]>
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 | |
1d3ba0bf | 9 | * or https://opensource.org/licenses/CDDL-1.0. |
34dc7c2f BB |
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> |
957c7aa2 | 55 | #include <sys/zfs_quota.h> |
34dc7c2f | 56 | #include <sys/sunddi.h> |
34dc7c2f | 57 | #include <sys/dmu_objset.h> |
5a42ef04 | 58 | #include <sys/dsl_dir.h> |
30af21b0 | 59 | #include <sys/objlist.h> |
3558fd73 | 60 | #include <sys/zpl.h> |
05805494 | 61 | #include <linux/vfs_compat.h> |
428870ff | 62 | #include "zfs_comutil.h" |
34dc7c2f | 63 | |
1c2555ef BB |
64 | enum { |
65 | TOKEN_RO, | |
66 | TOKEN_RW, | |
67 | TOKEN_SETUID, | |
68 | TOKEN_NOSETUID, | |
69 | TOKEN_EXEC, | |
70 | TOKEN_NOEXEC, | |
71 | TOKEN_DEVICES, | |
72 | TOKEN_NODEVICES, | |
73 | TOKEN_DIRXATTR, | |
74 | TOKEN_SAXATTR, | |
75 | TOKEN_XATTR, | |
76 | TOKEN_NOXATTR, | |
77 | TOKEN_ATIME, | |
78 | TOKEN_NOATIME, | |
79 | TOKEN_RELATIME, | |
80 | TOKEN_NORELATIME, | |
81 | TOKEN_NBMAND, | |
82 | TOKEN_NONBMAND, | |
83 | TOKEN_MNTPOINT, | |
84 | TOKEN_LAST, | |
85 | }; | |
86 | ||
87 | static const match_table_t zpl_tokens = { | |
88 | { TOKEN_RO, MNTOPT_RO }, | |
89 | { TOKEN_RW, MNTOPT_RW }, | |
90 | { TOKEN_SETUID, MNTOPT_SETUID }, | |
91 | { TOKEN_NOSETUID, MNTOPT_NOSETUID }, | |
92 | { TOKEN_EXEC, MNTOPT_EXEC }, | |
93 | { TOKEN_NOEXEC, MNTOPT_NOEXEC }, | |
94 | { TOKEN_DEVICES, MNTOPT_DEVICES }, | |
95 | { TOKEN_NODEVICES, MNTOPT_NODEVICES }, | |
96 | { TOKEN_DIRXATTR, MNTOPT_DIRXATTR }, | |
97 | { TOKEN_SAXATTR, MNTOPT_SAXATTR }, | |
98 | { TOKEN_XATTR, MNTOPT_XATTR }, | |
99 | { TOKEN_NOXATTR, MNTOPT_NOXATTR }, | |
100 | { TOKEN_ATIME, MNTOPT_ATIME }, | |
101 | { TOKEN_NOATIME, MNTOPT_NOATIME }, | |
102 | { TOKEN_RELATIME, MNTOPT_RELATIME }, | |
103 | { TOKEN_NORELATIME, MNTOPT_NORELATIME }, | |
104 | { TOKEN_NBMAND, MNTOPT_NBMAND }, | |
105 | { TOKEN_NONBMAND, MNTOPT_NONBMAND }, | |
106 | { TOKEN_MNTPOINT, MNTOPT_MNTPOINT "=%s" }, | |
107 | { TOKEN_LAST, NULL }, | |
108 | }; | |
109 | ||
110 | static void | |
111 | zfsvfs_vfs_free(vfs_t *vfsp) | |
112 | { | |
113 | if (vfsp != NULL) { | |
114 | if (vfsp->vfs_mntpoint != NULL) | |
e4f5fa12 | 115 | kmem_strfree(vfsp->vfs_mntpoint); |
1c2555ef BB |
116 | |
117 | kmem_free(vfsp, sizeof (vfs_t)); | |
118 | } | |
119 | } | |
120 | ||
121 | static int | |
122 | zfsvfs_parse_option(char *option, int token, substring_t *args, vfs_t *vfsp) | |
123 | { | |
124 | switch (token) { | |
125 | case TOKEN_RO: | |
126 | vfsp->vfs_readonly = B_TRUE; | |
127 | vfsp->vfs_do_readonly = B_TRUE; | |
128 | break; | |
129 | case TOKEN_RW: | |
130 | vfsp->vfs_readonly = B_FALSE; | |
131 | vfsp->vfs_do_readonly = B_TRUE; | |
132 | break; | |
133 | case TOKEN_SETUID: | |
134 | vfsp->vfs_setuid = B_TRUE; | |
135 | vfsp->vfs_do_setuid = B_TRUE; | |
136 | break; | |
137 | case TOKEN_NOSETUID: | |
138 | vfsp->vfs_setuid = B_FALSE; | |
139 | vfsp->vfs_do_setuid = B_TRUE; | |
140 | break; | |
141 | case TOKEN_EXEC: | |
142 | vfsp->vfs_exec = B_TRUE; | |
143 | vfsp->vfs_do_exec = B_TRUE; | |
144 | break; | |
145 | case TOKEN_NOEXEC: | |
146 | vfsp->vfs_exec = B_FALSE; | |
147 | vfsp->vfs_do_exec = B_TRUE; | |
148 | break; | |
149 | case TOKEN_DEVICES: | |
150 | vfsp->vfs_devices = B_TRUE; | |
151 | vfsp->vfs_do_devices = B_TRUE; | |
152 | break; | |
153 | case TOKEN_NODEVICES: | |
154 | vfsp->vfs_devices = B_FALSE; | |
155 | vfsp->vfs_do_devices = B_TRUE; | |
156 | break; | |
157 | case TOKEN_DIRXATTR: | |
158 | vfsp->vfs_xattr = ZFS_XATTR_DIR; | |
159 | vfsp->vfs_do_xattr = B_TRUE; | |
160 | break; | |
161 | case TOKEN_SAXATTR: | |
162 | vfsp->vfs_xattr = ZFS_XATTR_SA; | |
163 | vfsp->vfs_do_xattr = B_TRUE; | |
164 | break; | |
165 | case TOKEN_XATTR: | |
166 | vfsp->vfs_xattr = ZFS_XATTR_DIR; | |
167 | vfsp->vfs_do_xattr = B_TRUE; | |
168 | break; | |
169 | case TOKEN_NOXATTR: | |
170 | vfsp->vfs_xattr = ZFS_XATTR_OFF; | |
171 | vfsp->vfs_do_xattr = B_TRUE; | |
172 | break; | |
173 | case TOKEN_ATIME: | |
174 | vfsp->vfs_atime = B_TRUE; | |
175 | vfsp->vfs_do_atime = B_TRUE; | |
176 | break; | |
177 | case TOKEN_NOATIME: | |
178 | vfsp->vfs_atime = B_FALSE; | |
179 | vfsp->vfs_do_atime = B_TRUE; | |
180 | break; | |
181 | case TOKEN_RELATIME: | |
182 | vfsp->vfs_relatime = B_TRUE; | |
183 | vfsp->vfs_do_relatime = B_TRUE; | |
184 | break; | |
185 | case TOKEN_NORELATIME: | |
186 | vfsp->vfs_relatime = B_FALSE; | |
187 | vfsp->vfs_do_relatime = B_TRUE; | |
188 | break; | |
189 | case TOKEN_NBMAND: | |
190 | vfsp->vfs_nbmand = B_TRUE; | |
191 | vfsp->vfs_do_nbmand = B_TRUE; | |
192 | break; | |
193 | case TOKEN_NONBMAND: | |
194 | vfsp->vfs_nbmand = B_FALSE; | |
195 | vfsp->vfs_do_nbmand = B_TRUE; | |
196 | break; | |
197 | case TOKEN_MNTPOINT: | |
198 | vfsp->vfs_mntpoint = match_strdup(&args[0]); | |
199 | if (vfsp->vfs_mntpoint == NULL) | |
200 | return (SET_ERROR(ENOMEM)); | |
201 | ||
202 | break; | |
203 | default: | |
204 | break; | |
205 | } | |
206 | ||
207 | return (0); | |
208 | } | |
209 | ||
210 | /* | |
211 | * Parse the raw mntopts and return a vfs_t describing the options. | |
212 | */ | |
213 | static int | |
214 | zfsvfs_parse_options(char *mntopts, vfs_t **vfsp) | |
215 | { | |
216 | vfs_t *tmp_vfsp; | |
217 | int error; | |
218 | ||
219 | tmp_vfsp = kmem_zalloc(sizeof (vfs_t), KM_SLEEP); | |
220 | ||
221 | if (mntopts != NULL) { | |
222 | substring_t args[MAX_OPT_ARGS]; | |
223 | char *tmp_mntopts, *p, *t; | |
224 | int token; | |
225 | ||
e4f5fa12 | 226 | tmp_mntopts = t = kmem_strdup(mntopts); |
1c2555ef BB |
227 | if (tmp_mntopts == NULL) |
228 | return (SET_ERROR(ENOMEM)); | |
229 | ||
230 | while ((p = strsep(&t, ",")) != NULL) { | |
231 | if (!*p) | |
232 | continue; | |
233 | ||
234 | args[0].to = args[0].from = NULL; | |
235 | token = match_token(p, zpl_tokens, args); | |
236 | error = zfsvfs_parse_option(p, token, args, tmp_vfsp); | |
237 | if (error) { | |
e4f5fa12 | 238 | kmem_strfree(tmp_mntopts); |
1c2555ef BB |
239 | zfsvfs_vfs_free(tmp_vfsp); |
240 | return (error); | |
241 | } | |
242 | } | |
243 | ||
e4f5fa12 | 244 | kmem_strfree(tmp_mntopts); |
1c2555ef BB |
245 | } |
246 | ||
247 | *vfsp = tmp_vfsp; | |
248 | ||
249 | return (0); | |
250 | } | |
251 | ||
252 | boolean_t | |
253 | zfs_is_readonly(zfsvfs_t *zfsvfs) | |
254 | { | |
05805494 | 255 | return (!!(zfsvfs->z_sb->s_flags & SB_RDONLY)); |
1c2555ef BB |
256 | } |
257 | ||
34dc7c2f | 258 | int |
03f9ba9d | 259 | zfs_sync(struct super_block *sb, int wait, cred_t *cr) |
34dc7c2f | 260 | { |
ef70eff1 | 261 | (void) cr; |
0037b49e | 262 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
03f9ba9d | 263 | |
03f9ba9d BB |
264 | /* |
265 | * Semantically, the only requirement is that the sync be initiated. | |
266 | * The DMU syncs out txgs frequently, so there's nothing to do. | |
267 | */ | |
268 | if (!wait) | |
269 | return (0); | |
270 | ||
0037b49e | 271 | if (zfsvfs != NULL) { |
34dc7c2f BB |
272 | /* |
273 | * Sync a specific filesystem. | |
274 | */ | |
9babb374 | 275 | dsl_pool_t *dp; |
768eaced | 276 | int error; |
34dc7c2f | 277 | |
768eaced CC |
278 | if ((error = zfs_enter(zfsvfs, FTAG)) != 0) |
279 | return (error); | |
0037b49e | 280 | dp = dmu_objset_pool(zfsvfs->z_os); |
9babb374 BB |
281 | |
282 | /* | |
283 | * If the system is shutting down, then skip any | |
284 | * filesystems which may exist on a suspended pool. | |
285 | */ | |
03f9ba9d | 286 | if (spa_suspended(dp->dp_spa)) { |
768eaced | 287 | zfs_exit(zfsvfs, FTAG); |
9babb374 BB |
288 | return (0); |
289 | } | |
290 | ||
0037b49e BB |
291 | if (zfsvfs->z_log != NULL) |
292 | zil_commit(zfsvfs->z_log, 0); | |
428870ff | 293 | |
768eaced | 294 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
295 | } else { |
296 | /* | |
297 | * Sync all ZFS filesystems. This is what happens when you | |
76d04993 | 298 | * run sync(1). Unlike other filesystems, ZFS honors the |
34dc7c2f BB |
299 | * request by waiting for all pools to commit all dirty data. |
300 | */ | |
301 | spa_sync_allpools(); | |
302 | } | |
303 | ||
304 | return (0); | |
305 | } | |
306 | ||
34dc7c2f BB |
307 | static void |
308 | atime_changed_cb(void *arg, uint64_t newval) | |
309 | { | |
9c53e516 TK |
310 | zfsvfs_t *zfsvfs = arg; |
311 | struct super_block *sb = zfsvfs->z_sb; | |
312 | ||
313 | if (sb == NULL) | |
314 | return; | |
315 | /* | |
316 | * Update SB_NOATIME bit in VFS super block. Since atime update is | |
317 | * determined by atime_needs_update(), atime_needs_update() needs to | |
318 | * return false if atime is turned off, and not unconditionally return | |
319 | * false if atime is turned on. | |
320 | */ | |
321 | if (newval) | |
322 | sb->s_flags &= ~SB_NOATIME; | |
323 | else | |
324 | sb->s_flags |= SB_NOATIME; | |
34dc7c2f BB |
325 | } |
326 | ||
6d111134 TC |
327 | static void |
328 | relatime_changed_cb(void *arg, uint64_t newval) | |
329 | { | |
0037b49e | 330 | ((zfsvfs_t *)arg)->z_relatime = newval; |
6d111134 TC |
331 | } |
332 | ||
34dc7c2f BB |
333 | static void |
334 | xattr_changed_cb(void *arg, uint64_t newval) | |
335 | { | |
0037b49e | 336 | zfsvfs_t *zfsvfs = arg; |
34dc7c2f | 337 | |
82a37189 | 338 | if (newval == ZFS_XATTR_OFF) { |
0037b49e | 339 | zfsvfs->z_flags &= ~ZSB_XATTR; |
82a37189 | 340 | } else { |
0037b49e | 341 | zfsvfs->z_flags |= ZSB_XATTR; |
82a37189 BB |
342 | |
343 | if (newval == ZFS_XATTR_SA) | |
0037b49e | 344 | zfsvfs->z_xattr_sa = B_TRUE; |
82a37189 | 345 | else |
0037b49e | 346 | zfsvfs->z_xattr_sa = B_FALSE; |
82a37189 | 347 | } |
34dc7c2f BB |
348 | } |
349 | ||
023699cd MM |
350 | static void |
351 | acltype_changed_cb(void *arg, uint64_t newval) | |
352 | { | |
0037b49e | 353 | zfsvfs_t *zfsvfs = arg; |
023699cd MM |
354 | |
355 | switch (newval) { | |
485b50bb | 356 | case ZFS_ACLTYPE_NFSV4: |
023699cd | 357 | case ZFS_ACLTYPE_OFF: |
0037b49e | 358 | zfsvfs->z_acl_type = ZFS_ACLTYPE_OFF; |
05805494 | 359 | zfsvfs->z_sb->s_flags &= ~SB_POSIXACL; |
023699cd | 360 | break; |
7ead2be3 | 361 | case ZFS_ACLTYPE_POSIX: |
b695c34e | 362 | #ifdef CONFIG_FS_POSIX_ACL |
7ead2be3 | 363 | zfsvfs->z_acl_type = ZFS_ACLTYPE_POSIX; |
05805494 | 364 | zfsvfs->z_sb->s_flags |= SB_POSIXACL; |
b695c34e | 365 | #else |
0037b49e | 366 | zfsvfs->z_acl_type = ZFS_ACLTYPE_OFF; |
05805494 | 367 | zfsvfs->z_sb->s_flags &= ~SB_POSIXACL; |
b695c34e | 368 | #endif /* CONFIG_FS_POSIX_ACL */ |
023699cd MM |
369 | break; |
370 | default: | |
371 | break; | |
372 | } | |
373 | } | |
374 | ||
34dc7c2f BB |
375 | static void |
376 | blksz_changed_cb(void *arg, uint64_t newval) | |
377 | { | |
0037b49e BB |
378 | zfsvfs_t *zfsvfs = arg; |
379 | ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zfsvfs->z_os))); | |
f1512ee6 MA |
380 | ASSERT3U(newval, >=, SPA_MINBLOCKSIZE); |
381 | ASSERT(ISP2(newval)); | |
34dc7c2f | 382 | |
0037b49e | 383 | zfsvfs->z_max_blksz = newval; |
34dc7c2f BB |
384 | } |
385 | ||
386 | static void | |
387 | readonly_changed_cb(void *arg, uint64_t newval) | |
388 | { | |
0037b49e BB |
389 | zfsvfs_t *zfsvfs = arg; |
390 | struct super_block *sb = zfsvfs->z_sb; | |
34dc7c2f | 391 | |
2cf7f52b BB |
392 | if (sb == NULL) |
393 | return; | |
394 | ||
395 | if (newval) | |
05805494 | 396 | sb->s_flags |= SB_RDONLY; |
2cf7f52b | 397 | else |
05805494 | 398 | sb->s_flags &= ~SB_RDONLY; |
34dc7c2f BB |
399 | } |
400 | ||
401 | static void | |
402 | devices_changed_cb(void *arg, uint64_t newval) | |
403 | { | |
34dc7c2f BB |
404 | } |
405 | ||
406 | static void | |
407 | setuid_changed_cb(void *arg, uint64_t newval) | |
408 | { | |
34dc7c2f BB |
409 | } |
410 | ||
411 | static void | |
412 | exec_changed_cb(void *arg, uint64_t newval) | |
413 | { | |
34dc7c2f BB |
414 | } |
415 | ||
34dc7c2f BB |
416 | static void |
417 | nbmand_changed_cb(void *arg, uint64_t newval) | |
418 | { | |
0037b49e BB |
419 | zfsvfs_t *zfsvfs = arg; |
420 | struct super_block *sb = zfsvfs->z_sb; | |
3558fd73 | 421 | |
2cf7f52b BB |
422 | if (sb == NULL) |
423 | return; | |
424 | ||
425 | if (newval == TRUE) | |
05805494 | 426 | sb->s_flags |= SB_MANDLOCK; |
2cf7f52b | 427 | else |
05805494 | 428 | sb->s_flags &= ~SB_MANDLOCK; |
34dc7c2f BB |
429 | } |
430 | ||
431 | static void | |
432 | snapdir_changed_cb(void *arg, uint64_t newval) | |
433 | { | |
0037b49e | 434 | ((zfsvfs_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
435 | } |
436 | ||
a1af567b PH |
437 | static void |
438 | acl_mode_changed_cb(void *arg, uint64_t newval) | |
439 | { | |
440 | zfsvfs_t *zfsvfs = arg; | |
441 | ||
442 | zfsvfs->z_acl_mode = newval; | |
443 | } | |
444 | ||
34dc7c2f BB |
445 | static void |
446 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
447 | { | |
0037b49e | 448 | ((zfsvfs_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
449 | } |
450 | ||
1c2555ef BB |
451 | static int |
452 | zfs_register_callbacks(vfs_t *vfsp) | |
34dc7c2f BB |
453 | { |
454 | struct dsl_dataset *ds = NULL; | |
1c2555ef BB |
455 | objset_t *os = NULL; |
456 | zfsvfs_t *zfsvfs = NULL; | |
34dc7c2f BB |
457 | int error = 0; |
458 | ||
1c2555ef BB |
459 | ASSERT(vfsp); |
460 | zfsvfs = vfsp->vfs_data; | |
0037b49e | 461 | ASSERT(zfsvfs); |
1c2555ef | 462 | os = zfsvfs->z_os; |
0282c413 BB |
463 | |
464 | /* | |
465 | * The act of registering our callbacks will destroy any mount | |
466 | * options we may have. In order to enable temporary overrides | |
467 | * of mount options, we stash away the current values and | |
468 | * restore them after we register the callbacks. | |
469 | */ | |
0037b49e | 470 | if (zfs_is_readonly(zfsvfs) || !spa_writeable(dmu_objset_spa(os))) { |
1c2555ef BB |
471 | vfsp->vfs_do_readonly = B_TRUE; |
472 | vfsp->vfs_readonly = B_TRUE; | |
0282c413 | 473 | } |
34dc7c2f BB |
474 | |
475 | /* | |
476 | * Register property callbacks. | |
477 | * | |
478 | * It would probably be fine to just check for i/o error from | |
479 | * the first prop_register(), but I guess I like to go | |
480 | * overboard... | |
481 | */ | |
482 | ds = dmu_objset_ds(os); | |
13fe0198 | 483 | dsl_pool_config_enter(dmu_objset_pool(os), FTAG); |
3558fd73 | 484 | error = dsl_prop_register(ds, |
0037b49e | 485 | zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zfsvfs); |
98fad862 | 486 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 487 | zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zfsvfs); |
34dc7c2f | 488 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 489 | zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zfsvfs); |
34dc7c2f | 490 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 491 | zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zfsvfs); |
34dc7c2f | 492 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 493 | zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zfsvfs); |
34dc7c2f | 494 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 495 | zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zfsvfs); |
34dc7c2f | 496 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 497 | zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zfsvfs); |
34dc7c2f | 498 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 499 | zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zfsvfs); |
34dc7c2f | 500 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 501 | zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zfsvfs); |
34dc7c2f | 502 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 503 | zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zfsvfs); |
a1af567b PH |
504 | error = error ? error : dsl_prop_register(ds, |
505 | zfs_prop_to_name(ZFS_PROP_ACLMODE), acl_mode_changed_cb, zfsvfs); | |
023699cd | 506 | error = error ? error : dsl_prop_register(ds, |
0037b49e BB |
507 | zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, |
508 | zfsvfs); | |
2cf7f52b | 509 | error = error ? error : dsl_prop_register(ds, |
0037b49e | 510 | zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zfsvfs); |
13fe0198 | 511 | dsl_pool_config_exit(dmu_objset_pool(os), FTAG); |
34dc7c2f BB |
512 | if (error) |
513 | goto unregister; | |
514 | ||
0282c413 BB |
515 | /* |
516 | * Invoke our callbacks to restore temporary mount options. | |
517 | */ | |
1c2555ef BB |
518 | if (vfsp->vfs_do_readonly) |
519 | readonly_changed_cb(zfsvfs, vfsp->vfs_readonly); | |
520 | if (vfsp->vfs_do_setuid) | |
521 | setuid_changed_cb(zfsvfs, vfsp->vfs_setuid); | |
522 | if (vfsp->vfs_do_exec) | |
523 | exec_changed_cb(zfsvfs, vfsp->vfs_exec); | |
524 | if (vfsp->vfs_do_devices) | |
525 | devices_changed_cb(zfsvfs, vfsp->vfs_devices); | |
526 | if (vfsp->vfs_do_xattr) | |
527 | xattr_changed_cb(zfsvfs, vfsp->vfs_xattr); | |
528 | if (vfsp->vfs_do_atime) | |
529 | atime_changed_cb(zfsvfs, vfsp->vfs_atime); | |
530 | if (vfsp->vfs_do_relatime) | |
531 | relatime_changed_cb(zfsvfs, vfsp->vfs_relatime); | |
532 | if (vfsp->vfs_do_nbmand) | |
533 | nbmand_changed_cb(zfsvfs, vfsp->vfs_nbmand); | |
c9ada6d5 | 534 | |
34dc7c2f BB |
535 | return (0); |
536 | ||
537 | unregister: | |
0037b49e | 538 | dsl_prop_unregister_all(ds, zfsvfs); |
3558fd73 | 539 | return (error); |
34dc7c2f BB |
540 | } |
541 | ||
2516a878 MM |
542 | /* |
543 | * Takes a dataset, a property, a value and that value's setpoint as | |
544 | * found in the ZAP. Checks if the property has been changed in the vfs. | |
545 | * If so, val and setpoint will be overwritten with updated content. | |
546 | * Otherwise, they are left unchanged. | |
547 | */ | |
548 | int | |
549 | zfs_get_temporary_prop(dsl_dataset_t *ds, zfs_prop_t zfs_prop, uint64_t *val, | |
550 | char *setpoint) | |
551 | { | |
552 | int error; | |
553 | zfsvfs_t *zfvp; | |
554 | vfs_t *vfsp; | |
555 | objset_t *os; | |
556 | uint64_t tmp = *val; | |
557 | ||
558 | error = dmu_objset_from_ds(ds, &os); | |
559 | if (error != 0) | |
560 | return (error); | |
561 | ||
562 | if (dmu_objset_type(os) != DMU_OST_ZFS) | |
563 | return (EINVAL); | |
564 | ||
565 | mutex_enter(&os->os_user_ptr_lock); | |
566 | zfvp = dmu_objset_get_user(os); | |
567 | mutex_exit(&os->os_user_ptr_lock); | |
568 | if (zfvp == NULL) | |
569 | return (ESRCH); | |
570 | ||
571 | vfsp = zfvp->z_vfs; | |
572 | ||
573 | switch (zfs_prop) { | |
574 | case ZFS_PROP_ATIME: | |
575 | if (vfsp->vfs_do_atime) | |
576 | tmp = vfsp->vfs_atime; | |
577 | break; | |
578 | case ZFS_PROP_RELATIME: | |
579 | if (vfsp->vfs_do_relatime) | |
580 | tmp = vfsp->vfs_relatime; | |
581 | break; | |
582 | case ZFS_PROP_DEVICES: | |
583 | if (vfsp->vfs_do_devices) | |
584 | tmp = vfsp->vfs_devices; | |
585 | break; | |
586 | case ZFS_PROP_EXEC: | |
587 | if (vfsp->vfs_do_exec) | |
588 | tmp = vfsp->vfs_exec; | |
589 | break; | |
590 | case ZFS_PROP_SETUID: | |
591 | if (vfsp->vfs_do_setuid) | |
592 | tmp = vfsp->vfs_setuid; | |
593 | break; | |
594 | case ZFS_PROP_READONLY: | |
595 | if (vfsp->vfs_do_readonly) | |
596 | tmp = vfsp->vfs_readonly; | |
597 | break; | |
598 | case ZFS_PROP_XATTR: | |
599 | if (vfsp->vfs_do_xattr) | |
600 | tmp = vfsp->vfs_xattr; | |
601 | break; | |
602 | case ZFS_PROP_NBMAND: | |
603 | if (vfsp->vfs_do_nbmand) | |
604 | tmp = vfsp->vfs_nbmand; | |
605 | break; | |
606 | default: | |
607 | return (ENOENT); | |
608 | } | |
609 | ||
610 | if (tmp != *val) { | |
3a7d2a0c RY |
611 | if (setpoint) |
612 | (void) strcpy(setpoint, "temporary"); | |
2516a878 MM |
613 | *val = tmp; |
614 | } | |
615 | return (0); | |
616 | } | |
617 | ||
8614ddf9 MA |
618 | /* |
619 | * Associate this zfsvfs with the given objset, which must be owned. | |
620 | * This will cache a bunch of on-disk state from the objset in the | |
621 | * zfsvfs. | |
622 | */ | |
623 | static int | |
624 | zfsvfs_init(zfsvfs_t *zfsvfs, objset_t *os) | |
9babb374 | 625 | { |
8614ddf9 MA |
626 | int error; |
627 | uint64_t val; | |
0282c413 | 628 | |
0037b49e BB |
629 | zfsvfs->z_max_blksz = SPA_OLD_MAXBLOCKSIZE; |
630 | zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; | |
631 | zfsvfs->z_os = os; | |
9babb374 | 632 | |
0037b49e | 633 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); |
8614ddf9 MA |
634 | if (error != 0) |
635 | return (error); | |
636 | if (zfsvfs->z_version > | |
637 | zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { | |
638 | (void) printk("Can't mount a version %lld file system " | |
639 | "on a version %lld pool\n. Pool must be upgraded to mount " | |
b689de85 | 640 | "this file system.\n", (u_longlong_t)zfsvfs->z_version, |
8614ddf9 MA |
641 | (u_longlong_t)spa_version(dmu_objset_spa(os))); |
642 | return (SET_ERROR(ENOTSUP)); | |
9babb374 | 643 | } |
8614ddf9 MA |
644 | error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &val); |
645 | if (error != 0) | |
646 | return (error); | |
647 | zfsvfs->z_norm = (int)val; | |
9babb374 | 648 | |
8614ddf9 MA |
649 | error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &val); |
650 | if (error != 0) | |
651 | return (error); | |
652 | zfsvfs->z_utf8 = (val != 0); | |
9babb374 | 653 | |
8614ddf9 MA |
654 | error = zfs_get_zplprop(os, ZFS_PROP_CASE, &val); |
655 | if (error != 0) | |
656 | return (error); | |
657 | zfsvfs->z_case = (uint_t)val; | |
9babb374 | 658 | |
8614ddf9 MA |
659 | if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &val)) != 0) |
660 | return (error); | |
661 | zfsvfs->z_acl_type = (uint_t)val; | |
023699cd | 662 | |
9babb374 BB |
663 | /* |
664 | * Fold case on file systems that are always or sometimes case | |
665 | * insensitive. | |
666 | */ | |
0037b49e BB |
667 | if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || |
668 | zfsvfs->z_case == ZFS_CASE_MIXED) | |
669 | zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 670 | |
0037b49e BB |
671 | zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); |
672 | zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); | |
428870ff | 673 | |
8614ddf9 | 674 | uint64_t sa_obj = 0; |
0037b49e | 675 | if (zfsvfs->z_use_sa) { |
428870ff BB |
676 | /* should either have both of these objects or none */ |
677 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
678 | &sa_obj); | |
8614ddf9 MA |
679 | if (error != 0) |
680 | return (error); | |
82a37189 | 681 | |
8614ddf9 MA |
682 | error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &val); |
683 | if ((error == 0) && (val == ZFS_XATTR_SA)) | |
0037b49e | 684 | zfsvfs->z_xattr_sa = B_TRUE; |
428870ff BB |
685 | } |
686 | ||
9babb374 | 687 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, |
0037b49e | 688 | &zfsvfs->z_root); |
8614ddf9 MA |
689 | if (error != 0) |
690 | return (error); | |
0037b49e | 691 | ASSERT(zfsvfs->z_root != 0); |
9babb374 BB |
692 | |
693 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
0037b49e | 694 | &zfsvfs->z_unlinkedobj); |
8614ddf9 MA |
695 | if (error != 0) |
696 | return (error); | |
9babb374 BB |
697 | |
698 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
699 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
0037b49e | 700 | 8, 1, &zfsvfs->z_userquota_obj); |
8614ddf9 MA |
701 | if (error == ENOENT) |
702 | zfsvfs->z_userquota_obj = 0; | |
703 | else if (error != 0) | |
704 | return (error); | |
9babb374 BB |
705 | |
706 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
707 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
0037b49e | 708 | 8, 1, &zfsvfs->z_groupquota_obj); |
8614ddf9 MA |
709 | if (error == ENOENT) |
710 | zfsvfs->z_groupquota_obj = 0; | |
711 | else if (error != 0) | |
712 | return (error); | |
9babb374 | 713 | |
9c5167d1 NF |
714 | error = zap_lookup(os, MASTER_NODE_OBJ, |
715 | zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTQUOTA], | |
716 | 8, 1, &zfsvfs->z_projectquota_obj); | |
717 | if (error == ENOENT) | |
718 | zfsvfs->z_projectquota_obj = 0; | |
719 | else if (error != 0) | |
720 | return (error); | |
721 | ||
1de321e6 JX |
722 | error = zap_lookup(os, MASTER_NODE_OBJ, |
723 | zfs_userquota_prop_prefixes[ZFS_PROP_USEROBJQUOTA], | |
0037b49e | 724 | 8, 1, &zfsvfs->z_userobjquota_obj); |
8614ddf9 MA |
725 | if (error == ENOENT) |
726 | zfsvfs->z_userobjquota_obj = 0; | |
727 | else if (error != 0) | |
728 | return (error); | |
1de321e6 JX |
729 | |
730 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
731 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPOBJQUOTA], | |
0037b49e | 732 | 8, 1, &zfsvfs->z_groupobjquota_obj); |
8614ddf9 MA |
733 | if (error == ENOENT) |
734 | zfsvfs->z_groupobjquota_obj = 0; | |
735 | else if (error != 0) | |
736 | return (error); | |
1de321e6 | 737 | |
9c5167d1 NF |
738 | error = zap_lookup(os, MASTER_NODE_OBJ, |
739 | zfs_userquota_prop_prefixes[ZFS_PROP_PROJECTOBJQUOTA], | |
740 | 8, 1, &zfsvfs->z_projectobjquota_obj); | |
741 | if (error == ENOENT) | |
742 | zfsvfs->z_projectobjquota_obj = 0; | |
743 | else if (error != 0) | |
744 | return (error); | |
745 | ||
9babb374 | 746 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, |
0037b49e | 747 | &zfsvfs->z_fuid_obj); |
8614ddf9 MA |
748 | if (error == ENOENT) |
749 | zfsvfs->z_fuid_obj = 0; | |
750 | else if (error != 0) | |
751 | return (error); | |
9babb374 BB |
752 | |
753 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
0037b49e | 754 | &zfsvfs->z_shares_dir); |
8614ddf9 MA |
755 | if (error == ENOENT) |
756 | zfsvfs->z_shares_dir = 0; | |
757 | else if (error != 0) | |
758 | return (error); | |
759 | ||
04a3b079 TC |
760 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
761 | &zfsvfs->z_attr_table); | |
762 | if (error != 0) | |
763 | return (error); | |
764 | ||
765 | if (zfsvfs->z_version >= ZPL_VERSION_SA) | |
766 | sa_register_update_callback(os, zfs_sa_upgrade); | |
767 | ||
8614ddf9 MA |
768 | return (0); |
769 | } | |
770 | ||
771 | int | |
163a8c28 | 772 | zfsvfs_create(const char *osname, boolean_t readonly, zfsvfs_t **zfvp) |
8614ddf9 MA |
773 | { |
774 | objset_t *os; | |
775 | zfsvfs_t *zfsvfs; | |
776 | int error; | |
163a8c28 | 777 | boolean_t ro = (readonly || (strchr(osname, '@') != NULL)); |
8614ddf9 MA |
778 | |
779 | zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); | |
780 | ||
163a8c28 | 781 | error = dmu_objset_own(osname, DMU_OST_ZFS, ro, B_TRUE, zfsvfs, &os); |
d99a0153 | 782 | if (error != 0) { |
8614ddf9 MA |
783 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); |
784 | return (error); | |
785 | } | |
786 | ||
d99a0153 | 787 | error = zfsvfs_create_impl(zfvp, zfsvfs, os); |
891ac937 | 788 | |
d99a0153 CW |
789 | return (error); |
790 | } | |
791 | ||
04a3b079 TC |
792 | |
793 | /* | |
794 | * Note: zfsvfs is assumed to be malloc'd, and will be freed by this function | |
795 | * on a failure. Do not pass in a statically allocated zfsvfs. | |
796 | */ | |
d99a0153 CW |
797 | int |
798 | zfsvfs_create_impl(zfsvfs_t **zfvp, zfsvfs_t *zfsvfs, objset_t *os) | |
799 | { | |
800 | int error; | |
801 | ||
8614ddf9 MA |
802 | zfsvfs->z_vfs = NULL; |
803 | zfsvfs->z_sb = NULL; | |
804 | zfsvfs->z_parent = zfsvfs; | |
9babb374 | 805 | |
0037b49e BB |
806 | mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
807 | mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
808 | list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), | |
9babb374 | 809 | offsetof(znode_t, z_link_node)); |
5ebe425a | 810 | ZFS_TEARDOWN_INIT(zfsvfs); |
0037b49e BB |
811 | rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); |
812 | rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
d07b7c7f | 813 | |
8614ddf9 MA |
814 | int size = MIN(1 << (highbit64(zfs_object_mutex_size) - 1), |
815 | ZFS_OBJ_MTX_MAX); | |
0037b49e BB |
816 | zfsvfs->z_hold_size = size; |
817 | zfsvfs->z_hold_trees = vmem_zalloc(sizeof (avl_tree_t) * size, | |
818 | KM_SLEEP); | |
819 | zfsvfs->z_hold_locks = vmem_zalloc(sizeof (kmutex_t) * size, KM_SLEEP); | |
8614ddf9 | 820 | for (int i = 0; i != size; i++) { |
0037b49e | 821 | avl_create(&zfsvfs->z_hold_trees[i], zfs_znode_hold_compare, |
c96c36fa | 822 | sizeof (znode_hold_t), offsetof(znode_hold_t, zh_node)); |
0037b49e | 823 | mutex_init(&zfsvfs->z_hold_locks[i], NULL, MUTEX_DEFAULT, NULL); |
c96c36fa | 824 | } |
9babb374 | 825 | |
8614ddf9 MA |
826 | error = zfsvfs_init(zfsvfs, os); |
827 | if (error != 0) { | |
891ac937 | 828 | dmu_objset_disown(os, B_TRUE, zfsvfs); |
8614ddf9 | 829 | *zfvp = NULL; |
04a3b079 | 830 | zfsvfs_free(zfsvfs); |
8614ddf9 MA |
831 | return (error); |
832 | } | |
833 | ||
dcec0a12 AP |
834 | zfsvfs->z_drain_task = TASKQID_INVALID; |
835 | zfsvfs->z_draining = B_FALSE; | |
836 | zfsvfs->z_drain_cancel = B_TRUE; | |
837 | ||
0037b49e | 838 | *zfvp = zfsvfs; |
9babb374 | 839 | return (0); |
9babb374 BB |
840 | } |
841 | ||
8614ddf9 | 842 | static int |
f298b24d | 843 | zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) |
34dc7c2f | 844 | { |
34dc7c2f | 845 | int error; |
ae76f45c TC |
846 | boolean_t readonly = zfs_is_readonly(zfsvfs); |
847 | ||
1c2555ef | 848 | error = zfs_register_callbacks(zfsvfs->z_vfs); |
34dc7c2f BB |
849 | if (error) |
850 | return (error); | |
851 | ||
34dc7c2f BB |
852 | /* |
853 | * If we are not mounting (ie: online recv), then we don't | |
854 | * have to worry about replaying the log as we blocked all | |
855 | * operations out since we closed the ZIL. | |
856 | */ | |
857 | if (mounting) { | |
dcec0a12 | 858 | ASSERT3P(zfsvfs->z_kstat.dk_kstats, ==, NULL); |
fb087146 AH |
859 | error = dataset_kstats_create(&zfsvfs->z_kstat, zfsvfs->z_os); |
860 | if (error) | |
861 | return (error); | |
862 | zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data, | |
863 | &zfsvfs->z_kstat.dk_zil_sums); | |
dcec0a12 | 864 | |
34dc7c2f BB |
865 | /* |
866 | * During replay we remove the read only flag to | |
867 | * allow replays to succeed. | |
868 | */ | |
dcec0a12 | 869 | if (readonly != 0) { |
0037b49e | 870 | readonly_changed_cb(zfsvfs, B_FALSE); |
dcec0a12 AP |
871 | } else { |
872 | zap_stats_t zs; | |
873 | if (zap_get_stats(zfsvfs->z_os, zfsvfs->z_unlinkedobj, | |
874 | &zs) == 0) { | |
875 | dataset_kstats_update_nunlinks_kstat( | |
876 | &zfsvfs->z_kstat, zs.zs_num_entries); | |
7da304bb AJ |
877 | dprintf_ds(zfsvfs->z_os->os_dsl_dataset, |
878 | "num_entries in unlinked set: %llu", | |
879 | zs.zs_num_entries); | |
dcec0a12 | 880 | } |
0037b49e | 881 | zfs_unlinked_drain(zfsvfs); |
5a42ef04 PD |
882 | dsl_dir_t *dd = zfsvfs->z_os->os_dsl_dataset->ds_dir; |
883 | dd->dd_activity_cancelled = B_FALSE; | |
dcec0a12 | 884 | } |
34dc7c2f | 885 | |
428870ff BB |
886 | /* |
887 | * Parse and replay the intent log. | |
888 | * | |
889 | * Because of ziltest, this must be done after | |
890 | * zfs_unlinked_drain(). (Further note: ziltest | |
891 | * doesn't use readonly mounts, where | |
892 | * zfs_unlinked_drain() isn't called.) This is because | |
893 | * ziltest causes spa_sync() to think it's committed, | |
894 | * but actually it is not, so the intent log contains | |
895 | * many txg's worth of changes. | |
896 | * | |
897 | * In particular, if object N is in the unlinked set in | |
898 | * the last txg to actually sync, then it could be | |
899 | * actually freed in a later txg and then reallocated | |
900 | * in a yet later txg. This would write a "create | |
901 | * object N" record to the intent log. Normally, this | |
902 | * would be fine because the spa_sync() would have | |
903 | * written out the fact that object N is free, before | |
904 | * we could write the "create object N" intent log | |
905 | * record. | |
906 | * | |
907 | * But when we are in ziltest mode, we advance the "open | |
908 | * txg" without actually spa_sync()-ing the changes to | |
909 | * disk. So we would see that object N is still | |
910 | * allocated and in the unlinked set, and there is an | |
911 | * intent log record saying to allocate it. | |
912 | */ | |
0037b49e | 913 | if (spa_writeable(dmu_objset_spa(zfsvfs->z_os))) { |
572e2857 | 914 | if (zil_replay_disable) { |
0037b49e | 915 | zil_destroy(zfsvfs->z_log, B_FALSE); |
572e2857 | 916 | } else { |
0037b49e BB |
917 | zfsvfs->z_replay = B_TRUE; |
918 | zil_replay(zfsvfs->z_os, zfsvfs, | |
572e2857 | 919 | zfs_replay_vector); |
0037b49e | 920 | zfsvfs->z_replay = B_FALSE; |
572e2857 | 921 | } |
fb5f0bc8 | 922 | } |
2cf7f52b BB |
923 | |
924 | /* restore readonly bit */ | |
925 | if (readonly != 0) | |
0037b49e | 926 | readonly_changed_cb(zfsvfs, B_TRUE); |
fb087146 AH |
927 | } else { |
928 | ASSERT3P(zfsvfs->z_kstat.dk_kstats, !=, NULL); | |
929 | zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data, | |
930 | &zfsvfs->z_kstat.dk_zil_sums); | |
34dc7c2f BB |
931 | } |
932 | ||
9775e988 | 933 | /* |
0037b49e | 934 | * Set the objset user_ptr to track its zfsvfs. |
9775e988 | 935 | */ |
0037b49e BB |
936 | mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); |
937 | dmu_objset_set_user(zfsvfs->z_os, zfsvfs); | |
938 | mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); | |
9775e988 | 939 | |
34dc7c2f BB |
940 | return (0); |
941 | } | |
942 | ||
9babb374 | 943 | void |
f298b24d | 944 | zfsvfs_free(zfsvfs_t *zfsvfs) |
34dc7c2f | 945 | { |
0037b49e | 946 | int i, size = zfsvfs->z_hold_size; |
9babb374 | 947 | |
0037b49e | 948 | zfs_fuid_destroy(zfsvfs); |
9babb374 | 949 | |
0037b49e BB |
950 | mutex_destroy(&zfsvfs->z_znodes_lock); |
951 | mutex_destroy(&zfsvfs->z_lock); | |
952 | list_destroy(&zfsvfs->z_all_znodes); | |
5ebe425a | 953 | ZFS_TEARDOWN_DESTROY(zfsvfs); |
0037b49e BB |
954 | rw_destroy(&zfsvfs->z_teardown_inactive_lock); |
955 | rw_destroy(&zfsvfs->z_fuid_lock); | |
c96c36fa | 956 | for (i = 0; i != size; i++) { |
0037b49e BB |
957 | avl_destroy(&zfsvfs->z_hold_trees[i]); |
958 | mutex_destroy(&zfsvfs->z_hold_locks[i]); | |
c96c36fa | 959 | } |
0037b49e BB |
960 | vmem_free(zfsvfs->z_hold_trees, sizeof (avl_tree_t) * size); |
961 | vmem_free(zfsvfs->z_hold_locks, sizeof (kmutex_t) * size); | |
1c2555ef | 962 | zfsvfs_vfs_free(zfsvfs->z_vfs); |
a448a255 | 963 | dataset_kstats_destroy(&zfsvfs->z_kstat); |
0037b49e | 964 | kmem_free(zfsvfs, sizeof (zfsvfs_t)); |
34dc7c2f BB |
965 | } |
966 | ||
9babb374 | 967 | static void |
0037b49e | 968 | zfs_set_fuid_feature(zfsvfs_t *zfsvfs) |
9babb374 | 969 | { |
0037b49e BB |
970 | zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); |
971 | zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); | |
9babb374 | 972 | } |
34dc7c2f | 973 | |
65c7cc49 | 974 | static void |
0037b49e | 975 | zfs_unregister_callbacks(zfsvfs_t *zfsvfs) |
34dc7c2f | 976 | { |
0037b49e | 977 | objset_t *os = zfsvfs->z_os; |
34dc7c2f | 978 | |
0eb21616 | 979 | if (!dmu_objset_is_snapshot(os)) |
0037b49e | 980 | dsl_prop_unregister_all(dmu_objset_ds(os), zfsvfs); |
34dc7c2f BB |
981 | } |
982 | ||
bc3e15e3 | 983 | #ifdef HAVE_MLSLABEL |
428870ff | 984 | /* |
d3cc8b15 WA |
985 | * Check that the hex label string is appropriate for the dataset being |
986 | * mounted into the global_zone proper. | |
428870ff | 987 | * |
d3cc8b15 WA |
988 | * Return an error if the hex label string is not default or |
989 | * admin_low/admin_high. For admin_low labels, the corresponding | |
990 | * dataset must be readonly. | |
428870ff BB |
991 | */ |
992 | int | |
993 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
994 | { | |
995 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
996 | return (0); | |
997 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
998 | return (0); | |
999 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
1000 | /* must be readonly */ | |
1001 | uint64_t rdonly; | |
1002 | ||
1003 | if (dsl_prop_get_integer(dsname, | |
1004 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
2e528b49 | 1005 | return (SET_ERROR(EACCES)); |
28caa74b | 1006 | return (rdonly ? 0 : SET_ERROR(EACCES)); |
428870ff | 1007 | } |
2e528b49 | 1008 | return (SET_ERROR(EACCES)); |
428870ff | 1009 | } |
bc3e15e3 | 1010 | #endif /* HAVE_MLSLABEL */ |
428870ff | 1011 | |
9c5167d1 NF |
1012 | static int |
1013 | zfs_statfs_project(zfsvfs_t *zfsvfs, znode_t *zp, struct kstatfs *statp, | |
1014 | uint32_t bshift) | |
1015 | { | |
1016 | char buf[20 + DMU_OBJACCT_PREFIX_LEN]; | |
1017 | uint64_t offset = DMU_OBJACCT_PREFIX_LEN; | |
1018 | uint64_t quota; | |
1019 | uint64_t used; | |
1020 | int err; | |
1021 | ||
1022 | strlcpy(buf, DMU_OBJACCT_PREFIX, DMU_OBJACCT_PREFIX_LEN + 1); | |
957c7aa2 | 1023 | err = zfs_id_to_fuidstr(zfsvfs, NULL, zp->z_projid, buf + offset, |
c9e319fa | 1024 | sizeof (buf) - offset, B_FALSE); |
9c5167d1 NF |
1025 | if (err) |
1026 | return (err); | |
1027 | ||
1028 | if (zfsvfs->z_projectquota_obj == 0) | |
1029 | goto objs; | |
1030 | ||
1031 | err = zap_lookup(zfsvfs->z_os, zfsvfs->z_projectquota_obj, | |
1032 | buf + offset, 8, 1, "a); | |
1033 | if (err == ENOENT) | |
1034 | goto objs; | |
1035 | else if (err) | |
1036 | return (err); | |
1037 | ||
1038 | err = zap_lookup(zfsvfs->z_os, DMU_PROJECTUSED_OBJECT, | |
1039 | buf + offset, 8, 1, &used); | |
1040 | if (unlikely(err == ENOENT)) { | |
1041 | uint32_t blksize; | |
1042 | u_longlong_t nblocks; | |
1043 | ||
1044 | /* | |
1045 | * Quota accounting is async, so it is possible race case. | |
1046 | * There is at least one object with the given project ID. | |
1047 | */ | |
1048 | sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); | |
1049 | if (unlikely(zp->z_blksz == 0)) | |
1050 | blksize = zfsvfs->z_max_blksz; | |
1051 | ||
1052 | used = blksize * nblocks; | |
1053 | } else if (err) { | |
1054 | return (err); | |
1055 | } | |
1056 | ||
1057 | statp->f_blocks = quota >> bshift; | |
1058 | statp->f_bfree = (quota > used) ? ((quota - used) >> bshift) : 0; | |
1059 | statp->f_bavail = statp->f_bfree; | |
1060 | ||
1061 | objs: | |
1062 | if (zfsvfs->z_projectobjquota_obj == 0) | |
1063 | return (0); | |
1064 | ||
1065 | err = zap_lookup(zfsvfs->z_os, zfsvfs->z_projectobjquota_obj, | |
1066 | buf + offset, 8, 1, "a); | |
1067 | if (err == ENOENT) | |
1068 | return (0); | |
1069 | else if (err) | |
1070 | return (err); | |
1071 | ||
1072 | err = zap_lookup(zfsvfs->z_os, DMU_PROJECTUSED_OBJECT, | |
1073 | buf, 8, 1, &used); | |
1074 | if (unlikely(err == ENOENT)) { | |
1075 | /* | |
1076 | * Quota accounting is async, so it is possible race case. | |
1077 | * There is at least one object with the given project ID. | |
1078 | */ | |
1079 | used = 1; | |
1080 | } else if (err) { | |
1081 | return (err); | |
1082 | } | |
1083 | ||
1084 | statp->f_files = quota; | |
1085 | statp->f_ffree = (quota > used) ? (quota - used) : 0; | |
1086 | ||
1087 | return (0); | |
1088 | } | |
1089 | ||
e5c39b95 | 1090 | int |
f734301d | 1091 | zfs_statvfs(struct inode *ip, struct kstatfs *statp) |
34dc7c2f | 1092 | { |
f734301d | 1093 | zfsvfs_t *zfsvfs = ITOZSB(ip); |
34dc7c2f | 1094 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
9c5167d1 | 1095 | int err = 0; |
34dc7c2f | 1096 | |
768eaced CC |
1097 | if ((err = zfs_enter(zfsvfs, FTAG)) != 0) |
1098 | return (err); | |
34dc7c2f | 1099 | |
0037b49e | 1100 | dmu_objset_space(zfsvfs->z_os, |
34dc7c2f BB |
1101 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
1102 | ||
e897a23e | 1103 | uint64_t fsid = dmu_objset_fsid_guid(zfsvfs->z_os); |
34dc7c2f | 1104 | /* |
05ff35c6 BB |
1105 | * The underlying storage pool actually uses multiple block |
1106 | * size. Under Solaris frsize (fragment size) is reported as | |
1107 | * the smallest block size we support, and bsize (block size) | |
1108 | * as the filesystem's maximum block size. Unfortunately, | |
1109 | * under Linux the fragment size and block size are often used | |
1110 | * interchangeably. Thus we are forced to report both of them | |
1111 | * as the filesystem's maximum block size. | |
34dc7c2f | 1112 | */ |
0037b49e BB |
1113 | statp->f_frsize = zfsvfs->z_max_blksz; |
1114 | statp->f_bsize = zfsvfs->z_max_blksz; | |
e897a23e | 1115 | uint32_t bshift = fls(statp->f_bsize) - 1; |
34dc7c2f BB |
1116 | |
1117 | /* | |
3558fd73 BB |
1118 | * The following report "total" blocks of various kinds in |
1119 | * the file system, but reported in terms of f_bsize - the | |
1120 | * "preferred" size. | |
34dc7c2f BB |
1121 | */ |
1122 | ||
e1cfd73f | 1123 | /* Round up so we never have a filesystem using 0 blocks. */ |
83c796c5 | 1124 | refdbytes = P2ROUNDUP(refdbytes, statp->f_bsize); |
3558fd73 BB |
1125 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
1126 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
1127 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
1128 | ||
1129 | /* | |
1130 | * statvfs() should really be called statufs(), because it assumes | |
1131 | * static metadata. ZFS doesn't preallocate files, so the best | |
1132 | * we can do is report the max that could possibly fit in f_files, | |
1133 | * and that minus the number actually used in f_ffree. | |
e897a23e | 1134 | * For f_ffree, report the smaller of the number of objects available |
34dc7c2f BB |
1135 | * and the number of blocks (each object will take at least a block). |
1136 | */ | |
baab0630 | 1137 | statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT); |
34dc7c2f | 1138 | statp->f_files = statp->f_ffree + usedobjs; |
04f9432d CP |
1139 | statp->f_fsid.val[0] = (uint32_t)fsid; |
1140 | statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); | |
3558fd73 | 1141 | statp->f_type = ZFS_SUPER_MAGIC; |
eca7b760 | 1142 | statp->f_namelen = MAXNAMELEN - 1; |
34dc7c2f BB |
1143 | |
1144 | /* | |
3558fd73 | 1145 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
1146 | * Is there anything useful we could/should provide? |
1147 | */ | |
861166b0 | 1148 | memset(statp->f_spare, 0, sizeof (statp->f_spare)); |
34dc7c2f | 1149 | |
9c5167d1 NF |
1150 | if (dmu_objset_projectquota_enabled(zfsvfs->z_os) && |
1151 | dmu_objset_projectquota_present(zfsvfs->z_os)) { | |
f734301d | 1152 | znode_t *zp = ITOZ(ip); |
9c5167d1 NF |
1153 | |
1154 | if (zp->z_pflags & ZFS_PROJINHERIT && zp->z_projid && | |
1155 | zpl_is_valid_projid(zp->z_projid)) | |
1156 | err = zfs_statfs_project(zfsvfs, zp, statp, bshift); | |
1157 | } | |
1158 | ||
768eaced | 1159 | zfs_exit(zfsvfs, FTAG); |
9c5167d1 | 1160 | return (err); |
34dc7c2f BB |
1161 | } |
1162 | ||
65c7cc49 | 1163 | static int |
0037b49e | 1164 | zfs_root(zfsvfs_t *zfsvfs, struct inode **ipp) |
34dc7c2f | 1165 | { |
34dc7c2f BB |
1166 | znode_t *rootzp; |
1167 | int error; | |
1168 | ||
768eaced CC |
1169 | if ((error = zfs_enter(zfsvfs, FTAG)) != 0) |
1170 | return (error); | |
34dc7c2f | 1171 | |
0037b49e | 1172 | error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp); |
34dc7c2f | 1173 | if (error == 0) |
3558fd73 | 1174 | *ipp = ZTOI(rootzp); |
34dc7c2f | 1175 | |
768eaced | 1176 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1177 | return (error); |
1178 | } | |
1179 | ||
218b4e0a BB |
1180 | /* |
1181 | * Linux kernels older than 3.1 do not support a per-filesystem shrinker. | |
1182 | * To accommodate this we must improvise and manually walk the list of znodes | |
1183 | * attempting to prune dentries in order to be able to drop the inodes. | |
1184 | * | |
1185 | * To avoid scanning the same znodes multiple times they are always rotated | |
1186 | * to the end of the z_all_znodes list. New znodes are inserted at the | |
1187 | * end of the list so we're always scanning the oldest znodes first. | |
1188 | */ | |
1189 | static int | |
f298b24d | 1190 | zfs_prune_aliases(zfsvfs_t *zfsvfs, unsigned long nr_to_scan) |
218b4e0a BB |
1191 | { |
1192 | znode_t **zp_array, *zp; | |
1193 | int max_array = MIN(nr_to_scan, PAGE_SIZE * 8 / sizeof (znode_t *)); | |
1194 | int objects = 0; | |
1195 | int i = 0, j = 0; | |
1196 | ||
1142362f | 1197 | zp_array = vmem_zalloc(max_array * sizeof (znode_t *), KM_SLEEP); |
218b4e0a | 1198 | |
0037b49e BB |
1199 | mutex_enter(&zfsvfs->z_znodes_lock); |
1200 | while ((zp = list_head(&zfsvfs->z_all_znodes)) != NULL) { | |
218b4e0a BB |
1201 | |
1202 | if ((i++ > nr_to_scan) || (j >= max_array)) | |
1203 | break; | |
1204 | ||
1205 | ASSERT(list_link_active(&zp->z_link_node)); | |
0037b49e BB |
1206 | list_remove(&zfsvfs->z_all_znodes, zp); |
1207 | list_insert_tail(&zfsvfs->z_all_znodes, zp); | |
218b4e0a BB |
1208 | |
1209 | /* Skip active znodes and .zfs entries */ | |
1210 | if (MUTEX_HELD(&zp->z_lock) || zp->z_is_ctldir) | |
1211 | continue; | |
1212 | ||
1213 | if (igrab(ZTOI(zp)) == NULL) | |
1214 | continue; | |
1215 | ||
1216 | zp_array[j] = zp; | |
1217 | j++; | |
1218 | } | |
0037b49e | 1219 | mutex_exit(&zfsvfs->z_znodes_lock); |
218b4e0a BB |
1220 | |
1221 | for (i = 0; i < j; i++) { | |
1222 | zp = zp_array[i]; | |
1223 | ||
1224 | ASSERT3P(zp, !=, NULL); | |
1225 | d_prune_aliases(ZTOI(zp)); | |
1226 | ||
1227 | if (atomic_read(&ZTOI(zp)->i_count) == 1) | |
1228 | objects++; | |
1229 | ||
657ce253 | 1230 | zrele(zp); |
218b4e0a BB |
1231 | } |
1232 | ||
1142362f | 1233 | vmem_free(zp_array, max_array * sizeof (znode_t *)); |
218b4e0a BB |
1234 | |
1235 | return (objects); | |
1236 | } | |
218b4e0a | 1237 | |
2cbb06b5 BB |
1238 | /* |
1239 | * The ARC has requested that the filesystem drop entries from the dentry | |
1240 | * and inode caches. This can occur when the ARC needs to free meta data | |
1241 | * blocks but can't because they are all pinned by entries in these caches. | |
1242 | */ | |
ab26409d | 1243 | int |
f298b24d | 1244 | zfs_prune(struct super_block *sb, unsigned long nr_to_scan, int *objects) |
ab26409d | 1245 | { |
0037b49e | 1246 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
2cbb06b5 | 1247 | int error = 0; |
ab26409d BB |
1248 | struct shrinker *shrinker = &sb->s_shrink; |
1249 | struct shrink_control sc = { | |
1250 | .nr_to_scan = nr_to_scan, | |
1251 | .gfp_mask = GFP_KERNEL, | |
1252 | }; | |
1253 | ||
768eaced CC |
1254 | if ((error = zfs_enter(zfsvfs, FTAG)) != 0) |
1255 | return (error); | |
2cbb06b5 | 1256 | |
90947b23 TC |
1257 | #if defined(HAVE_SPLIT_SHRINKER_CALLBACK) && \ |
1258 | defined(SHRINK_CONTROL_HAS_NID) && \ | |
1259 | defined(SHRINKER_NUMA_AWARE) | |
1260 | if (sb->s_shrink.flags & SHRINKER_NUMA_AWARE) { | |
1261 | *objects = 0; | |
02730c33 | 1262 | for_each_online_node(sc.nid) { |
90947b23 | 1263 | *objects += (*shrinker->scan_objects)(shrinker, &sc); |
2a149775 | 1264 | /* |
1265 | * reset sc.nr_to_scan, modified by | |
1266 | * scan_objects == super_cache_scan | |
1267 | */ | |
1268 | sc.nr_to_scan = nr_to_scan; | |
02730c33 | 1269 | } |
90947b23 TC |
1270 | } else { |
1271 | *objects = (*shrinker->scan_objects)(shrinker, &sc); | |
1272 | } | |
09fb30e5 | 1273 | |
90947b23 | 1274 | #elif defined(HAVE_SPLIT_SHRINKER_CALLBACK) |
3c832b8c | 1275 | *objects = (*shrinker->scan_objects)(shrinker, &sc); |
066e8252 | 1276 | #elif defined(HAVE_SINGLE_SHRINKER_CALLBACK) |
ab26409d | 1277 | *objects = (*shrinker->shrink)(shrinker, &sc); |
218b4e0a | 1278 | #elif defined(HAVE_D_PRUNE_ALIASES) |
09fb30e5 | 1279 | #define D_PRUNE_ALIASES_IS_DEFAULT |
f298b24d | 1280 | *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); |
2cbb06b5 | 1281 | #else |
218b4e0a | 1282 | #error "No available dentry and inode cache pruning mechanism." |
3c832b8c | 1283 | #endif |
09fb30e5 TC |
1284 | |
1285 | #if defined(HAVE_D_PRUNE_ALIASES) && !defined(D_PRUNE_ALIASES_IS_DEFAULT) | |
1286 | #undef D_PRUNE_ALIASES_IS_DEFAULT | |
1287 | /* | |
f298b24d | 1288 | * Fall back to zfs_prune_aliases if the kernel's per-superblock |
09fb30e5 TC |
1289 | * shrinker couldn't free anything, possibly due to the inodes being |
1290 | * allocated in a different memcg. | |
1291 | */ | |
1292 | if (*objects == 0) | |
f298b24d | 1293 | *objects = zfs_prune_aliases(zfsvfs, nr_to_scan); |
09fb30e5 TC |
1294 | #endif |
1295 | ||
768eaced | 1296 | zfs_exit(zfsvfs, FTAG); |
ab26409d | 1297 | |
0037b49e | 1298 | dprintf_ds(zfsvfs->z_os->os_dsl_dataset, |
2cbb06b5 BB |
1299 | "pruning, nr_to_scan=%lu objects=%d error=%d\n", |
1300 | nr_to_scan, *objects, error); | |
1301 | ||
1302 | return (error); | |
ab26409d | 1303 | } |
ab26409d | 1304 | |
34dc7c2f | 1305 | /* |
0037b49e | 1306 | * Teardown the zfsvfs_t. |
34dc7c2f | 1307 | * |
a08abc1b | 1308 | * Note, if 'unmounting' is FALSE, we return with the 'z_teardown_lock' |
34dc7c2f BB |
1309 | * and 'z_teardown_inactive_lock' held. |
1310 | */ | |
f298b24d | 1311 | static int |
0037b49e | 1312 | zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) |
34dc7c2f BB |
1313 | { |
1314 | znode_t *zp; | |
1315 | ||
dcec0a12 AP |
1316 | zfs_unlinked_drain_stop_wait(zfsvfs); |
1317 | ||
fd23720a BB |
1318 | /* |
1319 | * If someone has not already unmounted this file system, | |
657ce253 | 1320 | * drain the zrele_taskq to ensure all active references to the |
0037b49e | 1321 | * zfsvfs_t have been handled only then can it be safely destroyed. |
fd23720a | 1322 | */ |
0037b49e | 1323 | if (zfsvfs->z_os) { |
f0da4d15 CD |
1324 | /* |
1325 | * If we're unmounting we have to wait for the list to | |
1326 | * drain completely. | |
1327 | * | |
1328 | * If we're not unmounting there's no guarantee the list | |
1329 | * will drain completely, but iputs run from the taskq | |
1330 | * may add the parents of dir-based xattrs to the taskq | |
1331 | * so we want to wait for these. | |
1332 | * | |
1333 | * We can safely read z_nr_znodes without locking because the | |
1334 | * VFS has already blocked operations which add to the | |
1335 | * z_all_znodes list and thus increment z_nr_znodes. | |
1336 | */ | |
1337 | int round = 0; | |
0037b49e | 1338 | while (zfsvfs->z_nr_znodes > 0) { |
657ce253 | 1339 | taskq_wait_outstanding(dsl_pool_zrele_taskq( |
0037b49e | 1340 | dmu_objset_pool(zfsvfs->z_os)), 0); |
f0da4d15 CD |
1341 | if (++round > 1 && !unmounting) |
1342 | break; | |
1343 | } | |
1344 | } | |
fd23720a | 1345 | |
5ebe425a | 1346 | ZFS_TEARDOWN_ENTER_WRITE(zfsvfs, FTAG); |
34dc7c2f BB |
1347 | |
1348 | if (!unmounting) { | |
1349 | /* | |
ceb43b93 BB |
1350 | * We purge the parent filesystem's super block as the |
1351 | * parent filesystem and all of its snapshots have their | |
1352 | * inode's super block set to the parent's filesystem's | |
1353 | * super block. Note, 'z_parent' is self referential | |
1354 | * for non-snapshots. | |
34dc7c2f | 1355 | */ |
0037b49e | 1356 | shrink_dcache_sb(zfsvfs->z_parent->z_sb); |
34dc7c2f BB |
1357 | } |
1358 | ||
1359 | /* | |
1360 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1361 | * threads are blocked as zil_close can call zfs_inactive. | |
1362 | */ | |
0037b49e BB |
1363 | if (zfsvfs->z_log) { |
1364 | zil_close(zfsvfs->z_log); | |
1365 | zfsvfs->z_log = NULL; | |
34dc7c2f BB |
1366 | } |
1367 | ||
0037b49e | 1368 | rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1369 | |
1370 | /* | |
1371 | * If we are not unmounting (ie: online recv) and someone already | |
1372 | * unmounted this file system while we were doing the switcheroo, | |
1373 | * or a reopen of z_os failed then just bail out now. | |
1374 | */ | |
0037b49e BB |
1375 | if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) { |
1376 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
5ebe425a | 1377 | ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); |
2e528b49 | 1378 | return (SET_ERROR(EIO)); |
34dc7c2f BB |
1379 | } |
1380 | ||
1381 | /* | |
7b3e34ba BB |
1382 | * At this point there are no VFS ops active, and any new VFS ops |
1383 | * will fail with EIO since we have z_teardown_lock for writer (only | |
1384 | * relevant for forced unmount). | |
34dc7c2f | 1385 | * |
e7a2fa70 TC |
1386 | * Release all holds on dbufs. We also grab an extra reference to all |
1387 | * the remaining inodes so that the kernel does not attempt to free | |
1388 | * any inodes of a suspended fs. This can cause deadlocks since the | |
1389 | * zfs_resume_fs() process may involve starting threads, which might | |
1390 | * attempt to free unreferenced inodes to free up memory for the new | |
1391 | * thread. | |
34dc7c2f | 1392 | */ |
f0da4d15 | 1393 | if (!unmounting) { |
0037b49e BB |
1394 | mutex_enter(&zfsvfs->z_znodes_lock); |
1395 | for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL; | |
1396 | zp = list_next(&zfsvfs->z_all_znodes, zp)) { | |
f0da4d15 CD |
1397 | if (zp->z_sa_hdl) |
1398 | zfs_znode_dmu_fini(zp); | |
e7a2fa70 TC |
1399 | if (igrab(ZTOI(zp)) != NULL) |
1400 | zp->z_suspended = B_TRUE; | |
1401 | ||
f0da4d15 | 1402 | } |
0037b49e | 1403 | mutex_exit(&zfsvfs->z_znodes_lock); |
7b3e34ba | 1404 | } |
34dc7c2f BB |
1405 | |
1406 | /* | |
7b3e34ba | 1407 | * If we are unmounting, set the unmounted flag and let new VFS ops |
34dc7c2f | 1408 | * unblock. zfs_inactive will have the unmounted behavior, and all |
7b3e34ba | 1409 | * other VFS ops will fail with EIO. |
34dc7c2f BB |
1410 | */ |
1411 | if (unmounting) { | |
0037b49e BB |
1412 | zfsvfs->z_unmounted = B_TRUE; |
1413 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
5ebe425a | 1414 | ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); |
34dc7c2f BB |
1415 | } |
1416 | ||
1417 | /* | |
1418 | * z_os will be NULL if there was an error in attempting to reopen | |
0037b49e | 1419 | * zfsvfs, so just return as the properties had already been |
3558fd73 | 1420 | * |
34dc7c2f BB |
1421 | * unregistered and cached data had been evicted before. |
1422 | */ | |
0037b49e | 1423 | if (zfsvfs->z_os == NULL) |
34dc7c2f BB |
1424 | return (0); |
1425 | ||
1426 | /* | |
1427 | * Unregister properties. | |
1428 | */ | |
0037b49e | 1429 | zfs_unregister_callbacks(zfsvfs); |
34dc7c2f BB |
1430 | |
1431 | /* | |
47ab01a1 TC |
1432 | * Evict cached data. We must write out any dirty data before |
1433 | * disowning the dataset. | |
34dc7c2f | 1434 | */ |
e2b31b58 PZ |
1435 | objset_t *os = zfsvfs->z_os; |
1436 | boolean_t os_dirty = B_FALSE; | |
1437 | for (int t = 0; t < TXG_SIZE; t++) { | |
1438 | if (dmu_objset_is_dirty(os, t)) { | |
1439 | os_dirty = B_TRUE; | |
1440 | break; | |
1441 | } | |
1442 | } | |
1443 | if (!zfs_is_readonly(zfsvfs) && os_dirty) { | |
0037b49e | 1444 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); |
e2b31b58 | 1445 | } |
0037b49e | 1446 | dmu_objset_evict_dbufs(zfsvfs->z_os); |
5a42ef04 PD |
1447 | dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; |
1448 | dsl_dir_cancel_waiters(dd); | |
34dc7c2f BB |
1449 | |
1450 | return (0); | |
1451 | } | |
1452 | ||
066e8252 | 1453 | #if defined(HAVE_SUPER_SETUP_BDI_NAME) |
5547c2f1 | 1454 | atomic_long_t zfs_bdi_seq = ATOMIC_LONG_INIT(0); |
8c45def2 | 1455 | #endif |
76659dc1 | 1456 | |
e5c39b95 | 1457 | int |
1c2555ef | 1458 | zfs_domount(struct super_block *sb, zfs_mnt_t *zm, int silent) |
34dc7c2f | 1459 | { |
1c2555ef | 1460 | const char *osname = zm->mnt_osname; |
e5f732ed | 1461 | struct inode *root_inode = NULL; |
3558fd73 | 1462 | uint64_t recordsize; |
1c2555ef | 1463 | int error = 0; |
163a8c28 TC |
1464 | zfsvfs_t *zfsvfs = NULL; |
1465 | vfs_t *vfs = NULL; | |
4ed5e250 WA |
1466 | int canwrite; |
1467 | int dataset_visible_zone; | |
1c2555ef BB |
1468 | |
1469 | ASSERT(zm); | |
1470 | ASSERT(osname); | |
34dc7c2f | 1471 | |
4ed5e250 WA |
1472 | dataset_visible_zone = zone_dataset_visible(osname, &canwrite); |
1473 | ||
1474 | /* | |
1475 | * Refuse to mount a filesystem if we are in a namespace and the | |
1476 | * dataset is not visible or writable in that namespace. | |
1477 | */ | |
1478 | if (!INGLOBALZONE(curproc) && | |
1479 | (!dataset_visible_zone || !canwrite)) { | |
1480 | return (SET_ERROR(EPERM)); | |
1481 | } | |
1482 | ||
163a8c28 | 1483 | error = zfsvfs_parse_options(zm->mnt_data, &vfs); |
3558fd73 BB |
1484 | if (error) |
1485 | return (error); | |
1486 | ||
4ed5e250 WA |
1487 | /* |
1488 | * If a non-writable filesystem is being mounted without the | |
1489 | * read-only flag, pretend it was set, as done for snapshots. | |
1490 | */ | |
1491 | if (!canwrite) | |
1492 | vfs->vfs_readonly = true; | |
1493 | ||
163a8c28 TC |
1494 | error = zfsvfs_create(osname, vfs->vfs_readonly, &zfsvfs); |
1495 | if (error) { | |
1496 | zfsvfs_vfs_free(vfs); | |
1c2555ef | 1497 | goto out; |
163a8c28 | 1498 | } |
1c2555ef | 1499 | |
3558fd73 | 1500 | if ((error = dsl_prop_get_integer(osname, "recordsize", |
163a8c28 TC |
1501 | &recordsize, NULL))) { |
1502 | zfsvfs_vfs_free(vfs); | |
3558fd73 | 1503 | goto out; |
163a8c28 | 1504 | } |
3558fd73 | 1505 | |
163a8c28 TC |
1506 | vfs->vfs_data = zfsvfs; |
1507 | zfsvfs->z_vfs = vfs; | |
0037b49e BB |
1508 | zfsvfs->z_sb = sb; |
1509 | sb->s_fs_info = zfsvfs; | |
3558fd73 BB |
1510 | sb->s_magic = ZFS_SUPER_MAGIC; |
1511 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1512 | sb->s_time_gran = 1; | |
1513 | sb->s_blocksize = recordsize; | |
1514 | sb->s_blocksize_bits = ilog2(recordsize); | |
5547c2f1 | 1515 | |
7dae2c81 | 1516 | error = -zpl_bdi_setup(sb, "zfs"); |
5547c2f1 BB |
1517 | if (error) |
1518 | goto out; | |
3558fd73 | 1519 | |
7dae2c81 BB |
1520 | sb->s_bdi->ra_pages = 0; |
1521 | ||
3558fd73 BB |
1522 | /* Set callback operations for the file system. */ |
1523 | sb->s_op = &zpl_super_operations; | |
1524 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1525 | sb->s_export_op = &zpl_export_operations; |
3558fd73 BB |
1526 | |
1527 | /* Set features for file system. */ | |
0037b49e | 1528 | zfs_set_fuid_feature(zfsvfs); |
3558fd73 | 1529 | |
0037b49e | 1530 | if (dmu_objset_is_snapshot(zfsvfs->z_os)) { |
3558fd73 BB |
1531 | uint64_t pval; |
1532 | ||
0037b49e BB |
1533 | atime_changed_cb(zfsvfs, B_FALSE); |
1534 | readonly_changed_cb(zfsvfs, B_TRUE); | |
d1d7e268 MK |
1535 | if ((error = dsl_prop_get_integer(osname, |
1536 | "xattr", &pval, NULL))) | |
3558fd73 | 1537 | goto out; |
0037b49e | 1538 | xattr_changed_cb(zfsvfs, pval); |
d1d7e268 MK |
1539 | if ((error = dsl_prop_get_integer(osname, |
1540 | "acltype", &pval, NULL))) | |
023699cd | 1541 | goto out; |
0037b49e BB |
1542 | acltype_changed_cb(zfsvfs, pval); |
1543 | zfsvfs->z_issnap = B_TRUE; | |
1544 | zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1545 | zfsvfs->z_snap_defer_time = jiffies; | |
1546 | ||
1547 | mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); | |
1548 | dmu_objset_set_user(zfsvfs->z_os, zfsvfs); | |
1549 | mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); | |
3558fd73 | 1550 | } else { |
f298b24d | 1551 | if ((error = zfsvfs_setup(zfsvfs, B_TRUE))) |
be88e733 | 1552 | goto out; |
34dc7c2f BB |
1553 | } |
1554 | ||
3558fd73 | 1555 | /* Allocate a root inode for the filesystem. */ |
0037b49e | 1556 | error = zfs_root(zfsvfs, &root_inode); |
3558fd73 BB |
1557 | if (error) { |
1558 | (void) zfs_umount(sb); | |
4d5aef3b | 1559 | zfsvfs = NULL; /* avoid double-free; first in zfs_umount */ |
3558fd73 | 1560 | goto out; |
34dc7c2f BB |
1561 | } |
1562 | ||
3558fd73 | 1563 | /* Allocate a root dentry for the filesystem */ |
6a0936ba | 1564 | sb->s_root = d_make_root(root_inode); |
3558fd73 BB |
1565 | if (sb->s_root == NULL) { |
1566 | (void) zfs_umount(sb); | |
4d5aef3b | 1567 | zfsvfs = NULL; /* avoid double-free; first in zfs_umount */ |
2e528b49 | 1568 | error = SET_ERROR(ENOMEM); |
3558fd73 BB |
1569 | goto out; |
1570 | } | |
ebe7e575 | 1571 | |
0037b49e BB |
1572 | if (!zfsvfs->z_issnap) |
1573 | zfsctl_create(zfsvfs); | |
2cbb06b5 | 1574 | |
0037b49e | 1575 | zfsvfs->z_arc_prune = arc_add_prune_callback(zpl_prune_sb, sb); |
3558fd73 BB |
1576 | out: |
1577 | if (error) { | |
163a8c28 TC |
1578 | if (zfsvfs != NULL) { |
1579 | dmu_objset_disown(zfsvfs->z_os, B_TRUE, zfsvfs); | |
1580 | zfsvfs_free(zfsvfs); | |
1581 | } | |
be88e733 CC |
1582 | /* |
1583 | * make sure we don't have dangling sb->s_fs_info which | |
1584 | * zfs_preumount will use. | |
1585 | */ | |
1586 | sb->s_fs_info = NULL; | |
3558fd73 | 1587 | } |
34dc7c2f | 1588 | |
3558fd73 BB |
1589 | return (error); |
1590 | } | |
3558fd73 | 1591 | |
ebe7e575 BB |
1592 | /* |
1593 | * Called when an unmount is requested and certain sanity checks have | |
1594 | * already passed. At this point no dentries or inodes have been reclaimed | |
1595 | * from their respective caches. We drop the extra reference on the .zfs | |
1596 | * control directory to allow everything to be reclaimed. All snapshots | |
1597 | * must already have been unmounted to reach this point. | |
1598 | */ | |
1599 | void | |
1600 | zfs_preumount(struct super_block *sb) | |
1601 | { | |
0037b49e | 1602 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
ebe7e575 | 1603 | |
0037b49e BB |
1604 | /* zfsvfs is NULL when zfs_domount fails during mount */ |
1605 | if (zfsvfs) { | |
dcec0a12 | 1606 | zfs_unlinked_drain_stop_wait(zfsvfs); |
278bee93 | 1607 | zfsctl_destroy(sb->s_fs_info); |
be88e733 | 1608 | /* |
657ce253 | 1609 | * Wait for zrele_async before entering evict_inodes in |
be88e733 CC |
1610 | * generic_shutdown_super. The reason we must finish before |
1611 | * evict_inodes is when lazytime is on, or when zfs_purgedir | |
657ce253 | 1612 | * calls zfs_zget, zrele would bump i_count from 0 to 1. This |
be88e733 CC |
1613 | * would race with the i_count check in evict_inodes. This means |
1614 | * it could destroy the inode while we are still using it. | |
1615 | * | |
1616 | * We wait for two passes. xattr directories in the first pass | |
1617 | * may add xattr entries in zfs_purgedir, so in the second pass | |
1618 | * we wait for them. We don't use taskq_wait here because it is | |
1619 | * a pool wide taskq. Other mounted filesystems can constantly | |
657ce253 | 1620 | * do zrele_async and there's no guarantee when taskq will be |
be88e733 CC |
1621 | * empty. |
1622 | */ | |
657ce253 | 1623 | taskq_wait_outstanding(dsl_pool_zrele_taskq( |
0037b49e | 1624 | dmu_objset_pool(zfsvfs->z_os)), 0); |
657ce253 | 1625 | taskq_wait_outstanding(dsl_pool_zrele_taskq( |
0037b49e | 1626 | dmu_objset_pool(zfsvfs->z_os)), 0); |
be88e733 | 1627 | } |
ebe7e575 | 1628 | } |
ebe7e575 BB |
1629 | |
1630 | /* | |
1631 | * Called once all other unmount released tear down has occurred. | |
1632 | * It is our responsibility to release any remaining infrastructure. | |
1633 | */ | |
3558fd73 BB |
1634 | int |
1635 | zfs_umount(struct super_block *sb) | |
1636 | { | |
0037b49e | 1637 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
3558fd73 BB |
1638 | objset_t *os; |
1639 | ||
b5256303 TC |
1640 | if (zfsvfs->z_arc_prune != NULL) |
1641 | arc_remove_prune_callback(zfsvfs->z_arc_prune); | |
0037b49e BB |
1642 | VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0); |
1643 | os = zfsvfs->z_os; | |
7dae2c81 | 1644 | zpl_bdi_destroy(sb); |
76659dc1 | 1645 | |
34dc7c2f BB |
1646 | /* |
1647 | * z_os will be NULL if there was an error in | |
0037b49e | 1648 | * attempting to reopen zfsvfs. |
34dc7c2f BB |
1649 | */ |
1650 | if (os != NULL) { | |
1651 | /* | |
1652 | * Unset the objset user_ptr. | |
1653 | */ | |
428870ff | 1654 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1655 | dmu_objset_set_user(os, NULL); |
428870ff | 1656 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1657 | |
1658 | /* | |
b128c09f | 1659 | * Finally release the objset |
34dc7c2f | 1660 | */ |
b5256303 | 1661 | dmu_objset_disown(os, B_TRUE, zfsvfs); |
34dc7c2f BB |
1662 | } |
1663 | ||
f298b24d | 1664 | zfsvfs_free(zfsvfs); |
2d8a2b51 | 1665 | sb->s_fs_info = NULL; |
34dc7c2f BB |
1666 | return (0); |
1667 | } | |
1668 | ||
0de19dad | 1669 | int |
1c2555ef | 1670 | zfs_remount(struct super_block *sb, int *flags, zfs_mnt_t *zm) |
0de19dad | 1671 | { |
0037b49e | 1672 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
1c2555ef | 1673 | vfs_t *vfsp; |
08de8c16 | 1674 | boolean_t issnap = dmu_objset_is_snapshot(zfsvfs->z_os); |
0282c413 BB |
1675 | int error; |
1676 | ||
08de8c16 | 1677 | if ((issnap || !spa_writeable(dmu_objset_spa(zfsvfs->z_os))) && |
05805494 TH |
1678 | !(*flags & SB_RDONLY)) { |
1679 | *flags |= SB_RDONLY; | |
08de8c16 | 1680 | return (EROFS); |
1681 | } | |
1682 | ||
1c2555ef BB |
1683 | error = zfsvfs_parse_options(zm->mnt_data, &vfsp); |
1684 | if (error) | |
1685 | return (error); | |
1686 | ||
05805494 | 1687 | if (!zfs_is_readonly(zfsvfs) && (*flags & SB_RDONLY)) |
47ab01a1 TC |
1688 | txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); |
1689 | ||
0037b49e | 1690 | zfs_unregister_callbacks(zfsvfs); |
1c2555ef BB |
1691 | zfsvfs_vfs_free(zfsvfs->z_vfs); |
1692 | ||
1693 | vfsp->vfs_data = zfsvfs; | |
1694 | zfsvfs->z_vfs = vfsp; | |
08de8c16 | 1695 | if (!issnap) |
1696 | (void) zfs_register_callbacks(vfsp); | |
0282c413 BB |
1697 | |
1698 | return (error); | |
0de19dad | 1699 | } |
0de19dad | 1700 | |
e5c39b95 | 1701 | int |
2cf7f52b | 1702 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1703 | { |
0037b49e | 1704 | zfsvfs_t *zfsvfs = sb->s_fs_info; |
34dc7c2f BB |
1705 | znode_t *zp; |
1706 | uint64_t object = 0; | |
1707 | uint64_t fid_gen = 0; | |
1708 | uint64_t gen_mask; | |
1709 | uint64_t zp_gen; | |
3558fd73 | 1710 | int i, err; |
34dc7c2f | 1711 | |
3558fd73 | 1712 | *ipp = NULL; |
34dc7c2f | 1713 | |
9b77d1c9 CC |
1714 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { |
1715 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1716 | ||
1717 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1718 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
34dc7c2f | 1719 | |
9b77d1c9 CC |
1720 | for (i = 0; i < sizeof (zfid->zf_gen); i++) |
1721 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1722 | } else { | |
1723 | return (SET_ERROR(EINVAL)); | |
1724 | } | |
1725 | ||
1726 | /* LONG_FID_LEN means snapdirs */ | |
34dc7c2f BB |
1727 | if (fidp->fid_len == LONG_FID_LEN) { |
1728 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1729 | uint64_t objsetid = 0; | |
1730 | uint64_t setgen = 0; | |
1731 | ||
1732 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1733 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1734 | ||
1735 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1736 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1737 | ||
9b77d1c9 CC |
1738 | if (objsetid != ZFSCTL_INO_SNAPDIRS - object) { |
1739 | dprintf("snapdir fid: objsetid (%llu) != " | |
1740 | "ZFSCTL_INO_SNAPDIRS (%llu) - object (%llu)\n", | |
1741 | objsetid, ZFSCTL_INO_SNAPDIRS, object); | |
34dc7c2f | 1742 | |
2e528b49 | 1743 | return (SET_ERROR(EINVAL)); |
9b77d1c9 | 1744 | } |
ebe7e575 | 1745 | |
9b77d1c9 CC |
1746 | if (fid_gen > 1 || setgen != 0) { |
1747 | dprintf("snapdir fid: fid_gen (%llu) and setgen " | |
1748 | "(%llu)\n", fid_gen, setgen); | |
1749 | return (SET_ERROR(EINVAL)); | |
1750 | } | |
34dc7c2f | 1751 | |
9b77d1c9 | 1752 | return (zfsctl_snapdir_vget(sb, objsetid, fid_gen, ipp)); |
34dc7c2f BB |
1753 | } |
1754 | ||
768eaced CC |
1755 | if ((err = zfs_enter(zfsvfs, FTAG)) != 0) |
1756 | return (err); | |
34dc7c2f BB |
1757 | /* A zero fid_gen means we are in the .zfs control directories */ |
1758 | if (fid_gen == 0 && | |
1759 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
0037b49e | 1760 | *ipp = zfsvfs->z_ctldir; |
3558fd73 | 1761 | ASSERT(*ipp != NULL); |
34dc7c2f | 1762 | if (object == ZFSCTL_INO_SNAPDIR) { |
ebe7e575 BB |
1763 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, |
1764 | 0, kcred, NULL, NULL) == 0); | |
34dc7c2f | 1765 | } else { |
1daad981 AM |
1766 | /* |
1767 | * Must have an existing ref, so igrab() | |
1768 | * cannot return NULL | |
1769 | */ | |
1770 | VERIFY3P(igrab(*ipp), !=, NULL); | |
34dc7c2f | 1771 | } |
768eaced | 1772 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1773 | return (0); |
1774 | } | |
1775 | ||
1776 | gen_mask = -1ULL >> (64 - 8 * i); | |
1777 | ||
29e57d15 | 1778 | dprintf("getting %llu [%llu mask %llx]\n", object, fid_gen, gen_mask); |
0037b49e | 1779 | if ((err = zfs_zget(zfsvfs, object, &zp))) { |
768eaced | 1780 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1781 | return (err); |
1782 | } | |
7938c2ac CC |
1783 | |
1784 | /* Don't export xattr stuff */ | |
1785 | if (zp->z_pflags & ZFS_XATTR) { | |
657ce253 | 1786 | zrele(zp); |
768eaced | 1787 | zfs_exit(zfsvfs, FTAG); |
7938c2ac CC |
1788 | return (SET_ERROR(ENOENT)); |
1789 | } | |
1790 | ||
0037b49e | 1791 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen, |
428870ff BB |
1792 | sizeof (uint64_t)); |
1793 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1794 | if (zp_gen == 0) |
1795 | zp_gen = 1; | |
0037b49e | 1796 | if ((fid_gen == 0) && (zfsvfs->z_root == object)) |
0500e835 | 1797 | fid_gen = zp_gen; |
34dc7c2f | 1798 | if (zp->z_unlinked || zp_gen != fid_gen) { |
29e57d15 NB |
1799 | dprintf("znode gen (%llu) != fid gen (%llu)\n", zp_gen, |
1800 | fid_gen); | |
657ce253 | 1801 | zrele(zp); |
768eaced | 1802 | zfs_exit(zfsvfs, FTAG); |
6c253064 | 1803 | return (SET_ERROR(ENOENT)); |
34dc7c2f BB |
1804 | } |
1805 | ||
3558fd73 BB |
1806 | *ipp = ZTOI(zp); |
1807 | if (*ipp) | |
fc273894 | 1808 | zfs_znode_update_vfs(ITOZ(*ipp)); |
960e08fe | 1809 | |
768eaced | 1810 | zfs_exit(zfsvfs, FTAG); |
34dc7c2f BB |
1811 | return (0); |
1812 | } | |
1813 | ||
1814 | /* | |
0037b49e | 1815 | * Block out VFS ops and close zfsvfs_t |
34dc7c2f BB |
1816 | * |
1817 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
831baf06 KW |
1818 | * 'z_teardown_inactive_lock' write held. We leave ownership of the underlying |
1819 | * dataset and objset intact so that they can be atomically handed off during | |
1820 | * a subsequent rollback or recv operation and the resume thereafter. | |
34dc7c2f BB |
1821 | */ |
1822 | int | |
0037b49e | 1823 | zfs_suspend_fs(zfsvfs_t *zfsvfs) |
34dc7c2f BB |
1824 | { |
1825 | int error; | |
1826 | ||
0037b49e | 1827 | if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0) |
34dc7c2f | 1828 | return (error); |
7b3e34ba | 1829 | |
34dc7c2f BB |
1830 | return (0); |
1831 | } | |
1832 | ||
1833 | /* | |
8614ddf9 MA |
1834 | * Rebuild SA and release VOPs. Note that ownership of the underlying dataset |
1835 | * is an invariant across any of the operations that can be performed while the | |
1836 | * filesystem was suspended. Whether it succeeded or failed, the preconditions | |
1837 | * are the same: the relevant objset and associated dataset are owned by | |
1838 | * zfsvfs, held, and long held on entry. | |
34dc7c2f BB |
1839 | */ |
1840 | int | |
0037b49e | 1841 | zfs_resume_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) |
34dc7c2f | 1842 | { |
2008e920 | 1843 | int err, err2; |
831baf06 | 1844 | znode_t *zp; |
34dc7c2f | 1845 | |
5ebe425a | 1846 | ASSERT(ZFS_TEARDOWN_WRITE_HELD(zfsvfs)); |
0037b49e | 1847 | ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); |
34dc7c2f | 1848 | |
831baf06 | 1849 | /* |
ec923db2 GM |
1850 | * We already own this, so just update the objset_t, as the one we |
1851 | * had before may have been evicted. | |
831baf06 | 1852 | */ |
8614ddf9 | 1853 | objset_t *os; |
0037b49e | 1854 | VERIFY3P(ds->ds_owner, ==, zfsvfs); |
ec923db2 | 1855 | VERIFY(dsl_dataset_long_held(ds)); |
0fdd6106 MA |
1856 | dsl_pool_t *dp = spa_get_dsl(dsl_dataset_get_spa(ds)); |
1857 | dsl_pool_config_enter(dp, FTAG); | |
8614ddf9 | 1858 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
0fdd6106 | 1859 | dsl_pool_config_exit(dp, FTAG); |
c1fabe79 | 1860 | |
8614ddf9 MA |
1861 | err = zfsvfs_init(zfsvfs, os); |
1862 | if (err != 0) | |
831baf06 | 1863 | goto bail; |
34dc7c2f | 1864 | |
5a42ef04 | 1865 | ds->ds_dir->dd_activity_cancelled = B_FALSE; |
f298b24d | 1866 | VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); |
c1fabe79 | 1867 | |
0037b49e BB |
1868 | zfs_set_fuid_feature(zfsvfs); |
1869 | zfsvfs->z_rollback_time = jiffies; | |
34dc7c2f | 1870 | |
831baf06 KW |
1871 | /* |
1872 | * Attempt to re-establish all the active inodes with their | |
1873 | * dbufs. If a zfs_rezget() fails, then we unhash the inode | |
1874 | * and mark it stale. This prevents a collision if a new | |
1875 | * inode/object is created which must use the same inode | |
1876 | * number. The stale inode will be be released when the | |
1877 | * VFS prunes the dentry holding the remaining references | |
1878 | * on the stale inode. | |
1879 | */ | |
0037b49e BB |
1880 | mutex_enter(&zfsvfs->z_znodes_lock); |
1881 | for (zp = list_head(&zfsvfs->z_all_znodes); zp; | |
1882 | zp = list_next(&zfsvfs->z_all_znodes, zp)) { | |
2008e920 TC |
1883 | err2 = zfs_rezget(zp); |
1884 | if (err2) { | |
86db35c4 | 1885 | zpl_d_drop_aliases(ZTOI(zp)); |
831baf06 | 1886 | remove_inode_hash(ZTOI(zp)); |
34dc7c2f | 1887 | } |
e7a2fa70 TC |
1888 | |
1889 | /* see comment in zfs_suspend_fs() */ | |
1890 | if (zp->z_suspended) { | |
657ce253 | 1891 | zfs_zrele_async(zp); |
e7a2fa70 TC |
1892 | zp->z_suspended = B_FALSE; |
1893 | } | |
34dc7c2f | 1894 | } |
0037b49e | 1895 | mutex_exit(&zfsvfs->z_znodes_lock); |
34dc7c2f | 1896 | |
dcec0a12 AP |
1897 | if (!zfs_is_readonly(zfsvfs) && !zfsvfs->z_unmounted) { |
1898 | /* | |
1899 | * zfs_suspend_fs() could have interrupted freeing | |
1900 | * of dnodes. We need to restart this freeing so | |
1901 | * that we don't "leak" the space. | |
1902 | */ | |
1903 | zfs_unlinked_drain(zfsvfs); | |
1904 | } | |
1905 | ||
5a6ac4cf PS |
1906 | /* |
1907 | * Most of the time zfs_suspend_fs is used for changing the contents | |
1908 | * of the underlying dataset. ZFS rollback and receive operations | |
1909 | * might create files for which negative dentries are present in | |
1910 | * the cache. Since walking the dcache would require a lot of GPL-only | |
1911 | * code duplication, it's much easier on these rather rare occasions | |
1912 | * just to flush the whole dcache for the given dataset/filesystem. | |
1913 | */ | |
1914 | shrink_dcache_sb(zfsvfs->z_sb); | |
1915 | ||
428870ff | 1916 | bail: |
30af21b0 PD |
1917 | if (err != 0) |
1918 | zfsvfs->z_unmounted = B_TRUE; | |
1919 | ||
7b3e34ba | 1920 | /* release the VFS ops */ |
0037b49e | 1921 | rw_exit(&zfsvfs->z_teardown_inactive_lock); |
5ebe425a | 1922 | ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); |
34dc7c2f | 1923 | |
30af21b0 | 1924 | if (err != 0) { |
34dc7c2f | 1925 | /* |
831baf06 KW |
1926 | * Since we couldn't setup the sa framework, try to force |
1927 | * unmount this file system. | |
34dc7c2f | 1928 | */ |
0037b49e BB |
1929 | if (zfsvfs->z_os) |
1930 | (void) zfs_umount(zfsvfs->z_sb); | |
34dc7c2f BB |
1931 | } |
1932 | return (err); | |
1933 | } | |
1934 | ||
30af21b0 PD |
1935 | /* |
1936 | * Release VOPs and unmount a suspended filesystem. | |
1937 | */ | |
1938 | int | |
1939 | zfs_end_fs(zfsvfs_t *zfsvfs, dsl_dataset_t *ds) | |
1940 | { | |
5ebe425a | 1941 | ASSERT(ZFS_TEARDOWN_WRITE_HELD(zfsvfs)); |
30af21b0 PD |
1942 | ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); |
1943 | ||
1944 | /* | |
1945 | * We already own this, so just hold and rele it to update the | |
1946 | * objset_t, as the one we had before may have been evicted. | |
1947 | */ | |
1948 | objset_t *os; | |
1949 | VERIFY3P(ds->ds_owner, ==, zfsvfs); | |
1950 | VERIFY(dsl_dataset_long_held(ds)); | |
0fdd6106 MA |
1951 | dsl_pool_t *dp = spa_get_dsl(dsl_dataset_get_spa(ds)); |
1952 | dsl_pool_config_enter(dp, FTAG); | |
30af21b0 | 1953 | VERIFY0(dmu_objset_from_ds(ds, &os)); |
0fdd6106 | 1954 | dsl_pool_config_exit(dp, FTAG); |
30af21b0 PD |
1955 | zfsvfs->z_os = os; |
1956 | ||
1957 | /* release the VOPs */ | |
1958 | rw_exit(&zfsvfs->z_teardown_inactive_lock); | |
5ebe425a | 1959 | ZFS_TEARDOWN_EXIT(zfsvfs, FTAG); |
30af21b0 PD |
1960 | |
1961 | /* | |
1962 | * Try to force unmount this file system. | |
1963 | */ | |
1964 | (void) zfs_umount(zfsvfs->z_sb); | |
1965 | zfsvfs->z_unmounted = B_TRUE; | |
1966 | return (0); | |
1967 | } | |
1968 | ||
5a6ac4cf PS |
1969 | /* |
1970 | * Automounted snapshots rely on periodic revalidation | |
1971 | * to defer snapshots from being automatically unmounted. | |
1972 | */ | |
1973 | ||
1974 | inline void | |
1975 | zfs_exit_fs(zfsvfs_t *zfsvfs) | |
1976 | { | |
1977 | if (!zfsvfs->z_issnap) | |
1978 | return; | |
1979 | ||
1980 | if (time_after(jiffies, zfsvfs->z_snap_defer_time + | |
1981 | MAX(zfs_expire_snapshot * HZ / 2, HZ))) { | |
1982 | zfsvfs->z_snap_defer_time = jiffies; | |
1983 | zfsctl_snapshot_unmount_delay(zfsvfs->z_os->os_spa, | |
1984 | dmu_objset_id(zfsvfs->z_os), | |
1985 | zfs_expire_snapshot); | |
1986 | } | |
1987 | } | |
1988 | ||
34dc7c2f | 1989 | int |
0037b49e | 1990 | zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) |
34dc7c2f BB |
1991 | { |
1992 | int error; | |
0037b49e | 1993 | objset_t *os = zfsvfs->z_os; |
34dc7c2f | 1994 | dmu_tx_t *tx; |
34dc7c2f BB |
1995 | |
1996 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
2e528b49 | 1997 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 1998 | |
0037b49e | 1999 | if (newvers < zfsvfs->z_version) |
2e528b49 | 2000 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 2001 | |
428870ff | 2002 | if (zfs_spa_version_map(newvers) > |
0037b49e | 2003 | spa_version(dmu_objset_spa(zfsvfs->z_os))) |
2e528b49 | 2004 | return (SET_ERROR(ENOTSUP)); |
428870ff | 2005 | |
34dc7c2f | 2006 | tx = dmu_tx_create(os); |
9babb374 | 2007 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
0037b49e | 2008 | if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { |
428870ff BB |
2009 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
2010 | ZFS_SA_ATTRS); | |
2011 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
2012 | } | |
34dc7c2f BB |
2013 | error = dmu_tx_assign(tx, TXG_WAIT); |
2014 | if (error) { | |
2015 | dmu_tx_abort(tx); | |
9babb374 BB |
2016 | return (error); |
2017 | } | |
428870ff | 2018 | |
9babb374 BB |
2019 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
2020 | 8, 1, &newvers, tx); | |
2021 | ||
2022 | if (error) { | |
2023 | dmu_tx_commit(tx); | |
2024 | return (error); | |
34dc7c2f | 2025 | } |
34dc7c2f | 2026 | |
0037b49e | 2027 | if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { |
428870ff BB |
2028 | uint64_t sa_obj; |
2029 | ||
0037b49e | 2030 | ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=, |
428870ff BB |
2031 | SPA_VERSION_SA); |
2032 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
2033 | DMU_OT_NONE, 0, tx); | |
2034 | ||
2035 | error = zap_add(os, MASTER_NODE_OBJ, | |
2036 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
c99c9001 | 2037 | ASSERT0(error); |
428870ff BB |
2038 | |
2039 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
2040 | sa_register_update_callback(os, zfs_sa_upgrade); | |
2041 | } | |
2042 | ||
6f1ffb06 | 2043 | spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, |
0037b49e | 2044 | "from %llu to %llu", zfsvfs->z_version, newvers); |
9babb374 | 2045 | |
34dc7c2f BB |
2046 | dmu_tx_commit(tx); |
2047 | ||
0037b49e | 2048 | zfsvfs->z_version = newvers; |
2e5dc449 | 2049 | os->os_version = newvers; |
9babb374 | 2050 | |
0037b49e | 2051 | zfs_set_fuid_feature(zfsvfs); |
9babb374 BB |
2052 | |
2053 | return (0); | |
34dc7c2f BB |
2054 | } |
2055 | ||
a08abc1b | 2056 | /* |
e1cfd73f | 2057 | * Return true if the corresponding vfs's unmounted flag is set. |
a08abc1b GM |
2058 | * Otherwise return false. |
2059 | * If this function returns true we know VFS unmount has been initiated. | |
2060 | */ | |
2061 | boolean_t | |
2062 | zfs_get_vfs_flag_unmounted(objset_t *os) | |
2063 | { | |
0037b49e | 2064 | zfsvfs_t *zfvp; |
a08abc1b GM |
2065 | boolean_t unmounted = B_FALSE; |
2066 | ||
2067 | ASSERT(dmu_objset_type(os) == DMU_OST_ZFS); | |
2068 | ||
2069 | mutex_enter(&os->os_user_ptr_lock); | |
2070 | zfvp = dmu_objset_get_user(os); | |
2071 | if (zfvp != NULL && zfvp->z_unmounted) | |
2072 | unmounted = B_TRUE; | |
2073 | mutex_exit(&os->os_user_ptr_lock); | |
2074 | ||
2075 | return (unmounted); | |
2076 | } | |
2077 | ||
7b4e2723 RM |
2078 | void |
2079 | zfsvfs_update_fromname(const char *oldname, const char *newname) | |
2080 | { | |
2081 | /* | |
2082 | * We don't need to do anything here, the devname is always current by | |
2083 | * virtue of zfsvfs->z_sb->s_op->show_devname. | |
2084 | */ | |
ef70eff1 | 2085 | (void) oldname, (void) newname; |
7b4e2723 RM |
2086 | } |
2087 | ||
3558fd73 BB |
2088 | void |
2089 | zfs_init(void) | |
2090 | { | |
ebe7e575 | 2091 | zfsctl_init(); |
3558fd73 | 2092 | zfs_znode_init(); |
7bcb7f08 | 2093 | dmu_objset_register_type(DMU_OST_ZFS, zpl_get_file_info); |
3558fd73 BB |
2094 | register_filesystem(&zpl_fs_type); |
2095 | } | |
2096 | ||
2097 | void | |
2098 | zfs_fini(void) | |
2099 | { | |
8e71ab99 CC |
2100 | /* |
2101 | * we don't use outstanding because zpl_posix_acl_free might add more. | |
2102 | */ | |
57ddcda1 | 2103 | taskq_wait(system_delay_taskq); |
8e71ab99 | 2104 | taskq_wait(system_taskq); |
3558fd73 BB |
2105 | unregister_filesystem(&zpl_fs_type); |
2106 | zfs_znode_fini(); | |
ebe7e575 | 2107 | zfsctl_fini(); |
3558fd73 | 2108 | } |
f298b24d | 2109 | |
93ce2b4c | 2110 | #if defined(_KERNEL) |
f298b24d BB |
2111 | EXPORT_SYMBOL(zfs_suspend_fs); |
2112 | EXPORT_SYMBOL(zfs_resume_fs); | |
f298b24d BB |
2113 | EXPORT_SYMBOL(zfs_set_version); |
2114 | EXPORT_SYMBOL(zfsvfs_create); | |
2115 | EXPORT_SYMBOL(zfsvfs_free); | |
2116 | EXPORT_SYMBOL(zfs_is_readonly); | |
2117 | EXPORT_SYMBOL(zfs_domount); | |
2118 | EXPORT_SYMBOL(zfs_preumount); | |
2119 | EXPORT_SYMBOL(zfs_umount); | |
2120 | EXPORT_SYMBOL(zfs_remount); | |
2121 | EXPORT_SYMBOL(zfs_statvfs); | |
2122 | EXPORT_SYMBOL(zfs_vget); | |
2123 | EXPORT_SYMBOL(zfs_prune); | |
2124 | #endif |