int offset = 0; | |
9babb374 BB |
725 | int err; |
726 | uint64_t obj; | |
727 | ||
728 | *valp = 0; | |
729 | ||
0037b49e | 730 | if (!dmu_objset_userspace_present(zfsvfs->z_os)) |
2e528b49 | 731 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 732 | |
1de321e6 JX |
733 | if ((type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || |
734 | type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA) && | |
0037b49e | 735 | !dmu_objset_userobjspace_present(zfsvfs->z_os)) |
1de321e6 JX |
736 | return (SET_ERROR(ENOTSUP)); |
737 | ||
0037b49e | 738 | obj = zfs_userquota_prop_to_obj(zfsvfs, type); |
1de321e6 | 739 | if (obj == ZFS_NO_OBJECT) |
9babb374 BB |
740 | return (0); |
741 | ||
1de321e6 | 742 | if (type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED) { |
1b81ab46 | 743 | strlcpy(buf, DMU_OBJACCT_PREFIX, DMU_OBJACCT_PREFIX_LEN); |
1de321e6 JX |
744 | offset = DMU_OBJACCT_PREFIX_LEN; |
745 | } | |
746 | ||
0037b49e | 747 | err = id_to_fuidstr(zfsvfs, domain, rid, buf + offset, B_FALSE); |
9babb374 BB |
748 | if (err) |
749 | return (err); | |
750 | ||
0037b49e | 751 | err = zap_lookup(zfsvfs->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
752 | if (err == ENOENT) |
753 | err = 0; | |
754 | return (err); | |
755 | } | |
756 | ||
757 | int | |
0037b49e | 758 | zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, |
9babb374 BB |
759 | const char *domain, uint64_t rid, uint64_t quota) |
760 | { | |
761 | char buf[32]; | |
762 | int err; | |
763 | dmu_tx_t *tx; | |
764 | uint64_t *objp; | |
765 | boolean_t fuid_dirtied; | |
766 | ||
0037b49e | 767 | if (zfsvfs->z_version < ZPL_VERSION_USERSPACE) |
2e528b49 | 768 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 769 | |
1de321e6 JX |
770 | switch (type) { |
771 | case ZFS_PROP_USERQUOTA: | |
0037b49e | 772 | objp = &zfsvfs->z_userquota_obj; |
1de321e6 JX |
773 | break; |
774 | case ZFS_PROP_GROUPQUOTA: | |
0037b49e | 775 | objp = &zfsvfs->z_groupquota_obj; |
1de321e6 JX |
776 | break; |
777 | case ZFS_PROP_USEROBJQUOTA: | |
0037b49e | 778 | objp = &zfsvfs->z_userobjquota_obj; |
1de321e6 JX |
779 | break; |
780 | case ZFS_PROP_GROUPOBJQUOTA: | |
0037b49e | 781 | objp = &zfsvfs->z_groupobjquota_obj; |
1de321e6 JX |
782 | break; |
783 | default: | |
784 | return (SET_ERROR(EINVAL)); | |
785 | } | |
9babb374 | 786 | |
0037b49e | 787 | err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_TRUE); |
9babb374 BB |
788 | if (err) |
789 | return (err); | |
0037b49e | 790 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 791 | |
0037b49e | 792 | tx = dmu_tx_create(zfsvfs->z_os); |
9babb374 BB |
793 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
794 | if (*objp == 0) { | |
795 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
796 | zfs_userquota_prop_prefixes[type]); | |
797 | } | |
798 | if (fuid_dirtied) | |
0037b49e | 799 | zfs_fuid_txhold(zfsvfs, tx); |
9babb374 BB |
800 | err = dmu_tx_assign(tx, TXG_WAIT); |
801 | if (err) { | |
802 | dmu_tx_abort(tx); | |
803 | return (err); | |
804 | } | |
805 | ||
0037b49e | 806 | mutex_enter(&zfsvfs->z_lock); |
9babb374 | 807 | if (*objp == 0) { |
0037b49e | 808 | *objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 809 | DMU_OT_NONE, 0, tx); |
0037b49e | 810 | VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
811 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
812 | } | |
0037b49e | 813 | mutex_exit(&zfsvfs->z_lock); |
9babb374 BB |
814 | |
815 | if (quota == 0) { | |
0037b49e | 816 | err = zap_remove(zfsvfs->z_os, *objp, buf, tx); |
9babb374 BB |
817 | if (err == ENOENT) |
818 | err = 0; | |
819 | } else { | |
0037b49e | 820 | err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
821 | } |
822 | ASSERT(err == 0); | |
823 | if (fuid_dirtied) | |
0037b49e | 824 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 BB |
825 | dmu_tx_commit(tx); |
826 | return (err); | |
827 | } | |
828 | ||
1de321e6 | 829 | boolean_t |
0037b49e | 830 | zfs_fuid_overobjquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) |
1de321e6 JX |
831 | { |
832 | char buf[20 + DMU_OBJACCT_PREFIX_LEN]; | |
833 | uint64_t used, quota, usedobj, quotaobj; | |
834 | int err; | |
835 | ||
0037b49e | 836 | if (!dmu_objset_userobjspace_present(zfsvfs->z_os)) { |
c0daec32 AB |
837 | if (dmu_objset_userobjspace_upgradable(zfsvfs->z_os)) { |
838 | dsl_pool_config_enter( | |
839 | dmu_objset_pool(zfsvfs->z_os), FTAG); | |
0037b49e | 840 | dmu_objset_userobjspace_upgrade(zfsvfs->z_os); |
c0daec32 AB |
841 | dsl_pool_config_exit( |
842 | dmu_objset_pool(zfsvfs->z_os), FTAG); | |
843 | } | |
1de321e6 JX |
844 | return (B_FALSE); |
845 | } | |
846 | ||
847 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
0037b49e BB |
848 | quotaobj = isgroup ? zfsvfs->z_groupobjquota_obj : |
849 | zfsvfs->z_userobjquota_obj; | |
850 | if (quotaobj == 0 || zfsvfs->z_replay) | |
1de321e6 JX |
851 | return (B_FALSE); |
852 | ||
853 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
0037b49e | 854 | err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); |
1de321e6 JX |
855 | if (err != 0) |
856 | return (B_FALSE); | |
857 | ||
858 | (void) sprintf(buf, DMU_OBJACCT_PREFIX "%llx", (longlong_t)fuid); | |
0037b49e | 859 | err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); |
1de321e6 JX |
860 | if (err != 0) |
861 | return (B_FALSE); | |
862 | return (used >= quota); | |
863 | } | |
864 | ||
9babb374 | 865 | boolean_t |
0037b49e | 866 | zfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) |
9babb374 | 867 | { |
1de321e6 | 868 | char buf[20]; |
9babb374 BB |
869 | uint64_t used, quota, usedobj, quotaobj; |
870 | int err; | |
871 | ||
872 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
0037b49e | 873 | quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; |
9babb374 | 874 | |
0037b49e | 875 | if (quotaobj == 0 || zfsvfs->z_replay) |
9babb374 BB |
876 | return (B_FALSE); |
877 | ||
878 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
0037b49e | 879 | err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
880 | if (err != 0) |
881 | return (B_FALSE); | |
882 | ||
0037b49e | 883 | err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
884 | if (err != 0) |
885 | return (B_FALSE); | |
886 | return (used >= quota); | |
887 | } | |
888 | ||
428870ff | 889 | boolean_t |
0037b49e | 890 | zfs_owner_overquota(zfsvfs_t *zfsvfs, znode_t *zp, boolean_t isgroup) |
428870ff BB |
891 | { |
892 | uint64_t fuid; | |
893 | uint64_t quotaobj; | |
2c6abf15 | 894 | struct inode *ip = ZTOI(zp); |
428870ff | 895 | |
0037b49e | 896 | quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; |
428870ff | 897 | |
2c6abf15 | 898 | fuid = isgroup ? KGID_TO_SGID(ip->i_gid) : KUID_TO_SUID(ip->i_uid); |
428870ff | 899 | |
0037b49e | 900 | if (quotaobj == 0 || zfsvfs->z_replay) |
428870ff BB |
901 | return (B_FALSE); |
902 | ||
0037b49e | 903 | return (zfs_fuid_overquota(zfsvfs, isgroup, fuid)); |
428870ff BB |
904 | } |
905 | ||
8614ddf9 MA |
906 | /* |
907 | * Associate this zfsvfs with the given objset, which must be owned. | |
908 | * This will cache a bunch of on-disk state from the objset in the | |
909 | * zfsvfs. | |
910 | */ | |
911 | static int | |
912 | zfsvfs_init(zfsvfs_t *zfsvfs, objset_t *os) | |
9babb374 | 913 | { |
8614ddf9 MA |
914 | int error; |
915 | uint64_t val; | |
0282c413 | 916 | |
0037b49e BB |
917 | zfsvfs->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; |
918 | zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; | |
919 | zfsvfs->z_os = os; | |
9babb374 | 920 | |
0037b49e | 921 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); |
8614ddf9 MA |
922 | if (error != 0) |
923 | return (error); | |
924 | if (zfsvfs->z_version > | |
925 | zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { | |
926 | (void) printk("Can't mount a version %lld file system " | |
927 | "on a version %lld pool\n. Pool must be upgraded to mount " | |
928 | "this file system.", (u_longlong_t)zfsvfs->z_version, | |
929 | (u_longlong_t)spa_version(dmu_objset_spa(os))); | |
930 | return (SET_ERROR(ENOTSUP)); | |
9babb374 | 931 | } |
8614ddf9 MA |
932 | error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &val); |
933 | if (error != 0) | |
934 | return (error); | |
935 | zfsvfs->z_norm = (int)val; | |
9babb374 | 936 | |
8614ddf9 MA |
937 | error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &val); |
938 | if (error != 0) | |
939 | return (error); | |
940 | zfsvfs->z_utf8 = (val != 0); | |
9babb374 | 941 | |
8614ddf9 MA |
942 | error = zfs_get_zplprop(os, ZFS_PROP_CASE, &val); |
943 | if (error != 0) | |
944 | return (error); | |
945 | zfsvfs->z_case = (uint_t)val; | |
9babb374 | 946 | |
8614ddf9 MA |
947 | if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &val)) != 0) |
948 | return (error); | |
949 | zfsvfs->z_acl_type = (uint_t)val; | |
023699cd | 950 | |
9babb374 BB |
951 | /* |
952 | * Fold case on file systems that are always or sometimes case | |
953 | * insensitive. | |
954 | */ | |
0037b49e BB |
955 | if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
956 | zfsvfs->z_case == ZFS_CASE_MIXED) | |
957 | zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 958 | |
0037b49e BB |
959 | zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); |
960 | zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); | |
428870ff | 961 | |
8614ddf9 | 962 | uint64_t sa_obj = 0; |
0037b49e | 963 | if (zfsvfs->z_use_sa) { |
428870ff BB |
964 | /* should either have both of these objects or none */ |
965 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
966 | &sa_obj); | |
8614ddf9 MA |
967 | if (error != 0) |
968 | return (error); | |
82a37189 | 969 | |
8614ddf9 MA |
970 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &val); |
971 | if ((error == 0) && (val == ZFS_XATTR_SA)) | |
0037b49e | 972 | zfsvfs->z_xattr_sa = B_TRUE; |
428870ff BB |
973 | } |
974 | ||
572e2857 | 975 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
0037b49e | 976 | &zfsvfs->z_attr_table); |
8614ddf9 MA |
977 | if (error != 0) |
978 | return (error); | |
428870ff | 979 | |
0037b49e | 980 | if (zfsvfs->z_version >= ZPL_VERSION_SA) |
428870ff | 981 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
982 | |
983 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
0037b49e | 984 | &zfsvfs->z_root); |
8614ddf9 MA |
985 | if (error != 0) |
986 | return (error); | |
0037b49e | 987 | ASSERT(zfsvfs->z_root != 0); |
9babb374 BB |
988 | |
989 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
0037b49e | 990 | &zfsvfs->z_unlinkedobj); |
8614ddf9 MA |
991 | if (error != 0) |
992 | return (error); | |
9babb374 BB |
993 | |
994 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
995 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
0037b49e | 996 | 8, 1, &zfsvfs->z_userquota_obj); |
8614ddf9 MA |
997 | if (error == ENOENT) |
998 | zfsvfs->z_userquota_obj = 0; | |
999 | else if (error != 0) | |
1000 | return (error); | |
9babb374 BB |
1001 | |
1002 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
1003 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
0037b49e | 1004 | 8, 1, &zfsvfs->z_groupquota_obj); |
8614ddf9 MA |
1005 | if (error == ENOENT) |
1006 | zfsvfs->z_groupquota_obj = 0; | |
1007 | else if (error != 0) | |
1008 | return (error); | |
9babb374 | 1009 | |
1de321e6 JX |
1010 | error = zap_lookup(os, MASTER_NODE_OBJ, |
1011 | zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA], | |
0037b49e | 1012 | 8, 1, &zfsvfs->z_userobjquota_obj); |
8614ddf9 MA |
1013 | if (error == ENOENT) |
1014 | zfsvfs->z_userobjquota_obj = 0; | |
1015 | else if (error != 0) | |
1016 | return (error); | |
1de321e6 JX |
1017 | |
1018 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
1019 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA], | |
0037b49e | 1020 | 8, 1, &zfsvfs->z_groupobjquota_obj); |
8614ddf9 MA |
1021 | if (error == ENOENT) |
1022 | zfsvfs->z_groupobjquota_obj = 0; | |
1023 | else if (error != 0) | |
1024 | return (error); | |
1de321e6 | 1025 | |
9babb374 | 1026 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, |
0037b49e | 1027 | &zfsvfs->z_fuid_obj); |
8614ddf9 MA |
1028 | if (error == ENOENT) |
1029 | zfsvfs->z_fuid_obj = 0; | |
1030 | else if (error != 0) | |
1031 | return (error); | |
9babb374 BB |
1032 | |
1033 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
0037b49e | 1034 | &zfsvfs->z_shares_dir); |
8614ddf9 MA |
1035 | if (error == ENOENT) |
1036 | zfsvfs->z_shares_dir = 0; | |
1037 | else if (error != 0) | |
1038 | return (error); | |
1039 | ||
1040 | return (0); | |
1041 | } | |
1042 | ||
1043 | int | |
1044 | zfsvfs_create(const char *osname, zfsvfs_t **zfvp) | |
1045 | { | |
1046 | objset_t *os; | |
1047 | zfsvfs_t *zfsvfs; | |
1048 | int error; | |
1049 | ||
1050 | zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); | |
1051 | ||
1052 | /* | |
1053 | * We claim to always be readonly so we can open snapshots; | |
1054 | * other ZPL code will prevent us from writing to snapshots. | |
1055 | */ | |
b5256303 TC |
1056 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, B_TRUE, |
1057 | zfsvfs, &os); | |
8614ddf9 MA |
1058 | if (error) { |
1059 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); | |
1060 | return (error); | |
1061 | } | |
1062 | ||
1063 | zfsvfs->z_vfs = NULL; | |
1064 | zfsvfs->z_sb = NULL; | |
1065 | zfsvfs->z_parent = zfsvfs; | |
9babb374 | 1066 | |
0037b49e BB |
1067 | mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
1068 | mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
1069 | list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), | |
9babb374 | 1070 | offsetof(znode_t, z_link_node)); |
0037b49e BB |
1071 | rrm_init(&zfsvfs->z_teardown_lock, B_FALSE); |
1072 | rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); | |
1073 | rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
d07b7c7f | 1074 | |
8614ddf9 MA |
1075 | int size = MIN(1 << (highbit64(zfs_object_mutex_size) - 1), |
1076 | ZFS_OBJ_MTX_MAX); | |
0037b49e BB |
1077 | zfsvfs->z_hold_size = size; |
1078 | zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, | |
1079 | KM_SLEEP); | |
1080 | zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
8614ddf9 | 1081 | for (int i = 0; i != size; i++) { |
0037b49e | 1082 | avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare, |
c96c36fa | 1083 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); |
0037b49e | 1084 | mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); |
c96c36fa | 1085 | } |
9babb374 | 1086 | |
8614ddf9 MA |
1087 | error = zfsvfs_init(zfsvfs, os); |
1088 | if (error != 0) { | |
b5256303 | 1089 | dmu_objset_disown(os, B_TRUE, zfsvfs); |
8614ddf9 MA |
1090 | *zfvp = NULL; |
1091 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); | |
1092 | return (error); | |
1093 | } | |
1094 | ||
0037b49e | 1095 | *zfvp = zfsvfs; |
9babb374 | 1096 | return (0); |
9babb374 BB |
1097 | } |
1098 | ||
8614ddf9 | 1099 | static int |
f298b24d | 1100 | zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) |
34dc7c2f | 1101 | { |
34dc7c2f BB |
1102 | int error; |
1103 | ||
1c2555ef | 1104 | error = zfs_register_callbacks(zfsvfs->z_vfs); |
34dc7c2f BB |
1105 | if (error) |
1106 | return (error); | |
1107 | ||
0037b49e | 1108 | zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); |
9babb374 | 1109 | |
34dc7c2f BB |
1110 | /* |
1111 | * If we are not mounting (ie: online recv), then we don't | |
1112 | * have to worry about replaying the log as we blocked all | |
1113 | * operations out since we closed the ZIL. | |
1114 | */ | |
1115 | if (mounting) { | |
b128c09f BB |
1116 | boolean_t readonly; |
1117 | ||
34dc7c2f BB |
1118 | /* |
1119 | * During replay we remove the read only flag to | |
1120 | * allow replays to succeed. | |
1121 | */ | |
0037b49e | 1122 | readonly = zfs_is_readonly(zfsvfs); |
fb5f0bc8 | 1123 | if (readonly != 0) |
0037b49e | 1124 | readonly_changed_cb(zfsvfs, B_FALSE); |
fb5f0bc8 | 1125 | else |
0037b49e | 1126 | zfs_unlinked_drain(zfsvfs); |
34dc7c2f | 1127 | |
428870ff BB |
1128 | /* |
1129 | * Parse and replay the intent log. | |
1130 | * | |
1131 | * Because of ziltest, this must be done after | |
1132 | * zfs_unlinked_drain(). (Further note: ziltest | |
1133 | * doesn't use readonly mounts, where | |
1134 | * zfs_unlinked_drain() isn't called.) This is because | |
1135 | * ziltest causes spa_sync() to think it's committed, | |
1136 | * but actually it is not, so the intent log contains | |
1137 | * many txg's worth of changes. | |
1138 | * | |
1139 | * In particular, if object N is in the unlinked set in | |
1140 | * the last txg to actually sync, then it could be | |
1141 | * actually freed in a later txg and then reallocated | |
1142 | * in a yet later txg. This would write a "create | |
1143 | * object N" record to the intent log. Normally, this | |
1144 | * would be fine because the spa_sync() would have | |
1145 | * written out the fact that object N is free, before | |
1146 | * we could write the "create object N" intent log | |
1147 | * record. | |
1148 | * | |
1149 | * But when we are in ziltest mode, we advance the "open | |
1150 | * txg" without actually spa_sync()-ing the changes to | |
1151 | * disk. So we would see that object N is still | |
1152 | * allocated and in the unlinked set, and there is an | |
1153 | * intent log record saying to allocate it. | |
1154 | */ | |
0037b49e | 1155 | if (spa_writeable(dmu_objset_spa(zfsvfs->z_os))) { |
572e2857 | 1156 | if (zil_replay_disable) { |
0037b49e | 1157 | zil_destroy(zfsvfs->z_log, B_FALSE); |
572e2857 | 1158 | } else { |
0037b49e BB |
1159 | zfsvfs->z_replay = B_TRUE; |
1160 | zil_replay(zfsvfs->z_os, zfsvfs, | |
572e2857 | 1161 | zfs_replay_vector); |
0037b49e | 1162 | zfsvfs->z_replay = B_FALSE; |
572e2857 | 1163 | } |
fb5f0bc8 | 1164 | } |
2cf7f52b BB |
1165 | |
1166 | /* restore readonly bit */ | |
1167 | if (readonly != 0) | |
0037b49e | 1168 | readonly_changed_cb(zfsvfs, B_TRUE); |
34dc7c2f BB |
1169 | } |
1170 | ||
9775e988 | 1171 | /* |
0037b49e | 1172 | * Set the objset user_ptr to track its zfsvfs. |
9775e988 | 1173 | */ |
0037b49e BB |
1174 | mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); |
1175 | dmu_objset_set_user(zfsvfs->z_os, zfsvfs); | |
1176 | mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); | |
9775e988 | 1177 | |
34dc7c2f BB |
1178 | return (0); |
1179 | } | |
1180 | ||
9babb374 | 1181 | void |
f298b24d | 1182 | zfsvfs_free(zfsvfs_t *zfsvfs) |
34dc7c2f | 1183 | { |
0037b49e | 1184 | int i, size = zfsvfs->z_hold_size; |
9babb374 | 1185 | |
0037b49e | 1186 | zfs_fuid_destroy(zfsvfs); |
9babb374 | 1187 | |
0037b49e BB |
1188 | mutex_destroy(&zfsvfs->z_znodes_lock); |
1189 | mutex_destroy(&zfsvfs->z_lock); | |
1190 | list_destroy(&zfsvfs->z_all_znodes); | |
1191 | rrm_destroy(&zfsvfs->z_teardown_lock); | |
1192 | rw_destroy(&zfsvfs->z_teardown_inactive_lock); | |
1193 | rw_destroy(&zfsvfs->z_fuid_lock); | |
c96c36fa | 1194 | for (i = 0; i != size; i++) { |
0037b49e BB |
1195 | avl_destroy(&zfsvfs->z_hold_trees[i]); |
1196 | mutex_destroy(&zfsvfs->z_hold_locks[i]); | |
c96c36fa | 1197 | } |
0037b49e BB |
1198 | vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size); |
1199 | vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size); | |
1c2555ef | 1200 | zfsvfs_vfs_free(zfsvfs->z_vfs); |
0037b49e | 1201 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); |
34dc7c2f BB |
1202 | } |
1203 | ||
9babb374 | 1204 | static void |
0037b49e | 1205 | zfs_set_fuid_feature(zfsvfs_t *zfsvfs) |
9babb374 | 1206 | { |
0037b49e BB |
1207 | zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); |
1208 | zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); | |
9babb374 | 1209 | } |
34dc7c2f BB |
1210 | |
1211 | void | |
0037b49e | 1212 | zfs_unregister_callbacks(zfsvfs_t *zfsvfs) |
34dc7c2f | 1213 | { |
0037b49e | 1214 | objset_t *os = zfsvfs->z_os; |
34dc7c2f | 1215 | |
0eb21616 | 1216 | if (!dmu_objset_is_snapshot(os)) |
0037b49e | 1217 | dsl_prop_unregister_all(dmu_objset_ds(os), zfsvfs); |
34dc7c2f BB |
1218 | } |
1219 | ||
bc3e15e3 | 1220 | #ifdef HAVE_MLSLABEL |
428870ff | 1221 | /* |
d3cc8b15 WA |
1222 | * Check that the hex label string is appropriate for the dataset being |
1223 | * mounted into the global_zone proper. | |
428870ff | 1224 | * |
d3cc8b15 WA |
1225 | * Return an error if the hex label string is not default or |
1226 | * admin_low/admin_high. For admin_low labels, the corresponding | |
1227 | * dataset must be readonly. | |
428870ff BB |
1228 | */ |
1229 | int | |
1230 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
1231 | { | |
1232 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
1233 | return (0); | |
1234 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
1235 | return (0); | |
1236 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
1237 | /* must be readonly */ | |
1238 | uint64_t rdonly; | |
1239 | ||
1240 | if (dsl_prop_get_integer(dsname, | |
1241 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
2e528b49 | 1242 | return (SET_ERROR(EACCES)); |
428870ff BB |
1243 | return (rdonly ? 0 : EACCES); |
1244 | } | |
2e528b49 | 1245 | return (SET_ERROR(EACCES)); |
428870ff | 1246 | } |
bc3e15e3 | 1247 | #endif /* HAVE_MLSLABEL */ |
428870ff | 1248 | |
e5c39b95 | 1249 | int |
3558fd73 | 1250 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 1251 | { |
0037b49e | 1252 | zfsvfs_t *zfsvfs = dentry->d_sb->s_fs_info; |
34dc7c2f | 1253 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
04f9432d | 1254 | uint64_t fsid; |
3558fd73 | 1255 | uint32_t bshift; |
34dc7c2f | 1256 | |
0037b49e | 1257 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1258 | |
0037b49e | 1259 | dmu_objset_space(zfsvfs->z_os, |
34dc7c2f BB |
1260 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
1261 | ||
0037b49e | 1262 | fsid = dmu_objset_fsid_guid(zfsvfs->z_os); |
34dc7c2f | 1263 | /* |
05ff35c6 BB |
1264 | * The underlying storage pool actually uses multiple block |
1265 | * size. Under Solaris frsize (fragment size) is reported as | |
1266 | * the smallest block size we support, and bsize (block size) | |
1267 | * as the filesystem's maximum block size. Unfortunately, | |
1268 | * under Linux the fragment size and block size are often used | |
1269 | * interchangeably. Thus we are forced to report both of them | |
1270 | * as the filesystem's maximum block size. | |
34dc7c2f | 1271 | */ |
0037b49e BB |
1272 | statp->f_frsize = zfsvfs->z_max_blksz; |
1273 | statp->f_bsize = zfsvfs->z_max_blksz; | |
3558fd73 | 1274 | bshift = fls(statp->f_bsize) - 1; |
34dc7c2f BB |
1275 | |
1276 | /* | |
3558fd73 BB |
1277 | * The following report "total" blocks of various kinds in |
1278 | * the file system, but reported in terms of f_bsize - the | |
1279 | * "preferred" size. | |
34dc7c2f BB |
1280 | */ |
1281 | ||
3558fd73 BB |
1282 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
1283 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
1284 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
1285 | ||
1286 | /* | |
1287 | * statvfs() should really be called statufs(), because it assumes | |
1288 | * static metadata. ZFS doesn't preallocate files, so the best | |
1289 | * we can do is report the max that could possibly fit in f_files, | |
1290 | * and that minus the number actually used in f_ffree. | |
1291 | * For f_ffree, report the smaller of the number of object available | |
1292 | * and the number of blocks (each object will take at least a block). | |
1293 | */ | |
baab0630 | 1294 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 1295 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
1296 | statp->f_fsid.val[0] = (uint32_t)fsid; |
1297 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 | 1298 | statp->f_type = ZFS_SUPER_MAGIC; |
eca7b760 | 1299 | statp->f_namelen = MAXNAMELEN - 1; |
34dc7c2f BB |
1300 | |
1301 | /* | |
3558fd73 | 1302 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
1303 | * Is there anything useful we could/should provide? |
1304 | */ | |
3558fd73 | 1305 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 1306 | |
0037b49e | 1307 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1308 | return (0); |
1309 | } | |
1310 | ||
e5c39b95 | 1311 | int |
0037b49e | 1312 | zfs_root(zfsvfs_t *zfsvfs, struct inode **ipp) |
34dc7c2f | 1313 | { |
34dc7c2f BB |
1314 | znode_t *rootzp; |
1315 | int error; | |
1316 | ||
0037b49e | 1317 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1318 | |
0037b49e | 1319 | error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp); |
34dc7c2f | 1320 | if (error == 0) |
3558fd73 | 1321 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1322 | |
0037b49e | 1323 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1324 | return (error); |
1325 | } | |
1326 | ||
09fb30e5 | 1327 | #ifdef HAVE_D_PRUNE_ALIASES |
218b4e0a BB |
1328 | /* |
1329 | * Linux kernels older than 3.1 do not support a per-filesystem shrinker. | |
1330 | * To accommodate this we must improvise and manually walk the list of znodes | |
1331 | * attempting to prune dentries in order to be able to drop the inodes. | |
1332 | * | |
1333 | * To avoid scanning the same znodes multiple times they are always rotated | |
1334 | * to the end of the z_all_znodes list. New znodes are inserted at the | |
1335 | * end of the list so we're always scanning the oldest znodes first. | |
1336 | */ | |
1337 | static int | |
f298b24d | 1338 | zfs_prune_aliases(zfsvfs_t *zfsvfs, unsigned long nr_to_scan) |
218b4e0a BB |
1339 | { |
1340 | znode_t **zp_array, *zp; | |
1341 | int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *)); | |
1342 | int objects = 0; | |
1343 | int i = 0, j = 0; | |
1344 | ||
1345 | zp_array = kmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP); | |
1346 | ||
0037b49e BB |
1347 | mutex_enter(&zfsvfs->z_znodes_lock); |
1348 | while ((zp = list_head(&zfsvfs->z_all_znodes)) != NULL) { | |
218b4e0a BB |
1349 | |
1350 | if ((i++ > nr_to_scan) || (j >= max_array)) | |
1351 | break; | |
1352 | ||
1353 | ASSERT(list_link_active(&zp->z_link_node)); | |
0037b49e BB |
1354 | list_remove(&zfsvfs->z_all_znodes, zp); |
1355 | list_insert_tail(&zfsvfs->z_all_znodes, zp); | |
218b4e0a BB |
1356 | |
1357 | /* Skip active znodes and .zfs entries */ | |
1358 | if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir) | |
1359 | continue; | |
1360 | ||
1361 | if (igrab(ZTOI(zp)) == NULL) | |
1362 | continue; | |
1363 | ||
1364 | zp_array[j] = zp; | |
1365 | j++; | |
1366 | } | |
0037b49e | 1367 | mutex_exit(&zfsvfs->z_znodes_lock); |
218b4e0a BB |
1368 | |
1369 | for (i = 0; i < j; i++) { | |
1370 | zp = zp_array[i]; | |
1371 | ||
1372 | ASSERT3P(zp, !=, NULL); | |
1373 | d_prune_aliases(ZTOI(zp)); | |
1374 | ||
1375 | if (atomic_read(&ZTOI(zp)->i_count) == 1) | |
1376 | objects++; | |
1377 | ||
1378 | iput(ZTOI(zp)); | |
1379 | } | |
1380 | ||
1381 | kmem_free(zp_array, max_array * sizeof (znode_t *)); | |
1382 | ||
1383 | return (objects); | |
1384 | } | |
1385 | #endif /* HAVE_D_PRUNE_ALIASES */ | |
1386 | ||
2cbb06b5 BB |
1387 | /* |
1388 | * The ARC has requested that the filesystem drop entries from the dentry | |
1389 | * and inode caches. This can occur when the ARC needs to free meta data | |
1390 | * blocks but can't because they are all pinned by entries in these caches. | |
1391 | */ | |
ab26409d | 1392 | int |
f298b24d | 1393 | zfs_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) |
ab26409d | 1394 | { |
0037b49e | 1395 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
2cbb06b5 BB |
1396 | int error = 0; |
1397 | #if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK) | |
ab26409d BB |
1398 | struct shrinker *shrinker = &sb->s_shrink; |
1399 | struct shrink_control sc = { | |
1400 | .nr_to_scan = nr_to_scan, | |
1401 | .gfp_mask = GFP_KERNEL, | |
1402 | }; | |
2cbb06b5 | 1403 | #endif |
ab26409d | 1404 | |
0037b49e | 1405 | ZFS_ENTER(zfsvfs); |
2cbb06b5 | 1406 | |
90947b23 TC |
1407 | #if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \ |
1408 | defined(SHRINK_CONTROL_HAS_NID) && \ | |
1409 | defined(SHRINKER_NUMA_AWARE) | |
1410 | if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) { | |
1411 | *objects = 0; | |
02730c33 | 1412 | for_each_online_node(sc.nid) { |
90947b23 | 1413 | *objects += (*shrinker->scan_objects)(shrinker, &sc); |
02730c33 | 1414 | } |
90947b23 TC |
1415 | } else { |
1416 | *objects = (*shrinker->scan_objects)(shrinker, &sc); | |
1417 | } | |
09fb30e5 | 1418 | |
90947b23 | 1419 | #elif defined(HAVE_SPLIT_SHRINKER_CALLBACK) |
3c832b8c | 1420 | *objects = (*shrinker->scan_objects)(shrinker, &sc); |
2cbb06b5 | 1421 | #elif defined(HAVE_SHRINK) |
ab26409d | 1422 | *objects = (*shrinker->shrink)(shrinker, &sc); |
218b4e0a | 1423 | #elif defined(HAVE_D_PRUNE_ALIASES) |
09fb30e5 | 1424 | #define D_PRUNE_ALIASES_IS_DEFAULT |
f298b24d | 1425 | *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); |
2cbb06b5 | 1426 | #else |
218b4e0a | 1427 | #error "No available dentry and inode cache pruning mechanism." |
3c832b8c | 1428 | #endif |
09fb30e5 TC |
1429 | |
1430 | #if defined(HAVE_D_PRUNE_ALIASES) && !defined(D_PRUNE_ALIASES_IS_DEFAULT) | |
1431 | #undef D_PRUNE_ALIASES_IS_DEFAULT | |
1432 | /* | |
f298b24d | 1433 | * Fall back to zfs_prune_aliases if the kernel's per-superblock |
09fb30e5 TC |
1434 | * shrinker couldn't free anything, possibly due to the inodes being |
1435 | * allocated in a different memcg. | |
1436 | */ | |
1437 | if (*objects == 0) | |
f298b24d | 1438 | *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); |
09fb30e5 TC |
1439 | #endif |
1440 | ||
0037b49e | 1441 | ZFS_EXIT(zfsvfs); |
ab26409d | 1442 | |
0037b49e | 1443 | dprintf_ds(zfsvfs->z_os->os_dsl_dataset, |
2cbb06b5 BB |
1444 | "pruning, nr_to_scan=%lu objects=%d error=%d\n", |
1445 | nr_to_scan, *objects, error); | |
1446 | ||
1447 | return (error); | |
ab26409d | 1448 | } |
ab26409d | 1449 | |
34dc7c2f | 1450 | /* |
0037b49e | 1451 | * Teardown the zfsvfs_t. |
34dc7c2f | 1452 | * |
a08abc1b | 1453 | * Note, if 'unmounting' is FALSE, we return with the 'z_teardown_lock' |
34dc7c2f BB |
1454 | * and 'z_teardown_inactive_lock' held. |
1455 | */ | |
f298b24d | 1456 | static int |
0037b49e | 1457 | zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) |
34dc7c2f BB |
1458 | { |
1459 | znode_t *zp; | |
1460 | ||
fd23720a BB |
1461 | /* |
1462 | * If someone has not already unmounted this file system, | |
1463 | * drain the iput_taskq to ensure all active references to the | |
0037b49e | 1464 | * zfsvfs_t have been handled only then can it be safely destroyed. |
fd23720a | 1465 | */ |
0037b49e | 1466 | if (zfsvfs->z_os) { |
f0da4d15 CD |
1467 | /* |
1468 | * If we're unmounting we have to wait for the list to | |
1469 | * drain completely. | |
1470 | * | |
1471 | * If we're not unmounting there's no guarantee the list | |
1472 | * will drain completely, but iputs run from the taskq | |
1473 | * may add the parents of dir-based xattrs to the taskq | |
1474 | * so we want to wait for these. | |
1475 | * | |
1476 | * We can safely read z_nr_znodes without locking because the | |
1477 | * VFS has already blocked operations which add to the | |
1478 | * z_all_znodes list and thus increment z_nr_znodes. | |
1479 | */ | |
1480 | int round = 0; | |
0037b49e | 1481 | while (zfsvfs->z_nr_znodes > 0) { |
c5528b9b | 1482 | taskq_wait_outstanding(dsl_pool_iput_taskq( |
0037b49e | 1483 | dmu_objset_pool(zfsvfs->z_os)), 0); |
f0da4d15 CD |
1484 | if (++round > 1 && !unmounting) |
1485 | break; | |
1486 | } | |
1487 | } | |
fd23720a | 1488 | |
0037b49e | 1489 | rrm_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1490 | |
1491 | if (!unmounting) { | |
1492 | /* | |
ceb43b93 BB |
1493 | * We purge the parent filesystem's super block as the |
1494 | * parent filesystem and all of its snapshots have their | |
1495 | * inode's super block set to the parent's filesystem's | |
1496 | * super block. Note, 'z_parent' is self referential | |
1497 | * for non-snapshots. | |
34dc7c2f | 1498 | */ |
0037b49e | 1499 | shrink_dcache_sb(zfsvfs->z_parent->z_sb); |
34dc7c2f BB |
1500 | } |
1501 | ||
1502 | /* | |
1503 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1504 | * threads are blocked as zil_close can call zfs_inactive. | |
1505 | */ | |
0037b49e BB |
1506 | if (zfsvfs->z_log) { |
1507 | zil_close(zfsvfs->z_log); | |
1508 | zfsvfs->z_log = NULL; | |
34dc7c2f BB |
1509 | } |
1510 | ||
0037b49e | 1511 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1512 | |
1513 | /* | |
1514 | * If we are not unmounting (ie: online recv) and someone already | |
1515 | * unmounted this file system while we were doing the switcheroo, | |
1516 | * or a reopen of z_os failed then just bail out now. | |
1517 | */ | |
0037b49e BB |
1518 | if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) { |
1519 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
1520 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
2e528b49 | 1521 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1522 | } |
1523 | ||
1524 | /* | |
7b3e34ba BB |
1525 | * At this point there are no VFS ops active, and any new VFS ops |
1526 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1527 | * relevant for forced unmount). | |
34dc7c2f BB |
1528 | * |
1529 | * Release all holds on dbufs. | |
1530 | */ | |
f0da4d15 | 1531 | if (!unmounting) { |
0037b49e BB |
1532 | mutex_enter(&zfsvfs->z_znodes_lock); |
1533 | for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL; | |
1534 | zp = list_next(&zfsvfs->z_all_znodes, zp)) { | |
f0da4d15 CD |
1535 | if (zp->z_sa_hdl) |
1536 | zfs_znode_dmu_fini(zp); | |
1537 | } | |
0037b49e | 1538 | mutex_exit(&zfsvfs->z_znodes_lock); |
7b3e34ba | 1539 | } |
34dc7c2f BB |
1540 | |
1541 | /* | |
7b3e34ba | 1542 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1543 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1544 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1545 | */ |
1546 | if (unmounting) { | |
0037b49e BB |
1547 | zfsvfs->z_unmounted = B_TRUE; |
1548 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
1549 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1550 | } |
1551 | ||
1552 | /* | |
1553 | * z_os will be NULL if there was an error in attempting to reopen | |
0037b49e | 1554 | * zfsvfs, so just return as the properties had already been |
3558fd73 | 1555 | * |
34dc7c2f BB |
1556 | * unregistered and cached data had been evicted before. |
1557 | */ | |
0037b49e | 1558 | if (zfsvfs->z_os == NULL) |
34dc7c2f BB |
1559 | return (0); |
1560 | ||
1561 | /* | |
1562 | * Unregister properties. | |
1563 | */ | |
0037b49e | 1564 | zfs_unregister_callbacks(zfsvfs); |
34dc7c2f BB |
1565 | |
1566 | /* | |
1567 | * Evict cached data | |
1568 | */ | |
0037b49e BB |
1569 | if (dsl_dataset_is_dirty(dmu_objset_ds(zfsvfs->z_os)) && |
1570 | !zfs_is_readonly(zfsvfs)) | |
1571 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); | |
1572 | dmu_objset_evict_dbufs(zfsvfs->z_os); | |
34dc7c2f BB |
1573 | |
1574 | return (0); | |
1575 | } | |
1576 | ||
3e6c9433 | 1577 | #if !defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER) && \ |
8c45def2 | 1578 | !defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER) |
5547c2f1 | 1579 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); |
8c45def2 | 1580 | #endif |
76659dc1 | 1581 | |
e5c39b95 | 1582 | int |
1c2555ef | 1583 | zfs_domount(struct super_block *sb, zfs_mnt_t *zm, int silent) |
34dc7c2f | 1584 | { |
1c2555ef | 1585 | const char *osname = zm->mnt_osname; |
3558fd73 BB |
1586 | struct inode *root_inode; |
1587 | uint64_t recordsize; | |
1c2555ef BB |
1588 | int error = 0; |
1589 | zfsvfs_t *zfsvfs; | |
1590 | ||
1591 | ASSERT(zm); | |
1592 | ASSERT(osname); | |
34dc7c2f | 1593 | |
1c2555ef | 1594 | error = zfsvfs_create(osname, &zfsvfs); |
3558fd73 BB |
1595 | if (error) |
1596 | return (error); | |
1597 | ||
1c2555ef BB |
1598 | error = zfsvfs_parse_options(zm->mnt_data, &zfsvfs->z_vfs); |
1599 | if (error) | |
1600 | goto out; | |
1601 | ||
3558fd73 BB |
1602 | if ((error = dsl_prop_get_integer(osname, "recordsize", |
1603 | &recordsize, NULL))) | |
1604 | goto out; | |
1605 | ||
1c2555ef | 1606 | zfsvfs->z_vfs->vfs_data = zfsvfs; |
0037b49e BB |
1607 | zfsvfs->z_sb = sb; |
1608 | sb->s_fs_info = zfsvfs; | |
3558fd73 BB |
1609 | sb->s_magic = ZFS_SUPER_MAGIC; |
1610 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1611 | sb->s_time_gran = 1; | |
1612 | sb->s_blocksize = recordsize; | |
1613 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 | 1614 | |
7dae2c81 | 1615 | error = -zpl_bdi_setup(sb, "zfs"); |
5547c2f1 BB |
1616 | if (error) |
1617 | goto out; | |
3558fd73 | 1618 | |
7dae2c81 BB |
1619 | sb->s_bdi->ra_pages = 0; |
1620 | ||
3558fd73 BB |
1621 | /* Set callback operations for the file system. */ |
1622 | sb->s_op = &zpl_super_operations; | |
1623 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1624 | sb->s_export_op = &zpl_export_operations; |
ee930353 BB |
1625 | #ifdef HAVE_S_D_OP |
1626 | sb->s_d_op = &zpl_dentry_operations; | |
1627 | #endif /* HAVE_S_D_OP */ | |
3558fd73 BB |
1628 | |
1629 | /* Set features for file system. */ | |
0037b49e | 1630 | zfs_set_fuid_feature(zfsvfs); |
3558fd73 | 1631 | |
0037b49e | 1632 | if (dmu_objset_is_snapshot(zfsvfs->z_os)) { |
3558fd73 BB |
1633 | uint64_t pval; |
1634 | ||
0037b49e BB |
1635 | atime_changed_cb(zfsvfs, B_FALSE); |
1636 | readonly_changed_cb(zfsvfs, B_TRUE); | |
d1d7e268 MK |
1637 | if ((error = dsl_prop_get_integer(osname, |
1638 | "xattr", &pval, NULL))) | |
3558fd73 | 1639 | goto out; |
0037b49e | 1640 | xattr_changed_cb(zfsvfs, pval); |
d1d7e268 MK |
1641 | if ((error = dsl_prop_get_integer(osname, |
1642 | "acltype", &pval, NULL))) | |
023699cd | 1643 | goto out; |
0037b49e BB |
1644 | acltype_changed_cb(zfsvfs, pval); |
1645 | zfsvfs->z_issnap = B_TRUE; | |
1646 | zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1647 | zfsvfs->z_snap_defer_time = jiffies; | |
1648 | ||
1649 | mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); | |
1650 | dmu_objset_set_user(zfsvfs->z_os, zfsvfs); | |
1651 | mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); | |
3558fd73 | 1652 | } else { |
f298b24d | 1653 | if ((error = zfsvfs_setup(zfsvfs, B_TRUE))) |
be88e733 | 1654 | goto out; |
34dc7c2f BB |
1655 | } |
1656 | ||
3558fd73 | 1657 | /* Allocate a root inode for the filesystem. */ |
0037b49e | 1658 | error = zfs_root(zfsvfs, &root_inode); |
3558fd73 BB |
1659 | if (error) { |
1660 | (void) zfs_umount(sb); | |
1661 | goto out; | |
34dc7c2f BB |
1662 | } |
1663 | ||
3558fd73 | 1664 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1665 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1666 | if (sb->s_root == NULL) { |
1667 | (void) zfs_umount(sb); | |
2e528b49 | 1668 | error = SET_ERROR(ENOMEM); |
3558fd73 BB |
1669 | goto out; |
1670 | } | |
ebe7e575 | 1671 | |
0037b49e BB |
1672 | if (!zfsvfs->z_issnap) |
1673 | zfsctl_create(zfsvfs); | |
2cbb06b5 | 1674 | |
0037b49e | 1675 | zfsvfs->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); |
3558fd73 BB |
1676 | out: |
1677 | if (error) { | |
b5256303 | 1678 | dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs); |
f298b24d | 1679 | zfsvfs_free(zfsvfs); |
be88e733 CC |
1680 | /* |
1681 | * make sure we don't have dangling sb->s_fs_info which | |
1682 | * zfs_preumount will use. | |
1683 | */ | |
1684 | sb->s_fs_info = NULL; | |
3558fd73 | 1685 | } |
34dc7c2f | 1686 | |
3558fd73 BB |
1687 | return (error); |
1688 | } | |
3558fd73 | 1689 | |
ebe7e575 BB |
1690 | /* |
1691 | * Called when an unmount is requested and certain sanity checks have | |
1692 | * already passed. At this point no dentries or inodes have been reclaimed | |
1693 | * from their respective caches. We drop the extra reference on the .zfs | |
1694 | * control directory to allow everything to be reclaimed. All snapshots | |
1695 | * must already have been unmounted to reach this point. | |
1696 | */ | |
1697 | void | |
1698 | zfs_preumount(struct super_block *sb) | |
1699 | { | |
0037b49e | 1700 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
ebe7e575 | 1701 | |
0037b49e BB |
1702 | /* zfsvfs is NULL when zfs_domount fails during mount */ |
1703 | if (zfsvfs) { | |
278bee93 | 1704 | zfsctl_destroy(sb->s_fs_info); |
be88e733 CC |
1705 | /* |
1706 | * Wait for iput_async before entering evict_inodes in | |
1707 | * generic_shutdown_super. The reason we must finish before | |
1708 | * evict_inodes is when lazytime is on, or when zfs_purgedir | |
1709 | * calls zfs_zget, iput would bump i_count from 0 to 1. This | |
1710 | * would race with the i_count check in evict_inodes. This means | |
1711 | * it could destroy the inode while we are still using it. | |
1712 | * | |
1713 | * We wait for two passes. xattr directories in the first pass | |
1714 | * may add xattr entries in zfs_purgedir, so in the second pass | |
1715 | * we wait for them. We don't use taskq_wait here because it is | |
1716 | * a pool wide taskq. Other mounted filesystems can constantly | |
1717 | * do iput_async and there's no guarantee when taskq will be | |
1718 | * empty. | |
1719 | */ | |
1720 | taskq_wait_outstanding(dsl_pool_iput_taskq( | |
0037b49e | 1721 | dmu_objset_pool(zfsvfs->z_os)), 0); |
be88e733 | 1722 | taskq_wait_outstanding(dsl_pool_iput_taskq( |
0037b49e | 1723 | dmu_objset_pool(zfsvfs->z_os)), 0); |
be88e733 | 1724 | } |
ebe7e575 | 1725 | } |
ebe7e575 BB |
1726 | |
1727 | /* | |
1728 | * Called once all other unmount released tear down has occurred. | |
1729 | * It is our responsibility to release any remaining infrastructure. | |
1730 | */ | |
3558fd73 BB |
1731 | /*ARGSUSED*/ |
1732 | int | |
1733 | zfs_umount(struct super_block *sb) | |
1734 | { | |
0037b49e | 1735 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
3558fd73 BB |
1736 | objset_t *os; |
1737 | ||
b5256303 TC |
1738 | if (zfsvfs->z_arc_prune != NULL) |
1739 | arc_remove_prune_callback(zfsvfs->z_arc_prune); | |
0037b49e BB |
1740 | VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0); |
1741 | os = zfsvfs->z_os; | |
7dae2c81 | 1742 | zpl_bdi_destroy(sb); |
76659dc1 | 1743 | |
34dc7c2f BB |
1744 | /* |
1745 | * z_os will be NULL if there was an error in | |
0037b49e | 1746 | * attempting to reopen zfsvfs. |
34dc7c2f BB |
1747 | */ |
1748 | if (os != NULL) { | |
1749 | /* | |
1750 | * Unset the objset user_ptr. | |
1751 | */ | |
428870ff | 1752 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1753 | dmu_objset_set_user(os, NULL); |
428870ff | 1754 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1755 | |
1756 | /* | |
b128c09f | 1757 | * Finally release the objset |
34dc7c2f | 1758 | */ |
b5256303 | 1759 | dmu_objset_disown(os, B_TRUE, zfsvfs); |
34dc7c2f BB |
1760 | } |
1761 | ||
f298b24d | 1762 | zfsvfs_free(zfsvfs); |
34dc7c2f BB |
1763 | return (0); |
1764 | } | |
1765 | ||
0de19dad | 1766 | int |
1c2555ef | 1767 | zfs_remount(struct super_block *sb, int *flags, zfs_mnt_t *zm) |
0de19dad | 1768 | { |
0037b49e | 1769 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
1c2555ef | 1770 | vfs_t *vfsp; |
08de8c16 | 1771 | boolean_t issnap = dmu_objset_is_snapshot(zfsvfs->z_os); |
0282c413 BB |
1772 | int error; |
1773 | ||
08de8c16 | 1774 | if ((issnap || !spa_writeable(dmu_objset_spa(zfsvfs->z_os))) && |
1775 | !(*flags & MS_RDONLY)) { | |
1776 | *flags |= MS_RDONLY; | |
1777 | return (EROFS); | |
1778 | } | |
1779 | ||
1c2555ef BB |
1780 | error = zfsvfs_parse_options(zm->mnt_data, &vfsp); |
1781 | if (error) | |
1782 | return (error); | |
1783 | ||
0037b49e | 1784 | zfs_unregister_callbacks(zfsvfs); |
1c2555ef BB |
1785 | zfsvfs_vfs_free(zfsvfs->z_vfs); |
1786 | ||
1787 | vfsp->vfs_data = zfsvfs; | |
1788 | zfsvfs->z_vfs = vfsp; | |
08de8c16 | 1789 | if (!issnap) |
1790 | (void) zfs_register_callbacks(vfsp); | |
0282c413 BB |
1791 | |
1792 | return (error); | |
0de19dad | 1793 | } |
0de19dad | 1794 | |
e5c39b95 | 1795 | int |
2cf7f52b | 1796 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1797 | { |
0037b49e | 1798 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
34dc7c2f BB |
1799 | znode_t *zp; |
1800 | uint64_t object = 0; | |
1801 | uint64_t fid_gen = 0; | |
1802 | uint64_t gen_mask; | |
1803 | uint64_t zp_gen; | |
3558fd73 | 1804 | int i, err; |
34dc7c2f | 1805 | |
3558fd73 | 1806 | *ipp = NULL; |
34dc7c2f | 1807 | |
9b77d1c9 CC |
1808 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { |
1809 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1810 | ||
1811 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1812 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
34dc7c2f | 1813 | |
9b77d1c9 CC |
1814 | for (i = 0; i < sizeof (zfid->zf_gen); i++) |
1815 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1816 | } else { | |
1817 | return (SET_ERROR(EINVAL)); | |
1818 | } | |
1819 | ||
1820 | /* LONG_FID_LEN means snapdirs */ | |
34dc7c2f BB |
1821 | if (fidp->fid_len == LONG_FID_LEN) { |
1822 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1823 | uint64_t objsetid = 0; | |
1824 | uint64_t setgen = 0; | |
1825 | ||
1826 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1827 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1828 | ||
1829 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1830 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1831 | ||
9b77d1c9 CC |
1832 | if (objsetid != ZFSCTL_INO_SNAPDIRS - object) { |
1833 | dprintf("snapdir fid: objsetid (%llu) != " | |
1834 | "ZFSCTL_INO_SNAPDIRS (%llu) - object (%llu)\n", | |
1835 | objsetid, ZFSCTL_INO_SNAPDIRS, object); | |
34dc7c2f | 1836 | |
2e528b49 | 1837 | return (SET_ERROR(EINVAL)); |
9b77d1c9 | 1838 | } |
ebe7e575 | 1839 | |
9b77d1c9 CC |
1840 | if (fid_gen > 1 || setgen != 0) { |
1841 | dprintf("snapdir fid: fid_gen (%llu) and setgen " | |
1842 | "(%llu)\n", fid_gen, setgen); | |
1843 | return (SET_ERROR(EINVAL)); | |
1844 | } | |
34dc7c2f | 1845 | |
9b77d1c9 | 1846 | return (zfsctl_snapdir_vget(sb, objsetid, fid_gen, ipp)); |
34dc7c2f BB |
1847 | } |
1848 | ||
0037b49e | 1849 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1850 | /* A zero fid_gen means we are in the .zfs control directories */ |
1851 | if (fid_gen == 0 && | |
1852 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
0037b49e | 1853 | *ipp = zfsvfs->z_ctldir; |
3558fd73 | 1854 | ASSERT(*ipp != NULL); |
34dc7c2f | 1855 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1856 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1857 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1858 | } else { |
3558fd73 | 1859 | igrab(*ipp); |
34dc7c2f | 1860 | } |
0037b49e | 1861 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1862 | return (0); |
1863 | } | |
1864 | ||
1865 | gen_mask = -1ULL >> (64 - 8 * i); | |
1866 | ||
29e57d15 | 1867 | dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); |
0037b49e BB |
1868 | if ((err = zfs_zget(zfsvfs, object, &zp))) { |
1869 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
1870 | return (err); |
1871 | } | |
7938c2ac CC |
1872 | |
1873 | /* Don't export xattr stuff */ | |
1874 | if (zp->z_pflags & ZFS_XATTR) { | |
1875 | iput(ZTOI(zp)); | |
0037b49e | 1876 | ZFS_EXIT(zfsvfs); |
7938c2ac CC |
1877 | return (SET_ERROR(ENOENT)); |
1878 | } | |
1879 | ||
0037b49e | 1880 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen, |
428870ff BB |
1881 | sizeof (uint64_t)); |
1882 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1883 | if (zp_gen == 0) |
1884 | zp_gen = 1; | |
0037b49e | 1885 | if ((fid_gen == 0) && (zfsvfs->z_root == object)) |
0500e835 | 1886 | fid_gen = zp_gen; |
34dc7c2f | 1887 | if (zp->z_unlinked || zp_gen != fid_gen) { |
29e57d15 NB |
1888 | dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, |
1889 | fid_gen); | |
3558fd73 | 1890 | iput(ZTOI(zp)); |
0037b49e | 1891 | ZFS_EXIT(zfsvfs); |
6c253064 | 1892 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1893 | } |
1894 | ||
3558fd73 BB |
1895 | *ipp = ZTOI(zp); |
1896 | if (*ipp) | |
1897 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1898 | |
0037b49e | 1899 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1900 | return (0); |
1901 | } | |
1902 | ||
1903 | /* | |
0037b49e | 1904 | * Block out VFS ops and close zfsvfs_t |
34dc7c2f BB |
1905 | * |
1906 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
831baf06 KW |
1907 | * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying |
1908 | * dataset and objset intact so that they can be atomically handed off during | |
1909 | * a subsequent rollback or recv operation and the resume thereafter. | |
34dc7c2f BB |
1910 | */ |
1911 | int | |
0037b49e | 1912 | zfs_suspend_fs(zfsvfs_t *zfsvfs) |
34dc7c2f BB |
1913 | { |
1914 | int error; | |
1915 | ||
0037b49e | 1916 | if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0) |
34dc7c2f | 1917 | return (error); |
7b3e34ba | 1918 | |
34dc7c2f BB |
1919 | return (0); |
1920 | } | |
1921 | ||
1922 | /* | |
8614ddf9 MA |
1923 | * Rebuild SA and release VOPs. Note that ownership of the underlying dataset |
1924 | * is an invariant across any of the operations that can be performed while the | |
1925 | * filesystem was suspended. Whether it succeeded or failed, the preconditions | |
1926 | * are the same: the relevant objset and associated dataset are owned by | |
1927 | * zfsvfs, held, and long held on entry. | |
34dc7c2f BB |
1928 | */ |
1929 | int | |
0037b49e | 1930 | zfs_resume_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) |
34dc7c2f | 1931 | { |
2008e920 | 1932 | int err, err2; |
831baf06 | 1933 | znode_t *zp; |
34dc7c2f | 1934 | |
0037b49e BB |
1935 | ASSERT(RRM_WRITE_HELD(&zfsvfs->z_teardown_lock)); |
1936 | ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); | |
34dc7c2f | 1937 | |
831baf06 | 1938 | /* |
ec923db2 GM |
1939 | * We already own this, so just update the objset_t, as the one we |
1940 | * had before may have been evicted. | |
831baf06 | 1941 | */ |
8614ddf9 | 1942 | objset_t *os; |
0037b49e | 1943 | VERIFY3P(ds->ds_owner, ==, zfsvfs); |
ec923db2 | 1944 | VERIFY(dsl_dataset_long_held(ds)); |
8614ddf9 | 1945 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
c1fabe79 | 1946 | |
8614ddf9 MA |
1947 | err = zfsvfs_init(zfsvfs, os); |
1948 | if (err != 0) | |
831baf06 | 1949 | goto bail; |
34dc7c2f | 1950 | |
f298b24d | 1951 | VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); |
c1fabe79 | 1952 | |
0037b49e BB |
1953 | zfs_set_fuid_feature(zfsvfs); |
1954 | zfsvfs->z_rollback_time = jiffies; | |
34dc7c2f | 1955 | |
831baf06 KW |
1956 | /* |
1957 | * Attempt to re-establish all the active inodes with their | |
1958 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
1959 | * and mark it stale. This prevents a collision if a new | |
1960 | * inode/object is created which must use the same inode | |
1961 | * number. The stale inode will be be released when the | |
1962 | * VFS prunes the dentry holding the remaining references | |
1963 | * on the stale inode. | |
1964 | */ | |
0037b49e BB |
1965 | mutex_enter(&zfsvfs->z_znodes_lock); |
1966 | for (zp = list_head(&zfsvfs->z_all_znodes); zp; | |
1967 | zp = list_next(&zfsvfs->z_all_znodes, zp)) { | |
2008e920 TC |
1968 | err2 = zfs_rezget(zp); |
1969 | if (err2) { | |
831baf06 KW |
1970 | remove_inode_hash(ZTOI(zp)); |
1971 | zp->z_is_stale = B_TRUE; | |
34dc7c2f | 1972 | } |
34dc7c2f | 1973 | } |
0037b49e | 1974 | mutex_exit(&zfsvfs->z_znodes_lock); |
34dc7c2f | 1975 | |
428870ff | 1976 | bail: |
7b3e34ba | 1977 | /* release the VFS ops */ |
0037b49e BB |
1978 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
1979 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1980 | |
1981 | if (err) { | |
1982 | /* | |
831baf06 KW |
1983 | * Since we couldn't setup the sa framework, try to force |
1984 | * unmount this file system. | |
34dc7c2f | 1985 | */ |
0037b49e BB |
1986 | if (zfsvfs->z_os) |
1987 | (void) zfs_umount(zfsvfs->z_sb); | |
34dc7c2f BB |
1988 | } |
1989 | return (err); | |
1990 | } | |
1991 | ||
34dc7c2f | 1992 | int |
0037b49e | 1993 | zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) |
34dc7c2f BB |
1994 | { |
1995 | int error; | |
0037b49e | 1996 | objset_t *os = zfsvfs->z_os; |
34dc7c2f | 1997 | dmu_tx_t *tx; |
34dc7c2f BB |
1998 | |
1999 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
2e528b49 | 2000 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 2001 | |
0037b49e | 2002 | if (newvers < zfsvfs->z_version) |
2e528b49 | 2003 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 2004 | |
428870ff | 2005 | if (zfs_spa_version_map(newvers) > |
0037b49e | 2006 | spa_version(dmu_objset_spa(zfsvfs->z_os))) |
2e528b49 | 2007 | return (SET_ERROR(ENOTSUP)); |
428870ff | 2008 | |
34dc7c2f | 2009 | tx = dmu_tx_create(os); |
9babb374 | 2010 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
0037b49e | 2011 | if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { |
428870ff BB |
2012 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
2013 | ZFS_SA_ATTRS); | |
2014 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
2015 | } | |
34dc7c2f BB |
2016 | error = dmu_tx_assign(tx, TXG_WAIT); |
2017 | if (error) { | |
2018 | dmu_tx_abort(tx); | |
9babb374 BB |
2019 | return (error); |
2020 | } | |
428870ff | 2021 | |
9babb374 BB |
2022 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
2023 | 8, 1, &newvers, tx); | |
2024 | ||
2025 | if (error) { | |
2026 | dmu_tx_commit(tx); | |
2027 | return (error); | |
34dc7c2f | 2028 | } |
34dc7c2f | 2029 | |
0037b49e | 2030 | if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { |
428870ff BB |
2031 | uint64_t sa_obj; |
2032 | ||
0037b49e | 2033 | ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=, |
428870ff BB |
2034 | SPA_VERSION_SA); |
2035 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
2036 | DMU_OT_NONE, 0, tx); | |
2037 | ||
2038 | error = zap_add(os, MASTER_NODE_OBJ, | |
2039 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c99c9001 | 2040 | ASSERT0(error); |
428870ff BB |
2041 | |
2042 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
2043 | sa_register_update_callback(os, zfs_sa_upgrade); | |
2044 | } | |
2045 | ||
6f1ffb06 | 2046 | spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, |
0037b49e | 2047 | "from %llu to %llu", zfsvfs->z_version, newvers); |
9babb374 | 2048 | |
34dc7c2f BB |
2049 | dmu_tx_commit(tx); |
2050 | ||
0037b49e | 2051 | zfsvfs->z_version = newvers; |
9babb374 | 2052 | |
0037b49e | 2053 | zfs_set_fuid_feature(zfsvfs); |
9babb374 BB |
2054 | |
2055 | return (0); | |
34dc7c2f BB |
2056 | } |
2057 | ||
2058 | /* | |
2059 | * Read a property stored within the master node. | |
2060 | */ | |
2061 | int | |
2062 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
2063 | { | |
2064 | const char *pname; | |
2e528b49 | 2065 | int error = SET_ERROR(ENOENT); |
34dc7c2f BB |
2066 | |
2067 | /* | |
2068 | * Look up the file system's value for the property. For the | |
2069 | * version property, we look up a slightly different string. | |
2070 | */ | |
2071 | if (prop == ZFS_PROP_VERSION) | |
2072 | pname = ZPL_VERSION_STR; | |
2073 | else | |
2074 | pname = zfs_prop_to_name(prop); | |
2075 | ||
87a275d9 AG |
2076 | if (os != NULL) { |
2077 | ASSERT3U(os->os_phys->os_type, ==, DMU_OST_ZFS); | |
b128c09f | 2078 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); |
87a275d9 | 2079 | } |
34dc7c2f BB |
2080 | |
2081 | if (error == ENOENT) { | |
2082 | /* No value set, use the default value */ | |
2083 | switch (prop) { | |
2084 | case ZFS_PROP_VERSION: | |
2085 | *value = ZPL_VERSION; | |
2086 | break; | |
2087 | case ZFS_PROP_NORMALIZE: | |
2088 | case ZFS_PROP_UTF8ONLY: | |
2089 | *value = 0; | |
2090 | break; | |
2091 | case ZFS_PROP_CASE: | |
2092 | *value = ZFS_CASE_SENSITIVE; | |
2093 | break; | |
023699cd MM |
2094 | case ZFS_PROP_ACLTYPE: |
2095 | *value = ZFS_ACLTYPE_OFF; | |
2096 | break; | |
34dc7c2f BB |
2097 | default: |
2098 | return (error); | |
2099 | } | |
2100 | error = 0; | |
2101 | } | |
2102 | return (error); | |
2103 | } | |
3558fd73 | 2104 | |
a08abc1b GM |
2105 | /* |
2106 | * Return true if the coresponding vfs's unmounted flag is set. | |
2107 | * Otherwise return false. | |
2108 | * If this function returns true we know VFS unmount has been initiated. | |
2109 | */ | |
2110 | boolean_t | |
2111 | zfs_get_vfs_flag_unmounted(objset_t *os) | |
2112 | { | |
0037b49e | 2113 | zfsvfs_t *zfvp; |
a08abc1b GM |
2114 | boolean_t unmounted = B_FALSE; |
2115 | ||
2116 | ASSERT(dmu_objset_type(os) == DMU_OST_ZFS); | |
2117 | ||
2118 | mutex_enter(&os->os_user_ptr_lock); | |
2119 | zfvp = dmu_objset_get_user(os); | |
2120 | if (zfvp != NULL && zfvp->z_unmounted) | |
2121 | unmounted = B_TRUE; | |
2122 | mutex_exit(&os->os_user_ptr_lock); | |
2123 | ||
2124 | return (unmounted); | |
2125 | } | |
2126 | ||
3558fd73 BB |
2127 | void |
2128 | zfs_init(void) | |
2129 | { | |
ebe7e575 | 2130 | zfsctl_init(); |
3558fd73 BB |
2131 | zfs_znode_init(); |
2132 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
2133 | register_filesystem(&zpl_fs_type); | |
2134 | } | |
2135 | ||
2136 | void | |
2137 | zfs_fini(void) | |
2138 | { | |
8e71ab99 CC |
2139 | /* |
2140 | * we don't use outstanding because zpl_posix_acl_free might add more. | |
2141 | */ | |
57ddcda1 | 2142 | taskq_wait(system_delay_taskq); |
8e71ab99 | 2143 | taskq_wait(system_taskq); |
3558fd73 BB |
2144 | unregister_filesystem(&zpl_fs_type); |
2145 | zfs_znode_fini(); | |
ebe7e575 | 2146 | zfsctl_fini(); |
3558fd73 | 2147 | } |
f298b24d BB |
2148 | |
2149 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
2150 | EXPORT_SYMBOL(zfs_suspend_fs); | |
2151 | EXPORT_SYMBOL(zfs_resume_fs); | |
2152 | EXPORT_SYMBOL(zfs_userspace_one); | |
2153 | EXPORT_SYMBOL(zfs_userspace_many); | |
2154 | EXPORT_SYMBOL(zfs_set_userquota); | |
2155 | EXPORT_SYMBOL(zfs_owner_overquota); | |
2156 | EXPORT_SYMBOL(zfs_fuid_overquota); | |
2157 | EXPORT_SYMBOL(zfs_fuid_overobjquota); | |
2158 | EXPORT_SYMBOL(zfs_set_version); | |
2159 | EXPORT_SYMBOL(zfsvfs_create); | |
2160 | EXPORT_SYMBOL(zfsvfs_free); | |
2161 | EXPORT_SYMBOL(zfs_is_readonly); | |
2162 | EXPORT_SYMBOL(zfs_domount); | |
2163 | EXPORT_SYMBOL(zfs_preumount); | |
2164 | EXPORT_SYMBOL(zfs_umount); | |
2165 | EXPORT_SYMBOL(zfs_remount); | |
2166 | EXPORT_SYMBOL(zfs_statvfs); | |
2167 | EXPORT_SYMBOL(zfs_vget); | |
2168 | EXPORT_SYMBOL(zfs_prune); | |
2169 | #endif |