]>
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 | ||
607 | fuid = strtonum(fuidstr, NULL); | |
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 BB |
836 | if (!dmu_objset_userobjspace_present(zfsvfs->z_os)) { |
837 | if (dmu_objset_userobjspace_upgradable(zfsvfs->z_os)) | |
838 | dmu_objset_userobjspace_upgrade(zfsvfs->z_os); | |
1de321e6 JX |
839 | return (B_FALSE); |
840 | } | |
841 | ||
842 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
0037b49e BB |
843 | quotaobj = isgroup ? zfsvfs->z_groupobjquota_obj : |
844 | zfsvfs->z_userobjquota_obj; | |
845 | if (quotaobj == 0 || zfsvfs->z_replay) | |
1de321e6 JX |
846 | return (B_FALSE); |
847 | ||
848 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
0037b49e | 849 | err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); |
1de321e6 JX |
850 | if (err != 0) |
851 | return (B_FALSE); | |
852 | ||
853 | (void) sprintf(buf, DMU_OBJACCT_PREFIX "%llx", (longlong_t)fuid); | |
0037b49e | 854 | err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); |
1de321e6 JX |
855 | if (err != 0) |
856 | return (B_FALSE); | |
857 | return (used >= quota); | |
858 | } | |
859 | ||
9babb374 | 860 | boolean_t |
0037b49e | 861 | zfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) |
9babb374 | 862 | { |
1de321e6 | 863 | char buf[20]; |
9babb374 BB |
864 | uint64_t used, quota, usedobj, quotaobj; |
865 | int err; | |
866 | ||
867 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
0037b49e | 868 | quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; |
9babb374 | 869 | |
0037b49e | 870 | if (quotaobj == 0 || zfsvfs->z_replay) |
9babb374 BB |
871 | return (B_FALSE); |
872 | ||
873 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
0037b49e | 874 | err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
875 | if (err != 0) |
876 | return (B_FALSE); | |
877 | ||
0037b49e | 878 | err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
879 | if (err != 0) |
880 | return (B_FALSE); | |
881 | return (used >= quota); | |
882 | } | |
883 | ||
428870ff | 884 | boolean_t |
0037b49e | 885 | zfs_owner_overquota(zfsvfs_t *zfsvfs, znode_t *zp, boolean_t isgroup) |
428870ff BB |
886 | { |
887 | uint64_t fuid; | |
888 | uint64_t quotaobj; | |
2c6abf15 | 889 | struct inode *ip = ZTOI(zp); |
428870ff | 890 | |
0037b49e | 891 | quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; |
428870ff | 892 | |
2c6abf15 | 893 | fuid = isgroup ? KGID_TO_SGID(ip->i_gid) : KUID_TO_SUID(ip->i_uid); |
428870ff | 894 | |
0037b49e | 895 | if (quotaobj == 0 || zfsvfs->z_replay) |
428870ff BB |
896 | return (B_FALSE); |
897 | ||
0037b49e | 898 | return (zfs_fuid_overquota(zfsvfs, isgroup, fuid)); |
428870ff BB |
899 | } |
900 | ||
8614ddf9 MA |
901 | /* |
902 | * Associate this zfsvfs with the given objset, which must be owned. | |
903 | * This will cache a bunch of on-disk state from the objset in the | |
904 | * zfsvfs. | |
905 | */ | |
906 | static int | |
907 | zfsvfs_init(zfsvfs_t *zfsvfs, objset_t *os) | |
9babb374 | 908 | { |
8614ddf9 MA |
909 | int error; |
910 | uint64_t val; | |
0282c413 | 911 | |
0037b49e BB |
912 | zfsvfs->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; |
913 | zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; | |
914 | zfsvfs->z_os = os; | |
9babb374 | 915 | |
0037b49e | 916 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); |
8614ddf9 MA |
917 | if (error != 0) |
918 | return (error); | |
919 | if (zfsvfs->z_version > | |
920 | zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { | |
921 | (void) printk("Can't mount a version %lld file system " | |
922 | "on a version %lld pool\n. Pool must be upgraded to mount " | |
923 | "this file system.", (u_longlong_t)zfsvfs->z_version, | |
924 | (u_longlong_t)spa_version(dmu_objset_spa(os))); | |
925 | return (SET_ERROR(ENOTSUP)); | |
9babb374 | 926 | } |
8614ddf9 MA |
927 | error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &val); |
928 | if (error != 0) | |
929 | return (error); | |
930 | zfsvfs->z_norm = (int)val; | |
9babb374 | 931 | |
8614ddf9 MA |
932 | error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &val); |
933 | if (error != 0) | |
934 | return (error); | |
935 | zfsvfs->z_utf8 = (val != 0); | |
9babb374 | 936 | |
8614ddf9 MA |
937 | error = zfs_get_zplprop(os, ZFS_PROP_CASE, &val); |
938 | if (error != 0) | |
939 | return (error); | |
940 | zfsvfs->z_case = (uint_t)val; | |
9babb374 | 941 | |
8614ddf9 MA |
942 | if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &val)) != 0) |
943 | return (error); | |
944 | zfsvfs->z_acl_type = (uint_t)val; | |
023699cd | 945 | |
9babb374 BB |
946 | /* |
947 | * Fold case on file systems that are always or sometimes case | |
948 | * insensitive. | |
949 | */ | |
0037b49e BB |
950 | if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
951 | zfsvfs->z_case == ZFS_CASE_MIXED) | |
952 | zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 953 | |
0037b49e BB |
954 | zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); |
955 | zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); | |
428870ff | 956 | |
8614ddf9 | 957 | uint64_t sa_obj = 0; |
0037b49e | 958 | if (zfsvfs->z_use_sa) { |
428870ff BB |
959 | /* should either have both of these objects or none */ |
960 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
961 | &sa_obj); | |
8614ddf9 MA |
962 | if (error != 0) |
963 | return (error); | |
82a37189 | 964 | |
8614ddf9 MA |
965 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &val); |
966 | if ((error == 0) && (val == ZFS_XATTR_SA)) | |
0037b49e | 967 | zfsvfs->z_xattr_sa = B_TRUE; |
428870ff BB |
968 | } |
969 | ||
572e2857 | 970 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
0037b49e | 971 | &zfsvfs->z_attr_table); |
8614ddf9 MA |
972 | if (error != 0) |
973 | return (error); | |
428870ff | 974 | |
0037b49e | 975 | if (zfsvfs->z_version >= ZPL_VERSION_SA) |
428870ff | 976 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
977 | |
978 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
0037b49e | 979 | &zfsvfs->z_root); |
8614ddf9 MA |
980 | if (error != 0) |
981 | return (error); | |
0037b49e | 982 | ASSERT(zfsvfs->z_root != 0); |
9babb374 BB |
983 | |
984 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
0037b49e | 985 | &zfsvfs->z_unlinkedobj); |
8614ddf9 MA |
986 | if (error != 0) |
987 | return (error); | |
9babb374 BB |
988 | |
989 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
990 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
0037b49e | 991 | 8, 1, &zfsvfs->z_userquota_obj); |
8614ddf9 MA |
992 | if (error == ENOENT) |
993 | zfsvfs->z_userquota_obj = 0; | |
994 | else if (error != 0) | |
995 | return (error); | |
9babb374 BB |
996 | |
997 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
998 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
0037b49e | 999 | 8, 1, &zfsvfs->z_groupquota_obj); |
8614ddf9 MA |
1000 | if (error == ENOENT) |
1001 | zfsvfs->z_groupquota_obj = 0; | |
1002 | else if (error != 0) | |
1003 | return (error); | |
9babb374 | 1004 | |
1de321e6 JX |
1005 | error = zap_lookup(os, MASTER_NODE_OBJ, |
1006 | zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA], | |
0037b49e | 1007 | 8, 1, &zfsvfs->z_userobjquota_obj); |
8614ddf9 MA |
1008 | if (error == ENOENT) |
1009 | zfsvfs->z_userobjquota_obj = 0; | |
1010 | else if (error != 0) | |
1011 | return (error); | |
1de321e6 JX |
1012 | |
1013 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
1014 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA], | |
0037b49e | 1015 | 8, 1, &zfsvfs->z_groupobjquota_obj); |
8614ddf9 MA |
1016 | if (error == ENOENT) |
1017 | zfsvfs->z_groupobjquota_obj = 0; | |
1018 | else if (error != 0) | |
1019 | return (error); | |
1de321e6 | 1020 | |
9babb374 | 1021 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, |
0037b49e | 1022 | &zfsvfs->z_fuid_obj); |
8614ddf9 MA |
1023 | if (error == ENOENT) |
1024 | zfsvfs->z_fuid_obj = 0; | |
1025 | else if (error != 0) | |
1026 | return (error); | |
9babb374 BB |
1027 | |
1028 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
0037b49e | 1029 | &zfsvfs->z_shares_dir); |
8614ddf9 MA |
1030 | if (error == ENOENT) |
1031 | zfsvfs->z_shares_dir = 0; | |
1032 | else if (error != 0) | |
1033 | return (error); | |
1034 | ||
1035 | return (0); | |
1036 | } | |
1037 | ||
1038 | int | |
1039 | zfsvfs_create(const char *osname, zfsvfs_t **zfvp) | |
1040 | { | |
1041 | objset_t *os; | |
1042 | zfsvfs_t *zfsvfs; | |
1043 | int error; | |
1044 | ||
1045 | zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); | |
1046 | ||
1047 | /* | |
1048 | * We claim to always be readonly so we can open snapshots; | |
1049 | * other ZPL code will prevent us from writing to snapshots. | |
1050 | */ | |
1051 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zfsvfs, &os); | |
1052 | if (error) { | |
1053 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); | |
1054 | return (error); | |
1055 | } | |
1056 | ||
1057 | zfsvfs->z_vfs = NULL; | |
1058 | zfsvfs->z_sb = NULL; | |
1059 | zfsvfs->z_parent = zfsvfs; | |
9babb374 | 1060 | |
0037b49e BB |
1061 | mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
1062 | mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
1063 | list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), | |
9babb374 | 1064 | offsetof(znode_t, z_link_node)); |
0037b49e BB |
1065 | rrm_init(&zfsvfs->z_teardown_lock, B_FALSE); |
1066 | rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); | |
1067 | rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
d07b7c7f | 1068 | |
8614ddf9 MA |
1069 | int size = MIN(1 << (highbit64(zfs_object_mutex_size) - 1), |
1070 | ZFS_OBJ_MTX_MAX); | |
0037b49e BB |
1071 | zfsvfs->z_hold_size = size; |
1072 | zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, | |
1073 | KM_SLEEP); | |
1074 | zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
8614ddf9 | 1075 | for (int i = 0; i != size; i++) { |
0037b49e | 1076 | avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare, |
c96c36fa | 1077 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); |
0037b49e | 1078 | mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); |
c96c36fa | 1079 | } |
9babb374 | 1080 | |
8614ddf9 MA |
1081 | error = zfsvfs_init(zfsvfs, os); |
1082 | if (error != 0) { | |
1083 | dmu_objset_disown(os, zfsvfs); | |
1084 | *zfvp = NULL; | |
1085 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); | |
1086 | return (error); | |
1087 | } | |
1088 | ||
0037b49e | 1089 | *zfvp = zfsvfs; |
9babb374 | 1090 | return (0); |
9babb374 BB |
1091 | } |
1092 | ||
8614ddf9 | 1093 | static int |
f298b24d | 1094 | zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) |
34dc7c2f | 1095 | { |
34dc7c2f BB |
1096 | int error; |
1097 | ||
1c2555ef | 1098 | error = zfs_register_callbacks(zfsvfs->z_vfs); |
34dc7c2f BB |
1099 | if (error) |
1100 | return (error); | |
1101 | ||
0037b49e | 1102 | zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); |
9babb374 | 1103 | |
34dc7c2f BB |
1104 | /* |
1105 | * If we are not mounting (ie: online recv), then we don't | |
1106 | * have to worry about replaying the log as we blocked all | |
1107 | * operations out since we closed the ZIL. | |
1108 | */ | |
1109 | if (mounting) { | |
b128c09f BB |
1110 | boolean_t readonly; |
1111 | ||
34dc7c2f BB |
1112 | /* |
1113 | * During replay we remove the read only flag to | |
1114 | * allow replays to succeed. | |
1115 | */ | |
0037b49e | 1116 | readonly = zfs_is_readonly(zfsvfs); |
fb5f0bc8 | 1117 | if (readonly != 0) |
0037b49e | 1118 | readonly_changed_cb(zfsvfs, B_FALSE); |
fb5f0bc8 | 1119 | else |
0037b49e | 1120 | zfs_unlinked_drain(zfsvfs); |
34dc7c2f | 1121 | |
428870ff BB |
1122 | /* |
1123 | * Parse and replay the intent log. | |
1124 | * | |
1125 | * Because of ziltest, this must be done after | |
1126 | * zfs_unlinked_drain(). (Further note: ziltest | |
1127 | * doesn't use readonly mounts, where | |
1128 | * zfs_unlinked_drain() isn't called.) This is because | |
1129 | * ziltest causes spa_sync() to think it's committed, | |
1130 | * but actually it is not, so the intent log contains | |
1131 | * many txg's worth of changes. | |
1132 | * | |
1133 | * In particular, if object N is in the unlinked set in | |
1134 | * the last txg to actually sync, then it could be | |
1135 | * actually freed in a later txg and then reallocated | |
1136 | * in a yet later txg. This would write a "create | |
1137 | * object N" record to the intent log. Normally, this | |
1138 | * would be fine because the spa_sync() would have | |
1139 | * written out the fact that object N is free, before | |
1140 | * we could write the "create object N" intent log | |
1141 | * record. | |
1142 | * | |
1143 | * But when we are in ziltest mode, we advance the "open | |
1144 | * txg" without actually spa_sync()-ing the changes to | |
1145 | * disk. So we would see that object N is still | |
1146 | * allocated and in the unlinked set, and there is an | |
1147 | * intent log record saying to allocate it. | |
1148 | */ | |
0037b49e | 1149 | if (spa_writeable(dmu_objset_spa(zfsvfs->z_os))) { |
572e2857 | 1150 | if (zil_replay_disable) { |
0037b49e | 1151 | zil_destroy(zfsvfs->z_log, B_FALSE); |
572e2857 | 1152 | } else { |
0037b49e BB |
1153 | zfsvfs->z_replay = B_TRUE; |
1154 | zil_replay(zfsvfs->z_os, zfsvfs, | |
572e2857 | 1155 | zfs_replay_vector); |
0037b49e | 1156 | zfsvfs->z_replay = B_FALSE; |
572e2857 | 1157 | } |
fb5f0bc8 | 1158 | } |
2cf7f52b BB |
1159 | |
1160 | /* restore readonly bit */ | |
1161 | if (readonly != 0) | |
0037b49e | 1162 | readonly_changed_cb(zfsvfs, B_TRUE); |
34dc7c2f BB |
1163 | } |
1164 | ||
9775e988 | 1165 | /* |
0037b49e | 1166 | * Set the objset user_ptr to track its zfsvfs. |
9775e988 | 1167 | */ |
0037b49e BB |
1168 | mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); |
1169 | dmu_objset_set_user(zfsvfs->z_os, zfsvfs); | |
1170 | mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); | |
9775e988 | 1171 | |
34dc7c2f BB |
1172 | return (0); |
1173 | } | |
1174 | ||
9babb374 | 1175 | void |
f298b24d | 1176 | zfsvfs_free(zfsvfs_t *zfsvfs) |
34dc7c2f | 1177 | { |
0037b49e | 1178 | int i, size = zfsvfs->z_hold_size; |
9babb374 | 1179 | |
0037b49e | 1180 | zfs_fuid_destroy(zfsvfs); |
9babb374 | 1181 | |
0037b49e BB |
1182 | mutex_destroy(&zfsvfs->z_znodes_lock); |
1183 | mutex_destroy(&zfsvfs->z_lock); | |
1184 | list_destroy(&zfsvfs->z_all_znodes); | |
1185 | rrm_destroy(&zfsvfs->z_teardown_lock); | |
1186 | rw_destroy(&zfsvfs->z_teardown_inactive_lock); | |
1187 | rw_destroy(&zfsvfs->z_fuid_lock); | |
c96c36fa | 1188 | for (i = 0; i != size; i++) { |
0037b49e BB |
1189 | avl_destroy(&zfsvfs->z_hold_trees[i]); |
1190 | mutex_destroy(&zfsvfs->z_hold_locks[i]); | |
c96c36fa | 1191 | } |
0037b49e BB |
1192 | vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size); |
1193 | vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size); | |
1c2555ef | 1194 | zfsvfs_vfs_free(zfsvfs->z_vfs); |
0037b49e | 1195 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); |
34dc7c2f BB |
1196 | } |
1197 | ||
9babb374 | 1198 | static void |
0037b49e | 1199 | zfs_set_fuid_feature(zfsvfs_t *zfsvfs) |
9babb374 | 1200 | { |
0037b49e BB |
1201 | zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); |
1202 | zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); | |
9babb374 | 1203 | } |
34dc7c2f BB |
1204 | |
1205 | void | |
0037b49e | 1206 | zfs_unregister_callbacks(zfsvfs_t *zfsvfs) |
34dc7c2f | 1207 | { |
0037b49e | 1208 | objset_t *os = zfsvfs->z_os; |
34dc7c2f | 1209 | |
0eb21616 | 1210 | if (!dmu_objset_is_snapshot(os)) |
0037b49e | 1211 | dsl_prop_unregister_all(dmu_objset_ds(os), zfsvfs); |
34dc7c2f BB |
1212 | } |
1213 | ||
bc3e15e3 | 1214 | #ifdef HAVE_MLSLABEL |
428870ff | 1215 | /* |
d3cc8b15 WA |
1216 | * Check that the hex label string is appropriate for the dataset being |
1217 | * mounted into the global_zone proper. | |
428870ff | 1218 | * |
d3cc8b15 WA |
1219 | * Return an error if the hex label string is not default or |
1220 | * admin_low/admin_high. For admin_low labels, the corresponding | |
1221 | * dataset must be readonly. | |
428870ff BB |
1222 | */ |
1223 | int | |
1224 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
1225 | { | |
1226 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
1227 | return (0); | |
1228 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
1229 | return (0); | |
1230 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
1231 | /* must be readonly */ | |
1232 | uint64_t rdonly; | |
1233 | ||
1234 | if (dsl_prop_get_integer(dsname, | |
1235 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
2e528b49 | 1236 | return (SET_ERROR(EACCES)); |
428870ff BB |
1237 | return (rdonly ? 0 : EACCES); |
1238 | } | |
2e528b49 | 1239 | return (SET_ERROR(EACCES)); |
428870ff | 1240 | } |
bc3e15e3 | 1241 | #endif /* HAVE_MLSLABEL */ |
428870ff | 1242 | |
e5c39b95 | 1243 | int |
3558fd73 | 1244 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 1245 | { |
0037b49e | 1246 | zfsvfs_t *zfsvfs = dentry->d_sb->s_fs_info; |
34dc7c2f | 1247 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
04f9432d | 1248 | uint64_t fsid; |
3558fd73 | 1249 | uint32_t bshift; |
34dc7c2f | 1250 | |
0037b49e | 1251 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1252 | |
0037b49e | 1253 | dmu_objset_space(zfsvfs->z_os, |
34dc7c2f BB |
1254 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
1255 | ||
0037b49e | 1256 | fsid = dmu_objset_fsid_guid(zfsvfs->z_os); |
34dc7c2f | 1257 | /* |
05ff35c6 BB |
1258 | * The underlying storage pool actually uses multiple block |
1259 | * size. Under Solaris frsize (fragment size) is reported as | |
1260 | * the smallest block size we support, and bsize (block size) | |
1261 | * as the filesystem's maximum block size. Unfortunately, | |
1262 | * under Linux the fragment size and block size are often used | |
1263 | * interchangeably. Thus we are forced to report both of them | |
1264 | * as the filesystem's maximum block size. | |
34dc7c2f | 1265 | */ |
0037b49e BB |
1266 | statp->f_frsize = zfsvfs->z_max_blksz; |
1267 | statp->f_bsize = zfsvfs->z_max_blksz; | |
3558fd73 | 1268 | bshift = fls(statp->f_bsize) - 1; |
34dc7c2f BB |
1269 | |
1270 | /* | |
3558fd73 BB |
1271 | * The following report "total" blocks of various kinds in |
1272 | * the file system, but reported in terms of f_bsize - the | |
1273 | * "preferred" size. | |
34dc7c2f BB |
1274 | */ |
1275 | ||
3558fd73 BB |
1276 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
1277 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
1278 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
1279 | ||
1280 | /* | |
1281 | * statvfs() should really be called statufs(), because it assumes | |
1282 | * static metadata. ZFS doesn't preallocate files, so the best | |
1283 | * we can do is report the max that could possibly fit in f_files, | |
1284 | * and that minus the number actually used in f_ffree. | |
1285 | * For f_ffree, report the smaller of the number of object available | |
1286 | * and the number of blocks (each object will take at least a block). | |
1287 | */ | |
baab0630 | 1288 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 1289 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
1290 | statp->f_fsid.val[0] = (uint32_t)fsid; |
1291 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 | 1292 | statp->f_type = ZFS_SUPER_MAGIC; |
eca7b760 | 1293 | statp->f_namelen = MAXNAMELEN - 1; |
34dc7c2f BB |
1294 | |
1295 | /* | |
3558fd73 | 1296 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
1297 | * Is there anything useful we could/should provide? |
1298 | */ | |
3558fd73 | 1299 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 1300 | |
0037b49e | 1301 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1302 | return (0); |
1303 | } | |
1304 | ||
e5c39b95 | 1305 | int |
0037b49e | 1306 | zfs_root(zfsvfs_t *zfsvfs, struct inode **ipp) |
34dc7c2f | 1307 | { |
34dc7c2f BB |
1308 | znode_t *rootzp; |
1309 | int error; | |
1310 | ||
0037b49e | 1311 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1312 | |
0037b49e | 1313 | error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp); |
34dc7c2f | 1314 | if (error == 0) |
3558fd73 | 1315 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1316 | |
0037b49e | 1317 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1318 | return (error); |
1319 | } | |
1320 | ||
09fb30e5 | 1321 | #ifdef HAVE_D_PRUNE_ALIASES |
218b4e0a BB |
1322 | /* |
1323 | * Linux kernels older than 3.1 do not support a per-filesystem shrinker. | |
1324 | * To accommodate this we must improvise and manually walk the list of znodes | |
1325 | * attempting to prune dentries in order to be able to drop the inodes. | |
1326 | * | |
1327 | * To avoid scanning the same znodes multiple times they are always rotated | |
1328 | * to the end of the z_all_znodes list. New znodes are inserted at the | |
1329 | * end of the list so we're always scanning the oldest znodes first. | |
1330 | */ | |
1331 | static int | |
f298b24d | 1332 | zfs_prune_aliases(zfsvfs_t *zfsvfs, unsigned long nr_to_scan) |
218b4e0a BB |
1333 | { |
1334 | znode_t **zp_array, *zp; | |
1335 | int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *)); | |
1336 | int objects = 0; | |
1337 | int i = 0, j = 0; | |
1338 | ||
1339 | zp_array = kmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP); | |
1340 | ||
0037b49e BB |
1341 | mutex_enter(&zfsvfs->z_znodes_lock); |
1342 | while ((zp = list_head(&zfsvfs->z_all_znodes)) != NULL) { | |
218b4e0a BB |
1343 | |
1344 | if ((i++ > nr_to_scan) || (j >= max_array)) | |
1345 | break; | |
1346 | ||
1347 | ASSERT(list_link_active(&zp->z_link_node)); | |
0037b49e BB |
1348 | list_remove(&zfsvfs->z_all_znodes, zp); |
1349 | list_insert_tail(&zfsvfs->z_all_znodes, zp); | |
218b4e0a BB |
1350 | |
1351 | /* Skip active znodes and .zfs entries */ | |
1352 | if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir) | |
1353 | continue; | |
1354 | ||
1355 | if (igrab(ZTOI(zp)) == NULL) | |
1356 | continue; | |
1357 | ||
1358 | zp_array[j] = zp; | |
1359 | j++; | |
1360 | } | |
0037b49e | 1361 | mutex_exit(&zfsvfs->z_znodes_lock); |
218b4e0a BB |
1362 | |
1363 | for (i = 0; i < j; i++) { | |
1364 | zp = zp_array[i]; | |
1365 | ||
1366 | ASSERT3P(zp, !=, NULL); | |
1367 | d_prune_aliases(ZTOI(zp)); | |
1368 | ||
1369 | if (atomic_read(&ZTOI(zp)->i_count) == 1) | |
1370 | objects++; | |
1371 | ||
1372 | iput(ZTOI(zp)); | |
1373 | } | |
1374 | ||
1375 | kmem_free(zp_array, max_array * sizeof (znode_t *)); | |
1376 | ||
1377 | return (objects); | |
1378 | } | |
1379 | #endif /* HAVE_D_PRUNE_ALIASES */ | |
1380 | ||
2cbb06b5 BB |
1381 | /* |
1382 | * The ARC has requested that the filesystem drop entries from the dentry | |
1383 | * and inode caches. This can occur when the ARC needs to free meta data | |
1384 | * blocks but can't because they are all pinned by entries in these caches. | |
1385 | */ | |
ab26409d | 1386 | int |
f298b24d | 1387 | zfs_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) |
ab26409d | 1388 | { |
0037b49e | 1389 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
2cbb06b5 BB |
1390 | int error = 0; |
1391 | #if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK) | |
ab26409d BB |
1392 | struct shrinker *shrinker = &sb->s_shrink; |
1393 | struct shrink_control sc = { | |
1394 | .nr_to_scan = nr_to_scan, | |
1395 | .gfp_mask = GFP_KERNEL, | |
1396 | }; | |
2cbb06b5 | 1397 | #endif |
ab26409d | 1398 | |
0037b49e | 1399 | ZFS_ENTER(zfsvfs); |
2cbb06b5 | 1400 | |
90947b23 TC |
1401 | #if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \ |
1402 | defined(SHRINK_CONTROL_HAS_NID) && \ | |
1403 | defined(SHRINKER_NUMA_AWARE) | |
1404 | if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) { | |
1405 | *objects = 0; | |
02730c33 | 1406 | for_each_online_node(sc.nid) { |
90947b23 | 1407 | *objects += (*shrinker->scan_objects)(shrinker, &sc); |
02730c33 | 1408 | } |
90947b23 TC |
1409 | } else { |
1410 | *objects = (*shrinker->scan_objects)(shrinker, &sc); | |
1411 | } | |
09fb30e5 | 1412 | |
90947b23 | 1413 | #elif defined(HAVE_SPLIT_SHRINKER_CALLBACK) |
3c832b8c | 1414 | *objects = (*shrinker->scan_objects)(shrinker, &sc); |
2cbb06b5 | 1415 | #elif defined(HAVE_SHRINK) |
ab26409d | 1416 | *objects = (*shrinker->shrink)(shrinker, &sc); |
218b4e0a | 1417 | #elif defined(HAVE_D_PRUNE_ALIASES) |
09fb30e5 | 1418 | #define D_PRUNE_ALIASES_IS_DEFAULT |
f298b24d | 1419 | *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); |
2cbb06b5 | 1420 | #else |
218b4e0a | 1421 | #error "No available dentry and inode cache pruning mechanism." |
3c832b8c | 1422 | #endif |
09fb30e5 TC |
1423 | |
1424 | #if defined(HAVE_D_PRUNE_ALIASES) && !defined(D_PRUNE_ALIASES_IS_DEFAULT) | |
1425 | #undef D_PRUNE_ALIASES_IS_DEFAULT | |
1426 | /* | |
f298b24d | 1427 | * Fall back to zfs_prune_aliases if the kernel's per-superblock |
09fb30e5 TC |
1428 | * shrinker couldn't free anything, possibly due to the inodes being |
1429 | * allocated in a different memcg. | |
1430 | */ | |
1431 | if (*objects == 0) | |
f298b24d | 1432 | *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); |
09fb30e5 TC |
1433 | #endif |
1434 | ||
0037b49e | 1435 | ZFS_EXIT(zfsvfs); |
ab26409d | 1436 | |
0037b49e | 1437 | dprintf_ds(zfsvfs->z_os->os_dsl_dataset, |
2cbb06b5 BB |
1438 | "pruning, nr_to_scan=%lu objects=%d error=%d\n", |
1439 | nr_to_scan, *objects, error); | |
1440 | ||
1441 | return (error); | |
ab26409d | 1442 | } |
ab26409d | 1443 | |
34dc7c2f | 1444 | /* |
0037b49e | 1445 | * Teardown the zfsvfs_t. |
34dc7c2f | 1446 | * |
a08abc1b | 1447 | * Note, if 'unmounting' is FALSE, we return with the 'z_teardown_lock' |
34dc7c2f BB |
1448 | * and 'z_teardown_inactive_lock' held. |
1449 | */ | |
f298b24d | 1450 | static int |
0037b49e | 1451 | zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) |
34dc7c2f BB |
1452 | { |
1453 | znode_t *zp; | |
1454 | ||
fd23720a BB |
1455 | /* |
1456 | * If someone has not already unmounted this file system, | |
1457 | * drain the iput_taskq to ensure all active references to the | |
0037b49e | 1458 | * zfsvfs_t have been handled only then can it be safely destroyed. |
fd23720a | 1459 | */ |
0037b49e | 1460 | if (zfsvfs->z_os) { |
f0da4d15 CD |
1461 | /* |
1462 | * If we're unmounting we have to wait for the list to | |
1463 | * drain completely. | |
1464 | * | |
1465 | * If we're not unmounting there's no guarantee the list | |
1466 | * will drain completely, but iputs run from the taskq | |
1467 | * may add the parents of dir-based xattrs to the taskq | |
1468 | * so we want to wait for these. | |
1469 | * | |
1470 | * We can safely read z_nr_znodes without locking because the | |
1471 | * VFS has already blocked operations which add to the | |
1472 | * z_all_znodes list and thus increment z_nr_znodes. | |
1473 | */ | |
1474 | int round = 0; | |
0037b49e | 1475 | while (zfsvfs->z_nr_znodes > 0) { |
c5528b9b | 1476 | taskq_wait_outstanding(dsl_pool_iput_taskq( |
0037b49e | 1477 | dmu_objset_pool(zfsvfs->z_os)), 0); |
f0da4d15 CD |
1478 | if (++round > 1 && !unmounting) |
1479 | break; | |
1480 | } | |
1481 | } | |
fd23720a | 1482 | |
0037b49e | 1483 | rrm_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1484 | |
1485 | if (!unmounting) { | |
1486 | /* | |
ceb43b93 BB |
1487 | * We purge the parent filesystem's super block as the |
1488 | * parent filesystem and all of its snapshots have their | |
1489 | * inode's super block set to the parent's filesystem's | |
1490 | * super block. Note, 'z_parent' is self referential | |
1491 | * for non-snapshots. | |
34dc7c2f | 1492 | */ |
0037b49e | 1493 | shrink_dcache_sb(zfsvfs->z_parent->z_sb); |
34dc7c2f BB |
1494 | } |
1495 | ||
1496 | /* | |
1497 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1498 | * threads are blocked as zil_close can call zfs_inactive. | |
1499 | */ | |
0037b49e BB |
1500 | if (zfsvfs->z_log) { |
1501 | zil_close(zfsvfs->z_log); | |
1502 | zfsvfs->z_log = NULL; | |
34dc7c2f BB |
1503 | } |
1504 | ||
0037b49e | 1505 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1506 | |
1507 | /* | |
1508 | * If we are not unmounting (ie: online recv) and someone already | |
1509 | * unmounted this file system while we were doing the switcheroo, | |
1510 | * or a reopen of z_os failed then just bail out now. | |
1511 | */ | |
0037b49e BB |
1512 | if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) { |
1513 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
1514 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
2e528b49 | 1515 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1516 | } |
1517 | ||
1518 | /* | |
7b3e34ba BB |
1519 | * At this point there are no VFS ops active, and any new VFS ops |
1520 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1521 | * relevant for forced unmount). | |
34dc7c2f BB |
1522 | * |
1523 | * Release all holds on dbufs. | |
1524 | */ | |
f0da4d15 | 1525 | if (!unmounting) { |
0037b49e BB |
1526 | mutex_enter(&zfsvfs->z_znodes_lock); |
1527 | for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL; | |
1528 | zp = list_next(&zfsvfs->z_all_znodes, zp)) { | |
f0da4d15 CD |
1529 | if (zp->z_sa_hdl) |
1530 | zfs_znode_dmu_fini(zp); | |
1531 | } | |
0037b49e | 1532 | mutex_exit(&zfsvfs->z_znodes_lock); |
7b3e34ba | 1533 | } |
34dc7c2f BB |
1534 | |
1535 | /* | |
7b3e34ba | 1536 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1537 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1538 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1539 | */ |
1540 | if (unmounting) { | |
0037b49e BB |
1541 | zfsvfs->z_unmounted = B_TRUE; |
1542 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
1543 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1544 | } |
1545 | ||
1546 | /* | |
1547 | * z_os will be NULL if there was an error in attempting to reopen | |
0037b49e | 1548 | * zfsvfs, so just return as the properties had already been |
3558fd73 | 1549 | * |
34dc7c2f BB |
1550 | * unregistered and cached data had been evicted before. |
1551 | */ | |
0037b49e | 1552 | if (zfsvfs->z_os == NULL) |
34dc7c2f BB |
1553 | return (0); |
1554 | ||
1555 | /* | |
1556 | * Unregister properties. | |
1557 | */ | |
0037b49e | 1558 | zfs_unregister_callbacks(zfsvfs); |
34dc7c2f BB |
1559 | |
1560 | /* | |
1561 | * Evict cached data | |
1562 | */ | |
0037b49e BB |
1563 | if (dsl_dataset_is_dirty(dmu_objset_ds(zfsvfs->z_os)) && |
1564 | !zfs_is_readonly(zfsvfs)) | |
1565 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); | |
1566 | dmu_objset_evict_dbufs(zfsvfs->z_os); | |
34dc7c2f BB |
1567 | |
1568 | return (0); | |
1569 | } | |
1570 | ||
8c45def2 BB |
1571 | #if !defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER) && \ |
1572 | !defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER) | |
5547c2f1 | 1573 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); |
8c45def2 | 1574 | #endif |
76659dc1 | 1575 | |
e5c39b95 | 1576 | int |
1c2555ef | 1577 | zfs_domount(struct super_block *sb, zfs_mnt_t *zm, int silent) |
34dc7c2f | 1578 | { |
1c2555ef | 1579 | const char *osname = zm->mnt_osname; |
3558fd73 BB |
1580 | struct inode *root_inode; |
1581 | uint64_t recordsize; | |
1c2555ef BB |
1582 | int error = 0; |
1583 | zfsvfs_t *zfsvfs; | |
1584 | ||
1585 | ASSERT(zm); | |
1586 | ASSERT(osname); | |
34dc7c2f | 1587 | |
1c2555ef | 1588 | error = zfsvfs_create(osname, &zfsvfs); |
3558fd73 BB |
1589 | if (error) |
1590 | return (error); | |
1591 | ||
1c2555ef BB |
1592 | error = zfsvfs_parse_options(zm->mnt_data, &zfsvfs->z_vfs); |
1593 | if (error) | |
1594 | goto out; | |
1595 | ||
3558fd73 BB |
1596 | if ((error = dsl_prop_get_integer(osname, "recordsize", |
1597 | &recordsize, NULL))) | |
1598 | goto out; | |
1599 | ||
1c2555ef | 1600 | zfsvfs->z_vfs->vfs_data = zfsvfs; |
0037b49e BB |
1601 | zfsvfs->z_sb = sb; |
1602 | sb->s_fs_info = zfsvfs; | |
3558fd73 BB |
1603 | sb->s_magic = ZFS_SUPER_MAGIC; |
1604 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1605 | sb->s_time_gran = 1; | |
1606 | sb->s_blocksize = recordsize; | |
1607 | sb->s_blocksize_bits = ilog2(recordsize); | |
0037b49e BB |
1608 | zfsvfs->z_bdi.ra_pages = 0; |
1609 | sb->s_bdi = &zfsvfs->z_bdi; | |
5547c2f1 | 1610 | |
0037b49e | 1611 | error = -zpl_bdi_setup_and_register(&zfsvfs->z_bdi, "zfs"); |
5547c2f1 BB |
1612 | if (error) |
1613 | goto out; | |
3558fd73 BB |
1614 | |
1615 | /* Set callback operations for the file system. */ | |
1616 | sb->s_op = &zpl_super_operations; | |
1617 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1618 | sb->s_export_op = &zpl_export_operations; |
ee930353 BB |
1619 | #ifdef HAVE_S_D_OP |
1620 | sb->s_d_op = &zpl_dentry_operations; | |
1621 | #endif /* HAVE_S_D_OP */ | |
3558fd73 BB |
1622 | |
1623 | /* Set features for file system. */ | |
0037b49e | 1624 | zfs_set_fuid_feature(zfsvfs); |
3558fd73 | 1625 | |
0037b49e | 1626 | if (dmu_objset_is_snapshot(zfsvfs->z_os)) { |
3558fd73 BB |
1627 | uint64_t pval; |
1628 | ||
0037b49e BB |
1629 | atime_changed_cb(zfsvfs, B_FALSE); |
1630 | readonly_changed_cb(zfsvfs, B_TRUE); | |
d1d7e268 MK |
1631 | if ((error = dsl_prop_get_integer(osname, |
1632 | "xattr", &pval, NULL))) | |
3558fd73 | 1633 | goto out; |
0037b49e | 1634 | xattr_changed_cb(zfsvfs, pval); |
d1d7e268 MK |
1635 | if ((error = dsl_prop_get_integer(osname, |
1636 | "acltype", &pval, NULL))) | |
023699cd | 1637 | goto out; |
0037b49e BB |
1638 | acltype_changed_cb(zfsvfs, pval); |
1639 | zfsvfs->z_issnap = B_TRUE; | |
1640 | zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1641 | zfsvfs->z_snap_defer_time = jiffies; | |
1642 | ||
1643 | mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); | |
1644 | dmu_objset_set_user(zfsvfs->z_os, zfsvfs); | |
1645 | mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); | |
3558fd73 | 1646 | } else { |
f298b24d | 1647 | if ((error = zfsvfs_setup(zfsvfs, B_TRUE))) |
be88e733 | 1648 | goto out; |
34dc7c2f BB |
1649 | } |
1650 | ||
3558fd73 | 1651 | /* Allocate a root inode for the filesystem. */ |
0037b49e | 1652 | error = zfs_root(zfsvfs, &root_inode); |
3558fd73 BB |
1653 | if (error) { |
1654 | (void) zfs_umount(sb); | |
1655 | goto out; | |
34dc7c2f BB |
1656 | } |
1657 | ||
3558fd73 | 1658 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1659 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1660 | if (sb->s_root == NULL) { |
1661 | (void) zfs_umount(sb); | |
2e528b49 | 1662 | error = SET_ERROR(ENOMEM); |
3558fd73 BB |
1663 | goto out; |
1664 | } | |
ebe7e575 | 1665 | |
0037b49e BB |
1666 | if (!zfsvfs->z_issnap) |
1667 | zfsctl_create(zfsvfs); | |
2cbb06b5 | 1668 | |
0037b49e | 1669 | zfsvfs->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); |
3558fd73 BB |
1670 | out: |
1671 | if (error) { | |
0037b49e | 1672 | dmu_objset_disown(zfsvfs->z_os, zfsvfs); |
f298b24d | 1673 | zfsvfs_free(zfsvfs); |
be88e733 CC |
1674 | /* |
1675 | * make sure we don't have dangling sb->s_fs_info which | |
1676 | * zfs_preumount will use. | |
1677 | */ | |
1678 | sb->s_fs_info = NULL; | |
3558fd73 | 1679 | } |
34dc7c2f | 1680 | |
3558fd73 BB |
1681 | return (error); |
1682 | } | |
3558fd73 | 1683 | |
ebe7e575 BB |
1684 | /* |
1685 | * Called when an unmount is requested and certain sanity checks have | |
1686 | * already passed. At this point no dentries or inodes have been reclaimed | |
1687 | * from their respective caches. We drop the extra reference on the .zfs | |
1688 | * control directory to allow everything to be reclaimed. All snapshots | |
1689 | * must already have been unmounted to reach this point. | |
1690 | */ | |
1691 | void | |
1692 | zfs_preumount(struct super_block *sb) | |
1693 | { | |
0037b49e | 1694 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
ebe7e575 | 1695 | |
0037b49e BB |
1696 | /* zfsvfs is NULL when zfs_domount fails during mount */ |
1697 | if (zfsvfs) { | |
278bee93 | 1698 | zfsctl_destroy(sb->s_fs_info); |
be88e733 CC |
1699 | /* |
1700 | * Wait for iput_async before entering evict_inodes in | |
1701 | * generic_shutdown_super. The reason we must finish before | |
1702 | * evict_inodes is when lazytime is on, or when zfs_purgedir | |
1703 | * calls zfs_zget, iput would bump i_count from 0 to 1. This | |
1704 | * would race with the i_count check in evict_inodes. This means | |
1705 | * it could destroy the inode while we are still using it. | |
1706 | * | |
1707 | * We wait for two passes. xattr directories in the first pass | |
1708 | * may add xattr entries in zfs_purgedir, so in the second pass | |
1709 | * we wait for them. We don't use taskq_wait here because it is | |
1710 | * a pool wide taskq. Other mounted filesystems can constantly | |
1711 | * do iput_async and there's no guarantee when taskq will be | |
1712 | * empty. | |
1713 | */ | |
1714 | taskq_wait_outstanding(dsl_pool_iput_taskq( | |
0037b49e | 1715 | dmu_objset_pool(zfsvfs->z_os)), 0); |
be88e733 | 1716 | taskq_wait_outstanding(dsl_pool_iput_taskq( |
0037b49e | 1717 | dmu_objset_pool(zfsvfs->z_os)), 0); |
be88e733 | 1718 | } |
ebe7e575 | 1719 | } |
ebe7e575 BB |
1720 | |
1721 | /* | |
1722 | * Called once all other unmount released tear down has occurred. | |
1723 | * It is our responsibility to release any remaining infrastructure. | |
1724 | */ | |
3558fd73 BB |
1725 | /*ARGSUSED*/ |
1726 | int | |
1727 | zfs_umount(struct super_block *sb) | |
1728 | { | |
0037b49e | 1729 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
3558fd73 BB |
1730 | objset_t *os; |
1731 | ||
0037b49e BB |
1732 | arc_remove_prune_callback(zfsvfs->z_arc_prune); |
1733 | VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0); | |
1734 | os = zfsvfs->z_os; | |
5547c2f1 | 1735 | bdi_destroy(sb->s_bdi); |
76659dc1 | 1736 | |
34dc7c2f BB |
1737 | /* |
1738 | * z_os will be NULL if there was an error in | |
0037b49e | 1739 | * attempting to reopen zfsvfs. |
34dc7c2f BB |
1740 | */ |
1741 | if (os != NULL) { | |
1742 | /* | |
1743 | * Unset the objset user_ptr. | |
1744 | */ | |
428870ff | 1745 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1746 | dmu_objset_set_user(os, NULL); |
428870ff | 1747 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1748 | |
1749 | /* | |
b128c09f | 1750 | * Finally release the objset |
34dc7c2f | 1751 | */ |
0037b49e | 1752 | dmu_objset_disown(os, zfsvfs); |
34dc7c2f BB |
1753 | } |
1754 | ||
f298b24d | 1755 | zfsvfs_free(zfsvfs); |
34dc7c2f BB |
1756 | return (0); |
1757 | } | |
1758 | ||
0de19dad | 1759 | int |
1c2555ef | 1760 | zfs_remount(struct super_block *sb, int *flags, zfs_mnt_t *zm) |
0de19dad | 1761 | { |
0037b49e | 1762 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
1c2555ef | 1763 | vfs_t *vfsp; |
0282c413 BB |
1764 | int error; |
1765 | ||
1c2555ef BB |
1766 | error = zfsvfs_parse_options(zm->mnt_data, &vfsp); |
1767 | if (error) | |
1768 | return (error); | |
1769 | ||
0037b49e | 1770 | zfs_unregister_callbacks(zfsvfs); |
1c2555ef BB |
1771 | zfsvfs_vfs_free(zfsvfs->z_vfs); |
1772 | ||
1773 | vfsp->vfs_data = zfsvfs; | |
1774 | zfsvfs->z_vfs = vfsp; | |
1775 | (void) zfs_register_callbacks(vfsp); | |
0282c413 BB |
1776 | |
1777 | return (error); | |
0de19dad | 1778 | } |
0de19dad | 1779 | |
e5c39b95 | 1780 | int |
2cf7f52b | 1781 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1782 | { |
0037b49e | 1783 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
34dc7c2f BB |
1784 | znode_t *zp; |
1785 | uint64_t object = 0; | |
1786 | uint64_t fid_gen = 0; | |
1787 | uint64_t gen_mask; | |
1788 | uint64_t zp_gen; | |
3558fd73 | 1789 | int i, err; |
34dc7c2f | 1790 | |
3558fd73 | 1791 | *ipp = NULL; |
34dc7c2f | 1792 | |
9b77d1c9 CC |
1793 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { |
1794 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1795 | ||
1796 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1797 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
34dc7c2f | 1798 | |
9b77d1c9 CC |
1799 | for (i = 0; i < sizeof (zfid->zf_gen); i++) |
1800 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1801 | } else { | |
1802 | return (SET_ERROR(EINVAL)); | |
1803 | } | |
1804 | ||
1805 | /* LONG_FID_LEN means snapdirs */ | |
34dc7c2f BB |
1806 | if (fidp->fid_len == LONG_FID_LEN) { |
1807 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1808 | uint64_t objsetid = 0; | |
1809 | uint64_t setgen = 0; | |
1810 | ||
1811 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1812 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1813 | ||
1814 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1815 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1816 | ||
9b77d1c9 CC |
1817 | if (objsetid != ZFSCTL_INO_SNAPDIRS - object) { |
1818 | dprintf("snapdir fid: objsetid (%llu) != " | |
1819 | "ZFSCTL_INO_SNAPDIRS (%llu) - object (%llu)\n", | |
1820 | objsetid, ZFSCTL_INO_SNAPDIRS, object); | |
34dc7c2f | 1821 | |
2e528b49 | 1822 | return (SET_ERROR(EINVAL)); |
9b77d1c9 | 1823 | } |
ebe7e575 | 1824 | |
9b77d1c9 CC |
1825 | if (fid_gen > 1 || setgen != 0) { |
1826 | dprintf("snapdir fid: fid_gen (%llu) and setgen " | |
1827 | "(%llu)\n", fid_gen, setgen); | |
1828 | return (SET_ERROR(EINVAL)); | |
1829 | } | |
34dc7c2f | 1830 | |
9b77d1c9 | 1831 | return (zfsctl_snapdir_vget(sb, objsetid, fid_gen, ipp)); |
34dc7c2f BB |
1832 | } |
1833 | ||
0037b49e | 1834 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
1835 | /* A zero fid_gen means we are in the .zfs control directories */ |
1836 | if (fid_gen == 0 && | |
1837 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
0037b49e | 1838 | *ipp = zfsvfs->z_ctldir; |
3558fd73 | 1839 | ASSERT(*ipp != NULL); |
34dc7c2f | 1840 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1841 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1842 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1843 | } else { |
3558fd73 | 1844 | igrab(*ipp); |
34dc7c2f | 1845 | } |
0037b49e | 1846 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1847 | return (0); |
1848 | } | |
1849 | ||
1850 | gen_mask = -1ULL >> (64 - 8 * i); | |
1851 | ||
29e57d15 | 1852 | dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); |
0037b49e BB |
1853 | if ((err = zfs_zget(zfsvfs, object, &zp))) { |
1854 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
1855 | return (err); |
1856 | } | |
7938c2ac CC |
1857 | |
1858 | /* Don't export xattr stuff */ | |
1859 | if (zp->z_pflags & ZFS_XATTR) { | |
1860 | iput(ZTOI(zp)); | |
0037b49e | 1861 | ZFS_EXIT(zfsvfs); |
7938c2ac CC |
1862 | return (SET_ERROR(ENOENT)); |
1863 | } | |
1864 | ||
0037b49e | 1865 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen, |
428870ff BB |
1866 | sizeof (uint64_t)); |
1867 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1868 | if (zp_gen == 0) |
1869 | zp_gen = 1; | |
0037b49e | 1870 | if ((fid_gen == 0) && (zfsvfs->z_root == object)) |
0500e835 | 1871 | fid_gen = zp_gen; |
34dc7c2f | 1872 | if (zp->z_unlinked || zp_gen != fid_gen) { |
29e57d15 NB |
1873 | dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, |
1874 | fid_gen); | |
3558fd73 | 1875 | iput(ZTOI(zp)); |
0037b49e | 1876 | ZFS_EXIT(zfsvfs); |
6c253064 | 1877 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1878 | } |
1879 | ||
3558fd73 BB |
1880 | *ipp = ZTOI(zp); |
1881 | if (*ipp) | |
1882 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1883 | |
0037b49e | 1884 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1885 | return (0); |
1886 | } | |
1887 | ||
1888 | /* | |
0037b49e | 1889 | * Block out VFS ops and close zfsvfs_t |
34dc7c2f BB |
1890 | * |
1891 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
831baf06 KW |
1892 | * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying |
1893 | * dataset and objset intact so that they can be atomically handed off during | |
1894 | * a subsequent rollback or recv operation and the resume thereafter. | |
34dc7c2f BB |
1895 | */ |
1896 | int | |
0037b49e | 1897 | zfs_suspend_fs(zfsvfs_t *zfsvfs) |
34dc7c2f BB |
1898 | { |
1899 | int error; | |
1900 | ||
0037b49e | 1901 | if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0) |
34dc7c2f | 1902 | return (error); |
7b3e34ba | 1903 | |
34dc7c2f BB |
1904 | return (0); |
1905 | } | |
1906 | ||
1907 | /* | |
8614ddf9 MA |
1908 | * Rebuild SA and release VOPs. Note that ownership of the underlying dataset |
1909 | * is an invariant across any of the operations that can be performed while the | |
1910 | * filesystem was suspended. Whether it succeeded or failed, the preconditions | |
1911 | * are the same: the relevant objset and associated dataset are owned by | |
1912 | * zfsvfs, held, and long held on entry. | |
34dc7c2f BB |
1913 | */ |
1914 | int | |
0037b49e | 1915 | zfs_resume_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) |
34dc7c2f | 1916 | { |
2008e920 | 1917 | int err, err2; |
831baf06 | 1918 | znode_t *zp; |
34dc7c2f | 1919 | |
0037b49e BB |
1920 | ASSERT(RRM_WRITE_HELD(&zfsvfs->z_teardown_lock)); |
1921 | ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); | |
34dc7c2f | 1922 | |
831baf06 | 1923 | /* |
ec923db2 GM |
1924 | * We already own this, so just update the objset_t, as the one we |
1925 | * had before may have been evicted. | |
831baf06 | 1926 | */ |
8614ddf9 | 1927 | objset_t *os; |
0037b49e | 1928 | VERIFY3P(ds->ds_owner, ==, zfsvfs); |
ec923db2 | 1929 | VERIFY(dsl_dataset_long_held(ds)); |
8614ddf9 | 1930 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
c1fabe79 | 1931 | |
8614ddf9 MA |
1932 | err = zfsvfs_init(zfsvfs, os); |
1933 | if (err != 0) | |
831baf06 | 1934 | goto bail; |
34dc7c2f | 1935 | |
f298b24d | 1936 | VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); |
c1fabe79 | 1937 | |
0037b49e BB |
1938 | zfs_set_fuid_feature(zfsvfs); |
1939 | zfsvfs->z_rollback_time = jiffies; | |
34dc7c2f | 1940 | |
831baf06 KW |
1941 | /* |
1942 | * Attempt to re-establish all the active inodes with their | |
1943 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
1944 | * and mark it stale. This prevents a collision if a new | |
1945 | * inode/object is created which must use the same inode | |
1946 | * number. The stale inode will be be released when the | |
1947 | * VFS prunes the dentry holding the remaining references | |
1948 | * on the stale inode. | |
1949 | */ | |
0037b49e BB |
1950 | mutex_enter(&zfsvfs->z_znodes_lock); |
1951 | for (zp = list_head(&zfsvfs->z_all_znodes); zp; | |
1952 | zp = list_next(&zfsvfs->z_all_znodes, zp)) { | |
2008e920 TC |
1953 | err2 = zfs_rezget(zp); |
1954 | if (err2) { | |
831baf06 KW |
1955 | remove_inode_hash(ZTOI(zp)); |
1956 | zp->z_is_stale = B_TRUE; | |
34dc7c2f | 1957 | } |
34dc7c2f | 1958 | } |
0037b49e | 1959 | mutex_exit(&zfsvfs->z_znodes_lock); |
34dc7c2f | 1960 | |
428870ff | 1961 | bail: |
7b3e34ba | 1962 | /* release the VFS ops */ |
0037b49e BB |
1963 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
1964 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1965 | |
1966 | if (err) { | |
1967 | /* | |
831baf06 KW |
1968 | * Since we couldn't setup the sa framework, try to force |
1969 | * unmount this file system. | |
34dc7c2f | 1970 | */ |
0037b49e BB |
1971 | if (zfsvfs->z_os) |
1972 | (void) zfs_umount(zfsvfs->z_sb); | |
34dc7c2f BB |
1973 | } |
1974 | return (err); | |
1975 | } | |
1976 | ||
34dc7c2f | 1977 | int |
0037b49e | 1978 | zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) |
34dc7c2f BB |
1979 | { |
1980 | int error; | |
0037b49e | 1981 | objset_t *os = zfsvfs->z_os; |
34dc7c2f | 1982 | dmu_tx_t *tx; |
34dc7c2f BB |
1983 | |
1984 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
2e528b49 | 1985 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1986 | |
0037b49e | 1987 | if (newvers < zfsvfs->z_version) |
2e528b49 | 1988 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1989 | |
428870ff | 1990 | if (zfs_spa_version_map(newvers) > |
0037b49e | 1991 | spa_version(dmu_objset_spa(zfsvfs->z_os))) |
2e528b49 | 1992 | return (SET_ERROR(ENOTSUP)); |
428870ff | 1993 | |
34dc7c2f | 1994 | tx = dmu_tx_create(os); |
9babb374 | 1995 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
0037b49e | 1996 | if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { |
428870ff BB |
1997 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1998 | ZFS_SA_ATTRS); | |
1999 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
2000 | } | |
34dc7c2f BB |
2001 | error = dmu_tx_assign(tx, TXG_WAIT); |
2002 | if (error) { | |
2003 | dmu_tx_abort(tx); | |
9babb374 BB |
2004 | return (error); |
2005 | } | |
428870ff | 2006 | |
9babb374 BB |
2007 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
2008 | 8, 1, &newvers, tx); | |
2009 | ||
2010 | if (error) { | |
2011 | dmu_tx_commit(tx); | |
2012 | return (error); | |
34dc7c2f | 2013 | } |
34dc7c2f | 2014 | |
0037b49e | 2015 | if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { |
428870ff BB |
2016 | uint64_t sa_obj; |
2017 | ||
0037b49e | 2018 | ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=, |
428870ff BB |
2019 | SPA_VERSION_SA); |
2020 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
2021 | DMU_OT_NONE, 0, tx); | |
2022 | ||
2023 | error = zap_add(os, MASTER_NODE_OBJ, | |
2024 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c99c9001 | 2025 | ASSERT0(error); |
428870ff BB |
2026 | |
2027 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
2028 | sa_register_update_callback(os, zfs_sa_upgrade); | |
2029 | } | |
2030 | ||
6f1ffb06 | 2031 | spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, |
0037b49e | 2032 | "from %llu to %llu", zfsvfs->z_version, newvers); |
9babb374 | 2033 | |
34dc7c2f BB |
2034 | dmu_tx_commit(tx); |
2035 | ||
0037b49e | 2036 | zfsvfs->z_version = newvers; |
9babb374 | 2037 | |
0037b49e | 2038 | zfs_set_fuid_feature(zfsvfs); |
9babb374 BB |
2039 | |
2040 | return (0); | |
34dc7c2f BB |
2041 | } |
2042 | ||
2043 | /* | |
2044 | * Read a property stored within the master node. | |
2045 | */ | |
2046 | int | |
2047 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
2048 | { | |
2049 | const char *pname; | |
2e528b49 | 2050 | int error = SET_ERROR(ENOENT); |
34dc7c2f BB |
2051 | |
2052 | /* | |
2053 | * Look up the file system's value for the property. For the | |
2054 | * version property, we look up a slightly different string. | |
2055 | */ | |
2056 | if (prop == ZFS_PROP_VERSION) | |
2057 | pname = ZPL_VERSION_STR; | |
2058 | else | |
2059 | pname = zfs_prop_to_name(prop); | |
2060 | ||
b128c09f BB |
2061 | if (os != NULL) |
2062 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
2063 | |
2064 | if (error == ENOENT) { | |
2065 | /* No value set, use the default value */ | |
2066 | switch (prop) { | |
2067 | case ZFS_PROP_VERSION: | |
2068 | *value = ZPL_VERSION; | |
2069 | break; | |
2070 | case ZFS_PROP_NORMALIZE: | |
2071 | case ZFS_PROP_UTF8ONLY: | |
2072 | *value = 0; | |
2073 | break; | |
2074 | case ZFS_PROP_CASE: | |
2075 | *value = ZFS_CASE_SENSITIVE; | |
2076 | break; | |
023699cd MM |
2077 | case ZFS_PROP_ACLTYPE: |
2078 | *value = ZFS_ACLTYPE_OFF; | |
2079 | break; | |
34dc7c2f BB |
2080 | default: |
2081 | return (error); | |
2082 | } | |
2083 | error = 0; | |
2084 | } | |
2085 | return (error); | |
2086 | } | |
3558fd73 | 2087 | |
a08abc1b GM |
2088 | /* |
2089 | * Return true if the coresponding vfs's unmounted flag is set. | |
2090 | * Otherwise return false. | |
2091 | * If this function returns true we know VFS unmount has been initiated. | |
2092 | */ | |
2093 | boolean_t | |
2094 | zfs_get_vfs_flag_unmounted(objset_t *os) | |
2095 | { | |
0037b49e | 2096 | zfsvfs_t *zfvp; |
a08abc1b GM |
2097 | boolean_t unmounted = B_FALSE; |
2098 | ||
2099 | ASSERT(dmu_objset_type(os) == DMU_OST_ZFS); | |
2100 | ||
2101 | mutex_enter(&os->os_user_ptr_lock); | |
2102 | zfvp = dmu_objset_get_user(os); | |
2103 | if (zfvp != NULL && zfvp->z_unmounted) | |
2104 | unmounted = B_TRUE; | |
2105 | mutex_exit(&os->os_user_ptr_lock); | |
2106 | ||
2107 | return (unmounted); | |
2108 | } | |
2109 | ||
3558fd73 BB |
2110 | void |
2111 | zfs_init(void) | |
2112 | { | |
ebe7e575 | 2113 | zfsctl_init(); |
3558fd73 BB |
2114 | zfs_znode_init(); |
2115 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
2116 | register_filesystem(&zpl_fs_type); | |
2117 | } | |
2118 | ||
2119 | void | |
2120 | zfs_fini(void) | |
2121 | { | |
8e71ab99 CC |
2122 | /* |
2123 | * we don't use outstanding because zpl_posix_acl_free might add more. | |
2124 | */ | |
57ddcda1 | 2125 | taskq_wait(system_delay_taskq); |
8e71ab99 | 2126 | taskq_wait(system_taskq); |
3558fd73 BB |
2127 | unregister_filesystem(&zpl_fs_type); |
2128 | zfs_znode_fini(); | |
ebe7e575 | 2129 | zfsctl_fini(); |
3558fd73 | 2130 | } |
f298b24d BB |
2131 | |
2132 | #if defined(_KERNEL) && defined(HAVE_SPL) | |
2133 | EXPORT_SYMBOL(zfs_suspend_fs); | |
2134 | EXPORT_SYMBOL(zfs_resume_fs); | |
2135 | EXPORT_SYMBOL(zfs_userspace_one); | |
2136 | EXPORT_SYMBOL(zfs_userspace_many); | |
2137 | EXPORT_SYMBOL(zfs_set_userquota); | |
2138 | EXPORT_SYMBOL(zfs_owner_overquota); | |
2139 | EXPORT_SYMBOL(zfs_fuid_overquota); | |
2140 | EXPORT_SYMBOL(zfs_fuid_overobjquota); | |
2141 | EXPORT_SYMBOL(zfs_set_version); | |
2142 | EXPORT_SYMBOL(zfsvfs_create); | |
2143 | EXPORT_SYMBOL(zfsvfs_free); | |
2144 | EXPORT_SYMBOL(zfs_is_readonly); | |
2145 | EXPORT_SYMBOL(zfs_domount); | |
2146 | EXPORT_SYMBOL(zfs_preumount); | |
2147 | EXPORT_SYMBOL(zfs_umount); | |
2148 | EXPORT_SYMBOL(zfs_remount); | |
2149 | EXPORT_SYMBOL(zfs_statvfs); | |
2150 | EXPORT_SYMBOL(zfs_vget); | |
2151 | EXPORT_SYMBOL(zfs_prune); | |
2152 | #endif |