]>
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. |
34dc7c2f BB |
23 | */ |
24 | ||
428870ff BB |
25 | /* Portions Copyright 2010 Robert Milkowski */ |
26 | ||
34dc7c2f BB |
27 | #include <sys/types.h> |
28 | #include <sys/param.h> | |
29 | #include <sys/systm.h> | |
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> | |
35 | #include <sys/vfs_opreg.h> | |
36 | #include <sys/mntent.h> | |
37 | #include <sys/mount.h> | |
38 | #include <sys/cmn_err.h> | |
39 | #include "fs/fs_subr.h" | |
40 | #include <sys/zfs_znode.h> | |
3558fd73 | 41 | #include <sys/zfs_vnops.h> |
34dc7c2f BB |
42 | #include <sys/zfs_dir.h> |
43 | #include <sys/zil.h> | |
44 | #include <sys/fs/zfs.h> | |
45 | #include <sys/dmu.h> | |
46 | #include <sys/dsl_prop.h> | |
47 | #include <sys/dsl_dataset.h> | |
48 | #include <sys/dsl_deleg.h> | |
49 | #include <sys/spa.h> | |
50 | #include <sys/zap.h> | |
428870ff | 51 | #include <sys/sa.h> |
34dc7c2f BB |
52 | #include <sys/varargs.h> |
53 | #include <sys/policy.h> | |
54 | #include <sys/atomic.h> | |
55 | #include <sys/mkdev.h> | |
56 | #include <sys/modctl.h> | |
57 | #include <sys/refstr.h> | |
58 | #include <sys/zfs_ioctl.h> | |
34dc7c2f BB |
59 | #include <sys/zfs_fuid.h> |
60 | #include <sys/bootconf.h> | |
61 | #include <sys/sunddi.h> | |
62 | #include <sys/dnlc.h> | |
63 | #include <sys/dmu_objset.h> | |
64 | #include <sys/spa_boot.h> | |
428870ff | 65 | #include <sys/sa.h> |
3558fd73 | 66 | #include <sys/zpl.h> |
428870ff | 67 | #include "zfs_comutil.h" |
34dc7c2f | 68 | |
9babb374 | 69 | |
34dc7c2f BB |
70 | /*ARGSUSED*/ |
71 | int | |
03f9ba9d | 72 | zfs_sync(struct super_block *sb, int wait, cred_t *cr) |
34dc7c2f | 73 | { |
03f9ba9d BB |
74 | zfs_sb_t *zsb = sb->s_fs_info; |
75 | ||
34dc7c2f BB |
76 | /* |
77 | * Data integrity is job one. We don't want a compromised kernel | |
78 | * writing to the storage pool, so we never sync during panic. | |
79 | */ | |
d5e53f9d | 80 | if (unlikely(oops_in_progress)) |
34dc7c2f BB |
81 | return (0); |
82 | ||
03f9ba9d BB |
83 | /* |
84 | * Semantically, the only requirement is that the sync be initiated. | |
85 | * The DMU syncs out txgs frequently, so there's nothing to do. | |
86 | */ | |
87 | if (!wait) | |
88 | return (0); | |
89 | ||
3558fd73 | 90 | if (zsb != NULL) { |
34dc7c2f BB |
91 | /* |
92 | * Sync a specific filesystem. | |
93 | */ | |
9babb374 | 94 | dsl_pool_t *dp; |
34dc7c2f | 95 | |
3558fd73 BB |
96 | ZFS_ENTER(zsb); |
97 | dp = dmu_objset_pool(zsb->z_os); | |
9babb374 BB |
98 | |
99 | /* | |
100 | * If the system is shutting down, then skip any | |
101 | * filesystems which may exist on a suspended pool. | |
102 | */ | |
03f9ba9d | 103 | if (spa_suspended(dp->dp_spa)) { |
3558fd73 | 104 | ZFS_EXIT(zsb); |
9babb374 BB |
105 | return (0); |
106 | } | |
107 | ||
3558fd73 BB |
108 | if (zsb->z_log != NULL) |
109 | zil_commit(zsb->z_log, 0); | |
428870ff | 110 | |
3558fd73 | 111 | ZFS_EXIT(zsb); |
34dc7c2f BB |
112 | } else { |
113 | /* | |
114 | * Sync all ZFS filesystems. This is what happens when you | |
115 | * run sync(1M). Unlike other filesystems, ZFS honors the | |
116 | * request by waiting for all pools to commit all dirty data. | |
117 | */ | |
118 | spa_sync_allpools(); | |
119 | } | |
120 | ||
121 | return (0); | |
122 | } | |
e5c39b95 | 123 | EXPORT_SYMBOL(zfs_sync); |
34dc7c2f | 124 | |
2cf7f52b BB |
125 | boolean_t |
126 | zfs_is_readonly(zfs_sb_t *zsb) | |
127 | { | |
128 | return (!!(zsb->z_sb->s_flags & MS_RDONLY)); | |
129 | } | |
130 | EXPORT_SYMBOL(zfs_is_readonly); | |
131 | ||
34dc7c2f BB |
132 | static void |
133 | atime_changed_cb(void *arg, uint64_t newval) | |
134 | { | |
2cf7f52b | 135 | ((zfs_sb_t *)arg)->z_atime = newval; |
34dc7c2f BB |
136 | } |
137 | ||
138 | static void | |
139 | xattr_changed_cb(void *arg, uint64_t newval) | |
140 | { | |
3558fd73 | 141 | zfs_sb_t *zsb = arg; |
34dc7c2f | 142 | |
2cf7f52b BB |
143 | if (newval == TRUE) |
144 | zsb->z_flags |= ZSB_XATTR; | |
145 | else | |
146 | zsb->z_flags &= ~ZSB_XATTR; | |
34dc7c2f BB |
147 | } |
148 | ||
149 | static void | |
150 | blksz_changed_cb(void *arg, uint64_t newval) | |
151 | { | |
3558fd73 | 152 | zfs_sb_t *zsb = arg; |
34dc7c2f BB |
153 | |
154 | if (newval < SPA_MINBLOCKSIZE || | |
155 | newval > SPA_MAXBLOCKSIZE || !ISP2(newval)) | |
156 | newval = SPA_MAXBLOCKSIZE; | |
157 | ||
3558fd73 | 158 | zsb->z_max_blksz = newval; |
34dc7c2f BB |
159 | } |
160 | ||
161 | static void | |
162 | readonly_changed_cb(void *arg, uint64_t newval) | |
163 | { | |
3558fd73 BB |
164 | zfs_sb_t *zsb = arg; |
165 | struct super_block *sb = zsb->z_sb; | |
34dc7c2f | 166 | |
2cf7f52b BB |
167 | if (sb == NULL) |
168 | return; | |
169 | ||
170 | if (newval) | |
3558fd73 | 171 | sb->s_flags |= MS_RDONLY; |
2cf7f52b | 172 | else |
3558fd73 | 173 | sb->s_flags &= ~MS_RDONLY; |
34dc7c2f BB |
174 | } |
175 | ||
176 | static void | |
177 | devices_changed_cb(void *arg, uint64_t newval) | |
178 | { | |
34dc7c2f BB |
179 | } |
180 | ||
181 | static void | |
182 | setuid_changed_cb(void *arg, uint64_t newval) | |
183 | { | |
34dc7c2f BB |
184 | } |
185 | ||
186 | static void | |
187 | exec_changed_cb(void *arg, uint64_t newval) | |
188 | { | |
34dc7c2f BB |
189 | } |
190 | ||
34dc7c2f BB |
191 | static void |
192 | nbmand_changed_cb(void *arg, uint64_t newval) | |
193 | { | |
3558fd73 BB |
194 | zfs_sb_t *zsb = arg; |
195 | struct super_block *sb = zsb->z_sb; | |
196 | ||
2cf7f52b BB |
197 | if (sb == NULL) |
198 | return; | |
199 | ||
200 | if (newval == TRUE) | |
3558fd73 | 201 | sb->s_flags |= MS_MANDLOCK; |
2cf7f52b | 202 | else |
3558fd73 | 203 | sb->s_flags &= ~MS_MANDLOCK; |
34dc7c2f BB |
204 | } |
205 | ||
206 | static void | |
207 | snapdir_changed_cb(void *arg, uint64_t newval) | |
208 | { | |
3558fd73 | 209 | ((zfs_sb_t *)arg)->z_show_ctldir = newval; |
34dc7c2f BB |
210 | } |
211 | ||
212 | static void | |
213 | vscan_changed_cb(void *arg, uint64_t newval) | |
214 | { | |
3558fd73 | 215 | ((zfs_sb_t *)arg)->z_vscan = newval; |
34dc7c2f BB |
216 | } |
217 | ||
34dc7c2f BB |
218 | static void |
219 | acl_inherit_changed_cb(void *arg, uint64_t newval) | |
220 | { | |
3558fd73 | 221 | ((zfs_sb_t *)arg)->z_acl_inherit = newval; |
34dc7c2f BB |
222 | } |
223 | ||
e5c39b95 | 224 | int |
3558fd73 | 225 | zfs_register_callbacks(zfs_sb_t *zsb) |
34dc7c2f BB |
226 | { |
227 | struct dsl_dataset *ds = NULL; | |
3558fd73 | 228 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
229 | int error = 0; |
230 | ||
2cf7f52b BB |
231 | if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) |
232 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
233 | |
234 | /* | |
235 | * Register property callbacks. | |
236 | * | |
237 | * It would probably be fine to just check for i/o error from | |
238 | * the first prop_register(), but I guess I like to go | |
239 | * overboard... | |
240 | */ | |
241 | ds = dmu_objset_ds(os); | |
3558fd73 BB |
242 | error = dsl_prop_register(ds, |
243 | "atime", atime_changed_cb, zsb); | |
34dc7c2f | 244 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 245 | "xattr", xattr_changed_cb, zsb); |
34dc7c2f | 246 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 247 | "recordsize", blksz_changed_cb, zsb); |
34dc7c2f | 248 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 249 | "readonly", readonly_changed_cb, zsb); |
34dc7c2f | 250 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 251 | "devices", devices_changed_cb, zsb); |
34dc7c2f | 252 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 253 | "setuid", setuid_changed_cb, zsb); |
34dc7c2f | 254 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 255 | "exec", exec_changed_cb, zsb); |
34dc7c2f | 256 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 257 | "snapdir", snapdir_changed_cb, zsb); |
34dc7c2f | 258 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 259 | "aclinherit", acl_inherit_changed_cb, zsb); |
34dc7c2f | 260 | error = error ? error : dsl_prop_register(ds, |
3558fd73 | 261 | "vscan", vscan_changed_cb, zsb); |
2cf7f52b BB |
262 | error = error ? error : dsl_prop_register(ds, |
263 | "nbmand", nbmand_changed_cb, zsb); | |
34dc7c2f BB |
264 | if (error) |
265 | goto unregister; | |
266 | ||
34dc7c2f BB |
267 | return (0); |
268 | ||
269 | unregister: | |
270 | /* | |
271 | * We may attempt to unregister some callbacks that are not | |
272 | * registered, but this is OK; it will simply return ENOMSG, | |
273 | * which we will ignore. | |
274 | */ | |
3558fd73 BB |
275 | (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb); |
276 | (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb); | |
277 | (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb); | |
278 | (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb); | |
279 | (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb); | |
280 | (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb); | |
281 | (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb); | |
282 | (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb); | |
34dc7c2f | 283 | (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, |
3558fd73 BB |
284 | zsb); |
285 | (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb); | |
2cf7f52b | 286 | (void) dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb); |
34dc7c2f | 287 | |
3558fd73 | 288 | return (error); |
34dc7c2f | 289 | } |
e5c39b95 | 290 | EXPORT_SYMBOL(zfs_register_callbacks); |
34dc7c2f | 291 | |
428870ff BB |
292 | static int |
293 | zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, | |
294 | uint64_t *userp, uint64_t *groupp) | |
9babb374 | 295 | { |
428870ff BB |
296 | znode_phys_t *znp = data; |
297 | int error = 0; | |
9babb374 | 298 | |
428870ff BB |
299 | /* |
300 | * Is it a valid type of object to track? | |
301 | */ | |
302 | if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) | |
303 | return (ENOENT); | |
9babb374 | 304 | |
428870ff BB |
305 | /* |
306 | * If we have a NULL data pointer | |
307 | * then assume the id's aren't changing and | |
308 | * return EEXIST to the dmu to let it know to | |
309 | * use the same ids | |
310 | */ | |
311 | if (data == NULL) | |
312 | return (EEXIST); | |
9babb374 | 313 | |
428870ff BB |
314 | if (bonustype == DMU_OT_ZNODE) { |
315 | *userp = znp->zp_uid; | |
316 | *groupp = znp->zp_gid; | |
9babb374 | 317 | } else { |
428870ff | 318 | int hdrsize; |
9babb374 | 319 | |
428870ff BB |
320 | ASSERT(bonustype == DMU_OT_SA); |
321 | hdrsize = sa_hdrsize(data); | |
322 | ||
323 | if (hdrsize != 0) { | |
324 | *userp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
325 | SA_UID_OFFSET)); | |
326 | *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + | |
327 | SA_GID_OFFSET)); | |
328 | } else { | |
329 | /* | |
330 | * This should only happen for newly created | |
331 | * files that haven't had the znode data filled | |
332 | * in yet. | |
333 | */ | |
334 | *userp = 0; | |
335 | *groupp = 0; | |
336 | } | |
9babb374 | 337 | } |
428870ff | 338 | return (error); |
9babb374 BB |
339 | } |
340 | ||
341 | static void | |
3558fd73 | 342 | fuidstr_to_sid(zfs_sb_t *zsb, const char *fuidstr, |
9babb374 BB |
343 | char *domainbuf, int buflen, uid_t *ridp) |
344 | { | |
9babb374 BB |
345 | uint64_t fuid; |
346 | const char *domain; | |
347 | ||
348 | fuid = strtonum(fuidstr, NULL); | |
349 | ||
3558fd73 | 350 | domain = zfs_fuid_find_by_idx(zsb, FUID_INDEX(fuid)); |
9babb374 BB |
351 | if (domain) |
352 | (void) strlcpy(domainbuf, domain, buflen); | |
353 | else | |
354 | domainbuf[0] = '\0'; | |
355 | *ridp = FUID_RID(fuid); | |
356 | } | |
357 | ||
358 | static uint64_t | |
3558fd73 | 359 | zfs_userquota_prop_to_obj(zfs_sb_t *zsb, zfs_userquota_prop_t type) |
9babb374 BB |
360 | { |
361 | switch (type) { | |
362 | case ZFS_PROP_USERUSED: | |
363 | return (DMU_USERUSED_OBJECT); | |
364 | case ZFS_PROP_GROUPUSED: | |
365 | return (DMU_GROUPUSED_OBJECT); | |
366 | case ZFS_PROP_USERQUOTA: | |
3558fd73 | 367 | return (zsb->z_userquota_obj); |
9babb374 | 368 | case ZFS_PROP_GROUPQUOTA: |
3558fd73 | 369 | return (zsb->z_groupquota_obj); |
149e873a BB |
370 | default: |
371 | return (ENOTSUP); | |
9babb374 BB |
372 | } |
373 | return (0); | |
374 | } | |
375 | ||
376 | int | |
3558fd73 | 377 | zfs_userspace_many(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
378 | uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) |
379 | { | |
380 | int error; | |
381 | zap_cursor_t zc; | |
382 | zap_attribute_t za; | |
383 | zfs_useracct_t *buf = vbuf; | |
384 | uint64_t obj; | |
385 | ||
3558fd73 | 386 | if (!dmu_objset_userspace_present(zsb->z_os)) |
9babb374 BB |
387 | return (ENOTSUP); |
388 | ||
3558fd73 | 389 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
390 | if (obj == 0) { |
391 | *bufsizep = 0; | |
392 | return (0); | |
393 | } | |
394 | ||
3558fd73 | 395 | for (zap_cursor_init_serialized(&zc, zsb->z_os, obj, *cookiep); |
9babb374 BB |
396 | (error = zap_cursor_retrieve(&zc, &za)) == 0; |
397 | zap_cursor_advance(&zc)) { | |
398 | if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > | |
399 | *bufsizep) | |
400 | break; | |
401 | ||
3558fd73 | 402 | fuidstr_to_sid(zsb, za.za_name, |
9babb374 BB |
403 | buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); |
404 | ||
405 | buf->zu_space = za.za_first_integer; | |
406 | buf++; | |
407 | } | |
408 | if (error == ENOENT) | |
409 | error = 0; | |
410 | ||
411 | ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); | |
412 | *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; | |
413 | *cookiep = zap_cursor_serialize(&zc); | |
414 | zap_cursor_fini(&zc); | |
415 | return (error); | |
416 | } | |
e5c39b95 | 417 | EXPORT_SYMBOL(zfs_userspace_many); |
9babb374 BB |
418 | |
419 | /* | |
420 | * buf must be big enough (eg, 32 bytes) | |
421 | */ | |
422 | static int | |
3558fd73 | 423 | id_to_fuidstr(zfs_sb_t *zsb, const char *domain, uid_t rid, |
9babb374 BB |
424 | char *buf, boolean_t addok) |
425 | { | |
426 | uint64_t fuid; | |
427 | int domainid = 0; | |
428 | ||
429 | if (domain && domain[0]) { | |
3558fd73 | 430 | domainid = zfs_fuid_find_by_domain(zsb, domain, NULL, addok); |
9babb374 BB |
431 | if (domainid == -1) |
432 | return (ENOENT); | |
433 | } | |
434 | fuid = FUID_ENCODE(domainid, rid); | |
435 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
436 | return (0); | |
437 | } | |
438 | ||
439 | int | |
3558fd73 | 440 | zfs_userspace_one(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
441 | const char *domain, uint64_t rid, uint64_t *valp) |
442 | { | |
443 | char buf[32]; | |
444 | int err; | |
445 | uint64_t obj; | |
446 | ||
447 | *valp = 0; | |
448 | ||
3558fd73 | 449 | if (!dmu_objset_userspace_present(zsb->z_os)) |
9babb374 BB |
450 | return (ENOTSUP); |
451 | ||
3558fd73 | 452 | obj = zfs_userquota_prop_to_obj(zsb, type); |
9babb374 BB |
453 | if (obj == 0) |
454 | return (0); | |
455 | ||
3558fd73 | 456 | err = id_to_fuidstr(zsb, domain, rid, buf, B_FALSE); |
9babb374 BB |
457 | if (err) |
458 | return (err); | |
459 | ||
3558fd73 | 460 | err = zap_lookup(zsb->z_os, obj, buf, 8, 1, valp); |
9babb374 BB |
461 | if (err == ENOENT) |
462 | err = 0; | |
463 | return (err); | |
464 | } | |
e5c39b95 | 465 | EXPORT_SYMBOL(zfs_userspace_one); |
9babb374 BB |
466 | |
467 | int | |
3558fd73 | 468 | zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type, |
9babb374 BB |
469 | const char *domain, uint64_t rid, uint64_t quota) |
470 | { | |
471 | char buf[32]; | |
472 | int err; | |
473 | dmu_tx_t *tx; | |
474 | uint64_t *objp; | |
475 | boolean_t fuid_dirtied; | |
476 | ||
477 | if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) | |
478 | return (EINVAL); | |
479 | ||
3558fd73 | 480 | if (zsb->z_version < ZPL_VERSION_USERSPACE) |
9babb374 BB |
481 | return (ENOTSUP); |
482 | ||
3558fd73 BB |
483 | objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj : |
484 | &zsb->z_groupquota_obj; | |
9babb374 | 485 | |
3558fd73 | 486 | err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE); |
9babb374 BB |
487 | if (err) |
488 | return (err); | |
3558fd73 | 489 | fuid_dirtied = zsb->z_fuid_dirty; |
9babb374 | 490 | |
3558fd73 | 491 | tx = dmu_tx_create(zsb->z_os); |
9babb374 BB |
492 | dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); |
493 | if (*objp == 0) { | |
494 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, | |
495 | zfs_userquota_prop_prefixes[type]); | |
496 | } | |
497 | if (fuid_dirtied) | |
3558fd73 | 498 | zfs_fuid_txhold(zsb, tx); |
9babb374 BB |
499 | err = dmu_tx_assign(tx, TXG_WAIT); |
500 | if (err) { | |
501 | dmu_tx_abort(tx); | |
502 | return (err); | |
503 | } | |
504 | ||
3558fd73 | 505 | mutex_enter(&zsb->z_lock); |
9babb374 | 506 | if (*objp == 0) { |
3558fd73 | 507 | *objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA, |
9babb374 | 508 | DMU_OT_NONE, 0, tx); |
3558fd73 | 509 | VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ, |
9babb374 BB |
510 | zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); |
511 | } | |
3558fd73 | 512 | mutex_exit(&zsb->z_lock); |
9babb374 BB |
513 | |
514 | if (quota == 0) { | |
3558fd73 | 515 | err = zap_remove(zsb->z_os, *objp, buf, tx); |
9babb374 BB |
516 | if (err == ENOENT) |
517 | err = 0; | |
518 | } else { | |
3558fd73 | 519 | err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx); |
9babb374 BB |
520 | } |
521 | ASSERT(err == 0); | |
522 | if (fuid_dirtied) | |
3558fd73 | 523 | zfs_fuid_sync(zsb, tx); |
9babb374 BB |
524 | dmu_tx_commit(tx); |
525 | return (err); | |
526 | } | |
e5c39b95 | 527 | EXPORT_SYMBOL(zfs_set_userquota); |
9babb374 BB |
528 | |
529 | boolean_t | |
3558fd73 | 530 | zfs_fuid_overquota(zfs_sb_t *zsb, boolean_t isgroup, uint64_t fuid) |
9babb374 BB |
531 | { |
532 | char buf[32]; | |
533 | uint64_t used, quota, usedobj, quotaobj; | |
534 | int err; | |
535 | ||
536 | usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; | |
3558fd73 | 537 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
9babb374 | 538 | |
3558fd73 | 539 | if (quotaobj == 0 || zsb->z_replay) |
9babb374 BB |
540 | return (B_FALSE); |
541 | ||
542 | (void) sprintf(buf, "%llx", (longlong_t)fuid); | |
3558fd73 | 543 | err = zap_lookup(zsb->z_os, quotaobj, buf, 8, 1, "a); |
9babb374 BB |
544 | if (err != 0) |
545 | return (B_FALSE); | |
546 | ||
3558fd73 | 547 | err = zap_lookup(zsb->z_os, usedobj, buf, 8, 1, &used); |
9babb374 BB |
548 | if (err != 0) |
549 | return (B_FALSE); | |
550 | return (used >= quota); | |
551 | } | |
e5c39b95 | 552 | EXPORT_SYMBOL(zfs_fuid_overquota); |
9babb374 | 553 | |
428870ff | 554 | boolean_t |
3558fd73 | 555 | zfs_owner_overquota(zfs_sb_t *zsb, znode_t *zp, boolean_t isgroup) |
428870ff BB |
556 | { |
557 | uint64_t fuid; | |
558 | uint64_t quotaobj; | |
428870ff | 559 | |
3558fd73 | 560 | quotaobj = isgroup ? zsb->z_groupquota_obj : zsb->z_userquota_obj; |
428870ff | 561 | |
572e2857 | 562 | fuid = isgroup ? zp->z_gid : zp->z_uid; |
428870ff | 563 | |
3558fd73 | 564 | if (quotaobj == 0 || zsb->z_replay) |
428870ff BB |
565 | return (B_FALSE); |
566 | ||
3558fd73 | 567 | return (zfs_fuid_overquota(zsb, isgroup, fuid)); |
428870ff | 568 | } |
e5c39b95 | 569 | EXPORT_SYMBOL(zfs_owner_overquota); |
428870ff | 570 | |
9babb374 | 571 | int |
3558fd73 | 572 | zfs_sb_create(const char *osname, zfs_sb_t **zsbp) |
9babb374 BB |
573 | { |
574 | objset_t *os; | |
3558fd73 | 575 | zfs_sb_t *zsb; |
9babb374 BB |
576 | uint64_t zval; |
577 | int i, error; | |
428870ff | 578 | uint64_t sa_obj; |
9babb374 | 579 | |
3558fd73 | 580 | zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP); |
9babb374 | 581 | |
428870ff BB |
582 | /* |
583 | * We claim to always be readonly so we can open snapshots; | |
584 | * other ZPL code will prevent us from writing to snapshots. | |
585 | */ | |
3558fd73 | 586 | error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os); |
428870ff | 587 | if (error) { |
3558fd73 | 588 | kmem_free(zsb, sizeof (zfs_sb_t)); |
9babb374 | 589 | return (error); |
428870ff | 590 | } |
9babb374 BB |
591 | |
592 | /* | |
593 | * Initialize the zfs-specific filesystem structure. | |
594 | * Should probably make this a kmem cache, shuffle fields, | |
595 | * and just bzero up to z_hold_mtx[]. | |
596 | */ | |
2cf7f52b | 597 | zsb->z_sb = NULL; |
3558fd73 BB |
598 | zsb->z_parent = zsb; |
599 | zsb->z_max_blksz = SPA_MAXBLOCKSIZE; | |
600 | zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; | |
601 | zsb->z_os = os; | |
9babb374 | 602 | |
3558fd73 | 603 | error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version); |
9babb374 BB |
604 | if (error) { |
605 | goto out; | |
3558fd73 | 606 | } else if (zsb->z_version > |
428870ff | 607 | zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { |
149e873a | 608 | (void) printk("Can't mount a version %lld file system " |
428870ff | 609 | "on a version %lld pool\n. Pool must be upgraded to mount " |
3558fd73 | 610 | "this file system.", (u_longlong_t)zsb->z_version, |
428870ff | 611 | (u_longlong_t)spa_version(dmu_objset_spa(os))); |
9babb374 BB |
612 | error = ENOTSUP; |
613 | goto out; | |
614 | } | |
9babb374 BB |
615 | if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) |
616 | goto out; | |
3558fd73 | 617 | zsb->z_norm = (int)zval; |
9babb374 BB |
618 | |
619 | if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) | |
620 | goto out; | |
3558fd73 | 621 | zsb->z_utf8 = (zval != 0); |
9babb374 BB |
622 | |
623 | if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) | |
624 | goto out; | |
3558fd73 | 625 | zsb->z_case = (uint_t)zval; |
9babb374 BB |
626 | |
627 | /* | |
628 | * Fold case on file systems that are always or sometimes case | |
629 | * insensitive. | |
630 | */ | |
3558fd73 BB |
631 | if (zsb->z_case == ZFS_CASE_INSENSITIVE || |
632 | zsb->z_case == ZFS_CASE_MIXED) | |
633 | zsb->z_norm |= U8_TEXTPREP_TOUPPER; | |
9babb374 | 634 | |
3558fd73 BB |
635 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
636 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
428870ff | 637 | |
3558fd73 | 638 | if (zsb->z_use_sa) { |
428870ff BB |
639 | /* should either have both of these objects or none */ |
640 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, | |
641 | &sa_obj); | |
642 | if (error) | |
643 | return (error); | |
644 | } else { | |
645 | /* | |
646 | * Pre SA versions file systems should never touch | |
647 | * either the attribute registration or layout objects. | |
648 | */ | |
649 | sa_obj = 0; | |
650 | } | |
651 | ||
572e2857 | 652 | error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, |
3558fd73 | 653 | &zsb->z_attr_table); |
572e2857 BB |
654 | if (error) |
655 | goto out; | |
428870ff | 656 | |
3558fd73 | 657 | if (zsb->z_version >= ZPL_VERSION_SA) |
428870ff | 658 | sa_register_update_callback(os, zfs_sa_upgrade); |
9babb374 BB |
659 | |
660 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, | |
3558fd73 | 661 | &zsb->z_root); |
9babb374 BB |
662 | if (error) |
663 | goto out; | |
3558fd73 | 664 | ASSERT(zsb->z_root != 0); |
9babb374 BB |
665 | |
666 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, | |
3558fd73 | 667 | &zsb->z_unlinkedobj); |
9babb374 BB |
668 | if (error) |
669 | goto out; | |
670 | ||
671 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
672 | zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], | |
3558fd73 | 673 | 8, 1, &zsb->z_userquota_obj); |
9babb374 BB |
674 | if (error && error != ENOENT) |
675 | goto out; | |
676 | ||
677 | error = zap_lookup(os, MASTER_NODE_OBJ, | |
678 | zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], | |
3558fd73 | 679 | 8, 1, &zsb->z_groupquota_obj); |
9babb374 BB |
680 | if (error && error != ENOENT) |
681 | goto out; | |
682 | ||
683 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, | |
3558fd73 | 684 | &zsb->z_fuid_obj); |
9babb374 BB |
685 | if (error && error != ENOENT) |
686 | goto out; | |
687 | ||
688 | error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, | |
3558fd73 | 689 | &zsb->z_shares_dir); |
9babb374 BB |
690 | if (error && error != ENOENT) |
691 | goto out; | |
692 | ||
3558fd73 BB |
693 | mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); |
694 | mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL); | |
695 | list_create(&zsb->z_all_znodes, sizeof (znode_t), | |
9babb374 | 696 | offsetof(znode_t, z_link_node)); |
3558fd73 BB |
697 | rrw_init(&zsb->z_teardown_lock); |
698 | rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); | |
699 | rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL); | |
9babb374 | 700 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 | 701 | mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); |
9babb374 | 702 | |
3558fd73 | 703 | *zsbp = zsb; |
9babb374 BB |
704 | return (0); |
705 | ||
706 | out: | |
3558fd73 BB |
707 | dmu_objset_disown(os, zsb); |
708 | *zsbp = NULL; | |
709 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
9babb374 BB |
710 | return (error); |
711 | } | |
712 | ||
34dc7c2f | 713 | static int |
3558fd73 | 714 | zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting) |
34dc7c2f | 715 | { |
34dc7c2f BB |
716 | int error; |
717 | ||
3558fd73 | 718 | error = zfs_register_callbacks(zsb); |
34dc7c2f BB |
719 | if (error) |
720 | return (error); | |
721 | ||
722 | /* | |
3558fd73 | 723 | * Set the objset user_ptr to track its zsb. |
34dc7c2f | 724 | */ |
3558fd73 BB |
725 | mutex_enter(&zsb->z_os->os_user_ptr_lock); |
726 | dmu_objset_set_user(zsb->z_os, zsb); | |
727 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
34dc7c2f | 728 | |
3558fd73 | 729 | zsb->z_log = zil_open(zsb->z_os, zfs_get_data); |
9babb374 | 730 | |
34dc7c2f BB |
731 | /* |
732 | * If we are not mounting (ie: online recv), then we don't | |
733 | * have to worry about replaying the log as we blocked all | |
734 | * operations out since we closed the ZIL. | |
735 | */ | |
736 | if (mounting) { | |
b128c09f BB |
737 | boolean_t readonly; |
738 | ||
34dc7c2f BB |
739 | /* |
740 | * During replay we remove the read only flag to | |
741 | * allow replays to succeed. | |
742 | */ | |
2cf7f52b | 743 | readonly = zfs_is_readonly(zsb); |
fb5f0bc8 | 744 | if (readonly != 0) |
2cf7f52b | 745 | readonly_changed_cb(zsb, B_FALSE); |
fb5f0bc8 | 746 | else |
3558fd73 | 747 | zfs_unlinked_drain(zsb); |
34dc7c2f | 748 | |
428870ff BB |
749 | /* |
750 | * Parse and replay the intent log. | |
751 | * | |
752 | * Because of ziltest, this must be done after | |
753 | * zfs_unlinked_drain(). (Further note: ziltest | |
754 | * doesn't use readonly mounts, where | |
755 | * zfs_unlinked_drain() isn't called.) This is because | |
756 | * ziltest causes spa_sync() to think it's committed, | |
757 | * but actually it is not, so the intent log contains | |
758 | * many txg's worth of changes. | |
759 | * | |
760 | * In particular, if object N is in the unlinked set in | |
761 | * the last txg to actually sync, then it could be | |
762 | * actually freed in a later txg and then reallocated | |
763 | * in a yet later txg. This would write a "create | |
764 | * object N" record to the intent log. Normally, this | |
765 | * would be fine because the spa_sync() would have | |
766 | * written out the fact that object N is free, before | |
767 | * we could write the "create object N" intent log | |
768 | * record. | |
769 | * | |
770 | * But when we are in ziltest mode, we advance the "open | |
771 | * txg" without actually spa_sync()-ing the changes to | |
772 | * disk. So we would see that object N is still | |
773 | * allocated and in the unlinked set, and there is an | |
774 | * intent log record saying to allocate it. | |
775 | */ | |
3558fd73 | 776 | if (spa_writeable(dmu_objset_spa(zsb->z_os))) { |
572e2857 | 777 | if (zil_replay_disable) { |
3558fd73 | 778 | zil_destroy(zsb->z_log, B_FALSE); |
572e2857 | 779 | } else { |
3558fd73 BB |
780 | zsb->z_replay = B_TRUE; |
781 | zil_replay(zsb->z_os, zsb, | |
572e2857 | 782 | zfs_replay_vector); |
3558fd73 | 783 | zsb->z_replay = B_FALSE; |
572e2857 | 784 | } |
fb5f0bc8 | 785 | } |
2cf7f52b BB |
786 | |
787 | /* restore readonly bit */ | |
788 | if (readonly != 0) | |
789 | readonly_changed_cb(zsb, B_TRUE); | |
34dc7c2f BB |
790 | } |
791 | ||
34dc7c2f BB |
792 | return (0); |
793 | } | |
794 | ||
9babb374 | 795 | void |
3558fd73 | 796 | zfs_sb_free(zfs_sb_t *zsb) |
34dc7c2f | 797 | { |
9babb374 | 798 | int i; |
9babb374 | 799 | |
3558fd73 | 800 | zfs_fuid_destroy(zsb); |
9babb374 | 801 | |
3558fd73 BB |
802 | mutex_destroy(&zsb->z_znodes_lock); |
803 | mutex_destroy(&zsb->z_lock); | |
804 | list_destroy(&zsb->z_all_znodes); | |
805 | rrw_destroy(&zsb->z_teardown_lock); | |
806 | rw_destroy(&zsb->z_teardown_inactive_lock); | |
807 | rw_destroy(&zsb->z_fuid_lock); | |
9babb374 | 808 | for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) |
3558fd73 BB |
809 | mutex_destroy(&zsb->z_hold_mtx[i]); |
810 | kmem_free(zsb, sizeof (zfs_sb_t)); | |
34dc7c2f BB |
811 | } |
812 | ||
9babb374 | 813 | static void |
3558fd73 | 814 | zfs_set_fuid_feature(zfs_sb_t *zsb) |
9babb374 | 815 | { |
3558fd73 BB |
816 | zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os); |
817 | zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os); | |
9babb374 | 818 | } |
34dc7c2f BB |
819 | |
820 | void | |
3558fd73 | 821 | zfs_unregister_callbacks(zfs_sb_t *zsb) |
34dc7c2f | 822 | { |
3558fd73 | 823 | objset_t *os = zsb->z_os; |
34dc7c2f BB |
824 | struct dsl_dataset *ds; |
825 | ||
826 | /* | |
827 | * Unregister properties. | |
828 | */ | |
829 | if (!dmu_objset_is_snapshot(os)) { | |
830 | ds = dmu_objset_ds(os); | |
831 | VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, | |
3558fd73 | 832 | zsb) == 0); |
34dc7c2f BB |
833 | |
834 | VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, | |
3558fd73 | 835 | zsb) == 0); |
34dc7c2f BB |
836 | |
837 | VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, | |
3558fd73 | 838 | zsb) == 0); |
34dc7c2f BB |
839 | |
840 | VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, | |
3558fd73 | 841 | zsb) == 0); |
34dc7c2f BB |
842 | |
843 | VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, | |
3558fd73 | 844 | zsb) == 0); |
34dc7c2f BB |
845 | |
846 | VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, | |
3558fd73 | 847 | zsb) == 0); |
34dc7c2f BB |
848 | |
849 | VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, | |
3558fd73 | 850 | zsb) == 0); |
34dc7c2f BB |
851 | |
852 | VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, | |
3558fd73 | 853 | zsb) == 0); |
34dc7c2f | 854 | |
34dc7c2f | 855 | VERIFY(dsl_prop_unregister(ds, "aclinherit", |
3558fd73 | 856 | acl_inherit_changed_cb, zsb) == 0); |
34dc7c2f BB |
857 | |
858 | VERIFY(dsl_prop_unregister(ds, "vscan", | |
3558fd73 | 859 | vscan_changed_cb, zsb) == 0); |
2cf7f52b BB |
860 | |
861 | VERIFY(dsl_prop_unregister(ds, "nbmand", | |
862 | nbmand_changed_cb, zsb) == 0); | |
34dc7c2f BB |
863 | } |
864 | } | |
e5c39b95 | 865 | EXPORT_SYMBOL(zfs_unregister_callbacks); |
34dc7c2f | 866 | |
bc3e15e3 | 867 | #ifdef HAVE_MLSLABEL |
428870ff BB |
868 | /* |
869 | * zfs_check_global_label: | |
870 | * Check that the hex label string is appropriate for the dataset | |
871 | * being mounted into the global_zone proper. | |
872 | * | |
873 | * Return an error if the hex label string is not default or | |
874 | * admin_low/admin_high. For admin_low labels, the corresponding | |
875 | * dataset must be readonly. | |
876 | */ | |
877 | int | |
878 | zfs_check_global_label(const char *dsname, const char *hexsl) | |
879 | { | |
880 | if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) | |
881 | return (0); | |
882 | if (strcasecmp(hexsl, ADMIN_HIGH) == 0) | |
883 | return (0); | |
884 | if (strcasecmp(hexsl, ADMIN_LOW) == 0) { | |
885 | /* must be readonly */ | |
886 | uint64_t rdonly; | |
887 | ||
888 | if (dsl_prop_get_integer(dsname, | |
889 | zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) | |
890 | return (EACCES); | |
891 | return (rdonly ? 0 : EACCES); | |
892 | } | |
893 | return (EACCES); | |
894 | } | |
bc3e15e3 | 895 | #endif /* HAVE_MLSLABEL */ |
428870ff | 896 | |
e5c39b95 | 897 | int |
3558fd73 | 898 | zfs_statvfs(struct dentry *dentry, struct kstatfs *statp) |
34dc7c2f | 899 | { |
3558fd73 | 900 | zfs_sb_t *zsb = dentry->d_sb->s_fs_info; |
34dc7c2f | 901 | uint64_t refdbytes, availbytes, usedobjs, availobjs; |
3558fd73 | 902 | uint32_t bshift; |
34dc7c2f | 903 | |
3558fd73 | 904 | ZFS_ENTER(zsb); |
34dc7c2f | 905 | |
3558fd73 | 906 | dmu_objset_space(zsb->z_os, |
34dc7c2f BB |
907 | &refdbytes, &availbytes, &usedobjs, &availobjs); |
908 | ||
909 | /* | |
05ff35c6 BB |
910 | * The underlying storage pool actually uses multiple block |
911 | * size. Under Solaris frsize (fragment size) is reported as | |
912 | * the smallest block size we support, and bsize (block size) | |
913 | * as the filesystem's maximum block size. Unfortunately, | |
914 | * under Linux the fragment size and block size are often used | |
915 | * interchangeably. Thus we are forced to report both of them | |
916 | * as the filesystem's maximum block size. | |
34dc7c2f | 917 | */ |
05ff35c6 | 918 | statp->f_frsize = zsb->z_max_blksz; |
3558fd73 BB |
919 | statp->f_bsize = zsb->z_max_blksz; |
920 | bshift = fls(statp->f_bsize) - 1; | |
34dc7c2f BB |
921 | |
922 | /* | |
3558fd73 BB |
923 | * The following report "total" blocks of various kinds in |
924 | * the file system, but reported in terms of f_bsize - the | |
925 | * "preferred" size. | |
34dc7c2f BB |
926 | */ |
927 | ||
3558fd73 BB |
928 | statp->f_blocks = (refdbytes + availbytes) >> bshift; |
929 | statp->f_bfree = availbytes >> bshift; | |
34dc7c2f BB |
930 | statp->f_bavail = statp->f_bfree; /* no root reservation */ |
931 | ||
932 | /* | |
933 | * statvfs() should really be called statufs(), because it assumes | |
934 | * static metadata. ZFS doesn't preallocate files, so the best | |
935 | * we can do is report the max that could possibly fit in f_files, | |
936 | * and that minus the number actually used in f_ffree. | |
937 | * For f_ffree, report the smaller of the number of object available | |
938 | * and the number of blocks (each object will take at least a block). | |
939 | */ | |
940 | statp->f_ffree = MIN(availobjs, statp->f_bfree); | |
34dc7c2f | 941 | statp->f_files = statp->f_ffree + usedobjs; |
53cf50e0 | 942 | statp->f_fsid.val[0] = dentry->d_sb->s_dev; |
3558fd73 BB |
943 | statp->f_fsid.val[1] = 0; |
944 | statp->f_type = ZFS_SUPER_MAGIC; | |
945 | statp->f_namelen = ZFS_MAXNAMELEN; | |
34dc7c2f BB |
946 | |
947 | /* | |
3558fd73 | 948 | * We have all of 40 characters to stuff a string here. |
34dc7c2f BB |
949 | * Is there anything useful we could/should provide? |
950 | */ | |
3558fd73 | 951 | bzero(statp->f_spare, sizeof (statp->f_spare)); |
34dc7c2f | 952 | |
3558fd73 | 953 | ZFS_EXIT(zsb); |
34dc7c2f BB |
954 | return (0); |
955 | } | |
e5c39b95 | 956 | EXPORT_SYMBOL(zfs_statvfs); |
34dc7c2f | 957 | |
e5c39b95 | 958 | int |
3558fd73 | 959 | zfs_root(zfs_sb_t *zsb, struct inode **ipp) |
34dc7c2f | 960 | { |
34dc7c2f BB |
961 | znode_t *rootzp; |
962 | int error; | |
963 | ||
3558fd73 | 964 | ZFS_ENTER(zsb); |
34dc7c2f | 965 | |
3558fd73 | 966 | error = zfs_zget(zsb, zsb->z_root, &rootzp); |
34dc7c2f | 967 | if (error == 0) |
3558fd73 | 968 | *ipp = ZTOI(rootzp); |
34dc7c2f | 969 | |
3558fd73 | 970 | ZFS_EXIT(zsb); |
34dc7c2f BB |
971 | return (error); |
972 | } | |
e5c39b95 | 973 | EXPORT_SYMBOL(zfs_root); |
34dc7c2f BB |
974 | |
975 | /* | |
3558fd73 | 976 | * Teardown the zfs_sb_t::z_os. |
34dc7c2f BB |
977 | * |
978 | * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' | |
979 | * and 'z_teardown_inactive_lock' held. | |
980 | */ | |
3558fd73 BB |
981 | int |
982 | zfsvfs_teardown(zfs_sb_t *zsb, boolean_t unmounting) | |
34dc7c2f BB |
983 | { |
984 | znode_t *zp; | |
985 | ||
3558fd73 | 986 | rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG); |
34dc7c2f BB |
987 | |
988 | if (!unmounting) { | |
989 | /* | |
ceb43b93 BB |
990 | * We purge the parent filesystem's super block as the |
991 | * parent filesystem and all of its snapshots have their | |
992 | * inode's super block set to the parent's filesystem's | |
993 | * super block. Note, 'z_parent' is self referential | |
994 | * for non-snapshots. | |
34dc7c2f | 995 | */ |
ceb43b93 | 996 | shrink_dcache_sb(zsb->z_parent->z_sb); |
e30c0ada | 997 | (void) spl_invalidate_inodes(zsb->z_parent->z_sb, 0); |
34dc7c2f BB |
998 | } |
999 | ||
b00131d4 GB |
1000 | /* |
1001 | * Drain the iput_taskq to ensure all active references to the | |
1002 | * zfs_sb_t have been handled only then can it be safely destroyed. | |
1003 | */ | |
1004 | taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os))); | |
1005 | ||
34dc7c2f BB |
1006 | /* |
1007 | * Close the zil. NB: Can't close the zil while zfs_inactive | |
1008 | * threads are blocked as zil_close can call zfs_inactive. | |
1009 | */ | |
3558fd73 BB |
1010 | if (zsb->z_log) { |
1011 | zil_close(zsb->z_log); | |
1012 | zsb->z_log = NULL; | |
34dc7c2f BB |
1013 | } |
1014 | ||
3558fd73 | 1015 | rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER); |
34dc7c2f BB |
1016 | |
1017 | /* | |
1018 | * If we are not unmounting (ie: online recv) and someone already | |
1019 | * unmounted this file system while we were doing the switcheroo, | |
1020 | * or a reopen of z_os failed then just bail out now. | |
1021 | */ | |
3558fd73 BB |
1022 | if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) { |
1023 | rw_exit(&zsb->z_teardown_inactive_lock); | |
1024 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1025 | return (EIO); |
1026 | } | |
1027 | ||
1028 | /* | |
1029 | * At this point there are no vops active, and any new vops will | |
1030 | * fail with EIO since we have z_teardown_lock for writer (only | |
1031 | * relavent for forced unmount). | |
1032 | * | |
1033 | * Release all holds on dbufs. | |
1034 | */ | |
3558fd73 BB |
1035 | mutex_enter(&zsb->z_znodes_lock); |
1036 | for (zp = list_head(&zsb->z_all_znodes); zp != NULL; | |
1037 | zp = list_next(&zsb->z_all_znodes, zp)) | |
428870ff | 1038 | if (zp->z_sa_hdl) { |
3558fd73 | 1039 | ASSERT(atomic_read(&ZTOI(zp)->i_count) > 0); |
34dc7c2f BB |
1040 | zfs_znode_dmu_fini(zp); |
1041 | } | |
3558fd73 | 1042 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f BB |
1043 | |
1044 | /* | |
1045 | * If we are unmounting, set the unmounted flag and let new vops | |
1046 | * unblock. zfs_inactive will have the unmounted behavior, and all | |
1047 | * other vops will fail with EIO. | |
1048 | */ | |
1049 | if (unmounting) { | |
3558fd73 BB |
1050 | zsb->z_unmounted = B_TRUE; |
1051 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
1052 | rw_exit(&zsb->z_teardown_inactive_lock); | |
34dc7c2f BB |
1053 | } |
1054 | ||
1055 | /* | |
1056 | * z_os will be NULL if there was an error in attempting to reopen | |
3558fd73 BB |
1057 | * zsb, so just return as the properties had already been |
1058 | * | |
34dc7c2f BB |
1059 | * unregistered and cached data had been evicted before. |
1060 | */ | |
3558fd73 | 1061 | if (zsb->z_os == NULL) |
34dc7c2f BB |
1062 | return (0); |
1063 | ||
1064 | /* | |
1065 | * Unregister properties. | |
1066 | */ | |
3558fd73 | 1067 | zfs_unregister_callbacks(zsb); |
34dc7c2f BB |
1068 | |
1069 | /* | |
1070 | * Evict cached data | |
1071 | */ | |
3558fd73 | 1072 | if (dmu_objset_is_dirty_anywhere(zsb->z_os)) |
2cf7f52b | 1073 | if (!zfs_is_readonly(zsb)) |
3558fd73 BB |
1074 | txg_wait_synced(dmu_objset_pool(zsb->z_os), 0); |
1075 | (void) dmu_objset_evict_dbufs(zsb->z_os); | |
34dc7c2f BB |
1076 | |
1077 | return (0); | |
1078 | } | |
1079 | ||
e5c39b95 | 1080 | int |
3558fd73 | 1081 | zfs_domount(struct super_block *sb, void *data, int silent) |
34dc7c2f | 1082 | { |
3558fd73 BB |
1083 | zpl_mount_data_t *zmd = data; |
1084 | const char *osname = zmd->z_osname; | |
1085 | zfs_sb_t *zsb; | |
1086 | struct inode *root_inode; | |
1087 | uint64_t recordsize; | |
1088 | int error; | |
34dc7c2f | 1089 | |
3558fd73 BB |
1090 | error = zfs_sb_create(osname, &zsb); |
1091 | if (error) | |
1092 | return (error); | |
1093 | ||
1094 | if ((error = dsl_prop_get_integer(osname, "recordsize", | |
1095 | &recordsize, NULL))) | |
1096 | goto out; | |
1097 | ||
1098 | zsb->z_sb = sb; | |
3558fd73 BB |
1099 | sb->s_fs_info = zsb; |
1100 | sb->s_magic = ZFS_SUPER_MAGIC; | |
1101 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
1102 | sb->s_time_gran = 1; | |
1103 | sb->s_blocksize = recordsize; | |
1104 | sb->s_blocksize_bits = ilog2(recordsize); | |
1105 | ||
1106 | /* Set callback operations for the file system. */ | |
1107 | sb->s_op = &zpl_super_operations; | |
1108 | sb->s_xattr = zpl_xattr_handlers; | |
055656d4 | 1109 | sb->s_export_op = &zpl_export_operations; |
3558fd73 BB |
1110 | |
1111 | /* Set features for file system. */ | |
1112 | zfs_set_fuid_feature(zsb); | |
1113 | ||
1114 | if (dmu_objset_is_snapshot(zsb->z_os)) { | |
1115 | uint64_t pval; | |
1116 | ||
1117 | atime_changed_cb(zsb, B_FALSE); | |
1118 | readonly_changed_cb(zsb, B_TRUE); | |
1119 | if ((error = dsl_prop_get_integer(osname,"xattr",&pval,NULL))) | |
1120 | goto out; | |
1121 | xattr_changed_cb(zsb, pval); | |
1122 | zsb->z_issnap = B_TRUE; | |
1123 | zsb->z_os->os_sync = ZFS_SYNC_DISABLED; | |
1124 | ||
1125 | mutex_enter(&zsb->z_os->os_user_ptr_lock); | |
1126 | dmu_objset_set_user(zsb->z_os, zsb); | |
1127 | mutex_exit(&zsb->z_os->os_user_ptr_lock); | |
1128 | } else { | |
1129 | error = zfs_sb_setup(zsb, B_TRUE); | |
1130 | #ifdef HAVE_SNAPSHOT | |
1131 | (void) zfs_snap_create(zsb); | |
1132 | #endif /* HAVE_SNAPSHOT */ | |
34dc7c2f BB |
1133 | } |
1134 | ||
3558fd73 BB |
1135 | /* Allocate a root inode for the filesystem. */ |
1136 | error = zfs_root(zsb, &root_inode); | |
1137 | if (error) { | |
1138 | (void) zfs_umount(sb); | |
1139 | goto out; | |
34dc7c2f BB |
1140 | } |
1141 | ||
3558fd73 BB |
1142 | /* Allocate a root dentry for the filesystem */ |
1143 | sb->s_root = d_alloc_root(root_inode); | |
1144 | if (sb->s_root == NULL) { | |
1145 | (void) zfs_umount(sb); | |
1146 | error = ENOMEM; | |
1147 | goto out; | |
1148 | } | |
1149 | out: | |
1150 | if (error) { | |
1151 | dmu_objset_disown(zsb->z_os, zsb); | |
1152 | zfs_sb_free(zsb); | |
1153 | } | |
34dc7c2f | 1154 | |
3558fd73 BB |
1155 | return (error); |
1156 | } | |
1157 | EXPORT_SYMBOL(zfs_domount); | |
1158 | ||
1159 | /*ARGSUSED*/ | |
1160 | int | |
1161 | zfs_umount(struct super_block *sb) | |
1162 | { | |
1163 | zfs_sb_t *zsb = sb->s_fs_info; | |
1164 | objset_t *os; | |
1165 | ||
1166 | VERIFY(zfsvfs_teardown(zsb, B_TRUE) == 0); | |
1167 | os = zsb->z_os; | |
34dc7c2f BB |
1168 | |
1169 | /* | |
1170 | * z_os will be NULL if there was an error in | |
3558fd73 | 1171 | * attempting to reopen zsb. |
34dc7c2f BB |
1172 | */ |
1173 | if (os != NULL) { | |
1174 | /* | |
1175 | * Unset the objset user_ptr. | |
1176 | */ | |
428870ff | 1177 | mutex_enter(&os->os_user_ptr_lock); |
34dc7c2f | 1178 | dmu_objset_set_user(os, NULL); |
428870ff | 1179 | mutex_exit(&os->os_user_ptr_lock); |
34dc7c2f BB |
1180 | |
1181 | /* | |
b128c09f | 1182 | * Finally release the objset |
34dc7c2f | 1183 | */ |
3558fd73 | 1184 | dmu_objset_disown(os, zsb); |
34dc7c2f BB |
1185 | } |
1186 | ||
3558fd73 | 1187 | zfs_sb_free(zsb); |
34dc7c2f BB |
1188 | return (0); |
1189 | } | |
e5c39b95 | 1190 | EXPORT_SYMBOL(zfs_umount); |
34dc7c2f | 1191 | |
0de19dad BB |
1192 | int |
1193 | zfs_remount(struct super_block *sb, int *flags, char *data) | |
1194 | { | |
0de19dad | 1195 | /* |
2cf7f52b BB |
1196 | * All namespace flags (MNT_*) and super block flags (MS_*) will |
1197 | * be handled by the Linux VFS. Only handle custom options here. | |
0de19dad | 1198 | */ |
0de19dad BB |
1199 | return (0); |
1200 | } | |
1201 | EXPORT_SYMBOL(zfs_remount); | |
1202 | ||
e5c39b95 | 1203 | int |
2cf7f52b | 1204 | zfs_vget(struct super_block *sb, struct inode **ipp, fid_t *fidp) |
34dc7c2f | 1205 | { |
2cf7f52b | 1206 | zfs_sb_t *zsb = sb->s_fs_info; |
34dc7c2f BB |
1207 | znode_t *zp; |
1208 | uint64_t object = 0; | |
1209 | uint64_t fid_gen = 0; | |
1210 | uint64_t gen_mask; | |
1211 | uint64_t zp_gen; | |
3558fd73 | 1212 | int i, err; |
34dc7c2f | 1213 | |
3558fd73 | 1214 | *ipp = NULL; |
34dc7c2f | 1215 | |
3558fd73 | 1216 | ZFS_ENTER(zsb); |
34dc7c2f BB |
1217 | |
1218 | if (fidp->fid_len == LONG_FID_LEN) { | |
1219 | zfid_long_t *zlfid = (zfid_long_t *)fidp; | |
1220 | uint64_t objsetid = 0; | |
1221 | uint64_t setgen = 0; | |
1222 | ||
1223 | for (i = 0; i < sizeof (zlfid->zf_setid); i++) | |
1224 | objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); | |
1225 | ||
1226 | for (i = 0; i < sizeof (zlfid->zf_setgen); i++) | |
1227 | setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); | |
1228 | ||
3558fd73 | 1229 | ZFS_EXIT(zsb); |
34dc7c2f | 1230 | |
3558fd73 BB |
1231 | #ifdef HAVE_SNAPSHOT |
1232 | err = zfsctl_lookup_objset(vfsp, objsetid, &zsb); | |
34dc7c2f BB |
1233 | if (err) |
1234 | return (EINVAL); | |
3558fd73 BB |
1235 | #endif /* HAVE_SNAPSHOT */ |
1236 | ZFS_ENTER(zsb); | |
34dc7c2f BB |
1237 | } |
1238 | ||
1239 | if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { | |
1240 | zfid_short_t *zfid = (zfid_short_t *)fidp; | |
1241 | ||
1242 | for (i = 0; i < sizeof (zfid->zf_object); i++) | |
1243 | object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); | |
1244 | ||
1245 | for (i = 0; i < sizeof (zfid->zf_gen); i++) | |
1246 | fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); | |
1247 | } else { | |
3558fd73 | 1248 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1249 | return (EINVAL); |
1250 | } | |
1251 | ||
3558fd73 | 1252 | #ifdef HAVE_SNAPSHOT |
34dc7c2f BB |
1253 | /* A zero fid_gen means we are in the .zfs control directories */ |
1254 | if (fid_gen == 0 && | |
1255 | (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { | |
3558fd73 BB |
1256 | *ipp = zsb->z_ctldir; |
1257 | ASSERT(*ipp != NULL); | |
34dc7c2f | 1258 | if (object == ZFSCTL_INO_SNAPDIR) { |
3558fd73 | 1259 | VERIFY(zfsctl_root_lookup(*ipp, "snapshot", ipp, NULL, |
34dc7c2f BB |
1260 | 0, NULL, NULL, NULL, NULL, NULL) == 0); |
1261 | } else { | |
3558fd73 | 1262 | igrab(*ipp); |
34dc7c2f | 1263 | } |
3558fd73 | 1264 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1265 | return (0); |
1266 | } | |
3558fd73 | 1267 | #endif /* HAVE_SNAPSHOT */ |
34dc7c2f BB |
1268 | |
1269 | gen_mask = -1ULL >> (64 - 8 * i); | |
1270 | ||
1271 | dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask); | |
3558fd73 BB |
1272 | if ((err = zfs_zget(zsb, object, &zp))) { |
1273 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1274 | return (err); |
1275 | } | |
3558fd73 | 1276 | (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zsb), &zp_gen, |
428870ff BB |
1277 | sizeof (uint64_t)); |
1278 | zp_gen = zp_gen & gen_mask; | |
34dc7c2f BB |
1279 | if (zp_gen == 0) |
1280 | zp_gen = 1; | |
1281 | if (zp->z_unlinked || zp_gen != fid_gen) { | |
1282 | dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen); | |
3558fd73 BB |
1283 | iput(ZTOI(zp)); |
1284 | ZFS_EXIT(zsb); | |
34dc7c2f BB |
1285 | return (EINVAL); |
1286 | } | |
1287 | ||
3558fd73 BB |
1288 | *ipp = ZTOI(zp); |
1289 | if (*ipp) | |
1290 | zfs_inode_update(ITOZ(*ipp)); | |
960e08fe | 1291 | |
3558fd73 | 1292 | ZFS_EXIT(zsb); |
34dc7c2f BB |
1293 | return (0); |
1294 | } | |
e5c39b95 | 1295 | EXPORT_SYMBOL(zfs_vget); |
34dc7c2f BB |
1296 | |
1297 | /* | |
3558fd73 | 1298 | * Block out VOPs and close zfs_sb_t::z_os |
34dc7c2f BB |
1299 | * |
1300 | * Note, if successful, then we return with the 'z_teardown_lock' and | |
1301 | * 'z_teardown_inactive_lock' write held. | |
1302 | */ | |
1303 | int | |
3558fd73 | 1304 | zfs_suspend_fs(zfs_sb_t *zsb) |
34dc7c2f BB |
1305 | { |
1306 | int error; | |
1307 | ||
3558fd73 | 1308 | if ((error = zfsvfs_teardown(zsb, B_FALSE)) != 0) |
34dc7c2f | 1309 | return (error); |
3558fd73 | 1310 | dmu_objset_disown(zsb->z_os, zsb); |
34dc7c2f BB |
1311 | |
1312 | return (0); | |
1313 | } | |
e5c39b95 | 1314 | EXPORT_SYMBOL(zfs_suspend_fs); |
34dc7c2f BB |
1315 | |
1316 | /* | |
3558fd73 | 1317 | * Reopen zfs_sb_t::z_os and release VOPs. |
34dc7c2f BB |
1318 | */ |
1319 | int | |
3558fd73 | 1320 | zfs_resume_fs(zfs_sb_t *zsb, const char *osname) |
34dc7c2f | 1321 | { |
428870ff | 1322 | int err, err2; |
34dc7c2f | 1323 | |
3558fd73 BB |
1324 | ASSERT(RRW_WRITE_HELD(&zsb->z_teardown_lock)); |
1325 | ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock)); | |
34dc7c2f | 1326 | |
3558fd73 | 1327 | err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zsb, &zsb->z_os); |
34dc7c2f | 1328 | if (err) { |
3558fd73 | 1329 | zsb->z_os = NULL; |
34dc7c2f BB |
1330 | } else { |
1331 | znode_t *zp; | |
428870ff BB |
1332 | uint64_t sa_obj = 0; |
1333 | ||
3558fd73 | 1334 | err2 = zap_lookup(zsb->z_os, MASTER_NODE_OBJ, |
428870ff BB |
1335 | ZFS_SA_ATTRS, 8, 1, &sa_obj); |
1336 | ||
3558fd73 | 1337 | if ((err || err2) && zsb->z_version >= ZPL_VERSION_SA) |
428870ff BB |
1338 | goto bail; |
1339 | ||
1340 | ||
3558fd73 BB |
1341 | if ((err = sa_setup(zsb->z_os, sa_obj, |
1342 | zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0) | |
572e2857 | 1343 | goto bail; |
34dc7c2f | 1344 | |
3558fd73 | 1345 | VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0); |
34dc7c2f BB |
1346 | |
1347 | /* | |
1348 | * Attempt to re-establish all the active znodes with | |
1349 | * their dbufs. If a zfs_rezget() fails, then we'll let | |
1350 | * any potential callers discover that via ZFS_ENTER_VERIFY_VP | |
1351 | * when they try to use their znode. | |
1352 | */ | |
3558fd73 BB |
1353 | mutex_enter(&zsb->z_znodes_lock); |
1354 | for (zp = list_head(&zsb->z_all_znodes); zp; | |
1355 | zp = list_next(&zsb->z_all_znodes, zp)) { | |
34dc7c2f BB |
1356 | (void) zfs_rezget(zp); |
1357 | } | |
3558fd73 | 1358 | mutex_exit(&zsb->z_znodes_lock); |
34dc7c2f BB |
1359 | |
1360 | } | |
1361 | ||
428870ff | 1362 | bail: |
34dc7c2f | 1363 | /* release the VOPs */ |
3558fd73 BB |
1364 | rw_exit(&zsb->z_teardown_inactive_lock); |
1365 | rrw_exit(&zsb->z_teardown_lock, FTAG); | |
34dc7c2f BB |
1366 | |
1367 | if (err) { | |
1368 | /* | |
3558fd73 | 1369 | * Since we couldn't reopen zfs_sb_t::z_os, force |
34dc7c2f BB |
1370 | * unmount this file system. |
1371 | */ | |
3558fd73 | 1372 | (void) zfs_umount(zsb->z_sb); |
34dc7c2f BB |
1373 | } |
1374 | return (err); | |
1375 | } | |
e5c39b95 | 1376 | EXPORT_SYMBOL(zfs_resume_fs); |
34dc7c2f | 1377 | |
34dc7c2f | 1378 | int |
3558fd73 | 1379 | zfs_set_version(zfs_sb_t *zsb, uint64_t newvers) |
34dc7c2f BB |
1380 | { |
1381 | int error; | |
3558fd73 | 1382 | objset_t *os = zsb->z_os; |
34dc7c2f | 1383 | dmu_tx_t *tx; |
34dc7c2f BB |
1384 | |
1385 | if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) | |
1386 | return (EINVAL); | |
1387 | ||
3558fd73 | 1388 | if (newvers < zsb->z_version) |
9babb374 | 1389 | return (EINVAL); |
34dc7c2f | 1390 | |
428870ff | 1391 | if (zfs_spa_version_map(newvers) > |
3558fd73 | 1392 | spa_version(dmu_objset_spa(zsb->z_os))) |
428870ff BB |
1393 | return (ENOTSUP); |
1394 | ||
34dc7c2f | 1395 | tx = dmu_tx_create(os); |
9babb374 | 1396 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); |
3558fd73 | 1397 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1398 | dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, |
1399 | ZFS_SA_ATTRS); | |
1400 | dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); | |
1401 | } | |
34dc7c2f BB |
1402 | error = dmu_tx_assign(tx, TXG_WAIT); |
1403 | if (error) { | |
1404 | dmu_tx_abort(tx); | |
9babb374 BB |
1405 | return (error); |
1406 | } | |
428870ff | 1407 | |
9babb374 BB |
1408 | error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, |
1409 | 8, 1, &newvers, tx); | |
1410 | ||
1411 | if (error) { | |
1412 | dmu_tx_commit(tx); | |
1413 | return (error); | |
34dc7c2f | 1414 | } |
34dc7c2f | 1415 | |
3558fd73 | 1416 | if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) { |
428870ff BB |
1417 | uint64_t sa_obj; |
1418 | ||
3558fd73 | 1419 | ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, |
428870ff BB |
1420 | SPA_VERSION_SA); |
1421 | sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, | |
1422 | DMU_OT_NONE, 0, tx); | |
1423 | ||
1424 | error = zap_add(os, MASTER_NODE_OBJ, | |
1425 | ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); | |
1426 | ASSERT3U(error, ==, 0); | |
1427 | ||
1428 | VERIFY(0 == sa_set_sa_object(os, sa_obj)); | |
1429 | sa_register_update_callback(os, zfs_sa_upgrade); | |
1430 | } | |
1431 | ||
1432 | spa_history_log_internal(LOG_DS_UPGRADE, | |
1433 | dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu", | |
3558fd73 | 1434 | zsb->z_version, newvers, dmu_objset_id(os)); |
9babb374 | 1435 | |
34dc7c2f BB |
1436 | dmu_tx_commit(tx); |
1437 | ||
3558fd73 | 1438 | zsb->z_version = newvers; |
9babb374 | 1439 | |
3558fd73 BB |
1440 | if (zsb->z_version >= ZPL_VERSION_FUID) |
1441 | zfs_set_fuid_feature(zsb); | |
9babb374 BB |
1442 | |
1443 | return (0); | |
34dc7c2f | 1444 | } |
e5c39b95 | 1445 | EXPORT_SYMBOL(zfs_set_version); |
34dc7c2f BB |
1446 | |
1447 | /* | |
1448 | * Read a property stored within the master node. | |
1449 | */ | |
1450 | int | |
1451 | zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) | |
1452 | { | |
1453 | const char *pname; | |
b128c09f | 1454 | int error = ENOENT; |
34dc7c2f BB |
1455 | |
1456 | /* | |
1457 | * Look up the file system's value for the property. For the | |
1458 | * version property, we look up a slightly different string. | |
1459 | */ | |
1460 | if (prop == ZFS_PROP_VERSION) | |
1461 | pname = ZPL_VERSION_STR; | |
1462 | else | |
1463 | pname = zfs_prop_to_name(prop); | |
1464 | ||
b128c09f BB |
1465 | if (os != NULL) |
1466 | error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); | |
34dc7c2f BB |
1467 | |
1468 | if (error == ENOENT) { | |
1469 | /* No value set, use the default value */ | |
1470 | switch (prop) { | |
1471 | case ZFS_PROP_VERSION: | |
1472 | *value = ZPL_VERSION; | |
1473 | break; | |
1474 | case ZFS_PROP_NORMALIZE: | |
1475 | case ZFS_PROP_UTF8ONLY: | |
1476 | *value = 0; | |
1477 | break; | |
1478 | case ZFS_PROP_CASE: | |
1479 | *value = ZFS_CASE_SENSITIVE; | |
1480 | break; | |
1481 | default: | |
1482 | return (error); | |
1483 | } | |
1484 | error = 0; | |
1485 | } | |
1486 | return (error); | |
1487 | } | |
3558fd73 BB |
1488 | |
1489 | void | |
1490 | zfs_init(void) | |
1491 | { | |
1492 | zfs_znode_init(); | |
1493 | dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); | |
1494 | register_filesystem(&zpl_fs_type); | |
1495 | } | |
1496 | ||
1497 | void | |
1498 | zfs_fini(void) | |
1499 | { | |
1500 | unregister_filesystem(&zpl_fs_type); | |
1501 | zfs_znode_fini(); | |
1502 | } |