]>
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. |
a448a255 | 23 | * Copyright (c) 2012, 2018 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> | |
34dc7c2f BB |
30 | #include <sys/sysmacros.h> |
31 | #include <sys/kmem.h> | |
32 | #include <sys/pathname.h> | |
33 | #include <sys/vnode.h> | |
34 | #include <sys/vfs.h> | |
34dc7c2f | 35 | #include <sys/mntent.h> |
34dc7c2f | 36 | #include <sys/cmn_err.h> |
34dc7c2f | 37 | #include <sys/zfs_znode.h> |
3558fd73 | 38 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
39 | #include <sys/zfs_dir.h> |
40 | #include <sys/zil.h> | |
41 | #include <sys/fs/zfs.h> | |
42 | #include <sys/dmu.h> | |
43 | #include <sys/dsl_prop.h> | |
44 | #include <sys/dsl_dataset.h> | |
45 | #include <sys/dsl_deleg.h> | |
46 | #include <sys/spa.h> | |
47 | #include <sys/zap.h> | |
428870ff | 48 | #include <sys/sa.h> |
a94addd9 | 49 | #include <sys/sa_impl.h> |
34dc7c2f BB |
50 | #include <sys/policy.h> |
51 | #include <sys/atomic.h> | |
34dc7c2f | 52 | #include <sys/zfs_ioctl.h> |
ebe7e575 | 53 | #include <sys/zfs_ctldir.h> |
34dc7c2f | 54 | #include <sys/zfs_fuid.h> |
34dc7c2f | 55 | #include <sys/sunddi.h> |
34dc7c2f BB |
56 | #include <sys/dmu_objset.h> |
57 | #include <sys/spa_boot.h> | |
30af21b0 | 58 | #include <sys/objlist.h> |
3558fd73 | 59 | #include <sys/zpl.h> |
05805494 | 60 | #include <linux/vfs_compat.h> |
428870ff | 61 | #include "zfs_comutil.h" |
34dc7c2f | 62 | |
1c2555ef BB |
63 | enum { |
64 | TOKEN_RO, | |
65 | TOKEN_RW, | |
66 | TOKEN_SETUID, | |
67 | TOKEN_NOSETUID, | |
68 | TOKEN_EXEC, | |
69 | TOKEN_NOEXEC, | |
70 | TOKEN_DEVICES, | |
71 | TOKEN_NODEVICES, | |
72 | TOKEN_DIRXATTR, | |
73 | TOKEN_SAXATTR, | |
74 | TOKEN_XATTR, | |
75 | TOKEN_NOXATTR, | |
76 | TOKEN_ATIME, | |
77 | TOKEN_NOATIME, | |
78 | TOKEN_RELATIME, | |
79 | TOKEN_NORELATIME, | |
80 | TOKEN_NBMAND, | |
81 | TOKEN_NONBMAND, | |
82 | TOKEN_MNTPOINT, | |
83 | TOKEN_LAST, | |
84 | }; | |
85 | ||
86 | static const match_table_t zpl_tokens = { | |
87 | { TOKEN_RO, MNTOPT_RO }, | |
88 | { TOKEN_RW, MNTOPT_RW }, | |
89 | { TOKEN_SETUID, MNTOPT_SETUID }, | |
90 | { TOKEN_NOSETUID, MNTOPT_NOSETUID }, | |
91 | { TOKEN_EXEC, MNTOPT_EXEC }, | |
92 | { TOKEN_NOEXEC, MNTOPT_NOEXEC }, | |
93 | { TOKEN_DEVICES, MNTOPT_DEVICES }, | |
94 | { TOKEN_NODEVICES, MNTOPT_NODEVICES }, | |
95 | { TOKEN_DIRXATTR, MNTOPT_DIRXATTR }, | |
96 | { TOKEN_SAXATTR, MNTOPT_SAXATTR }, | |
97 | { TOKEN_XATTR, MNTOPT_XATTR }, | |
98 | { TOKEN_NOXATTR, MNTOPT_NOXATTR }, | |
99 | { TOKEN_ATIME, MNTOPT_ATIME }, | |
100 | { TOKEN_NOATIME, MNTOPT_NOATIME }, | |
101 | { TOKEN_RELATIME, MNTOPT_RELATIME }, | |
102 | { TOKEN_NORELATIME, MNTOPT_NORELATIME }, | |
103 | { TOKEN_NBMAND, MNTOPT_NBMAND }, | |
104 | { TOKEN_NONBMAND, MNTOPT_NONBMAND }, | |
105 | { TOKEN_MNTPOINT, MNTOPT_MNTPOINT "=%s" }, | |
106 | { TOKEN_LAST, NULL }, | |
107 | }; | |
108 | ||
109 | static void | |
110 | zfsvfs_vfs_free(vfs_t *vfsp) | |
111 | { | |
112 | if (vfsp != NULL) { | |
113 | if (vfsp->vfs_mntpoint != NULL) | |
e4f5fa12 | 114 | kmem_strfree(vfsp->vfs_mntpoint); |
1c2555ef BB |
115 | |
116 | kmem_free(vfsp, sizeof (vfs_t)); | |
117 | } | |
118 | } | |
119 | ||
120 | static int | |
121 | zfsvfs_parse_option(char *option, int token, substring_t *args, vfs_t *vfsp) | |
122 | { | |
123 | switch (token) { | |
124 | case TOKEN_RO: | |
125 | vfsp->vfs_readonly = B_TRUE; | |
126 | vfsp->vfs_do_readonly = B_TRUE; | |
127 | break; | |
128 | case TOKEN_RW: | |
129 | vfsp->vfs_readonly = B_FALSE; | |
130 | vfsp->vfs_do_readonly = B_TRUE; | |
131 | break; | |
132 | case TOKEN_SETUID: | |
133 | vfsp->vfs_setuid = B_TRUE; | |
134 | vfsp->vfs_do_setuid = B_TRUE; | |
135 | break; | |
136 | case TOKEN_NOSETUID: | |
137 | vfsp->vfs_setuid = B_FALSE; | |
138 | vfsp->vfs_do_setuid = B_TRUE; | |
139 | break; | |
140 | case TOKEN_EXEC: | |
141 | vfsp->vfs_exec = B_TRUE; | |
142 | vfsp->vfs_do_exec = B_TRUE; | |
143 | break; | |
144 | case TOKEN_NOEXEC: | |
145 | vfsp->vfs_exec = B_FALSE; | |
146 | vfsp->vfs_do_exec = B_TRUE; | |
147 | break; | |
148 | case TOKEN_DEVICES: | |
149 | vfsp->vfs_devices = B_TRUE; | |
150 | vfsp->vfs_do_devices = B_TRUE; | |
151 | break; | |
152 | case TOKEN_NODEVICES: | |
153 | vfsp->vfs_devices = B_FALSE; | |
154 | vfsp->vfs_do_devices = B_TRUE; | |
155 | break; | |
156 | case TOKEN_DIRXATTR: | |
157 | vfsp->vfs_xattr = ZFS_XATTR_DIR; | |
158 | vfsp->vfs_do_xattr = B_TRUE; | |
159 | break; | |
160 | case TOKEN_SAXATTR: | |
161 | vfsp->vfs_xattr = ZFS_XATTR_SA; | |
162 | vfsp->vfs_do_xattr = B_TRUE; | |
163 | break; | |
164 | case TOKEN_XATTR: | |
165 | vfsp->vfs_xattr = ZFS_XATTR_DIR; | |
166 | vfsp->vfs_do_xattr = B_TRUE; | |
167 | break; | |
168 | case TOKEN_NOXATTR: | |
169 | vfsp->vfs_xattr = ZFS_XATTR_OFF; | |
170 | vfsp->vfs_do_xattr = B_TRUE; | |
171 | break; | |
172 | case TOKEN_ATIME: | |
173 | vfsp->vfs_atime = B_TRUE; | |
174 | vfsp->vfs_do_atime = B_TRUE; | |
175 | break; | |
176 | case TOKEN_NOATIME: | |
177 | vfsp->vfs_atime = B_FALSE; | |
178 | vfsp->vfs_do_atime = B_TRUE; | |
179 | break; | |
180 | case TOKEN_RELATIME: | |
181 | vfsp->vfs_relatime = B_TRUE; | |
182 | vfsp->vfs_do_relatime = B_TRUE; | |
183 | break; | |
184 | case TOKEN_NORELATIME: | |
185 | vfsp->vfs_relatime = B_FALSE; | |
186 | vfsp->vfs_do_relatime = B_TRUE; | |
187 | break; | |
188 | case TOKEN_NBMAND: | |
189 | vfsp->vfs_nbmand = B_TRUE; | |
190 | vfsp->vfs_do_nbmand = B_TRUE; | |
191 | break; | |
192 | case TOKEN_NONBMAND: | |
193 | vfsp->vfs_nbmand = B_FALSE; | |
194 | vfsp->vfs_do_nbmand = B_TRUE; | |
195 | break; | |
196 | case TOKEN_MNTPOINT: | |
197 | vfsp->vfs_mntpoint = match_strdup(&args[0]); | |
198 | if (vfsp->vfs_mntpoint == NULL) | |
199 | return (SET_ERROR(ENOMEM)); | |
200 | ||
201 | break; | |
202 | default: | |
203 | break; | |
204 | } | |
205 | ||
206 | return (0); | |
207 | } | |
208 | ||
209 | /* | |
210 | * Parse the raw mntopts and return a vfs_t describing the options. | |
211 | */ | |
212 | static int | |
213 | zfsvfs_parse_options(char *mntopts, vfs_t **vfsp) | |
214 | { | |
215 | vfs_t *tmp_vfsp; | |
216 | int error; | |
217 | ||
218 | tmp_vfsp = kmem_zalloc(sizeof (vfs_t), KM_SLEEP); | |
219 | ||
220 | if (mntopts != NULL) { | |
221 | substring_t args[MAX_OPT_ARGS]; | |
222 | char *tmp_mntopts, *p, *t; | |
223 | int token; | |
224 | ||
e4f5fa12 | 225 | tmp_mntopts = t = kmem_strdup(mntopts); |
1c2555ef BB |
226 | if (tmp_mntopts == NULL) |
227 | return (SET_ERROR(ENOMEM)); | |
228 | ||
229 | while ((p = strsep(&t, ",")) != NULL) { | |
230 | if (!*p) | |
231 | continue; | |
232 | ||
233 | args[0].to = args[0].from = NULL; | |
234 | token = match_token(p, zpl_tokens, args); | |
235 | error = zfsvfs_parse_option(p, token, args, tmp_vfsp); | |
236 | if (error) { | |
e4f5fa12 | 237 | kmem_strfree(tmp_mntopts); |
1c2555ef BB |
238 | zfsvfs_vfs_free(tmp_vfsp); |
239 | return (error); | |
240 | } | |
241 | } | |
242 | ||
e4f5fa12 | 243 | kmem_strfree(tmp_mntopts); |
1c2555ef BB |
244 | } |
245 | ||
246 | *vfsp = tmp_vfsp; | |
247 | ||
248 | return (0); | |
249 | } | |
250 | ||
251 | boolean_t | |
252 | zfs_is_readonly(zfsvfs_t *zfsvfs) | |
253 | { | |
05805494 | 254 | return (!!(zfsvfs->z_sb->s_flags & SB_RDONLY)); |
1c2555ef BB |
255 | } |
256 | ||
34dc7c2f BB |
257 | /*ARGSUSED*/ |
258 | int | |
03f9ba9d | 259 | zfs_sync(struct super_block *sb, int wait, cred_t *cr) |
34dc7c2f | 260 | { |
0037b49e | 261 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
03f9ba9d | 262 | |
03f9ba9d BB |
263 | /* |
264 | * Semantically, the only requirement is that the sync be initiated. | |
265 | * The DMU syncs out txgs frequently, so there's nothing to do. | |
266 | */ | |
267 | if (!wait) | |
268 | return (0); | |
269 | ||
0037b49e | 270 | if (zfsvfs != NULL) { |
34dc7c2f BB |
271 | /* |
272 | * Sync a specific filesystem. | |
273 | */ | |
9babb374 | 274 | dsl_pool_t *dp; |
34dc7c2f | 275 | |
0037b49e BB |
276 | ZFS_ENTER(zfsvfs); |
277 | dp = dmu_objset_pool(zfsvfs->z_os); | |
9babb374 BB |
278 | |
279 | /* | |
280 | * If the system is shutting down, then skip any | |
281 | * filesystems which may exist on a suspended pool. | |
282 | */ | |
03f9ba9d | 283 | if (spa_suspended(dp->dp_spa)) { |
0037b49e | 284 | ZFS_EXIT(zfsvfs); |
9babb374 BB |
285 | return (0); |
286 | } | |
287 | ||
0037b49e BB |
288 | if (zfsvfs->z_log != NULL) |
289 | zil_commit(zfsvfs->z_log, 0); | |
428870ff | 290 | |
0037b49e | 291 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
292 | } else { |
293 | /* | |
294 | * Sync all ZFS filesystems. This is what happens when you | |
295 | * run sync(1M). Unlike other filesystems, ZFS honors the | |
296 | * request by waiting for all pools to commit all dirty data. | |
297 | */ | |
298 | spa_sync_allpools(); | |
299 | } | |
300 | ||
301 | return (0); | |
302 | } | |
303 | ||
34dc7c2f BB |
304 | static void |
305 | atime_changed_cb(void *arg, uint64_t newval) | |
306 | { | |
9c53e516 TK |
307 | zfsvfs_t *zfsvfs = arg; |
308 | struct super_block *sb = zfsvfs->z_sb; | |
309 | ||
310 | if (sb == NULL) | |
311 | return; | |
312 | /* | |
313 | * Update SB_NOATIME bit in VFS super block. Since atime update is | |
314 | * determined by atime_needs_update(), atime_needs_update() needs to | |
315 | * return false if atime is turned off, and not unconditionally return | |
316 | * false if atime is turned on. | |
317 | */ | |
318 | if (newval) | |
319 | sb->s_flags &= ~SB_NOATIME; | |
320 | else | |
321 | sb->s_flags |= SB_NOATIME; | |
34dc7c2f BB |
322 | } |
323 | ||
6d111134 TC |
324 | static void |
325 | relatime_changed_cb(void *arg, uint64_t newval) | |
326 | { | |
0037b49e | 327 | ((zfsvfs_t *)arg)->z_relatime = newval; |
6d111134 TC |
328 | } |
329 | ||
34dc7c2f BB |
330 | static void |
331 | xattr_changed_cb(void *arg, uint64_t newval) | |
332 | { | |
0037b49e | 333 | zfsvfs_t *zfsvfs = arg; |
34dc7c2f | 334 | |
82a37189 | 335 | if (newval == ZFS_XATTR_OFF) { |
0037b49e | 336 | zfsvfs->z_flags &= ~ZSB_XATTR; |
82a37189 | 337 | } else { |
0037b49e | 338 | zfsvfs->z_flags |= ZSB_XATTR; |
82a37189 BB |
339 | |
340 | if (newval == ZFS_XATTR_SA) | |
0037b49e | 341 | zfsvfs->z_xattr_sa = B_TRUE; |
82a37189 | 342 | else |
0037b49e | 343 | zfsvfs->z_xattr_sa = B_FALSE; |
82a37189 | 344 | } |
34dc7c2f BB |
345 | } |
346 | ||
023699cd MM |
347 | static void |
348 | acltype_changed_cb(void *arg, uint64_t newval) | |
349 | { | |
0037b49e | 350 | zfsvfs_t *zfsvfs = arg; |
023699cd MM |
351 | |
352 | switch (newval) { | |
353 | case ZFS_ACLTYPE_OFF: | |
0037b49e | 354 | zfsvfs->z_acl_type = ZFS_ACLTYPE_OFF; |
05805494 | 355 | zfsvfs->z_sb->s_flags &= ~SB_POSIXACL; |
023699cd MM |
356 | break; |
357 | case ZFS_ACLTYPE_POSIXACL: | |
b695c34e | 358 | #ifdef CONFIG_FS_POSIX_ACL |
0037b49e | 359 | zfsvfs->z_acl_type = ZFS_ACLTYPE_POSIXACL; |
05805494 | 360 | zfsvfs->z_sb->s_flags |= SB_POSIXACL; |
b695c34e | 361 | #else |
0037b49e | 362 | zfsvfs->z_acl_type = ZFS_ACLTYPE_OFF; |
05805494 | 363 | zfsvfs->z_sb->s_flags &= ~SB_POSIXACL; |
b695c34e | 364 | #endif /* CONFIG_FS_POSIX_ACL */ |
023699cd MM |
365 | break; |
366 | default: | |
367 | break; | |
368 | } | |
369 | } | |
370 | ||
34dc7c2f BB |
371 | static void |
372 | blksz_changed_cb(void *arg, uint64_t newval) | |
373 | { | |
0037b49e BB |
374 | zfsvfs_t *zfsvfs = arg; |
375 | ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zfsvfs->z_os))); | |
f1512ee6 MA |
376 | ASSERT3U(newval, >=, SPA_MINBLOCKSIZE); |
377 | ASSERT(ISP2(newval)); | |
34dc7c2f | 378 | |
0037b49e | 379 | zfsvfs->z_max_blksz = newval; |
34dc7c2f BB |
380 | } |
381 | ||
382 | static void | |
383 | readonly_changed_cb(void *arg, uint64_t newval) | |
384 | { | |
0037b49e BB |
385 | zfsvfs_t *zfsvfs = arg; |
386 | struct super_block *sb = zfsvfs->z_sb; | |
34dc7c2f | 387 | |
2cf7f52b BB |
388 | if (sb == NULL) |
389 | return; | |
390 | ||
391 | if (newval) | |
05805494 | 392 | sb->s_flags |= SB_RDONLY; |
2cf7f52b | 393 | else |
05805494 | 394 | sb->s_flags &= ~SB_RDONLY; |
34dc7c2f BB |
395 | } |
396 | ||
397 | static void | |
398 | devices_changed_cb(void *arg, uint64_t newval) | |
399 | { | |
34dc7c2f BB |
400 | } |
401 | ||
402 | static void | |
403 | setuid_changed_cb(void *arg, uint64_t newval) | |
404 | { | |
34dc7c2f BB |
405 | } |
406 | ||
407 | static void | |
408 | exec_changed_cb(void *arg, uint64_t newval) | |
409 | { | |
34dc7c2f BB |
410 | } |
411 | ||
34dc7c2f BB |
412 | static void |
413 | nbmand_changed_cb(void *arg, uint64_t newval) | |
414 | { | |
0037b49e BB |
415 | zfsvfs_t *zfsvfs = arg; |
416 | struct super_block *sb = zfsvfs->z_sb; | |
3558fd73 | 417 | |
2cf7f52b BB |
418 | if (sb == NULL) |
419 | return; | |
420 | ||
421 | if (newval == TRUE) | |
05805494 | 422 | sb->s_flags |= SB_MANDLOCK; |
2cf7f52b | 423 | else |
05805494 | 424 | sb->s_flags &= ~SB_MANDLOCK; |
34dc7c2f BB |
425 | } |
426 | ||
427 | static void | |
428 | snapdir_changed_cb(void *arg, uint64_t newval) | |
429 | { | |
0037b49e | 430 | ((zfsvfs_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
431 | } |
432 | ||
433 | static void | |
434 | vscan_changed_cb(void *arg, uint64_t newval) | |
435 | { | |
0037b49e | 436 | ((zfsvfs_t *)arg)->z_vscan = newval; |
34dc7c2f BB |
437 | } |
438 | ||
34dc7c2f BB |
439 | static void |
440 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
441 | { | |
0037b49e | 442 | ((zfsvfs_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
443 | } |
444 | ||
1c2555ef BB |
445 | static int |
446 | zfs_register_callbacks(vfs_t *vfsp) | |
34dc7c2f BB |
447 | { |
448 | struct dsl_dataset *ds = NULL; | |
1c2555ef BB |
449 | objset_t *os = NULL; |
450 | zfsvfs_t *zfsvfs = NULL; | |
34dc7c2f BB |
451 | int error = 0; |
452 | ||
1c2555ef BB |
453 | ASSERT(vfsp); |
454 | zfsvfs = vfsp->vfs_data; | |
0037b49e | 455 | ASSERT(zfsvfs); |
1c2555ef | 456 | os = zfsvfs->z_os; |
0282c413 BB |
457 | |
458 | /* | |
459 | * The act of registering our callbacks will destroy any mount | |
460 | * options we may have. In order to enable temporary overrides | |
461 | * of mount options, we stash away the current values and | |
462 | * restore them after we register the callbacks. | |
463 | */ | |
0037b49e | 464 | if (zfs_is_readonly(zfsvfs) || !spa_writeable(dmu_objset_spa(os))) { |
1c2555ef BB |
465 | vfsp->vfs_do_readonly = B_TRUE; |
466 | vfsp->vfs_readonly = B_TRUE; | |
0282c413 | 467 | } |
34dc7c2f BB |
468 | |
469 | /* | |
470 | * Register property callbacks. | |
471 | * | |
472 | * It would probably be fine to just check for i/o error from | |
473 | * the first prop_register(), but I guess I like to go | |
474 | * overboard... | |
475 | */ | |
476 | ds = dmu_objset_ds(os); | |
13fe0198 | 477 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
3558fd73 | 478 | error = dsl_prop_register(ds, |
0037b49e | 479 | zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zfsvfs); |
98fad862 | 480 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 481 | zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zfsvfs); |
34dc7c2f | 482 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 483 | zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zfsvfs); |
34dc7c2f | 484 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 485 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zfsvfs); |
34dc7c2f | 486 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 487 | zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zfsvfs); |
34dc7c2f | 488 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 489 | zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zfsvfs); |
34dc7c2f | 490 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 491 | zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zfsvfs); |
34dc7c2f | 492 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 493 | zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zfsvfs); |
34dc7c2f | 494 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 495 | zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zfsvfs); |
34dc7c2f | 496 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 497 | zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zfsvfs); |
023699cd | 498 | error = error ? error : dsl_prop_register(ds, |
0037b49e BB |
499 | zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, |
500 | zfsvfs); | |
34dc7c2f | 501 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 502 | zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zfsvfs); |
2cf7f52b | 503 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 504 | zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zfsvfs); |
13fe0198 | 505 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); |
34dc7c2f BB |
506 | if (error) |
507 | goto unregister; | |
508 | ||
0282c413 BB |
509 | /* |
510 | * Invoke our callbacks to restore temporary mount options. | |
511 | */ | |
1c2555ef BB |
512 | if (vfsp->vfs_do_readonly) |
513 | readonly_changed_cb(zfsvfs, vfsp->vfs_readonly); | |
514 | if (vfsp->vfs_do_setuid) | |
515 | setuid_changed_cb(zfsvfs, vfsp->vfs_setuid); | |
516 | if (vfsp->vfs_do_exec) | |
517 | exec_changed_cb(zfsvfs, vfsp->vfs_exec); | |
518 | if (vfsp->vfs_do_devices) | |
519 | devices_changed_cb(zfsvfs, vfsp->vfs_devices); | |
520 | if (vfsp->vfs_do_xattr) | |
521 | xattr_changed_cb(zfsvfs, vfsp->vfs_xattr); | |
522 | if (vfsp->vfs_do_atime) | |
523 | atime_changed_cb(zfsvfs, vfsp->vfs_atime); | |
524 | if (vfsp->vfs_do_relatime) | |
525 | relatime_changed_cb(zfsvfs, vfsp->vfs_relatime); | |
526 | if (vfsp->vfs_do_nbmand) | |
527 | nbmand_changed_cb(zfsvfs, vfsp->vfs_nbmand); | |
c9ada6d5 | 528 | |
34dc7c2f BB |
529 | return (0); |
530 | ||
531 | unregister: | |
0037b49e | 532 | dsl_prop_unregister_all(ds, zfsvfs); |
3558fd73 | 533 | return (error); |
34dc7c2f BB |
534 | } |
535 | ||
428870ff BB |
536 | static int |
537 | zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, | |
9c5167d1 | 538 | uint64_t *userp, uint64_t *groupp, uint64_t *projectp) |
9babb374 | 539 | { |
9c5167d1 NF |
540 | sa_hdr_phys_t sa; |
541 | sa_hdr_phys_t *sap = data; | |
542 | uint64_t flags; | |
543 | int hdrsize; | |
544 | boolean_t swap = B_FALSE; | |
545 | ||
428870ff BB |
546 | /* |
547 | * Is it a valid type of object to track? | |
548 | */ | |
549 | if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) | |
2e528b49 | 550 | return (SET_ERROR(ENOENT)); |
9babb374 | 551 | |
428870ff BB |
552 | /* |
553 | * If we have a NULL data pointer | |
554 | * then assume the id's aren't changing and | |
555 | * return EEXIST to the dmu to let it know to | |
556 | * use the same ids | |
557 | */ | |
558 | if (data == NULL) | |
2e528b49 | 559 | return (SET_ERROR(EEXIST)); |
9babb374 | 560 | |
428870ff | 561 | if (bonustype == DMU_OT_ZNODE) { |
a94addd9 | 562 | znode_phys_t *znp = data; |
428870ff BB |
563 | *userp = znp->zp_uid; |
564 | *groupp = znp->zp_gid; | |
9c5167d1 NF |
565 | *projectp = ZFS_DEFAULT_PROJID; |
566 | return (0); | |
567 | } | |
568 | ||
569 | if (sap->sa_magic == 0) { | |
570 | /* | |
571 | * This should only happen for newly created files | |
572 | * that haven't had the znode data filled in yet. | |
573 | */ | |
574 | *userp = 0; | |
575 | *groupp = 0; | |
576 | *projectp = ZFS_DEFAULT_PROJID; | |
577 | return (0); | |
578 | } | |
579 | ||
580 | sa = *sap; | |
581 | if (sa.sa_magic == BSWAP_32(SA_MAGIC)) { | |
582 | sa.sa_magic = SA_MAGIC; | |
583 | sa.sa_layout_info = BSWAP_16(sa.sa_layout_info); | |
584 | swap = B_TRUE; | |
9babb374 | 585 | } else { |
9c5167d1 NF |
586 | VERIFY3U(sa.sa_magic, ==, SA_MAGIC); |
587 | } | |
a94addd9 | 588 | |
9c5167d1 NF |
589 | hdrsize = sa_hdrsize(&sa); |
590 | VERIFY3U(hdrsize, >=, sizeof (sa_hdr_phys_t)); | |
591 | ||
592 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + SA_UID_OFFSET)); | |
593 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + SA_GID_OFFSET)); | |
594 | flags = *((uint64_t *)((uintptr_t)data + hdrsize + SA_FLAGS_OFFSET)); | |
595 | if (swap) | |
596 | flags = BSWAP_64(flags); | |
597 | ||
598 | if (flags & ZFS_PROJID) | |
599 | *projectp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
600 | SA_PROJID_OFFSET)); | |
601 | else | |
602 | *projectp = ZFS_DEFAULT_PROJID; | |
603 | ||
604 | if (swap) { | |
605 | *userp = BSWAP_64(*userp); | |
606 | *groupp = BSWAP_64(*groupp); | |
607 | *projectp = BSWAP_64(*projectp); | |
9babb374 | 608 | } |
13fe0198 | 609 | return (0); |
9babb374 BB |
610 | } |
611 | ||
612 | static void | |
0037b49e | 613 | fuidstr_to_sid(zfsvfs_t *zfsvfs, const char *fuidstr, |
9babb374 BB |
614 | char *domainbuf, int buflen, uid_t *ridp) |
615 | { | |
9babb374 BB |
616 | uint64_t fuid; |
617 | const char *domain; | |
618 | ||
e19572e4 | 619 | fuid = zfs_strtonum(fuidstr, NULL); |
9babb374 | 620 | |
0037b49e | 621 | domain = zfs_fuid_find_by_idx(zfsvfs, FUID_INDEX(fuid)); |
9babb374 BB |
622 | if (domain) |
623 | (void) strlcpy(domainbuf, domain, buflen); | |
624 | else | |
625 | domainbuf[0] = '\0'; | |
626 | *ridp = FUID_RID(fuid); | |
627 | } | |
628 | ||
629 | static uint64_t | |
0037b49e | 630 | zfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type) |
9babb374 BB |
631 | { |
632 | switch (type) { | |
633 | case ZFS_PROP_USERUSED: | |
1de321e6 | 634 | case ZFS_PROP_USEROBJUSED: |
9babb374 BB |
635 | return (DMU_USERUSED_OBJECT); |
636 | case ZFS_PROP_GROUPUSED: | |
1de321e6 | 637 | case ZFS_PROP_GROUPOBJUSED: |
9babb374 | 638 | return (DMU_GROUPUSED_OBJECT); |
9c5167d1 NF |
639 | case ZFS_PROP_PROJECTUSED: |
640 | case ZFS_PROP_PROJECTOBJUSED: | |
641 | return (DMU_PROJECTUSED_OBJECT); | |
9babb374 | 642 | case ZFS_PROP_USERQUOTA: |
0037b49e | 643 | return (zfsvfs->z_userquota_obj); |
9babb374 | 644 | case ZFS_PROP_GROUPQUOTA: |
0037b49e | 645 | return (zfsvfs->z_groupquota_obj); |
1de321e6 | 646 | case ZFS_PROP_USEROBJQUOTA: |
0037b49e | 647 | return (zfsvfs->z_userobjquota_obj); |
1de321e6 | 648 | case ZFS_PROP_GROUPOBJQUOTA: |
0037b49e | 649 | return (zfsvfs->z_groupobjquota_obj); |
9c5167d1 NF |
650 | case ZFS_PROP_PROJECTQUOTA: |
651 | return (zfsvfs->z_projectquota_obj); | |
652 | case ZFS_PROP_PROJECTOBJQUOTA: | |
653 | return (zfsvfs->z_projectobjquota_obj); | |
149e873a | 654 | default: |
1de321e6 | 655 | return (ZFS_NO_OBJECT); |
9babb374 | 656 | } |
9babb374 BB |
657 | } |
658 | ||
659 | int | |
0037b49e | 660 | zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, |
9babb374 BB |
661 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
662 | { | |
663 | int error; | |
664 | zap_cursor_t zc; | |
665 | zap_attribute_t za; | |
666 | zfs_useracct_t *buf = vbuf; | |
667 | uint64_t obj; | |
1de321e6 | 668 | int offset = 0; |
9babb374 | 669 | |
0037b49e | 670 | if (!dmu_objset_userspace_present(zfsvfs->z_os)) |
2e528b49 | 671 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 672 | |
9c5167d1 NF |
673 | if ((type == ZFS_PROP_PROJECTQUOTA || type == ZFS_PROP_PROJECTUSED || |
674 | type == ZFS_PROP_PROJECTOBJQUOTA || | |
675 | type == ZFS_PROP_PROJECTOBJUSED) && | |
676 | !dmu_objset_projectquota_present(zfsvfs->z_os)) | |
677 | return (SET_ERROR(ENOTSUP)); | |
678 | ||
1de321e6 | 679 | if ((type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || |
9c5167d1 NF |
680 | type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA || |
681 | type == ZFS_PROP_PROJECTOBJUSED || | |
682 | type == ZFS_PROP_PROJECTOBJQUOTA) && | |
0037b49e | 683 | !dmu_objset_userobjspace_present(zfsvfs->z_os)) |
1de321e6 JX |
684 | return (SET_ERROR(ENOTSUP)); |
685 | ||
0037b49e | 686 | obj = zfs_userquota_prop_to_obj(zfsvfs, type); |
1de321e6 | 687 | if (obj == ZFS_NO_OBJECT) { |
9babb374 BB |
688 | *bufsizep = 0; |
689 | return (0); | |
690 | } | |
691 | ||
9c5167d1 NF |
692 | if (type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || |
693 | type == ZFS_PROP_PROJECTOBJUSED) | |
1de321e6 JX |
694 | offset = DMU_OBJACCT_PREFIX_LEN; |
695 | ||
0037b49e | 696 | for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep); |
9babb374 BB |
697 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
698 | zap_cursor_advance(&zc)) { | |
699 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
700 | *bufsizep) | |
701 | break; | |
702 | ||
1de321e6 JX |
703 | /* |
704 | * skip object quota (with zap name prefix DMU_OBJACCT_PREFIX) | |
705 | * when dealing with block quota and vice versa. | |
706 | */ | |
707 | if ((offset > 0) != (strncmp(za.za_name, DMU_OBJACCT_PREFIX, | |
708 | DMU_OBJACCT_PREFIX_LEN) == 0)) | |
709 | continue; | |
710 | ||
0037b49e | 711 | fuidstr_to_sid(zfsvfs, za.za_name + offset, |
9babb374 BB |
712 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
713 | ||
714 | buf->zu_space = za.za_first_integer; | |
715 | buf++; | |
716 | } | |
717 | if (error == ENOENT) | |
718 | error = 0; | |
719 | ||
720 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
721 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
722 | *cookiep = zap_cursor_serialize(&zc); | |
723 | zap_cursor_fini(&zc); | |
724 | return (error); | |
725 | } | |
726 | ||
727 | /* | |
728 | * buf must be big enough (eg, 32 bytes) | |
729 | */ | |
730 | static int | |
0037b49e | 731 | id_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid, |
9babb374 BB |
732 | char *buf, boolean_t addok) |
733 | { | |
734 | uint64_t fuid; | |
735 | int domainid = 0; | |
736 | ||
737 | if (domain && domain[0]) { | |
0037b49e | 738 | domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok); |
9babb374 | 739 | if (domainid == -1) |
2e528b49 | 740 | return (SET_ERROR(ENOENT)); |
9babb374 BB |
741 | } |
742 | fuid = FUID_ENCODE(domainid, rid); | |
743 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
744 | return (0); | |
745 | } | |
746 | ||
747 | int | |
0037b49e | 748 | zfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, |
9babb374 BB |
749 | const char *domain, uint64_t rid, uint64_t *valp) |
750 | { | |
1de321e6 JX |
751 | char buf[20 + DMU_OBJACCT_PREFIX_LEN]; |
752 | int offset = 0; | |
9babb374 BB |
753 | int err; |
754 | uint64_t obj; | |
755 | ||
756 | *valp = 0; | |
757 | ||
0037b49e | 758 | if (!dmu_objset_userspace_present(zfsvfs->z_os)) |
2e528b49 | 759 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 760 | |
1de321e6 | 761 | if ((type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || |
9c5167d1 NF |
762 | type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA || |
763 | type == ZFS_PROP_PROJECTOBJUSED || | |
764 | type == ZFS_PROP_PROJECTOBJQUOTA) && | |
0037b49e | 765 | !dmu_objset_userobjspace_present(zfsvfs->z_os)) |
1de321e6 JX |
766 | return (SET_ERROR(ENOTSUP)); |
767 | ||
9c5167d1 NF |
768 | if (type == ZFS_PROP_PROJECTQUOTA || type == ZFS_PROP_PROJECTUSED || |
769 | type == ZFS_PROP_PROJECTOBJQUOTA || | |
770 | type == ZFS_PROP_PROJECTOBJUSED) { | |
771 | if (!dmu_objset_projectquota_present(zfsvfs->z_os)) | |
772 | return (SET_ERROR(ENOTSUP)); | |
773 | if (!zpl_is_valid_projid(rid)) | |
774 | return (SET_ERROR(EINVAL)); | |
775 | } | |
776 | ||
0037b49e | 777 | obj = zfs_userquota_prop_to_obj(zfsvfs, type); |
1de321e6 | 778 | if (obj == ZFS_NO_OBJECT) |
9babb374 BB |
779 | return (0); |
780 | ||
9c5167d1 NF |
781 | if (type == ZFS_PROP_USEROBJUSED || type == ZFS_PROP_GROUPOBJUSED || |
782 | type == ZFS_PROP_PROJECTOBJUSED) { | |
ed15d544 | 783 | strlcpy(buf, DMU_OBJACCT_PREFIX, DMU_OBJACCT_PREFIX_LEN + 1); |
1de321e6 JX |
784 | offset = DMU_OBJACCT_PREFIX_LEN; |
785 | } | |
786 | ||
0037b49e | 787 | err = id_to_fuidstr(zfsvfs, domain, rid, buf + offset, B_FALSE); |
9babb374 BB |
788 | if (err) |
789 | return (err); | |
790 | ||
0037b49e | 791 | err = zap_lookup(zfsvfs->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
792 | if (err == ENOENT) |
793 | err = 0; | |
794 | return (err); | |
795 | } | |
796 | ||
797 | int | |
0037b49e | 798 | zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, |
9babb374 BB |
799 | const char *domain, uint64_t rid, uint64_t quota) |
800 | { | |
801 | char buf[32]; | |
802 | int err; | |
803 | dmu_tx_t *tx; | |
804 | uint64_t *objp; | |
805 | boolean_t fuid_dirtied; | |
806 | ||
0037b49e | 807 | if (zfsvfs->z_version < ZPL_VERSION_USERSPACE) |
2e528b49 | 808 | return (SET_ERROR(ENOTSUP)); |
9babb374 | 809 | |
1de321e6 JX |
810 | switch (type) { |
811 | case ZFS_PROP_USERQUOTA: | |
0037b49e | 812 | objp = &zfsvfs->z_userquota_obj; |
1de321e6 JX |
813 | break; |
814 | case ZFS_PROP_GROUPQUOTA: | |
0037b49e | 815 | objp = &zfsvfs->z_groupquota_obj; |
1de321e6 JX |
816 | break; |
817 | case ZFS_PROP_USEROBJQUOTA: | |
0037b49e | 818 | objp = &zfsvfs->z_userobjquota_obj; |
1de321e6 JX |
819 | break; |
820 | case ZFS_PROP_GROUPOBJQUOTA: | |
0037b49e | 821 | objp = &zfsvfs->z_groupobjquota_obj; |
1de321e6 | 822 | break; |
9c5167d1 NF |
823 | case ZFS_PROP_PROJECTQUOTA: |
824 | if (!dmu_objset_projectquota_enabled(zfsvfs->z_os)) | |
825 | return (SET_ERROR(ENOTSUP)); | |
826 | if (!zpl_is_valid_projid(rid)) | |
827 | return (SET_ERROR(EINVAL)); | |
828 | ||
829 | objp = &zfsvfs->z_projectquota_obj; | |
830 | break; | |
831 | case ZFS_PROP_PROJECTOBJQUOTA: | |
832 | if (!dmu_objset_projectquota_enabled(zfsvfs->z_os)) | |
833 | return (SET_ERROR(ENOTSUP)); | |
834 | if (!zpl_is_valid_projid(rid)) | |
835 | return (SET_ERROR(EINVAL)); | |
836 | ||
837 | objp = &zfsvfs->z_projectobjquota_obj; | |
838 | break; | |
1de321e6 JX |
839 | default: |
840 | return (SET_ERROR(EINVAL)); | |
841 | } | |
9babb374 | 842 | |
0037b49e | 843 | err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_TRUE); |
9babb374 BB |
844 | if (err) |
845 | return (err); | |
0037b49e | 846 | fuid_dirtied = zfsvfs->z_fuid_dirty; |
9babb374 | 847 | |
0037b49e | 848 | tx = dmu_tx_create(zfsvfs->z_os); |
9babb374 BB |
849 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
850 | if (*objp == 0) { | |
851 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
852 | zfs_userquota_prop_prefixes[type]); | |
853 | } | |
854 | if (fuid_dirtied) | |
0037b49e | 855 | zfs_fuid_txhold(zfsvfs, tx); |
9babb374 BB |
856 | err = dmu_tx_assign(tx, TXG_WAIT); |
857 | if (err) { | |
858 | dmu_tx_abort(tx); | |
859 | return (err); | |
860 | } | |
861 | ||
0037b49e | 862 | mutex_enter(&zfsvfs->z_lock); |
9babb374 | 863 | if (*objp == 0) { |
0037b49e | 864 | *objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 865 | DMU_OT_NONE, 0, tx); |
0037b49e | 866 | VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
867 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
868 | } | |
0037b49e | 869 | mutex_exit(&zfsvfs->z_lock); |
9babb374 BB |
870 | |
871 | if (quota == 0) { | |
0037b49e | 872 | err = zap_remove(zfsvfs->z_os, *objp, buf, tx); |
9babb374 BB |
873 | if (err == ENOENT) |
874 | err = 0; | |
875 | } else { | |
0037b49e | 876 | err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
877 | } |
878 | ASSERT(err == 0); | |
879 | if (fuid_dirtied) | |
0037b49e | 880 | zfs_fuid_sync(zfsvfs, tx); |
9babb374 BB |
881 | dmu_tx_commit(tx); |
882 | return (err); | |
883 | } | |
884 | ||
1de321e6 | 885 | boolean_t |
9c5167d1 | 886 | zfs_id_overobjquota(zfsvfs_t *zfsvfs, uint64_t usedobj, uint64_t id) |
1de321e6 JX |
887 | { |
888 | char buf[20 + DMU_OBJACCT_PREFIX_LEN]; | |
9c5167d1 | 889 | uint64_t used, quota, quotaobj; |
1de321e6 JX |
890 | int err; |
891 | ||
0037b49e | 892 | if (!dmu_objset_userobjspace_present(zfsvfs->z_os)) { |
c0daec32 AB |
893 | if (dmu_objset_userobjspace_upgradable(zfsvfs->z_os)) { |
894 | dsl_pool_config_enter( | |
895 | dmu_objset_pool(zfsvfs->z_os), FTAG); | |
9c5167d1 | 896 | dmu_objset_id_quota_upgrade(zfsvfs->z_os); |
c0daec32 AB |
897 | dsl_pool_config_exit( |
898 | dmu_objset_pool(zfsvfs->z_os), FTAG); | |
899 | } | |
1de321e6 JX |
900 | return (B_FALSE); |
901 | } | |
902 | ||
9c5167d1 NF |
903 | if (usedobj == DMU_PROJECTUSED_OBJECT) { |
904 | if (!dmu_objset_projectquota_present(zfsvfs->z_os)) { | |
905 | if (dmu_objset_projectquota_upgradable(zfsvfs->z_os)) { | |
906 | dsl_pool_config_enter( | |
907 | dmu_objset_pool(zfsvfs->z_os), FTAG); | |
908 | dmu_objset_id_quota_upgrade(zfsvfs->z_os); | |
909 | dsl_pool_config_exit( | |
910 | dmu_objset_pool(zfsvfs->z_os), FTAG); | |
911 | } | |
912 | return (B_FALSE); | |
913 | } | |
914 | quotaobj = zfsvfs->z_projectobjquota_obj; | |
915 | } else if (usedobj == DMU_USERUSED_OBJECT) { | |
916 | quotaobj = zfsvfs->z_userobjquota_obj; | |
917 | } else if (usedobj == DMU_GROUPUSED_OBJECT) { | |
918 | quotaobj = zfsvfs->z_groupobjquota_obj; | |
919 | } else { | |
920 | return (B_FALSE); | |
921 | } | |
0037b49e | 922 | if (quotaobj == 0 || zfsvfs->z_replay) |
1de321e6 JX |
923 | return (B_FALSE); |
924 | ||
9c5167d1 | 925 | (void) sprintf(buf, "%llx", (longlong_t)id); |
0037b49e | 926 | err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); |
1de321e6 JX |
927 | if (err != 0) |
928 | return (B_FALSE); | |
929 | ||
9c5167d1 | 930 | (void) sprintf(buf, DMU_OBJACCT_PREFIX "%llx", (longlong_t)id); |
0037b49e | 931 | err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); |
1de321e6 JX |
932 | if (err != 0) |
933 | return (B_FALSE); | |
934 | return (used >= quota); | |
935 | } | |
936 | ||
9babb374 | 937 | boolean_t |
9c5167d1 | 938 | zfs_id_overblockquota(zfsvfs_t *zfsvfs, uint64_t usedobj, uint64_t id) |
9babb374 | 939 | { |
1de321e6 | 940 | char buf[20]; |
9c5167d1 | 941 | uint64_t used, quota, quotaobj; |
9babb374 BB |
942 | int err; |
943 | ||
9c5167d1 NF |
944 | if (usedobj == DMU_PROJECTUSED_OBJECT) { |
945 | if (!dmu_objset_projectquota_present(zfsvfs->z_os)) { | |
946 | if (dmu_objset_projectquota_upgradable(zfsvfs->z_os)) { | |
947 | dsl_pool_config_enter( | |
948 | dmu_objset_pool(zfsvfs->z_os), FTAG); | |
949 | dmu_objset_id_quota_upgrade(zfsvfs->z_os); | |
950 | dsl_pool_config_exit( | |
951 | dmu_objset_pool(zfsvfs->z_os), FTAG); | |
952 | } | |
953 | return (B_FALSE); | |
954 | } | |
955 | quotaobj = zfsvfs->z_projectquota_obj; | |
956 | } else if (usedobj == DMU_USERUSED_OBJECT) { | |
957 | quotaobj = zfsvfs->z_userquota_obj; | |
958 | } else if (usedobj == DMU_GROUPUSED_OBJECT) { | |
959 | quotaobj = zfsvfs->z_groupquota_obj; | |
960 | } else { | |
961 | return (B_FALSE); | |
962 | } | |
0037b49e | 963 | if (quotaobj == 0 || zfsvfs->z_replay) |
9babb374 BB |
964 | return (B_FALSE); |
965 | ||
9c5167d1 | 966 | (void) sprintf(buf, "%llx", (longlong_t)id); |
0037b49e | 967 | err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
968 | if (err != 0) |
969 | return (B_FALSE); | |
970 | ||
0037b49e | 971 | err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
972 | if (err != 0) |
973 | return (B_FALSE); | |
974 | return (used >= quota); | |
975 | } | |
976 | ||
428870ff | 977 | boolean_t |
9c5167d1 | 978 | zfs_id_overquota(zfsvfs_t *zfsvfs, uint64_t usedobj, uint64_t id) |
428870ff | 979 | { |
9c5167d1 NF |
980 | return (zfs_id_overblockquota(zfsvfs, usedobj, id) || |
981 | zfs_id_overobjquota(zfsvfs, usedobj, id)); | |
428870ff BB |
982 | } |
983 | ||
8614ddf9 MA |
984 | /* |
985 | * Associate this zfsvfs with the given objset, which must be owned. | |
986 | * This will cache a bunch of on-disk state from the objset in the | |
987 | * zfsvfs. | |
988 | */ | |
989 | static int | |
990 | zfsvfs_init(zfsvfs_t *zfsvfs, objset_t *os) | |
9babb374 | 991 | { |
8614ddf9 MA |
992 | int error; |
993 | uint64_t val; | |
0282c413 | 994 | |
0037b49e BB |
995 | zfsvfs->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; |
996 | zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; | |
997 | zfsvfs->z_os = os; | |
9babb374 | 998 | |
0037b49e | 999 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); |
8614ddf9 MA |
1000 | if (error != 0) |
1001 | return (error); | |
1002 | if (zfsvfs->z_version > | |
1003 | zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { | |
1004 | (void) printk("Can't mount a version %lld file system " | |
1005 | "on a version %lld pool\n. Pool must be upgraded to mount " | |
b689de85 | 1006 | "this file system.\n", (u_longlong_t)zfsvfs->z_version, |
8614ddf9 MA |
1007 | (u_longlong_t)spa_version(dmu_objset_spa(os))); |
1008 | return (SET_ERROR(ENOTSUP)); | |
9babb374 | 1009 | } |
8614ddf9 MA |
1010 | error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &val); |
1011 | if (error != 0) | |
1012 | return (error); | |
1013 | zfsvfs->z_norm = (int)val; | |
9babb374 | 1014 | |
8614ddf9 MA |
1015 | error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &val); |
1016 | if (error != 0) | |
1017 | return (error); | |
1018 | zfsvfs->z_utf8 = (val != 0); | |
9babb374 | 1019 | |
8614ddf9 MA |
1020 | error = zfs_get_zplprop(os, ZFS_PROP_CASE, &val); |
1021 | if (error != 0) | |
1022 | return (error); | |
1023 | zfsvfs->z_case = (uint_t)val; | |
9babb374 | 1024 | |
8614ddf9 MA |
1025 | if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &val)) != 0) |
1026 | return (error); | |
1027 | zfsvfs->z_acl_type = (uint_t)val; | |
023699cd | 1028 | |
9babb374 BB |
1029 | /* |
1030 | * Fold case on file systems that are always or sometimes case | |
1031 | * insensitive. | |
1032 | */ | |
0037b49e BB |
1033 | if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
1034 | zfsvfs->z_case == ZFS_CASE_MIXED) | |
1035 | zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 1036 | |
0037b49e BB |
1037 | zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); |
1038 | zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); | |
428870ff | 1039 | |
8614ddf9 | 1040 | uint64_t sa_obj = 0; |
0037b49e | 1041 | if (zfsvfs->z_use_sa) { |
428870ff BB |
1042 | /* should either have both of these objects or none */ |
1043 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
1044 | &sa_obj); | |
8614ddf9 MA |
1045 | if (error != 0) |
1046 | return (error); | |
82a37189 | 1047 | |
8614ddf9 MA |
1048 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &val); |
1049 | if ((error == 0) && (val == ZFS_XATTR_SA)) | |
0037b49e | 1050 | zfsvfs->z_xattr_sa = B_TRUE; |
428870ff BB |
1051 | } |
1052 | ||
9babb374 | 1053 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, |
0037b49e | 1054 | &zfsvfs->z_root); |
8614ddf9 MA |
1055 | if (error != 0) |
1056 | return (error); | |
0037b49e | 1057 | ASSERT(zfsvfs->z_root != 0); |
9babb374 BB |
1058 | |
1059 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
0037b49e | 1060 | &zfsvfs->z_unlinkedobj); |
8614ddf9 MA |
1061 | if (error != 0) |
1062 | return (error); | |
9babb374 BB |
1063 | |
1064 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
1065 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
0037b49e | 1066 | 8, 1, &zfsvfs->z_userquota_obj); |
8614ddf9 MA |
1067 | if (error == ENOENT) |
1068 | zfsvfs->z_userquota_obj = 0; | |
1069 | else if (error != 0) | |
1070 | return (error); | |
9babb374 BB |
1071 | |
1072 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
1073 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
0037b49e | 1074 | 8, 1, &zfsvfs->z_groupquota_obj); |
8614ddf9 MA |
1075 | if (error == ENOENT) |
1076 | zfsvfs->z_groupquota_obj = 0; | |
1077 | else if (error != 0) | |
1078 | return (error); | |
9babb374 | 1079 | |
9c5167d1 NF |
1080 | error = zap_lookup(os, MASTER_NODE_OBJ, |
1081 | zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA], | |
1082 | 8, 1, &zfsvfs->z_projectquota_obj); | |
1083 | if (error == ENOENT) | |
1084 | zfsvfs->z_projectquota_obj = 0; | |
1085 | else if (error != 0) | |
1086 | return (error); | |
1087 | ||
1de321e6 JX |
1088 | error = zap_lookup(os, MASTER_NODE_OBJ, |
1089 | zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA], | |
0037b49e | 1090 | 8, 1, &zfsvfs->z_userobjquota_obj); |
8614ddf9 MA |
1091 | if (error == ENOENT) |
1092 | zfsvfs->z_userobjquota_obj = 0; | |
1093 | else if (error != 0) | |
1094 | return (error); | |
1de321e6 JX |
1095 | |
1096 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
1097 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA], | |
0037b49e | 1098 | 8, 1, &zfsvfs->z_groupobjquota_obj); |
8614ddf9 MA |
1099 | if (error == ENOENT) |
1100 | zfsvfs->z_groupobjquota_obj = 0; | |
1101 | else if (error != 0) | |
1102 | return (error); | |
1de321e6 | 1103 | |
9c5167d1 NF |
1104 | error = zap_lookup(os, MASTER_NODE_OBJ, |
1105 | zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTOBJQUOTA], | |
1106 | 8, 1, &zfsvfs->z_projectobjquota_obj); | |
1107 | if (error == ENOENT) | |
1108 | zfsvfs->z_projectobjquota_obj = 0; | |
1109 | else if (error != 0) | |
1110 | return (error); | |
1111 | ||
9babb374 | 1112 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, |
0037b49e | 1113 | &zfsvfs->z_fuid_obj); |
8614ddf9 MA |
1114 | if (error == ENOENT) |
1115 | zfsvfs->z_fuid_obj = 0; | |
1116 | else if (error != 0) | |
1117 | return (error); | |
9babb374 BB |
1118 | |
1119 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
0037b49e | 1120 | &zfsvfs->z_shares_dir); |
8614ddf9 MA |
1121 | if (error == ENOENT) |
1122 | zfsvfs->z_shares_dir = 0; | |
1123 | else if (error != 0) | |
1124 | return (error); | |
1125 | ||
04a3b079 TC |
1126 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
1127 | &zfsvfs->z_attr_table); | |
1128 | if (error != 0) | |
1129 | return (error); | |
1130 | ||
1131 | if (zfsvfs->z_version >= ZPL_VERSION_SA) | |
1132 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1133 | ||
8614ddf9 MA |
1134 | return (0); |
1135 | } | |
1136 | ||
1137 | int | |
163a8c28 | 1138 | zfsvfs_create(const char *osname, boolean_t readonly, zfsvfs_t **zfvp) |
8614ddf9 MA |
1139 | { |
1140 | objset_t *os; | |
1141 | zfsvfs_t *zfsvfs; | |
1142 | int error; | |
163a8c28 | 1143 | boolean_t ro = (readonly || (strchr(osname, '@') != NULL)); |
8614ddf9 MA |
1144 | |
1145 | zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); | |
1146 | ||
163a8c28 | 1147 | error = dmu_objset_own(osname, DMU_OST_ZFS, ro, B_TRUE, zfsvfs, &os); |
d99a0153 | 1148 | if (error != 0) { |
8614ddf9 MA |
1149 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); |
1150 | return (error); | |
1151 | } | |
1152 | ||
d99a0153 CW |
1153 | error = zfsvfs_create_impl(zfvp, zfsvfs, os); |
1154 | if (error != 0) { | |
1155 | dmu_objset_disown(os, B_TRUE, zfsvfs); | |
1156 | } | |
1157 | return (error); | |
1158 | } | |
1159 | ||
04a3b079 TC |
1160 | |
1161 | /* | |
1162 | * Note: zfsvfs is assumed to be malloc'd, and will be freed by this function | |
1163 | * on a failure. Do not pass in a statically allocated zfsvfs. | |
1164 | */ | |
d99a0153 CW |
1165 | int |
1166 | zfsvfs_create_impl(zfsvfs_t **zfvp, zfsvfs_t *zfsvfs, objset_t *os) | |
1167 | { | |
1168 | int error; | |
1169 | ||
8614ddf9 MA |
1170 | zfsvfs->z_vfs = NULL; |
1171 | zfsvfs->z_sb = NULL; | |
1172 | zfsvfs->z_parent = zfsvfs; | |
9babb374 | 1173 | |
0037b49e BB |
1174 | mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
1175 | mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
1176 | list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), | |
9babb374 | 1177 | offsetof(znode_t, z_link_node)); |
0037b49e BB |
1178 | rrm_init(&zfsvfs->z_teardown_lock, B_FALSE); |
1179 | rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); | |
1180 | rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
d07b7c7f | 1181 | |
8614ddf9 MA |
1182 | int size = MIN(1 << (highbit64(zfs_object_mutex_size) - 1), |
1183 | ZFS_OBJ_MTX_MAX); | |
0037b49e BB |
1184 | zfsvfs->z_hold_size = size; |
1185 | zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, | |
1186 | KM_SLEEP); | |
1187 | zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
8614ddf9 | 1188 | for (int i = 0; i != size; i++) { |
0037b49e | 1189 | avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare, |
c96c36fa | 1190 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); |
0037b49e | 1191 | mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); |
c96c36fa | 1192 | } |
9babb374 | 1193 | |
8614ddf9 MA |
1194 | error = zfsvfs_init(zfsvfs, os); |
1195 | if (error != 0) { | |
8614ddf9 | 1196 | *zfvp = NULL; |
04a3b079 | 1197 | zfsvfs_free(zfsvfs); |
8614ddf9 MA |
1198 | return (error); |
1199 | } | |
1200 | ||
dcec0a12 AP |
1201 | zfsvfs->z_drain_task = TASKQID_INVALID; |
1202 | zfsvfs->z_draining = B_FALSE; | |
1203 | zfsvfs->z_drain_cancel = B_TRUE; | |
1204 | ||
0037b49e | 1205 | *zfvp = zfsvfs; |
9babb374 | 1206 | return (0); |
9babb374 BB |
1207 | } |
1208 | ||
8614ddf9 | 1209 | static int |
f298b24d | 1210 | zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) |
34dc7c2f | 1211 | { |
34dc7c2f | 1212 | int error; |
ae76f45c TC |
1213 | boolean_t readonly = zfs_is_readonly(zfsvfs); |
1214 | ||
1c2555ef | 1215 | error = zfs_register_callbacks(zfsvfs->z_vfs); |
34dc7c2f BB |
1216 | if (error) |
1217 | return (error); | |
1218 | ||
0037b49e | 1219 | zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); |
9babb374 | 1220 | |
34dc7c2f BB |
1221 | /* |
1222 | * If we are not mounting (ie: online recv), then we don't | |
1223 | * have to worry about replaying the log as we blocked all | |
1224 | * operations out since we closed the ZIL. | |
1225 | */ | |
1226 | if (mounting) { | |
dcec0a12 AP |
1227 | ASSERT3P(zfsvfs->z_kstat.dk_kstats, ==, NULL); |
1228 | dataset_kstats_create(&zfsvfs->z_kstat, zfsvfs->z_os); | |
1229 | ||
34dc7c2f BB |
1230 | /* |
1231 | * During replay we remove the read only flag to | |
1232 | * allow replays to succeed. | |
1233 | */ | |
dcec0a12 | 1234 | if (readonly != 0) { |
0037b49e | 1235 | readonly_changed_cb(zfsvfs, B_FALSE); |
dcec0a12 AP |
1236 | } else { |
1237 | zap_stats_t zs; | |
1238 | if (zap_get_stats(zfsvfs->z_os, zfsvfs->z_unlinkedobj, | |
1239 | &zs) == 0) { | |
1240 | dataset_kstats_update_nunlinks_kstat( | |
1241 | &zfsvfs->z_kstat, zs.zs_num_entries); | |
1242 | } | |
1243 | dprintf_ds(zfsvfs->z_os->os_dsl_dataset, | |
1244 | "num_entries in unlinked set: %llu", | |
1245 | zs.zs_num_entries); | |
0037b49e | 1246 | zfs_unlinked_drain(zfsvfs); |
dcec0a12 | 1247 | } |
34dc7c2f | 1248 | |
428870ff BB |
1249 | /* |
1250 | * Parse and replay the intent log. | |
1251 | * | |
1252 | * Because of ziltest, this must be done after | |
1253 | * zfs_unlinked_drain(). (Further note: ziltest | |
1254 | * doesn't use readonly mounts, where | |
1255 | * zfs_unlinked_drain() isn't called.) This is because | |
1256 | * ziltest causes spa_sync() to think it's committed, | |
1257 | * but actually it is not, so the intent log contains | |
1258 | * many txg's worth of changes. | |
1259 | * | |
1260 | * In particular, if object N is in the unlinked set in | |
1261 | * the last txg to actually sync, then it could be | |
1262 | * actually freed in a later txg and then reallocated | |
1263 | * in a yet later txg. This would write a "create | |
1264 | * object N" record to the intent log. Normally, this | |
1265 | * would be fine because the spa_sync() would have | |
1266 | * written out the fact that object N is free, before | |
1267 | * we could write the "create object N" intent log | |
1268 | * record. | |
1269 | * | |
1270 | * But when we are in ziltest mode, we advance the "open | |
1271 | * txg" without actually spa_sync()-ing the changes to | |
1272 | * disk. So we would see that object N is still | |
1273 | * allocated and in the unlinked set, and there is an | |
1274 | * intent log record saying to allocate it. | |
1275 | */ | |
0037b49e | 1276 | if (spa_writeable(dmu_objset_spa(zfsvfs->z_os))) { |
572e2857 | 1277 | if (zil_replay_disable) { |
0037b49e | 1278 | zil_destroy(zfsvfs->z_log, B_FALSE); |
572e2857 | 1279 | } else { |
0037b49e BB |
1280 | zfsvfs->z_replay = B_TRUE; |
1281 | zil_replay(zfsvfs->z_os, zfsvfs, | |
572e2857 | 1282 | zfs_replay_vector); |
0037b49e | 1283 | zfsvfs->z_replay = B_FALSE; |
572e2857 | 1284 | } |
fb5f0bc8 | 1285 | } |
2cf7f52b BB |
1286 | |
1287 | /* restore readonly bit */ | |
1288 | if (readonly != 0) | |
0037b49e | 1289 | readonly_changed_cb(zfsvfs, B_TRUE); |
34dc7c2f BB |
1290 | } |
1291 | ||
9775e988 | 1292 | /* |
0037b49e | 1293 | * Set the objset user_ptr to track its zfsvfs. |
9775e988 | 1294 | */ |
0037b49e BB |
1295 | mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); |
1296 | dmu_objset_set_user(zfsvfs->z_os, zfsvfs); | |
1297 | mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); | |
9775e988 | 1298 | |
34dc7c2f BB |
1299 | return (0); |
1300 | } | |
1301 | ||
9babb374 | 1302 | void |
f298b24d | 1303 | zfsvfs_free(zfsvfs_t *zfsvfs) |
34dc7c2f | 1304 | { |
0037b49e | 1305 | int i, size = zfsvfs->z_hold_size; |
9babb374 | 1306 | |
0037b49e | 1307 | zfs_fuid_destroy(zfsvfs); |
9babb374 | 1308 | |
0037b49e BB |
1309 | mutex_destroy(&zfsvfs->z_znodes_lock); |
1310 | mutex_destroy(&zfsvfs->z_lock); | |
1311 | list_destroy(&zfsvfs->z_all_znodes); | |
1312 | rrm_destroy(&zfsvfs->z_teardown_lock); | |
1313 | rw_destroy(&zfsvfs->z_teardown_inactive_lock); | |
1314 | rw_destroy(&zfsvfs->z_fuid_lock); | |
c96c36fa | 1315 | for (i = 0; i != size; i++) { |
0037b49e BB |
1316 | avl_destroy(&zfsvfs->z_hold_trees[i]); |
1317 | mutex_destroy(&zfsvfs->z_hold_locks[i]); | |
c96c36fa | 1318 | } |
0037b49e BB |
1319 | vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size); |
1320 | vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size); | |
1c2555ef | 1321 | zfsvfs_vfs_free(zfsvfs->z_vfs); |
a448a255 | 1322 | dataset_kstats_destroy(&zfsvfs->z_kstat); |
0037b49e | 1323 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); |
34dc7c2f BB |
1324 | } |
1325 | ||
9babb374 | 1326 | static void |
0037b49e | 1327 | zfs_set_fuid_feature(zfsvfs_t *zfsvfs) |
9babb374 | 1328 | { |
0037b49e BB |
1329 | zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); |
1330 | zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); | |
9babb374 | 1331 | } |
34dc7c2f BB |
1332 | |
1333 | void | |
0037b49e | 1334 | zfs_unregister_callbacks(zfsvfs_t *zfsvfs) |
34dc7c2f | 1335 | { |
0037b49e | 1336 | objset_t *os = zfsvfs->z_os; |
34dc7c2f | 1337 | |
0eb21616 | 1338 | if (!dmu_objset_is_snapshot(os)) |
0037b49e | 1339 | dsl_prop_unregister_all(dmu_objset_ds(os), zfsvfs); |
34dc7c2f BB |
1340 | } |
1341 | ||
bc3e15e3 | 1342 | #ifdef HAVE_MLSLABEL |
428870ff | 1343 | /* |
d3cc8b15 WA |
1344 | * Check that the hex label string is appropriate for the dataset being |
1345 | * mounted into the global_zone proper. | |
428870ff | 1346 | * |
d3cc8b15 WA |
1347 | * Return an error if the hex label string is not default or |
1348 | * admin_low/admin_high. For admin_low labels, the corresponding | |
1349 | * dataset must be readonly. | |
428870ff BB |
1350 | */ |
1351 | int | |
1352 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
1353 | { | |
1354 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
1355 | return (0); | |
1356 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
1357 | return (0); | |
1358 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
1359 | /* must be readonly */ | |
1360 | uint64_t rdonly; | |
1361 | ||
1362 | if (dsl_prop_get_integer(dsname, | |
1363 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
2e528b49 | 1364 | return (SET_ERROR(EACCES)); |
428870ff BB |
1365 | return (rdonly ? 0 : EACCES); |
1366 | } | |
2e528b49 | 1367 | return (SET_ERROR(EACCES)); |
428870ff | 1368 | } |
bc3e15e3 | 1369 | #endif /* HAVE_MLSLABEL */ |
428870ff | 1370 | |
9c5167d1 NF |
1371 | static int |
1372 | zfs_statfs_project(zfsvfs_t *zfsvfs, znode_t *zp, struct kstatfs *statp, | |
1373 | uint32_t bshift) | |
1374 | { | |
1375 | char buf[20 + DMU_OBJACCT_PREFIX_LEN]; | |
1376 | uint64_t offset = DMU_OBJACCT_PREFIX_LEN; | |
1377 | uint64_t quota; | |
1378 | uint64_t used; | |
1379 | int err; | |
1380 | ||
1381 | strlcpy(buf, DMU_OBJACCT_PREFIX, DMU_OBJACCT_PREFIX_LEN + 1); | |
1382 | err = id_to_fuidstr(zfsvfs, NULL, zp->z_projid, buf + offset, B_FALSE); | |
1383 | if (err) | |
1384 | return (err); | |
1385 | ||
1386 | if (zfsvfs->z_projectquota_obj == 0) | |
1387 | goto objs; | |
1388 | ||
1389 | err = zap_lookup(zfsvfs->z_os, zfsvfs->z_projectquota_obj, | |
1390 | buf + offset, 8, 1, "a); | |
1391 | if (err == ENOENT) | |
1392 | goto objs; | |
1393 | else if (err) | |
1394 | return (err); | |
1395 | ||
1396 | err = zap_lookup(zfsvfs->z_os, DMU_PROJECTUSED_OBJECT, | |
1397 | buf + offset, 8, 1, &used); | |
1398 | if (unlikely(err == ENOENT)) { | |
1399 | uint32_t blksize; | |
1400 | u_longlong_t nblocks; | |
1401 | ||
1402 | /* | |
1403 | * Quota accounting is async, so it is possible race case. | |
1404 | * There is at least one object with the given project ID. | |
1405 | */ | |
1406 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); | |
1407 | if (unlikely(zp->z_blksz == 0)) | |
1408 | blksize = zfsvfs->z_max_blksz; | |
1409 | ||
1410 | used = blksize * nblocks; | |
1411 | } else if (err) { | |
1412 | return (err); | |
1413 | } | |
1414 | ||
1415 | statp->f_blocks = quota >> bshift; | |
1416 | statp->f_bfree = (quota > used) ? ((quota - used) >> bshift) : 0; | |
1417 | statp->f_bavail = statp->f_bfree; | |
1418 | ||
1419 | objs: | |
1420 | if (zfsvfs->z_projectobjquota_obj == 0) | |
1421 | return (0); | |
1422 | ||
1423 | err = zap_lookup(zfsvfs->z_os, zfsvfs->z_projectobjquota_obj, | |
1424 | buf + offset, 8, 1, "a); | |
1425 | if (err == ENOENT) | |
1426 | return (0); | |
1427 | else if (err) | |
1428 | return (err); | |
1429 | ||
1430 | err = zap_lookup(zfsvfs->z_os, DMU_PROJECTUSED_OBJECT, | |
1431 | buf, 8, 1, &used); | |
1432 | if (unlikely(err == ENOENT)) { | |
1433 | /* | |
1434 | * Quota accounting is async, so it is possible race case. | |
1435 | * There is at least one object with the given project ID. | |
1436 | */ | |
1437 | used = 1; | |
1438 | } else if (err) { | |
1439 | return (err); | |
1440 | } | |
1441 | ||
1442 | statp->f_files = quota; | |
1443 | statp->f_ffree = (quota > used) ? (quota - used) : 0; | |
1444 | ||
1445 | return (0); | |
1446 | } | |
1447 | ||
e5c39b95 | 1448 | int |
3558fd73 | 1449 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 1450 | { |
0037b49e | 1451 | zfsvfs_t *zfsvfs = dentry->d_sb->s_fs_info; |
34dc7c2f | 1452 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
9c5167d1 | 1453 | int err = 0; |
34dc7c2f | 1454 | |
0037b49e | 1455 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1456 | |
0037b49e | 1457 | dmu_objset_space(zfsvfs->z_os, |
34dc7c2f BB |
1458 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
1459 | ||
e897a23e | 1460 | uint64_t fsid = dmu_objset_fsid_guid(zfsvfs->z_os); |
34dc7c2f | 1461 | /* |
05ff35c6 BB |
1462 | * The underlying storage pool actually uses multiple block |
1463 | * size. Under Solaris frsize (fragment size) is reported as | |
1464 | * the smallest block size we support, and bsize (block size) | |
1465 | * as the filesystem's maximum block size. Unfortunately, | |
1466 | * under Linux the fragment size and block size are often used | |
1467 | * interchangeably. Thus we are forced to report both of them | |
1468 | * as the filesystem's maximum block size. | |
34dc7c2f | 1469 | */ |
0037b49e BB |
1470 | statp->f_frsize = zfsvfs->z_max_blksz; |
1471 | statp->f_bsize = zfsvfs->z_max_blksz; | |
e897a23e | 1472 | uint32_t bshift = fls(statp->f_bsize) - 1; |
34dc7c2f BB |
1473 | |
1474 | /* | |
3558fd73 BB |
1475 | * The following report "total" blocks of various kinds in |
1476 | * the file system, but reported in terms of f_bsize - the | |
1477 | * "preferred" size. | |
34dc7c2f BB |
1478 | */ |
1479 | ||
e1cfd73f | 1480 | /* Round up so we never have a filesystem using 0 blocks. */ |
83c796c5 | 1481 | refdbytes = P2ROUNDUP(refdbytes, statp->f_bsize); |
3558fd73 BB |
1482 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
1483 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
1484 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
1485 | ||
1486 | /* | |
1487 | * statvfs() should really be called statufs(), because it assumes | |
1488 | * static metadata. ZFS doesn't preallocate files, so the best | |
1489 | * we can do is report the max that could possibly fit in f_files, | |
1490 | * and that minus the number actually used in f_ffree. | |
e897a23e | 1491 | * For f_ffree, report the smaller of the number of objects available |
34dc7c2f BB |
1492 | * and the number of blocks (each object will take at least a block). |
1493 | */ | |
baab0630 | 1494 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 1495 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
1496 | statp->f_fsid.val[0] = (uint32_t)fsid; |
1497 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 | 1498 | statp->f_type = ZFS_SUPER_MAGIC; |
eca7b760 | 1499 | statp->f_namelen = MAXNAMELEN - 1; |
34dc7c2f BB |
1500 | |
1501 | /* | |
3558fd73 | 1502 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
1503 | * Is there anything useful we could/should provide? |
1504 | */ | |
3558fd73 | 1505 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 1506 | |
9c5167d1 NF |
1507 | if (dmu_objset_projectquota_enabled(zfsvfs->z_os) && |
1508 | dmu_objset_projectquota_present(zfsvfs->z_os)) { | |
1509 | znode_t *zp = ITOZ(dentry->d_inode); | |
1510 | ||
1511 | if (zp->z_pflags & ZFS_PROJINHERIT && zp->z_projid && | |
1512 | zpl_is_valid_projid(zp->z_projid)) | |
1513 | err = zfs_statfs_project(zfsvfs, zp, statp, bshift); | |
1514 | } | |
1515 | ||
0037b49e | 1516 | ZFS_EXIT(zfsvfs); |
9c5167d1 | 1517 | return (err); |
34dc7c2f BB |
1518 | } |
1519 | ||
e5c39b95 | 1520 | int |
0037b49e | 1521 | zfs_root(zfsvfs_t *zfsvfs, struct inode **ipp) |
34dc7c2f | 1522 | { |
34dc7c2f BB |
1523 | znode_t *rootzp; |
1524 | int error; | |
1525 | ||
0037b49e | 1526 | ZFS_ENTER(zfsvfs); |
34dc7c2f | 1527 | |
0037b49e | 1528 | error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp); |
34dc7c2f | 1529 | if (error == 0) |
3558fd73 | 1530 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1531 | |
0037b49e | 1532 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
1533 | return (error); |
1534 | } | |
1535 | ||
09fb30e5 | 1536 | #ifdef HAVE_D_PRUNE_ALIASES |
218b4e0a BB |
1537 | /* |
1538 | * Linux kernels older than 3.1 do not support a per-filesystem shrinker. | |
1539 | * To accommodate this we must improvise and manually walk the list of znodes | |
1540 | * attempting to prune dentries in order to be able to drop the inodes. | |
1541 | * | |
1542 | * To avoid scanning the same znodes multiple times they are always rotated | |
1543 | * to the end of the z_all_znodes list. New znodes are inserted at the | |
1544 | * end of the list so we're always scanning the oldest znodes first. | |
1545 | */ | |
1546 | static int | |
f298b24d | 1547 | zfs_prune_aliases(zfsvfs_t *zfsvfs, unsigned long nr_to_scan) |
218b4e0a BB |
1548 | { |
1549 | znode_t **zp_array, *zp; | |
1550 | int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *)); | |
1551 | int objects = 0; | |
1552 | int i = 0, j = 0; | |
1553 | ||
1554 | zp_array = kmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP); | |
1555 | ||
0037b49e BB |
1556 | mutex_enter(&zfsvfs->z_znodes_lock); |
1557 | while ((zp = list_head(&zfsvfs->z_all_znodes)) != NULL) { | |
218b4e0a BB |
1558 | |
1559 | if ((i++ > nr_to_scan) || (j >= max_array)) | |
1560 | break; | |
1561 | ||
1562 | ASSERT(list_link_active(&zp->z_link_node)); | |
0037b49e BB |
1563 | list_remove(&zfsvfs->z_all_znodes, zp); |
1564 | list_insert_tail(&zfsvfs->z_all_znodes, zp); | |
218b4e0a BB |
1565 | |
1566 | /* Skip active znodes and .zfs entries */ | |
1567 | if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir) | |
1568 | continue; | |
1569 | ||
1570 | if (igrab(ZTOI(zp)) == NULL) | |
1571 | continue; | |
1572 | ||
1573 | zp_array[j] = zp; | |
1574 | j++; | |
1575 | } | |
0037b49e | 1576 | mutex_exit(&zfsvfs->z_znodes_lock); |
218b4e0a BB |
1577 | |
1578 | for (i = 0; i < j; i++) { | |
1579 | zp = zp_array[i]; | |
1580 | ||
1581 | ASSERT3P(zp, !=, NULL); | |
1582 | d_prune_aliases(ZTOI(zp)); | |
1583 | ||
1584 | if (atomic_read(&ZTOI(zp)->i_count) == 1) | |
1585 | objects++; | |
1586 | ||
1587 | iput(ZTOI(zp)); | |
1588 | } | |
1589 | ||
1590 | kmem_free(zp_array, max_array * sizeof (znode_t *)); | |
1591 | ||
1592 | return (objects); | |
1593 | } | |
1594 | #endif /* HAVE_D_PRUNE_ALIASES */ | |
1595 | ||
2cbb06b5 BB |
1596 | /* |
1597 | * The ARC has requested that the filesystem drop entries from the dentry | |
1598 | * and inode caches. This can occur when the ARC needs to free meta data | |
1599 | * blocks but can't because they are all pinned by entries in these caches. | |
1600 | */ | |
ab26409d | 1601 | int |
f298b24d | 1602 | zfs_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) |
ab26409d | 1603 | { |
0037b49e | 1604 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
2cbb06b5 BB |
1605 | int error = 0; |
1606 | #if defined(HAVE_SHRINK) || defined(HAVE_SPLIT_SHRINKER_CALLBACK) | |
ab26409d BB |
1607 | struct shrinker *shrinker = &sb->s_shrink; |
1608 | struct shrink_control sc = { | |
1609 | .nr_to_scan = nr_to_scan, | |
1610 | .gfp_mask = GFP_KERNEL, | |
1611 | }; | |
2cbb06b5 | 1612 | #endif |
ab26409d | 1613 | |
0037b49e | 1614 | ZFS_ENTER(zfsvfs); |
2cbb06b5 | 1615 | |
90947b23 TC |
1616 | #if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \ |
1617 | defined(SHRINK_CONTROL_HAS_NID) && \ | |
1618 | defined(SHRINKER_NUMA_AWARE) | |
1619 | if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) { | |
1620 | *objects = 0; | |
02730c33 | 1621 | for_each_online_node(sc.nid) { |
90947b23 | 1622 | *objects += (*shrinker->scan_objects)(shrinker, &sc); |
02730c33 | 1623 | } |
90947b23 TC |
1624 | } else { |
1625 | *objects = (*shrinker->scan_objects)(shrinker, &sc); | |
1626 | } | |
09fb30e5 | 1627 | |
90947b23 | 1628 | #elif defined(HAVE_SPLIT_SHRINKER_CALLBACK) |
3c832b8c | 1629 | *objects = (*shrinker->scan_objects)(shrinker, &sc); |
2cbb06b5 | 1630 | #elif defined(HAVE_SHRINK) |
ab26409d | 1631 | *objects = (*shrinker->shrink)(shrinker, &sc); |
218b4e0a | 1632 | #elif defined(HAVE_D_PRUNE_ALIASES) |
09fb30e5 | 1633 | #define D_PRUNE_ALIASES_IS_DEFAULT |
f298b24d | 1634 | *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); |
2cbb06b5 | 1635 | #else |
218b4e0a | 1636 | #error "No available dentry and inode cache pruning mechanism." |
3c832b8c | 1637 | #endif |
09fb30e5 TC |
1638 | |
1639 | #if defined(HAVE_D_PRUNE_ALIASES) && !defined(D_PRUNE_ALIASES_IS_DEFAULT) | |
1640 | #undef D_PRUNE_ALIASES_IS_DEFAULT | |
1641 | /* | |
f298b24d | 1642 | * Fall back to zfs_prune_aliases if the kernel's per-superblock |
09fb30e5 TC |
1643 | * shrinker couldn't free anything, possibly due to the inodes being |
1644 | * allocated in a different memcg. | |
1645 | */ | |
1646 | if (*objects == 0) | |
f298b24d | 1647 | *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); |
09fb30e5 TC |
1648 | #endif |
1649 | ||
0037b49e | 1650 | ZFS_EXIT(zfsvfs); |
ab26409d | 1651 | |
0037b49e | 1652 | dprintf_ds(zfsvfs->z_os->os_dsl_dataset, |
2cbb06b5 BB |
1653 | "pruning, nr_to_scan=%lu objects=%d error=%d\n", |
1654 | nr_to_scan, *objects, error); | |
1655 | ||
1656 | return (error); | |
ab26409d | 1657 | } |
ab26409d | 1658 | |
34dc7c2f | 1659 | /* |
0037b49e | 1660 | * Teardown the zfsvfs_t. |
34dc7c2f | 1661 | * |
a08abc1b | 1662 | * Note, if 'unmounting' is FALSE, we return with the 'z_teardown_lock' |
34dc7c2f BB |
1663 | * and 'z_teardown_inactive_lock' held. |
1664 | */ | |
f298b24d | 1665 | static int |
0037b49e | 1666 | zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) |
34dc7c2f BB |
1667 | { |
1668 | znode_t *zp; | |
1669 | ||
dcec0a12 AP |
1670 | zfs_unlinked_drain_stop_wait(zfsvfs); |
1671 | ||
fd23720a BB |
1672 | /* |
1673 | * If someone has not already unmounted this file system, | |
1674 | * drain the iput_taskq to ensure all active references to the | |
0037b49e | 1675 | * zfsvfs_t have been handled only then can it be safely destroyed. |
fd23720a | 1676 | */ |
0037b49e | 1677 | if (zfsvfs->z_os) { |
f0da4d15 CD |
1678 | /* |
1679 | * If we're unmounting we have to wait for the list to | |
1680 | * drain completely. | |
1681 | * | |
1682 | * If we're not unmounting there's no guarantee the list | |
1683 | * will drain completely, but iputs run from the taskq | |
1684 | * may add the parents of dir-based xattrs to the taskq | |
1685 | * so we want to wait for these. | |
1686 | * | |
1687 | * We can safely read z_nr_znodes without locking because the | |
1688 | * VFS has already blocked operations which add to the | |
1689 | * z_all_znodes list and thus increment z_nr_znodes. | |
1690 | */ | |
1691 | int round = 0; | |
0037b49e | 1692 | while (zfsvfs->z_nr_znodes > 0) { |
c5528b9b | 1693 | taskq_wait_outstanding(dsl_pool_iput_taskq( |
0037b49e | 1694 | dmu_objset_pool(zfsvfs->z_os)), 0); |
f0da4d15 CD |
1695 | if (++round > 1 && !unmounting) |
1696 | break; | |
1697 | } | |
1698 | } | |
fd23720a | 1699 | |
0037b49e | 1700 | rrm_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
1701 | |
1702 | if (!unmounting) { | |
1703 | /* | |
ceb43b93 BB |
1704 | * We purge the parent filesystem's super block as the |
1705 | * parent filesystem and all of its snapshots have their | |
1706 | * inode's super block set to the parent's filesystem's | |
1707 | * super block. Note, 'z_parent' is self referential | |
1708 | * for non-snapshots. | |
34dc7c2f | 1709 | */ |
0037b49e | 1710 | shrink_dcache_sb(zfsvfs->z_parent->z_sb); |
34dc7c2f BB |
1711 | } |
1712 | ||
1713 | /* | |
1714 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1715 | * threads are blocked as zil_close can call zfs_inactive. | |
1716 | */ | |
0037b49e BB |
1717 | if (zfsvfs->z_log) { |
1718 | zil_close(zfsvfs->z_log); | |
1719 | zfsvfs->z_log = NULL; | |
34dc7c2f BB |
1720 | } |
1721 | ||
0037b49e | 1722 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1723 | |
1724 | /* | |
1725 | * If we are not unmounting (ie: online recv) and someone already | |
1726 | * unmounted this file system while we were doing the switcheroo, | |
1727 | * or a reopen of z_os failed then just bail out now. | |
1728 | */ | |
0037b49e BB |
1729 | if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) { |
1730 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
1731 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
2e528b49 | 1732 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1733 | } |
1734 | ||
1735 | /* | |
7b3e34ba BB |
1736 | * At this point there are no VFS ops active, and any new VFS ops |
1737 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1738 | * relevant for forced unmount). | |
34dc7c2f | 1739 | * |
e7a2fa70 TC |
1740 | * Release all holds on dbufs. We also grab an extra reference to all |
1741 | * the remaining inodes so that the kernel does not attempt to free | |
1742 | * any inodes of a suspended fs. This can cause deadlocks since the | |
1743 | * zfs_resume_fs() process may involve starting threads, which might | |
1744 | * attempt to free unreferenced inodes to free up memory for the new | |
1745 | * thread. | |
34dc7c2f | 1746 | */ |
f0da4d15 | 1747 | if (!unmounting) { |
0037b49e BB |
1748 | mutex_enter(&zfsvfs->z_znodes_lock); |
1749 | for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL; | |
1750 | zp = list_next(&zfsvfs->z_all_znodes, zp)) { | |
f0da4d15 CD |
1751 | if (zp->z_sa_hdl) |
1752 | zfs_znode_dmu_fini(zp); | |
e7a2fa70 TC |
1753 | if (igrab(ZTOI(zp)) != NULL) |
1754 | zp->z_suspended = B_TRUE; | |
1755 | ||
f0da4d15 | 1756 | } |
0037b49e | 1757 | mutex_exit(&zfsvfs->z_znodes_lock); |
7b3e34ba | 1758 | } |
34dc7c2f BB |
1759 | |
1760 | /* | |
7b3e34ba | 1761 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1762 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1763 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1764 | */ |
1765 | if (unmounting) { | |
0037b49e BB |
1766 | zfsvfs->z_unmounted = B_TRUE; |
1767 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
1768 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1769 | } |
1770 | ||
1771 | /* | |
1772 | * z_os will be NULL if there was an error in attempting to reopen | |
0037b49e | 1773 | * zfsvfs, so just return as the properties had already been |
3558fd73 | 1774 | * |
34dc7c2f BB |
1775 | * unregistered and cached data had been evicted before. |
1776 | */ | |
0037b49e | 1777 | if (zfsvfs->z_os == NULL) |
34dc7c2f BB |
1778 | return (0); |
1779 | ||
1780 | /* | |
1781 | * Unregister properties. | |
1782 | */ | |
0037b49e | 1783 | zfs_unregister_callbacks(zfsvfs); |
34dc7c2f BB |
1784 | |
1785 | /* | |
47ab01a1 TC |
1786 | * Evict cached data. We must write out any dirty data before |
1787 | * disowning the dataset. | |
34dc7c2f | 1788 | */ |
e2b31b58 PZ |
1789 | objset_t *os = zfsvfs->z_os; |
1790 | boolean_t os_dirty = B_FALSE; | |
1791 | for (int t = 0; t < TXG_SIZE; t++) { | |
1792 | if (dmu_objset_is_dirty(os, t)) { | |
1793 | os_dirty = B_TRUE; | |
1794 | break; | |
1795 | } | |
1796 | } | |
1797 | if (!zfs_is_readonly(zfsvfs) && os_dirty) { | |
0037b49e | 1798 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); |
e2b31b58 | 1799 | } |
0037b49e | 1800 | dmu_objset_evict_dbufs(zfsvfs->z_os); |
34dc7c2f BB |
1801 | |
1802 | return (0); | |
1803 | } | |
1804 | ||
3e6c9433 | 1805 | #if !defined(HAVE_2ARGS_BDI_SETUP_AND_REGISTER) && \ |
8c45def2 | 1806 | !defined(HAVE_3ARGS_BDI_SETUP_AND_REGISTER) |
5547c2f1 | 1807 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); |
8c45def2 | 1808 | #endif |
76659dc1 | 1809 | |
e5c39b95 | 1810 | int |
1c2555ef | 1811 | zfs_domount(struct super_block *sb, zfs_mnt_t *zm, int silent) |
34dc7c2f | 1812 | { |
1c2555ef | 1813 | const char *osname = zm->mnt_osname; |
3558fd73 BB |
1814 | struct inode *root_inode; |
1815 | uint64_t recordsize; | |
1c2555ef | 1816 | int error = 0; |
163a8c28 TC |
1817 | zfsvfs_t *zfsvfs = NULL; |
1818 | vfs_t *vfs = NULL; | |
1c2555ef BB |
1819 | |
1820 | ASSERT(zm); | |
1821 | ASSERT(osname); | |
34dc7c2f | 1822 | |
163a8c28 | 1823 | error = zfsvfs_parse_options(zm->mnt_data, &vfs); |
3558fd73 BB |
1824 | if (error) |
1825 | return (error); | |
1826 | ||
163a8c28 TC |
1827 | error = zfsvfs_create(osname, vfs->vfs_readonly, &zfsvfs); |
1828 | if (error) { | |
1829 | zfsvfs_vfs_free(vfs); | |
1c2555ef | 1830 | goto out; |
163a8c28 | 1831 | } |
1c2555ef | 1832 | |
3558fd73 | 1833 | if ((error = dsl_prop_get_integer(osname, "recordsize", |
163a8c28 TC |
1834 | &recordsize, NULL))) { |
1835 | zfsvfs_vfs_free(vfs); | |
3558fd73 | 1836 | goto out; |
163a8c28 | 1837 | } |
3558fd73 | 1838 | |
163a8c28 TC |
1839 | vfs->vfs_data = zfsvfs; |
1840 | zfsvfs->z_vfs = vfs; | |
0037b49e BB |
1841 | zfsvfs->z_sb = sb; |
1842 | sb->s_fs_info = zfsvfs; | |
3558fd73 BB |
1843 | sb->s_magic = ZFS_SUPER_MAGIC; |
1844 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1845 | sb->s_time_gran = 1; | |
1846 | sb->s_blocksize = recordsize; | |
1847 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 | 1848 | |
7dae2c81 | 1849 | error = -zpl_bdi_setup(sb, "zfs"); |
5547c2f1 BB |
1850 | if (error) |
1851 | goto out; | |
3558fd73 | 1852 | |
7dae2c81 BB |
1853 | sb->s_bdi->ra_pages = 0; |
1854 | ||
3558fd73 BB |
1855 | /* Set callback operations for the file system. */ |
1856 | sb->s_op = &zpl_super_operations; | |
1857 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1858 | sb->s_export_op = &zpl_export_operations; |
ee930353 BB |
1859 | #ifdef HAVE_S_D_OP |
1860 | sb->s_d_op = &zpl_dentry_operations; | |
1861 | #endif /* HAVE_S_D_OP */ | |
3558fd73 BB |
1862 | |
1863 | /* Set features for file system. */ | |
0037b49e | 1864 | zfs_set_fuid_feature(zfsvfs); |
3558fd73 | 1865 | |
0037b49e | 1866 | if (dmu_objset_is_snapshot(zfsvfs->z_os)) { |
3558fd73 BB |
1867 | uint64_t pval; |
1868 | ||
0037b49e BB |
1869 | atime_changed_cb(zfsvfs, B_FALSE); |
1870 | readonly_changed_cb(zfsvfs, B_TRUE); | |
d1d7e268 MK |
1871 | if ((error = dsl_prop_get_integer(osname, |
1872 | "xattr", &pval, NULL))) | |
3558fd73 | 1873 | goto out; |
0037b49e | 1874 | xattr_changed_cb(zfsvfs, pval); |
d1d7e268 MK |
1875 | if ((error = dsl_prop_get_integer(osname, |
1876 | "acltype", &pval, NULL))) | |
023699cd | 1877 | goto out; |
0037b49e BB |
1878 | acltype_changed_cb(zfsvfs, pval); |
1879 | zfsvfs->z_issnap = B_TRUE; | |
1880 | zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1881 | zfsvfs->z_snap_defer_time = jiffies; | |
1882 | ||
1883 | mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); | |
1884 | dmu_objset_set_user(zfsvfs->z_os, zfsvfs); | |
1885 | mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); | |
3558fd73 | 1886 | } else { |
f298b24d | 1887 | if ((error = zfsvfs_setup(zfsvfs, B_TRUE))) |
be88e733 | 1888 | goto out; |
34dc7c2f BB |
1889 | } |
1890 | ||
3558fd73 | 1891 | /* Allocate a root inode for the filesystem. */ |
0037b49e | 1892 | error = zfs_root(zfsvfs, &root_inode); |
3558fd73 BB |
1893 | if (error) { |
1894 | (void) zfs_umount(sb); | |
1895 | goto out; | |
34dc7c2f BB |
1896 | } |
1897 | ||
3558fd73 | 1898 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1899 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1900 | if (sb->s_root == NULL) { |
1901 | (void) zfs_umount(sb); | |
2e528b49 | 1902 | error = SET_ERROR(ENOMEM); |
3558fd73 BB |
1903 | goto out; |
1904 | } | |
ebe7e575 | 1905 | |
0037b49e BB |
1906 | if (!zfsvfs->z_issnap) |
1907 | zfsctl_create(zfsvfs); | |
2cbb06b5 | 1908 | |
0037b49e | 1909 | zfsvfs->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); |
3558fd73 BB |
1910 | out: |
1911 | if (error) { | |
163a8c28 TC |
1912 | if (zfsvfs != NULL) { |
1913 | dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs); | |
1914 | zfsvfs_free(zfsvfs); | |
1915 | } | |
be88e733 CC |
1916 | /* |
1917 | * make sure we don't have dangling sb->s_fs_info which | |
1918 | * zfs_preumount will use. | |
1919 | */ | |
1920 | sb->s_fs_info = NULL; | |
3558fd73 | 1921 | } |
34dc7c2f | 1922 | |
3558fd73 BB |
1923 | return (error); |
1924 | } | |
3558fd73 | 1925 | |
ebe7e575 BB |
1926 | /* |
1927 | * Called when an unmount is requested and certain sanity checks have | |
1928 | * already passed. At this point no dentries or inodes have been reclaimed | |
1929 | * from their respective caches. We drop the extra reference on the .zfs | |
1930 | * control directory to allow everything to be reclaimed. All snapshots | |
1931 | * must already have been unmounted to reach this point. | |
1932 | */ | |
1933 | void | |
1934 | zfs_preumount(struct super_block *sb) | |
1935 | { | |
0037b49e | 1936 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
ebe7e575 | 1937 | |
0037b49e BB |
1938 | /* zfsvfs is NULL when zfs_domount fails during mount */ |
1939 | if (zfsvfs) { | |
dcec0a12 | 1940 | zfs_unlinked_drain_stop_wait(zfsvfs); |
278bee93 | 1941 | zfsctl_destroy(sb->s_fs_info); |
be88e733 CC |
1942 | /* |
1943 | * Wait for iput_async before entering evict_inodes in | |
1944 | * generic_shutdown_super. The reason we must finish before | |
1945 | * evict_inodes is when lazytime is on, or when zfs_purgedir | |
1946 | * calls zfs_zget, iput would bump i_count from 0 to 1. This | |
1947 | * would race with the i_count check in evict_inodes. This means | |
1948 | * it could destroy the inode while we are still using it. | |
1949 | * | |
1950 | * We wait for two passes. xattr directories in the first pass | |
1951 | * may add xattr entries in zfs_purgedir, so in the second pass | |
1952 | * we wait for them. We don't use taskq_wait here because it is | |
1953 | * a pool wide taskq. Other mounted filesystems can constantly | |
1954 | * do iput_async and there's no guarantee when taskq will be | |
1955 | * empty. | |
1956 | */ | |
1957 | taskq_wait_outstanding(dsl_pool_iput_taskq( | |
0037b49e | 1958 | dmu_objset_pool(zfsvfs->z_os)), 0); |
be88e733 | 1959 | taskq_wait_outstanding(dsl_pool_iput_taskq( |
0037b49e | 1960 | dmu_objset_pool(zfsvfs->z_os)), 0); |
be88e733 | 1961 | } |
ebe7e575 | 1962 | } |
ebe7e575 BB |
1963 | |
1964 | /* | |
1965 | * Called once all other unmount released tear down has occurred. | |
1966 | * It is our responsibility to release any remaining infrastructure. | |
1967 | */ | |
3558fd73 BB |
1968 | /*ARGSUSED*/ |
1969 | int | |
1970 | zfs_umount(struct super_block *sb) | |
1971 | { | |
0037b49e | 1972 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
3558fd73 BB |
1973 | objset_t *os; |
1974 | ||
b5256303 TC |
1975 | if (zfsvfs->z_arc_prune != NULL) |
1976 | arc_remove_prune_callback(zfsvfs->z_arc_prune); | |
0037b49e BB |
1977 | VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0); |
1978 | os = zfsvfs->z_os; | |
7dae2c81 | 1979 | zpl_bdi_destroy(sb); |
76659dc1 | 1980 | |
34dc7c2f BB |
1981 | /* |
1982 | * z_os will be NULL if there was an error in | |
0037b49e | 1983 | * attempting to reopen zfsvfs. |
34dc7c2f BB |
1984 | */ |
1985 | if (os != NULL) { | |
1986 | /* | |
1987 | * Unset the objset user_ptr. | |
1988 | */ | |
428870ff | 1989 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1990 | dmu_objset_set_user(os, NULL); |
428870ff | 1991 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1992 | |
1993 | /* | |
b128c09f | 1994 | * Finally release the objset |
34dc7c2f | 1995 | */ |
b5256303 | 1996 | dmu_objset_disown(os, B_TRUE, zfsvfs); |
34dc7c2f BB |
1997 | } |
1998 | ||
f298b24d | 1999 | zfsvfs_free(zfsvfs); |
34dc7c2f BB |
2000 | return (0); |
2001 | } | |
2002 | ||
0de19dad | 2003 | int |
1c2555ef | 2004 | zfs_remount(struct super_block *sb, int *flags, zfs_mnt_t *zm) |
0de19dad | 2005 | { |
0037b49e | 2006 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
1c2555ef | 2007 | vfs_t *vfsp; |
08de8c16 | 2008 | boolean_t issnap = dmu_objset_is_snapshot(zfsvfs->z_os); |
0282c413 BB |
2009 | int error; |
2010 | ||
08de8c16 | 2011 | if ((issnap || !spa_writeable(dmu_objset_spa(zfsvfs->z_os))) && |
05805494 TH |
2012 | !(*flags & SB_RDONLY)) { |
2013 | *flags |= SB_RDONLY; | |
08de8c16 | 2014 | return (EROFS); |
2015 | } | |
2016 | ||
1c2555ef BB |
2017 | error = zfsvfs_parse_options(zm->mnt_data, &vfsp); |
2018 | if (error) | |
2019 | return (error); | |
2020 | ||
05805494 | 2021 | if (!zfs_is_readonly(zfsvfs) && (*flags & SB_RDONLY)) |
47ab01a1 TC |
2022 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); |
2023 | ||
0037b49e | 2024 | zfs_unregister_callbacks(zfsvfs); |
1c2555ef BB |
2025 | zfsvfs_vfs_free(zfsvfs->z_vfs); |
2026 | ||
2027 | vfsp->vfs_data = zfsvfs; | |
2028 | zfsvfs->z_vfs = vfsp; | |
08de8c16 | 2029 | if (!issnap) |
2030 | (void) zfs_register_callbacks(vfsp); | |
0282c413 BB |
2031 | |
2032 | return (error); | |
0de19dad | 2033 | } |
0de19dad | 2034 | |
e5c39b95 | 2035 | int |
2cf7f52b | 2036 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 2037 | { |
0037b49e | 2038 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
34dc7c2f BB |
2039 | znode_t *zp; |
2040 | uint64_t object = 0; | |
2041 | uint64_t fid_gen = 0; | |
2042 | uint64_t gen_mask; | |
2043 | uint64_t zp_gen; | |
3558fd73 | 2044 | int i, err; |
34dc7c2f | 2045 | |
3558fd73 | 2046 | *ipp = NULL; |
34dc7c2f | 2047 | |
9b77d1c9 CC |
2048 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { |
2049 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
2050 | ||
2051 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
2052 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
34dc7c2f | 2053 | |
9b77d1c9 CC |
2054 | for (i = 0; i < sizeof (zfid->zf_gen); i++) |
2055 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
2056 | } else { | |
2057 | return (SET_ERROR(EINVAL)); | |
2058 | } | |
2059 | ||
2060 | /* LONG_FID_LEN means snapdirs */ | |
34dc7c2f BB |
2061 | if (fidp->fid_len == LONG_FID_LEN) { |
2062 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
2063 | uint64_t objsetid = 0; | |
2064 | uint64_t setgen = 0; | |
2065 | ||
2066 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
2067 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
2068 | ||
2069 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
2070 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
2071 | ||
9b77d1c9 CC |
2072 | if (objsetid != ZFSCTL_INO_SNAPDIRS - object) { |
2073 | dprintf("snapdir fid: objsetid (%llu) != " | |
2074 | "ZFSCTL_INO_SNAPDIRS (%llu) - object (%llu)\n", | |
2075 | objsetid, ZFSCTL_INO_SNAPDIRS, object); | |
34dc7c2f | 2076 | |
2e528b49 | 2077 | return (SET_ERROR(EINVAL)); |
9b77d1c9 | 2078 | } |
ebe7e575 | 2079 | |
9b77d1c9 CC |
2080 | if (fid_gen > 1 || setgen != 0) { |
2081 | dprintf("snapdir fid: fid_gen (%llu) and setgen " | |
2082 | "(%llu)\n", fid_gen, setgen); | |
2083 | return (SET_ERROR(EINVAL)); | |
2084 | } | |
34dc7c2f | 2085 | |
9b77d1c9 | 2086 | return (zfsctl_snapdir_vget(sb, objsetid, fid_gen, ipp)); |
34dc7c2f BB |
2087 | } |
2088 | ||
0037b49e | 2089 | ZFS_ENTER(zfsvfs); |
34dc7c2f BB |
2090 | /* A zero fid_gen means we are in the .zfs control directories */ |
2091 | if (fid_gen == 0 && | |
2092 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
0037b49e | 2093 | *ipp = zfsvfs->z_ctldir; |
3558fd73 | 2094 | ASSERT(*ipp != NULL); |
34dc7c2f | 2095 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
2096 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
2097 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 2098 | } else { |
3558fd73 | 2099 | igrab(*ipp); |
34dc7c2f | 2100 | } |
0037b49e | 2101 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2102 | return (0); |
2103 | } | |
2104 | ||
2105 | gen_mask = -1ULL >> (64 - 8 * i); | |
2106 | ||
29e57d15 | 2107 | dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); |
0037b49e BB |
2108 | if ((err = zfs_zget(zfsvfs, object, &zp))) { |
2109 | ZFS_EXIT(zfsvfs); | |
34dc7c2f BB |
2110 | return (err); |
2111 | } | |
7938c2ac CC |
2112 | |
2113 | /* Don't export xattr stuff */ | |
2114 | if (zp->z_pflags & ZFS_XATTR) { | |
2115 | iput(ZTOI(zp)); | |
0037b49e | 2116 | ZFS_EXIT(zfsvfs); |
7938c2ac CC |
2117 | return (SET_ERROR(ENOENT)); |
2118 | } | |
2119 | ||
0037b49e | 2120 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen, |
428870ff BB |
2121 | sizeof (uint64_t)); |
2122 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
2123 | if (zp_gen == 0) |
2124 | zp_gen = 1; | |
0037b49e | 2125 | if ((fid_gen == 0) && (zfsvfs->z_root == object)) |
0500e835 | 2126 | fid_gen = zp_gen; |
34dc7c2f | 2127 | if (zp->z_unlinked || zp_gen != fid_gen) { |
29e57d15 NB |
2128 | dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, |
2129 | fid_gen); | |
3558fd73 | 2130 | iput(ZTOI(zp)); |
0037b49e | 2131 | ZFS_EXIT(zfsvfs); |
6c253064 | 2132 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
2133 | } |
2134 | ||
3558fd73 BB |
2135 | *ipp = ZTOI(zp); |
2136 | if (*ipp) | |
2137 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 2138 | |
0037b49e | 2139 | ZFS_EXIT(zfsvfs); |
34dc7c2f BB |
2140 | return (0); |
2141 | } | |
2142 | ||
2143 | /* | |
0037b49e | 2144 | * Block out VFS ops and close zfsvfs_t |
34dc7c2f BB |
2145 | * |
2146 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
831baf06 KW |
2147 | * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying |
2148 | * dataset and objset intact so that they can be atomically handed off during | |
2149 | * a subsequent rollback or recv operation and the resume thereafter. | |
34dc7c2f BB |
2150 | */ |
2151 | int | |
0037b49e | 2152 | zfs_suspend_fs(zfsvfs_t *zfsvfs) |
34dc7c2f BB |
2153 | { |
2154 | int error; | |
2155 | ||
0037b49e | 2156 | if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0) |
34dc7c2f | 2157 | return (error); |
7b3e34ba | 2158 | |
34dc7c2f BB |
2159 | return (0); |
2160 | } | |
2161 | ||
2162 | /* | |
8614ddf9 MA |
2163 | * Rebuild SA and release VOPs. Note that ownership of the underlying dataset |
2164 | * is an invariant across any of the operations that can be performed while the | |
2165 | * filesystem was suspended. Whether it succeeded or failed, the preconditions | |
2166 | * are the same: the relevant objset and associated dataset are owned by | |
2167 | * zfsvfs, held, and long held on entry. | |
34dc7c2f BB |
2168 | */ |
2169 | int | |
0037b49e | 2170 | zfs_resume_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) |
34dc7c2f | 2171 | { |
2008e920 | 2172 | int err, err2; |
831baf06 | 2173 | znode_t *zp; |
34dc7c2f | 2174 | |
0037b49e BB |
2175 | ASSERT(RRM_WRITE_HELD(&zfsvfs->z_teardown_lock)); |
2176 | ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); | |
34dc7c2f | 2177 | |
831baf06 | 2178 | /* |
ec923db2 GM |
2179 | * We already own this, so just update the objset_t, as the one we |
2180 | * had before may have been evicted. | |
831baf06 | 2181 | */ |
8614ddf9 | 2182 | objset_t *os; |
0037b49e | 2183 | VERIFY3P(ds->ds_owner, ==, zfsvfs); |
ec923db2 | 2184 | VERIFY(dsl_dataset_long_held(ds)); |
8614ddf9 | 2185 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
c1fabe79 | 2186 | |
8614ddf9 MA |
2187 | err = zfsvfs_init(zfsvfs, os); |
2188 | if (err != 0) | |
831baf06 | 2189 | goto bail; |
34dc7c2f | 2190 | |
f298b24d | 2191 | VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); |
c1fabe79 | 2192 | |
0037b49e BB |
2193 | zfs_set_fuid_feature(zfsvfs); |
2194 | zfsvfs->z_rollback_time = jiffies; | |
34dc7c2f | 2195 | |
831baf06 KW |
2196 | /* |
2197 | * Attempt to re-establish all the active inodes with their | |
2198 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
2199 | * and mark it stale. This prevents a collision if a new | |
2200 | * inode/object is created which must use the same inode | |
2201 | * number. The stale inode will be be released when the | |
2202 | * VFS prunes the dentry holding the remaining references | |
2203 | * on the stale inode. | |
2204 | */ | |
0037b49e BB |
2205 | mutex_enter(&zfsvfs->z_znodes_lock); |
2206 | for (zp = list_head(&zfsvfs->z_all_znodes); zp; | |
2207 | zp = list_next(&zfsvfs->z_all_znodes, zp)) { | |
2008e920 TC |
2208 | err2 = zfs_rezget(zp); |
2209 | if (err2) { | |
831baf06 KW |
2210 | remove_inode_hash(ZTOI(zp)); |
2211 | zp->z_is_stale = B_TRUE; | |
34dc7c2f | 2212 | } |
e7a2fa70 TC |
2213 | |
2214 | /* see comment in zfs_suspend_fs() */ | |
2215 | if (zp->z_suspended) { | |
2216 | zfs_iput_async(ZTOI(zp)); | |
2217 | zp->z_suspended = B_FALSE; | |
2218 | } | |
34dc7c2f | 2219 | } |
0037b49e | 2220 | mutex_exit(&zfsvfs->z_znodes_lock); |
34dc7c2f | 2221 | |
dcec0a12 AP |
2222 | if (!zfs_is_readonly(zfsvfs) && !zfsvfs->z_unmounted) { |
2223 | /* | |
2224 | * zfs_suspend_fs() could have interrupted freeing | |
2225 | * of dnodes. We need to restart this freeing so | |
2226 | * that we don't "leak" the space. | |
2227 | */ | |
2228 | zfs_unlinked_drain(zfsvfs); | |
2229 | } | |
2230 | ||
428870ff | 2231 | bail: |
30af21b0 PD |
2232 | if (err != 0) |
2233 | zfsvfs->z_unmounted = B_TRUE; | |
2234 | ||
7b3e34ba | 2235 | /* release the VFS ops */ |
0037b49e BB |
2236 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
2237 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
34dc7c2f | 2238 | |
30af21b0 | 2239 | if (err != 0) { |
34dc7c2f | 2240 | /* |
831baf06 KW |
2241 | * Since we couldn't setup the sa framework, try to force |
2242 | * unmount this file system. | |
34dc7c2f | 2243 | */ |
0037b49e BB |
2244 | if (zfsvfs->z_os) |
2245 | (void) zfs_umount(zfsvfs->z_sb); | |
34dc7c2f BB |
2246 | } |
2247 | return (err); | |
2248 | } | |
2249 | ||
30af21b0 PD |
2250 | /* |
2251 | * Release VOPs and unmount a suspended filesystem. | |
2252 | */ | |
2253 | int | |
2254 | zfs_end_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) | |
2255 | { | |
2256 | ASSERT(RRM_WRITE_HELD(&zfsvfs->z_teardown_lock)); | |
2257 | ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); | |
2258 | ||
2259 | /* | |
2260 | * We already own this, so just hold and rele it to update the | |
2261 | * objset_t, as the one we had before may have been evicted. | |
2262 | */ | |
2263 | objset_t *os; | |
2264 | VERIFY3P(ds->ds_owner, ==, zfsvfs); | |
2265 | VERIFY(dsl_dataset_long_held(ds)); | |
2266 | VERIFY0(dmu_objset_from_ds(ds, &os)); | |
2267 | zfsvfs->z_os = os; | |
2268 | ||
2269 | /* release the VOPs */ | |
2270 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
2271 | rrm_exit(&zfsvfs->z_teardown_lock, FTAG); | |
2272 | ||
2273 | /* | |
2274 | * Try to force unmount this file system. | |
2275 | */ | |
2276 | (void) zfs_umount(zfsvfs->z_sb); | |
2277 | zfsvfs->z_unmounted = B_TRUE; | |
2278 | return (0); | |
2279 | } | |
2280 | ||
34dc7c2f | 2281 | int |
0037b49e | 2282 | zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) |
34dc7c2f BB |
2283 | { |
2284 | int error; | |
0037b49e | 2285 | objset_t *os = zfsvfs->z_os; |
34dc7c2f | 2286 | dmu_tx_t *tx; |
34dc7c2f BB |
2287 | |
2288 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
2e528b49 | 2289 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 2290 | |
0037b49e | 2291 | if (newvers < zfsvfs->z_version) |
2e528b49 | 2292 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 2293 | |
428870ff | 2294 | if (zfs_spa_version_map(newvers) > |
0037b49e | 2295 | spa_version(dmu_objset_spa(zfsvfs->z_os))) |
2e528b49 | 2296 | return (SET_ERROR(ENOTSUP)); |
428870ff | 2297 | |
34dc7c2f | 2298 | tx = dmu_tx_create(os); |
9babb374 | 2299 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
0037b49e | 2300 | if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { |
428870ff BB |
2301 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
2302 | ZFS_SA_ATTRS); | |
2303 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
2304 | } | |
34dc7c2f BB |
2305 | error = dmu_tx_assign(tx, TXG_WAIT); |
2306 | if (error) { | |
2307 | dmu_tx_abort(tx); | |
9babb374 BB |
2308 | return (error); |
2309 | } | |
428870ff | 2310 | |
9babb374 BB |
2311 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
2312 | 8, 1, &newvers, tx); | |
2313 | ||
2314 | if (error) { | |
2315 | dmu_tx_commit(tx); | |
2316 | return (error); | |
34dc7c2f | 2317 | } |
34dc7c2f | 2318 | |
0037b49e | 2319 | if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { |
428870ff BB |
2320 | uint64_t sa_obj; |
2321 | ||
0037b49e | 2322 | ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=, |
428870ff BB |
2323 | SPA_VERSION_SA); |
2324 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
2325 | DMU_OT_NONE, 0, tx); | |
2326 | ||
2327 | error = zap_add(os, MASTER_NODE_OBJ, | |
2328 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c99c9001 | 2329 | ASSERT0(error); |
428870ff BB |
2330 | |
2331 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
2332 | sa_register_update_callback(os, zfs_sa_upgrade); | |
2333 | } | |
2334 | ||
6f1ffb06 | 2335 | spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, |
0037b49e | 2336 | "from %llu to %llu", zfsvfs->z_version, newvers); |
9babb374 | 2337 | |
34dc7c2f BB |
2338 | dmu_tx_commit(tx); |
2339 | ||
0037b49e | 2340 | zfsvfs->z_version = newvers; |
2e5dc449 | 2341 | os->os_version = newvers; |
9babb374 | 2342 | |
0037b49e | 2343 | zfs_set_fuid_feature(zfsvfs); |
9babb374 BB |
2344 | |
2345 | return (0); | |
34dc7c2f BB |
2346 | } |
2347 | ||
2348 | /* | |
2349 | * Read a property stored within the master node. | |
2350 | */ | |
2351 | int | |
2352 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
2353 | { | |
2e5dc449 MA |
2354 | uint64_t *cached_copy = NULL; |
2355 | ||
2356 | /* | |
2357 | * Figure out where in the objset_t the cached copy would live, if it | |
2358 | * is available for the requested property. | |
2359 | */ | |
2360 | if (os != NULL) { | |
2361 | switch (prop) { | |
2362 | case ZFS_PROP_VERSION: | |
2363 | cached_copy = &os->os_version; | |
2364 | break; | |
2365 | case ZFS_PROP_NORMALIZE: | |
2366 | cached_copy = &os->os_normalization; | |
2367 | break; | |
2368 | case ZFS_PROP_UTF8ONLY: | |
2369 | cached_copy = &os->os_utf8only; | |
2370 | break; | |
2371 | case ZFS_PROP_CASE: | |
2372 | cached_copy = &os->os_casesensitivity; | |
2373 | break; | |
2374 | default: | |
2375 | break; | |
2376 | } | |
2377 | } | |
2378 | if (cached_copy != NULL && *cached_copy != OBJSET_PROP_UNINITIALIZED) { | |
2379 | *value = *cached_copy; | |
2380 | return (0); | |
2381 | } | |
34dc7c2f BB |
2382 | |
2383 | /* | |
2e5dc449 MA |
2384 | * If the property wasn't cached, look up the file system's value for |
2385 | * the property. For the version property, we look up a slightly | |
2386 | * different string. | |
34dc7c2f | 2387 | */ |
2e5dc449 MA |
2388 | const char *pname; |
2389 | int error = ENOENT; | |
34dc7c2f BB |
2390 | if (prop == ZFS_PROP_VERSION) |
2391 | pname = ZPL_VERSION_STR; | |
2392 | else | |
2393 | pname = zfs_prop_to_name(prop); | |
2394 | ||
87a275d9 AG |
2395 | if (os != NULL) { |
2396 | ASSERT3U(os->os_phys->os_type, ==, DMU_OST_ZFS); | |
b128c09f | 2397 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); |
87a275d9 | 2398 | } |
34dc7c2f BB |
2399 | |
2400 | if (error == ENOENT) { | |
2401 | /* No value set, use the default value */ | |
2402 | switch (prop) { | |
2403 | case ZFS_PROP_VERSION: | |
2404 | *value = ZPL_VERSION; | |
2405 | break; | |
2406 | case ZFS_PROP_NORMALIZE: | |
2407 | case ZFS_PROP_UTF8ONLY: | |
2408 | *value = 0; | |
2409 | break; | |
2410 | case ZFS_PROP_CASE: | |
2411 | *value = ZFS_CASE_SENSITIVE; | |
2412 | break; | |
023699cd MM |
2413 | case ZFS_PROP_ACLTYPE: |
2414 | *value = ZFS_ACLTYPE_OFF; | |
2415 | break; | |
34dc7c2f BB |
2416 | default: |
2417 | return (error); | |
2418 | } | |
2419 | error = 0; | |
2420 | } | |
2e5dc449 MA |
2421 | |
2422 | /* | |
2423 | * If one of the methods for getting the property value above worked, | |
2424 | * copy it into the objset_t's cache. | |
2425 | */ | |
2426 | if (error == 0 && cached_copy != NULL) { | |
2427 | *cached_copy = *value; | |
2428 | } | |
2429 | ||
34dc7c2f BB |
2430 | return (error); |
2431 | } | |
3558fd73 | 2432 | |
a08abc1b | 2433 | /* |
e1cfd73f | 2434 | * Return true if the corresponding vfs's unmounted flag is set. |
a08abc1b GM |
2435 | * Otherwise return false. |
2436 | * If this function returns true we know VFS unmount has been initiated. | |
2437 | */ | |
2438 | boolean_t | |
2439 | zfs_get_vfs_flag_unmounted(objset_t *os) | |
2440 | { | |
0037b49e | 2441 | zfsvfs_t *zfvp; |
a08abc1b GM |
2442 | boolean_t unmounted = B_FALSE; |
2443 | ||
2444 | ASSERT(dmu_objset_type(os) == DMU_OST_ZFS); | |
2445 | ||
2446 | mutex_enter(&os->os_user_ptr_lock); | |
2447 | zfvp = dmu_objset_get_user(os); | |
2448 | if (zfvp != NULL && zfvp->z_unmounted) | |
2449 | unmounted = B_TRUE; | |
2450 | mutex_exit(&os->os_user_ptr_lock); | |
2451 | ||
2452 | return (unmounted); | |
2453 | } | |
2454 | ||
3558fd73 BB |
2455 | void |
2456 | zfs_init(void) | |
2457 | { | |
ebe7e575 | 2458 | zfsctl_init(); |
3558fd73 BB |
2459 | zfs_znode_init(); |
2460 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
2461 | register_filesystem(&zpl_fs_type); | |
2462 | } | |
2463 | ||
2464 | void | |
2465 | zfs_fini(void) | |
2466 | { | |
8e71ab99 CC |
2467 | /* |
2468 | * we don't use outstanding because zpl_posix_acl_free might add more. | |
2469 | */ | |
57ddcda1 | 2470 | taskq_wait(system_delay_taskq); |
8e71ab99 | 2471 | taskq_wait(system_taskq); |
3558fd73 BB |
2472 | unregister_filesystem(&zpl_fs_type); |
2473 | zfs_znode_fini(); | |
ebe7e575 | 2474 | zfsctl_fini(); |
3558fd73 | 2475 | } |
f298b24d | 2476 | |
93ce2b4c | 2477 | #if defined(_KERNEL) |
f298b24d BB |
2478 | EXPORT_SYMBOL(zfs_suspend_fs); |
2479 | EXPORT_SYMBOL(zfs_resume_fs); | |
2480 | EXPORT_SYMBOL(zfs_userspace_one); | |
2481 | EXPORT_SYMBOL(zfs_userspace_many); | |
2482 | EXPORT_SYMBOL(zfs_set_userquota); | |
9c5167d1 NF |
2483 | EXPORT_SYMBOL(zfs_id_overblockquota); |
2484 | EXPORT_SYMBOL(zfs_id_overobjquota); | |
2485 | EXPORT_SYMBOL(zfs_id_overquota); | |
f298b24d BB |
2486 | EXPORT_SYMBOL(zfs_set_version); |
2487 | EXPORT_SYMBOL(zfsvfs_create); | |
2488 | EXPORT_SYMBOL(zfsvfs_free); | |
2489 | EXPORT_SYMBOL(zfs_is_readonly); | |
2490 | EXPORT_SYMBOL(zfs_domount); | |
2491 | EXPORT_SYMBOL(zfs_preumount); | |
2492 | EXPORT_SYMBOL(zfs_umount); | |
2493 | EXPORT_SYMBOL(zfs_remount); | |
2494 | EXPORT_SYMBOL(zfs_statvfs); | |
2495 | EXPORT_SYMBOL(zfs_vget); | |
2496 | EXPORT_SYMBOL(zfs_prune); | |
2497 | #endif |